Transcription Start Site Heterogeneity and Preferential Packaging of Specific Full-Length RNA Species Are Conserved Features of Primate Lentiviruses.

HIV-1 must package its RNA genome to generate infectious viruses. Recent studies have revealed that during genome packaging, HIV-1 not only excludes cellular mRNAs, but also distinguishes among full-length viral RNAs. Using NL4-3 and MAL molecular clones, multiple transcription start sites (TSS) were identified, which generate full-length RNAs that differ by only a few nucleotides at the 5' end. However, HIV-1 selectively packages RNAs containing one guanosine (1G RNA) over RNAs with three guanosines (3G RNA) at the 5' end. Thus, the 5' context of HIV-1 full-length RNA can affect its function. To determine whether the regulation of genome packaging by TSS usage is unique to NL4-3 and MAL, we examined 15 primate lentiviruses including transmitted founder viruses of HIV-1, HIV-2, and several simian immunodeficiency viruses (SIVs). We found that all 15 viruses used multiple TSS to some extent. However, the level of TSS heterogeneity in infected cells varied greatly, even among closely related viruses belonging to the same subtype. Most viruses also exhibited selective packaging of specific full-length viral RNA species into particles. These findings demonstrate that TSS heterogeneity and selective packaging of certain full-length viral RNA species are conserved features of primate lentiviruses. In addition, an SIV strain closely related to the progenitor virus that gave rise to HIV-1 group M, the pandemic pathogen, exhibited TSS usage similar to some HIV-1 strains and preferentially packaged 1G RNA. These findings indicate that multiple TSS usage and selective packaging of a particular unspliced RNA species predate the emergence of HIV-1. IMPORTANCE Unspliced HIV-1 RNA serves two important roles during viral replication: as the virion genome and as the template for translation of Gag/Gag-Pol. Previous studies of two HIV-1 molecular clones have concluded that the TSS usage affects unspliced HIV-1 RNA structures and functions. To investigate the evolutionary origin of this replication strategy, we determined TSS of HIV-1 RNA in infected cells and virions for 15 primate lentiviruses. All HIV-1 isolates examined, including several transmitted founder viruses, utilized multiple TSS and selected a particular RNA species for packaging. Furthermore, these features were observed in SIVs related to the progenitors of HIV-1, suggesting that these characteristics originated from the ancestral viruses. HIV-2, SIVs related to HIV-2, and other SIVs also exhibited multiple TSS and preferential packaging of specific unspliced RNA species, demonstrating that this replication strategy is broadly conserved across primate lentiviruses.

INTRODUCTION

Upon infection, human immunodeficiency virus type-1 (HIV-1) reverse transcribes its RNA genome into DNA and integrates the DNA copy into host chromosomes, forming a provirus (1). Host machinery transcribes the provirus to generate viral RNAs, some of which are spliced and translated to express various viral proteins. Other viral RNAs remain unspliced and serve two important roles in viral replication: template for translation to produce the Gag/Gag-Pol polyproteins and packaging into nascent virions as the viral genome (1). Recent studies of HIV-1 molecular clones NL4-3 and MAL have demonstrated that multiple neighboring transcription start sites (TSS) are used during the transcription of HIV-1 RNA, resulting in RNAs with three, two, or one guanosines at the 5′ end (referred to as 3G, 2G, and 1G RNA, respectively) (2–7). Interestingly, while the most abundant unspliced HIV-1 RNA in cells is 3G RNA, 1G RNA is significantly enriched in virions (2–7). Therefore, HIV-1 distinguishes between ~9-kb RNAs that differ by only a couple of nucleotides (nts) and selects a minor RNA species during genome packaging.

HIV-1 has been classified into four groups (M, N, O, and P) that arose from independent cross-species transmission events (8). HIV-1 group M is primarily responsible for the HIV pandemic and has further been categorized into numerous subtypes and circulating recombinant forms (CRFs). At this time, HIV-1 TSS have only been characterized in NL4-3 and MAL, which belong to group M subtypes B and A, respectively. However, HIV-1 is quite diverse in the human population, and these two viruses might not fully represent the breadth of TSS usage among HIV-1 groups and subtypes. In addition, the evolutionary origins of HIV-1 TSS heterogeneity and selective packaging of 1G RNA have not been investigated. HIV-1 groups M and N resulted from cross-species transmission of simian immunodeficiency virus from chimpanzees (SIVcpz) into humans, whereas HIV-1 groups O and P resulted from cross-species transmission of SIV from gorillas (SIVgor) into humans (8–12). Furthermore, SIVgor also originated from SIVcpz (11). SIVcpz, in turn, is thought to have arisen from recombination of at least two other SIVs: SIV from red-capped mangabeys (SIVrcm) and SIV from mustached, mona, or greater spot-nosed monkeys (SIVmus/mon/gsn) (13, 14). TSS usage in these SIVs has not yet been examined.

In addition to the HIV-1/SIVcpz/SIVgor lineage, there are many other distinct lineages of SIVs in primates. Of these, HIV-2 is a known human pathogen and causes AIDS. However, HIV-2 is significantly less pathogenic, has a different genome structure, and replicates more slowly in cell culture than HIV-1 (15, 16). HIV-2 resulted from cross-species transmission of SIV from sooty mangabey monkeys (SIVsmm) into humans. Likewise, cross-species transmission of SIVsmm into macaques generated SIVmac (17, 18). Thus, HIV-2, SIVsmm, and SIVmac represent a group of closely related viruses that are distinct from HIV-1. Furthermore, there are several other SIV lineages including SIVagm from African green monkeys that are clearly distinct from HIV-1- or HIV-2-related viruses. In total, there are at least 45 known SIVs, with evidence of frequent cross-species transmission and recombination events (14, 19). It is currently unclear whether TSS heterogeneity and selective packaging of a particular RNA species are conserved features of these primate lentiviruses.

