Pathogen Discovery, Detection, and Diagnostics
David A. Relman, M.D.
Departments of Microbiology and Immunology and of Medicine
Stanford University, Stanford, California
VA Palo Alto Health Care System, Palo Alto, California
Microbial Diversity and the Limitations of Cultivation Methods
Beginning in the late 1970s and early 1980s, explorations of the natural microbial world focused on extreme environments and exploited the use of newly described molecular approaches for phylogenetic analysis and classification. Recovery of sequence-based signatures of life directly from these environments confirmed revolutionary proposals for three aboriginal lines of descent (Woese and Fox, 1977), and led to the realization that nearly all microbial life is resistant to cultivation in the laboratory. With increasing reliance on molecular methods in environmental microbiology, a picture of microbial diversity emerged that currently includes as many as 40 major divisions of bacteria; a broad and cosmopolitan domain of life known as the Archaea; and an intertwined early history of endosymbiotic prokaryotes, eukaryotic protists, and lateral gene transfer events (Pace, 1997; Hugenholtz et al., 1998). The inadequacies of available cultivation techniques are reflected in the fact that nearly 90 percent of all known cultivated bacterial species are contained within just 4 of the 40 divisions, even though many of the other divisions are equally or more diverse and well populated. The fact that 65 percent of all published microbiological research over a 6-year period concerned just 8 bacterial genera is a dramatic
illustration of our strong bias toward bacteria that are amenable to cultivation (Hugenholtz, 2002).
Given these facts, it is concerning but perhaps not surprising that clinical microbiology continues to rely heavily upon cultivation-based methods. In fact, cultivation methods have improved considerably over the past several decades with advances in the scope and diversity of media components, control of environmental conditions, use of heterologous host cells, and use of growth-promoting factors (Mukamolova et al., 1998) A number of recently recognized and newly described microbial pathogens have been cultivated successfully in the laboratory, including spirochetes, rickettsia, actinomycetes, and a variety of viruses. Because the internal environmental conditions of the human body are viewed as more hospitable to life and are more easily replicated in the laboratory than are many external environmental conditions, we often assume that microbial cultivation efforts have been relatively successful. To be sure, there is no dearth of known, cultivated microbial pathogens.
On the other hand, we should not be so complacent about the completeness of our inventory of microbial pathogens, or about the sensitivity of detection methods for cultivation-amenable microorganisms. When traditional diagnostic methods are rigorously applied to syndromes of suspected infectious etiology, such as pneumonia, encephalitis, lymphocyte-predominant meningitis, pericarditis, acute diarrhea, and sepsis, only a minority of cases can be explained microbiologically. The majority of emerging infectious disease agents are zoonotic organisms, and as such are better adapted to nonhuman environmental conditions. In addition, a long list of chronic inflammatory diseases with features of infection remain poorly understood and inadequately explained from a microbiological perspective. Thus, it seems fair to speculate that the identification of pathogens in only seven bacterial divisions and the absence of any known pathogens within the domain Archaea may represent an imperfect understanding of the true diversity of microbes capable of causing human disease. Furthermore, it seems reasonable to conclude that current methods are lacking in sensitivity.
Molecular Approaches for Microbial Pathogen Detection and Identification: Seeking Signatures
In an effort to avoid reliance on cultivation and to establish alternative and complementary approaches, one might view the goal of microbial detection and pathogen discovery as identifying molecular signatures of infection. These signatures must be reliable in identifying a microorganism and in establishing the relationships between a previously uncharacterized organism and those that have previously been characterized. Molecular signa-
tures can be based directly upon the features of the microbe itself or upon the features of the host response to a pathogen. And there are a variety of methods and techniques with which to acquire each of these two types of signatures (Relman, 1999).
Genomic sequence is the most frequently used “currency” in the identification of microbial signatures, and broad-range (or consensus) polymerase chain reaction (PCR) is the most practical tool for generating this currency (Relman, 1999). Furthermore, ribosomal DNA (rDNA) is among the most useful kinds of genome sequence from which organismal ancestry and interrelationships can be reliably inferred and pathogen discovery approaches designed (Relman et al., 1990, 1992). With recent improvements in the speed with which primary genome sequence can be acquired and analyzed, other detection or screening formats may become widely available and additional regions of microbial (and viral) genomes more commonly targeted. For example, high-density microarrays of oligonucleotides or amplified DNA products can be designed to screen complex pools of microbial nucleic acid for specific agents in a massively parallel and efficient manner. This technical platform obviates the need to clone and sequence large numbers of variant microbial molecules, which is particularly relevant to the analysis of clinical specimens with a significant burden of “background” microorganisms (see the further discussion below), and facilitates more sophisticated uses of pattern recognition analysis as a tool for microbial signature identification.
There are at least two alternative kinds of approaches for detecting diagnostic signatures of microbial origin that incorporate features to help discriminate between signal and noise. The first relies on differential analysis of microbial sequences in specimens from host sites that are involved and uninvolved by disease. Differential display is a screen for differences in sequence diversity and abundance between involved and uninvolved sites; representational difference analysis in essence selects for sequences of differential abundance, using PCR (Chang et al., 1994; Gao and Moore, 1996). Both approaches share the disadvantage that the sequences of differential abundance that are revealed by these methods may not be useful markers for microbial identification. A interesting study of Crohn’s disease illustrates this problem (Sutton et al., 2000). The second kind of alternative approach relies upon the host immune response to identify sequences that may originate from a putative pathogen. These techniques include screening of expression or phage display peptide libraries with patient antisera or reactive T cells (Hemmer et al., 1999).
