Summary of Workshops
Pathobiology. Primary herpes simplex virus (HSV) infection occurs through mucosal surfaces, followed by Kaposi’s varicelliform eruption. Immune host responses control this primary infection within 14 to 21 days. What is unique is the retrograde neuronal transport of the virus to the dorsal root ganglia, where a cycle of replication occurs and the virus becomes latent until a provocative stimulus leads to reactivation. At that time, anterior grade transport returns the virus to the mucosal surfaces, where replication ensues.
Incidence and Burden. By adulthood, about 80 percent of the U.S. population has been exposed to nongenital HSV type 1 and is at risk for reactivation, leading to herpes simplex labialis. There are about 750,000 new cases of genital infection per year in the United States, and overall seroprevalence of HSV-2 is about 60 million individuals. Genital herpes generates about 600,000 physician visits and 20,000 unnecessary caesarian sections annually. Two more serious infections are herpes simplex encephalitis and neonatal herpes, each with about 1,500 cases per year. The cost of neonatal herpes is at least $750 million per year, primarily in indirect costs for the long-term management of neurological sequelae from which these babies suffer.
Seroprevalence of HSV type 2 is increasing rapidly. In 1992, seroprevalence for the U.S. population at large was about 31 percent; for persons of color, between 50 and 55 percent. Acquisition is a function of the number of sexual partners to whom an individual is exposed: for heterosexual men with greater than 50 partners, the probability of acquiring HSV-2 is 80 percent or higher; for heterosexual women with more than 50 partners, over 90 percent. Women are more likely to acquire HSV-2 than men. As with other genital ulcerative disease, there is also a four- or fivefold increase in the risk of acquiring HIV infection.
Rationale and Goals of Vaccine. Studies of couples with discordant serostatus suggest that prior HSV-1 infection confers some degree of protection from acquiring HSV-2 infection. When the male is positive for HSV-2 and the female is seronegative, her probability of acquiring HSV-2 is 20 percent during the calendar year. This becomes particularly significant if she becomes pregnant during that year. If the female is seropositive for HSV-1, however, the probability drops to 10 percent. (The reverse also holds: if the male is positive for HSV-1 rather than seronegative, his probability of acquiring HSV-2 drops from 10 percent to 5 percent.) These findings suggest that a vaccine that could seroconvert the mother from at-risk to lower-risk could lower the probability of transmission to the newborn.
The goal of this vaccine would be to prevent or at least reduce the severity of primary infection, and possibly to reduce the frequency and clinical symptoms of reoccurrences. We know that exogenous reinfection is extremely
uncommon in the immune-competent host, and that HSV-1 antibodies reduce both the probability and the clinical symptoms of primary HSV-2 infection. Similarly, fetal transmission is 10 times less likely when there are preexisting HSV-2 antibodies, compared with primary infection. And the duration of primary HSV-2 infections are reduced in seropositive individuals from 21–28 days to 10–14 days.
Approaches to Vaccine Development. Four approaches have been attempted in developing such a vaccine. Live virus by autoinoculation has been defined as a failure in studies over the past 100 years. HSV has 13 glycoproteins in its envelope, two of which (B and D) are required for infectivity and potent inducers of neutralizing antibodies. Subunit vaccines for glycoprotein B, glycoprotein D, and a combination of B and D are under investigation. Vector glycoprotein genes using either canary pox or a vaccinia vector have been. Engineered herpes simplex virus is a potential future candidate for clinical trials.
Subunit Vaccines. Clinical trials are currently evaluating monoclonal antibody therapy using glycoprotein B and D combinations to treat neonatal HSV infection and HSV encephalitis. Other subunit trials underway include one therapeutic trial of approximately 800 volunteers and 2 primary prevention trials of approximately 500 couples each. In the latter prophylactic trial, one partner in each monogamous relationship has recurrent HSV-2 and the other is HSV-2-seronegative. The former, therapeutic trial is based on the results of a preliminary trial of 100 volunteers between the ages 18 and 55, who averaged 4 to 14 outbreaks per year and had not received acyclovir for 3 months prior to therapy. Half received 100 micrograms of glycoprotein D-2 with alum adjuvant; half received placebo; and there was a booster at 2 months. In those who received the vaccine, the average number of recurrences dropped from 0.5 to 0.42 per month, and the median from 6 to 4 per year.
Genetically Engineered HSV Virus. Genetic stability is a concern in all genetically engineered vaccines; reversions to or recombinations with wild-type viruses, when they do occur, should be less virulent than the current strain. In addition, live attenuated viruses abdicated the ability to become latent and therefore, under some circumstances, might be reactivated. This is why deletions in the gamma 134.5 gene are particularly promising: they debilitate the virus’ ability to become latent and subsequently reactivate (see below).
The herpes simplex virus has a genome that is only 150,000 base pairs (150 kB) long, in an architecture consisting of a unique long segment and a unique short segment. There are internal repeats bonding the unique long and short segments so that the virus could invert upon itself, leaving four equimotor isomers in any population. There is a 70-percent homology between HSV-1 and HSV-2.
The starting point for an engineered vaccine is an intratypic hybrid identified as R-70–20 and was developed about 10 years ago. It consists of the unique long segment of HSV-1, a segment of the unique long segment of HSV-
2, and the unique short segment of HSV-2. A 12-kB deletion of internal repeats from the HSV-2 short domain, mapping for a series of glycoproteins (including G, J, GD, GI, and GE), was inserted. This vaccine considerably reduces virulence in rodent models, is stable upon serial passage in mouse brains, and is safe and efficacious in aotis monkeys at concentrations up to 107 PFU administered by any route, including intracerebrally. It was safe in humans at up to 104.5 PFU (higher doses were not studied). This vaccine required two doses, implying that it was overly attenuated for administration to humans.
HSV can also be attenuated through deletions and stop codes. The gamma 134.5 gene resides in the inverted repeats in two copies on the unique long segment of the HSV-2 genome; when 134.5 is deleted, it leaves a virus with an LD-50 upon inoculation directly into the central nervous system of a mouse of 106 PFU, compared with 102 PFU for the restored wild-type virus. A stop-code on the carboxy terminus of the genome can similarly attenuate the virus.
Using the latter technique with HSV-2, a recombinant vaccine was generated that also deleted the structural components of UL-55 and UL-56. This vaccine has significantly reduced neurovirulence in mice—LD-50 is 5.6 X 105 PFU, compared with less than 50 PFU for the wild type in that circumstance—and is safe in the aotis monkey at up to 106 PFU. In terms of efficacy, the vaccine appears to protect guinea pigs from disease at dosages in the range of 104 to 105 PFU. Immunized aotis monkeys survive challenges with wild-type virus at up to 105 PFU when given intravaginally. These results suggests that this is, at least potentially, a genetically engineered vaccine that will provide a broader immune response than is encountered with subunit vaccines.
Both the method of attenuation and the route of administration seem to influence the efficacy of these candidate vaccines. When mice were inoculated intranasally and then challenged intranasally with either wild virus or HSV-2, there was a reduction of mortality of 20 percent for those inoculated with 134.5-deletion mutants, and 13 percent when a stop code is place at the carboxy terminus. When mice are immunized intranasally and challenged intravaginally, however, the effect is not as great. Researchers are currently looking for IgA and IgG2a in the vaginal secretions of these mice to learn more about these immune responses.
Animal Models. Two new animal models have emerged from this work. The first is a test of the virulence of the virus; the aotis monkey is exquisitely sensitive for the evaluation of genetically engineered viruses, which are inoculated directly into the eye. The second is a rodent model to establish the genetic stability of attenuated vaccines. It involves inoculating virus into the mouse brain, harvesting tissue, raising the virus again, and reinoculating virus into mouse brains on subsequent occasions; nine passages will usually select unstable variants or a reversion to wild type. Inoculation into mouse brain can also be used to evaluate the virulence of attenuated virus.
In response to questions from the audience, Dr. Whitley added the following:
Researchers do not at present have a marker of host immune response that is predictive of vaccine efficacy. Both neutralizing antibodies and cell-mediated immune responses have been investigated and eliminated.
Gamma-134.5-deletion vaccines will eventually be tested in immunocompromised models.
Subunit vaccines currently require three doses to provide persistent antibody levels, although the duration and intensity of immune response may increase with a new adjuvant. R-70–20 acted more like a dead vaccine, requiring two doses. There are hopes that replication-competent viruses like the 134.5-deletion mutants will provide enhanced immunogenic effects with one shot rather than two.
There has been little work on cellular responses to these vaccines or indeed to primary and recurrent herpes at all. Some researchers are doing CTLs on mice, and others are looking at T-cell responses in humans who are seropositive but have no clinical recurrences. T-cell responses following primary infection are robust, but no one has done comprehensive CTLs or tried to dissect out specific T-cell response in all of these human populations.
The target population for HSV-2 immunization is adolescents as they are about to begin sexual activity, perhaps 10 or 12 years old, and the vaccine should provide at least 10 years of resistance to the wild-type virus. It will probably be impractical to test for serologic status for HSV-1 or HSV-2 prior to immunization.
Epstein-Barr virus (EBV) is another extremely common virus, but it is linked to a growing list of pathologies. Unlike HSV, there have been few vaccine trials, but a great deal is known about the disease processes. This knowledge could be used to generate partial or complete immunity to block these diseases.
Incidence and Burden. Worldwide, about 95 percent of the adult population is infected with EBV. If the primary infection comes in late adolescence, about 50 percent of seronegatives will develop the clinical manifestation called infectious mononucleosis (IM). IM is a significant disease, with at least 125,000 new cases recorded each year in the United States. About 70 percent of patient will resolve their symptoms in 2 to 4 weeks and is the major cause of lost time for new Army recruits. However, 30 percent develop more extensive impairment that may not resolve for 3 to 4 months. About 1 percent of cases develop complications, including neurological, bone marrow, liver involvement, and
even fatal IM. This extensive morbidity might be decreased by appropriate immunotherapy, but there is at present no effective antiviral therapy.
In most cases EBV establishes latency in the B-lymphocytes. However, EBV is associated with a variety of aplastic diseases, including 80 percent of AIDS-related lymphomas. Among post-transplant proliferative disorders, anywhere from 25 percent to 100 percent (depending on organ) occur in patients who are having a primary EBV infection. Similarly, there are many similarities between IM and Hodgkin’s disease—patients with a history of IM are 2 to 3 times more likely to develop Hodgkin’s lymphoma, and 70 percent of Hodgkin’s-like lymphocytes are EBV-positive. EBV is also associated with Burkitt’s lymphoma, which is relatively rare in the United States but is endemic in Africa. EBV is associated with the vast majority of nasopharyngeal carcinoma, which occurs when the virus is reactivated in mucosal lymphocytes and infects epithelial cells, which develop abnormalities and rapidly becomes dysplasia or carcinoma in situ. EBV is now being described in an increasing number of T-cell lymphomas, a significant proportion of Hodgkin’s disease, and in discrete substantive gastric carcinomas. Other related diseases include hairy leukoplakia, which was originally described in AIDS patients.
Pathobiology. EBV can establish both permissive infection in epithelial cells and lymphocytes and latent, nonpermissive transforming infections in lymphocytes. Pathogenesis involves introduction of the virus into the oral epithelial cells, which are usually permissive to viral infection, and virus is then secreted into the saliva. Secondarily to this epithelial infection, virus enters the B-lymphocyte, where it circularizes, forming an extra chromosomal episome in the nucleus. One week after the initial infection, as many as 5 percent to 20 percent of peripheral blood lymphocytes are infected with EBV, including a wide variety of cell types: B-lymphocytes, Reed-Sternberg-like cells, plasmacytoid cells, and even a small percentage of T-lymphocytes. This is basically a latent infection: most of these cells do not make viruses but instead begin to express viral gene products that are associated with the process of cellular transformation. Many of these gene products are very potent targets of cytotoxic T-lymphocytes (CTLs), which must be generated to control the infection and the transformed lymphocytes. In the absence of T-cell response, these lymphocytes grow uncontrolled and eventually develop into a lymphoma.
Viral Genome. Analysis of the viral genome can distinguish latent from replicative infection and is informative in other ways. Like HSV, the EBV genome is a double-stranded DNA molecule with variants, but with a simpler structure. It is about 190 kB long, but instead of inverted repeats, it has multiple copies of a 500-base-pair direct repeat at each end of the genome. Variants are highly heterogenous, with anywhere from 1 to 20 copies of the terminal fragments at either end of the genome. This assay was able to distinguish between the variant form from the larger episomal form, but it also revealed that (within each cell) each of the up to 100 copies of the EBV episome was identical with regard to the number of terminal repeats. This also suggested that every cell within a tumor would be identical, from which researchers deduced that the
epithelial tumors associated with EBV (e.g., nasopharyngeal and salivary gland carcinomas) were also clonal. This was confirmed by comparing the clonality predicted by EBV terminal studies and the clonality found in immunoglobulin rearrangements.
This indicated that most of the tumors associated with EBV were clonal proliferations and, more importantly, that these proliferations had developed from a single EBV-infected progenitor cell. This has been confirmed in post-transplant proliferative disorders, Hodgkin’s lymphoma, Burkitt’s lymphoma, gastrocarcinomas, cancers of the salivary gland, and the new T-cell lymphomas that are being reported in Taiwan and Japan. In nasopharyngeal carcinoma, patients have extremely high titers to replicative antigens and, while the Southern blot test does show a single clonal EBV fragment, there is also a faint ladder array indicating that at least a few cells go into permissive infection and make additional virus. The principal exception is hairy leukoplakia, for which the tests show no evidence of a fused band representing an episomal compartment, but rather an abundant ladder array indicating that this is a permissive infection. In the majority of cases, however, the diseases associated with EBV are associated with latent infection.
EBV Gene Functions. As noted, EBV gene products are expressed by infected cells, and researchers have identified the function of many of these products. One of the most important is Epstein-Barr nuclear antigen type one (EBNA-1), which binds to a specific DNA sequence in the EBV genome (the origin of plasma replication) and allows replication by DNA prolimerase. As a result, these infections are not susceptible to any kind of antiviral therapy directed toward the prolimerase. EBNA-1 also has the ability to induce the expression of the rad genes, which may be critical in some of the genomic rearrangements that have been described in Burkitt’s lymphoma. No CTLs have been identified that are specific for EBNA-1, apparently due to a unique sequence in the EBNA-1 gene that mimics a cellular gene and prevents it from being recognized by class one antigens.
Five additional nuclear antigens (EBNA-2 through 6) have viral regulatory functions. Three of them are essential for transformation of B-lymphocytes. EBNA-2 transactivates these other genes, and three of them regulate the expression of latent membrane protein one (LMP-1) in the transformed lymphocytes. Almost all of the CTL targets are in these genes, which are expressed uniquely in B-lymphocytes.
LMP-1 is considered to be the viral oncogene: it promotes transformation and cellular division; it is the only gene that transforms rodent cells and cultures; and it is essential to the activation of a B-lymphocyte. As a result, it is able to inhibit apoptosis both in B-lymphocytes (by inducing expression of BCL-2) and in epithelial cells (through a p53-dependent mechanism that occurs through the induction of the A-20 gene). The LMP-1 gene is expressed in post-transplant lymphomas and other malignancies associated with EBV; it appears to be a
critical gene for neoplastic growth. LMP-2 associates with LMP-1 in transformed cells and appears to block activation of the B-lymphocyte by interfering with the signals that would induce antibody synthesis. In doing so, LMP-2 helps to maintain the latent infection.
A last set of EBV genes expressed in infected cells are called Epstein-Barr encoded RNAs (EBERs). They are RNA polymerase-3 transfers, but their function is not yet known. They are not essential for transformation, but they are extremely abundant (50,000 to 1 million copies per infected cell) and very stable. EBERs can be very useful for identifying infected tissues, and because they are not expressed in hairy leukoplakia (a permissive infection) they may be useful in identifying repositories of latent infection.
Levels of Latency. Research on the actual behavior of EBV in each of the associated pathologies has shown that there are actually three levels of latency. Type one latency is a minimal level of expression, characteristic of Burkitt’s lymphoma. Only EBNA-1 and the EBERs are expressed, so the infected cells do not become targets for cytotoxic T-cells. The peripheral blood lymphocytes of normal individuals who have been exposed to EBV also express EBNA-1 and EBERs, and they are also believed to express LMP-2, helping to maintain them in a state of latent infection.
In type two latency, additional genes are expressed. Specifically, LMP-1 and LMP-2, two genes that are regulated by the EBNAs, are expressed in the absence of the EBNAs. In other words, the critical transforming genes are being expressed but not the key CTL targets, which may be important in the development of disease for which this is the characteristic state of latency, including nasopharyngeal carcinoma, Hodgkin’s disease, and T-cell lymphoma.
In type three latency, the entire array of EBV latent genes are expressed, including not only the oncogenes but also EBNAs 2 through 6, which are the key CTL targets, as well as high levels of EBERs whose function is as yet unknown. This type of infection, which is typical of peripheral blood lymphocytes in infectious mononucleosis and most of the cells in post-transplant proliferative disease, should be very amenable to control by the immune system.
In replicative infection, one vital gene called ZEBRA is the replication activator that turns on the expression of the viral replicative antigens (including polymerase and thymidine kinase) and the structural proteins (including the viral caps antigen, or VCA, and glycoproteins 350/220, 110, 85, 42, and 27). The glycoproteins of EBV are much less complex than those of HSV, and one in particular—gp350/220—is the main target of neutralizing antibodies and hence the principal focus of research attention.
Approaches for Vaccine Development. One approach that is being pursued is recombinant gp350, which does induce antibody-dependent cytotoxicity. In cotton-top tamarinds, this vaccine protects against lymphoma after a parenteral challenge. This approach would probably protect humans in pathologies that are dependent on replication, such as IM and nasopharyngeal carcinoma. In a very limited trial in China, 10 children were infected with gp350 in vaccinia and 10 got placebo; after a year, all 10 on placebo became infected with EBV,
but only 2 of the 10 who got gp350 were infected. Much has been done in terms of designing gp350 to be produced in the cells and working on a strategy for clarification in high quantities. Abnormal viral response in the absence of antibodies to fusion protein is a concern.
T-cell epitope vaccines are also being investigated, because researchers have identified so many of the CTL epitopes that are critical for EBV infections. Experiments with SCID mice indicate that it is possible to transfer CTLs and thereby control viral proliferations. Such a vaccine might provide prophylactic protection for transplant recipients and (in combination with gp350) in nasopharyngeal carcinoma. Tumors may not express key antigens, or viral functions may impair peptide presentation.
Ultimately, however, the goal would be to make a genetically altered EBV vaccine. This would be essential to protect against EBV-related malignancies such as Burkitt’s lymphoma and nasopharyngeal carcinoma, where the infection would have to be eliminated to prevent the disease. As detailed above, many of the necessary steps have been taken in this direction. Many people are infected with both type 1 and type 2 EBV, suggesting that wild-type infection is not protective.
Animal Models. Animal models are somewhat limited. Cotton-top tamarinds develop lymphomas when EBV is injected parenterally, and they can be protected with gp350 vaccine. SCID-human chimeric mice also develop EBV lymphoproliferative diseases, and can be protected with transferred CTLs; but they may not be very useful for evaluating vaccines because of transient T-cell function and the absence of appropriate lymphokine synthesis.
A new and promising prototype is a mouse EBV homologue, although it might be more similar to herpes saimiri. You can induce infection by internasal inoculation, and the mice develop lymphoid proliferations that have some similarities to EBV infection. However, the most encouraging animal model is a rhesus EBV that is highly homologous to EBV, and has identical patterns of mucosal infection and disease. Also, SIV-infected animals develop lymphomas. This may be the ideal system to test a genetically altered EBV virus.
In response to questions from the audience, Dr. Raab-Traub added the following:
No research is being done to analyze the mucosal infection or to induce mucosal immune response. There have been anecdotal reports of defective viral genomes in some seronegative subjects. These genomes, which lack the transforming EBV genes that generated immunoglobulin-A response in saliva, might offer some protection.
100 percent of infected B-lymphocytes become immortalized.
Recombinant gp350 vaccine should be ready for testing in a small seronegative population soon. It would be effective against both type 1 and type 2 EBV.
HEPATITIS C VIRUS3
Hepatitis C virus (HCV) is a recent field of research but a major cause of disease. What used to be known as “non-A and non-B hepatitis” was recognized in the mid-1970s with the development of serological tests for the hepatitis A and B viruses, and the cause was identified and named in 1988. Researchers are only beginning to understand the interaction of HCV with the host and the ensuing immune responses. With the technology now available in the areas of recombinant antigen production, adjuvants, and genetic immunization, researchers hope to have at least some impact on transmission and disease development at some point in the future.
Burden and Epidemiology. HCV is a truly global problem and is the major infectious disease problem in Japan. In the United States, there have been 150,000 new cases of HCV infection per year for the past decade. From 50 percent to 60 percent of these infections progress to persistent viremia and chronic persistent hepatitis. Of patients with chronic hepatitis, about 20 percent will progress to cirrhosis, and about 20 percent of cirrhotics with HCV will undergo liver failure. As a result, HCV is most common cause for liver transplantation in the United States. In addition, it is now clear that hepatocellular carcinoma is associated with HCV, with cirrhosis as a precondition, and that about 10 percent of cirrhotics with HCV will develop cancer over a period of time.
Research in Japan indicates that mean time from infection to cirrhosis is 20 years, and to cancer 30 years, so this is a very indolent type of disease. At the same time, research at NIH indicates that in a minority of patients cirrhosis can develop in a few years. Conversely, around 15 percent of post-transfusion HCV infections become negative over the long term, so there are some spontaneous resolvers.
The biggest risk factor for contracting hepatitis C is intravenous drug use, constituting about half of U.S. infections. About 30 percent to 40 percent of cases have no known risk factor. Transfusion is now a negligible risk, with the introduction of a blood screening test. There is a measurable incidence in health care workers who are exposed to infected blood, and data from Japan indicated that mothers can transmit the virus to their babies. The issue of sexual transmission is controversial, but studies show that multiple heterosexual partners are a risk factor, while the incidence in homosexual men is extremely low; apparently HCV can be transmitted sexually, but it is inefficiently transmitted that way.
Viral Genome. HCV belongs to the Flaviviridae family, along with flavivirus and pestiviruses. It is an RNA virus that does not integrate into cellular DNA, and instead replicates only through RNA replication intermediates. The virus is highly heterogenous: it can change rapidly in the infected host, and at
least 6 major types and 40 subtypes have been identified around the world. The 1A and 1B subtypes account for most infections in the United States.
The genome consists of about 10,000 nucleotides, with a 5' terminal iris and an internal ribosome entry site. It makes a large polyprotein that is cleaved post-translationally and co-translationally by a combination of host enzymes and viral proteases. It seems that the host signal peptidase is primarily involved in processing the nuclear capsid and the two envelope glycoproteins (gpE1 and gpE2). The presumed nonstructural proteins (NS) appear to be processed by the action of two viral proteases, one in the NS3 domain that is a trypsin-like protease, and another that spans the NS2 and NS3 genes that appears to be a metallic protease.
Immunology. Preliminary work on the correlates of immunity have shown that peripheral CD4-positive T-cells respond to HCV nonstructural protein 3 (NS3) very early in the infection, and that this response persists following recovery. In chronic patients, however, there is very little T-cell response. There is evidence that the protective immune response is short-lived and weak, at best. For example, a study of polytransfused thalassemic children found that patients who normalized after an initial episode of acute hepatitis nevertheless developed a second infection that progressed to chronic persistent hepatitis. Both infections involved the HCV-1-B subtype.
Viral Persistence. HCV is remarkably adept at persisting in the host in the face of an apparently substantial immune response. In the livers of about 50 percent of patients with chronic hepatitis, researchers are able to identify CTLs of varying specificity to either structural protein or to nonstructural proteins in the polyprotein precursor. CTLs also infiltrate the livers of chimpanzees with chronic infections. It is not known how the virus persists in the face of such a response. One theory is that HCV has evolved a mechanism for abrogating a lymphokine action; another is that the virus inhibits CTL induction in vivo. It is possible that there may be immune-privilege sites in the host that have not been identified.
Another theory is based on recent work suggesting that there are CTL escape mutants in chronically infected chimpanzees. An otherwise conserved epitope in nonstructural protein 3 (NS3), which is the target for a strong CTL response, mutated over time, leading researchers to speculate that escape variants might emerge.
Immune Escape. Virtually every patient with chronic non-A and non-B hepatitis, whether from transfusion or IV drug use, has circulating levels of antibodies to the envelope glycoproteins gpE1 and gpE2. HCV can be difficult to grow in vitro, but work done in Japan suggests that the virus mutates over time to evade this humeral immune response. Using an RNA binding assay, researchers were able to show that serum taken from 1978 to 1982 contained antibodies that would neutralize HCV taken in 1977. In the same patient, however, the antibodies that neutralized the 1977 virus did not neutralize virus
taken in 1990. By 1991, the patient did have antibodies to neutralize the 1990 virus.
This finding was consistent with the process of immune escape. Further research revealed that there were multiple mutations between the 1977 and 1990 virus, as would be expected from an RNA virus. The rate of mutation is at the level of about .01 percent per nucleotide per year, but there was a cluster of mutations in a very small region at the terminus of gpE2. These mutations, between residues 395 and 487, resulted in a variety of nonconservative amino acid changes.
This region appears to be under severe immune pressure, and isolates of HCV from patients around the world show that the end terminal region of gpE2 is different in virtually every isolate, often involving nonconservative neoacid changes. This region is a target for B-lymphocytes, and a recent study in Germany indicates that antibodies to this end-terminal hypervariable region prevented binding of HCV inoculum to human fibroblasts. This leads researchers to speculate that the end terminus of gpE2 may be the principal neutralizing domain of the virus. Given the difficulty of growing HCV in vitro, it will take some time to confirm this.
Vaccine Development. Many of the approaches that might be pursued aren’t yet possible because of the early stage of the research on HCV.
The virus is very difficult to grow in tissue culture, and the only animal model is the chimpanzee, so it will also be difficult to attenuate the virus. There are few subunits circulating in these patients. As a result, there are only three approaches available at present: recombinant subunit vaccines, naked DNA vaccines, and vector DNA vaccines. Researchers are pursuing all three approaches in order to produce a very complete immune response. The following discussion focuses on subunit vaccines.
The subunit vaccine that has been tested in chimpanzees was derived by expressing the viral gpE1, gpE2, and NS-2 genes in mammalian cells, initially through the use of recombinant vaccinia donated by NIH. The pure subunits are combined with an adjuvant called MF-59 that has been used extensively in the herpes clinical trials. Usually the chimpanzee was immunized first with wildtype vaccinia, to rule out any residual live recombinant vaccinia in any subsequent protection, and then given up to 40 micrograms of subunits at months 0, 1, and 7. A total of 3 weeks after the final boost the animal was challenged with 10 infectious doses of homologous virus.
Out of 12 experimental chimpanzees, 5 were completely protected—there was no trace of viral RNA in the plasma, in the liver, or in the PBLs. Of the other 7 animals, 5 went through acute infection and resolved, after which there was no trace of virus. Only 2 experimental chimpanzees went on to develop chronic infection, and one of these had ameliorated acute hepatitis. By contrast, out of 7 unimmunized controls that were challenged with virus, 6 developed chronic infection following acute hepatitis and only one experienced resolution of the infection.
These results suggest that the vaccine not only gives some protection against infection but also causes resolution of infection when it does occur. Vaccination achieved a high level of immune response: antibody titers prior to challenge were higher than those usually seen in infected blood donors and patients with chronic hepatitis. Further analysis showed that complete protection against infection was correlated with antibody levels, but not with antibodies to the hypervariable region at the end terminus of gpE2.
However, researchers don’t know how efficient this subunit vaccine, based on HCV subtype 1A, would be against heterologous viral isolates. Initial experiments with a related viral isolate produced rather low levels of antibodies following reboost, and a slight inhibition of the onset of viremia. But it was impossible to conclude from this study that a vaccine based on HCV-1 antigens can prevent chronic infection by heterologous virus. Additional studies are ongoing.
Goals and Problems. The objective for a prophylactic vaccine would be to prime the humeral immune response, using recombinant gpE1 and gpE2 subunits to produce neutralizing antibodies that will restrict viral spread and load and thereby allow cell-mediated immune (CMI) response to clear the infection. In addition, it should also prime the CMI response using antigens to conserved viral proteins—such as the T-cell response to viral NS-3—to enhance its ability to resolve infection.
A constant problem is the knowledge that chimpanzees and humans alike appear to have weak immunity against reinfection. As a result, researchers are hoping to use naked DNA and vector DNA immunizations to broaden the immune response and increase the protective efficacy of the vaccines. At present, however, it seems unlikely that they will be able to achieve a vaccine as effective as those against hepatitis A and B. Still, it will be highly beneficial if they can ameliorate the disease by slowing down or preventing the progression to chronic hepatitis in a significant fraction of subjects.
In response to questions, Dr. Houghton added the following:
The first population to vaccinate would be high-risk groups such as health care workers, family members of patients, dialysis patients, etc.
Researchers have not yet identified major differences in the antibody or CTL response of patients who resolve following acute infection compared with those who develop chronic infection. Nor do they understand why they are getting resolution in experimental chimpanzees, although they suspect that the vaccine limits the viral load and spreads sufficiently to allow the host CMI response to deal with the infection.
The only studies showing resolution have involved HCV subtypes 1-A and 1-B. There seem to be few differences between these two subtypes, but it is impossible to say anything about resolution versus chronicity in the other major strains of HVC.
Incidence and Burden. Human papillomaviruses (HPVs) are extremely common and widespread. Genital HPV infection is a sexually transmitted disease with an incidence of 5 percent to 40 percent among sexually active women. HPVs are etiologically linked to genital cancers, especially cervical cancer, and infection with a high-risk HPV is by far the most significant risk factor for developing cervical cancer—more than 90 percent of cervical cancers contain HPV DNA. Cervical cancer is approximately the number-six cancer among women in developed countries but its the number-one killer from cancer in developing countries, where Pap smears are not readily available. On a worldwide basis, cervical cancer is the number-two cause of death from cancer in women, after cancer of the breast. It has been estimated that HPV infection is involved in approximately 15 percent of all human cancers.
Pathobiology. HPVs are small, nonenveloped DNA viruses that replicate in the nucleus of the host cell, leading to lesions, warts, and tumors. They are epitheliotropic: replication occurs only in epithelial cells, primarily in the differentiated layers of the epidermis. More than 70 different HPV genotypes have been identified and classified into three large groups according to the region of epithelia they tend to infect:
Cutaneous nongenital HPVs are commonly seen in dermatological practices and do not appear to have malignant potential. Several million cases are present in the United States at any given time. It is not clear that there would be demand for a prophylactic vaccine, although a therapeutic vaccine might well be important.
Epidermodysplasia verruciformis-specific HPVs include almost one-half of all known HPV types and are found principally in patients with a predisposition to develop widespread, chronic, nongenital lesions. Some of these patients go on to develop malignancies, but this is a very rare condition.
Mucosal HPVs include almost 20 different types identified to date. Some are low-risk types associated with nonmalignant disease, particularly HPV-6 and 11. Others are high-risk types that seem to have malignant potential, particularly HPV-16, 18, 31, and 45. In a study of viral DNA in 1,000 cervical cancers from around the world, HPV-16 is by far the most common type, although HPV-18 is also common in Southeast Asia. However, the DNA of multiple HPV types is found in both benign and malignant genital lesions, for which reason a vaccine against genital HPV will need to be polyvalent.
About 50 percent of cervical infections are with high-risk HPVs, but the majority are clinically inapparent and self-limited. Only about 10 percent of the infected women have cytopathological changes on Pap smears, and most of
these are early or mild dysplasias. Severe dysplasia is less common and tends to occur in older women, as though there were changes in the cells during persistent infection over a period of years that leads to dysplasia. Presumably this includes virus-specific changes, such as integration of the viral DNA into the host genome, and possibly some cell-specific changes as well.