In this study, we have investigated the TSS usage of cellular and virion RNA in 16 primate lentiviruses. In addition to NL4-3, we examined six HIV-1 viruses including multiple clinical isolates and four SIVs including several that are closely related to the progenitor viruses that were zoonotically transmitted into humans. All of the HIV-1 and HIV-1-related viruses we examined have three guanosines at the TSS, and they used multiple neighboring nts to initiate transcription. Furthermore, all of them preferentially packaged 1G RNA. These findings indicate that multiple TSS and selective 1G RNA packaging are features that preexisted in SIV prior to cross-species transmission and the generation of HIV-1. Intriguingly, the proportions of the transcribed RNA species varied greatly across viruses, even among those belonging to the same subtype. We also examined TSS usage in two HIV-2 molecular clones and two SIVs closely related to HIV-2 (i.e., SIVmac and SIVsmm). We found that these viruses also use multiple TSS and, except for SIVsmm, selectively package specific RNA species. Taken together, our study shows that primate lentiviruses use heterogenous TSS and most of these viruses also distinguish among unspliced RNAs with minor differences at the 5′ end, indicating that these features are part of a conserved replication strategy.

RESULTS

HIV-1 NL4-3 exhibits TSS heterogeneity and preferentially packages 1G RNA in multiple cell types.

Using modified HIV-1 constructs, we and others have previously shown that transcription of NL4-3 initiates at multiple neighboring sites, generating several RNA species (Fig. 1A) (2–4, 7). First, we examined whether NL4-3 TSS usage is affected by experimental conditions such as cell types from which viruses are produced or by modifications to the viral genome intended to limit the replication cycle (Fig. 1B and C). To recapitulate our previous studies, we used the same single-cycle vector called H0 (20), an NL4-3-derived construct that contains intact long terminal repeats (LTRs), 5′ untranscribed region (UTR), gag-pol, tat, and rev, but has inactivating deletions in vif, vpr, vpu, and env. Additionally, this construct contains a heat stable antigen (hsa) marker gene in nef (Fig. 1B). To ensure capture of the authentic TSS distribution, we infected 293T cells with HIV-1 virions pseudotyped with vesicular stomatitis virus G protein (VSV G) (Fig. 1B) and used flow cytometry to detect HSA to determine the multiplicity of infection. H0 expresses functional Gag/GagPol; thus, H0-infected cells produce viral particles that are noninfectious due to the absence of Env. We isolated total RNA from infected cells and viral particles produced from these cells and analyzed the TSS using 5′ rapid amplification of cDNA ends (5′ RACE). Primers that anneal to the gag gene were used for cDNA synthesis so that only unspliced RNAs were analyzed. For both cellular and virion RNA, ~100 clones were sequenced per virus, combined across at least two independent experiments. We obtained results similar to those previously published (7); briefly, 3G RNA was the most abundant species in cells, whereas 1G RNA was enriched in virions (Fig. 1D).

FIG 1

System used to study the TSS of HIV-1 RNA. (A) The sequences near the TSS of NL4-3 and RNA species generated from using several TSS. (B and C) Experimental workflow used to determine TSS in cells infected with and viruses produced by a modified construct H0 (B) and unmodified replication-competent HIV-1 (C). H0 expresses Gag/Gag-Pol, Tat, and Rev and can carry out a single round of virus replication when supplemented with Env. (D) Proportions of HIV-1 full-length RNA species in different cells infected with and viruses produced by NL4-3 as determined by 5′ RACE. The results are summarized from at least two independent experiments with sequencing the following total number of clones: 100 (293T cell), 99 (293T virus), 119 (HUT/R5 cell), 101 (HUT/R5 viruses), 137 (CEM-SS cell), 135 (CEM-SS viruses), 127 (GHOST-X4 cell), and 101 (GHOST-X4 virus). ***, P < 0.001, chi-square test.

To examine TSS usage in intact, replication-competent NL4-3 during spreading infection, we used two T cell lines, HUT/R5 and CEM-SS, as well as an engineered osteosarcoma cell line GHOST-X4 that expresses CD4 and coreceptor CXCR4 (Fig. 1C). After infection, cells and virus supernatants were collected every 2 days, and viral replication was monitored by p24 ELISA. Cell and virion RNAs were extracted from samples collected at the peak of virus production and analyzed using 5′ RACE. In all cases, we found that in the cytoplasm, 3G RNA was the most abundant species, although a significant proportion of 1G RNA was also detected. Additionally, we identified several other RNA species starting at TSS slightly downstream of the three guanosines at T4, C5, and T6 (Fig. 1A). Most transcripts with these downstream TSS had 5′ guanosines consistent with 5′ capping, indicating that they are not artifacts of 5′ RACE. These results agree with previous observations (7) and confirm the validity of our 5′ RACE method.

We next determined the TSS in NL4-3 virions produced from infected HUT/R5, CEM-SS, and GHOST-X4 cells. In all cases, NL4-3 packaged mostly 1G RNA into virions (Fig. 1D). Additionally, TSS distributions were significantly different between cells and virions in all four cell types (P < 0.001, chi-square test; Fig. 1D). Specifically, 1G RNA was enriched in virions, while 3G RNA was depleted (P < 0.001 in all cases, Fisher’s exact test). The proportions of most other RNA species did not significantly differ between cells and virions, except for C5 RNA, which was significantly depleted in virions from 293T, HUT/R5, and CEM-SS cells (P = 0.01, 0.03, and < 0.001, respectively, Fisher’s exact test), but not GHOST/X4 cells (P = 0.09, Fisher’s exact test). These results show that the distribution of NL4-3 RNA species in cells and virions is largely consistent across cell types and between single-cycle vectors and intact replication-competent viruses. Because results from the H0 vector and replication-competent NL4-3 were similar, in our analyses below for other viruses we used either H0-like constructs to infect 293T cells or replication-competent viruses to infect T cells or GHOST cells. In all cases, infection was used instead of transfection to better recapitulate authentic TSS patterns. Furthermore, viral genomes were packaged by the appropriate authentic Gag/Gag-Pol proteins, and the LTRs, 5′ UTR, and gag-pol sequences were unaltered to reflect authentic transcription regulation and genome packaging.