Despite a greater emphasis on efforts to discover and detect microbes by targeting them directly, the host response to infection offers attractive features for pathogen detection and classification that are unique and complementary. In theory, the host response provides microbial signatures
that are, by definition, clinically relevant—that is, the host response largely defines whether infection has led to disease. It is also more intimately connected to clinical outcome and may provide signatures with prognostic value. Finally, signatures based on host response do not presuppose the presence of the putative pathogen for specimen analysis. There is further discussion of these approaches and their relative value below.
Exploring the Human Microbiome in Health and Disease
Surveys of the microbial communities associated with humans during states of health, using molecular techniques, have been initiated, albeit belatedly. These surveys are important for a number of reasons. In addition to the numerous but poorly characterized beneficial effects of the endogenous microflora on human health, a proper understanding of community membership, relative abundance, and variations therein will be critical for recognizing potential pathogens and patterns that are predictive of disease. Basic principles and paradigms in the field of ecology have not yet been applied to the study of the human microbial ecosystem. For example, we have virtually no information on the levels of microbial diversity (“richness”) and abundance (“evenness”) that are optimal for maintenance of local health, or of those that are associated with disease. We are only just recently learning about microbial partitioning within human micro-environments, and still understand little about interindividual variability or variability as a function of time. The validity of the “intermediate disturbance theory” (Sousa, 1984; Buckling et al., 2000) and its application to human endogenous microbial communities have not been explored. Furthermore, our capabilities to predict the effects of physical or chemical (e.g., antibiotic) perturbation on these microbial communities is extremely limited.
The subgingival crevice in the mouth is one of the more intensively studied and better understood colonized sites of the human body. Molecular surveys using broad-range rDNA PCR suggest that approximately 50– 60 percent of the bacteria present at this site are distinct from all of those previously described at the taxonomic level of species—albeit a term that is loosely defined (Kroes et al., 1999; Paster et al., 2001). Some of these bacteria are not assigned to the dominant four divisions—actinobacteria, firmicutes, bacteroidetes, and proteobacteria—and belong to divisions such as TM7 and OP11 that have not previously been recognized or discussed by clinicians or clinical microbiologists, probably because they contain no known cultivated members. Molecular surveys of the endogenous flora of the human intestinal tract have only just begun. Although the clinical significance of the newly discovered community membership has not yet been established, it is clear that the human endogenous microflora play an im-
portant role in a variety of important disease states involving the skin and mucosal tissues. What is not clear, but widely speculated, is the possible role of the endogenous flora in either provoking or propagating disease at distant sites, including the central nervous system. The proposed associations between viral respiratory infections and subsequent flares of multiple sclerosis and between Campylobacter jejuni enterocolitis and Guillain-Barre syndrome (Ang et al., 2002) are but two examples.
Seeking Evidence of Causation
The increasing availability of molecular pathogen discovery methods and the ease with which molecular signatures are generated create a pressing problem of a different kind: How can one build a convincing body of evidence for a causative role of the putative pathogen in a disease process when the pathogen is identified with molecular signatures and has not been isolated or purified? The issues surrounding this problem are familiar to epidemiologists and have been addressed during the past half-century using distinct terminology that nonetheless is quite relevant today. One can adapt the same concepts to the kinds of data and techniques generated and used in modern approaches to pathogen detection and discovery (Fredericks, 2001). In particular, the ability to connect a signature physically to the sites of pathology where one most expects to find the putative disease agent is a helpful evidentiary component. Fluorescent in situ hybridization allows one to correlate a specific sequence with areas of pathology and tissue-based microbial structures (Fredricks and Relman, 1996; Fredricks et al., 2000). This approach also allows one to examine signature “dosage” effects. Alternatively, anatomic sites of interest can be targeted specifically for signature detection using laser capture microdissection (Emmert-Buck et al., 1996; Becich, 2000).
Recognition and Classification of Microbial Disease Based on Host Gene Expression Patterns
The limitations of methods for analysis of microbial signatures and the emergence of technology platforms for rapid, highly parallel gene expression measurements have facilitated a potentially important, independent approach for identification of microbial disease. The basic question raised is whether one can recognize and classify clinical (and preclinical) states of infection by examining host gene response patterns (Cummings and Relman, 2000; Diehn et al., 2000a). This approach offers several advantages. First, changes in gene transcript abundance occur within minutes of a new exogenous stimulus. Second, the complexity and diversity of signal transduction
mechanisms that impact on human gene expression and the complexity of the output (at a genome-wide level) are extensive; therefore, discrimination between numerous diverse stimuli (e.g., different classes of pathogens) may be discernible. Third, a clinical specimen need not contain the exogenous stimulus, i.e., the infectious agent. Fourth, the intrinsic nature of the host response may be directly informative about the clinical relevance of the stimulus (host–microbe interaction) and the clinical outcome. At the present time, however, the answer to this basic question is not yet available.