Viral Gene Products. In cell lines derived from cervical tumors, there is a preferential retention and expression of three oncogenes from the high-risk HPV. Two genes (E6 and E7) in collaboration with each other can establish immortalization of human keratocytes grown in culture: the E6 product binds to and inactivates the p53 tumor-suppressor protein; the E7 product binds to and inactivates the pRB tumor-suppressor protein. These properties are significantly more active in high-risk than in low-risk HPVs. A third transforming gene of HPV is E5, whose product appears to activate the growth-factor receptors; it is not always found in the malignant tumors themselves, but it may play a role in initiating the lesions.
Animal studies suggest that immunization with E6 and E7 does not prevent lesions, but the lesions do regress faster. As a result, trials are now being conducted to evaluate a therapeutic vaccine based on differing combinations of E6 and E7.
Barriers to Vaccine Development. Several problems limit attempts to understand HPV infection and develop a vaccine. Perhaps the most important is the lack of a system to propagate HPV in vitro. Second, researchers have not been able to produce preparative quantities of purified viral capsid proteins. Third, immune parameters have not generally correlated with the benign infection antibodies to E6, E7, E2, and other viral proteins. In addition, there is no animal model of HPV infection—the productive infection is species-specific, although there are several animal systems that might eventually be used as models.
Approaches to Vaccine Development. Most attention has focused on the development of a subunit vaccine. Work at the Laboratory of Cellular Oncology at the National Cancer Institute (NCI) has focused on the two viral structural proteins, major capsid protein L1 and minor capsid protein L2. Researchers found that when they expressed the bovine papillomavirus (BPV) L1 gene in recombinant baculovirus in insect cells, it was sufficient to cause self-assembly of virus-like particles (VLPs) —essentially empty capsid that do not contain the viral genome. When L2 was added to this insect cell system, there was no major differences in immunogenicity, but there was more efficient assembly of VLPs. This self-assembly process can be carried out with HPV, producing VLPs that do not contain the HPV genome. The BPV system now provides the major in vitro test for infection, a conformational test for antibodies to HPV-16 VLPs.
When rabbits are immunized with authentic BPV virions, it induces very high levels of neutralizing antibodies, more than 105 titer. VLPs, whether L1 or L1-plus-L2, induce similarly high levels of neutralizing antibodies. When the
VLPs are denatured before immunization, however, they fail to induce the neutralizing antibodies. From this it is possible to conclude that the neutralizing antibodies are directed against conformationally dependent epitopes on L1 that mimic authentic virus. This model seemed most relevant to the development of prophylactic rather than therapeutic vaccine.
In collaboration with the Pasteur Institute, NCI researchers tested this model in a vaccine for cottontail rabbit papilloma virus (CRPV). A total of 39 experimental rabbits were immunized 3 times with VLPs containing CRPV L1 or L1-plus-L2 (in Freund’s adjuvant) or L1-plus-L2 (in alum) and then, with 39 controls, challenged with high-dose virus. The results indicated that 90 percent of the controls developed the expected disease, whereas 90 percent of the experimentals developed either no disease or a mild regressing disease that was very self-limited, and only one experimental developed persistent disease. In addition, about 40 percent of the controls developed invasive squamous cell cancer, but none of the experimental animals developed this. Importantly, animals inoculated with bovine L1 and L2 became infected despite high levels of BPV antibodies; the protection against CRPV was type-specific.
Researchers were able to transfer protection passively by transferring serum from immunized to unimmunized animals, thereby confirming that the neutralizing antibodies are the protective mechanism. Other laboratories have demonstrated different techniques for immunizing rabbits, including DNA transfer and the use of vaccinia vectors. The common denominator is the ability to obtain the appropriate conformationally dependent epitopes for viral capsid protein L1.
One drawback of this CRPV model is that it involves a cutaneous infection, whereas the genital HPV infections associated with cervical cancer are mucosal infections. However, an experiment with BPV type 4, which causes oral mucosal disease in cows, showed protection similar to that seen in the CRPV cutaneous system. Another experiment with canine oral papillomavirus also demonstrated very substantial protection from mucosal infection using L1 VLPs that had been expressed in insect cells. In the latter case, researchers were able to wait for as much as a year after immunization before challenge and still obtain very substantial protection, indicating a considerable duration of immunity in this mucosal model. A Rhesus monkey papillomavirus model would be superb, since it would be a mucosal infection and some monkeys develop invasive cervical cancer; however, the virus is not yet available, so this model remains theoretical rather than actual.
Serological Assay. Because there are so many types of mucosal HPVs, the vaccine against genital HPV infection should in principal be polyvalent. Some of the relevant HPVs will probably represent different serotypes. For example, genetic classification according to DNA homology shows that HPV-6 and 11 are closely related; HPV-16 and 31 are closely related; and HPV-18 and 45 are closely related. Whether they represent distinct serotype, however, remains to be established. The serotyping of HPVs is constrained by the unavailability of the appropriate conformationally dependent epitopes for the various types.
NCI researchers have developed an ELISA assay for VLPs of HPV-16, based on reaction to the conformationally dependent epitopes of the L1 protein, but there seems to be some crossreactivity to other high-risk types. That is, women who are infected with HPV-18 and 31 are more likely to be positive on this assay than women infected with the low-risk HPV-6 and 11. The latter show little difference from uninfected women, so the assay does not seem to be detecting antibodies that would be directed against low-risk HPV types. In a prospective study, this assay proved to be about 90 percent accurate in measuring current or past infection with HPV-16, irrespective of cytology, with a maximum false positive of about 3 percent.
This ELISA assay for HPV-16 has been used to look at different groups with various cancers. As expected, there is a highly significant odds ratio for being positive on the ELISA and cervical cancer. However, there is also a significant odds ratio with urethral cancer, vulvar cancer, and possibly, cancer of the esophagus. The association with esophageal cancer is still controversial, in part because most studies fail to find HPV DNA in these cancers. There is well-documented evidence that BPV in conjunction with a carcinogen, is responsible for esophageal cancers in cattle. The viral DNA may be absent from human esophageal cancers because a “hit and run” mechanism is at work.
Hemagglutination Assay. To provide additional information, and to overcome some of the disadvantages of the serological assay, researchers have developed a very sensitive hemagglutination assay using BVP and VLPs. Data published about 20 years ago indicated that incubation with BPV causes agglutination of mouse red blood cells. Researchers found the same response to L1 and L1-plus-L2 VLPs, and that the response is sensitive to differing concentrations and combinations and BPV proteins. Similar results were obtained with other papillomaviruses, including CRPV and several varieties of HPV.
Further research led to the development of a sensitive hemagglutination inhibition (HI) assay, in which the presence of antibodies to papillomavirus will disrupt the viruses and prevent agglutination. The HI assay is type-specific, in that BVP antibodies will prevent hemagglutination in response to VLPs of BVP but not HPV-16, and conversely HPV-16 antibodies will give a negative response for VLPs of HPV-16 but not BVP. This assay has been used to show that rabbits immunized with disrupted VLPs were not protected from CRPV, despite high antibody titers on the ELISA test, while rabbits immunized with intact VLPs were protected. This assay should be useful in measuring natural exposure to papillomaviruses, evaluating immune response after VLP vaccination, and investigating cross-protection from heterologous types after VLP vaccination.
In response to questions from the audience, Dr. Lowy added the following:
Researchers are studying the role of E7 protease in inducing CTL response—for example, inducing CD8-positive T-cells—but it unclear whether these cells are effective in viral infection or in the treatment of existing human cancer.
When these studies move into humans for efficacy testing, a logical clinical endpoint would be cytological abnormalities—not cervical cancer. Investigators could also monitor DNA in the cervix.
There is as yet no data on mucosal routes of administration, although this will be an important issue in the future.
There is as yet inadequate knowledge of the epitopes of the various papillomavirus subtypes to assess the possibility of making chimeric VLPs containing L1 proteins from multiple serotypes.
The timetable is uncertain. Some pharmaceutical companies may try to go into Phase I clinical trials in the next couple of years, but it will take a period of time to demonstrate immunological reactivity and safety, and a longer time following women to demonstrate efficacy. Controlled Phase III trials might be completed in more than 5 but less than 10 years.
The target population for a prophylactic vaccine would be women, hopefully before they become sexually active.
DENGUE HEMORRHAGIC FEVER5
Pathobiology. Dengue is a positive-strand RNA virus, like HCV a member of the flavivirus family, and is transmitted by mosquitoes. There are four serotypes, and infection with one serotype results in long-lived immunity against that serotype, as well as sensitization to other serotypes. Dengue virus infection presents as two clinical syndromes: dengue fever (DF) and dengue hemorrhagic fever (DHF). Several days pass between transmission and the appearance of fever. In uncomplicated DF, the fever lasts for 3 or 4 days and then resolves successfully. In a few cases, however, as the fever is resolving, the patient develops capillary leak syndrome, which is a more serious illness commonly called DHF. There is little real hemorrhage but a lot of leakage from the capillaries, and hematocrit rises and body temperature falls. In its most severe form, DHF can result in shock and death. Depending on the speed of diagnosis and treatment, mortality can be anywhere from less than 1 percent to as high as 20 percent. Fortunately, most children respond well to aggressive volume expansion.
Incidence and Burden. Dengue fever (DF) is a frequent childhood infection in tropical and subtropical regions of the world where the mosquito Aetes aegypti is prevalent. There are an estimated 100 million cases per year,
based on the expectation of a 10-percent annual infection rate among children in endemic areas.
The far more severe form of dengue hemorrhagic fever (DHF) has been recognized as a problem in Southeast Asia for the past 40 years. During the 25-year period 1956–1980, there were 3 million cases of DHF in Southeast Asia and 20,000 deaths (0.67 percent). During the first half of the 1980s there were 800,000 cases and 10,000 deaths (1.25 percent). During the second half of the 1980s, there were 900,000 to 1 million cases and 10,000 to 11,000 deaths (1.00 to 1.22 percent).
In the past few years, DHF has been occurring with increasing frequency in other parts of the world, including Cuba and the Caribbean, Central America, and South America—areas that previously had DF but not DHF. The reasons for this are not known. One possibility is a change in the virulence of the various strains of the virus. For example, the dengue type 2 virus that is present in the Caribbean today is different from what was there in the past and more closely resembles the type 2 genotype of Southeast Asia, which may have been transported to the Western Hemisphere.
Epidemiology. However, the current hypothesis, based on epidemiological data, suggests that increased risk of DHF comes from pre-existing infection rather than a more virulent virus. Studies of children aged 1–14 in Thailand indicate that there is an extremely small risk of developing DHF during primary infection with any of the four serotypes. In secondary infection, the vast majority have asymptomatic, self-limited DF, but there is about a 50-fold increased risk of developing the more severe form of DHF. However, DHF is very seldom seen during the first 2 years of life.
This is confirmed by data from Cuba. The island had been free of dengue for generations until an outbreak of dengue type 1 in 1976–1977. The country mobilized in anticipation of an outbreak of DHF, but they observed very little serious illness and no deaths. Then, in 1981, there was an outbreak of dengue type 2 and many cases of DHF, mostly of which occurred in children aged 3 to 12—those who had been sensitized by the preceding type 1 infection. Only one child born between the two outbreaks developed serious disease.
Immunopathogenesis. Recent research has emphasized the immunopathogenesis of dengue, and specifically the concept of antibody-dependent enhancement, rather than vaccine development per se. The current hypothesis is that, during secondary infection, previously existing crossreactive antibodies are binding the dengue virus and helping it into FC receptor-bearing cells. The receptor hasn’t been defined yet, but it seems clear that human monocytes would be the most permissive cell, and that binding with non-neutralizing antibodies will increase the number of infected cells about tenfold. That is, only about 2 percent of human monocytes will normally become infected with dengue virus, but if the virus is bound with crossreactive non-neutralizing antibodies, about 20 percent become infected, and if FC receptors are up-regulated by exposure to
interferon-gamma, perhaps 60 percent become infected, with virus yields going up proportionately.
In this case, memory T-cells and CD8 T-cells are present in high numbers because of the recent infection with a crossreactive strain. These T-cells are stimulating their MHC molecules, and as a result they wind up acting at least in part as antigen-presenting cells. They also bear cross-one, cross-two FC receptors on their surface; secondary infection is more efficient because crossreactive antibodies bind to the viral protein and help it enter via the FC receptors. The result is more infected monocytes, which function as virus factories.
Activated T-cells produce cytokines and other products that can affect capillary action, and killed monocyte may also release vasoactive compounds. The target cell for dengue infection appears to be the RE monocyte, and virus antigen is not seen in endothelial cells. Researchers currently believe that endothelial cells are leaking because of cytokine alteration of their function. Hence, research is focusing on the serum, T-cells, and other peripheral monocytes in order to understand the basic mechanism of immunopathogenesis.
Human T-cell Responses to Dengue. The three-dimensional structure of an HLA class II MHC molecule shows a so-called A-2 crypt that holds an endogenous peptide—the antigen. In the case of the dengue virus, CD4 T-cells initially recognize the infection in the form of dengue peptides being presented in HLA class II molecules on the surface of antigen-presenting cells.
Following primary infection with dengue type 1, stimulated peripheral blood monocytes will respond most strongly to the dengue-1 antigen, but there is also a lesser response to the three other serotypes. Analysis shows that about 1 cell in 1,000 will be positive for dengue-1, and about 1 cell in 10,000 will be positive for dengue-2, 3, and 4. To determine the smallest amino acid sequence that CD4 T-cells would recognize, researchers identified and reproduced these crossreactive T-cells using limiting dilution cloning. They then conducted a series of experiments in which the cells were infected with vaccinia virus that expresses the dengue protein. As the protein was truncated, peptide analysis identified the amino acid sequence to which the CD4 T-cell clone responded—in this case, viral nonstructural protein 3 (NS-3).
The results confirm the polymorphism of T-cell response to dengue virus NS-3. As with hepatitis C, dengue NS-3 apparently contains multiple T-cell epitopes with very different specificities. In the case of a subject who had been immunized with dengue-3, all T-cell clones recognized the epitope for dengue-3, but with four degrees of crossreactivity: (1) just dengue-3; (2) dengue-2, 3, and 4, but not 1; (3) dengue-1, 2, and 3, but not 4; and (4) dengue-1 through -4 plus West Nile and yellow fever virus, both of which are also flaviviruses. In another subject immunized with dengue-4, analysis found T-cell clones that reacted to (5) just dengue-4; (6) dengue-2 and 4; and (7) dengue-1, 2, 3, and 4. Within each serotype, there were three or four distinct epitopes on the viral NS-3.
This polymorphism—multiple T-cell epitopes to the viral protein—probably is not unique to dengue. But in dengue it contributes to the T-cell activation level because of the epidemiology of the close circulating viruses.
Comparison of sera from children with DF, DHF, and controls showed significant elevation of soluble CD8 and IL-2 receptor, in those with DHF, evidence of a marked activation of CD8 T-cells and IL-2 receptor-bearing cells. Soluble CD4 was elevated to a lesser degree. Interferon-gamma was elevated in both DF and DHF.
Based on these results, researchers have mounted a prospective study of children recruited from a Thai outpatient clinic. Subjects had fever of 24 to 72 hours duration and no other source of infection. During the first year, 180 children were recruited; 60 were subsequently diagnosed with DF and showed dengue-positive antibody response, and 29 developed DHF. The study is not yet complete, but several interesting findings have emerged. One is that 100 percent of the children with dengue had viremia, while the literature suggests that the positive isolation rate would be 50 percent for DF and 25 percent for DHF. This finding actually makes more sense, in view of the theory that enhancing antibodies were contributing to the increased risk for severe disease. Researchers plan to titrate plasma to see whether the viral burden is greater in children with DHF.
Using T-cell clones from these children, researchers also plan to measure the cytokine production (particularly interferon-gamma) and the selective activation of T-cell receptors. The activation of specific T-cell receptors is of interest in regard to the pathogenesis of many infectious diseases. In previous studies, stimulating with dengue led to the preferential activation of V-beta-17-bearing T-cells, and the majority of dengue-specific T-cell clones also bear the V-beta-17 receptor. It is unclear whether this is an integral part of DHF, but it bears further investigation.
Vaccine Development. Classical studies by Sabin showed long-lived immunity to homologous virus serotype, so the concept of vaccination seems reasonable. The concern is to prevent DHF and not sensitize a population that might later be exposed to a different serotype and find itself at greater risk for more serious disease. For this reason, there is consensus that a successful vaccine must induce immunity to all four serotypes. From a public health point of view, it is also very important to immunize with one dose of a polyvalent vaccine, if at all possible.
A group in Thailand has performed Phase I immunogenicity studies in adults, using live attenuated strains of dengue-1 through 4. They believe that there is adequate immunogenicity in terms of neutralizing antibody response, although enhancing antibody response has not been measured. The group is now beginning studies with combination vaccines, but it is too early to say what the results will be. Polyvalent vaccination appears to be more difficult in children, possibly because of interference between serotypes, and a number of additional studies will be required. Lots of vaccines are being prepared for clinical trials, and there is an expectation that efficacy studies might be conducted by the end
of this decade. However, far more information is needed on the safety and antigenicity of the polyvalent vaccine.
Other approaches under investigation might also provide good results, including subunit vaccine, plasma DNA vaccine, and infectious cloned vaccine (e.g., chimeric dengue-4 backbone), and recombinant dengue vaccine containing NS-1 (which is a potent antigen in experimental vaccines against yellow fever). However, there appears to be no practical way to produce neutralizing antibodies without at least some enhancing activity; this will be a safety concern in all dengue vaccines.
Another major concern is the duration of protection. Most cases of DHF occur between ages 3 and 15, at which point children seem to develop an immunity. Hence, dengue vaccines should have a very solid long-term memory, with the ability to induce both B-cell and T-cell responses, as well as good antibody response.
One of the biggest barriers to vaccine development is that there is no suitable animal model for DHF. One team looked at over 100 primate species and couldn’t find a model. Primates can be infected, and they develop viremia, but they don’t progress to the more serious syndrome of DHF.
In response to questions from the audience, Dr. Ennis added the following:
Although T-cell activation is V-beta-17 specific, it does not appear to be a super-antigen effect. Instead, host factors such as HLA seem to be contributing to increased risk. HLA typing of the children in Thailand should shed light on this question.
Antigenic variation doesn’t seem to be a major issue. There are four serotypes, but within the serotypes there doesn’t seem to be much drift in the antigens or epitopes.
Immune response to primary infection is serotype-specific, with antibodies appearing after a day or two, about when the virus is cleared. At the beginning of a secondary infection, therefore, the only antibodies that will recognize the new virus will be the crossreactive antibodies. Hence, the triggering of the increased number of infected cells takes place before the antibody response.
The risk of DHF increases with secondary infection, but it does not appear to increase further when the subject is infected with a third serotype. In fact, risk seems to go down after the second infection, possibly because of cross-neutralization.
While there is a strong emphasis on developing a quadravalent vaccine, additional studies will be needed to develop a combination vaccine that has 95 percent neutralizing-antibody responses to all four serotypes, instead of 60 percent for one and 95 percent for another.
Greater knowledge about the polymorphic nature of T-cell response should make it possible to engineer vaccines that contain T-cell epitopes as well as neutralizing epitopes.
Once considered to be a virus, Chlamydia is in fact a bacteria, although distantly related to other familiar bacteria. The genus is composed of three species:
C. psittaci is a common pathogen on a variety of mammals and birds. It causes psittacosis, or parrot fever, an opportunistic respiratory infection that can be life-threatening when transmitted to humans. However, this disease has been largely controlled by regulation of the trade in exotic birds.
C. pneumoniae is a new and emerging disease. It is a frequent cause of respiratory infections and associated with pneumonia, especially among young adults in northern Europe. It is also associated with arteriosclerosis, although the relation is not yet known.
C. trachomatis can be subdivided into three biovariants that cause very different kinds of disease: (a) the mouse pneumositis biovariant is not a pathogen of humans; (b) the lymphogranuloma venereum (LGV) biovariant, which consists of three serotypes that proliferate in the lymph nodes; and (c) the trachoma biovariant, which consists of 15 different serotypes that cause epithelial infections of the eye or of the genital tract.
Incidence and Burden. As an ocular infection, C. trachomatis causes trachoma, which remains the leading cause of preventable blindness in the world, affecting some 200 million people. Trachoma is no longer encountered in the United States, but exactly the same strains cause genital infections that have made chlamydia the fastest growing sexually transmitted disease (STD) in this country, with over 5 million new cases per year. (By comparison, there are only about 235 cases of LGV, although it remains an important STD in developing countries.) Most of these 5 million cases present as either nongonococcal urethritis or cervicitis. Significantly, many of these cases are largely asymptomatic, which is a problem both for treatment and for identifying individuals at risk for more serious sequelae.
Among men, chlamydial urethritis is a common cause of epididymitis, but the real focus of attention is on women. In addition to the complications of urethritis and cervicitis, infants born to infected women are also at risk for conjunctivitis and chlamydial pneumonia. Because so many infections are asymptomatic, they often go undiagnosed and develop into more serious problems. One of the most serious is pelvic inflammatory disease, which can lead to chronic pelvic pain, involuntary sterility, or ectopic pregnancy.
Diagnosis, Treatment, and Prevention. Over the past 10 years, researchers have improved substantially the ability of clinicians to diagnose chlamydial
infections, moving from self-culture-based assays to culture-independent assays. Nevertheless, even the best molecular assays to date have only a 70 percent sensitivity in high-prevalence populations or symptomatic infections, and less than that in low-prevalence populations or asymptomatic infections. Once diagnosed, chlamydia can be treated with antibiotics and there are as yet no signs of antibiotic resistance, although the infection hasn’t really been challenged to date.
Behavior change might be one preventative strategy, but like other STDs, chlamydia is predominantly an infection of teenagers and young adults, with all the problems that entails in terms of recognizing and changing high-risk behavior. There are two biological strategies for preventing the disease: one is the development of a competitive analogue, the other is the development of a vaccine.
Pathobiology. The chlamydiae are nonmotile, obligate intracellular parasites with a distinctive life cycle. The extracellular forms can be detected by direct fluorescent antibody and enzyme immunoassay, but while they are infectious, they are metabolically inactive. The extracellular forms attach to a cell by a mechanism that is not yet understood and, once attached, enter the cell by endocytotic mechanisms. Inside the cell, it remains inside an inclusion—a membrane-bound vacuole that inhibits fusion with lysosomes—throughout its growth cycle.
Chlamydiae can propagate only inside a cell. Once inside, it undergoes changes in its membrane and DNA and begins to differentiate. Researchers believe that it follows one of two pathways: it can multiply and grow like other bacteria, or it can remain in a cryptic state. The latter, which is one of the mechanisms by which this organism persists in the host, is poorly understood. If it multiplies, it will divide by binary fusion some 24 to 72 hours after infection, after which it again differentiates from the vegetative to the elementary or infectious form and is released, by another unknown mechanism, to infect new cells. Importantly, chlamydia does not seem to be able to transmit by cell-to-cell interaction but must leave the cell and thus be exposed to antibodies and other host activity.
Chlamydia is a mucosal pathogen that does not so much kill host cells as cause persistent inflammation. The inflammation causes scarring in both the ocular and genital models, and the scarring accounts for the disease problem—for example, deformation of the eyelashes and blindness. Infection of epithelial cells leads to a potent gamma T-cell response and the excretion of inflammatory cytokines, especially IL-8, which in this case fits the requirements for a mediator for pathogenesis. Other pathogens elicit a similar response early upon uptake, which then fades away; whereas with chlamydia it occurs late in the cycle and endures throughout the infection process.
Adhesion Molecules and Analogues. Researchers believe that the parasite attaches and enters the host cell using a carbohydrate-like ligand that is synthesized on the surface of the chlamydia and somehow mediates entry. This ligand is similar to the heparin sulfate molecule and apparently can be cleaved
with an enzyme that is specific to heparin sulfate. When this occurs, the chlamydia decreases or loses attachment and is no longer infectious. When exogenous heparin or heparin sulfate is added, the chlamydia regains its infectivity; indeed, the level of attachment increases by some 50 percent.
Researchers have been able to model this mechanism by coating polystyrene beads with either the natural ligand or heparin sulfate. The beads attach to and enter epithelial cells very efficiently in a manner similar to chlamydia. They also elicit the same pattern of tyrosine polyphospholated host-cell proteins upon uptake, suggesting that they are mimicking the chlamydia pathway. This mechanism may provide a way to target delivery of vaccines to the mucosal surfaces that chlamydia infect.
Understanding of these adhesion molecules also suggests the possibility of a second antimicrobial strategy, that of using analogues to compete with and block the binding activity of chlamydia. Researchers experimented with different forms of heparin and discovered that the N-desulfated, O-sulfated form of the molecule would inhibit binding to the eucaryotic host cells without binding to the chlamydia elementary bodies. This suggests that one could design compounds that would not bind to the organism (which would risk potentiating or rescuing infectivity) and would be potent competitors for the host-cell receptor for this ligand. Such a chlamydia inhibitor might be added to various spermicidal compounds.
Immunology. Epidemiological data, animal models, and early vaccine trials demonstrate serotype-specific immunity to the 15 known serotypes of C. trachomatis var. trachoma. Immunity is relatively short-lived, typically waning in 6 months to 1 year. It is commonly believed that immunity is related to serotype-specific antigens. Three molecules on the surface of chlamydia are candidates for these targets: (1) the lipopolysaccharide (LPS), (2) the major outer membrane protein (MOMP), and (3) the adhesion-and-invasion ligand (see above).
LPS is genospecific, not serospecific, and neutralization cannot be demonstrated, so it doesn’t fit this serovariant-specific model. MOMP, on the other hand, is known to carry serovariant-specific antigens, and antibodies to these molecules neutralize infectivity. Even in polyvalent sera there is a big difference between homologous and heterologous neutralization. Unfortunately, MOMP is the only major surface component that is well understood. There may be other antigens, and this area is worthy of continued investigation.
C. trachomatis has only one gene for MOMP, but the gene and the protein vary across strains. Variation in amino acid sequence occurs in each of four variable sequence (VS) regions called VS1 through VS4; changes are most frequent in VS4 and least frequent in VS3. CD4 T-cell determinants have also been mapped in this protein, including two major ones, one in a conserved region between VS1 and VS2, the other overlapping with VS3. Since there seem to be
no antibodies to the VS3 region, however, the serovariant-specific antigens are to be found in VS1 and VS2 or the more broadly reacting antigens of VS4.
Synthetic (linear) peptides from these regions elicit a strong immune response, both independently and in expression vectors such as poliovirus and vaccinia. Relatively little of this response confers on the organism, however, so there is no microbiologic protection. An increasing amount of data suggest that conformational-dependent determinants are important for this target molecule. In VS1, for example, a given 12-amino-acid sequence yields a peptide that is sometimes recognized in linear form but by other serovariants only in circular form. Similar variance occurs in VS2, in which there is good antiserologic response to circular peptides, but not linear.
CD4 and T-cell help is also essential for protection and resolution of infection. A variety of proteins can elicit stimulation for CD4 cells, including MOMP. Animal models also suggest that CD8 cells and cytotoxicity are also important in resolving infection, and possibly in protection. Recent studies have detected CD8-killing of chlamydia-infected epithelial cells in vitro, but it is unknown whether this involves MHC class IA or IB presentation. Research in this area is ongoing.
Issues in Vaccine Development. Unlike some of the viral infections, there are several animal models for chlamydia, including mice, guinea pigs, and monkeys. Murine models include respiratory, systemic, and even genital tract models, including infection with human strains of the pathogen. The guinea pig model is limited to a strain that is a natural pathogen to guinea pigs. Both ocular and genital tract models have been developed in monkeys.
At present, however, there is a need for a better understanding of the cellmediated immune mechanism, the role of CTLs and antibodies. There also needs to be a clearly differentiated understanding of the role of conformation determinants, the exact nature of the antigenic targets, and how to deliver them. And one of the biggest challenges is to find a way to genetically transform and manipulate these organisms, which currently cannot be grown in the quantities needed to support research.
In response to questions, Dr. Stephens added the following:
The time frame for developing a vaccine against either MOMP or the adhesion ligand is entirely a function of money. Most of the problems are fairly straightforward; given enough money, the time frame would be quite reasonable.
Differences in immune response to chlamydia peptides suggest that a small difference in primary sequence may result in a very large difference in the tertiary or quaternary structure of the molecules. In the guinea pig model, protection is totally dependent on conformational determinants. This suggests that researchers need to be more sophisticated in how they look at and present the antigens.
The hypothesis that heat shock protein 60 (HSP-60) plays a role in the pathogenesis of chlamydia has been called into question by recent experiments showing that immunization with HSP-60 makes no difference in response to
infective challenge. In other words, this is not a delayed-type hypersensitivity mechanism, and while HSP-60 may still play a role, it is not responsible for the persistence of inflammation.
While there may be 15 serovariants, typically 1 or 2 serotypes will be dominant in any given population. Researchers don’t understand why, and indeed they can’t always differentiate among strains. Nevertheless, in terms of vaccine development, this makes for a much simpler “cocktail”, rather than trying to include all 15.
Studies of antibody passive-immune therapy were conducted before a neu-tralizing assay was available and produced ambiguous results. These studies need to be repeated in light of new understanding of antibody response at mucosal sites.
Most researchers would agree that the route of administration should be mucosal. It is unclear whether intranasal, rectal, etc., would elicit the best response.
MOMP is down-regulated by interferon (IFN) gamma, but only in 3 of the 15 serotypes.
Incidence and Burden. A third of the world’s population is thought to be infected with tuberculosis, and 10 million die from TB each year. In the United States there are an estimated 10 million people infected with TB, and the numbers of infections and deaths has been rising rather than falling—there were more than 50,000 excess deaths between 1985 and 1992. And while a number of effective chemical therapies are available, clinicians are seeing increasing numbers of multiple-drug-resistant strains of TB. The massive epidemiology of this disease demands an effective preventative vaccine.
Problems with Existing Vaccine. Unlike many other infectious diseases, there is already a vaccine for tuberculosis (TB) —the Bacillus Calmette-GuJrin (BCG) vaccine, developed over 70 years ago in France. But while BCG is one of the oldest vaccines, it is also one of the most controversial. Its effectiveness varies considerably, and there is concern over the potential variability among the various strains of the vaccine that have developed over the years. Consequently, there is growing interest in finding a possible replacement for BCG as a vaccine against TB.
BCG is made with attenuated Mycobacterium bovis, the other species that, like M. tuberculosis, can cause TB. It was developed in the 1920s, but was never cloned and, as it passed through different laboratories throughout the world, a lot
of potential variability has developed among the different strains. Over 3 billion doses have been given worldwide, by multiple routes, and BCG has been shown to be an extremely safe vaccine, even in infants. The original route was oral, but in the past 20 or 30 years the intracutaneous route has become more common.
The efficacy of BCG in producing immunity ranges from 0 percent to 100 percent, depending on trial or study. A number of factors have been postulated to explain this variability. Trial methodology varies considerably and makes it difficult to determine the statistical validity of the results. Different strains of BCG were used in the various trials. Different routes and doses may also have contributed to variability. Crossreactive immunity to environmental mycobacteria may have masked or biased the protective effect of the vaccine. Background rates of TB were sometimes too low to show a significant effect.
A 1994 meta-analysis of the published literature on BCG efficacy found that, overall, BCG was 51-percent protective against pulmonary TB, even more effective against disseminated forms of the disease, and 71 percent protective against death. Comparison of 13 prospective trials and 10 case-controlled studies suggested that neither BCG strain nor age at vaccination was an important variable in terms of efficacy. However, geography was important: the further away from the Equator, the higher the efficacy of BCG vaccination. (This may be related to higher endemic rates of environmental mycobacterial colonization and infection in warmer climates.) Study design was also important: the higher the data validity, the higher the efficacy of BCG.
Special Problems in Developing TB Vaccines. Many of the problems with the current vaccine have to do with the special conditions of TB as a disease, rather than with BCG itself. For example, researchers are greatly hampered by their inability to measure infection rates in vaccine trials with BCG. The best way to identify infection is the purified protein derivative (PPD) delayed hypersensitivity test; unfortunately, BCG induces a positive PPD response. In addition, only 10 percent of people that become infected with M. tuberculosis go on to develop disease, and this is often associated with suppression of the immune system. Finally, the long latency of the disease is also a problem; efficacy trials require long-term follow-up, which presents its own difficulties.
