replication (14). When viewed in the context of the quantitative properties of HIV infection in vivo, the apparently inevitable development of resistance following drug monotherapy illustrates a potential advantage for gene therapy of HIV disease. By virtue of their action through Watson-Crick base pairing, only nucleic acids (antisense molecules and ribozymes) can presently be prospectively targeted at specific sites within the approximately 10-kb HIV genome, which are widely or universally conserved in natural isolates (15, 16). Such regions are likely to be less susceptible to escape mutations that simultaneously preserve viability. Furthermore, genes that target these multiple sites can be combined to virtually eliminate the possibility of virus escape, akin to the concept of multidrug combinations, but still delivered in a single vector. Basic science investigations of steric interaction are also the fundamental source for protease inhibitors, the most potent anti-HIV drugs yet, which are among the first clinically effective drugs developed from precise knowledge of three-dimensional protein structure. Molecular prediction is, however, more versatile and more specific with nucleic acid-based therapies.
Ribozymes are small, catalytic antisense RNAs that bind and cleave specific sites in target RNAs (17, 18). Cleavage, a cis reaction in the natural setting, can be engineered to occur in trans and results through the action of a central region containing secondary structure that is not base-paired with the substrate. The cleavage products are rapidly degraded in cells. The catalytic mechanism (one ribozyme molecule can cleave many substrate molecules in succession) may provide an advantage over antisense approaches. Our laboratory has concentrated on the hairpin ribozyme (19–27); other groups have employed hammerhead ribozymes for antiviral studies (28–31). Using Moloney murine leukemia virus-based retroviral vectors for delivery, hairpin ribozymes have been shown to confer protection from HIV-1 infection of T-cell lines, primary T cells, and macrophage-like progeny of CD34+ hematopoietic progenitor cells (19–25). The use of two ribozymes targeting the long terminal repeat (LTR) and env genes of HIV-1, each fused to an RNA decoy [the RRE (rev response) element), resulted in a potent antiviral vector that effectively inhibits replication of diverse clades of HIV-1 (F.W.-S. and A.Gervaix, unpublished data). Recently, a ribozyme-mediated inhibition of SIVmac was demonstrated in tissue culture (26). Furthermore, transduction of Rhesus macaque cord blood-derived CD34+ cells with this ribozyme conferred viral resistance to both the T cells and macrophage progenies (F.W.-S., M.Heush, G.Kraus, M.Rosenzweig, and P.Johnson, unpublished data). Application of this ribozyme to the SIVmac model, currently the most relevant animal model of AIDS pathogenesis, may allow testing of antiviral efficacy in vivo. In addition, a phase I trial for use of autologous T cells transduced with two hairpin ribozymes that cleave conserved sites in the HIV-1 LTR and pol has received FDA approval to enroll patients.
This and other ongoing gene therapy trials using anti-HIV molecules (32) entail the relatively cumbersome and expensive procedures of T-cell leukapheresis, ex vivo transduction with the Moloney murine leukemia virus vector, ex vivo expansion, and infusion of transduced cells (Fig. 1). The approach is feasible and currently necessary for proof-of-concept studies, but is not likely to be practical or comprehensive enough for routine use. In particular, it does not access the macrophage reservoir, an important component of the in vivo burden. While reconstitution with CD34+ hematopoietic progenitor cells has obvious advantages, the ability to reconstitute HIV-1-infected individuals, who exhibit complex derangements of hematopoiesis, remains uncertain.
If stem cell therapy proves workable for HIV disease, transduction of the most primitive precursor, probably a subset within the CD34+, CD38− population, has the most chance of success. Targeting this subset and converting from cumbersome ex vivo transduction processes to direct in vivo gene delivery are central goals.
Lentiviral vectors have attracted interest with respect to both of these aims. The capacity of lentiviruses to infect non-cycling cells probably resides in the ability of the lentiviral preintegration complex to traverse an intact nuclear envelope through the nuclear targeting properties of both the p17Gag protein and the accessory protein Vpr (33, 34). This property, which is not shared by oncoretroviruses or Moloney murine leukemia virus-derived retroviral vectors, has spurred efforts to develop lentiviral-based gene therapy vectors. The practical goal to which such investigations aspire is stable transfer of genes to rare (and rarely dividing) stem cells and to postmitotic cells in the hematopoietic, nervous, and other body systems.
Other properties of HIV vectors may be particularly desirable for treatment of HIV infection. First, vector systems employing an HIV envelope may allow direct lineage-specific targeting to CD4+ T cells and to non-cycling macrophages and glial cells in vivo. Second, rescue of the vector in vivo by patients’ HIV-1 may result in an effective amplfication of the vector through several cycles before lack of selection pressure results in reverse transcription-derived mutations. Third, the tat and rev regulatory cycles may be exploited to achieve inducible expression of delivered genes. These combined features could elevate gene transfer efficiency to the realm of in vivo therapy.
Notable progress has recently been made with an HIV-1-based system employing vesicular stomatitis virus G protein (VSV-G)-pseudotyped HIV-1 vectors (35); titers exceeding 105/ml and delivery of a lacZ marker gene to post-mitotic cells (neurons) in rodent brain were reported. This system relies upon transient transfection to generate the vector because expression of VSV-G lyses the producer cells.
Although other gene transfer vectors can transduce nondividing cells (e.g., adenovirus vectors), other limitations, chiefly the lack of a stable, consistent genomic integration mechanism, limits their applicability. Adeno-associated virus has been reported to integrate at a specific locus in chromosome 19, but proof of integration and stable gene transfer by engineered vectors in non-dividing cells remains elusive (36).
Other lentiviral vector systems have been studied (37–44). All are derived from HIV-1. Several use wild-type replication-competent helper virus as the source of virion proteins, and some represent simple pseudotyping of an env gene-mutated full-length provirus by VSV-G (37). In general, two problems have been troublesome in this field: (i) vector titers, with the exception of Naldini et al. (35), have been low (101–103) or not reported and (ii) stable packaging lines have been difficult to develop for these viruses, which have more genes and much more complex genetic regulation schemes than the simple retroviruses such as Moloney murine leukemia virus. Carrol et al. (42) reported the first packaging cell lines derived from HIV-1. However, the HXBc2-derived packaging construct expressed a defective Vpr protein, which may interfere with the normal karyophilic properties of the HIV pre-integration complex (33, 34), and the lines expressed predominantly unprocessed gag/pol precursor.
Our focus with lentiviral vectors has been 3-fold. First, we are experimenting with both the native HIV envelopes for lineage-specific gene delivery and with pseudotyped particles because of their higher stability and potential to transduce CD4-negative stem cells. Second, we have concentrated on