The vast majority of work in this area to date has focused on the response of host cells to microbial stimuli in vitro (Boldrick et al., 2002; Nau et al., 2002). A larger body of work predates this more recent emphasis on microbial stimuli and addresses the nature of the expression patterns associated with various forms of cancer (Alizadeh et al., 2000; Ross et al., 2000; Perou et al., 1999). As many might have predicted, findings from examination of host–microbial encounters in vitro indicate the predominance of shared gene expression patterns, suggesting a stereotyped temporally controlled response to microbes in human cells (Boldrick et al., 2002). Gene expression responses exhibit microbial dose dependence, yet universal, shared dose equivalence relationships are not apparent. From these early experiments, it appears that identification of discriminatory (diagnostic) signatures may be possible. Furthermore, active, virulence-associated mechanisms may provide the basis for specific pathogen class recognition.
The transition to an analysis of humans with and without known infectious diseases ex vivo is accompanied by a number of interesting but complex questions. What is the most useful and practical type of clinical specimen from which to record genome-wide expression patterns and discern meaningful information about infection? Blood cells are attractive, given that they circulate, make contact with a wide variety of microenvironments and other cell types, and are easy to obtain. But it is unclear how well they might reflect a localized infectious process, e.g., in the brain. How much variability occurs within and between individuals during various states of health and during noninfectious stimuli? Must each individual serve as his or her own control, for proper interpretation of infection-associated responses? What kinds of host-specific genetic information and proclivities are embedded in expression data? These questions are currently being explored, but are quite wide-ranging and will require extensive sampling before they can be answered in a comprehensive fashion. One of the most intriguing questions is on what basis humans classify noxious stimuli, and, in particular, microbial causes of disease. Among the most likely uses and practical outcomes of these investigations is the identification of patterns that predict disease outcome (Alizadeh et al., 2000). Furthermore, expression analysis can be used to identify predicted membrane-associated and
secreted proteins (Diehn et al., 2000b); with this approach, diagnostic and prognostic transcript abundance patterns can be converted to sets of easily measured proteins in body fluids.
MICROBIAL DETECTION AND DIAGNOSIS
Sample Collection and Processing
Current diagnostic approaches for the collection of environmental samples or clinical specimens involve primarily hands-on, ad hoc procedures using a variety of devices and instruments. In the clinical arena, these procedures result in specimens of variable quantity and quality. The process is relatively laborious and nonstandarized. A few common methods for specimen disruption are applied to each specimen type without particular regard for the possible diversity of pathogens and their various requirements, nor are special precautions used in a uniform manner to minimize degradation of pathogens or their viability. Problems with lack of standardization and nonuniformity of procedure are even more prevalent in the area of environmental microbial detection. Air and water are among the environmental specimen types that are currently collected and processed most effectively. Because of the time demands and resource constraints found in today’s clinical workplace, as well as recent laboratory downsizing, recovery of fastidious microbes from clinical specimens has almost certainly suffered (Bartlett et al., 2000). For example, increasing delays from the time of sputum specimen collection to the inoculation of appropriate culture media have probably contributed to the decreasing recovery rates of S. pneumoniae from cases of pneumonia. Some technology developments, such as the use of microsonicators for efficient rapid microbial lysis, are likely to improve the current situation (Belgrader et al., 1999b; Taylor et al., 2001).
Traditional approaches for microbial detection and identification include microbial cultivation, immunological (e.g., antibody-based) assays, and nucleic acid detection schemes—especially amplification methods such as PCR (Tang et al., 1997; Fredricks and Relman, 1999). Cultivation is the most widely used approach in laboratories, clinics, and health care facilities throughout the world, especially in developing countries, and hence is currently the most common microbial detection platform for international surveillance. Cultivation, despite being slow, limited in sensitivity for some clinically relevant microbes, and the least technologically sophisticated,
nevertheless provides the most ready assessment of complex microbial phenotypes (behaviors), such as drug resistance. Solid-phase immunological assays, such as dipsticks and optical immunoassays, have established a niche in the clinical workplace, but their utility has been demonstrated in only a limited number of infectious diseases settings (Needham et al., 1998). PCR is the most widely used method for microbial nucleic acid detection; other signal and target amplification techniques for nucleic acid detection, such as ligase chain reaction, have generated more limited commercial markets (Tang et al., 1997; Fredricks and Relman, 1999).
Although hundreds of different microbe-specific PCR assays have been described, and many of these have been applied to diverse environmental problems, a much smaller number of assays has entered routine clinical practice. Examples include assays for N. gonorrheae, C. trachomatis, herpes simplex virus, and HIV. PCR can be used to detect antibiotic resistance (Fluit et all., 2001); however, the diversity of genotypes and mechanisms associated with this phenotype and the difficulty of predicting expression from simple gene detection have hampered this approach. A modest number of recent studies have confirmed that the use of these molecular diagnostic tests can reduce patient-care costs and favorably impact patient management (Dumler et al., 1999; Ramers et al., 2000). Some of the factors that may have limited more widespread use of these theoretically appealing molecular approaches are specimen issues (see later discussion), a paucity of studies that address clinical validation, and the need for specialized expertise. Again, technology advances with PCR may in the near future shift further attention toward this platform. In particular, the development of rapid, real-time (semiquantitative) PCR with point-of-care microbial detection within 30 minutes (Belgrader et al., 1998, 1999b) may potentially alter the use of antibiotics on a widespread scale and reduce antibiotic resistance (Bergeron and Oullette, 1998). Overall, each of the three detection/diagnostic platforms provides complementary advantages and disadvantages. No one approach alone currently provides a rapid and reliable method for microbial detection in the real world.