It seems obvious that the most important immune response to concentrate on is mycobacteria-specific memory, the only response that could be induced by a vaccine that would persist in vivo and protect against a rechallenge. Much of the in vitro research has not taken this into account. Mycobacteria produce adjuvant-like effects, and it important to have negative lymphocytes as a control or baseline, in order to determine whether an in vitro response is related to memory immune response rather than to an adjuvant or a superantigen-like response that occurs only in vitro.
Four major mycobacteria-specific antigens have been studied as potential vaccine candidates over the past 20 years:
Heat shock proteins were seemingly dominant in immunoscreening strategies, but many researchers now think that they may be involved in pathology and that it would not be a good idea to use them as a subunit vaccine.
Actively secreted protein component of mycobacterium may provide the most protective antigens, and this has drawn increasing interest in the past 5 or 10 years.
Isopentenyl pyrophosphate and other pyrophosphates can stimulate gamma-delta cells, but it is not known whether this immune response can develop a memory.
Mycobacterial lipids can stimulate the so-called CD4-negative, CD8-double-negative T-cell populations, but again, it is not known whether this immune response can develop a memory. It may have a place in immunotherapy strategies.
Pyrophosphates and lipids will probably prove to be important in therapies where the immune response is being induced in the face of disease.
Interest in actively secreted proteins originated with the finding that live vaccines work better than killed vaccines. It has been shown that culture filtrate proteins, the antigens secreted by mycobacteria in vitro, can induce an immune response in mice and guinea pigs and can protect against experimental challenge. Three of these antigens are: MPT-57, which is similar to a heat shock protein; antigen 85 complex (Ag85), probably the hottest candidate (see below); and a so-called less-than-10 kD moiety, which is also secreted by the mycobacterium. Ag85 complex is currently going into clinical trials, but there has been no good comparison of culture filtrate proteins with BCG.
Animal models provide strong evidence that Th-1 CD4-positive cells are important in protective immunity. These cells produce interleukin 2 (IL-2) and interferon gamma (IFN-gamma), both of which activate macrophages to kill intracellular organisms. These findings have been confirmed in the three human models of protective immunity:
People with tuberculosis pleuritis are thought to have an immune response that is controlling the infection, since most of them go on to resolve the disease. Purified lymphocytes from the pleural fluid of these patients produced high levels of IFN-gamma in response to a protein from the cell wall of mycobacterium and to the less-than-10-Kd secreted moiety. This suggests that Th-1 cells are important in this model.
People who are PPD-positive but healthy also appear to have protective immunity, since they have not developed disease. Comparison with TB patients showed that these individuals showed a predominant Th-1 response: elevated levels of IL-2 and IFN-gamma, and depressed levels of IL-4. This suggests that it may be important both to induce Th-1 response and to inhibit Th-2 response.
People who have been vaccinated with BCG also demonstrate a Th-1-like response. In this case, however, there is a stronger response to whole lysate of mycobacteria (WL) than to culture filtrate (CF), suggesting that BCG induces an immune response to cell-associated antigens rather than to the actively secreted proteins of the CF.
The latter finding may explain why BCG fails to provide as much protection as clinicians might wish. Researchers offer three possible explanations. First, BCG does not induce a predominant immune response against the secreted antigens. Second, there may be other, possibly unidentified antigens that should be considered in developing a vaccine for TB. Third, and consistent with other data, immune response to mycobacterium is highly heterogeneous, making it difficult for a single subunit protein to induce the same levels of immunity that would be induced with multiple antigens or with a whole, attenuated organism.
Routes of Vaccination. Researchers compared the above findings, which were observed following intradermal vaccination, with the immune response following percutaneous vaccination. They found that intradermal vaccination induces statistically increased positive PPD response. Both T-cell proliferative response and T-cell IFN-gamma response to the whole lysate were significantly up-regulated following intradermal vaccination, but not after percutaneous vaccination. These findings are consistent with the finding that, despite the fact that intradermal uses one-tenth as much vaccine, it produces better immune responses.
Since the TB infection is mucosal, however, the vaccine will have to induce good mucosal immunity. Experiments with the mouse model indicate that oral vaccination is a good method of inducing both mucosal and systemic immunity. Internasal vaccination with BCG can also induce immune responses, and this is an important area for future research. It may be important to use a combination of intradermal and oral immunizations in order to induce a Th-1 response systemically, and a Th-2 response locally, within the mucosal surfaces.
Researchers are currently one-third of the way through a double-dose escalation trial using increasing doses of BCG orally versus a placebo; data will be available in 12 to 18 months. In addition, researchers at Colorado State University are doing a study of a mucosal homing molecule that would be expressed after BCG vaccination.
Possible Improvements in TB Vaccines. Among potential subunit vaccines, Ag85 is the best-characterized. It is known to protect guinea pigs, and it is going into clinical trials in humans. Human T-cell epitope mapping studies have been published. Preliminary results using purified Ag85 indicate that it does not stimulate an immune response from BCG. However, there are still no good comparisons of subunit vaccines with BCG itself.
Recombinant BCG vaccines are another possibility. For example, it might be possible to clone high-expression genes for secreted proteins into BCG, thereby inducing better protective immunity. An alternative would be to clone cytokine-expression genes into BCG, thereby potentially inducing higher Th-1
response and lower Th-2 response. A third possibility would be to delete a nonessential antigen from BCG in order to use that antigen, or its absence, as a diagnostic marker; this would be valuable in future trials of protective efficacy for any TB vaccine.
In response to questions from the audience, Dr. Hoft added the following points:
Efficacy studies show that BCG is more effective in infants than in adults, but this is partly because its easier to detect clinical effects in infants, who have a higher incidence of progressing to disseminated disease. Efficacy is harder to detect in vaccinated adults, who may take 10 or 20 years to progress to disseminated disease.
Rising rates of TB in adult populations, especially in tropical developing countries, do not point to a major failure in terms of BCG vaccination of adults. At the same time, they point to the need to do better. What is needed is a strategy that concentrates on preventing the spread of infection by detecting and treating asymptomatic as well as symptomatic cases.
Evidence suggests that Th-2 responses predominate in antibody production, while Th-1 responses induce Ig subsets. The two don’t necessarily conflict in the development of protective immunity, but it may be important to learn how to use different schedules of vaccination to induce the two responses in different sites.
As long as overall rates of infection remain low, the United States will probably not use BCG except for people who are at very high risk for infection with multiple-drug resistant TB.
There is data on crossreactive immune responses to other mycobacterial species such as M. vaccae. This may have a role as a potential adjuvant in immunotherapy. Further study is needed.
There are as yet no data on the use of BCG as a carrier for other antigens, but it promises to be a very useful vector. The technique has worked in animals, and MedImmune is currently conducting tests in humans.
There have at present been no controlled trials of intranasal or oral vaccination with BCG in humans.
HISTOPLASMOSIS AND COCCIDIOIDOMYCOSIS8
Pathobiology. Histoplasma capsulatum and Coccidioides immitis are soil-based dimorphic fungi. In the soil they exist as molds, but they can cause serious disease when inhaled by animals or humans.
In the lungs, histoplasma converts to a yeast phase, a vacuolated intracellular parasite similar to Leishmania donovani and L. mexicana, that then disseminates into the liver and spleen. It has a predilection for mononuclear phagocytes and grows readily in phagolysosomes. The survival strategy is unclear, but it appears that the organism can alkalinize the phagolysosome to a pH of 6.0 or 6.5, low enough that it can still scavenge iron but high enough to mitigate the effects of acid proteases that are present in the phagolysosome. If the inoculum is sufficiently high with the pulmonary infection, the individual can become moderately to severely ill. In most individuals the infection is self-resolving, often either asymptomatic or with flu-like symptoms. When the immune system is suppressed, however, the organism will disseminate and reactivate to cause progressive disseminated histoplasmosis that, if untreated, can be life-threatening.
When coccidioides enters the lungs it is transformed into spherules that can become quite large, containing up to 1,000 endospores. As with histoplasma, the infection is often self-resolving and asymptomatic, but in an immunocompromised host the organism can disseminate and reactivate, causing Valley fever—progressive disseminated coccidioidomycosis. It shows a predilection for the skin and the meninges, as well as the lungs, and like histoplasma it likes viscera that are rich in mononuclear phagocytes.
Both organisms pose the biggest threat to immunocompromised hosts, in the form of a reactivation disease. Primary infection is followed by a dormant phase that is held in check until a perturbation in the immune system allows the organism to flourish. Except where known, immunosuppressive agents are involved, researchers do not know why otherwise healthy immune systems should break down and allow the organism to replicate.
Incidence and Burden. Incidence can be very high in areas where the pathogen is endemic. During the 1950s, approximately 90 percent of Navy recruits from the Cincinnati area were skin-test positive for histoplasma. The figure may be lower today, yet clinical observations indicate that about 75 percent of the people in Cincinnati who undergo routine chest x-rays have calcifications in their lungs or spleens, and an autopsy series demonstrated that—at least in Cincinnati—most of those calcifications were correlated with the presence of organisms consistent with Histoplasma capsulatum. When the National Institute of Occupational Safety and Health (NIOSH), which is located in Cincinnati, tested for the presence of histoplasma in a building being renovated, they found the organism not only in accumulated bat guano inside the building, but also in soil samples from outside the building. After the bat guano was treated with hypochlorite solution, it still tested positive for histoplasma (the usual treatment is formaldehyde).
Immunology. With both histoplasma and coccidioides, cell-mediated immunity is of primary importance. CD4 cells appear to be the primary mediator, although CD8 cells play a smaller role. Humoral immunity has little or no role. In both cases, the transfer of hyperimmune serum results in very little enhancement of phagocytosis by mononuclear phagocytes. One reason is the
very rapid intrinsic phagocytosis of the organism—when unopsonized histoplasma are incubated with human mononuclear phagocytes, 80 percent to 90 percent of them are bound to the phagocyte within 5 minutes, and within 10 minutes the same proportion have been ingested.
Other evidence that histoplasma infection leads to cell-mediated immunity include (1) granuloma formation, (2) delayed hypersensitivity, (3) proliferation by peripheral blood mononuclear cells in vitro, (4) greater susceptibility by athymic mice than in their normal littermates, and (5) the ability of T-cells from immunized mice to transfer protection to unimmunized mice. Immunity to coccidioides can also be transferred with T-cells, while T-cell depletion or impairment is associated with poor prognosis. In both cases, however, as in tuberculosis, the result is not so much the killing of the organism as the inhibiting of its growth, since the organism remains dormant for many years following the initial exposure.
In both histoplasmosis and coccidioidymosis, as in tuberculosis, the Th-1 response seems to be most important in controlling infection, at least in the mouse models of these diseases. Resting phagocytes allow the organisms to grow until they are activated by a soluble signal from the T-cells. In mice, that signal seems to be IFN-gamma, but in humans (at least for histoplasma) the signals are IL-3 and the granulocyte and macrophage colony-stimulating factor (GM-CSF). Human mononuclear phagocytes on plastic are activated by neither IFN-gamma nor tumor necrosis factor alpha (TNF-alpha).
Isolating Protective Immunogens for Histoplasma. The literature on histoplasma indicated that a sublethal inoculate of either the conidia (spores) or yeast form of H. capsulatum could protect mice from subsequent challenge, and that inactivated conidia or heat-killed yeast could also provide protection. Further analysis showed that a detergent extract of the cell wall and cell membrane from this organism could confer protective immunity, and two antigens were isolated from an extract taken from the virulent strain H. capsulatum G217-B, the standard strain used in all animal models.
Researchers were able to demonstrate (1) that one of these extracts, called HIS-62, was recognized by immune sera from mice, (2) that it stimulated proliferation of sensitized lymphocytes in immunized mice, (3) that mouse T-cell clones recognized the crude extract, and finally (4) that vaccination with HIS-62 could confer protective immunity against pulmonary histoplasmosis in three different strains of mice. Upon further analysis, researchers found that the peptide sequence from HIS-62 was highly homologous to heat shock protein 60 (hsp 60) —about 70 percent identity with hsp 60 from Saccharomyces (the nearest yeast species), 50 percent identity with hsp 60 from bacteria, and 60 percent to 70 percent identity with mouse or human hsp 60, at the amino acid level. Crossreactivity with anti-GroEL serum confirmed that HIS-62 was a member of the hsp 60 family.
Researchers subsequently isolated the gene that expresses HIS-62 and cloned it into a bacterial expression vector, pET19b. Further tests confirmed that the rHIS-62 induced protective immunity in BALB/c mice against intranasal challenge with a lethal dose of histoplasma. Currently, they are trying to identify the smallest fragment of rHIS-62 that can confer protective immunity, using 4 overlapping fragments of about 250 to 300 amino acids in length. Unfortunately, the hierarchy of proliferation response has been totally different for BALB/c mice immunized with whole H. capsulatum (fragment 3 highest, followed by 2, 4, and 1) than for mice immunized with rHIS-62 (fragment 1 highest, followed by 2, 3, and 4). This suggests that the fragments are being recognized and processed differently by the differently immunized mice.
When these proliferation experiments were repeated with Black-6 mice, which differ from BALB/c mice at the MHC locus, the mice immunized with whole organism did not respond to any of the four fragments. For mice immunized with rHIS-62, the hierarchy was fragment 2 highest, followed by 3, 4, and 1. Researchers do not know what the conformation of the antigen is in the organism, but they do know that adjuvants (which were used in these experiments) tend to linearize the antigen; this may explain at least part of the difference in how the fragments are being processed.
On the other hand, when mice were immunized with the four rHIS-62 fragments and challenged intranasally with a sublethal inoculum, fragment 3 produced the best protection in both strains of mice—about a log reduction in colony-forming units in the lungs and livers, and a 30- to 40-percent reduction in the spleens. Thus, there may be no correlation between the degree of stimulation and the degree of protection.
But while whole rHIS-62 had induced protective immunity against a lethal challenge in BALB/c mice, immunization with the fragments provided only 50 percent protection for fragment 4 and no protection at all for 1, 2, or 3. There are two possible explanations. First, hsp 60 has some adjuvant effect, at least in mycobacteria, but the region that induces it may not be on a particular fragment. Second, it may take all of the fragments working together to induce a strong proliferative response. This experiment is being repeated in Black-6 mice against intranasal challenge with a lethal dose of histoplasma. Researchers believe they will be able to demonstrated that fragment 4 of rHIS-62 can induce a protective immune response in mice.
Isolating Protective Immunogens for Coccidioides. Researchers are using a different approach for coccidioides, screening potential antigens with antisera from humans who have recovered from an infection. The genes for these antigens will be introduced into DNA vectors and used to immunize mice in hopes of getting a protective immune response. One of those antigens is the hsp 60 from coccidioides, which has about 90 percent identity with hsp 60 from histoplasma. Researchers have sequenced and cloned this gene and introduced it into a bacterial expression vector, pET21a or b. They have also inserted it into PCNV to see if it can induce genetic vaccination. An interesting question for future research is whether hsp 60 is crossprotective for histoplasma and cocci-
dioides, given the differences in pathogenesis and pathophysiology between the two diseases.
Another potential antigen that has been isolated from the soluble fraction of C. immitis is HPPD, an enzyme that is involved in the degradation of aromatic amino acids and the production of pigments. The gene for HPPD is highly conserved from bacteria to humans, and that of coccidioides is about 50 percent identical to the human gene. Researchers have sequenced and cloned the gene and demonstrated that HPPD can evoke a proliferative response in mice immunized with coccidioides. At present, however, there are no data on the protective efficacy of either of these antigens.
Problems in Developing Fungal Vaccines. As with other vaccines, an important question is whom to vaccinate. Clinicians don’t know at what age these diseases are acquired. Another important question is when to vaccinate. Because cell-mediated immunity is primary, the timing of interdiction is crucial.
In response to questions from the audience, Dr. Deepe added the following:
Spherulin, the spherule-based vaccine against coccidioides that was developed 10 or so years ago, does not contain hsp 60. There are questions about its efficacy and side effects, but extensive animal studies have been done and the results of randomized clinical trials were published about 12 months ago.
One of the differences between the two pathogens is the importance of neutrophils in the histopathology of Coccidioides immitis. By the time there are symptoms and biopsies for histoplasmosis, however, there is no sign of neutrophils.
GROUP A STREPTOCOCCI9
Incidence and Burden. Group A streptococcus bacteria cause a wide array of clinical syndromes, ranging from the uncomplicated streptococcal pharyngitis (a very common infection among young children) and streptococcal pyoderma at one extreme, to necrotizing fasciitis, “flesh-eating” pyomyositis, and the newly described streptococcal toxic shock syndrome at the other. The resurgence of serious Group A streptococcal infections has provided a new impetus for the development of vaccines, but the major driving force over the years has been the acute rheumatic fever and chronic rheumatic carditis that can follow Group A infection, causing significant morbidity and mortality.
The past 10 years have seen a significant change in the epidemiology of Group A streptococcal infections in the United States and Europe. In certain U.S. cities, for example, the incidence of rheumatic fever has jumped from 1 per
100,000 to 10 or 15 per 100,000. An outbreak of rheumatic fever in Utah has been continuing since the mid-1980s. At the same time, there has been an increase in more serious infections—not just pharyngitis and impetigo without complications, but loss of limbs and even of lives. CDC estimates that there are 15,000 to 20,000 cases of serious streptococcal infection per year in the United States, with mortality of anywhere from 30 to 50 percent. In developing nations, meantime, acute rheumatic fever and life-threatening streptococcal infections continue to be as serious a problem as ever.
Rationale for a Surface M Protein-Based Vaccine. There are several potential antigens for vaccines, but over the years the surface M protein has produces the best evidence for protective immune responses. The surface M protein is a major virulence factor of Group A streptococci, since organisms that are rich in M protein are able to resist phagocytosis in the nonimmune host. Significantly, antibodies to M protein are opsonic.
Given these characteristics, the logical approach would be to extract M proteins from the various serotypes of Group A streptococci and combine them into a vaccine. There are three significant obstacles to this approach:
Some of the extracellular products that co-purify with M protein preparations are highly toxic. This obstacle has been overcome by recombinant vaccine technology.
There are over 80 serotypes of Group A streptococci, all of which express slightly different M proteins, and it would be extremely difficult to concoct a vaccine containing all 80 M proteins. However, there is evidence that the most serious infections can be prevented by a vaccine containing as few as 14 or 16 different M proteins.
Most serious of all, some M proteins contain autoimmune (tissue-crossreactive) epitopes and could theoretically trigger acute rheumatic fever, although there is no direct evidence of their role in the pathogenesis of that disease. For example, antibodies to type 5 M protein bind to human myocardium; antibodies to type 6 bind to neurofibers in human brain tissue; and antibodies to type 18 bind to the surface and chondrocytes of mouse joint bone. Among the other host antigens with which M proteins crossreact are myosin, vimentin, keratin, actin, tropomyosin, phosphorylase, DNA, and a large number of undefined antigens.
The latter obstacle led researchers to undertake 15 years of studies to identify which regions of the M protein evoke opsonic (i.e., protective) antibodies and which regions contain autoimmune or potentially harmful epitopes.
Structural and Functional Domains of M Proteins. A generic M protein, which protrudes from the surface of a Group A streptococcus with the amino terminus outward and the carboxy terminus buried in the cytoplasm, is a highly alpha-helical (coiled coil) molecule whose central rod typically contains internal repeats. However, the exposed amino terminus is nonhelical.
In general, the amino terminus usually contains the epitopes that evoke antibodies with the greatest serotypic specificity and bacteriocidal activity. By
contrast, most of the tissue-crossreactive epitopes that have been identified to date occur at a distance from the amino terminus, typically in the A, B, or C repeat regions. Consequently, researchers have attempted to develop a vaccine that incorporates amino terminal fragments of several different M proteins, so that they can evoke opsonic antibodies without evoking tissue-crossreactive antibodies. This approach might be characterized as a “fragment of subunit vaccine.” By incorporating these fragments into a single construct, they hope to minimize the total amount of protein injected while representing the largest possible number of serotypes.
Development of Complex, Type-Specific M Protein Vaccines. Using specific PCR primers, researchers amplified the 5-prime regions of the M protein gene from four different serotypes—M24, M5, M6, and M19. They purified the PCR products, ligated them in a hybrid M protein gene, and inserted the gene into an expression vector, in this case a transformed E. coli. Schematically, the resulting hybrid protein contained 110 amino acids from M24, 58 from M5, and 35 each from M6 and M19. Recombinant hybrid protein was extracted from the E. coli, purified, and used as a vaccine in a series of animal tests.
Three rabbits immunized with the tetravalent protein produced significant levels of antibodies against all four serotypes of native protein. However, the level of relevant (i.e., opsonic) antibodies tended to be higher for the first two fragments (M24 and M5, nearest the amino terminus) and lower for the last two (M6 and M19, nearest the carboxy terminus). Researchers were uncertain whether this difference was a function of the size of the fragments or their position on the hybrid protein. To resolve this uncertainty, researchers added another four fragments toward the carboxy terminus—35 amino acids each from M1, M3, M18, and M2—and injected another set of rabbits with the resulting octavalent protein. Again, the immunized rabbits produced significant levels of antibodies against all of the serotypes, and again the level of opsonic antibodies tended to decline toward the carboxy terminus.
To determine whether multivalent hybrid M protein would be immunogenic when administered locally by the mucosal route, researchers constructed additional hybrid proteins consisting of the entire B subunit of E. coli labile toxin (LT-B) ligated to 15 amino acids from type 5 M protein (M5), using a proline-and glycine-rich linker. Mice that were immunized intranasally with the resulting LT-B-M5 vaccine produced opsonic serum antibodies against the M5 component, and this response afforded significant protection when the mice were challenged peritoneally with type 5 streptococci. When this experiment was repeated with a LT-B-tetravalent vaccine, however, one of the components (M19, nearest the carboxy terminus) was not immunogenic at all. These mice also developed secretory immunoglobulin-A (IGA) in their saliva, a sign of mucosal immunity, but the IgA level was not particularly high.
An alternative to this approach concentrates not on the hypervariable amino terminus but instead on epitopes in the so-called C-repeats region, epitopes that
are highly conserved from one serotype to another. These conserved epitopes are exposed on the surface of the organism and available for antibody binding. Researchers have shown that antibodies to these C-repeat epitopes can block adherence and colonization, and thus prevent infection, in the mouse model. In this experiment, mice were immunized intranasally with peptide from the C-repeat region of type 5 streptococcus and then challenged with the heterologous type 24 streptococcus. The experimental group showed a higher level of survival and a lower level of colonization compared with controls. While this approach will work, however, it is questionable whether it should be the sole basis for streptococcal vaccines.
There are potential vaccine constructs that might be effective against Type A streptococci. Data on the octavalent vaccine suggest that these high complex hybrid molecules can maintain the conformations that stimulate B-cells to produce the relevant antibodies, and researchers have plans to construct a dodecavalent gene and protein. Linking this multivalent protein to LT-B or another carrier that serves as a mucosal adjuvant would provide for both opsonic and secretory antibodies, or a “cocktail” of monovalent terminal fragments could be mixed together according the epidemiology of the organism. C-repeat fragments linked to a carrier might provide broader, more flexible protection. A strategy that deserves attention in the future is the combination of amino terminal fragments (which evoke opsonic antibodies) and C-repeat epitopes (which evoke IgA that can block colonization); this combination might afford the broadest protective immunity in ultimate vaccine trials.
In response to questions from the audience, Dr. Dale added the following:
The best candidate at present is the type-specific amino terminal approach. Researchers have had unexpected success with these epitopes, which often evoke a better immune response when buried in the middle of a hybrid molecule than when they are on the end.
Researchers have not yet tested any of these approaches in humans. A great deal of pure development remains to be done, including the preclinical studies that will be needed to convince FDA that these complex constructs are safe.
It would be desirable to take a candidate vaccine into limited Phase I trials within 5 years.
The goal of the vaccine is not just to prevent rheumatic fever and other serious disease, but to prevent streptococcus infection in general. Types 2, 4, and 12 are not particularly rheumatogenic but are highly prevalent; by incorporating them in a multivalent vaccine, it might be possible to have a broad impact on disease.
The target population for the vaccine is preschool children. If they can be immunized before kindergarten, it might be possible to have an impact on overall incidence.
In addition to common C-repeats, the so-called M5 family (M5, M6, M18, M19, and others) also has common B-repeats, and these conserved epitopes are opsonic.
Several companies have expressed interest in becoming partners for further development and preparation for human trials.
Despite tremendous pressure, Group A streptococcus has not developed antibiotic resistance. Penicillin is still effective in treatment and in prophylaxis.
There is sometimes a fine line between immunizing epitopes and tissue-crossreactive epitopes. Researchers hope to clarify this issue when the final vaccine construct is tested in animals, especially primates whose immune response genes are most similar to humans.
Recent developments indicate that it may soon be possible to use vaccines not only prophylactically, to protect against infection and reinfection with Helicobacter pylori (HP), but also therapeutically, using active immunization to clear existing infection and prevent reinfection.
Incidence and Burden. HP is the cause of the vast majority of cases of peptic ulcer, over 90 percent of duodenal ulcers, and perhaps 80 percent of gastric ulcers. Once infected with HP, the lifetime risk of acquiring peptic ulcer disease is about 20 percent, with additional cases of atrophy. The risk of gastric cancer is also significant: overall, the lifetime risk of gastric cancer in the United States is about 1 percent, and 60 percent of these cases are attributable to HP.
The association with cancer is even more important elsewhere in the world. In the United States, seroprevalence is about 50 percent by age 50; in developing countries it approaches 100 percent and the infection is acquired earlier, often in childhood. Early acquisition and decades of chronic inflammation appear to be important in the genesis of gastric cancer. The World Health Organization has classified HP as a definite carcinogen.
Pathobiology. HP is a Gram-negative bacterium that is transmitted by saliva and vomitus; young children are the most active transmitters. Once ingested, the bacterium is well-adapted to penetrate the mucous lining of the stomach and colonize the gastric epithelium. It has a spiral shape designed for boring through this viscous environment, and a polar flagella that provides motility and also plays a role in virulence. HP does not invade the host; instead, it causes an “offshore” or surface infection of the epithelium and sometimes penetrates into the gastric glands. Once established, it is a lifelong persistent infection that is not cleared by host immune response.
The result of chronic infection is inflammation of the gastric epithelium leading ultimately to atrophy and destruction of the gastric glands, lymphoid follicles, and a massive accumulation of T and B cells. Secondary effects of
inflammation include gastric ulcer, gastric metaplasia of the duodenal bulb, duodenitis, and ultimately duodenal ulcer. A long-standing infection may lead to intestinal metaplasia and gastric carcinoma. Recently, B-cell lymphomas of the stomach have also been associated with HP.
Issues in Vaccine Development. Because HP is a noninvasive surface infection, vaccines should elicit strong mucosal or secretory IgA immunity against one or more prominent surface proteins of the bacterium. Because HP has several virulence factors, the vaccine should be designed to prevent infection rather than interfere with just one virulence factor. Because the infection persists despite a vigorous immune response by the host, and hosts cured of infection with antibiotics appear to be susceptible to reinfection or recrudescence, natural immunity does not seem to be sufficient to protect against or clear infection; hence the vaccine should elicit a response that is qualitatively or quantitatively different from natural immunity.
In addition, HP antigens may elicit hypersensitivity or autoreactive responses, as was the case with Group A streptococci; for this reason, the vaccine should be based on a well-characterized protein that avoids these reactions. The latter considerations—inadequacy of natural immunity and potential crossreactions—argue against the use of live attenuated vaccines or crude whole-cell preparations.
Finally, HP strains are extremely plastic, showing wide variation at both the genomic and antigenic levels. Consequently, the antigen or antigens used in a vaccine must be highly conserved and must be expressed in vivo, so as to be available as a target for immune response. Several candidate antigens have been investigated, including urease and vacuolating cytotoxin, and the ultimate vaccine may involve a combination of antigens. The balance of the presentation focused on urease.
Urease-Based Prophylactic Vaccine. Urease is an abundant protein on the surface of the bacteria and makes up over 6 percent of its total soluble protein. It is highly conserved across strains of HP and even across different species of Helicobacter that can be used in animal models (see below). It is a large molecule (550 kD) with a particulate structure, but natural immunity to HP includes only a weak or inconsistent response to urease. The enzyme functions as a virulence factor: it splits the urea found in gastric secretions into two molecules of ammonia, creating a neutralizing cloud around that bacteria that protects it as it passes through the acid environment of the stomach on its way to the pH-neutral environment under the mucous lining.
The operon that controls urease includes two structural genes, ure-A and ure-B, that code for proteins of about 25 and 60 kD respectively. Six units of each of these proteins make up the intact 550-kD molecule. The rest of the genes in this operon are involved in folding the molecule in such a way as to incorporate a molecule of nickel, the metalloenzyme required for its activity. By cloning the two structural genes and leaving out the rest, researchers were able to generate a recombinant urease that is structurally and antigenically intact but lacks the enzymatic activity that would be harmful to a host. Cloned into E. coli
and expressed in a high-density fermentation system, this modified operon produces extremely large quantities of this recombinant urease—over 3 grams per liter of culture, when immunization requires microgram or nanogram doses (see below).
Preclinical Results. There are several good animal models in which to test HP vaccines. H. pilus is a very similar species that infects mice and cats, and the human pathogen H. pylori can also be modified to infect mice. Ferrets are susceptible to a third species. However, the most interesting work has been done with cats, which are highly susceptible to H. pylori, naturally infected, and develop gastritis and even ulcer disease that is very similar to humans. Cat immunology is at a relatively advanced state of knowledge, which makes this model even more useful. Finally, researchers are beginning to work with primates as well.
Experience has shown that the most effective schedule for prophylactic vaccination is 4 applications 1 week apart. The animals are then tested for antibodies and challenged with either H. pilus or H. pylori after an appropriate interval—initially 2 weeks, but now up to 1 year. A total of 2 weeks or longer after challenge, the animals are sacrificed and tested in various ways: spectrophotometry (to measure urease activity in the stomach and hence the presence of bacteria), histology (a silver stain for bacteria), electromicroscopy (for antibodies), and immunocytochemistry (for antibody-secreting cells in the gastric mucosa).
Experiments with H. pilus in mice showed that mucosal immunity plays an important role in protective immunity. Experimental animals were immunized with recombinant urease by different routes: oral, oral with bicarbonate, intragastric, intragastric with bicarbonate, and subcutaneous. Following challenge, 100 percent of unimmunized controls became infected, while 100 percent of those immunized by the oral route were protected. Intragastric immunization, which bypasses potential induction sites in the oral cavity, was less protective, especially when given without bicarbonate to neutralize gastric acid. Parenteral immunization did not protect at all, although it did induce very high serum IgG antibodies.
Dose-response studies showed this is an incredibly potent immunogen, but that a mucosal adjuvant is required for immunization. When mice were given recombinant urease linked to E. coli labile toxin (LT, similar to the LT-B used as an adjuvant in streptococcus vaccines, above) by the oral or intranasal route, doses as low as 50 nanograms provided significant protection. Doses are usually in the microgram range or higher in animal experiments. Without the mucosal adjuvant, however, doses as high as 5 milligrams provided no protection: 100 percent of the animals became infected. Analysis of antibody response indicate that secretory IgA correlates best with protection: animals that were fully or partially protected had IgA in their saliva; those that became infected might have IgG in their serum but had no IgA in their saliva.
Durability of protection is another critical issue, and results to date show very durable immunity after oral immunization with recombinant urease. Animals challenged at various intervals show up to 94 percent protection when challenged 1-year after immunization. Significantly, these animals also have very high concentrations of salivary IgA antibodies one year after immunization. This distinguishes artificial immunization from natural immunity, which in this model is primarily serum IgG. Based on this finding, researchers now theorize that HP may avoid the natural host immune response (at least in part) through an antigen-specific suppression of a Th-2 response (other possible explanations include antigenic variation, molecular mimicry, privileged site of sequestration, and polyclonal stimulation).