TSS usage in HIV-1 group M subtype B viruses.

Although widely used in research settings, NL4-3 is a lab-adapted chimera that may not reflect the properties of subtype B viruses from patients. To address this, we determined the TSS usage of three subtype B transmitted founder viruses: TRJO, CH058, and CH106 (21–24). Transmitted founder molecular clones represent the sequences of transmitted viruses, as determined by sequencing samples from acutely infected patients and mathematical modeling (22). Similar to NL4-3, TRJO, CH058, and CH106 also have three guanosines near the putative TSS and identical 6-nucleotide (nt) sequences immediately surrounding the guanosines (Fig. 2A). In HUT/R5 cells, NL4-3 and TRJO had similar TSS usage, producing mostly 3G RNA but also a significant proportion of 1G RNA (Fig. 2B). However, CH058 and CH106 exhibited a strikingly distinct pattern of TSS usage. These viruses generated mostly C5 (rather than 3G) RNA in cells, but also a significant proportion of 1G RNA. Compared to NL4-3 and TRJO, both CH058 and CH106 contain a 1-nt deletion between the CATA/TATA box (25) and the TSS (Fig. 2A). This difference might explain the altered TSS usage for these viruses, as the TSS is partly determined by distance from the TATA box (see Discussion). Nonetheless, all four viruses had significantly different TSS distributions between cells and virions (P < 0.001 in all cases, chi-square test). Specifically, 1G RNA was enriched in virions, while 3G (NL4-3, TRJO) or 5C (CH058, CH106) RNAs were depleted (P < 0.001 in all cases, Fisher’s exact test). These results demonstrate that 1G RNA is selectively packaged over 3G or 5C RNAs in all the examined subtype B viruses.

FIG 2

TSS usage of HIV-1 group M subtype B transmitted founder viruses. (A) Alignment of sequences between the CATA/TATA box (25) and near the TSS compared to NL4-3. Green letters denote major start sites, which were defined as >20% of the sequences from cells or virions. Sequence differences relative to NL4-3 are shown in blue. (B) Proportions of RNA species in cells infected with and viruses produced by the indicated HIV-1 strains. TRJO, CH058, and CH106 are unmodified full-length viruses and were propagated in HUT/R5 cells. NL4-3 results from Fig. 1D (HUT/R5) are shown here for comparison. Results are summarized from two independent experiments with sequencing the following total number of clones: 110 (TRJO cell), 95 (TRJO virus), 127 (CH058 cell), 119 (CH058 virus), 109 (CH106 cell), and 112 (CH106 virus). ***, P < 0.001, chi-square test.

TSS usage in HIV-1 group M subtype C viruses.

To investigate whether other HIV-1 group M viruses have similar TSS usage and selective packaging of 1G RNA, we examined the 5′ ends of unspliced viral RNAs for MJ4 and CH185, two subtype C viruses (26, 27), which is the most prevalent subtype globally (28). In addition, CH185 is a transmitted founder virus identified from an acutely infected individual. Compared to NL4-3, MJ4 and CH185 viruses have the same three guanosines and immediate surrounding sequences at the putative TSS (Fig. 3A). Surprisingly, unlike NL4-3 and other subtype B viruses, the subtype C viruses MJ4 and CH185 produced mostly 1G RNA in cells (Fig. 3B), with small amounts of 3G RNA also detected. Similar results were observed using an MJ4-based H0-like vector in 293T cells and full-length replication-competent MJ4 in T cells (Fig. 3B). Thus, these two subtype C viruses had less transcriptional heterogeneity than the subtype B viruses we examined. Even so, MJ4 and CH185 had statistically significant differences in TSS distributions between cells and virions (P < 0.001 [MJ4 293T, CH185], P = 0.008 [MJ4 HUT/R5], chi-square test). In all cases, 3G RNA was depleted in virions (P < 0.001 [MJ4 293T], P = 0.005 [MJ4 HUT/R5], P = 0.02 [CH185], Fisher’s exact test). In addition, 1G RNA was enriched in virions in all cases (P < 0.001 [MJ4 293T], P = 0.049 [MJ4 HUT/R5], P = 0.045 [CH185], Fisher’s exact test). In sum, the subtype C viruses we examined predominantly generated 1G RNA and appeared to have less transcriptional heterogeneity than subtype B viruses; however, they still selectively packaged 1G RNA into virions.

FIG 3

TSS usage of HIV-1 subtype C viruses. (A) Alignment of sequences between the CATA/TATA box and near the TSS compared to NL4-3. Green letters denote major TSS, whereas blue letters denote sequence differences relative to NL4-3. (B) Proportions of RNA species in cells infected with and viruses produced by the indicated HIV-1 strains. MJ4 results in 293T cells were generated using the previously described constructs CH0 and CT6 (37). Full-length MJ4 and CH185 were propagated in HUT/R5 cells. NL4-3 results from Fig. 1D are shown for comparison. Results are summarized from at least two independent experiments with sequencing data from the following total number of clones: 106 (MJ4 cell, 293T), 97 (MJ4 virus, 293T), 104 (MJ4 cell, HUT/R5), 108 (MJ4 virus, HUT/R5), 101 (CH185 cell), and 100 (CH185 virus). ***, P < 0.001; **, P < 0.01, chi-square test.

TSS usage in other HIV-1 and HIV-1-related viruses.