The above comments are in general equally relevant to clinical and environmental microbial detection and identification. In the environmental detection arena, much emphasis has been placed on sensor technology. The results of these investments have led to a plethora of sensor types but relatively limited maturation of any one platform. Problems are similar to those in the clinical arena, and focus on a lack of real-world validation (see later discussion). Future efforts in environmental microbial detection will likely emphasize the integration of multiple types of environmental data, building on the principles established in the field of measurement and signature intelligence.
A disproportionate effort in microbial detection and diagnosis has been devoted to technology platform development, rather than to standardization or validation of sample collection and processing procedures, or to test validation in a real-world setting. While analytical performance characteristics have been deliberately pursued, “clinical” performance characteristics have been relatively ignored. For example, it is important to be able to anticipate false-positive test results in a reliable and quantitative fashion. However, a proper calculation of positive predictive value requires some understanding of the pretest probability of a true positive test. In the area of environmental detection, pretest probabilities for the presence of a variety of microbial agents is probably low and difficult to determine. One is unavoidably left with a situation in which the false-positive rate is likely to be high and poorly characterized. Many of these problems also apply to the situation with clinical and preclinical diagnosis.
ISSUES AND PROBLEMS
Clinical Relevance of Laboratory Findings
The inevitable result of increasingly sensitive detection platforms is difficulty in establishing the clinical relevance of positive test results. The important distinction between infection and disease, i.e., colonization or contamination of a host with a potential biothreat agent, and pathology (disease) has challenged clinicians for a century. The same problem arises in environmental analysis. Few data are available with which to infer the level of risk to a human host of acquiring disease after detection of a potential pathogen in the environment. Sensitive and specific diagnostic tests are vitally important adjuncts to clinical diagnosis; however, screening and diagnostic tests cannot replace the crucial need for careful studies that define likelihood of exposure or examine correlations between detection and disease.
One issue of particular importance concerns the complexity and widespread distribution of microbial sequence “background” or “noise” observed in the analysis of human clinical specimens (both experimental and biological) discussed earlier. The distribution and nature of this sequence background have as yet not been well characterized. Findings of bacterial rDNA in association with blood samples from healthy humans threaten to expand the extent of this problem, and involve the analysis of anatomic compartments that have traditionally been viewed as usually sterile (Nikkari et al., 2001). A different perspective on this same apparent problem was provided in an analysis of expressed sequence tag libraries from human
tissues (Weber et al., 2002; Relman, 2002). Some of the transcripts that were originally assumed to derive from the human genome appeared on closer inspection to be of microbial origin. Whether some of these molecules were intrinsic to the original specimen or introduced later remains unclear; but at least a portion can easily be attributed to agents that are common, persistent, or dormant infectious agents found within these human tissues.
Breadth of Current Assays and Approaches
There is a strong tendency in molecular diagnostics to focus disproportionate attention on a small number of pathogens that have proven to be amenable to detection. Positive findings tend to create a self-fulfilling prophecy. Few efforts are invested in improving methods for the detection of less commonly found pathogens. In addition, important lessons have been learned in recent years about the use and limitations of molecular methods for microbial pathogen discovery (see the earlier discussion). First, the assumption that sequences identified as universally conserved within a group of organisms are in fact found in all members of the group is not always justified. As previously unrecognized members of a group are revealed, small additional degrees of sequence variation are sometimes discovered. The small subunit rDNA sequences that were originally described as universal are now known to be conserved in only a subset of cellular life (Lane et al., 1985); revised sequence sites have taken their place. Second, PCR can exhibit bias and favor certain members of a mixed starting pool of molecules. The use of multiple broad-range primer pairs or reaction cosolvents may avoid a skewed perspective. Third, conserved sequences for use in broad-range PCR have not yet been identified and validated for all groups of viruses. This limitation most certainly contributed to the sizable number of cases that remained unexplained after investigation by the Unexplained Deaths and Critical Illnesses Working Group within the CDC’s Emerging Infections Program (Nikkari et al., 2002). Finally, a large number of detection and diagnostic tests rely upon a small number of specific antibodies or microbial genomic sequences. This reliance creates vulnerabilities. Microbial pathogens that have variant antibody epitopes (binding sites) or sequences will fail to be detected, and may in fact be selected over time.
The problems associated with clinical and other real-world specimens are substantial, and are currently underappreciated and inadequately targeted by funding agencies. These problems became highlighted during the investigations of the Unexplained Deaths and Critical Illnesses Project un-
der the direction of CDC and collaborating scientists (Nikkari et al., 2002; Hajjeh et al., 2002). The Unexplained Deaths Project is an enhanced passive surveillance system that identifies life-threatening cases of disease with features of infection for which routine laboratory tests fail to provide a microbiological diagnosis. It has served as a source of clinical specimens with which newer molecular diagnostic methods can be evaluated. Problems revealed in this study are common in most clinical settings. First, clinical specimens from cases of suspected but unproven infectious etiology are often obtained late in the disease course, at a time when the putative agent may no longer be present. Second, the site from which the specimen is obtained may not coincide optimally with the expected anatomic distribution of the agent. Third, the quantity of specimen may be insufficient for the expected concentration of the agent and a reasonable probability of its presence (as a single particle or genome equivalent) in the specimen. And fourth, in the real world of clinical medicine, specimen handling and storage may introduce exogenous contamination, spurious signals, or target degradation.