Other studies have confirmed that immunized mice produce IgA-class antibodies not only in their saliva but also directly in their stomachs following challenge. Immunocytochemistry reveals a marked increase in the number of IgA-secreting cells in immunized animals and a much lower response in infected controls. In fact, up to 20 percent of the IgA-secreting cells in the gastric mucosa of immunized mice were urease-specific. It was this vigorous local immune response that first suggested the possibility of developing a therapeutic vaccine.
The same results were achieved when these experiments were repeated in the mouse model using H. pylori instead of H. pilus, and similar studies are now being conducted with H. pylori in cats, which may be a better model of the human infection. Domestic cats immunized with urease and LT adjuvant developed salivary IgA responses to urease that peaked at 2 weeks and declined somewhat by 5 weeks. When challenged 8 weeks after immunization and sacrificed 8 weeks after challenge, immunized cats had significantly lower mean numbers of bacteria in the stomach compared with controls, and significant recruitment of urease-specific antibody-secreting cells in the gastric mucosa.
Therapeutic Vaccine. After discovering that immunized animals could become transiently colonized and then clear the infection, researchers designed an experiment to see if they could eradicate an infection by active immunization. Mice were infected, given the typical vaccine course, and then sacrificed after 2 or 8 weeks to determine outcome. Controls were 100-percent infected, but there was a significant decrease in the infection rate in immunized mice. When immunized mice were challenged a second time 4 weeks later, there was significant protection to rechallenge among animals that had cleared their infection.
The next question was how best to use the vaccine therapeutically. Current therapeutic regimes achieve cure rates of up to 90 percent using a combination of antacids and various antibiotics, and vaccine can transiently suppress the infection as well. But neither antibiotics nor vaccine by itself can achieve 100-percent clearance, and bacteria populations tend to drift upward again over time. Working with the mouse model, however, researchers found that the combination of amoxicillin, Pepto Bismol, and vaccine was able to achieve 100-percent clearance of the infection and to protect against recrudescence or reinfection.
Current Status of Vaccine. Researchers filed an Investigational New Drug application in 1995. A Phase I safety study indicated that the vaccine was
quite safe and well tolerated. Researchers are currently in the middle of an appropriate Phase II study of the safety and immunogenicity of this vaccine, using an appropriate adjuvant.
In response to questions from the audience, Dr. Monath added the following:
Other groups are investigating other potential immunogens, including vacuolating cytotoxin and other molecules. In the end, some combination of antigens including urease and others may be required to produce a successful vaccine.
Urease is present on the surface of the bacterium, but researchers don’t know how it gets there. They have not discovered the signal sequence. One theory is that as other bacteria are lysed and release cytoplasmic urease, the urease is scavenged by bystander bacteria and collected on their surface.
HP infection is densest in the antrum of the stomach, but there are also bacteria present in the corpus and even the cardia. There is no clear or well-understood relationship between the site of infection and the subsequent site of ulceration. The mechanism is better understood: infection compromises the protective mucous layer as a result of HP virulence factors—mucinase, ammonia, etc., and this exposes the a region of mucosa to acid.
HP have been found colonizing dental plaque, as well as the stomach, but there is no evidence that HP replicates in the gut.
Incidence and Burden. The incidence of gonococcus infections has declined dramatically in Europe, by 80 to 99 percent in most countries and almost to extinction in Sweden. It has also declined by about 40 percent in the United States over the past few years, but the social economics of the disease are such that it is unlikely to disappear from its core regions, which include the inner cities and the rural South. Seroprevalence may be a few percent to 5 percent in these U.S. areas, and gonorrhoea remains a very common disease in much of the rest of the world. In Africa, for example, 5 or 10 percent of many study populations are carrying gonococcus at any one time.
Gonococcus is important not because of its acute symptoms but rather because it causes tubal infertility. In this country, it is probably the number-2 cause of tubal infections, after chlamydia, causing about 100,000 cases per year of salpingitis, ectopic pregnancy, and infertility.
More importantly, both gonococcus and chlamydia appear to be important co-factors for HIV transmission. Among African women, for example, gonococcus and chlamydia infections are associated with a fourfold increase in the risk of contracting HIV, and the vaginal secretions of infected HIV-positive women contain an increased amount of virus compared with uninfected HIV-positive women. Recent studies have shown that antibiotic treatment of gonococcal or chlamydial infection dramatically reduces the amount of HIV RNA in semen. In a prospective study, simple syndromic management with antibiotics of minor genital infections, urethritis, and genital ulcer syndrome dramatically decreased HIV transmission.
Gonococcus is developing resistance to antibiotics, however. Clinicians have already lost penicillin and tetracycline, which were the drugs of choice until about 2 years ago. Isolated cases of resistance to ciprofloxacin and chlorinated quinoline have appeared in the United States and will probably spread. If clinicians lose the second- and third-generation cephalosporins, treatment would becomes a very significant problem.
Natural Immunity. Like Helicobacter pylori, there is little or no naturally acquired immunity to gonococcus. Infections persist, and reinfections are common. Nevertheless, there are convincing data from the era before antibiotics indicating that a deep and persistent gonococcal infection resulted in an immunological cure. In fact, there are a few articles in the old literature on the use of therapeutic vaccines. Despite the shortcomings of these studies, they do suggest that gonococcus, however elegantly it evades the immune response, cannot completely evade a truly vigorous response, and that this effective immune response could be utilized in a vaccine.
This relative lack of a protective immune response to uncomplicated genital infection is surprising, at first glance, because such infection does result in a very vigorous systemic and mucosal antibody response against a number of gonococcal antigens. The mechanisms by which gonococci escape what might otherwise be an effective immune response include rapid antigenic variation of key surface proteins, masking of surface antigens by sialylation of gonococcal lipooligosaccharide (LOS), production of IgA protease, and stimulation of “blocking” antibodies.
In part because of the lack of a good animal model in which to study immune response, past research has focused on the genetics, molecular biology, and pathophysiology of gonococcus rather than functional or protective immune responses. Researchers have learned a good bit about several of these gonococcal antigens, and this knowledge may provide explanations for the lack of natural immunity, as well as candidate antigens that might become the basis for gonococcal vaccines.
Current Candidate Antigens. Most of the initial research concentrated on different forms of gonococcal pilus, a surface appendate that functions as an adherence ligand, and its main subunit (pilin). Antibodies to pilus block adherence and opsonize phagocytes, but gonococci generate a bewildering antigenic variation in pilus. In a human volunteer study, a pilus subunit vaccine adminis-
tered intramuscularly resulted in significant protection against urethral challenge by the exact strain from which the vaccine was made, but not against a heterogenous strain. When the U.S. military conducted a field trial of monovalent pilin vaccine in Korea, the vaccine failed completely, and predictably. Attention has therefore shifted to pil-C, a pilus-associated protein that is expressed at the sides and tip of the assembled pilus rod and is crucial for epithelial cell adherence. Pil-C is a relatively minor protein on the cell surface, and little is known about how variable it is or whether antibodies against it are protective. It may nevertheless prove to be a better candidate that pilin.
Another effort has focused on the principal outer membrane protein (Por), which is a more attractive candidate than pirin as a vaccine. Por is the major protein on the surface of the gonococcus; it is expressed constitutively and in thus always plentiful; and it shows relatively little antigenic variability (two major immunochemical classes, and minor variance within classes). Antibodies against Por have been shown to be bacteriocidal and opsonic and to block tissue-culture toxicity. Some antibodies crossreact broadly within one or the other immunotype. All of these properties suggest that a vaccine that stimulated these antibodies would be broadly protective. Several groups have worked on developing vaccines using recombinant porine and peptides based on porine.
One difficulty in developing a Por-based vaccine is that Por occurs on the cell surface in tight clusters with two other molecules, LOS and reduction modifiable protein (Rmp). When sialic acid, a sugar residue, is added to the core sugars in LOS, it physically masks the surface exposure of the critical Por epitopes. Rmp is highly immunogenic, and antibodies to Rmp functionally block the bacteriocidal activity of antibodies to Por. The latter mechanism is confirmed by epidemiological data showing a more or less quantitative relationship between the titer of anti-RMP antibody and the risk of a woman becoming infected through sexual contact with an infected man. This still suggests that a Por vaccine is possible, if it induces sufficiently high titers of antibodies to Por while inducing little or no Rmp antibodies. One proposed strategy is to prepare recombinant protein from an organism in which the Rmp has been genetically knocked out. Several companies are pursuing this strategy, and there may well be Phase I clinical trials of such a Por liposome vaccine within 12 to 18 months.
Another strategy has been to focus on the LOS itself. Certain of the core sugars in this molecule are conserved, and antibodies against them are bacteriocidal and opsonic. Anti-idiotype monoclonal antibodies that mimic these core sugars are also bacteriocidal and opsonic. One such antibody is mAb 2C7, which reacts with an oligosaccharide determinant of LOS. Importantly, this particular epitope is not masked when sialic acid is added to LOS. Screening tests show that this epitope is common and stable during serial passage, and antibodies against it are generated in natural human infection. When mAb 2C7 is used as an immunogen in mice and rabbits, the immune response is broadly bacteriocidal and opsonic. This epitope is clearly of interest for further studies.
Potential Future Candidate Antigens. Four other surface proteins, all of them expressed in vivo under conditions in which the bacteria are starved for nutritional iron, have also attracted attention as potential vaccine candidates. One of them, lactoferrin binding protein 1 (Lbp1) is found in only about 50 percent of gonococci; because it is not uniformly present, it is probably not a serious candidate. The other three molecules are present in virtually every gonococcus, and each has properties that suggest its potential value in a vaccine.
Transferrin binding protein 1 (Tbp1) is an integral outer membrane protein that may form a gated pore through which iron enters the cell. Transferrin binding protein 2 (Tbp2) is a lipoprotein that is loosely tethered to the outer membrane; there are multiples of Tbp2 for each Tbp1. Together they form a functional receptor for binding human transferrin and extracting the iron that the gonococcus must take up in order to grow. (This thesis was tested in human volunteers in early 1996, using mutant gonococcus that do not make Tbp1 and Tbp2, and therefore should not be virulent.)
Tbp1 is a potential vaccine candidate because (1) it is antigenically conserved across gonococci (about 90 percent identity across 3 variants characterized); (2) polyclonal antipeptide sera bind to the conserved surface loops; (3) this binding appears to block transferrin binding, an important cellular function; and (4) unlike Por, sialylation of LOS does not mask the functional exposure of this epitope.
Tbp2 is highly immunogenic in vivo, and it is not masked by sialylation of LOS, but it is strikingly antigenically variable—not a characteristic that would recommend it as a vaccine candidate. However, a consortium of European researchers has done considerable work on Tbp2 in both gonococcus (N. gonorrhoeae) and meningococcus (N. meningitidis). They have found that meningococcal monoclonal antibodies and polyclonal sera against meningococcal Tbp2 block transferrin binding, are bacteriocidal, and—unexpectedly—are relatively crossreactive among different strains. Sequencing of five different strains of meningococcal Tbp2 revealed that, despite differences throughout the molecules, two regions are highly conserved, both of them in the domain that is necessary for binding transferrin. If these antigenically conserved regions are accessible to antibodies, this would explain why antibodies are more crossreactive than expected. It would also make Tbp2 a candidate for vaccines against both meningococcus and gonococcus.
Finally, iron-repressed protein B (FrpB) is another vaccine candidate. This protein in abundant on the surface of some gonococci and present on all. Its physiological function is unknown, although it appears to be a member of the receptor family. Research on meningococcal FrpB, which is almost identical to the gonococcal protein, shows that antibodies are almost always bacteriocidal and surprisingly crossreactive, given that they fall into 10 different serovariant groups. However, sequencing of two gonococcal genes and one meningococcal gene show that this protein is highly conserved at the amino acid level, even in the surface exposed domains. Further research remains to be done, but this protein might also be included in an “bull’s eye” component vaccine.
Human Testing. Several animal models have been attempted, but none has been satisfactory, and this has impeded vaccine development. Because of this, however, researchers have learned a great deal about the genetics, molecular biology, and pathophysiology of the organism—how it works and how it evades the immune response. Many researchers have now decided that the time has come to move to human subjects, the only relevant model, in order to develop a vaccine in an intelligent, rational, and cost-effective manner.
One group has challenged over 80 male human experimental volunteers with gonococci in the past few years, establishing several important points. First of all, it is safe. Second, several proteins appear to be sufficient or facilitative for infection, but no single protein appears to be truly necessary or essential for infection. Third, the core polysaccharides of LOS do appear to be absolutely essential to infection, which is another argument for including antibodies to these surface molecules in a vaccine.
In response to questions from the audience, Dr. Sparling added the following:
It is possible that there will be Phase I and Phase II trials in the next few years, at least on the pilus- and Por-based vaccines. If these trials don’t go well, it will take additional years to develop the information needed to go forward with the other candidate antigens described above.
Progress is a factor of people and money. The field has been moving relatively slowly because there has been little focus on gonococcus vaccine per se.
The LOS antigen could be developed as an anti-idiotypic vaccine, but it wouldn’t have to be. Another possibility would be to combine that core sugar with another antigen in a complex protein-carbohydrate vaccine.
The LOS antigen has not yet been tested in humans. It does give a good booster response (IgA and IgM) in mice and rabbits, but this is only a surrogate for opsonic and bacteriocidal activity.
Work on the other antigens—Tbp1, Tbp2, and FrpB—is still at an early stage of concept development.
It would be desirable to administer the vaccine orally or intranasally. Shigella might be an attractive vector, but there are other potential vectors, as well as cytokines, to target the vaccine.
The target population would be, at the least, young people as soon as they’ve had their first incident of any sexually transmitted disease.
An adjuvant is any agent or substance that enhances an immune response. In most cases, adjuvants do not elicit immunity to themselves, but they can and do influence every aspect of the response to the antigen they accompany: the amount of antibody made, its specificity, which epitopes of a given protein are responded to, the isotype of the response, its avidity, duration, memory, and so on. In a very real sense, they are vital to vaccine development.
General Mechanisms of Adjuvant Action. Most adjuvants induce a protective or neutralizing antibody response, and over the years this has been the focus of most adjuvant research. Fewer adjuvants have been looked at for delayedtype hypersensitivity (DTH) or cellular immune response. And only recently have researchers begun to look at the small number of adjuvants that can induce Class I-restricted CTL responses. This research, going back to 1979, has identified a number of general mechanisms by which adjuvants work. One is the so-called “depot effect” —the antigen persists longer because it has been incorporated in an emulsion and is released slowly over time (e.g., Freund’s adjuvant). Another is the selective antigen localization in thymic-dependent areas.
However, the crucial mechanism is probably macrophage activation and the generation of inflammation. Indeed, this is how adjuvants were first discovered early in this century—horses that developed a sterile abscess at the injection site had much higher titers of antibody, so researchers started creating abscesses and injecting antigen into the abscess to boost the response. Freund’s adjuvant certainly induces inflammation, which is why it is not used in humans.
Other proposed mechanisms for adjuvant activity include increased uptake and presentation by antibody-presenting cells; processing pathway switching; specific or nonspecific stimulation of various helper T-cells; stimulation of increased cytokine production; B-cell isotype switching; and maturation of precursor cells. The specifics for any given adjuvant can be difficult to work out, because these are fairly complex substances that do several different things at once. In the past few decades, much activity has focussed on refining adjuvants down to a simpler, less toxic substance that can be used with greater safety.
On the molecular level, adjuvants appear to enhance the production of costimulatory cytokines. These include IL-1-beta, IL-6, and tumor necrosis factor (TNF), which are crucial for the initiation of immune response. This in turn upregulates costimulatory molecules such as B7, CD8–28, and various other adhesion molecules on T-cells and B-cells. In the future, it may be possible to formulate new adjuvants that selectively exploit these mechanisms. For example, a recent article indicates that the CTLA-4 ligand on T-cells gives a very inhibitory signal to the T-cell; inhibition of that pathway would probably have very good adjuvant effects.
Increased local production of cytokines such as GM-CSF, which activates macrophage IL-12 and IFN, will recruit increased numbers of macrophages to the lesion. If a “depot” antigen is sitting there, these larger numbers will mean a greater immune response. In many ways, the action of an adjuvant is to overcome the immune system’s basic tone or bias, which is to nonresponse rather than response.
Common Adjuvants. Almost every vaccine experiment in animals involves the use of some form of adjuvant, although it seldom receives prominent attention. The most commonly used adjuvants in experimental animals are Freund’s adjuvant, either complete or incomplete; Bordatella pertussis; lipopolysaccharide (LPS) or the more refined lipid-A; muramyl dipeptide (MDP, the smallest component of microbacteria that has an adjuvant effect); immunostimulating complexes; glycopolymers; liposomes; and a few others. Some of these adjuvants enhance the Th1 or cellular immune pathway (e.g., Freund’s, LPS, MDP); others enhance the Th2 or humoral immune pathway (e.g., alum, pertussis).
Only one adjuvant is currently approved by FDA for use in humans: alum, a mixture of aluminum salts. When injected, shards of crystal in the alum cause inflammation, recruiting macrophages that eat the antigen associated with the crystals, thereby presenting it to the immune system. However, researchers have identified at least six groups of potentially useful agents, most of which are already in human trials:
Monophosphoric lipid-A (MPL) might be called the active ingredient of LPS, a further refinement that has lower toxicity but retains adjuvant effect. In mice, it stimulates both antibody and cell-mediated immunity. It stimulates macrophage IL-1, TNF, various colony-stimulating factors, and IFN-gamma, and it also up-regulates Class II MHC expression. Like LPS, in enhances nonspecific resistance to bacterial infection and has a global effect on the immune system. One such adjuvant is currently in Phase 3 human trials.
Muramyl dipeptides (MDPs) include hundreds of derivatives that have been synthesized and tested over the years. Two products are currently in human trials with AIDS vaccines: SAF-1, which contains a blocker polymer (see below); and NF-59, which contains a small amount of lipid. Both products stimulate antibody as well as cell-mediated immunity, costimulating T-cells and up-regulating Class II macrophages. The exact cytokine they are producing is not really clear, but they are candidates for future vaccines.
Immune-stimulating complex (ISC) was developed by a veterinarian in Europe. It contains cholesterol and safranin (a detergent derived from plants), with which the antigen is mixed, and takes the form of small particles. In animal tests (mice, cats, sheep, cattle, and monkeys), it produces very strong antibody and CMI response, and it is effective for mucosal immunization, which also produces CTL responses. A formulation called QS-21 is in Phase 1 trails in humans.
Glycopolymer is a polyoximer related to common compounds found in mouthwash, toothpaste, and even food products. It consists of a central segment of hydrophobic polyoxyl propylene polymer, with a hydrophilic polyoxyl ethylene at each end; the lengths of these segments are critical to the activity of the molecules. They enhance antibody, IgG1 and IgG2a, and cell-mediated responses. They appear to have a depot effect and have been shown to be effective as an adjuvant with peptides, proteins, and polysaccharides in oil and water emulsions. Currently available as an animal preparation under the name Titer-Max, it is fairly expensive but may be cheaper in broader human use. It is currently in Phase 1 trials.
Cytokines are a logical target of research: if adjuvants stimulate cytokines, why not use the cytokines themselves? GM-CSF is a leading candidate because both stimulate the production of macrophages in bone marrow and activate the macrophages; it will soon be approved for use in cancer gene therapy. IL-12 preferentially stimulates the Th1 pathway in mice and might be attractive, but there are toxicity problems. IFN-gamma, IL-4, and anti-IL-4 might also be useful to preferentially activate or block certain response pathways, depending on the antigen.
Human Testing. One of the biggest hurdles in the development of new adjuvants is the difficulty of testing them in humans. Under FDA rules, new adjuvants can only be tested in cancer patients (where the standards of harm and side effects are lowered). Cancer vaccines aren’t really that similar to infectious disease vaccines, and adjuvants that fail in a cancer vaccine might still be excellent adjuvants for other applications. Moreover, no single adjuvant is going to be universally effective; many different adjuvants, each with their own particular application, might be needed.
Mucosal Immunization. Most pathogens invade or colonize the mucosal surfaces, and parenteral immunization does not protect mucosal surfaces. On the other hand, the majority of lymphoid cells in the body are at mucosal surfaces, particularly in the gut, which has more lymphoid cells than all other lymphoid organs put together. The lymphoid follicles in the gastrointestinal tract and the nose are important entry or inductive sites for mucosal immune responses. Cells that are induced in one site are transported to other mucosal surfaces, making it possible to immunize the respiratory epithelium (for example) by immunizing the gut or the nose. This has led to considerable interest in nasal immunization.
Unfortunately, there are at present no adjuvants for mucosal immunization that are approved for human use, nothing in human trials, and very little even in early stages of development. Experimentally, however, both cholera toxin (CT) and E. coli heat labile toxin (two molecules with similar three-dimensional structures) have been shown to have adjuvant effects at mucosal surfaces.
For example, when mice are fed keyhole limpet hemocyanin (KLH) there is no response in the gut, but when CT is mixed with the KLH there is an IgA response to KLH. This finding has been repeated with a wide spectrum on antigens, including not only proteins (e.g., KLH) but also polysaccharides,
whole bacteria (e.g., H. pylori), whole viruses (e.g., influenza, Sendai, measles), and even whole protozoa (e.g., Toxoplasma gondii). Unlike LPS, CT and the antigen have to be administered together by the mucosal route, and in mice this usually requires microgranules. But all the properties of CT as an immunogen seem to rub off on the accompanying antigen, notably a prolonged memory response. Researchers have also succeeded in attaching peptide antigens directly to the CT molecule, with similar results.
Adjuvant Action of Cholera Toxin. CT has stimulatory effects on macrophages for IL-6 production, about a 30-fold increase. LPS has a somewhat greater effect, but the two together have even greater effect than either alone. On the other hand, CT does not stimulate TNF production, and it inhibits the stimulatory effect of LPS on TNF production.
Researchers have also achieved interesting effects by varying the sequence in which antigen-presenting cells are stimulated. Using bone marrow macrophages grown in vitro with M-CSF, for example, they found that both IFN and CT stimulate IL-6, and the combination stimulates better than either alone. But when the macrophages are “primed” with IFN for as little as 30 minutes before CT is introduced, there is an enormous increase in IL-6 production. This finding suggests that the sequence in which the immune system is stimulated may be more important than the mixture of stimulants.
Researchers are finding similar effects on production of the costimulatory molecule B7: CT alone stimulates 1.7 percent to 10.3 percent of cells to express B7, while CT and IFN together stimulate 40 percent (as measured by flow cytometry). Further analysis shows that B7 comes in two varieties, B7–1 (thought to preferentially stimulate the Th1 pathway) and B7–2 (thought to stimulate the Th2 pathway). B7–1 remains fairly constant regardless of stimulus, while B7–2 increases from 4 percent to 20 percent with CT, to 20 percent with IFN, to 56 percent with CT and IFN together. In short, CT appears to have a preferential, up-regulation effect on B7–2, the molecule that appears to be responsible for the costimulatory effect of CT and IFN. CT produces these effects down to the nanogram range, so it is a potent, very-low-dose effect of CT.
In vivo, the net effect of cholera toxin is to enhance T-cell priming, probably through its effects on macrophages. A single injection of CT plus KLH produces enhanced proliferation in response to KLH from multiple tissues, with cells producing a variety of cytokines for both the Th1 and Th2 pathways. CT also increases precursor cell frequency from 1:23,000 to 1:900. The net effect of all these biologic effects is strong stimulation.
There is a single dominant epitope for T cells on the CT molecule, a peptide designated CTV–89–100. Experiments with this peptide in mice have shown that when the animal is tolerized to CT as an antigen, CT loses its adjuvant effect—both the T-cell and B-cell responses were blocked. Other experiments have shown that the adjuvant effect is also blocked unless both the A subunit and B subunit of the CT molecule are present.
Prospects for Future Adjuvants. Adjuvants will be required for most of the emerging vaccine candidates, and it is likely that different adjuvants will be required for different antigens, particularly to shape the immune response in the most beneficial direction. At present, however, there is no detailed understanding of how the adjuvant mechanism operates at the molecular level; this understanding is needed in the future. Almost no work is being done on adjuvants for mucosal vaccines, and this should be a particular priority.
At present, the number of candidate vaccines greatly exceeds the number of candidate adjuvants, and this imbalance will only increase as genomes are sequenced for increasing numbers of pathogens. In this situation, the lack of new adjuvants may become the principal limiting factor in the development of human vaccines in general and mucosal vaccines in particular. Another limiting factor, however, will be the necessity of testing new vaccines on cancer patients; after all, is their immune system so normal?
In response to questions from the audience, Dr. Elson added the following:
Researchers have not yet repeated the CT experiments in mice at lower doses, but in all likelihood when it no longer works as an adjuvant it will also loses it effect on T-cells. The mouse experiments were done in the B6 mouse, which has the best response to CTV-89–100.
In countries where cholera is endemic, the IgA system is fully developed by age 1 or 2, whereas in the United States you would not see adult levels of IgA until the teenage years.
CT-vibrio vaccine (CTV) appears to be a stronger immunogen than CT itself, in humans if not in mice, and it may also prove to be a better adjuvant in humans. It represents one way to avoid the potential toxicity of CT.
Another way to avoid toxicity is to use multiple emulsions to sequester the toxin from the epithelium and deliver it instead to the lymphoid follicles of the intestine. This technique has been tested in mice and does not produce fluid secretion. It is much faster than developing correct microencapsulation, which also requires that the antigen be denatured. In addition, multiple emulsions seem to require smaller amounts of antigen and can contain several antigens at once.
Mucosal immune memory for antigens presented with CT extends for up to 1 year.
The adjuvanticity of CT seems to be tied to its immunogenicity; mutants that lack adjuvant activity also aren’t immunized to themselves. It might to identify a single-amino-acid mutation that lacks toxicity but retains adjuvant activity; but it is not clear how this would be accomplished.
A tremendous amount of very basic work remains to be done before this knowledge could be used to produce vaccines for use by providers. University research can handle some of the questions, but commercial partners will be needed for development.
ANTIGEN DELIVERY SYSTEMS13
To dramatize the importance of mucosal antigen delivery systems, three important concepts bear repeating:
Most infectious diseases or agents enter through the mucosal tissues. In terms of incidence and burden, this includes a billion cases of diarrhea at any given time. Bacteria such as salmonella, Shigella, vibrio, etc., spread at roughly one-half million cases per hour, just among children, 500 of whom will die. Viruses such as hemophilus, pneumococcus, and measles infect some 14,000 case per hour. AIDS infects 340 people per hour, 80 percent of 90 percent of them through a mucosal surface.
The entire immune system is stimulated by what we eat and by what compromises our mucosal surfaces. There are roughly 1014 bacteria living in the human gut. Each person ingests about one metric ton of solids and liquids per year. About 10 milligrams of undigested protein are “injected” into the circulation every day. Some 70 percent of all lymphoid cells are associated with mucosal tissues, and the gut has more macrophages than the liver.
There is a common immune system. The concept of systemic immunization, first recognized by Erlich in 1891, involves the stimulation of cells from inductive sites (e.g., Peyer’s patches in the intestine) to remote sites (e.g., nasal passages, salivary glands, even the genital tract). In both monkeys and mice, for example, intranasal immunization produces a greater antibody response in the genital tract than any other route. The instrument of this response is the M-cell, which can internalize both particulate and soluble antigens.
Research on Antigen Delivery Systems. The central problem with mucosal application of antigens is the small quantity of antigen that is absorbed, due to degradation by enzymes, mucous coatings, etc. Typical absorption rates for proteins are only 1:100,000. Taking antigen with food can reduce degradation somewhat, as can encapsulation—protecting the antigen from the stomach in order to deliver it to a desired segment of the intestine.
A few groups in Europe and the United States are using liposomes to deliver antigens. In the mouth, this can reduce the formation of plaque and the incidence of dental caries. However, liposomes are relatively unstable in the internal environment. Others are working with mucosal adhesives, microspheres, CT and CPB conjugates, and live pathogen delivery systems.
The goal of this research could be called the “dream vaccine” —something safe and stable that might be produced locally, could be administered orally in one dose to provide both systemic and mucosal immunity, possibly to several
different pathogens at the same time. At present, 60 percent of the cost of vaccine delivery is spent on refrigeration, and another large fraction on syringes and health professionals. Some of these vaccines would make it possible to immunize thousands or even millions of people in a single day without doctors, needles, or pain.
Mucosal Adhesives. Other researchers are developing new mucosal adhesives, which increase absorption by extending the period of time during which antigen is exposed to the mucosal surface. Substances such as carboxymethylcellulose adhere very nicely to the mucosal surfaces and could be used to deliver antigen to those tissues. Investigators in Iceland have tested mucosal adhesives containing influenza virus in the nasal cavity of animals and even humans. It produced stronger and longer-lasting immune response than the injectable vaccine.
Microspheres. Microspheres are already in common use for the delivery of drugs and are being developed for the delivery of hormones and insulin. Because M-cells preferentially internalize particles, it should be possible to use them to introduce soluble antigen to the macrophage. A company in Korea has applied to develop a hepatitis-B vaccine using microspheres that has the considerable advantage of requiring only one injection. (Compliance is a problem with the current vaccine, which requires 3 injections: only 60 percent of patients return for the second shot, and 40 percent for the third.) The spheres themselves are totally biodegradable and nontoxic; varying the composition regulates the rate (and hence the location) at which individual microspheres dissolve.
Injectable microspheres can be up to 50 microns and still be effective, and in the gut these larger spheres collect in the Peyer’s patch, where they induce a mucosal immune response (circulatory IgG). They must be less than 10 microns to be absorbed by the intestine. These smaller microspheres can later be found in the mesentery lymph nodes and spleen, as well as in circulation. In some cases (e.g., SIV) they can also induce T-cell-mediated immunity in cytotoxic T-cells, although this is not seen with other antigens (e.g., influenza and polio virus). The reasons for this difference should be addressed in future research.
Obviously, the prospect of a vaccine that could be eaten and would induce both mucosal and system immunity is very attractive. Because microspheres protect against degradation by gastric acid in the stomach, 100 percent of the antigen reaches the gut. At present, however, less than 1 percent of ingested microspheres are internalized. The rate of absorption might be increased through surface modifications or coadministration with CT or CPB, which has increased absorption in animals.
Other development problems also remain. In numerous animal experiments, antigen in microspheres induced much high titers of specific antigen that raw antigen. However, several vaccines that work in animals haven’t worked in humans. In some cases, the organic solvents used in preparing the microspheres can denature the antigen unless it is relatively stable. And improvements may be needed in the reproducibility of microsphere preparation.
Live Virus and Bacteria. Another delivery system that is attracting attention is the use of recombinant technology to create yeast viruses and bacteria that express antigen. This has already been done with hepatitis B, removing DNA from the pathogen and introducing it into a vector that then produces the immunizing antigen. Some potential vectors can be administered intranasally, orally, and/or rectally. It may also be possible to induce more than one antigen in a single vector, and the cost of producing these vaccines is fairly low.
Among the bacteria being explored as vectors are E. coli, BCG (attenuated Mycobacterium bovis), Shigella, and lactobacilli. These microbial vectors are immunogenic by themselves, and consequently they cannot be used for a second immunization. Lactobacilli show particular promise because children are usually colonized at an early age, so the immune response is minimal. There is some risk of contaminating the environment with recombinant bacteria.
Viruses may have many advantages, including better control over the conformation and glycosylation of the expressed antigen. Promising vectors include poliovirus, influenza, canarypox, and rhinovirus. Both influenza and canarypox have been used to express HIV antigens in human experiments.
These vectors have disadvantages that must be overcome. For example, the dominant response is always to the vector rather than the antigen. The major problem for immunology, however, is that the level of expression of the desired antigen is extremely low; this might be overcome by inserting multiple copies of the gene. In addition, bacterial vectors do not always produce antigen with the proper secondary and tertiary structure.