TSS heterogeneity and selective packaging of 1G RNA may have arisen during the spread of HIV-1 in humans or prior to zoonotic transmission of SIV in chimpanzees or gorillas. To distinguish between these possibilities, we next determined TSS in unspliced viral RNA from cells infected with and virions generated from SIVcpzMB897 and SIVcpzEK505. SIVcpzMB897 is closely related to the ancestral virus that jumped species and was transmitted into humans becoming HIV-1 group M, whereas SIVcpzEK505 is closely related to the ancestral virus that became HIV-1 group N (29). Both SIVcpzMB897 and SIVcpzEK505 have three guanosines at the putative TSS and flanking sequences identical to NL4-3 (Fig. 4A). In 293T cells, both SIVcpzMB897 and SIVcpzEK505 produced multiple RNA species with 3G RNA being the most abundant species (Fig. 4B). Additionally, in both viruses, 1G RNA was enriched in virions, whereas 3G RNA was depleted (P < 0.001 in all cases, Fisher’s exact test). These results indicate that TSS heterogeneity and selective packaging of 1G RNA likely predate the emergence of HIV-1 groups M and N.

FIG 4

TSS usage of SIVcpz, HIV-1 group O, SIVgor, and SIVrcm. (A) Alignment of sequences between the CATA/TATA box and near the TSS compared to NL4-3. Green letters denote major TSS, whereas blue letters denote sequence differences relative to NL4-3. (B) Proportions of RNA species in cells infected with and viruses produced by a given virus. SIVcpz MB897 molecular clone was engineered to be encoded by two plasmids containing an overlapping region and inactivated env gene. To generate virus, these two plasmids were cotransfected with a plasmid encoding VSV G protein into 293T cells, and the resulting viruses were used to infect 293T cells, from which cellular and virion RNAs were isolated and analyzed. The same approach was used to generate samples from SIVcpzEK505 and SIVgor2139. HIV-1 group O results were generated using a previously described vector (38) similar to H0 (Fig. 1B). Full-length unmodified SIVrcmGAB1 was propagated in GHOST-CCR2b cells. NL4-3 results from Fig. 1D are shown for comparison. Results are summarized from at least two independent experiments with sequencing data from the following total number of clones: 125 (MB897 cell), 125 (MB897 virus), 94 (EK505 cell), 94 (EK505 virus), 208 (HIV-1 group O cell), 125 (HIV-1 group O virus), 131 (SIVgor cell), 116 (SIVgor virus), 108 (SIVrcm cell), and 109 (SIVrcm virus). ***, P < 0.001; ns, not significant, chi-square test.

HIV-1 groups O and P were introduced into humans by cross-species transmission of SIVgor (8, 12, 30). However, like HIV-1 groups M and N, SIVgor is also thought to have arisen from SIVcpz in a cross-species transmission to gorilla (11, 12). Thus, we also investigated TSS usage in HIV-1 group O CMO2.5 and SIVgor2139 viruses (Fig. 4A) (11, 31). We found that both viruses produced multiple unspliced RNA species in 293T cells, including 3G and 1G RNA. However, 1G RNA was the most abundant species in both viruses. Like HIV-1 group M and SIVcpz, 3G RNA was depleted and 1G RNA was selectively packaged into virions (P < 0.001 in all cases, Fisher’s exact test). These results indicate that transcriptional heterogeneity and selective packaging of 1G RNA are highly conserved features of the HIV-1/SIVcpz/SIVgor lineage of primate lentiviruses.

Next, we analyzed SIVrcmGAB1, which also has three guanosines at the putative TSS (Fig. 4A). SIVrcm is thought to be one of the precursors of SIVcpz, although the origin of the SIVcpz LTR (which includes the TSS) is still uncertain (8, 13, 14). In GHOST-CCR2b cells, SIVrcmGAB1 produced almost exclusively 3G RNA (Fig. 4B), unlike any of the other viruses we analyzed. Furthermore, there was no significant difference in TSS distribution between cells and virions (P = 0.10, chi-square test). Thus, SIVrcmGAB1 did not exhibit selective packaging of 1G RNA or depletion of 3G RNA. These results clearly differ from those observed for SIVcpz and suggest that SIVcpz TSS heterogeneity and selective packaging of 1G RNA did not arise from SIVrcm.

TSS usage of HIV-2, SIVmac, SIVsmm, and SIVagm.

To determine whether other lineages of primate lentiviruses exhibit TSS heterogeneity and selective packaging of a particular RNA species, we next analyzed HIV-2 and SIVs related to HIV-2: SIVmac, SIVsmm, and SIVagm. We examined the TSS usage of unspliced RNA in infected cells and virions for HIV-2 ROD, HIV-2 ST, SIVmac239, and SIVsmmE543. HIV-1 sequences nearly always have GGG at the TSS. However, most HIV-2 sequences have CAG at the putative TSS, although some have CGG instead. Among the viruses we used, HIV-2 ROD has CGG, whereas HIV-2 ST, SIVmac239, and SIVsmmE543 have CAG at the putative TSS (Fig. 5A). In 293T and HUT/R5 cells, we found that HIV-2 ROD RNAs most often initiated at the 5′ guanosine in the CGG sequence (referred to as G1; Fig. 5B). These results are consistent with a previous report (32). However, we also observed a significant proportion of RNAs initiating from the 3′ guanosine (referred to as G2; Fig. 5B). HIV-2 ST, SIVmac239, and SIVsmmE543 RNAs in cells most often initiated at the adenosine in the CAG sequence (referred to as A1; Fig. 5B). However, HIV-2 ST also generated significant proportions of RNA from the cytosine (C-1) and guanosine (G2). Likewise, SIVmac239 also generated a significant proportion of RNA from the guanosine (G2). In contrast, SIVsmmE543 did not utilize alternative TSS to a significant extent. Thus, like most HIV-1-related viruses, most HIV-2-related viruses exhibited substantial transcriptional heterogeneity, although we identified exceptions in both cases.