Standardization of Procedures
The preceding discussion has suggested that current diagnostic and detection procedures lack sufficient standardization and validation. The barriers to a successful resolution of this problem are multiple. First, additional method development and optimization for specimen collection and processing are necessary. Second, a set of uniform reagents is needed, with which multiple competing procedures and platforms can be cross-evaluated. Third, additional funds and incentives will need to be provided before an effort of the necessary breadth and rigor is undertaken by public and private organizations, as well as academic, government, and commercial ventures.
To address the problems and issues outlined above, a number of needs must be met. First is validation of methods in the real world. To this end, an investment in the development of procedures and in the kinds of resources described earlier must be undertaken. Second, for any given assay for a specific agent, we need additional information on the distribution and abundance of that agent and its close relatives in a wide variety of environments. As broad-range assays become more widely used in the mid- to distant future, a proportionately larger-scale effort must be undertaken to describe the microbial background of clinical and environmental sites. Third, with increasing emphasis on the use of microbial sequences for detection and
diagnosis, a strong imperative is created for a broad investment in microbial genomics. In addition, the usefulness of genome sequences for microbial forensics further emphasizes the need for this investment (Cummings and Relman, 2002). Fourth, high-throughput sequencing technology and sample handling (robotics) have advanced recently in dramatic fashion. As a result, it becomes more practical and timely to consider major investments in laboratories that can handle large numbers of samples and provide surge capacity (Layne et al., 2001; Layne and Beugelsdijk, 1998). Such laboratories might facilitate standardization of methods and technology development. Finally, centralized repositories of diverse, high-affinity binding and detection reagents (e.g., antibodies, peptides, oligonucleotides, aptamers (Brody and Gold, 2000) should be established, as well as centralized repositories of genomic material and control samples. These resources would assist with standardization and validation of methods and help minimize reliance on an overly narrow set of detection reagents.
ANTICIPATING MICROBIAL THREATS
Intelligence Gathering: Human Intention
Deliberate release of a biological agent as a weapon poses a number of additional challenges in detection and diagnosis, including a much-expanded spectrum of agents and greater difficulty in calculating pretest probabilities and positive predictive values. In a timeline of events related to a bioterrorist attack, the earliest stage predates release of a bioagent and involves the evolution of a bioweapons plan or program from inception through weapon deployment. The greatest potential benefit might derive from preemptive efforts at this point in the timeline. Opportunities exist in the areas of human intelligence gathering. The science and technology communities have not been adequately tapped for their expertise in this regard. The challenge of understanding human intent in the area of biotechnology requires familiarity with science and technology culture, process, and procedure.
Intelligence Gathering: Nature
A variety of forms of evidence suggest that the diversity and number of microbial disease agents and virulence genes in nature is immense. Most of these agents and genes remain uncharacterized and unfamiliar to us, in part because they have not yet had opportunities for exposure to humans, or exposure has not been recognized. Examples of well-known reservoirs for emerging pathogens include rodents (Sin nombre virus or Lassa fever virus), birds (West Nile virus), and fruit bats (Hendra virus or Nipah virus). Within specific contexts and with appropriate epidemiological leads, one might
consider a directed survey of suspected reservoirs for potential viral and microbial pathogens. Broad-range PCR would be an attractive approach for such a project (see earlier discussion). However, high-throughput sequence analysis and large-volume sample analysis will be required for recognition of meaningful patterns and establishment of significant associations with disease. These needs may ultimately require the development and use of high-density DNA microarrays designed for broad-range microbial surveys.
A Second Human Genome Project: Microbial Community Genomics
As one broad approach to the problem of poorly understood endogenous flora, one might consider a second human genome project (Relman and Falkow, 2001). Such a project would entail a comprehensive inventory of microbial genes and genomes at the four major sites of microbial colonization in the human body: mouth, gut, vagina, and skin. Community microbial genomics is a rapidly emerging experimental approach toward understanding the composition, functional capabilities, coevolution, and interactions of complex groups of microbes, many of which are resistant to cultivation or purification. This approach has already been undertaken in the study of marine and soil microbial ecology (Beja et al., 2000, 2002; DeLong, 2001; Rondon et al., 2000). A community genomics analysis of the human microbiome would provide an equally rich data set from which critical issues in human health and emerging infectious diseases could be addressed. This second human genome project could be approached with a combination of random shotgun sequencing procedures, targeted large-insert clone sequencing, and assessments of intra- and interindividual variation using high-density microarrays. From these flora arise current and future opportunistic microbial pathogens, including those that are drug-resistant. From inventories of endogenous microbial community membership, gene content, and gene expression, it may be possible to identify patterns that indicate early stages of colonization or takeover by newly acquired pathogens. With increasing degrees of population sampling in well-characterized settings and with the integration of host genome-wide expression analysis (Relman and Falkow, 2001; Hooper et al., 2001), major insights into the role of endogenous flora in health and disease will be gained.
LOOKING TO THE FUTURE
The principles that are embedded in the use of genome-wide expression patterns for classification and characterization of infectious diseases (see earlier discussion) can be viewed as an important generic feature of future
directions in microbial diagnostics and detection. Complex biological signatures and pattern recognition are relevant to the analysis of the endogenous microbial flora, protein expression profiles, secreted or exhaled volatile small molecules, and spectral properties of human cells and tissues. Protein microarrays (Haab et al., 2001; Templin et al., 2002) are rapidly emerging as a complementary platform for the potential generation of important biological signatures from clinical specimens. The same is true for high-throughput mass spectroscopy methods (Petricoin et al., 2002). Patterns can be discerned that embed diagnostic and prognostic information without the need for a clear understanding of mechanism. As mentioned earlier, endogenous microbial flora might be the source of informative patterns that would indicate exposure to or incipient development of infectious disease, as well as prior behavior of the host. It is already clear that automated and miniaturized technology platforms, such as microfluidics chips and cartridges, will speed the development of point-of-care, rapid, hands-off diagnostic tests (Belgrader et al., 2000; Pourahamadi et al., 2000). Technology cannot substitute for a holistic understanding of biological systems, but exciting clinical investigation will be greatly accelerated by new and emerging technologies.
