Even though intersubtype recombinants are evident in humans co- or super-infected with two or more different HIV-1 isolates [31–35], the frequency and survival of intersubtype HIV-1 recombinants are highly variable during disease. The few studies on de novo emergence of intersubtype recombination in vivo reflect the difficulties of identifying dual or super-infections at time of actual occurrence as well as the careful follow-up required to identify possible recombinants. As a consequence, analyses of HIV-1 recombination are quite common in vitro but are limited as a model for various reasons. Based on in vitro studies, the frequency of retroviral recombination within the 9.7 kilonucleotides of HIV-1 genome fluctuates between three and thirty recombination events per round of replication and is highly dependent on (i) cell type , (ii) the use of primary HIV-1 isolates [12, 13, 17] versus defective strains [14–16], (iii) sequence identity between different HIV-1 strains/genomes [12, 13], and (iv) selection of all recombinants [14–16] or only those that are replication competent [12, 13, 17, 37]. Many studies including those of our group have helped to elucidate various factors driving intersubtype recombination following infections with a defective retrovirus harboring a heterodiploid genome [12–16, 36, 37]. The most striking observations may relate to an obvious increase in intersubtype recombination in genomic regions with the highest sequence identity . Although this might be expected, we also observed "hotspots" for intersubtype recombination breakpoints that are less dependent on sequence identity and may be more related to the mechanism(s) of strand transfer during reverse transcription [13, 38].
In vitro models likely identify the correct progenitors of intersubtype recombinants, but the final composition of intersubtype recombinants in a dually infected patient may reflect other factors, which include the obvious requirement of virus replication in the face of different host and immune selective pressures. To understand which intersubtype recombination breakpoints can lead to replication competent virus, we had to develop a system to enrich for intersubtype HIV-1 recombinants in tissue culture. In past studies, we have examined intersubtype recombination through a dual infection of cell lines or primary human cells with two or more primary HIV-1 isolates [12, 13, 17]. In the first round of dual infection, the frequency of co-infected cells (with two different viruses) reflects the initial multiplicities of infection (MOI). For example, in a flask with 100,000 cells, an MOI of 0.01 virus A and 0.01 virus B (i.e. 1 virus per 100 cells) would result in approximately 10 cells being co-infected with both virus A and B. However, previous reports suggest that with two identical virus strains (aside from a marker) the frequency of dual infection is often higher than expected from random virus-cell interactions . In these co-infect cells, Hardy-Weinberg equilibrium and an assumption of equal packaging of both RNA genomes would predict that half the virus would be heterodiploid. As virus titer of both A and B increased during multiple rounds of replication, so would heterodiploid virus production, but eventually all susceptible cells are exhausted for infection in the culture. Thus, parental viruses (e.g. A and B) always dominate a dual infection and basically obscuring the characterization of replication-competent intersubtype HIV-1 recombinants, i.e. present at very low levels. Adding the complexity to this system, recombinants are only generated following a de novo infection with a heterodiploid A+B virus. Even though the frequency A/B cross-over event during reverse transcription can be as high as 50% over a 9.7 nt genome, less than 10% of these resulting intersubtype HIV-1 recombinants survive due to generation of defective virus or simply being unable to compete with the parental strains .
To focus our analyses on intersubtype recombination, we developed a system to selectively inhibit parental virus replication in a dual infection and as consequence, to enrich for only intersubtype HIV-1 recombinants. This system first involved the design and testing of virus-specific siRNA that would not only inhibit replication of a single parental virus, but would also enrich for intersubtype recombination in a specific HIV-1 genomic region. For the purposes of this study, siRNA120a selectively inhibited a subtype A primary HIV-1 isolate (v120-A) by targeting a 5' sequence in the HIV-1 env genes whereas a 3' env sequence was targeted by siRNA126a to specifically inhibit a subtype D primary HIV-1 isolate (v126-D). This inhibition of HIV-1 replication by siRNAs is due to the targeting and degradation of newly transcribed HIV-1 mRNA. We and others have recently shown that incoming HIV-1 genome RNA is protected by the HIV-1 core proteins and cannot be degraded by the RISC-siRNA complex [23, 40]. Destablization of the core with the exogenous addition of TRIM5α will increase core dissociation, increase RISC-siRNA access, and result in HIV-1 RNA degradation. However, this process is not activated in normal conditions of human cell lines . Based on siRNA degradation of only HIV-1 mRNA and following reverse transcription, our dual siRNA treatment would not influence the retroviral recombination mechanism but only select for those intersubtype D/A recombinants with breakpoints that occurred at a frequency of >5%/1000 nt in the env gene  and up to 15-20% between the siRNA target sequences. These D/A env recombinants could be propagated and enriched in culture considering they are resistant to the siRNA inhibition of HIV-1 mRNA transcription, again at a step subsequent to reverse transcription and integration. As illustrated in Figure 4B, D/A env recombinant virus became the majority of replicating virus when treated with the two siRNAs during two rounds of propagation. In absence of this siRNA enrichment/selection, D/A env recombinant virus represented only 6.7% of the replicating virus population, which was dominated by the parental v120-A and/or v126-D virus.