Plants. Several groups are looking at plants as possible delivery systems. The tobacco plant in particular is called “the white mouse of botany” because its genes are well known and easy to manipulate using the tobacco mosaic virus, which can infect 400 other plants. Researchers in England have already succeeded in producing hepatitis B antigens in tobacco leaves. Other plants that are being considered as delivery systems include potato, beet, rice, lettuce, tomato, and bananas. Bananas would be particularly attractive in tropical areas, but the intended antigen is produced at low levels—100 micrograms of hepatitis B protein per banana, rather than the 10 milligrams researchers had hoped for.
Plants can also be used to produce specific antigens, including secretory IgA containing the secretory element. This has been done by introducing genes for the heavy chains of IgA into tobacco, and could conceivably be done with edible plants as well. The result could be an edible delivery system for highly specific secretory IgA antibodies against rotavirus, salmonella, etc., as well as antigens. This could become one of the vaccines of the future.
DNA Immunization. DNA vaccines are covered in greater detail in the following summary. However, this is a difficult route of immunization because of generally poor reproducibility and low uptake of DNA. It may be possible to increase uptake using microspheres or other systems. It is unclear at present
whether DNA introduced through the Peyer’s patch will lead to the synthesis of desired antigens, as it does when injected into muscle or skin.
Barriers to Vaccine Development. Tolerance is a major problem with some of these delivery systems. In addition to reducing the immune response, this can also lead to systemic unresponsiveness or even collapse at the T-cell level, the B-cell level, and even the antibody level. This can be addressed by changing the form or delivery of the antigen—globulin given in large oral doses induces tolerance, but globulin in microspheres continues to produce a good immune response. Frequency and timing of exposure also plays a role. Studies of humans immunized with keyhole limpet hemocyanin (KLH) demonstrated that oral administration “primes” the systemic response and reduces both hypersensitivity and tolerance. The nasal route appears to be more effective than eating the antigen.
Other Applications. These delivery systems might be used to introduce proteins that would treat or protect against autoimmune diseases such as rheumatoid arthritis and multiple sclerosis. Another possible application might be in the control of fertility—researchers working with rats have found that the sperm antigen SP-10, introduced orally, produces specific secretory antibodies against sperm in the female genital tract. The result is a temporary infertility mediated by specific IgA, temporary because IgA immune response is not long-lasting.
In response to questions from the audience, Dr. Mestecky added the following:
“Immune tolerance” is the wrong name for what happens in oral adminisration. “Mucosal deviation” might be better, since it changes the response rather than causing total unresponsiveness.
The intranasal route induced better tolerance that the oral route with the same antigen in mice. This work involved animal models of arthritis.
Antigens produced by plants may not be as “naked” as desired. Antigens like HIV, SIV, and GP-120 and 160 are generally heavily glycosylated, but glycosylation in plants is different from that in mammalian cells.
Since the immune response to the vector is a limiting factor, systems that are inert might be preferable to live vectors.
Parenteral immunization is relatively ineffective in inducing mucosal immunity. Studies with a variety of antigens demonstrated that it is almost never possible to boost the secretory IgA response by systemic immunization. Similarly, IgG in the genital tract is not derived from the circulation; systemic IgG cannot prevent the pathogen from crossing the mucosal membrane.
Systemic, intranasal, and internal immunization produce very different patterns of potential homing receptors on the surface of B-cells. Researchers have not looked at receptors on T-cells yet.
Unlike other vaccines, which deliver the antigen itself in some form, a DNA vaccine delivers only the genes that encode for the antigen of interest. In this sense, it is nonreplicating bacterial plasma in a fairly generic expression vector that is expressed by mammalian (and ultimately primate) muscle tissue. These are not yet clinical entities, since none have been approved as yet for use in humans. However, researchers have had success with bacterial and viral antigens in a number of animal experiments, and work is beginning in England on cancer antigens as well.
Research in this area began with a paper published in 1990, detailing a process for transfecting genes into cells in vitro. When reporter genes were put into expression vectors and injected intramuscularly, the tissue expressed the protein encoded in the gene. It is still not known why muscle tissue expressed protein better that other cell and tissues types. Researchers estimate that no more than 1,000 cells actually take up the DNA and express the protein. This technique did not appear promising for gene therapy, but since there was amplification of the immune system, it might be a good way to generate antigen in situ.
Researchers had been looking for a way to deliver antigens to the cytoplasm of the cell, where their epitopes would evoke class I restricted cytotoxic T-lymphocyte responses. Because every individual’s MHC molecules are different, this would allow different HLA haplotypes to select their own epitopes, instead of concocting a “cocktail” of different types to cover an entire population. Muscle cells aren’t usually thought of as antigen-presenting cells, but the DNA vaccine approach did provide a way to have the antigen produced endogenously by a host cell.
Viral DNA Vaccines. For a protection experiment based on cytotoxic T-cell response, researchers extracted the gene-encoding nucleate protein from influenza virus. This internal protein is highly conserved between different strains of the virus. Mice were immunized with the gene taken from an H1-N1 strain from 1934, and then challenged with a very different H3-NT strain called Compound 68. The results showed that there was a cell-mediated response, and that it was protective. Antibodies were generated as well, but since they were antibodies to an internal protein this response was not protective.
Researchers have since improved on both the vector and the expression of the target protein. In recent tests, 100 percent of immunized animals survive, while 90 percent of controls die. Controls receive either saline immunization or “control DNA” that does not encode for the nucleate protein. Two years after immunization, there is still cytotoxic T-lymphocyte activity that is specific for peptide target cells and influenza-infected cells. Since this is a class I restricted
peptide, this is a CD8 response. Protection does not appear to persist as long as the CTL response.
Another group has done similar experiments with aged mice. They were able to get CTLs with the same precursor frequencies in the aged mice as in younger mice, with good antibodies and good small-dose protection. This is an important issue in diseases like influenza, where the elderly are the population at greatest risk.
To determine whether is would also be possible to get efficacious B-cell and antibody responses, as well as CTL response, researchers repeated the experiment using the gene encoding for influenza hemagglutin, a surface glycoprotein that is the antigen against which the major neutralizing antibody is directed. (These proteins change frequently, which is why the current influenza vaccine must be reformulated every year.) The result was a very good level of functional antibodies, with titers comparable to live infection and probably high enough to be protective in humans. These antibody levels have been sustained for 20 months, although duration might not be as great in humans. Researchers are uncertain whether antigen is still being expressed after this interval; one study suggests that the DNA persists and expresses for up to 24 months.
One unexpected finding was that the DNA vaccine encoding for hemaglutin induced antibodies with a different profile from those induced by commercial vaccines containing inactivated influenza, notably a greater predominance of IgG2a. The latter is thought to be important in providing protection from influenza. Doses as low as 1 microgram, given twice 3 weeks apart, provide 100 percent survival in mice and little or no morbidity as measured by weight loss and grooming behavior. Further tests indicate that DNA vaccines also stimulated greater lymphocyte proliferation from splenocytes and higher levels of IL-2 and IFN-gamma on restimulation. This and the absence of IL-4 point to a Th1-like helper response.
HIV DNA Vaccines. It is increasingly accepted that the same kind of Th1-like helper response and CTLs are major factors in preventing HIV infection, or at least prolonging the phase of latent disease. This is supported by recent findings that high-risk individuals who remain seronegative have good Th blood-type responses, and that those with the less virulent HIV-2 are at decreased risk for HIV-1.
Initial work on HIV envelope proteins focused on surface antigens such as B3, GP-120, and GP-160, looking for a neutralizing epitope and trying to make it recombinant. It has become increasingly clear that the antibody response is more complicated, and that memorization in particular involves dimeric dimers of GP-160 that form conformational epitopes. Accordingly, effort has shifted from peptide vaccines to ways of making intact oligomeric GP-160. DNA vaccines offer the advantage that the protein is synthesized in situ, with normal glycosylation.
Researchers gave African green monkeys two injections of a DNA vaccine based on a truncated GP-160-like construct. Animals developed relatively high antibody titers that show diverse neutralizing responses, which is desirable given
the diversity of HIV types. However, antibodies are not at levels that would be protective against SIV challenge. Researchers hope to improve antibody titers in the future.
To look for CTL responses, researchers also vaccinated monkeys with a DNA vaccine based on envelope proteins with known CTL epitopes. The animals developed good CTL response, with precursor frequencies as high as animals that have been infected with SIV. This may not be high enough to challenge at present, but researchers are encouraged by the combination of antibody and CTL response. Future work may focus on conserved envelope protein such as GAG and POL.
Similar results were obtained in mice immunized with DNA GP-120 vaccine. When splenocytes were restimulated in vitro with GP-120, they proliferated and produced a lot of IFN-gamma and IL-4—again characteristic of a Th1-type response. Lymphocytes from other sites—mesenteric, iliac, inguinal, and peripheral blood—also proliferated in response to antigen, even 6 months after vaccination in the quadriceps. This kind of long-lived response, with both cytokines and helper CTLs, would be particularly desirable in an HIV vaccine.
TB DNA Vaccines. TB remains a major health problem, with perhaps one-third of the world’s population infected, and it is a growing problem with multidrug resistance in the United States. An effective vaccine would be important at least until new antibiotics are available. Researchers prepared a vaccine containing the DNA coding for antigen Ag-85, one of the secreted antigens of TB, and were able to induce CTLs. These lymphocytes were not antigen-restimulated, however, but rather were ConA- and IL-2-restimulated. Vaccinated animals also produced IFN-gamma and GM-CSF, the latter also being important for protective response.
In preliminary challenge studies, immunization with DNA Ag-85 protected just as well as BCG against intravenous and inhalation challenges. This may not be a very good vaccine—yet—but it is working as well as the vaccine that many people are getting now. The next step may be to add genes for additional antigens.
Researchers had thought that one of the advantages of DNA vaccines would be mammalian post-translational modifications in the antigen. In the case of TB, however, this was not necessarily desirable. Fortunately, it was fairly easy to remake the vectors—cleaving Ag-85 and altering the leader sequence, glycosylation points, etc., to get the desired form. Nevertheless, researchers still weren’t sure whether they were getting true oligomerization of the resulting envelope proteins.
HPV DNA Vaccine. L1, the major capsid protein of human papillomavirus, forms pentimers that come together to form the capsid. The neutralizing antibody is directed at conformational epitopes, so the protein must have the correct structure and oligomerization to act as antigen. This protein is targeted to
the nucleus, where new viruses are assembled and then excreted from the cell. For these reasons, it wasn’t clear that the DNA vaccine would work.
Researchers prepared DNA vaccine using the L1 protein from rabbit papilloma virus and immunized rabbits. Following a single immunization, animals made a good titer of antibody that has persisted out to 32 weeks, although this doesn’t predict what will happen in humans. The animals are challenged by scarifying the skin and applying virus, resulting in a wart. This is not how infection occurs in humans, but it does allow measurement of neutralizing antibodies. Animals immunized with L1 DNA were all protected from getting warts, so at least in this model it was possible to get neutralizing antibodies to a protein that forms a conformational epitope.
Herpes DNA Vaccine. HSV is an interesting case because there is a good animal model for mucosal challenge. Researchers would like to develop a mucosal delivery mechanism for DNA vaccine, since the current procedure does not produce mucosal antibodies.
In the first study, mice were vaccinated with DNA encoding for the surface glycoprotein GC and challenged intraparenterally with a lethal dose. About 90 percent of controls died, while animals immunized with as little as 800 nanograms of vaccine were protected to some extent.
In a second study, guinea pigs were immunized and then challenged intravaginally. This is thought to resemble closely what happens in humans. The results showed that immunizing the animals with one or two doses of DNA vaccine provided very good protection from the lesions that occur in the group injected with saline solution.
Other Approaches. The DNA vaccine technology is relatively simple: just put a gene into a vector and immunize with it. But in less than 3 years since the first study was published, the technique has been proven effective against a broad range of targets, including parasites as well as viruses, microbacteria, and cancer. One group has used the technique to construct “expression libraries,” in which an organism’s genome is broken into perhaps 2,000 fragments and then used to immunize experimental animals that are subsequently challenged. A decoding process identifies which antigens elicited protective immune responses. The process is slow and animal-intensive, and there may be problems with antigenic competition or partial conformations, but this approach does offer potential application for vaccine discovery efforts for antigens.
Issues in DNA Vaccine Development. Three sets of key issues raise questions about the ultimate importance of DNA vaccine technology: (1) safety, (2) human efficacy, and (3) process and stability. The stability issue in particular poses the danger of overselling the potential of the technology.
There is a need for a full safety profile. Some of the key points raised by reguatory personnel concerned subintegration: i.e., does insertionial mutagenesis occur, are pathogenic anti-DNA antibodies generated, and does tolerance occur? To date there is no evidence of any of these problems. Researchers have performed a very sensitive PCR-based assay on tissue from 13 different organs and found no evidence of the inserted DNA. Mutations are less likely in double-
stranded DNA than in single strands. Tolerance is a more difficult issue, but researchers are monitoring it. One advantage of this approach is that there is no danger of viral disease, since the vaccine does not contain replicating virus.
On the question of efficacy, native protein is made endogenously, with the appropriate mammalian post-translational modifications, and it induces neutralizing antibody even across conformational epitopes. There are cytotoxic T-lymphocytes and cross-strain protection. The muscle cells, while they may not be antigen-presenting cells, definitely do transfer antigen, and there is good helper T-cell response.
On the question of production, it is (or will be) a generic technology. Once the manufacturing process is in place, producing different vaccines will be no different than producing different flavors of ice cream. The process is the same, and so is the host strain. The technology will also be useful as a laboratory tool for producing and screening antibodies.
In response to questions from the audience, Dr. Liu added the following:
It would appear that the DNA itself is modulating the Th1-like response. Recombinant nucleate proteins induce antibodies to recombinant nucleate proteins, but when you add plasma DNA—even if it doesn’t encode for anything—you get a Th1-like response with predominance of IgG2A. In a sense, the DNA acts as an adjuvant.
Researchers don’t know whether the muscle cells that are generating antigen will be attacked and lysed by antigen-specific CTLs. For one thing, they are multinucleated; for another, they are very large. In any event, very few muscle cells are transfected: the number of plasmic copies drops from 2,000 initially to about 1,500 at 4 weeks and 500 at 18 weeks.
Researchers have tried various routes of immunization. Subcutaneous injection produces good antibody response but no CTL or helper T-cell response. They have also applied DNA directly to mucosal tissues; the result was IgG antibody response without IgA, and no CTL response. Nevertheless, their goal is to move toward mucosal delivery if only because of the ease of delivery.
Researchers have not yet investigated the class II response in detail.
Histologic studies of transfected muscle cells reveal no inflammatory myopathies. The needle track remains visible and is infiltrated with lymphocytes, but the response was much less marked than the response to an antibiotic that is injected intramuscularly and known to be a mild irritant.
Researchers still don’t understand the exact mechanism by which antigen expressed in muscle cells induces antibodies. Clearly it makes it out of the cell, but that could be due to (1) transfer by the myoblasts or (2) transfection of antigen-presenting cells that happen to be in the muscle tissue. They plan to pursue this question using experiments in which transfected myoblasts are transplanted to bone marrow chimeras.
The DNA plasma in the vaccine is not integrated, because the gene itself does not replicate, even though there is expression of the antigen that it encodes for.
The TB studies have looked at three major forms of Ag-85—A, B, and C—and at different variations of these forms (with and without leaders, etc.). They are just now beginning to work with ESET6, the other major secreted antigen that is believed to provide some protection.
It is very difficult to say how soon there might be a vaccine for humans. Work seems to be on track in terms of addressing safety concerns. The biggest unknown is whether they will be allowed to test an IND in healthy humans instead of cancer and HIV patients as is currently the case. A great deal depends on the regulatory agency.
The influenza vaccine is the most advanced candidate, even though it may not be the best choice for a DNA vaccine because of antigenic drift. And there are influenza vaccines already, but no vaccines for herpes, HIV, etc.
Studies have not been conducted with newborn mice, because of their very small size. Studies have used mice from 3 weeks to aged. There might be some advantage to using this technology for something like a measles vaccine to cover the period of 6 to 12 months, before maternal antibodies decline. But this will not be the first population that a commercial laboratory would focus on because of regulatory problems.
Tolerance is unlikely to be a problem unless expression is boosted considerably, or unless the dosage is increased significantly. However, more needs to be known about the phenomenon of tolerance induction.
From a safety point of view, it might be desirable to transfect a more differentiated but shorter-lasting cell, such as a dendritic cell, rather than a muscle cell. Researchers will probably continue to explore this question.
PREVENTIVE VACCINE FOR DIABETES15
Clinicians have noticed that when insulin replacement therapy is begun, it often appears to reverse the disease process. This led to experiments in which non-obese diabetic (NOD) mice were given daily prophylactic doses of insulin at the outset of the inflammatory lesion in the pancreatic islets. The results showed that diabetes is not only preventable, but also reversible. Photomicrographs of the pancreas of an NOD mouse showed some islets that were completely free of lymphocytic infiltration, while others were full of lymphocytes.
This raises the question of mechanism, and one possibility is that of beta cell rest: if the animal receives insulin every day, the islets don’t need to produce it, and the suppressed cells don’t produce autoantigens that are reactive to the immune system. In this case, insulin itself seems to be acting in some speci-
fic immunological way. Based on this experience in NOD mice, the researchers have organized a Diabetes Prevention Trial Type 1, which is ongoing. They have also conducted further experiments in antigen-specific immunotherapy in the NOD mouse.
The strategy in general is to tolerize against the harmful Th-1 immune cells, and if possible transform the response from a harmful Th-1 to an ostensibly harmless Th-2. This can be accomplished by immunizing with various antigens and adjuvants, either orally or intravenously. Three possible autoantigens have the greatest relevance for human diabetes:
Glutamic acid decarboxylase (GAD), usually GAD-65 rather than a higher molecular weight isoform;
Transmembrane tyrosine phosphatases, one called IA2 and the other IA2-beta; and
Insulin and insulin receptors.
Experiments with IA-2 and IA-2-beta are ongoing; results are available for experiments with GAD and insulin.
Oral Antigen Therapy Experiment. In one experiment, NOD mice were fed doses of 1.0 milligram (mg) of insulin or 0.5 mg of GAD (from pig brain) per day. Over time, most of the controls developed diabetes, while insulin and GAD both provided significant but not absolute protection. Feeding the two antigens together seems to have some additive effect, particularly in maintaining protection. At about 12 weeks of age, when diabetes begins to occur in the NOD mouse colony, the insulitis score (based on level of inflammation observed by microscope) is much lower in mice fed GAD, insulin, or both.
This anergy or suppression of T-cell response is antigen-specific: feeding insulin suppresses insulin responses; feeding GAD suppresses GAD responses; and feeding together suppresses both, but there is no bystander suppression. Most of the antibodies in the NOD mouse are of the IgG-B2 variety, and oral feeding doesn’t suppress the animal’s ability to make antibodies, or switch the subtype of antibodies made.
Two cytokines of interest are induced: tumor necrosis factor beta (TNF-beta) and interferon gamma (IFN-gamma). Feeding regimes did not change TNF-beta levels, but controls had much higher levels of IFN-gamma in islet infiltrates, most of it from Th-1 CD4 cells. This is evidence of down-regulation of Th-1 response in animals fed insulin and/or GAD. In the Peyer’s patch, too, researchers found an inhibition of elaboration of messenger RNA for IFN-gamma, as well as increases in interleukin four (IL-4) and possibly IL-10, evidence of down-regulation of Th-1 response and possibly of up-regulation of Th-2.
Intravenous Antigen Therapy Experiments. Large amounts of IV insulin will kill the mice, so researchers used biologically inactive forms, including
A chain, B chain, desoxyinsulin, and pro-insulin, as well as human GAD. In this case, only GAD was effective in providing protection.
Subcutaneous Immunization Experiments. Researchers immunized animals at 4, 8, and 12 weeks. Controls received incomplete Freund’s adjuvant (IFA) with a diluent, which had been shown to give no protection. (Mycobacterium antigen in a complete Freund’s does give marked but nonspecific protection.) Experimentals received A chain, B chain, or whole insulin with IFA. The A chain provided insignificant protection, but animals immunized with B chain developed marked protection that was transferrable to irradiated animals. The protective epitope turns out to be amino acids 9 through 23 on the B chain of the insulin molecule. Similar experiments are now under way with GAD peptides and various whole GAD molecules.
Since oily adjuvants such as IFA cannot be used in humans, researchers have investigated other adjuvants that might be approvable. Alum plus B chain proved to be effective, although alum by itself may have some effect as well. Diphtheria, tetanus toxoids, and pertussis vaccine (DTP), a common childhood vaccine that contains alum, also proved to be a good adjuvant, at least in mice. Clearly, however, DTP itself causes a nonspecific stimulation that biases the animal to a Th-2-type peripheral phenotype. Analysis of cytokines produced in DTP-treated animals shows an increase in IL-4 and a big increase in IL-10, but not the same decrease in IFN-gamma that was seen following oral administration.
This process is not anergy but instead an immune response. Responses to GAD are not decreased but actually enhanced. Researchers believe that this is an IL-4-driven response. Regardless of regime, including DTP by itself, the treated animals had increased responses to GAD. Consequently, this is an active process.
Conclusions. Oral administration of autoantigens may be leading to tolerance. Intravenous administration of insulin is relatively ineffective in delaying diabetes. Subcutaneous administration in IFA proved to be the easiest route and to have the longest-lasting effect. Alum and DTP are effective substitutes for IFA as an adjuvant. The process seems to involve, at least in part, switching from a destructive Th-1 to a protective Th-2 response. These findings could now be subjected to human trials, if the B chain epitope is similarly reproduced in humans.
Other groups have reported that the T-cells of NOD mice have a high density of receptors for insulin, and that transfer of these cells to irradiated mice uniformly transferred diabetes as well. It is possible that immunization against the B chain of insulin deviates these cells from homing on the pancreatic islets that express insulin. Researchers have already constructed a viral vaccine designed to immunize against the B chain of insulin experiments are ongoing, and they look forward to subjecting it to provisional human feasibility studies.
In response to questions from the audience, Dr. McLaren added the following:
Researchers repeated the oral experiments using pig brain GAD and eventually baculoviral-expressed GAD to remove questions about specificity of response. However, they also found that—while there isn’t an additive effect—the best results are obtained by giving both antigens.
Researchers are now looking at costimulation and using antibodies such as neutralizing IL-4 and neutralizing IL-10.
Transfer of protection depends on both CD4 and CD8 cells, but mainly CD4.
Researchers have not looked at the question of prior maternal sensitization, but they have found that age at time of immunization is important.
These experiments used human insulin, but insulin is a highly conserved molecule. Human and pig insulin differ by only one amino acid, human and mouse by only three amino acids. Human and pig brain GAD are somewhat less conserved, perhaps 70 percent homologous. The possibility of foreign protein response makes it difficult to factor in the results.
CYTOKINE MODIFICATION OF AUTOREACTIVITY16
There are two general theories with regard to autoimmune disease:
The antigen is the primary effector of the immune response, but a pathogenic response leads to disease; and
Disease is a problem of immune regulation.
It is well known that Th-1 and Th-2 cells counterregulate each other. Th-1 cells are thought to promote autoimmune disease, but their action can be counterre-gulated by Th-2 cytokines, particularly IL-10 and IL-4 and, to a lesser extent, tumor growth factor beta (TGF-beta).
Working with NOD mice, which are a good model for human diabetes, researchers first tried to determine whether IL-10—whose primary role is to down-regulate IFN-gamma (the prototypic Th-1 cytokine) —could regulate this autoimmune response. This was done by eliciting the cytokine locally, in the pancreatic islets of a transgenic (BALB/c) mouse, using the insulin promoter. This attracted inflammatory lymphocytes, but this response did not induce diabetes in a nonsusceptible mouse.
IL-10. The next step was to see whether IL-10, which down-regulates IFN-gamma, is capable of regulating the disease in the NOD mouse. Researchers introduced the genetic loci onto the BALB/c background by backcrossing to F1s, then backcrossed again to N2s, at which point they expected to
see a 10 percent incidence of diabetes. Instead, they found a 94 percent incidence of diabetes in MHC-susceptible animals and no diabetes in IL-10-negative, MHC-susceptible animals. Mice with BALB/c MHC did not develop diabetes.
In other words, the effect of IL-10 was completely the opposite of what they had expected—not only did it accelerate the disease process, it also seemed to overcome the requirement for a lot of susceptibility information. (There are numerous susceptibility alleles in the NOD mouse, and more are being mapped all the time.) It seemed that IL-10 was actually a very strong potentiator of disease, at least in this and other transgenic mouse models. To see whether or not IL-10 plays a role in the natural disease, researchers have conducted depletion experiments with neutralizing (anti-IL-10) antibodies. In one experiment, NOD mice that were treated with anti-IL-10 from a young age developed far less insulitis than controls. In another experiment, irradiated NOD mice were injected with sensitized lymphocytes from diabetic mice and then treated with anti-IL-10, but the neutralizing antibody had no effect in this transfer model, which involved older (9-week old) mice.
Researchers interpret these results to indicate that IL-10 is required early on in the natural disease, but that it is dispensable later on. NOD disease differs from other autoimmune diseases in its diversity—many new antigens are still being identified, and a corresponding variety of T cells mediate and regulate the disease. Hence, the progression of the disease seems to involve three stages:
Presentation (seeing the antigen or antigens);
Diversification (sensitizing lots of T cells to islet antigens); and
Effector phase (leading to destruction).
Researchers hypothesize that IL-10 acts in the diversification process, specifically through its ability to activate B cells and (through them) to diversify the immune response. Researchers are conducting further experiments to test this hypothesis, using anti-IL-10-treated NOD mice, but focussing on responses to peptides of GAD. The two immunodominant determinants, GAD 34 and GAD 35, respond similarly in the presence or absence of anti-IL-10, but the neutralizing antibody inhibits the spreading of the cryptic determinants, GAD 23 and GAD 17. Since there are high ratios of B cells in the islet infiltrates, this response may be critical to the clinical manifestation in the natural NOD disease, and possibly in the human disease as well.
IL-4. Unlike, IL-10, IL-4 totally blocks the disease. This colony of NOD mice has an 85-percent incidence of diabetes, usually developing the disease between 16 and 24 weeks. Targeted IL-4 to the pancreatic islets reduced the incidence of diabetes to zero, for over a year. To determine whether this reflects a state of tolerance, researchers transplanted NOD islets into transgenic mice that express IL-4 and looked for signs of rejection. The graft was accepted, indicating tolerance of an unknown mechanism.
When researchers reversed the transplant, however—putting islets expressing IL-4 into diabetic mice—the grafts were rejected. The beta cells in the transplanted islets were destroyed, while the alpha cells remained stupidly in the graft. This suggests that, once autoreactivity develops, you cannot stop committed or primed cells from killing the islet.
On the other hand, when researchers transplanted splenocytes from NOD diabetic mice into an NOD IL-4 transgenic recipient, they again saw protection, while disease was transferred when splenocytes were transplanted to NOD controls. Researchers recognize that these are contradictory results and suggest that the difference has to do with the “audience.” That is, circulating cells have a much higher proportion of primed, or memory cells, whereas the spleen contains both primed and unprimed, memory and naive cells; the presence of at least some naive cells in the tissue being transferred allows for some regulation. Researchers are currently testing this hypothesis.
These results suggest that the immune system must see antigen in order to produce tolerance, but only if it is seen in the presence of IL-4. If the antigen is seen alone, the result is destruction, and seeing it a second time in the presence of IL-4 doesn’t seem to give any kind of regulation. They also suggest that the T-cells enter the islet naive, and that priming occurs within the islet. This in turn suggests that therapy could be accomplished locally if it were possible to “speak” to the cells as they entered the islet. This could be done more elegantly if it were possible to merely add the proper determinants, but this will require a better understanding of natural Th-2 determinants, conditions for priming, site of priming, cytokines, etc.
In response to questions from the audience, Dr. Sarvetnick added the following:
The IL-4 in these experiments was murine IL-4, so this could not be the reason for the rejection.
An IL-4-producing islet from a BALB/c donor is not rejected when transplanted into a BALB/c recipient, only by an NOD recipient.
The results with TGF-beta are far less interesting—essentially a fibrosis of the pancreas without any sign of blocking diabetes, at least in their lab. Other researchers have found some signs of inhibition.
Researchers have not looked at the pattern of B-cell development or how it is altered in transgenic mice, except with regard to IFN-gamma. B7–1 and B7– 2 are in the infiltrates in the NOD mouse, but they have not looked at whether that is down-regulated. It would be a good experiment.
Slides of the islets of transgenic mice show lymphocytes surrounding but not infiltrating the islets. This suggests that the protective effect of IL-4 may come from altering the receptors or adhesion molecules on lymphocytes.
Researchers have not looked at endothelial markers extensively.
Researchers believe that the homing patterns of primed Th-1 cells are different from those of primed Th-2 cells. In addition, Th-2 cells seem to turn over very rapidly—there is circumstantial evidence of Th-2 cells homing away from the islet and dying in the adjacent lymphoid aggregate.
Cytotoxic T lymphocytes (CTLs) appear to be involved in the initial insult, but not enough to cause disease or killing. Experiments with NOD Class I knockouts showed that both CD8 and CD4 cells are required. Results suggest that IL-10 can attract CD8 cells and enhance CD8 CTL activity in-vitro. However, researchers found relatively few CD8 cells in the infiltrate compared to CD4s and B-cells. Breeding a Class I knockout onto an NOD mouse would apparently require nine backcrosses.
Noel Rose of Johns Hopkins University described a new model he and his coworkers have developed showing the role of IFN-gamma in inducing autoimmune thyroiditis. This involves transgenic mice that express IFN-gamma in response to thyroglobulin promoter. The results are dramatic: the mice develop severe thyroiditis, their thyroids are almost completely infiltrated and destroyed, and they become profoundly hypothyroid, with very low levels of circulating T4. All of this is caused by local expression of IFN-gamma in the thyroid. The mice in question were C57 blacks, a susceptible strain.
T-CELL SUBSET CHOICE ON THE OUTCOME OF AUTOIMMUNE DISEASE17
Transgenic T-cell models allow researchers to ask questions that cannot be asked in clinical disease, primarily because they limit the dramatic diversity of diabetes. In both humans and NOD mice, diabetes involves a plethora of antiens and T-cells and cytokines. Researchers try to constrain these variables in such a way as to gain some insight on the total disease.
When a naive T-cell first enters the islet or encounters antigen, it becomes a Th-0 or initially activated cell that can produce both IL-4 and IFN-gamma, as well as a host of other lymphokines. IFN-gamma will stimulate macrophages to make IL-12, which will drive the Th-0 cell towards the Th-1 phenotype. IL-4, on the other hand—either autocrine or paracrine (from a cell not yet defined, possibly a mast cell or CD 1-restricted CD4-positive T-cell) —will drive the Th-0 cell towards the Th-2 phenotype. These are the two very distinct arms of the immune system. IL-4 feeds back and down-regulates IL-12 responsiveness, and IL-12 down-regulates IL-4 responsiveness—a very nice polarizing system. Other researchers have shown that T-complement (Tc-1 and Tc-2) cells are very analogous to the Th-1 and Th-2 phenotypes, with the same signature lymphokines.
Other Disease Models. Researchers have found interesting dichotomies involving these same pathways and lymphokine profiles; in nematode infections, for example, giving IL-4, results in lower fecundity of eggs in the gut and more rapid shedding of the eggs, with a curative effect. In leprosy, the tuberculoid form of the disease produces more IFN-gamma; and IL-2, while the lepromatous form produces higher antigen levels and more CD8 cells, IL-4, and IL-10. In listeriosis, the pathogen induces macrophages to release IL-12 that drives the response down the Th-1 pathway.