Alizadeh, A. A., Eisen, M. B., Davis, R. E., Ma, C., Lossos, I. S., Rosenwald, A., Boldrick, J. C., Sabet, H., Tran, T., Yu, X., Powell, J. I., Yang, L., Marti, G. E., Moore, T., Hudson, J., Jr., Lu, L., Lewis, D. B., Tibshirani, R., Sherlock, G., Chan, W. C., Greiner, T. C., Weisenburger, D. D., Armitage, J. O., Warnke, R., Staudt, L. M., et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. (2000) Nature 403, 503-11.
Ang, C. W., Laman, J. D., Willison, H. J., Wagner, E. R., Endtz, H. P., De Klerk, M. A., Tio-Gillen, A. P., Van den Braak, N., Jacobs, B. C., and Van Doorn, P. A. Structure of Campylobacter jejuni lipopolysaccharides determines antiganglioside specificity and clinical features of Guillain-Barre and Miller Fisher patients. (2002) Infect Immun 70, 1202-8.
Bartlett, J. G., Dowell, S. F., Mandell, L. A., File Jr., T. M., Musher, D. M., and Fine, M. J. Practice guidelines for the management of community-acquired pneumonia in adults. Infectious Diseases Society of America. (2000) Clin Infect Dis 31, 347-82.
Becich, M. J. The role of the pathologist as tissue refiner and data miner. (2000) Mol Diagn 5, 287-99.
Beja, O., Suzuki, M. T., Koonin, E. V., Aravind, L., Hadd, A., Nguyen, L. P., Villacorta, R., Amjadi, M., Garrigues, C., Jovanovich, S. B., Feldman, R. A., and DeLong, E. F. Construction and analysis of bacterial artificial chromosome libraries from a marine microbial assemblage. (2000) Environ Microbiol 2, 516-29.
Beja, O., Suzuki, M. T., Heidelberg, J. F., Nelson, W. C., Preston, C. M., Hamada, T., Eisen, J. A., Fraser, C. M., and DeLong, E. F. Unsuspected diversity among marine aerobic anoxygenic phototrophs. (2002) Nature 415, 630-3.
Belgrader, P., Benett, W., Hadley, D., Long, G., Mariella, R., Jr., Milanovich, F., Nasarabadi, S., Nelson, W., Richards, J., and Stratton, P. Rapid pathogen detection using a microchip PCR array instrument. (1998) Clin Chem 44, 2191-4.
Belgrader, P., Benett, W., Hadley, D., Richards, J., Stratton, P., Mariella, R., Jr. and Milanovich, F. PCR detection of bacteria in seven minutes. (1999a) Science 284, 449-50.
Belgrader, P., Hansford, D., Kovacs, G. T., Venkateswaran, K., Mariella, R., Jr., Milanovich, F., Nasarabadi, S., Okuzumi, M., Pourahmadi, F., and Northrup, M. A. A minisonicator to rapidly disrupt bacterial spores for DNA analysis. (1999b) Anal Chem 71, 4232-6.
Belgrader, P., Okuzumi, M., Pourahmadi, F., Borkholder, D. A., and Northrup, M. A. A microfluidic cartridge to prepare spores for PCR analysis. (2000) Biosens Bioelectron 14, 849-52.
Bergeron, M. G., and Ouellette, M. Pathogenesis of pneumococcal pneumonia in cyclophosphamide-induced leukopenia in mice. (1998) Infect Control Hosp Epidemiol 19, 560-4.
Boldrick, J. C., Alizadeh, A. A., Diehn, M., Dudoit, S., Liu, C. L., Belcher, C. E., Botstein, D., Staudt, L. M., Brown, P. O., and Relman, D. A. Stereotyped and specific gene expression programs in human innate immune responses to bacteria. (2002) Proc Natl Acad Sci U S A 99, 972-7.
Brody, E. N., and Gold, L. Aptamers as therapeutic and diagnostic agents. (2000) J Biotechnol 74, 5-13.
Buckling, A., Kassen, R., Bell, G., and Rainey, P. B. Disturbance and diversity in experimental microcosms. (2000) Nature 408, 961-4.
Chang, Y., Cesarman, E., Pessin, M. S., Lee, F., Culpepper, J., Knowles, D. M., and Moore, P. S. Identification of herpesvirus-like DNA sequences in AIDS-associated Kaposi’s sarcoma. (1994) Science 266, 1865-9.
Cummings, C. A., and Relman, D. A. Using DNA microarrays to study host-microbe interactions. (2000) Emerg Infect Dis 6, 513-25.
Cummings, C. A., and Relman, D. A. Genomics and microbiology. Microbial forensics— “cross-examining pathogens.” (2002) Science.
DeLong, E. F. Microbial seascapes revisited. (2001) Curr Opin Microbiol 4, 290-5.
Diehn, M., Alizadeh, A. A., and Brown, P. O. Examining the living genome in health and disease with DNA microarrays. (2000a) Jama 283, 2298-9.