The frequencies of recombination in these propagated dual infections were initially estimated by PCR amplification using isolate-specific oligonucleotide primers. However, due to non-specific amplification, these recombination frequencies were likely over-estimates. To determine a more accurate level of recombination and more importantly, to map the site of recombination, the dual infections propagated in the presence or absence of siRNAs were PCR amplified with conserved env primers. Cloning and sequencing of these env products revealed that D/A recombinants could hardly be detected in the absence of siRNAs, but that dual siRNA treatment resulted in 39 v126-D/v120-A recombinants in 50 env clones. Interestingly, the remaining clones were v120-A or v120-A/v126-D env genes, and 10/11 harbored mutations in the siRNA120a target sequence. We suspect that the mutations in this target sequence conferred resistance to the siRNA120a and as consequence would be resistant to both siRNAs. Predominance of v120-A with siRNA120a resistant mutations as opposed to v126-D with siRNA126a mutations likely relates to the increased fitness of v120-A over v126-D in these dual infections . In fact, we did not observe v126-D/v120-A recombinants after eight rounds of dual siRNA selection, but instead v120-A dominated with a single mutation in the siRNA120a target sequence. We are now determining how dual infections with viruses of equal or different fitness might influence (1) the frequency of recombination, (2) the rate of intersubtype recombinant viruses in the presence of dual siRNA treatment, and possibly, (3) the sites of recombination breakpoints.
Our previous studies indicated that intersubtype recombination breakpoints were scattered across conserved regions, i.e. C1, C2, and C3, using a single cycle assay [12, 13]. As mentioned earlier, the multiple cycle assay in the absence of siRNA enrichment results in a very low level of recombinants co-circulating with the parental strains in culture. As a result, we suspect that the pattern of v126-D/v120-A recombination in our previous multiple-cycle/dual infection assays may be the result of some re-sampling of recombinant clones . In addition, obtaining replication-competent recombinants from the dual infection was less likely in the absence of siRNA enrichment due to the continuous generation of both defective and replication-competent recombinants and the ongoing parental strain replication. In the study presented herein, use of dual siRNA treatment inhibited the replication of parental strains, and only 126D/120A recombinants were likely to survive two rounds of propagation. In addition, we examined recombination sites across most of the env gene (~1700 nt) (this study) as opposed to just the C1-C4 regions (~1100 nt) . Nearly identical numbers of recombination breakpoints were identified in the C2 region as compared to other C1-C4 regions regardless of system, i.e. 45% 126D/120A recombination breakpoints in C2 with single cycle system versus 46% with multiple cycle  and 48% with multiple cycle in the presence of siRNA enrichment (this study). The differences relates to the positioning of the breakpoints in C2. Replication-competent 126D/120A recombinant viruses appeared to harbor more breakpoints near the V2/C2 junction than those from single cycle assays. Due to the siRNA enrichment, we could now perform more careful mapping of the 126D/120A breakpoints in the C2 region of env. It is now quite clear that recombination in C2 maps to two specific regions, nt positions 6813-6873 and nt positions 6938-7005. The C2 region is 299 nt in length and yet, aside from these 60 and 67 nt segments, the remaining 172 nt have nearly no recombination breakpoints. These findings clearly indicate that pattern of breakpoints in intersubtype HIV-1 recombinants is shaped by both the reverse transcription process and during subsequent selection for replication competent virus.
Although numerous selective forces could influence selection of these intersubtype HIV-1 recombinants within a host, several studies have now shown a clear overlap in the recombination breakpoints derived from intersubtype HIV-1 recombinants generated in tissue culture with those identified in unique and circulating intersubtype recombinant forms (URFs and CRFs) found in the HIV-1 epidemic [12, 17, 38]. Considering that HIV-1 intersubtype recombinants represent approximately 20% of all infections, vaccines and new drug therapies could be designed based on a better understanding of strand transfer mechanisms during reverse transcription and the actual breakpoints that give rise to functional, replication competent virus. It is also apparent that intersubtype HIV-1 recombination, following a dual infection or superinfection of already infected individual, could lead to rapid immune escape. A simple comparison of our A/D env sequences with HIV-1 epitope maps, restricted by common HLA alleles (http://www.hiv.lanl.gov/content/immunology/maps/ctl/gp160.html), revealed that 30/39 of the clones had recombination breakpoints within immunodominant epitopes.