FIG 5

TSS usage of HIV-2, SIVmac, SIVsmm, and SIVagm. (A) Alignment of sequences between the CATA/TATA box and near the TSS compared to HIV-2 ROD isolate. Green letters denote major TSS, whereas blue letters denote sequences differences relative to ROD. (B) Proportions of RNA species in cells infected with and viruses produced by a given virus. HIV-2 ROD and SIVagm results in 293T cells were generated using previously described constructs (39). Full-length unmodified HIV-2 ROD and HIV-2 ST were propagated in HUT/R5 cells, whereas SIVmac239 and SIVsmm were propagated in SUP-T1/R5 cells and BC7.6.1A10 cells, respectively. Results are summarized from at least two independent experiments with sequencing data from the following total number of clones: 163 (ROD cell, 293T), 166 (ROD virus, 293T), 105 (ROD cell, HUT/R5), 136 (ROD virus, HUT/R5), 105 (ST cell), 107 (ST virus), 134 (SIVmac cell), 113 (SIVmac virus), 113 (SIVsmm cell), 91 (SIVsmm virus), 150 (SIVagm cell), and 159 (SIVagm virus). ***, P < 0.001; ns, not significant, chi-square test.

We next determined the TSS of unspliced viral RNA in virions produced by HIV-2 and HIV-2-related SIVs. In all cases, we found that the predominant TSS in virions was G1 (HIV-2 ROD) or A1 (HIV-2 ST, SIVmac239, SIVsmmE543) (Fig. 5B). Furthermore, we found that TSS distributions were often significantly different between cells and virions. Specifically, for HIV-2 ROD, HIV-2 ST, and SIVmac239, G1 or A1 RNAs were enriched in virions (P = 0.01 [HIV-2 ROD 293T], P = 0.002 [HIV-2 ROD HUT/R5], P < 0.001 [HIV-2 ST, SIVmac239], Fisher’s exact test). However, TSS distributions did not significantly differ between cells and virions for SIVsmmE543 (P = 0.24, chi-square test), as A1 RNA comprised a higher proportion of cellular RNA for this virus. Thus, HIV-2 and some of the HIV-2-related viruses also selected certain RNA species for genome packaging based on TSS.

Lastly, we analyzed the TSS usage of SIVagm, which belongs to a lineage of SIV distinct from HIV-1- and HIV-2-related viruses. SIVagm sequences have CAG near the putative TSS (Fig. 5A). In 293T cells, SIVagmTAN1 generated mostly A1 RNA, similar to HIV-2-related viruses (Fig. 5B). However, small amounts of C-1 and G2 RNAs were also detected. In virions, A1 RNA was the predominant species, and it was slightly enriched in virions relative to cells (P = 0.003, Fisher’s exact test). Taken together, our results demonstrate that TSS heterogeneity and selective packaging of a particular RNA species are conserved features of primate lentiviruses. However, the extent of TSS heterogeneity and the degree of selection during packaging vary widely, even among closely related viruses.

DISCUSSION

Recent studies have demonstrated that HIV-1 generates multiple full-length viral RNAs via heterogenous TSS usage and selectively packages a specific full-length viral RNA species (1G RNA) (2–7), distinct from an earlier report (33). In this current study, we examined whether these features are conserved across primate lentiviruses by determining TSS of viral RNAs in cells and virions for 15 viruses. Based on phylogenetic analysis (Fig. 6), these viruses can be divided into 4 main groups: HIV-1-related viruses, HIV-2-related viruses, SIVagm, and SIVrcm. HIV-1-related viruses, including HIV-1, SIVcpz, and SIVgor, usually have GGG at the TSS (Fig. 6). We found that these viruses exhibited TSS heterogeneity and selectively packaged 1G RNA into virus particles. Nonetheless, we identified significant differences in TSS usage among HIV-1-related viruses, even within the same group or subtype. HIV-2-related viruses, including HIV-2, SIVmac, and SIVsmm, usually have CAG at the TSS, although HIV-2 ROD has CGG (Fig. 6). Like HIV-1, most of these viruses exhibited TSS heterogeneity and selective packaging of a particular RNA species. SIVagm has CAG at the TSS, like most HIV-2-related viruses (Fig. 6). SIVagm also displayed TSS heterogeneity and selective packaging of a specific RNA species. In contrast, SIVrcm has variable sequences near the TSS. Of the four SIVrcm sequences available (http://www.hiv.lanl.gov), three have CAG at the TSS (like HIV-2 and SIVagm), whereas one has GGG (like HIV-1) (Fig. 6). We analyzed SIVrcmGAB1, which has GGG at the TSS, because reagents were not available for the other strains. We found that SIVrcmGAB1 generated almost exclusively 3G RNA, with no significant difference in the distribution of RNA species between cells and virions. Collectively, our results demonstrate that heterogeneity in TSS usage and selective packaging of a particular RNA species are common features in primate lentiviruses and are not unique to HIV-1-related viruses or viruses with GGG TSS sequences. However, we did identify two exceptions: SIVrcm and SIVsmm used multiple TSS but did not select a specific RNA species for genome packaging.

FIG 6

Evolutionary relationship of HIV/SIV LTR sequences. Phylogenetic trees based on LTR (U3-R-U5) sequences were constructed using the neighbor-joining method. Blue labels indicate viruses that were analyzed in this report. Sequences near the putative TSS (based on alignment with the NL4-3 TSS) are indicated below the label for each group. Green underlined nucleotides indicate major TSS that were identified in at least one virus from the indicated group. Major TSS were defined as sites used by ≥20% of the sequences from either the cellular RNA sample or the virion RNA sample. Scale bar: number of substitutions/site.