Diehn, M., Eisen, M. B., Botstein, D., and Brown, P. O. Large-scale identification of secreted and membrane-associated gene products using DNA microarrays. (2000b) Nat Genet 25, 58-62.
Dumler, J. S., and Valsamakis, A. Molecular diagnostics for existing and emerging infections. Complementary tools for a new era of clinical microbiology. (1999) Am J Clin Pathol 112, S33-9.
Emmert-Buck, M. R., Bonner, R. F., Smith, P. D., Chuaqui, R. F., Zhuang, Z., Goldstein, S. R., Weiss, R. A., and Liotta, L. A. Laser capture microdissection. (1996) Science 274, 998-1001.
Fluit, A. C., Visser, M. R., and Schmitz, F. J. Molecular detection of antimicrobial resistance. (2001) Clin Microbiol Rev 14, 836-71, table of contents.
Fredericks, D. N., and Relman, D. A. Sequence-based identification of microbial pathogens: a reconsideration of Koch’s postulates. (1996) Clin Microbiol Rev 9, 18-33.
Fredricks, D. N., and Relman, D. A. Application of polymerase chain reaction to the diagnosis of infectious diseases. (1999) Clin Infect Dis 29, 475-86; quiz 487-8.
Fredricks, D. N., and Relman, D. A. Localization of Tropheryma whippelii rRNA in tissues from patients with Whipple’s disease. (2001) J Infect Dis 183, 1229-37.
Fredricks, D. N., Jolley, J. A., Lepp, P. W., Kosek, J. C., and Relman, D. A. Rhinosporidium seeberi: a human pathogen from a novel group of aquatic protistan parasites. (2000) Emerg Infect Dis 6, 273-82.
Gao, S. J., and Moore, P. S. Molecular approaches to the identification of unculturable infectious agents. (1996) Emerg Infect Dis 2, 159-67.
Haab, B. B., Dunham, M. J., and Brown, P. O. Protein microarrays for highly parallel detection and quantitation of specific proteins and antibodies in complex solutions. (2001) Genome Biol 2, RESEARCH0004.
Hajjeh, R. A., Relman, D., Cieslak, P. R., Sofair, A. N., Passaro, D., Flood, J., Johnson, J., Hacker, J. K., Shieh, W. J., Hendry, R. M., Nikkari, S., Ladd-Wilson, S., Hadler, J., Rainbow, J., Tappero, J. W., Woods, C. W., Conn, L., Reagan, S., Zaki, S., and Perkins, B. A. Surveillance for unexplained deaths and critical illnesses due to possibly infectious causes, United States, 1995-1998. (2002) Emerg Infect Dis 8, 145-53.
Hemmer, B., Gran, B., Zhao, Y., Marques, A., Pascal, J., Tzou, A., Kondo, T., Cortese, I., Bielekova, B., Straus, S. E., McFarland, H. F., Houghten, R., Simon, R., Pinilla, C., and Martin, R. Identification of candidate T-cell epitopes and molecular mimics in chronic Lyme disease. (1999) Nat Med 5, 1375-82.
Hooper, L. V., Wong, M. H., Thelin, A., Hansson, L., Falk, P. G., and Gordon, J. I. Molecular analysis of commensal host-microbial relationships in the intestine. (2001) Science 291, 881-4.
Hugenholtz, P. (2002) Genome Biol 3, REVIEWS0003.
Hugenholtz, P., Goebel, B. M., and Pace, N. R. Impact of culture-independent studies on the emerging phylogenetic view of bacterial diversity . (1998) J Bacteriol 180, 4765-74.
Kroes, I., Lepp, P. W. and Relman, D. A. Bacterial diversity within the human subgingival crevice. (1999) Proc Natl Acad Sci U S A 96, 14547-52.
Lane, D. J., Pace, B., Olsen, G. J., Stahl, D. A., Sogin, M. L., and Pace, N. R. Rapid determination of 16S ribosomal RNA sequences for phylogenetic analyses. (1985) Proc Natl Acad Sci U S A 82, 6955-9.
Layne, S. P., and Beugelsdijk, T. J. Laboratory firepower for infectious disease research. (1998) Nat Biotechnol 16, 825-9.
Layne, S. P., Beugelsdijk, T. J., Patel, C. K., Taubenberger, J. K., Cox, N. J., Gust, I. D., Hay, A. J., Tashiro, M., and Lavanchy, D. A global lab against influenza. (2001) Science 293, 1729.
Mukamolova, G. V., Kaprelyants, A. S., Young, D. I., Young, M., and Kell, D. B. A bacterial cytokine (1998) Proc Natl Acad Sci U S A 95, 8916-21.
Nau, G. J., Richmond, J. F., Schlesinger, A., Jennings, E. G., Lander, E. S., and Young, R. A. Human macrophage activation programs induced by bacterial pathogens. (2002) Proc Natl Acad Sci U S A 99, 1503-8.
Needham, C. A., McPherson, K. A., and Webb, K. H. Streptococcal pharyngitis: impact of a high-sensitivity antigen test on physician outcome. (1998) J Clin Microbiol 36, 3468-73.
Nikkari, S., McLaughlin, I. J., Bi, W., Dodge, D. E., and Relman, D. A. Does blood of healthy subjects contain bacterial ribosomal DNA? (2001) J Clin Microbiol 39, 1956-9.