Experiments have show that genetic influences may establish “default pathways” in certain individuals or strains—predispositions that researchers will need to consider when developing vaccine strategies. In response to the leishmaniasis pathogen, for example, B10.D2 mice produce an initial burst of IL-4, followed by rising levels of IL-12 and ultimately a Th-1 response with rising levels of IFN-gamma; while BALB/c mice produce a sustained level of IL-4 with rising levels of IL-12 and ultimately a Th-2 response with low levels of IFN-gamma. In this case, it appears that the polarizing variable is not the IL-12 per se but the IFN-unresponsiveness of the BALB/c mouse compared with the B10.D2 mouse. Further experiments confirmed that the difference between the two responses lay in a constitutive part of the IL-12 beta chain; a surface component (which is inducible by IL-12) allows the IL-12 molecule to bind to the Th-1 cell surface, actually competing with the signalling pathway that tells the Th-1 cell to produce IFN-gamma.
Diabetes. Type 1 diabetes accounts for about 7 percent of all diabetics in the United States, some 1.4 million individuals. Since this is a T-cell-dependent disease, the relevant questions are (1) what kind of T-cells transfer the disease and (2) how these T-cells develop. Sequencing the T-cell receptor of a Class II-restricted, CD4-positive clone revealed nothing that would mark it as a diabetogenic T-cell. Researchers developed a transgenic mouse model in which to observe T-cell development in the thymus and periphery, and in which to learn how to control that development. The transgenic was crossed onto the C-alpha knockout background, because they lacked a neutralizing antibody for the endogenous alpha chain.
The result is a line of “hypertransgenic” mice with CD4-positive T-cells that differ from the original phenotype only in the expression of the V-beta-4 T-cell receptor. These mice, as well as the regular transgenic mice, develop destructive insulitis at about 120 days, with infiltration of the beta granules and loss of insulin production. The next question is what type of cells are important in transferring disease.
In an earlier experiment, islet-specific T-cells were generated and then forced their development towards either Th-1 or Th-2. When transferred into the mice, diabetes developed only in the Th-1 recipients. This experiment had no control for antigen specificity or receptor affinity and avidity, which play a role in determining Th-1 vs. Th-2 responses. However, researchers could use the
transgenic mouse, which had only one particular T-cell receptor, to control for this diversity. This “mixing” experiment is described below.
Researchers removed peripheral T-cells from the spleens and the mesenteric and inguinal lymph nodes of transgenic mice and polarized them by stimulating with concanavalin A (conA) in the presence of recombinant IL-2, IFN-gamma, and anti-IL-4 (to generate Th-1) or in the presence of recombinant IL-4 and anti-IFN-gamma (to generate Th-2). The resulting T-cells had a high level of expression of D-beta-4 TCR, up-regulated the proper activation markers (CD44 and CD25), and down-regulated CD62L (also called L-select). After stimulating with islet cells and culturing for about a week, researchers transferred the cells into perinatal (5 to 7 days old) NOD mice and waited 14 to 28 days to see if insulitis and diabetes developed.
The results indicated that within 3 or 4 days after transfer, there is infiltraion of the islets by both Th-1 and Th-2 cells, with resulting peri-insulitis and insulitis. However, only the Th-1 recipients go on to develop diabetes. This suggests that, at some point in the future, it may be possible to regulate diabetes by forcing a Th-2 response, even when that cell carries the TCR for the as-yet-unknown antigen in the beta granules of the islets. Unfortunately, other experiments indicate that Th-2 does not provide this same protection when Th-1 is also present, even in small amounts.
Researchers are now conducting further experiments to test the hypothesis suggested above, namely that the real differentiation event is the loss of the beta chain of the IL-12 receptor. If it were possible to insert this receptor early on, it should be possible to generate a Th-0 cell that responds to IL-12 and makes IFN-gamma while still producing IL-4. It may be that the Th-2 cells produced in the mixing experiment have the IL-12 receptor and revert to Th-0 in an IL-12-producing environment, thereby preventing the desired suppression. Hence, researchers are trying to produce very heavily polarized Th-1 and Th-2 cells with no high-level expression of the receptor to see if this pattern will hold. Other possibilities for negative regulation of T-cell development include blocking the CD40 ligands or blocking the CD28 and B7 pathway, which may be important in the development of Th-1-type cells. Another is a soluble IL-12 receptor that might also switch the balance, even if the default is toward a Th-1 response.
In response to questions from the audience, Dr. Katz added the following:
It may be possible in the future to overcome genetic predispositions by choosing the correct immunogenic peptide, with correct affinity, to induce a strong Th-1 or Th-2 response. At present, however, researchers don’t know the antigens and peptides, and can’t answer questions about affinity.
The antigen in question copurifies with beta granule fraction by normal fractionation.
Researchers are conducting experiments to see whether they can activate a Th-2 population by immunization while blocking Th-1 cells, but they have no data as yet.
Knockouts and backcrossing with various mouse strains can be problematic, because many of the traits of interest are in the same cluster of genes.
It is now assumed that the effector cell in diabetes is the CD-4-positive ultra-T-cell.
Th-2 transfers also result in massive eosinophil infiltration. The Th-2 cells themselves persist for 5 to 7 weeks at the longest, and dye experiments to trace their fate show that, like a lot of transferred cells, they end up in the liver and get stuck there. It is unclear why; researchers are currently conduting more detailed histology.
ANTIGEN-INDUCED PROGRAMMED T-CELL DEATH AS A NEW APPROACH TO IMMUNE THERAPY18
Another new class of vaccines has an intended effect the opposite of all currently available vaccines—that is, to extinguish rather than activate an immune response—by means of eliminating the catalytic cell of immune responses—the T-cell—by using antigen to specifically program those T-cells to die. Such vaccines could play an important role in the treatment of disease in which T-cells play an important role in pathogenesis, including autoimmune diseases, graft rejection, allergies, and some others. Such vaccines would use the inherent specificity of the immune system itself to treat immunological diseases.
Researchers were prompted to propose this class of vaccines in response to the surprising observation that T-cells could be specifically programmed to die by antigen. More surprisingly, the agent that primed them for receptor-driven death was IL-2 T-cell growth factor: when cells that are cycling IL-2 are exposed to a peptide antigen on the antigen-presenting cells, a large fraction of them undergo programmed death. Almost any TCR ligand will have this effect; the original observation involve 2C11, an antibody against the CD3-epsilon chain of the TCR. That this was apoptosis was indicated by the fact that IL-2 or 2C11 alone did not disrupt genomic DNA, whereas IL-2 followed by 2C11 resulted in DNA fragmentation suggesting cleavage between nucleosomes, one of the hallmarks of programmed cell death.
The usual in vitro protocol for producing this effect involves (1) exposing the T-cells to conA, a nonspecific TCR ligand that up-regulates the high-affinity IL-2 receptor; (2) bathing the cells in various doses of IL-2; (3) restimulating the cells with various doses of antigen; and (4) recovering viable cells from the culture. The results are from a strain of mouse that is transgenic for a T-cell receptor that recognizes a peptide of myelin basic protein (MBP), by virtue of which these mice are susceptible to a disease called experimental allergic encephalomyelitis (EAE, see below). As expected, T-cells proliferate in
response to increasing doses of IL-2, but when restimulated by increasing doses of antigen there is a substantial decline in the number of viable cells. The exception is at low doses of IL-2, and even here proliferation drops off in response to higher doses of antigen, which induce endogenous IL-2 that supplements the existing level in the culture.
This finding—that the things that stimulate T-cells (i.e., antigen receptor occupancy and IL-2) can also program them to die—creates a paradox. After infection with a replicating pathogen, one might hope that T-cell response would increase along with the pathogen challenge. Researchers now believe that this is negative feedback regulation: the system appears to recognize that a strong immune response represents a potentially dangerous alteration in normal physiology, and therefore modulates the response in a very specific way that is tied to the particular antigen and the amount of stimulation received. There are, in fact, many examples of noncytopathic (parasitic) viruses in which the immune response is the detrimental component of the disease, rather than the replication of the microorganism itself. It is also known from clinical trials that several of the T-cell-derived cytokines are very toxic in the high doses that would result from an unregulated response to heavy antigen load.
Researchers tested this hypothesis by repeatedly challenging BALB/c mice with a superantigen, staphylococcal enterotoxin B (SEB), that is similar to toxic shock toxin. They administered large amounts of SEB to initiate immune response, followed by smaller doses on days 3 and 5 to maintain a large population of cycling cells, while also giving the animals regular doses of an antibody that would block the IL-2 receptor and thus prevent IL-2 responses. They sacrificed the animals at day 8, harvested the lymph nodes, and measured the populations of V-beta-8 T-cells (which specifically respond to SEB) and V-beta-6 T-cells (which do not). The results showed that the fraction of V-beta-8 cells dropped from 22 percent to 7.5 percent—a deletion of 60 to 70 percent of specifically responding T-cells through exposure to the very superantigen that caused them to undergo mitogenesis.
Evidence of Active Programmed T-Cell Death. This finding, which has been reproduced by other groups and with peptide antigens, fits a feedback regulation model that researchers call “propriocidal regulation,” a name borrowed from neural feedback regulatory loops. Activated T-cells both produce and respond to IL-2, and when cycling or proliferating T-cells are rechallenged by high doses of antigen a fraction of them will be shunted into an active programmed death pathway. This is not a mechanism to mop up T-cells at the end of an immune response, when the antigen is cleared—it occurs during the immune response and is a means of regulating that response when antigen is still present.
On the molecular level, active programmed death requires strong engagement of the T-cell receptor, which stimulates high-level production of death-inducing cytokines, including tumor necrosis factor (TNF) and Fas ligand. These cytokines engage specific receptors on the cell surface, engaging intracellular mechanisms that produce apoptotic proteins. Through several steps that have not yet been characterized, this leads to the activation of a cystine protease that is
related to IL-1-beta-converting enzyme (ICE) into an active tetramer that can cleave a number of molecules.
Genetic evidence for this pathway can be seen in mice that are homozygously deficient for the IL-2 receptor; such animals have a severe disregulation of peripheral T-cell homeostasis and are unable to eliminate mature T-cells as they are supposed to do. The SEB experiment described above demonstrated that animals lacking the IL-2 receptor are unable to delete V-beta-8 cells from their mature repertoire. Other experiments have shown that deficiencies in the TNF receptor will also prevent mature T-cell death. And the immunology community has known for some time that mice as well as children with mutations in the Fas molecule will develop lymphoproliferative autoimmune disease very early in life.
Controlled Use of Programmed T-Cell Death in Autoimmune Therapy. To determine whether the active programmed cell death pathway could be used in a controlled way, to eliminate T-cells that were causing T-cell disease, researchers used an animal model—experimental allergic encephalomyelitis (EAE) —in laboratory mice. EAE was first observed in patients who were vaccinated against rabies but developed a disease that looked a lot like multiple sclerosis (MS). It is a murine autoimmune disease in which CD4 cells react against various protein components of the myelin sheath, causing demyelinization and inflammation that leads to various neurological deficits, including paralysis and sensory defects. Certain forms of EAE are relapsing conditions very similar to human MS.
The protocol follows the paradigm of the SEB experiment described above to see if readministration of large doses of antigen at defined times would program a fraction of the activated T-cells for death and thereby prevent the autoimmune sequelae. The clinical results demonstrated that repeated doses of 400 micrograms of antigen, coadministered with 30,000 units of IL-2, can dramatically suppress the severity of the disease compared with untreated mice. In fact, two of the five animals in the experimental group showed no signs of disease and were indistinguishable from their litter mates.
Since pretreatment is unlikely in a clinical setting, researchers also delayed the administration of therapy for 9 days (until the first symptoms of disease) and then 17 days (until the first signs of chronic disease, namely the first wave of paralysis). Administering MBP alone at day 9 dramatically reduced the severity of disease, and even at day 17, when the disease was fully established, the treatment produced statistically significant improvement. Even after 40 days, following the second wave of paralysis, treatment produced statistically significant improvement in clinical scores, although it did not reverse the damage to the spinal cords. After several months, the treatment has proved to be very long-lasting, and most of the animals do not suffer relapses.
This does not seem to be a classic suppressive-type mechanism, however, since readministering encephalogenic T-cells causes re-exacerbation of the di-
sease. The data supports the idea that the treatment effect comes from eliminating the offending T-cells. Histology revealed that untreated animals show severe disruption of myelin architecture and inflammatory cell infiltrate in the subarachnoid space, while treated animals show little or no disruption of the normal architecture.
Researchers also wondered whether there is a fundamental immunoregulatory problem that might prohibit apoptosis and thus lead to autoimmune disease. The current state of knowledge may not be sufficient to rule this out. However, researchers have repeated the MBP experiment with T-cells from a number of human MS patients and found that they could indeed induce dramatic apoptosis in vitro. They are now initiating studies to look at other antigens.
Conclusions. Researchers now believe that there is a sound scientific basis for a new class of vaccines that will target T-cells for programmed cell death using specific antigen. They are encouraged by preclinical studies that show that this is effective and safe. They are moving on to study this question in nonhuman primates, establishing a marmoset model of EAE. They urge IOM to consider this type of vaccine in their priorities.
In response to questions from the audience, Dr. Leonardo added the following:
Activated T-cells may represent a reservoir of antigen; when they undergo apoptosis, they release the antigen, recruiting inflammatory cells and leading to relapses.
Evidence suggests that T-cells are in fact killed, not just switched off or pushed away from their target tissues (e.g., the brain).
Costimulatory molecules and interactions (e.g., B7-CD28, CTL-A4-IgE) seem to have little effect on the killing phase, or indeed any time after the immune response is initiated. Once the cell is primed, it appears that antigen itself is sufficient to kill it; in fact, a strong TCR ligand alone is enough to kill the cell.
It appears that different molecular parts of the TCR are important for activation and death. For example, the epsilon chain is capable of stimulating IL-2 production, but the zeta chain is particularly important for programming the cell to die through Fas or TNF. However, researchers do not yet know how to manipulate these differences.
This appears to be a high-affinity phenomenon, in which the most avid cells would be most quickly deleted, but while low-level stimulation does not eliminate, researchers have not looked into the exact affinity requirements. [Dr. Berzofsky volunteered that high doses of antigen with high epitope densities will selectively produce apoptosis in high-affinity CTLs, but will actually stimulate low-affinity CTLs; he and his coworkers have not looked at CD4 CTLs in this regard.]
INDUCTION, PROPAGATION, AND IMMUNOREGULATION OF AUTOIMMUNE DISEASES OF THE CENTRAL NERVOUS SYSTEM19
Possible Mechanisms of Pathogenesis of MS. There are two schools of thought about the etiology of MS. Most people believe that there is a loss of immune regulation, leading to an autoimmune response against neural antigens. This theory is modeled by the murine EAE model described above. However, many epidemiological studies suggest that there is an infectious or environmental component, possibly a virus, that triggers this response. The model for this theory is Tyler’s murine encephalomyelitis (TMEV).
If the putative virus is directly cytopathic to oligodendrocytes in the myelin, there would be a primary demyelination, but this doesn’t appear to be happening in either MS or TMEV. On the other hand, the virus might induce demyelination—directly or indirectly—as a result of the immune response against the putative infection. One possible mechanism for this process might be called “bystander demyelination”: provided the infection persists in the target organ, it can induce Th-1 responses and activate macrophages that somehow, through a process of “molecular mimicry,” begin to crossreact with the self-antigenic determinant, leading to a cascade of inflammatory events.
In an alternative mechanism that might be called “epitope spreading,” the virus causes the initial damage—either directly by targeting the oligodendrocytes, or indirectly by inducing a chronic inflammatory response in the target organ—and exposes self-antigenic proteins to the immune system, but the chronic stage of the disease is mediated at least in part by responses against the autoantigens that were released by the virus infection.
Murine Models of Virus-Induced MS. Researchers use different mouse models for the two forms of human MS. Both employ the SJL strain of mouse. Experimental allergic encephalomyelitis (EAE) can be induced with myelin basic protein (MBP) or proteolipid protein MOG (another myelin antigen), or by transferring T-cells specific for those epitopes; it develops into a relapsing-remitting form of autoimmune disease. In contrast, TMEV expresses a chronic progressive disease pattern that mimics the other, more serious form of human MS.
The progression of induction in these diseases at the Th-1 level is described below. A particular peptide is recognized in conjunction with MHC Class II and—provided the appropriate costimulatory signals are sent through the CD28 molecule—it induces the proliferation and activation of potential immunopathogenic T-cells. These primed T-cells leave the peripheral lymphoid tissue and penetrate the blood-brain barrier. When they again encounter the same epitope on APCs, they release chemokines and cytokines that activate and recruit mononu-
clear cells—microglial cells of the astrocytes, in the case of the central nervous system—which in turn produce the oxygen radicals, nitric oxide, etc., that destroy self-tissue.
Over the past 15 years, several groups of researchers have used immune tolerance as a tool for investigating the specificity of immune response in these two models of MS. This line of research began with the serendipitous observation that Th-1-type or delayed-type hypersensitivity responses could be inhibited by injecting naive animals with a population of MHC Class II-bearing APCs that had been incubated with a particular antigen or peptide in the presence of a crosslinking reagent called ethylene carbodiamide (ECDI). Further experiments revealed that this procedure altered the APC in such a way as to block the costimulatory signal 2, leading not to activation but to anergy and/or deletion of the antigen-specific T-cells.
Potential for Peptide Immunotherapy in Relapsing Autoimmune Disease. Building on these findings, researchers discovered that if it was possible to induce EAE by injecting mice with MBP-specific T-cells at day 0, it was also possible to effectively turn off that response by tolerizing the animals anywhere from 7 days prior to 5 days after injection—i.e., prior to initial clinical tissue damage—either with whole MBP (in a mouse spinal cord homogenate or MSCH) or with its immunodominant epitope (in the 84–104 region of the MBP molecule). If they delayed tolerance until after the acute phase of disease, however, it delayed the disease but the animals tolerized with 84–104 eventually relapsed at the same rate as controls, while the crude MSCH (containing not only MBP but every other potential neuroantigen) did a very effective job of inhibiting any further relapses.
The reason for these findings, researchers suspected, was that the initial phase of disease is directed primarily against the immunodominant epitope, but that tissue damage during the acute phase activates responses against endogenous epitopes, a process others have called epitope or determinant spreading. In a more defined experiment, they induced disease with the immunodominant epitope of PLP (amino acids 139–151) and then measured specificity of the responses that followed. The results show that, following the acute phase, there is a clear and consistent response against the secondary 178–191 epitope, a response that was not seen before the acute phase. Researchers also found that 139–151-specific T-cells could be reactivated with the noncrossreactive 178–191 epitope and would then transfer disease to naive animals, suggesting that—although 139–151 is the first self-response to arise—178–191 is primarily responsible for disease relapse. And indeed, tolerizing the animals with either 178 alone or 178 plus 139 protected the animals against further relapses, good evidence that the secondary epitopes have a major role in mediating relapses in this model.
This points to a major problem in using peptide vaccines and peptide therapies: the target keeps changing as time goes on. Other researchers have attempted to circumvent this problem by targeting the B7.1 and B7.2 molecules themselves, and hence inhibiting costimulation. They found that infusing anti-
B7.2 antibodies following the acute phase had no effect on relapses, either regulatory or enhancing, and that intact anti-B7.1 antibody actually exacerbated the disease. However, the Fab fragment of anti-B7.1 did protect animals from relapse, apparently by blockading the B7.1 molecule. Other studies have shown that the B7.1 molecule is dramatically up-regulated in the target organ during the preclinical stages of EAE, suggesting that it has a role in breaking down central nervous system (CNS) mononuclear cells into either F480 macrophages, B220 B-cells, or CD3 T-cells.
Potential for Peptide Immunotherapy in Progressive Autoimmune Disease. Researchers believe that tissue damage in TMEV is at least initiated because the virus is trophic for the CNS, where it lives in APC-like cells and can persist for long periods of time. Tolerizing animals with MSCH, as described above for MBP-induced EAE, has no effect at all on the development of TMEV-induced demyelinating disease. However, inducing tolerance against viral epiopes prior to infection does a very good job of shutting off the initiation of this disease. In fact, there are no neuroantigen-specific responses against MBP or PLP epitopes in the first 30 to 40 days after immunization, when the disease is already evident. By day 87, however, there is evidence of a response directed against the dominant 139–151 epitope of PLP, a response that first appears somewhere between days 42 and 52. And by day 164, there are responses not only against 139–151 but also against the secondary 178–191 epitope, the MOG epitope, and a third 56–70 epitope of PLP.
These findings strongly suggest that the tissue damage in the chronic, progressive TMEV model is initiated by virus-specific T-cells that target virus in the CNS and induce the initial inflammatory response. As tissue damage progresses, however, more and more self-epitopes seem to get recruited into this response. It is not yet clear whether these self-responses are playing a role in the chronic pathogenesis, or whether its course can be changed with peptide therapy; researchers are currently conducting reactivation, serial transfer, and tolerizing experiments—similar to those for EAE described above—to answer these questions.
For example, researchers are particularly interested in whether the initial self-response arises because of mimicry between 139–151 and some epitope in the virus, or because of the elaboration of myelin epitopes caused by the chronic inflammatory response. They identified three immunodominant epitopes of TMEV (lying on the VP1, VP2, and VP3 of the capsid) and developed T-cell clones and hybridomas specific for each. However, none of these hybridomas crossreacted with any of the neuroantigens or myelin epitopes that get recruited in the disease, nor do naive animals primed with neural epitopes develop any crossreactivity with the viral epitopes. This argues against molecular mimicry as an explanation of the disease, as does the fact that the response against self-antigens doesn’t arise until after tissue damage has occurred, whereas the
response to viral antigens arises within 2 or 3 weeks after infection, before the initial expression of disease.
Conclusions. T-cell responses in MS and other Th1-mediated autoimmune diseases appear to evolve dynamically during the course of both relapsing-remitting and chronic-progressive types of disease. Autoimmune reactivity in MS may in part be a secondary consequence of chronic CNS damage initiated by some putative persisting virus. Over the past 15 to 20 years, perhaps 15 or 20 different viruses have been associated with MS, but none has stood the test of time. One of the most recent is herpes virus type 6, which has been reported to be associated with oligodendrocytes in some cases of MS.
The dynamic nature of the T-cell repertoire has important implications for treatment modalities that employ antigen-specific strategies. Because the target changes as the disease progresses, for example, researchers hope to target the B7.1 molecule and costimulation generally, in order to inhibit disease progression without prior knowledge of the epitopes or T-cell receptors that are involved.
The model that emerges. The inducing epitope may be either a self-antigen or a viral antigen, provided the virus persists in the target organ. A Th1-type response leads to inflammation and tissue destruction that produces myelin debris, which stimulates T-cells expressed against endogenous myelin epitopes. Examples of such viruses in humans include Theiler’s virus, encephalomyocarditis virus (EMCV), and Coxsackie virus, all of which have been shown to persist in tissues and to be associated with autoimmune sequelae. With EMCV and Coxsackie the sequelae depends on the strain of virus, and the latter may be involved with human diabetes.
In response to questions from the audience, Dr. Miller added the following:
Researchers still know relatively little about what induces the remissions that follow the acute phase: a Th1-to-Th2 switch, antigen-induced programmed cell death, or some combination of processes. However, the persistence of some of the primed T-cells argues against propriocidal cell death as the sole explanation.
In Coxsackie B3-induced myocarditis, the inflammation produced by viral infection releases myosin, which is the antigen in the autoimmune phase. However, researchers have blocked the secondary autoimmune response by giving susceptible mice IL-1 receptor antagonist. They have also induced autoimmune myocarditis in normally nonsusceptible mice by administering IL-1. Hence it would appear that inflammation, and especially the cytokines induced by the viral infection, are critical in activating the true autoimmune process.
TMEV can be a lytic virus. It lives in some APCs, but primarily in F480 macrophages, where it undergoes defective replication, producing more viral antigen than infectious viral particles. The virus can persist in mice up to 18 months after infection. Because it lives in APCs, it may interfere with endoge-
nous IL-12 and IFN-gamma production to maintain a milieu that doesn’t lead to remission, but instead encourages inflammation to continue.
PEPTIDE-MEDIATED REGULATION OF AUTOIMMUNITY20
Unlike Jenner and Pasteur, who were trying to initiate a strong immune response in order to get rid of a pathogen, the problem in autoimmune diseases is to turn off a strong immune response against an autoantigen. And while there are many differences among autoimmune diseases, and a welter of potential autoantigens, it is nevertheless worth asking if there may not be some more generic form of treatment that would work for many or all of these conditions. It may be hard to envision a single antigen or a single T-cell receptor that could be used in a range of immunotherapies, but if there were a sufficient understanding of the mechanisms of the immune response, it might be possible to develop strategies that would deliver regulatory products—cytokines or other products—to the sites of inflammation.
Regulatory Peptides. Peptides or proteins delivered in nonaggregated form, in the absence of strong adjuvants, lead to a form of immune unresponsiveness that can be adapted for immunotherapeutic intervention. In at least three different animal models of autoimmune disease (EAE, diabetes, and type 2 collagen-induced arthritis), whether spontaneous or induced, it is possible to use peptides of major dominant autoantigens in a “tolerogenic” fashion to turn off the disease. This presentation focused on EAE, which had been discussed by several other speakers (see above).
The first step in developing vaccine strategies is to test the obvious sites of potential intervention in preclinical models, to ensure that the vaccine will not make the disease worse. This would be the worst possible outcome—to take a patient with a mild case of MS or arthritis and put them in a wheelchair. In the case of EAE, this intervention might come at two different stages in the development of the disease: (1) prior to the initiation of symptoms, and (2) during remission following the onset of disease.
Peptide Immunotherapy. In the first case, mice were given synthetic peptides of immunodominant determinants of myelin protein in a tolerizing regime to block the initiation of symptoms. Earlier studies had shown that mice tolerized prior to time zero with peptides of myelin sheath do not develop disease in response to MBP. More recent work has shown that a “cocktail” of peptides is more effective than either the major immunodominant (AC 1–11) or secondary immunodominant (AC35–47). Even AC 1–11 is a rather weak antigen, however, with very poor MHC binding. Researchers have determined that peptide substitution—using a tyrosine at the fourth position of the acetylated 11
N-terminal amino acids of MBP—produces a more efficient MHC binder, enhanced immune phenomena, and a more efficient immunotherapeutic.
More importantly, substituted peptides also worked at a later and more clinically relevant stage of the disease, i.e., during remission after the onset of clinical symptoms. When researchers took mice that had progressed to stage 2 of EAE and tolerized them with the 4-tyrosine-substituted peptide, the mice did not get worse; in fact, they got better and stayed better for a long time. The peptide did not cause a relapse, and seemed to block a relapse—in short, it appeared to induce remission in clinically ill animals. Obviously, this result suggests a more realistic and clinically relevant way to administer the vaccine, namely when patients present with clinical disease. This model would be particularly important in diseases marked by a similar exacerbation-and-remission cycle, including rheumatoid arthritis, systemic lupus erythematosus, and MS.
To address concerns about the “innocent bystander” effects that this approach might cause, researchers tolerized mice with a single determinant of MBP and then induced them with spinal cord homogenate containing the entire gamut of proteins from the myelin sheath. The mice were protected. And in a related experiment, separate T-cell clones recognizing determinants A and B of myelin were transferred to naive mice, but when they were then tolerized with a single determinant, it turned off both clones. Something other than the apoptosis or anergy of that single T-cell clone was occurring. Whatever the mechanism, however, the results were dramatic: sick mice looked and acted normal after 24 hours, and their large inflammatory lesions had disappeared.
Researchers speculate that the peptide was not just killing cells but also turning off the inflammatory milieu through some unknown counter-regulatory mechanism. This has important implications for vaccine development strategies: instead of identifying the autoantigens, and hence the regulatory peptide, it might be possible to understand the counterinflammatory mechanisms and thereby bypass peptide immunotherapy entirely. One candidate involves not only the recognition of the antigen but also the regulation of T-cells themselves. Research currently underway suggests that this T-cell immunoregulatory circuit is very active in the exacerbating-remitting characteristics of EAE and may have a role to play in immunotherapy.
The idea would be to enhance this regulatory role, possibly by using peptides of T-cell receptors, but at present researchers do not know which T-cell receptors to pursue. In the clonal transfer experiment cited above, peptide A administered when disease has progressed to stage 2 will turn the disease off; but if anti-IL-4 is administered at the same time, it completely blocks the therapeutic effect of the peptide. This suggests, but by no means proves, that regulatory cytokines might be a central focus in efforts to bypass peptide vaccination immunotherapy and treat autoimmune diseases more directly.
T-Cell Hybridoma Cytokine Regulation. Researchers have investigated this question in a line of retroviral gene products that make it possible to transduce or infect cells with selected genes. For example, researchers “tag” retroviruses with green fluorescent protein to tell which cells are infected, or
insert LAK-Z to target the retroviruses. In one study, researchers used this technique to produce a T-cell hybridoma that was prototypically Th1-like and had a receptor for MBP (i.e., was encephalogenic or at least had the capacity to traffic to the CNS) and then transduced it with gene sets for the expression of either IL-4, IL-10, or LAK-Z (as a control). When researchers administered transduced hybridoma T-cells to mice that had been immunized with MBP but had not yet developed EAE, the targeted IL-4 hybridoma appeared to delay the onset and decrease the incidence of disease. In a second experiment, the hybridoma was given a receptor for myoglobin, and hence targeted muscle rather than nerve tissue; in this case the systemic application of IL-4 was not protective, while the targeted CNS-localizing hybridoma was protective.
These results again suggest, but by no means prove, that immunodominant peptides may have a role in inducing immune unresponsiveness. If that role is generic, it might make possible therapeutic strategies that are relevant to multiple autoimmune disorders. An earlier presentation pointed out that the expression of IL-4 under the right insulin promoter seemed to block the development of NOD diabetes. There is some indication in the literature that IL-4 may also be a down-regulatory product in rheumatoid arthritis.
Retrovirus Transduction. Retrovirus transduction may provide a tool for targeted specificity, restoring regulatory cytokines to sites of inflammation. Other studies have suggested different strategies for targeting retrovirus, specifically to activated T-cells and lesions of autoimmunity, where they restore expression of the cytokine of choice. In the future—when funding becomes available—researchers would like to pursue this approach using a retrovirus that is targeted to OX40, a TNF-family receptor that is expressed on the surface of activated T-cells in lesions of EAE. OX40 is also expressed in rheumatoid arthritis and possibly in diabetes as well, although researchers have not yet looked for it in NOD mice. The plan is to add to the retroviral coat a chimeric protein that includes not only the usual gp120 but also some structure (e.g., a single-chain variable region of an antibody) targeting the OX40 ligand of the OX40 receptor. This would target the retrovirus on activated T-cells in sites of inflammation, where the expression of regulatory cytokines would turn off the inflammation. If it proves feasible, this approach promises to provide an alternative mechanism for the delivery of regulatory products that bypasses the vaccine approach altogether.
In response to questions from the audience, Dr. Fathman added the following:
Evidence from some studies, including those using IL-4 knockout mice, suggests that there may be multiple compensatory mechanisms in the immune response, mechanisms that could regulate inflammation in the absence of IL-4. At present, however, the significant finding is that IL-4 has the capacity to turn
off the inflammatory events in three separate preclinical animal models of autoimmune disease.
The alternative strategy of transferring huge numbers of fully activated Th1-type T-cells may simply overwhelm the immune system, when the goal is simply to restore homeostasis. More importantly, very few of the transferred T-cells would find their way to the CNS joints or beta cells. The transgene approach doesn’t allow for the restoration of homeostasis; for that, one needs immunoregulation.
An alternative approach is to develop preventive immunization that is harmless enough to use in genetically predisposed individuals. This approach has been used in thyroiditis, where the etiologic autoantigen is well known and studied. There is a long list of potential autoantigens in other diseases, however, and much remains to be learned about them.
Human studies have been limited. Chimpanzees with MS responded well to an extract of myelin sheath containing MBP and other peptides. Human trials of MBP were conducted in the 1970s, but the subcutaneous route proved to be disastrously wrong, and several patients died of fulminating encephalomyelitis. There will be Phase I clinical trials this fall using immunodominant determinants of myelin protein, as described above.
Researchers do not know if the retrovirus will persist after homeostasis is restored.