One surprising observation from our study is that closely related viruses can have striking differences in TSS usage. For example, among subtype B viruses, NL4-3 and TRJO produced mostly 3G and 1G RNAs, whereas CH058 and CH106 produced mostly 1G and 5C RNAs. Upon analyzing the promoter sequences of these viruses, we found that, compared to NL4-3, CH058 and CH106 have a 1-nt deletion between the TATA box and the TSS. In cellular genes that utilize TATA box-dependent transcription, TSS are determined by nt sequence as well as distance from the TATA box (34–36). Thus, it is possible that the 1-nt deletion altered the TSS usage of CH058 and CH106. However, not all of the differences in HIV-1 TSS usage can be explained by the distance between the TATA box and TSS. For example, the distances between the TATA box and TSS are identical among NL4-3, MJ4, and CH185, yet 1G RNA dominates the HIV-1 RNA species in cells infected with MJ4 and CH185, whereas 3G RNA dominates in cells infected with NL4-3. Further studies will be needed to determine the mechanism(s) that cause the distinct patterns of TSS usage in these viruses. Nonetheless, despite differences in TSS heterogeneity, all HIV-1-related viruses predominantly packaged 1G RNA into particles. In every case, 1G RNA was enriched in virions relative to cells, indicating that it was preferentially packaged over 3G and C5 RNAs. However, the degree to which 1G RNA was enriched in virions varied widely among viruses, mainly due to differences in the intracellular proportions of 1G RNA. Therefore, our findings suggest that selective packaging of 1G RNA is highly conserved among HIV-1-related viruses. The distinct usage of TSS in closely related viruses can also be observed in HIV-2-related viruses. Among the viruses we examined, HIV-2 ST exhibited the greatest degree of TSS heterogeneity and selective packaging, while SIVsmmE543 exhibited the least. Among HIV-2 related viruses, there were sequence differences in the region between the CATA/TATA box and the TSS (Fig. 5A), which may have led to differences in TSS usage.

Previous studies have shown that 1G RNA and 3G RNA fold into distinct structures: 1G RNA, but not 3G RNA, folds into structure(s) that expose the RNA dimerization signal and Gag-binding sites, facilitating genome packaging (3, 7). It has been proposed that the 5′ context of the HIV-1 RNA also affects other RNA functions. This model, based on NL4-3 (subtype B) and MAL (subtype A), proposed that the 5′ cap is sequestered in 1G RNA, thereby preventing its translation, whereas the 5′ cap is exposed in 3G RNA, thereby promoting its translation (3, 5, 6). In our study, we found that different HIV-1 strains had very different ratios of 3G and 1G RNA; in some viruses such as CH058, CH106, MJ4, and CH185, very little 3G RNA was generated (≤10%). Thus, it is unclear how these viruses produce sufficient RNA to serve as templates for Gag/Gag-Pol translation. One possibility is that in these viruses, RNA species other than 3G are preferentially translated. Alternatively, it is possible the 5′ context does not affect translation in these viruses. Further studies are needed to distinguish between these possibilities.

Lastly, our findings provide insights into the evolutionary origin of HIV-1 TSS heterogeneity and selective packaging of 1G RNA. In SIVcpz, the progenitor of HIV-1 groups M and N, we observed significant TSS heterogeneity, with mostly 3G and some 1G RNA in cells, and selective packaging of 1G RNA into virions. These results closely resemble those obtained for some HIV-1 strains (NL4-3 and TRJO), indicating that TSS heterogeneity and selective packaging of 1G RNA likely preceded the emergence of HIV-1. Thus, these features may have originated in SIVcpz or earlier SIV precursors. SIVcpz arose from recombination between at least two viruses: SIVrcm and SIVmus/mon/gsn (13). We found that SIVrcmGAB1 has GGG at the TSS like SIVcpz but generated and packaged almost exclusively 3G RNA. Thus, SIVrcm lacked TSS heterogeneity and selective packaging of 1G RNA, indicating that these features may have evolved directly in SIVcpz or may have arisen from a different SIV lineage.

MATERIALS AND METHODS

Molecular clones and viruses.

Two types of viruses were used in this study: unmodified replication-competent viruses and modified viral constructs. All modified viral constructs had intact LTRs, 5′ UTR, and functional gag-pol; thus, transcription regulation and genome packaging were carried out by authentic elements from the same virus. The previously described NL4-3-based construct H0 (20) contains all the cis-acting elements required for HIV-1 replication and expresses functional Gag/Gag-Pol, Tat, and Rev. In addition, an hsa gene, internal ribosomal entry site (IRES) from encephalomyocarditis virus, and a mutated green fluorescent protein (gfp) gene were inserted in the nef gene (20); for simplicity, IRES-gfp is not described in the text or Fig. 1. Results of HIV-1 studies from 293T cells were generated using previously described constructs derived from other subtypes/groups but having H0-like structures. Specifically, MJ4 results were from a 293T cell line infected with CH0 and CT6 (37), and HIV-1 group O results were from cells infected with OHIG (38). CT6 has the same general structure as CH0 except CH0 expresses hsa whereas CT6 expresses a Thy1.2 gene (37). The previously described construct HIV2-H0G (39) is based on the ROD12 isolate of HIV-2 (40) and has a frameshift mutation in vpr, an inactivating deletion in env, and an insertion of a fragment containing hsa-IRES-inactivated gfp in the nef gene. The previously described SIVagm construct pTan-H0G is based on SIVagm molecular clone Tan-1 which has inactivating mutations in vpr and env, and an hsa-IRES-inactivated gfp fragment inserted in the nef gene (39).

The full-length infectious molecular clones pNL4-3, pTRJO, pCH058, pCH106, pMJ4 (21–24, 26), and pHIV-2/ST were used for spreading infections and were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program (NIH ARP). The infectious molecular clone pROD12 is a kind gift from Keith Peden (FDA), whereas pCH185 is a kind gift from Christina Ochsenbauer and John Kappes (University of Alabama). SupT1/R5 cells stably infected with SIVmac239*, which contains a Q734stop mutation in the env gene and BC7.6.1A10 cells stably infected with SIVsmm-E543-3 were generous gifts from Julian Bess (AIDS and Cancer Virus Program, Frederick National Laboratory). Infectious virus stock for SIVrcm-GAB1 was obtained from NIH ARP.