Nikkari, S., Lopez, F. A., Lepp, P. W., Cieslak, P. R., Ladd-Wilson, S., Passaro, D., Danila, R., and Relman, D. A. Broad-range bacterial detection and the analysis of unexplained death and critical illness. (2002) Emerg Infect Dis 8, 188-94.
Pace, N. R. A molecular view of microbial diversity and the biosphere. (1997) Science 276, 734-40.
Paster, B. J., Boches, S. K., Galvin, J. L., Ericson, R. E., Lau, C. N., Levanos, V. A., Sahasrabudhe, A., and Dewhirst, F. E. Bacterial diversity in human subgingival plaque. (2001) J Bacteriol 183, 3770-83.
Perou, C. M., Jeffrey, S. S., van de Rijn, M., Rees, C. A., Eisen, M. B., Ross, D. T., Pergamenschikov, A., Williams, C. F., Zhu, S. X., Lee, J. C., Lashkari, D., Shalon, D., Brown, P. O., and Botstein, D. Distinctive gene expression patterns in human mammary epithelial cells and breast cancers. (1999) Proc Natl Acad Sci U S A 96, 9212-7.
Petricoin, E. F., Ardekani, A. M., Hitt, B. A., Levine, P. J., Fusaro, V. A., Steinberg, S. M., Mills, G. B., Simone, C., Fishman, D. A., Kohn, E. C., and Liotta, L. A. Use of proteomic patterns in serum to identify ovarian cancer. (2002) Lancet 359, 572-7.
Pourahmadi, F., Taylor, M., Kovacs, G., Lloyd, K., Sakai, S., Schafer, T., Helton, B., Western, L., Zaner, S., Ching, J., McMillan, B., Belgrader, P., and Northrup, M. A. Toward a rapid, integrated, and fully automated DNA diagnostic assay for Chlamydia trachomatis and Neisseria gonorrhoeae (2000) Clin Chem 46, 1511-3.
Ramers, C., Billman, G., Hartin, M., Ho, S., and Sawyer, M. H. Impact of a diagnostic cerebrospinal fluid enterovirus polymerase chain reaction test on patient management. (2000) Jama 283, 2680-5.
Relman, D. A. The search for unrecognized pathogens. (1999) Science 284, 1308-10.
Relman, D. A. New technologies, human-microbe interactions, and the search for previously unrecognized pathogens. (2002) J Infect Dis, in press.
Relman, D. A. The human body as microbial observatory. (2002) Nat Genet 30, 131-3.
Relman, D. A., and Falkow, S. The meaning and impact of the human genome sequence for microbiology. (2001) Trends Microbiol 9, 206-8.
Relman, D. A., Loutit, J. S., Schmidt, T. M., Falkow, S., and Tompkins, L. S. The agent of bacillary angiomatosis. An approach to the identification of uncultured pathogens. (1990) N Engl J Med 323, 1573-80.
Relman, D. A., Schmidt, T. M., MacDermott, R. P., and Falkow, S. Identification of the uncultured bacillus of Whipple’s disease. (1992) N Engl J Med 327, 293-301.
Rondon, M. R., August, P. R., Bettermann, A. D., Brady, S. F., Grossman, T. H., Liles, M. R., Loiacono, K. A., Lynch, B. A., MacNeil, I. A., Minor, C., Tiong, C. L., Gilman, M., Osburne, M. S., Clardy, J., Handelsman, J., and Goodman, R. M. Cloning the soil metagenome: a strategy for accessing the genetic and functional diversity of uncultured microorganisms. (2000) Appl Environ Microbiol 66, 2541-7.
Ross, D. T., Scherf, U., Eisen, M. B., Perou, C. M., Rees, C., Spellman, P., Iyer, V., Jeffrey, S. S., Van de Rijn, M., Waltham, M., Pergamenschikov, A., Lee, J. C., Lashkari, D., Shalon, D., Myers, T. G., Weinstein, J. N., Botstein, D., and Brown, P. O. Systematic variation in gene expression patterns in human cancer cell lines. (2000) Nat Genet 24, 227-35.
Sousa, W. P. Intertidal mosaics: catch size, propagule availability, and spacially variable patterns of succession. (1984) Ecology 65, 1918-1935.
Sutton, C. L., Kim, J., Yamane, A., Dalwadi, H., Wei, B., Landers, C., Targan, S. R., and Braun, J. Identification of a novel bacterial sequence associated with Crohn’s disease. (2000) Gastroenterology 119, 23-31.
Tang, Y. W., Procop, G. W., and Persing, D. H. Molecular diagnostics of infectious diseases. (1997) Clin Chem 43, 2021-38.
Taylor, M. T., Belgrader, P., Furman, B. J., Pourahmadi, F., Kovacs, G. T., and Northrup, M. A. Lysing bacterial spores by sonication through a flexible interface in a microfluidic system. (2001) Anal Chem 73, 492-6.
Templin, M. F., Stoll, D., Schrenk, M., Traub, P. C., Vohringer, C. F., and Joos, T. O. Protein microarray technology. (2002) Trends Biotechnol 20, 160-6.
Weber, G., Shendure, J., Tanenbaum, D. M., Church, G. M., and Meyerson, M. Identification of foreign gene sequences by transcript filtering against the human genome. (2002) Nat Genet 30, 141-2.
Woese, C. R., and Fox, G. E. Phylogenetic structure of the prokaryotic domain: the primary kingdoms. (1977) Proc Natl Acad Sci U S A 74, 5088-90.