There have been anecdotal reports of patients whose autoimmune diseases were cured when they received autologous bone marrow transplants for other reasons, such as cancer therapy, but this approach is far too radical to use pre-ventively—for example, in prediabetic children that aren’t yet sick. An alternative that may be much simpler is stem cell rescue: if the patient becomes heterozygous, this protects against diabetes, although it may allow other diseases to be induced. This approach has been successful in animal models, but humans will have to wait until the next century—at least 25 years—for well-thought-out strategies involving bone marrow transplantation or stem cell reconstitution to (1) induce chimerism and heterozygosity or (2) induce homozygosity for nonpermissive alleles.
Researchers got protection with IL-10, as well, and TGF-beta may be a good regulatory cytokine as well. They pursued IL-4 initially because they got the best results with it in early tests. Researchers will eventually look at a wider range of regulatory and counterregulatory inflammatory cytokines, alone and in combinations.
The genetics of human diabetes is similar to that of the NOD mouse. MHC typing reveals five genomic intervals related to the disease, none of which has been targeted. In NOD mice, there is a clear B-chain epitope on the MHC that, when used to immunize naive animals, does not cause inflammation. If this epitope is sufficiently conserved to be relevant to human disease, it will not cause harm there either. However, the targets include antigen processing and presentation, as well as MHC. Researchers know nothing about processing at present.
STIMULATION AND COSTIMULATION21
Full activation of a T-cell requires two different signals. The first, antigen-specific signal, comes when the T-cell receptor recognizes a peptide-MHC complex on the surface of an antigen-presenting cell (APC). This first signal causes events and reactions that can be quantified, such as the expression of the growth factor receptor or IL-2 receptor or the release of certain cytokine molecules. However, full activation of the T-cell requires a second, costimulatory signal involving a different ligand and receptor. Indeed, the antigen-specific signal by itself is a negative regulator of the system and can be used to induce tolerance, as shown in several of the preceding presentations.
At a more complex level, the costimulatory signal appears to be particularly important in activating the CD4 helper cells, which—in either the Th-1 or Th-2 form—plays a critical role in activating other cells of the immune system. And while the signal it transits to the B-lymphocyte may be antigen-nonspecific (e.g., cyryokines such as IL-4), the activation of the helper T-cell is antigen-specific. In this sense, both of the signals received by the B-cell are antigen-specific.
Molecular Signaling in Costimulation. A range of ligands and receptors have been shown to play various roles in costimulatory signals. Perhaps the best-known are the B7 family, which are expressed on the APC in two well-characterized forms, B7.1 and B7.2 (a third form has been proposed). These molecules interact with the CD28 and/or CTLA4 molecules, which are normally expressed on the surface of the responding T-cell. When B7 molecules engage CD28, they turn the T-cell on; when they engage CTLA4, they turn the cell off, thereby providing a feedback loop.
In their resting state, the APC expresses a low level of B7 on its surface, and the T-cell expresses a large amount of the CD28 receptor, and a low but detectable amount of CTLA4. The CTLA4 acts as a buffer, preventing nonspecific activation, because it has about a tenfold-higher affinity for the B7 ligands and, under the limiting conditions of B7, it preferentially occupies these receptors and gives negative signals to the cell. Activation up-regulates B7 on the surface of the APC (first B7.2, later B7.1), rapidly saturating the few CTLA4 receptors on the T-cell and then occupying CD28 molecules, sending positive signals and initiating costimulation. Between 24 and 48 hours later, there is a gradual shift from B7.2 to B7.1 on the APC, inducing an increase in CTLA4 molecules that compete for B7 molecules, send increasing negative signals, and turn off the T-cell—or at least this pathway.
Molecular Response to Costimulation. When CD28 is stimulated in vitro, the predominant effect in tissue culture is a 30- to 100-fold increase in the production of the T-cell growth hormone IL-2, relative to the amount produced
in the absence of costimulation. Two different mechanisms have been proposed for this dramatic effect:
Increased transactivation of the IL-2 gene. The initial signal transduction events are not clear, but it appears that the CD28 cytoplasmic tail will bind to PI-3 kinase after tyrosine phosphorylation. There is also one controversial report of acidic sphingomyelinase activity, leading (probably through several steps) to the activation of Jun N-terminal kinase.
Stability of the IL-2 message. This mechanism for augmenting IL-2 production has been well studied in the mouse and human systems. Early studies postulated that sequences in the 3-prime-1 translated region were important for message stability, but other studies are underway to identify additional critical factors.
Message stability has an effect on the duration as well as the intensity of the response. When the T-cell was stimulated with signal 1 alone (e.g., anti-TCR), the increase in IL-2 message peaks at 4 hours and then falls off fairly rapidly. When the T-cell also received signal 2 (e.g., anti-CD28), the IL-2 message was more prolonged, and there was about a 35-fold difference in the amount of IL-2 produced. Because of this, some researchers believe that message stability rather than transactivation is the dominant effect.
Studies with Knockout Mice. Researchers have also studied costimulation in vivo using CD28 and CTLA4 knockout mice. When they knocked out the CTLA4 gene, which provides the negative feedback loop, they observed massive lymphoproliferation and death at about 3 weeks as lymphocytes infiltrate and destroy multiple organs. These were the effects that might be expected in the absence of a negative feedback. There was some indication that it was actually cardiac problems that killed the mice.
When researchers knocked out the gene for CD28, however, the data were more ambiguous. IgG antibody responses were impaired, but there were normal cytotoxic responses to L, C, and V viruses. Subsequent studies showed impaired cytotoxic responses to VSV viruses, and a measurable but minor impairment of CD4 proliferative responses. These differences were not as dramatic as would be expected, were CD28 the key molecule in costimulation. This has led to further studies in search of the “missing component.”
Other Costimulatory Molecules. Given that the CD28-B7 mechanism does not appear to be the whole story, researchers’ attention will probably turn to additional biochemical mechanisms for costimulation. Among the ligands and receptors listed in Table 3 (above), the cell adhesion molecules such as LFA-1 and ICAMs are important because they allow the cells to come together, and adding antibodies to LFA-1 will block the initiation of T-cell response. The LFA-3-CD2 combination also appears to augment TCR signalling. Other outliers have been known for years, including heat-stable antigen (HSA) and the invariant chain (Ii-c.s.).
Several groups are currently studying the TNF super-family (IL-1, etc.), which includes several members that can send both positive and negative signals to cells. IL-1 was the first costimulatory molecule to be defined, and there is good evidence that it participates in initiating events in the immune response. The best-characterized of the family is the CD40-ligand interaction—antibodies that block this interaction will also impair immune responses, and researchers are now waiting for the combination studies and knockout studies that will demonstrate whether this is the “missing component.” IL-12 helps initiate Th-1 responses, and GM-CSF helps initiate macrophages. Even chemokines have been reported to give costimulation. In short, there are entire families of molecules that could potentially be manipulated in a vaccine strategy.
Cytokine-Induced Apoptosis and Anergy. TNF and fas can also induce apoptosis. But activation in the absence of costimulation sends a negative signal to the T-cell, driving it into a nonresponsive state. The T-cell does not die, although it does function less well than it did before it was given signal 1 alone, and the term “anergy” was borrowed from B-cell research to describe this state. If the anergic T-cell is restimulated by a normal APC with full costimulation, it generally fails to divide and proliferate, primarily because it fails to produce IL-2, the T-cell growth factor. Production of some additional cytokines such as IL-3 and GM-CSF are down by intermediate amounts, but others such as IFN-gamma don’t seem to be affected.
Only when researchers developed a mouse that produced Th-0 cells, which produce both IL-4 and IFN-gamma, were they able to investigate these effects. They found that anergy didn’t significantly decrease the production of either IL-4 or IFN-gamma, but at the same time the cell was blocked from proliferating in response to IL-4. IL-4 can be a growth factor in the same way as IL-2, but in this case the lack of response wasn’t related to shutting off production.
Researchers concluded from this that the cell has a special kind of regulation that stops proliferation—in short, T-cell anergy (at least in vitro) is really a state of growth arrest, possibly involved through a differentiation process. If it occurs very early, however—i.e., after the cell down-regulates IL-2, but before, has had a chance to turn on its IFN-gamma and IL-4 genes—it can also be a mechanism of tolerance, because it prevents cells from expanding and differentiating.
Molecular Basis for Cytokine-Induced Anergy. Researchers are gaining a clearer understanding of the mechanisms that block signal transduction and prevent IL-2 production (for example) in anergized cells. Early studies demonstrated that the intracellular calcium pathway was completely intact. Other studies showed that transactivation through AP-1 was inhibited, and most recently that the problem is in the activation of ras. In particular, they showed that the She, Grb2, and SOS activations were normal, but that downstream events from ras such as RAF and the ERK kinases can activate the Jun kinases and thereby block AP-1. There are a variety of other ways to regulate ras, including
exchange factors like SOS and VAB as well as GTG-ase activity and another pathway activated by protein kinase C (PKC). As researchers zero in on these molecular mechanisms, they will eventually have a full understanding of this process.
At bottom, then, anergy can be characterized as negative regulation induced by T-cell receptor occupancy in the absence of costimulation. Signal 1 alone actually induces an inhibitor that operates on the ras pathway, blocking ras activation. Because this happens at the same time that the signals activating the IL-2 gene, the initial response is to produce IL-2. But the inhibitor persists after the initial response dies down, and when the cell is restimulated it behaves like a negative feedback loop, preventing subsequent activation.
When signals 1 and 2 are both received, on the other hand, there is a far greater production of IL-2, followed by proliferation of the cell. At first researchers thought that cell division alone was enough to dilute out the inhibitor; it now appears that signal transduction through the IL-2 receptor, particularly the gamma chain, can antagonize the induction of anergy, presumably by inhibiting the production of the inhibitor. The biochemical basis of this mechanism has not yet been worked out. In this case, it would appear that costimulation antagonizes the anergic effect in two ways: (1) by augmenting IL-2 production, and (2) by inhibiting inhibitor production.
Finally, recent studies have demonstrated that, even in the presence of normal costimulation, it is possible to modify the peptides in certain MHC complexes by making certain amino acid substitutions that probably decrease the affinity of interaction with the TCR. These so-called “partial agonists” are also capable of inducing anergy, apparently by interfering with the transduction of signal 1 and preventing the downstream event, namely the production of IL-2. This result demonstrates that it is possible to achieve anergy through two different biochemical mechanisms—lack of costimulation, and partial signal transduction.
T-Cell Receptors and Apoptosis. Programmed cell death was described above. However, the resting T-cell has on its surface the CD95 molecule, which is fas. When the cell is activated through antigen stimulation and costimulation, one result is the production of additional fas ligands on the surface. These ligands form a trimeric complex that can interact with the receptor on the same cell, or those on other cells. Researchers hypothesize that this interaction may be what signals the cell to undergo apoptosis. This would explain why interfering with the CD28 system had no effect on apoptosis.
Another team of researchers investigated the roles of antigen-specific and costimulatory receptors by looking at T-cells stimulated with anti-CD3 and measuring the effect of anti-CD28, as reflected in the production of Bcl-2 family members, which are important in protecting the cell against apoptosis. They found that the combination of the two signals resulted in very high levels of Bcl-x. In this case, the antagonist to CD28 was solubilized CTLA4 receptor (with its greater affinity for B7) that has been fused with immunoglobulin for stability. In
vitro, this fusion protein (CTLA4-Ig) blocked the costimulatory pathway, Bcl-x production was negligible, and the cells died.
Immunotherapeutic Applications. The NZD mouse is an animal model for lupus. When the animal is treated with the solubilized CTLA4 receptor, which blocks costimulation through CD28-B7 interactions, autoantibody production is reduced and the life of the animal is prolonged. CTLA4-Ig had an ameliorative effect even when given late in the disease. In young NOD mice, on the other hand, CTLA4-Ig reduced the incidence of diabetes but not insulitis, and it had no effect in mice over 10 weeks of age. These partial effects may be related to the fact that CD28-B7 is not the only costimulatory pathway, but further research will be needed.
In tumor immunity, the APC must activate cytotoxic T-cells in order to eliminate tumor cells. Normally, this is done by stimulating T-helper cells, using both signals and producing IL-2 to encourage T-cell proliferation (as in A). In some cases, a precytotoxic T-cell can respond to antigen in the absence of costimulation, but this leads to anergy or apoptosis through the mechanisms described above (as in B). However, a number of studies have shown that when various costimulatory ligands are introduced into the tumor (e.g., B7, as in C), it is possible to enhance CD8 cytotoxicity in the absence of CD4 help, and even in the absence of costimulation by the tumor itself, so long as the peptide ligand is recognized. This strategy has worked in limited situations.
Another group has followed a slightly different track. Because CD28 turns on the costimulatory signal, but also leads to a negative feedback loop when the CTLA4 molecule is stimulated, they developed a monoclonal antibody against CTLA4 (as opposed to CTLA4-Ig, which is a soluble form of the molecule itself). By preventing the negative feedback signal, once the T-cell response has been activated naturally, this strategy results in a tremendous augmentation of T-cell cytotoxic responses that can eliminate certain tumors that wouldn’t be eliminated under other conditions.
In response to questions from the audience, Dr. Schwartz added the following:
IL-2 drives the T-cells into S phase. However, the immune response is not the physiological effect of IL-2, but rather what IL-2 does to the biology of the cell. It’s nothing special, just cell cycling.
Chemokines are part of a nonadaptive immune response in which their major role is to call in T-cells. Neutrophils in particular are the second line of defense, after the skin and mucosal tissue, and the earliest kind of hematopoietic cells to respond to inflammation, trying to destroy whatever organism or invader it finds. Chemokines are released both by the neutrophils and by local tissue. In a generalized sense, they too represent a kind of costimulation, and some investigators are beginning to think that the entire inflammatory process should be considered in up-regulating specific molecules to talk to T-lymphocytes.
VIRAL THERAPEUTIC VACCINES—HEPATITIS B22
Peptide-Mediated Vaccines. Researchers sought to develop therapeutic vaccines to treat tumors and chronic viral infections, focusing initially on chronic hepatitis B virus (HBV) infection. In such an infection, large amounts of antigen are present in host tissues but are presented by inappropriate cells. Their approach was to introduce antigen in a more efficacious form by using peptides as the source of antigen.
One advantage of this peptide-mediated approach is the ability to target the type of immunity by selecting peptides with either Class I- or Class II-restricted epitopes. It also avoids the confounding effects of producing a lot of antibody, which might have negative effects on cellular immunity. More importantly, it offers the possibility of using peptides that are conserved across various viral isolates, especially in RNA viruses such as hepatitis C and HIV, in which epitope drift is considerable. Finally, it should be possible to select epitopes that might not be tolerized in the host.
The disadvantage of peptides is that they are MHC-restricted, and for this reason a single peptide can only be expected to mount immune responses in a limited number of individuals in the population. As a result, there is a quantitative problem of how many epitopes are required to mount an effective immune response.
The strategy adopted by the researchers—identifying the epitopes that are most capable of inducing CTL responses—required them first to identify the binding motifs responsible for peptide-MHC interaction, then to investigate their immunogenicity, both in vitro and in vivo.
Peptide-MHC Binding Motifs. Researchers began with the most common alleles of human leukocyte antigen group A (HLA-A) and studied their MHC binding motifs. They discovered that some alleles have very similar motifs, such as A3 and A11, which favor hydrophobic residues in position 2 and lysine at the C terminal position. When they examined the binding activity of different peptides, roughly half of the peptides they tested bound significantly to both A3 and A11.
This led to the concept of “super-motifs,” which has been further investigated by several groups. The results show that a combination of peptides that share three of these super-motifs will provide coverage for a very large percentage of all human populations. In this case, the A2 proteins are hydrophobics in the B and F pockets; A3 is hydrophobic and basic; and B7 binds almost any peptide in the F pocket. This finding greatly reduces the number of peptides that must be isolated in order to immunize an outbred population.
The binding motif does not fully define the binding affinity of the peptide, however. Indeed, researchers have observed 10,000-fold differences in binding affinities among peptides that are identical at their anchor positions but differ at
other positions. Researchers analyzed the importance of each nonanchor residue by comparing a very large set of peptides. They found that charged residues (positive or negative) are deleterious to binding, whereas aromatic residues are generally favorable. This allows researchers to predict with high efficiency the binding capacity of peptides that bear a given motif. In general, a peptide bearing one of the favored residues has about a 25-percent chance of binding, but when it contains one favored residue and none of the deleterious residues the chances of high-affinity binding increase to approximately 80 percent.
Binding Affinity and Immunogenicity. Based on the literature, about 90 percent of normal T-cell epitopes bind to the restriction element with high affinity, defined as a KD of 50 nanomolar. Peptides that are naturally processed by MHC also tend to fall into this high-affinity category. High-affinity peptides from transgenic mice also tended to be preferred immunogens. When researchers repeated this analysis for known epitopes of viral and tumor antigens, however, they found that 90 percent of viral epitopes are high-affinity binders for MHC, while less than half of tumor epitopes are high-affinity binders. Because tolerance tends to favor high-affinity, immunodominant peptides, a possible vaccine in situations of significant tolerance would be to shift to subdominant, lower-affinity peptides as potential immunogens.
In order to vaccinate with these peptides, researchers decided to include a “helper epitope” along with the classical CTL epitope. To this they added palmitic acids, which according to the literature enhance the immunogenicity of peptides. The data showed that the lipidated helper-CTL epitopes was a very efficient immunogen, and that it is equally efficient to a similar construct without the lipid but using IFA as an adjuvant.
Peptide-Mediated HBV Vaccine. Based on these findings, researchers proceeded to construct a CTL vaccine for hepatitis B. The rationale for such a vaccine was that CTL responsiveness was known to be associated with (1) clearance of acute HBV infection, (2) spontaneous clearance of chronic HBV infection, and (3) successful IFN-alpha therapy.
The vaccine used in preliminary studies consisted of an immunodominant epitope derived from the HBV core (18–27), combined with a helper epitope from tetanus toxin (830–843, which is relatively MNH-unrestricted), plus lipids. Phase I clinical data showed a clear immune response (measured in CTL activity) in normal subjects at a dosage of 500 micrograms of antigen, administered with a single booster shot. Higher and lower doses produced corresponding responses.
Preliminary Phase II clinical data show similar responses in about one dozen patients with chronic HBV infection who were given two injections of lipidated peptide as antigen. The “ALT flare” refers to liver damage with release of transaminase enzymes, which would be expected to follow the induction of CTL and is in fact an important indicator of CTL response. A similar flare can be observed in patients treated with IFN-alpha who clear the virus. DNA levels
were also measured, often decreasing by anywhere from 50 percent to complete clearance. These results are very early, but they do show that—even in the face of chronic antigen load and viral infection—introducing the antigen in a suitably immunogenic form can lead to an immune response to the virus.
While the tetanus toxin (TT) peptide in this vaccine is relatively MHC-unrestricted, there are nevertheless several MHC haplotypes to which it will not bind, in which cases it is not immunogenic. Examples include DR4 and some of the DR2 splits. Researchers had already developed several peptides that were known to bind to most DR alleles, but when they tried to immunize peripheral blood in vitro with a “pan-DR” peptide, they got very little response compared with TT. The reason had to do with the topography of this particular peptide: the MHC contact residues were on a polyalanine background that gave the TCR very little to recognize except the short methyl groups on the side chain of alanine. Researchers substituted lysine and tryptophan for alanine in three positions, giving the TCR more interesting side chains to recognize, and the resulting peptide induces a better proliferative response in vitro than did TT.
Like TT, pan-DR represents a helper epitope that may be very useful in conjunction with CTL epitopes to augment the immune response. It also shows considerable usefulness in helping antibody responses, and it might be an interesting peptide to add to some of the prophylactic vaccines that are currently used to generate carbohydrate antigen antibodies.
In response to questions from the audience, Dr. Grey added the following:
Peripheral blood is admittedly a poor compartment in which to observe CTL response, and these in vitro results may not be predictive.
Normal subjects in the Phase I study were subsequently given traditional surface antigen vaccine and responded normally—decreasing EA, DMA, and surface antigen levels, indicating that the vaccine is in fact affecting viral expression.
Once they are linked to helper epitopes, the peptides no longer bind directly to the MHC Class I molecules on the APCs. However, they are being processed somewhere—intra- or extracellularly—and the linkage does not destroy their ability to prime CD8 cells. Perhaps phagocytic cells are taking up these lipopeptide adjuvants.
The original DR peptides were originally developed as MHC blockers in autoimmune disease. They worked well in vitro, but in vivo they have an extremely short half-life and quickly fall below the concentration needed to demonstrate MHC blockade.
HIV peptide vaccine studies have shown that the CTL epitope and helper or adjuvant must be covalently linked—a mixture doesn’t work. Processing appears to extracellular, and the small size of the peptides may allow them to bypass the Class I processing pathway.
These short peptides pulsed onto dendritic cells have effective properties in vitro, which Dr. Berzofsky will address in the following presentation (see below). Briefly, part of the function of CD4 help in inducing CD8 CTLs is in
regulating stimulated molecules and APCs. If the peptides are presented on dendritic cells, it might be possible to induce CTLs in a CD4-independent mechanism.
There is concern that inducing CTL may increase the severity of cell loss, but studies in HBV-transgenic mice indicate that passive transfer of large amounts of CTL clones does not lead to massive liver damage. There might be a patchy necrosis, in addition to a cytokine-mediated decrease in viral DNA due to IFN-gamma and TNF, but CTL does not appear to kill liver cells, even when 100 percent of those cells express antigen.
CD8 CTL TO MUTATED ONCOPROTEINS AND FUSION PROTEINS23
Mutant Proteins and Peptides as Target Antigens. Whatever the inciting carcinogenic event, most types of cancer involve either the inactivation of tumor suppressor genes or the activation of oncogenes. Often, but not always, that involves point mutations or translocations that potentially create neoantigenic determinants that might serve as tumor antigens and thereby serve as the targets of vaccines. Since these mutations or translocations would occur only in the tumor cell, they could provide unique markers that distinguish the tumor cell from normal host tissues to make a specific vaccine.
The problem with this approach, from the point of view of conventional tumor vaccine work, is that most tumor antigens have been described using antibodies, which can only recognize proteins on the surface of tumor cells. The products of these oncogenes and tumor suppressor genes are generally intracellular proteins, often nuclear proteins, that are not expressed on the surface of the cell. However, this limitation does not apply to CD8 cytotoxic T-lymphocytes (CTLs), which is able to “see” any protein synthesized in the cell. This is because those proteins, or some subset of them, are degraded into peptide fragments that are actively transported into the endoplasmic reticulum, where they bind to newly formed Class I MHC molecules and are transported to the cell surface, at which point the peptide fragments can be recognized by CD8 CTLs.
In this way, nucleo- and cytoplasmic proteins can operate as tumor antigens for CTLs. Among the many mutant proteins, researchers have thus far focused on p53 and ras, both of which occur in many of the most common types of cancer. This presentation focuses on p53; a later presentation focuses on ras (see below).
Mutant p53 Tumor Vaccine. The strategy followed was to make a short synthetic peptide spanning the site of a point mutation, and then immunize with
the peptide to raise CD8 CTLs against the peptide. This approach was based on earlier work with HIV peptides in mice, using peptide-pulsed spleen cells; however, researchers found that they could also use peripheral blood mononuclear cells (PBMCs). Both contain about 1 or 2 percent dendritic cells. Peptide can bind to the MHC Class I molecules on these dendritic cells without going through the intracellular processing pathway. When the cells are washed, irradiated, and reinjected intravenously, they induced CD8 CTLs and specific lysis.
The key cell in this process is the dendritic cell. Researchers found that immunizing with 100,000 purified dendritic cells produced as much or more specific lysis as 8 million peptide-pulsed spleen cells. This is consistent with the ratio of one or two dendritic cells per 100 spleen cells.
Previous experiments had shown that HIV peptide 18, part of the V3 loop of the envelope protein, was a major target for CTLs in BALB/c mice. Other researchers had successfully demonstrated that this same peptide-pulsed spleen cell immunization technique could protect against a mouse tumor that was expressing HIV gp160 (made tumorigenic by cotransfecting with ras). Based on this, investigators decided to use the same approach with peptides corresponding to mutations of p53, specifically a mutation called T1272, that came from a human lung carcinoma.
Researchers immunized BALB/c mice with peptides corresponding to this mutation and were able to induce CTLs that would kill targets. Using various truncated peptides, they mapped the minimal epitope to a tenmer that contained the point mutation (a cystine to tyrosine). Interestingly, this epitope avoids the differences between human p53 and mouse p53, and the induced CTLs are specific only for the mutation difference—they do not kill controls or targets pulsed with wild-type peptide.
In other words, the cystine-to-tyrosine mutation created a new antigenic determinant that was not present in natural p53. This peptide is presented by KFD, whose binding motif involves a tyrosine at position 2. Since natural p53 doesn’t bind to KFD, and isn’t recognized by the CTLs, there should be no danger of inducing autoimmune disease if the subject were a human with cancer.
The key question was whether endogenously expressed mutant p53 would be processed and presented in such a way, and in sufficient quantity, to be killed by the induced CTLs. Researchers transfected mouse tumor cells with mutant p53 and found that they were killed, whereas untransfected cells were killed only if peptide was added to the culture. In addition, the level of mutant p53 expressed by the transfected cells was at the low end of what is found in natural human and murine tumors. This indicated that the tumor antigen was not produced by overexpression of p53 and would in fact work at the levels found in natural tumors.
Researchers concluded from these findings that endogenously expressed mutant tumor suppressor oncogene product p53 can serve as a target antigen for CD8 CTLs, and that such CTLs can be induced by peptide immunization. They chose this method of immunization in part to avoid the need to attach a helper
epitope (see presentation by Ronald Schwartz, above). Adjuvants containing helper epitopes can be very important, but when peptides are pulsed into dendritic cells, helper epitopes seem to be less critical. Consequently, researchers would be able to immunize numerous patients, each with a different mutation of p53 requiring a unique peptide—challenge enough—without also having to attach a helper epitope to each unique peptide.
Tumor Immunotherapeutic Experiments. To test the ability of the mutant peptide vaccine to treat a mouse with an established tumor, researchers injected tumor cells, waited 8 days until they could see or palpate tumor nodules that were 2 to 4 millimeters in diameter, and then immunized with peptidepulsed dendritic cells either (1) a single time or (2) repeatedly every 4 or 5 days. The single immunization did not change the rate of growth of established tumors, but multiple immunizations significantly inhibited tumor growth. This protection lasts as long as immunizations continued; when immunization stopped, the tumors eventually began growing again. The animals were not cured, but the tumors were suppressed as long as researchers kept boosting their immunity.
Based on these results, researchers have started a clinical trial with human subjects. They performed a biopsy of the patient’s tumor to determine if there was a mutation of p53 or ras, synthesized the corresponding peptide, pulsed it onto autologous PBMCs, and then reinfused the cells into the patient to immunize. This trial is still at an early stage, and it would be premature to report any results, but in most cases the patients had bulky tumors and extensive prior treatment with chemotherapy, which has left them with very poor immune systems (e.g., many are unable to make a CTL response to flu). However, researchers have shown the safety of this approach, and a few patients with less severe disease they have seen hints of either cytokine or CTL responses. This clinical trial is now moving into a new stage involving patients with less tumor bulk, or no tumor bulk, in whom investigators expect to see a better response because the immune system is more intact. This new stage may also immunize with larger numbers of cells, or with purified dendritic cells.
Fusion Proteins as Target Antigens. A similar approach has been tried in a different disease system found in Ewing’s sarcoma and alveolar rhabdomyosarcoma (AR). These two pediatric sarcomas involve chromosomal translocations that create fusion proteins. For example, in about 90 percent of AR patients, there is a fusion between the PAX3 and FKHR genes, both of which are transcriptional regulators. In a chromosomal translocation between chromosomes 2 and 13, the DNA-binding domains of PAX3 are juxtaposed with the activation domain of FORCO, creating an aberrant transcription factor that is believed to be causative in this sarcoma. However, this break point also creates a potential neoantigenic determinant.
Researchers therefore asked whether there was a similar type of fusion in Ewing’s sarcoma, which involves translocations between chromosomes 11 and
22. In contrast to the single break point of AR, they found several different break points that can occur between the EWS chain and the FLI1 gene. But while several peptides might be needed, depending on the patient, the same principle does apply: the translocation events generate new tumor-specific antigens capable of binding to MHC molecules and eliciting T-cell responses.
Fortunately, the sequences of the peptides surrounding the break points contain a number of different binding motifs for both human and mouse Class I and Class II MHC molecules. Researchers do not yet have binding data and cannot be sure all of these motifs will actually lead to binding, but (1) all of these peptides span the break point, and (2) none of them are present in normal cells. Hence they have the same properties as point mutations: these MHC-binding motifs are unique to the tumor.
Researchers tested these results by immunizing mice with peptide-pulsed spleen cells and found that they could induce CTLs that would kill a tumor cell, in this case CT26, an H2D-positive BALB/c mouse colon carcinoma cell. Very importantly, a transfected cell that endogenously expressed the whole PAX3-FORCO fusion protein were also killed. This demonstrated that, as with mutant p53, endogenously expressed whole fusion protein is appropriately processed and presented on Class I molecules to be seen by CD8 CTLs.
Fusion Protein Protective Vaccines. Researchers “mock immunized” two groups of mice (A and B) as controls and immunized two other groups (C and D) with PAX3-FKHR peptide. They then injected all four groups with tumor cells from lung metastases—A and C with wild-type cells, but B and D with tumor cells that had been transfected to express fusion protein, called F8. Autopsy showed no difference in the number of number of nodules in the lungs of groups A and C, which received wild-type tumor cells. However, there was statistically significant protection in group D—immunized mice had substantially fewer nodules than unimmunized mice in group B. This experiment wasn’t perfect—transfected cells didn’t grow as well as wild-type in vivo, many mice in both arms had micrometastases that couldn’t be counted, and the results need to be repeated—but these preliminary results are encouraging.
In another experiment, researchers injected tumor cells first, allowing them to establish micrometastases, and then—24 hours later—adoptively transferred cells from immunized mice into the infected mice. Again, there was no protection against wild-type tumor cells, but there was a substantial reduction in the number of tumor modules in mice that received both transfected tumor cells and immunized CTLs. The results were not quite statistically significant, and there was the same problem with micrometastases, but researchers were encouraged by the trend, and by the fact that two of the animals had no detectable nodules whatsoever.
Conclusions. Researchers concluded from these experiments that the characteristic T-11–22 translocation of Ewing’s sarcoma and T-13 translocation of AR generate fusion proteins that act as neoantigens. Sequence analysis of these fusion proteins suggests the presence of MHC Class I and II binding motifs for both mouse and human. Immunization of experimental animals with synthetic
peptides corresponding to the fusion break point results in generation of specific CTL responses. Immunized animals have reduced tumor burden following tumor challenge, compared with controls. Adoptive transfer of bulk spleen cells from immunized animals mediated partial reduction in tumor burden in animals with established disease. These preliminary results need to be repeated with additional controls, and that work is in progress. Researchers are now planning a clinical trial in patients with these pediatric sarcomas.
Tumor Immunogenicity. These strategies would be unnecessary if the tumor itself were more immunogenic. At least five possible explanations have been offered for the failure of the tumor to elicit a CTL immune response:
Failure of the mutant epitope in the oncogene product or fusion protein to be presented by an individual’s MHC alleles. Research cannot overcome this problem, but it is not the only way to induce an immune response.
The tumor down-regulates the MHC allele responsible for presentation. Research has shown that it takes a much higher density of MHC-peptide complexes to induce an immune response than it does to be the target of a CTL immune response. In this case, even if MHC is expressed at a level insufficient to induce an immune response, it is still possible to induce an immune response on professional APCs that will induce CTLs to kill the tumor.
The tumor processes antigen inefficiently, and hence at insufficient levels to induce immune response.
Loss of costimulatory molecules. Several groups have shown that tumor cells transfected for the B7 costimulatory molecule become immunogenic and induce CD8 CTLs that will kill the tumor. That is, costimulation is necessary for the afferent (inducing) but not the efferent (killing) limb of the immune response.