Molecular clones for SIVcpz-MB897, SIVcpz-EK505, and SIVgor-2139 were generous gifts from Beatrice Hahn. As these plasmids frequently acquired deletions or insertions during propagation, we split each molecular clone into two plasmids. SIVcpz-MB897-3′ was created by deleting sequences between the MluI restriction site upstream of the 5′ LTR and the NheI site in env, whereas SIVcpz-MB897-5′ was created by deleting sequences between the StuI restriction site in env and the SnaBI site downstream of the 3′ LTR. The SIVcpz-MB897-5′ and SIVcpz-MB897-3′ plasmids have a 458-bp overlap in viral sequences to facilitate homologous recombination during transfection. In addition, the env gene in SIVcpz-MB897-3′ was inactivated by introducing a frameshift mutation using Klenow fill-in of an EcoRI site. To generate SIVcpz-EK505-3′, a region between the MluI restriction site upstream of the 5′ LTR and the BstZ17I site in pol was deleted; SIVcpz-EK505-5′ was generated by deleting sequences between the SbfI restriction site in vif and the NsiI site downstream of the 3′ LTR. The SIVcpz-EK505-5′ and SIVcpz-EK505-3′ plasmids contain a 2.5-kb overlap in viral sequences. In addition, a portion of the env gene in SIVcpz-EK505-3′ was deleted by digestion with ScaI and re-ligation. The SIVgor-2139 molecular clone we propagated contained a 1.3-kb bacterial sequence inserted in the env gene causing its inactivation; additionally, the TSS sequence was GTGG, which is atypical for SIVgor isolates. We generated SIVgor2139-3′ by deleting a region between the ZraI restriction site upstream of the 5′ LTR and the HpaI site in pol, whereas SIVgor2139-5′ was created by deleting sequences between SphI restriction sites in env and downstream of the 3′ LTR. The resulting SIVgor2139-5′ and SIVgor2139-3′ contain a 3.6-kb overlap in viral sequences. To generate SIVgor plasmids with the more representative sequence of TGGG at the TSS, 3′ and 5′ LTR fragments were synthesized (IDT) and used to replace the original sequences between the BbvCI and NotI sites in SIVgor2139-3′ or the MluI and StuI sites in SIVgor2139-5′, respectively. All cloning was performed using the NEBuilder HiFi DNA Assembly kit (NEB) or standard molecular cloning techniques. The viral genome portions in SIVcpz and SIVgor plasmids were sequenced in entirety to ensure sequence integrity.

Cell culture, infections, and flow cytometry.

Human embryonic kidney 293T cells were obtained from American Type Culture Collection. CEM-SS (a derivative of human T cell line CEM), GHOST-X4, and GHOST-CCR2b cells were obtained from the NIH AIDS Reagent Program. GHOST are human osteosarcoma cells derived from HOS that stably express CD4 and either CXCR4 or CCR2b, respectively. HUT/R5 is a derivative of human T-cell line HUT78 that stably expresses CCR5 (41). Both SUP-T1/R5 and BC7.6.1A10 are human T cell lines derived from SUP-T1. GHOST and 293T cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin. CEM-SS, HUT/R5, SUP-T1/R5, and BC7.6.1A10 cells were maintained in Roswell Park Memorial Institute (RPMI-1640) medium supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. All cells were maintained in humidified 37°C incubators with 5% CO2.

To generate VSV-G-pseudotyped viruses, 293T cells were transfected using TransIT-LT1 (Mirus Bio) with viral constructs and the helper plasmids pSynGagPol and pHCMV-G that express codon-optimized HIV-1 Gag/Gag-Pol and VSV-G, respectively (42, 43). These viruses were used to infect fresh 293T cells to generate infected cell pools. SIVcpz and SIVgor plasmids were linearized prior to transfection to facilitate homologous DNA recombination. For this purpose, SIVcpzMB897-5′ and -3′ were digested with NheI, SIVcpz-EK505-5′ and -3′ were digested with ApaI and AccI, respectively, and SIVgor-TGGG-5′ and -3′ were digested with SphI and KpnI, respectively. Supernatants were harvested 24 to 48 h after transfection, clarified through a 0.45-μm filter (Millex), and in some cases concentrated 10- to15-fold by ultracentrifugation at 25,000 × g for 90 min through a 20% sucrose cushion. To generate pools of infected cells, more than 1 million 293T cells were infected with viruses at MOIs of 0.1–1.9 and analyzed 3 days postinfection by flow cytometry to determine the proportion of infected cells. Pools of infected cells consisted of at least 95,000 independent infection events.

To prepare virus stocks of NL4-3, TRJO, CH058, CH106, MJ4, CH185, ROD, and ST for spreading infections, 293T cells were transfected in 6-well plates using 3 μg plasmid/well and TransIT-LT1 (Mirus Bio). Virus supernatants were collected 48 h posttransfection and clarified through 0.45-μm filters. HIV-1 virus stocks were quantified by p24 ELISA (XpressBio), whereas HIV-2 virus stocks (ROD, ST) were quantified by p27 ELISA (XpressBio). HUT/R5 or CEM-SS cells were infected with 100 μL of undiluted virus, using 1 million cells in 3 mL volume. Infections were performed in duplicate using independent virus stocks. Starting 3 days postinfection, cells and virus supernatants were collected every other day. Briefly, 1.5 mL of culture was removed from each infection and centrifuged at 16,000 × g for 10 min. Viral supernatants and cell pellets (resuspended in 150 μL RNAlater) were frozen at −80°C. Lastly, 1.5 mL of fresh medium was added to the cultures. GHOST-X4 and GHOST-CCR2b cells were infected with NL4-3 or SIVrcmGAB1, respectively, as follows: 30,000 cells/well were seeded in 48-well plates using 200 μL medium/well and infected the next day with 50 μL of undiluted virus by spinoculation at 1200 × g for 1 h at room temperature in the presence of 20 μg/mL Polybrene. At 24 h postinfection, the cells were split into 6-well plates; cells and viral supernatants were collected 3 and 6 days postinfection. Virus replication was monitored by p24 (NL4-3, TRJO, CH058, CH106, MJ4, and CH185) or p27 (ROD, ST, and SIVrcm) ELISA.