Tumor cells may be tolerogenic. They present signal 1 without signal 2, and so instead of prolonging memory they kill the CTLs induced by immunization. Evidence for this is seen in the need to immunize repeatedly to suppress the tumor. Even this may have a beneficial effect in terms of the longevity of the patient.
In response to questions from the audience, Dr. Berzofsky added the following:
Investigators have intentionally avoided the nonmutated portions of p53, etc., because they don’t want to induce autoimmune responses. Because mutant p53 is overexpressed (due to prolonged survival rather than increased synthesis), this risk is particularly great.
Tumor genotype will become an important component of designing specific tumor vaccines. In the case of ras, there is a handful of common mutations, and the peptide is probably already synthesized. There are hundreds of different
mutations of p53, however, and a new peptide must be synthesized for almost every new patient. For this approach to work, clinicians will need to have on the shelf a repertoire of at least the most common peptides.
Researchers know the binding motifs for only a handful of human MHC molecules. Preliminary binding studies suggest that there is a match between the mutation and the patient’s HLA allele in about 35 percent of cases. However, even 35 percent of cancer patients is more than can be treated at present.
Because different types of cancer will have different biologies and different escape routes (see above), strategies that fail with one cancer may still succeed with others.
Phase I trials were aimed at safety rather than efficacy, and there have been clinical remissions to date, but a few patients have unexpected stability.
Tumor immunotherapy may prove to be most useful as a “cancer adjuvant,” for patients who have already had the bulk of their disease removed by surgery or other therapy, but remain at high risk of recurrence from micrometastases.
More than 100 common mutations of p53 have been observed in lung, breast, and other cancers. The necessary “repertoire” of mutant p53 peptides is the correspondingly large. This may not be, most cost-effective approach, but it is a first step. It may eventually be possible to make a DNA vaccine of these constructs.
Patients in clinical studies are immunized with autologous peptide-pulsed PBMCs, a 2-hour process. If researchers move to purified dendritic cells, grown in GM-CSF or IL-4, quality control is better but several days are required for incubation, and the patient must be available for up to a week at a time.
The current route of immunization is intravenous. This is not the best way to make antibodies, but experiments in mice have shown that the intravenous route induces CTLs more effectively than either subcutaneous or intraperitoneal when working with either spleen or dendritic cells.
CTL SCREENING FOR TUMOR ANTIGENS24
Researchers attempted to characterize tumor antigens that are recognized by CTLs, in hopes of targeting the immune system for these antigens. Using mixed lymphocyte-tumor cell culture, they obtained CTL clones that lyse autologous /tumor cells. Most of this work has been done with melanoma, which proved to be easier to work with that other tumor types that are now under investigation (e.g., sarcoma, lung, blood, renal, and head and neck carcinoma).
Investigators used a genetic approach to isolate genes coding for the proeins from which tumor antigens are derived. There are three categories of human tumor antigens:
antigens encoded by genes expressed in different tumor types but not in most normal tissue;
antigens encoded by genes expressed in both tumors and normal tissues (e.g., melanoma and normal melanocytes); and
antigens derived from abnormal proteins such as mutated proteins and fusion proteins (see preceding presentation).
Tumor Antigens (MAGE, RAGE, GAGE, RAGE). The first gene they isolated was called melanoma antigen 1 (MAGE-1), a previously unknown gene whose sequence is the same in DNA from melanoma cells and blood lymphocytes of the same patient. MAGE-1 belongs to a family of genes that are expressed in melanoma and other tumors; the coding regions in a terminal segment include about 300 amino acids for the putative proteins.
All of the MAGE genes are located in the long arm of chromosome X. Other teams have recently identified a related cluster of genes in the short arm that present strong homology with the MAGE family. Homology is even stronger with a cluster of genes isolated in the mouse genome. Researchers do not currently know the function of the MAGE genes; they will attempt to develop knockout mice in which to study the function of the proteins they encode.
Researchers used a specific PCR approach to detect the expression of the MAGE genes, and they found that they are expressed in a wide range of tumor types. They are not expressed in renal carcinoma, nor in leukemia and lymphoma; MAGE-2 and MAGE-3 are more frequently expressed than MAGE-1; and they tend to be expressed in a higher percentage of metastases and infiltrating tumors. The latter have a relatively bad prognosis, but the higher expression of MAGE genes is probably related not to higher metabolic potential but rather to the methylation status of the gene promoter.
The transcription factors capable of activating the MAGE-1 promoter are present in most if not all cells, including those that don’t express MAGE-1. The MAGE-1 promoter region contains two major elements that have ETS binding sites. It also contains a number of methylation sites (vertical bars) and HPA2 restriction sites (H), where it will be cut and digested if it is undermethylated. PCR analysis of DNA from different tumors demonstrated an inverse correlation between expression of MAGE-1 and overall degree of DNA methylation. In general, however, MAGE genes are not expressed in normal tissues, with two exceptions: testes (male germ cell lines undergo genome-wide demethylation), an placenta (primarily MAGE-4).
Two additional gene families have been found to encode for antigens that were recognized on the melanoma of the same patient. Designated BAGE and GAGE, they are expressed over the same range of tumor types, although at lower levels than MAGE-1 through 3. They are not expressed in leukemia or lymphoma, renal carcinoma, or colorectal carcinoma; and like MAGE they are not expressed in normal tissues, except the testes.
All three families of genes have been found to encode for antigens that are expressed specifically by tumors, presented by different HLAs, and recognized by different CTLs:
MAGE-1 protein has two such epitopes, located at different regions of the protein; one is presented by HLA-1, the other by HLA-C16, and researchers have identified CTLs expressing at least three different T-cell receptors. MAGE-3 also encodes for one antigen presented on HLA-1, a second presented on HLA-B44, and four others that are expressed on HLA-2.
BAGE seems to belong to a family of several genes. The protein it encodes is very short, with the immunogenic peptide located at the amino terminus. The peptide is presented to CTL by HLA-C16, so again different CTLs with different TCRs are able to recognize the peptide-HLA combination.
GAGE also belongs to a family of several genes. Researchers have isolated six cDNAs; GAGE-1 and GAGE-2 encode a peptide presented by HLA-C6, while the others encode peptides presented by HLA-A29.
The only normal tissues in which these genes express themselves are the testes and placenta. However, it appears unlikely that immunization against one or more of these antigens will cause harmful side effects due to expression in the testes. First of all, expression was found only in germ line cells, spermatocytes and spermatogonia, and since these cells do not express MHC molecules, gene expression should not result in antigen expression. Secondly, the PGFB-type fas ligand is present normally in testes and would not contribute to destructive inflammatory reactions. Third, experiments with mouse p815 tumor antigen (also expressed in the testes) produce strong CTL response in male mice, with no signs of inflammation and no loss of fertility.
Finally, other investigators have isolated an antigen from a renal carcinoma gene, designated RAGE. The protein is very short and the antigen is presented by HLA-B7. Researchers have been able to obtain a CTL response that recognizes HLA-B7 renal tumors but not B7 normal tissues. RAGE is not expressed in normal tissues except the retina, which (like the testes) do not express MHC molecules—an immunologically privileged site. However, RAGE is expressed in only 2 percent of renal cell carcinomas.
Antigens of Both Tumors and Normal Tissue. CTL responses are readily generated against several different antigens encoded on normal melanocytes as well as on melanoma. This finding was unexpected several years ago, but several genes encoding such antigens have been identified, including tyrosinase, Pme117-gp100, Melan-A-MART-1, and TRP-1-gp75. Most of the antigenic peptides are presented by HLA-2, although other peptide combinations have been found. The pattern of CTL precursors against these antigens is very different from that observed with the MAGE-like antigens.
In fact, most melanoma patients have CTL precursors that can be readily restimulated in vitro by autologous tumor cells. This implies that immunization should be possible, and should increase the levels of such CTLs. On the other
hand, most of these patients have progressive disease, and that suggests that these CTL precursors are not very effective. There is also concern for side effects—not vitiligo (due to destruction of normal melanocytes), which can occur without unacceptable consequences, but rather in the pigmented cells in the choroid layer of the retina, where the expression of fas ligand and TGJ-beta could contribute to inflammatory reactions. Nevertheless, several groups are going forward with plans for carefully devised clinical trials of immunotherapy against these antigens.
Antigens of Mutated Protein. Point mutations can also generate antigens on melanoma that are recognized by CTLs. One of the most interesting is a protein that normally binds to p16 in the regulation of the cell cycle. The mutation prevents this binding, thereby increasing entry into the S phase of the cell cycle. This mutation is both antigenic and oncogenic.
Another mutation antigen was isolated in renal carcinoma using CTL clones that recognize renal tumors but not autologous Epstein-Barr virus-transformed B-lymphocytes. The gene was identified by transfecting cos cells in combination with HLA-A21, because the lysis of this clone was inhibited by anti-HLA-A2 antibodies. Surprisingly, the cDNA alone could confer recognition, and the sequence was the sequence of HLA-A21 molecule that was mutated at position 117.
Experimental Results. Identification of tumor antigens has three advantages: (1) the patients are easily identified; (2) the antigens can be engineered for optimal immunization; and (3) there are several modes of immunization. Of those that have been tried thus far, one of the most impressive is adenoviral injection.
One experiment used mouse tumor antigen p815-A, a known epitope presented to CTL by LD molecules. Researchers inserted the sequence encoding for this peptide into the genome of an E3, E1-deletion mutant of adenovirus serotype 5. The location of the insertion was just after the promoter, which is a strong promoter active in a wide range of mammalian cells. Varying amounts of the plaque-forming units of adeno. P1A were injected intradermally in the ears of experimental mice, while other animals received equal numbers of control adenovirus. After 14 days, researchers removed spleen cells and stimulated them in vitro in the presence of L1210 cells transfected with either P1 A or P1 A and B7.1. Experimental animals that received higher concentrations produced a strong CTL response, especially if the stimulated cells were also transfected with B7.1. Results were similar when IL-2 was used in place of B7.1. Otherwise, however, results were less impressive: there was a CTL response, but never enough to protect against challenge with living p815 tumor cells, and prior infection with MD adeno virus prevented CTL responses to p815.
Prospects for Human Immunotherapy. Given the frequency of expression of different tumor antigen genes and the frequency of different HLA alleles in the human population, up to 82 percent of melanoma patients might
theoretically be eligible for immunotherapy. In practice, because some tumors express different tumor antigens, about 60 percent of melanoma patients could be eligible, and lower percentages of patients with other tumor types.
In a preliminary study, patients received three different dosages (30, 100, or 300 micrograms) of MAGE-1 peptide, without adjuvant, subcutaneously, at monthly intervals. After three injections, there was little or no toxicity, no tumor response, and no CTL response. In a second study, patients received either 100 or 300 micrograms of MAGE-3 peptide, again subcutaneously and at monthly intervals. There was no toxicity, but significant tumor response and at least one possible case of CTL response. One melanoma patient showed considerable response by the day of the third injection, but unfortunately died a month later of a brain metastasis. A second patient with 100 metastases around a skin graft showed significant regression after three injections and one year later is tumorfree. A third patient, designated AVL3, had primary melanomas removed in 1990, but in April 1995 had several metastases in the lung; by October 1995 the patient was disease-free, although the tumor subsequently relapsed.
In response to questions from the audience, Dr. van der Bruggen added the following:
Human patients show tumor response but no CTL response, either spleen or peripheral blood; experimental mice show CTL responses but no regression of tumors.
Researchers have not had the opportunity to look for tumor-infiltrating lymphocytes, nor have they looked for antibodies against MAGE-1 and MAGE3, although this is planned.
Researchers cannot explain why regressions are observed only after the third injection. One member of the audience speculated that CTLs have fairly short memory, and since the tumor itself may not be a good source of antigen, multiple injections are needed to maintain a high level of CTLs.
It may be possible to elicit a CD4 restricted response (e.g., to tumor lysate) as a supplement to the initial CD8 response.
IMMUNITY TO ONCOGENIC SELF-PROTEINS25
Immunity to Mutated Ras. Researchers asked whether or not oncogenic proteins can be targeted for vaccine and T-cell therapy. They first examine mutated ras as a prototype. Ras is activated by point mutations that are common in diverse tumors. Ras is present in about 15 percent of all human tumors, including 50 percent of colon cancer and 95 percent of pancreas cancer.
Animal experiments demonstrated that ras can function as a tumor-specific antigen, eliciting both helper T-cell and CTL responses that can decrease the growth of tumor in vitro and in vivo. Researchers have also found existent immune responses in a few patients with pancreatic and colon cancers. Most are antibody responses to normal ras, but a small number have a very restricted antibody response to the mutated protein. Other patients have a very specific T-cell response to the mutated segment of ras, but there is a problem in trying to focus immune attack against a single epitope. More importantly, oncogenic proteins that are activated by mutation have increased function, and proteins that have increased function don’t have to be present in abundant amounts.
Immunity to HER-2/neu. Rather than looking at immune responses to proteins that weren’t abundant, researchers asked whether or not it was possible to focus an immune attack against an oncogenic protein that is present in large amounts. Their prototype was HER-2/neu, a very large nonmutated protein that is expressed in very low amounts in some normal tissues, but is amplified in about 25 percent of breast cancer patients. When amplified, it is overexpressed and substantially more abundant. The structure of HER-2/neu includes a very large extracellular domain, so large that there are potentially epitopes for every individual.
Initial experiments revealed that 15 percent or 16 percent of breast cancer patients have existent antibody responses to HER-2/neu, including 42 percent of patients with documented overexpression of the protein. In the latter case, the immune response is assumed to be elicited by virtue of the fact that the protein is overexpressed. Unfortunately, some normal patients also have a response, between 2 percent and 5 percent based on screening of blood donors. Experience has shown that by setting the cutoff level high enough—in this case, a titer of less than 1:500—it is possible to exclude virtually all responses in normal individuals while retaining a much more specific response for breast cancer patients.
In many cases, the responses are extremely low, between 1:100 and 1:500, but five patients out of 96 had very substantial antibody responses, with titers of greater than 1:12,000. As it happened, these same five patients also had stage I or stage II breast cancer, as opposed to more advanced disease. In general,
antibody titer tended to be highest in patients with early breast cancer, and to decrease with more advanced disease. However, HER-2/neu was overexpressed in about 25 percent of patients with advanced breast cancer, where it was associated with more aggressive disease. Researchers speculate that the existent immunity occurs early on in the course of disease and prevents the progression of some patients.
Researchers have shown that the antibody response is to the whole protein and to both the intracellular and extracellular domains. This appears to be due to cell breakdown, which releases previously sequestered segments of the protein. The extracellular domain functions as a growth factor receptor, so that when it is amplified and overexpressed, the signalling through this protein is part and parcel of the aggressiveness of the disease. Some monoclonal antibodies to HER-2/neu have agonistic effects, some have antagonistic effects; hence, antibodies to the extracellular domain might either inhibit or stimulate the growth of breast cancer cells.
Researchers studied the patients with the highest antibody titer in greater detail. In several of these patients, the IgG antibody binds to the same tumor cells that overexpress HER-2/neu. Using epitope mapping, they have discovered a segment of the extracellular domain that is rich in cystine and thus a potential binding site. The researchers are now investigating whether this is a functional antibody.
Some patients with an IgG antibody response also showed a proliferative T-cell response to HER-2/neu. The patient with the highest antibody response also had the greatest T-cell response, to both whole protein and to peptides from the intra- and extracellular domains. Other patients showed proliferative response to peptides but not whole protein. Researchers have not yet mapped all of the epitopes involved.
Animal Models. To learn how to immunize with the HER-2/neu protein, researchers needed to develop an animal model. In the mouse, the neu sequence is not evident and has not been closed. Rat neu protein is quite homologous to human HER-2/neu, but when researchers immunized rats with purified neu protein in complete Freund’s adjuvant, they could find neither proliferative T-cell response nor antibody response.
Earlier researchers had developed a vaccinia virus vector that expressed the extracellular domain of rat neu protein. While this vector was immunogenic in mice, however, it too failed to elicit a response in rats. Those researchers had concluded that the failure to elicit a response was due to tolerance to self. However, subsequent screening has identified patients with existent immune response, proving that is possible to overcome tolerance to HER-2/neu.
Researchers therefore began to focus on immunizing to fragments of the protein. They found that immunizing rats with groups of intracellular domain neu peptides resulted in peptide-specific T-cell and antibody responses, and that the T-cells that responded to peptide also responded to whole protein. They were also able to immunize to peptides from the extracellular domain, but the response was much weaker. This may be due to a biological principle—i.e.,
tolerance of extracellular domain is more stringent because it is a shed protein that is more available to the thymus for induction of tolerance—or it may be because researchers merely picked the wrong peptides to work with.
Sequencing shows that the peptides to which rats were immunized are identical in the human protein (rat neu and human HER-2/neu are about 89 percent homologous at the amino acid level). Hence, the antibodies that will immunoprecipitate the rat protein will also immunoprecipitate the human protein. These are also peptides to which at least some human patients have responded.
Researchers now plan to go forward with a vaccine trial in which humans patients will be immunized with peptides identified as immunogenic in rats. The adjuvant used in the rat studies was complete Freund’s, which is too toxic for standard use in humans, so they plan to use GM-CSF as an adjuvant. Instead of growing dendritic cells in vitro with GM-CSF, the cytokine is injected intradermally with the peptide. Animal tests have shown that it is possible to generate immune response to intra- and extracellular domain peptides with GM-CSF as the adjuvant, and indeed that the DTH assay is much stronger that when using complete Freund’s adjuvant. Researchers have submitted to an IRB a protocol to immunize patients with breast and ovarian cancers with peptides. FDA has signed off, and the protocol should begin within months. Data should be available before 2000 on whether these peptides are toxic and/or efficacious in vivo.
Autoimmune Cancer Therapy. Generating an immune response to self-protein does not resolve the issue of whether it is possible to induce an aggressive autoimmune response as a form of cancer therapy. To answer this question, researchers focused on the prostate—once the prostate becomes malignant, it is often removed, at which point any prostate tissue left in the body is by definition malignant. If it were possible to induce a rapidly destructive, aggressive autoimmune prostatitis, it would have therapeutic benefit.
There are several problems in this use of autoimmune prostatitis; not the least is immunological tolerance to self-proteins. There is no information available on which prostate-specific proteins are immunogenic; nor have experiments confirmed that autoimmunity can destroy normal prostate tissue. In addition, most autoimmune disease is relapsing and often resolves spontaneously, while this strategy requires a rapid and aggressive autoimmunity that can eradicate the organ. (Parenthetically, there is a general lack of attention to CTLs in the field of autoimmunity, possibly because autoimmunity isn’t mediated by CTLs, or possibly because CTLs are too difficult to deal with.)
Researchers first tried to immunize to prostatic acid phosphatase (PAP), a common and well-characterized marker for human prostate tumor for which the sequence of a rodent model is also known. PAP is a glycoprotein secreted exclusively by prostate epithelial cells. It is expressed by all normal prostate tissue and by most prostate cancers. And while portions of the molecule are similar to
acid phosphatases from other tissues, other portions are prostate-specific. All of these factors made it an appealing target.
As with HER-2/neu, purified rat PAP with complete Freund’s adjuvant induced neither helper T-cell nor antibody response. Peptides induced helper T-cell responses in female rats, which had little or no protein in their systems, and in two cases the response was peptide-specific. But rat peptides failed to induce any response in male rats. Trying another approach, researchers first immunized male rats to homologous human PAP peptides, then immunized with rat peptides, and this sequential immunization did induce helper T-cell and antibody responses to rat PAP. Researchers believe that their ability to immunize females but not males is related to the peptides they used, which are far less abundant in females than in males; they are currently investigating this supposition.
Even when they succeeded in immunizing male rats, however, researchers found that there was none of the inflammation to the prostate that would have been expected. This suggests that tumor immunotherapy vaccination is very ineffective for inducing any antitumor response. Researchers suspect that it will be necessary to use T-cell therapy to get the precursor frequency high enough to support an immune response. The researchers have not yet tested or tried to elicit CTLs.
Researchers are currently trying to determine which proteins are most immunogenic in autoimmune prostatitis, and will focus in the future on those among them that are expressed by prostate cancer. Relatively little is known about autoimmunity and prostate in either humans or rats—only that prostate inflammation can be induced in rats immunized to prostate homogenate. Previous studies used multiple injections of homogenate with repeated use of complete Freund’s, which is no longer allowed. This results in antibody response to a variety of different proteins, which are now being identified, but immunization with whole prostate doesn’t get a response to PAP.
By selecting out the fractions of protein that induce the greatest antibody response, and then immunizing to these fractions alone, researchers were able to generate a very substantial, rapidly progressive, destructive autoimmune prostatitis. The researchers are now concentrating their efforts on (1) identifying the exact targets of this response and (2) learning how to induce this response prospectively and at will.
Considerable additional research will be needed before these results can be used for therapeutic benefits in human prostate cancer. Autoimmune prostatitis is an ill-defined syndrome, and none of the relevant antigens have been identified. Animal experiments may reveal how to maximize the destructive response, but it will still be necessary to identify the human proteins that are homologous to rat proteins, and verify that the homologous antigen is immunogenic in humans, before instituting human vaccine and T-cell therapy trials for prostate cancer.
In response to questions from the audience, Dr. Cheever added the following:
Several strains of rats are used in these studies.
Whole prostate homogenate appears to contain fractions that somehow suppress or block the immune response, as well as fractions that induce a response. Researchers use the Western blot test to determine which induce an antibody response, and then select for them.
Ras and other oncoproteins still represent a valid approach, but it will apply only to a small subset of patients. The advantage of working on a protein like HER-2/neu is the greater likelihood of getting a response in every individual who has that very common protein.
CYTOKINES AND THEIR LOCAL ENVIRONMENTS26
Potential Energy Model. A certain threshold level of response against a particular tumor antigen is required for rejection of that tumor. The endogenous level of immunity against the antigen is below this threshold; vaccination must enhance existent immunity sufficiently to raise the response above this threshold level. This has implications for the choice of antigen and vaccine strategy:
Even if the endogenous level of immunity to antigen A is close to the thre-shold, if the vaccine approach is weak, the response will not reach the threshold, and the vaccine fails.
On the other hand, if the endogenous level of immunity to antigen B is much lower, even a strong vaccine may not be able to raise the response to the threshold, and the vaccine still fails.
An extreme example of the latter case is an antigen expressed in the thymus, with tolerance generated by clonal deletion; the endogenous immunity is negative, and no vaccine approach could hope to produce a therapeutic response.
Consequently, the ideal strategy is to identify a good antigen whose endogenous level of immunity is relative close to the threshold, along with a strong vaccine approach. This involves issues of endogenous self-tolerance, repertoire, and vaccine approaches. At the present time, however, there is almost an embarrassment of riches in terms of different approaches to cancer vaccines—peptides, recombinant adenovirus, pox virus, Listeria, BCG, etc. It may well be impossible to test each of them reasonably in patients unless there is first a concerted effort to compare them rigorously, in a head-to-head fashion, in the appropriate animal models.
Role of Cytokines. For many cancers these target antigens still aren’t known, although it is assumed that tumor cells themselves are important—the
antigens relevant to immune response are in there, somewhere. However, another avenue of research has concentrated on the lack of signal 2 as the missing ingredient in immune response against tumors. There was evidence from earlier studies that one way to activate T-cells against tumor antigens was to provide signal 1 and signal 2 on the same cell, and that an important signal 2 for CTLs was IL-2 (or lymphokines made by helper T-cells).
Based on this paradigm, researchers began introducing cytokine genes into tumor cells, beginning with IL-2, in order to produce a whole-cell vaccine that contained all of the antigens and presented peptide as MHC signal 1 and signal 2 on the same cell. Aided by the use of defective retroviral vectors, they inserted a wide range of cytokine genes into a weakly immunogenic tumor, vaccinated animals, and compared the resulting protection against challenge with wild-type tumor cells. The most effective vaccine involved tumor cells transduced with a GM-CSF gene.
This was a surprise at the time, but other groups soon reported that GM-CSF has the unique and interesting function of inducing hematopoietic progenitors to differentiate not only into granulocytes and macrophages, but also into dendritic cells. It may be that these high-potency APCs, differentiating locally in the presence of GM-CSF, have something to do with the enhanced systemic immune response to GM-CSF-transduced tumor cells. Subsequent research has tried to explain how this process works.
Paracrine Cytokine Adjuvant. What turned out to be important physiologically, in addition to the particular cytokine, is the elaboration of that cytokine at the site of the antigen. In a sense, the GM-CSF-transduced tumor cell actually represents a timed-release depot for two sets of molecules: its own antigens, and GM-CSF. Importantly, it also replaces the complex “black boxes” of conventional adjuvants (BCG, C.parvum, mycobacterium, TB) with a single molecule, and in doing so generated a systemic antitumor immune response that was more potent than mixing irradiated tumor cells with conventional adjuvants.
Researchers have learned that APCs that differentiated at the site of the vaccine, under control of GM-CSF, actually ingest antigens from tumor cells and process them into both the Class I and Class II pathways, an example of crosspriming. As the APCs are ingesting and processing antigens, they are travelling to the draining lymph node, where one can first identify activated Class I- and Class II-restricted CTLs and helper cells. Once activated, these cells leave the draining lymph node and circulate systemically.
An important implication is that effector-phase CD4 is very important, in addition to CD8-positive cells. Consequently, the best vaccines will involve epitopes that actually represent tumor antigens. For this reason, there should be concern over the use of “universal helper epitopes,” which are not expressed by the tumor cell.
Clinical Trials. These results led researchers to initiate a Phase I trial in patients with metastatic renal cancer who had undergone nephrectomy to remove the primary tumor. They used a retroviral vector to transduce human GM-CSF gene into tumor cells, expanded these cells, irradiated them at doses that
inhibited replication but not immunogenicity, and then vaccinated patients with three monthly injections at two different doses, half the dose intradermal and half subcutaneous. Researchers didn’t expect and didn’t find any toxicity; the important results were immunological.
In order to determine whether paracrine elaboration of GM-CSF generated human immune responses, patients were randomized in a double-blind fashion to receive either irradiated tumor cells or irradiated tumor cells transduced with the GM-CSF gene. DTH was chosen as the simplest and most reproducible assay for in vivo immune response. A total of 30 days after vaccination, 1 million nontransduced cells were injected at a distant site, and the diameter of induration, edema, and erythema was measured to determine DTH response.
When the blind was broken, the results indicated that the lower vaccine dose (4 million cells) did not induce significant immune responses. However, the higher dose (40 million cells) did produce some fairly impressive DTH responses in patients who had received the GM-CSF-transduced vaccine. Even more encouraging were the results of the biopsy analysis of infiltrates at the DTH site. In addition to the quantitative difference, roughly 50 percent of the infiltrating cells in GM-CSF-transduced patients were eosinophils, whereas no eosinophils were found infiltrating the DTH sites of nontransduced patients. This was the same result observed in animal tests.
Anecdotally, one of the three patients that received the higher dose of transduced vaccine showed a fairly significant clinical response. This patient had multiple pulmonary metastases from his renal cancer that had progressed very rapidly over the 2 months between surgery and initial vaccination. After 1 vaccination, there was evidence of regression, and after the third vaccination there was a 95-percent reduction in the volume of metastatic tumor. While anecdotal, this suggests that there may be a therapeutic correlate to the observed immune response.
Allogeneic Vaccines. In these renal cancer patients, as in the melanoma patients discussed above, the vaccine did not induce a response against normal tissue; patients showed no impairment or autoimmunity of their remaining kidney, just as melanoma patients did not develop vitiligo. This suggests that it should be possible to generate responses against them without generating a clinically prohibitive autoimmune disease. At the same time, there is mounting evidence that the immunorelevant antigens in tumors are shared, as was the case in the MAGE proteins (see above).
This may provide a rationale for using a genetically modified allogeneic vaccine. Tumor antigens are presented to T-cells not by the tumor but by APCs derived from host bone marrow. This implies that it may not be necessary to match HLA between the vaccine and the patient. Certainly, generic vaccines would be far less labor-intensive and less expensive than individualized vaccines, and—since quantities would no longer be limited by the growth potential
of the individual’s tumor—it would also be possible to vaccinate with higher doses.
Antigen-Specific Tumor Vaccines. Researchers also hope to use activated T-cells to identify the relevant antigens for antigen-specific vaccines. One group has been pursuing this concept in the mouse model using CT26, an NMU-induced colon cancer, in order to identify the repertoire of antigens that are being recognized by the CD8 arm of the immune response following vaccination with GM-CSF-transduced tumor cells. Their technique involves taking bulk T-cells from the draining lymph nodes (rather than tumor-specific CTLs), eluting peptides from Class I molecules, fractionating them with reverse-phase HPLC, and using surrogate targets to assay peptide fractions for bioactivity. The results indicated that there was only one bioactive fraction, suggesting that the majority of CD8-positive immune response was focused on a single peptide among the many presented by the CT26 tumor. These results have been repeated in 40 separate experiments.
This peptide, called AH1, had a molecular weight of 1,128 and was doubly charged. It sensitized surrogate target cells down to a concentration of 5 X 10-12 molar. Upon sequencing, it proved to be a peptide derived from an endogenous murine MuLV gene that is normally completely silent in the BALB/c genome but is reactivated by altered methylation, much like MAGE-1 (see above). (The latter finding has led to new interest in endogenous human retroviruses, which also seem to be reactivated in human tumors.)
AH1 is not expressed in normal tissues from BALB/c mice, including normal colon and small intestine epithelium, but it is turned on in a number of different tumors. Interestingly, if T-cells from vaccinated animals are stimulated for two rounds with AH1 plus IL-2 and then adoptively transferred back into animals with CT26 tumors, most of the animals are cured. However, there is no significant response when tumor-bearing animals are vaccinated with AH1-pulsed dendritic cells or other approaches.
Researchers believe that they should examine several other approaches for introducing these gene products in vivo, such as viruses that target them to both the Class I and Class III MHC pathways. In addition, there are some interesting recombinant viral and bacterial approaches that should be compared head-to-head with endogenous tumor antigen models before deciding which approaches will be taken to clinical trials, with their tremendous investment of time, effort, and expense.
In response to questions from the audience, Dr. Pardoll added the following:
Although GM-CSF up-regulates B7 in macrophages, it does not do so in tumor cells and, hence, this cannot explain their increased immunogenicity. In fact, the major effect of transducing B7 into tumors is to provide a target molecule for NK cells to more actively lyse the tumor cells. A B7-transduced tumor cell isn’t nearly as good an APC as a bone marrow progenitor that differentiates into a dendritic cell in the presence of GM-CSF.
Researchers have compared paracrine GM-CSF (either by transduction or by time-release microspheres) with BCG and C. parvum in seven or eight different tumor models. The resulting systemic immune response generated by GM-CSF is between 1.5 and 4.0 logs more potent than either adjuvant.
While researchers have learned a tremendous amount from their experiments, they are also certain that transducing autologous tumor explants is not feasible for large-scale application in the general patient population. Microsphere approaches obviate the need for GM-CSF transduction, and the question is moot if immunodominant tumor antigens are in fact shared.
However, it may be 15 to 30 years before investigators identify the relevant antigens for all of the important tumors.
GM-CSF significantly up-regulates both Th-1 and Th-2 lymphokines.
A single vaccination with B7-transduced tumor cells does not evoke a measurable response that maps to the tumor’s MHC type. There is a small response to a second vaccination.
Tumor antigens will ultimately prove to be very important, but at present the best strategy for identifying immunorelevant antigens is to use whole-cell vaccines and let the immune system indicate which of the 50,000 or 100,000 antigens in that tumor it is capable of responding to.
There is no data to support the assertion that viral sequences bind to MHC better than self-sequences. However, high-affinity binding generates a much more profound tolerance.
A total of 10 percent or 15 percent of tumors turn off MHC Class I, in which case CTL response is irrelevant. In such cases, NK cells can be brought into the response to replace CD8 responses.
Once the relevant antigens are defined, melanoma and cervical cancer are both logical targets for cancer vaccines. So, too, are cancers of dispensable tissues (e.g., ovary, prostate, breast). Eventually, tumor vaccines might be used prophylactically, to prevent tumors caused by viruses before they occur (e.g., human papilloma virus, hepatitis C virus).