To perform flow cytometry analysis of cells infected with HIV-1 or HIV-2 single-round viruses, cells were collected, washed twice with Dulbecco’s phosphate-buffered saline (PBS) containing 2.5% fetal bovine serum (FBS), then stained with 0.4 μg/mL phycoerythrin- or allophycocyanin-conjugated anti-HSA antibodies (BioLegend) for 30 min, and washed two additional times with PBS containing 2.5% FBS. Infections by SIVcpz and SIVgor were quantified using internal p24 staining; briefly, infected 293T cells were fixed for 30 min with 2% formaldehyde in PBS containing 2.5% FBS, washed twice in PBS with 2.5% FBS, stained with RD1-conjugated anti-p24 antibodies (KC57-RD1, Beckman Coulter) in permeabilization buffer (PBS with 2.5% FBS and 0.5% Saponin), washed twice in PBS with 2.5% FBS, and fixed in 2% formaldehyde in PBS. Flow cytometry was performed using an LSR II system (BD Biosciences) and data were analyzed using FlowJo software (TreeStar, LLC).

RNA isolation and 5′ RACE.

To isolate RNA from viral particles, supernatant from infected cells was harvested, clarified through a 0.45-μm filter (Millex), and concentrated by centrifugation at 25,000 × g for 90 min through a 20% sucrose cushion, and total RNA was isolated using a Mini Viral RNA kit (Qiagen) according to the manufacturer’s protocol. Total cellular RNA from infected cells was isolated using the RNeasy Plus minikit (Qiagen) according to the manufacturer’s protocol. Isolated RNA was converted into cDNA using the SMARTer RACE 5′/3′ kit (TaKaRa); RNA was denatured immediately prior to cDNA synthesis as described in the manufacturer’s protocol. All primers for cDNA preparations targeted the gag gene to ensure that only full-length RNA is analyzed. Next, PCR was performed using a 5′ universal primer from the SMARTer RACE 5′/3′ kit and a 3′ virus sequence-specific primer. Primers used in this study are shown in Table 1. The products of 5′ RACE were gel-purified and cloned into the pRACE plasmid (TaKaRa). For the 5′ RACE reactions, >1 × 106 copies of full-length viral RNA were typically used for cDNA synthesis; subsequently, >1 × 106 copies of viral cDNA were used for PCR, and >5 × 109 copies of the PCR product were used for cloning. Most experiments yielded >1,000 transformed bacterial colonies. On average, >100 clones per sample from 2 to 5 independent experiments were analyzed. The 5′ UTR sequences were analyzed manually in CloneManager v2 (Sci Ed Software LLC) or with a custom R script.

TABLE 1

RT and PCR Primers used for 5′ RACE

Virus	RT primer	Reverse PCR primera
NL4-3	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT - TCGTTCTAGCTCCCTGCTTG
CH058	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT-TCGTTCTAACTCCCTGCTTG
CH106	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT-TCGTTCTAGCTCCCTGCTTG
TRJO	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT-TCGTTCTAGCTCCCTGCTTG
MJ4	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT-AGCTCCCTGCTTGCCCATAC
CH185	GGTGGCTCCTTCTGATAATG	GATTACGCCAAGCTT-AGCTCCCTGCTTGCCCATAC
CMO2.5	GGGAGATGGCACTATGTACC	GATTACGCCAAGCTT-CACACCAGAGCACTGCTATTGTGTTC
HIV-2 ROD	TACCCAGGCATTTAGGGTTC	GATTACGCCAAGCTT-TCTGTCCAATTCATTCGCTGCCCACAC
HIV-2 ST	TACCCAAGCATTTAGGGTTC	GATTACGCCAAGCTT-TCTGTCCAATTCATTCGCTGCCCACAC
SIVgor2139	GAGATCGCAAGCCATCTGAC	GATTACGCCAAGCTT-CGTTCCCATGCATCCAATCG
SIVcpzMB897	TCTTGCCCACATTTCCA	GATTACGCCAAGCTT-AGCTCCCTGCTTGCCCATAC
SIVcpzEK505	TCTTGCCCACATTTCCA	GATTACGCCAAGCTT-AGCTCCCTGCTTGCCCATAC
SIVmac239	TACCCAGGCATTTAATGTTC	GATTACGCCAAGCTT-CTATCTAATTCATTTGCTGCCC
SIVsmmE543	TACCCAAGCATTTAATGTTC	GATTACGCCAAGCTT-TCTGTCCAATTCATTTGCTGC
SIVrcmGAB1	GACCCAGGCATTTAGAGTCC	GATTACGCCAAGCTT-CCATACCAAATGCTTCAGCATG
SIVagmTAN1	GCGGGCTCAATACTTCTATG	GATTACGCCAAGCTT-CCAAATTCCTCCCTGACAGTGCCGAGTG

5' extension with InFusion sequence for cloning into pRACE plasmid is in italic. The forward PCR primer was provided by the 5′ RACE kit.

Alignments, phylogenetic trees, and statistical analysis.

Most LTR (U3-R-U5) sequences were obtained from the Los Alamos HIV Sequence Database (http://www.hiv.lanl.gov). Several others were obtained from GenBank under the following accession numbers: MJ4 (HIV-1 group M, subtype C), AF321523.1; CMO2.5 (HIV-1 group O), AY623602; and RBF168 (HIV-1 group P), GU111555. LTR sequences were imported into Geneious Prime v2021.2.2 (Biomatters) and aligned with MUSCLE v3.8.425 (44). Neighbor-joining phylogenetic trees were constructed in Geneious Prime using the Tamura-Nei genetic distance model.

Chi-square tests were used to determine whether the distributions of TSS were different between cellular RNA and virion RNA. Fisher’s exact tests were used to determine whether the proportions of a particular RNA species were different between cellular RNA and virion RNA. Two-sided tests were used in all cases. Both chi-square and Fisher’s exact tests were performed in GraphPad Prism v9.2.0 (GraphPad Software, LLC).

Data availability.

All data required to support conclusions are presented in the article.