In Vivo SELEX of Single-Stranded Domains in the HIV-1 Leader RNA

ABSTRACT

The 5′ untranslated leader region of the human immunodeficiency virus type 1 (HIV-1) RNA genome is a strongly conserved sequence that encodes several regulatory motifs important for viral replication. Most of these motifs are exposed as hairpin structures, including the dimerization initiation signal (DIS), the major splice donor site (SD), and the packaging signal (Ψ), which are connected by short single-stranded regions. Mutational analysis revealed many functions of these hairpins, but only a few studies have focused on the single-stranded purine-rich sequences. Using the in vivo SELEX (systematic evolution of ligands by exponential enrichment) approach, we probed the sequence space in these regions that is compatible with efficient HIV-1 replication and analyzed the impact on the RNA secondary structure of the leader RNA. Our results show a strong sequence requirement for the DIS hairpin flanking regions. We postulate that these sequences are important for the binding of specific protein factors that support leader RNA-mediated functions. The sequence between the SD and Ψ hairpins seems to have a less prominent role, despite the strong conservation of the stretch of 5 A residues in natural isolates. We hypothesize that this may reflect the subtle evolutionary pressure on HIV-1 to acquire an A-rich RNA genome. In silico analyses indicate that sequences are avoided in all 3 single-stranded domains that affect the local or overall leader RNA folding.

IMPORTANCEin vivo

INTRODUCTION

The RNA genome of human immunodeficiency virus type 1 (HIV-1) contains 9 genes and 5′ and 3′ untranslated regions (UTRs) (Fig. 1A). The untranslated leader region is a highly conserved part of the HIV-1 genome that encodes several replication signals. These signals are often exposed as hairpins connected by short single-stranded regions (Fig. 1B). The hairpins have important regulatory functions in the HIV-1 replication cycle (1): the trans-acting response (TAR) element, the 5′ polyadenylation signal (polyA), the primer binding site (PBS), the dimerization initiation site (DIS), the splice donor (SD), and the packaging signal (Ψ). Other important regulatory elements include the primer activation signal (PAS) and the U5-AUG duplex, a long-distance base-pairing interaction that includes the start codon of the Gag open reading frame (2). TAR is essential for transcriptional activation by binding the viral Tat protein and the cellular positive transcriptional elongation factor (pTEFb) (3–6). Mutation of TAR has been shown to affect RNA dimerization and packaging (7–9), although this is likely due to indirect RNA misfolding effects (10, 11). The polyA hairpin comprises the polyadenylation signal (AAUAAA), which is inactive because it is inaccessible as part of the hairpin structure. A copy of this signal on the 3′ end of the viral RNA is activated by different mechanisms (12–15). The tRNA^lys3 primer binds to the PBS signal where reverse transcription is initiated (16, 17). The tRNA primer also interacts with the PAS domain to initiate fast and efficient reverse transcription (18, 19). The TAR-polyA cDNA product of the first phase of reverse transcription binds to the 3′ UTR during the first strand transfer, enabling minus-strand DNA synthesis to continue (20).

FIG 1.

The HIV-1 genome. (A) The untranslated leader of the HIV-1 RNA genome with enlarged the 5′ RNA leader domain (nt 1 to 368) that contains several important regulatory signals: TAR element, polyA, PAS, PBS, DIS, SD site, packaging signal (Ψ), and translational start codon of gag. (B) The leader RNA can fold the BMH and LDI conformations. The positions of the single-stranded regions 1, 2, and 3 are boxed. (C) HIV-1 libraries were generated in which regions 1, 2, and 3 were randomized. Population-based sequencing of these libraries was used to demonstrate that all 4 nucleotides are present at the intended nucleotide positions.

The DIS hairpin harbors a 6-nucleotide (nt) palindrome in the loop, via which kissing-loop RNA dimerization occurs by noncovalent base-pairing. The palindrome plays a major role in in vivo RNA partner selection (21, 22) and the DIS hairpin is the major site for in vitro RNA dimerization (23–26), but some studies imply that other dimerization sites must be present (reviewed in reference 26). For example, the genomic RNA dimers are much more stable in mature virion particles than in immature virions (27–29), and viruses with a deleted or altered DIS hairpin are able to replicate in certain cell types (30–32). The major SD site, which is located in the loop of the SD hairpin, is used for the production of all spliced HIV-1 RNAs (33). The activity of the SD may be modulated by the stability of the SD hairpin (34). The Ψ hairpin is important, but not sufficient, for packaging of the RNA dimer into viral particles. The minimal RNA signal that is sufficient for genomic RNA dimerization and packaging includes up- and downstream sequences (from the PBS up to the AUG start codon of Gag) (35–37).

For several retroviruses it has been suggested that the 5′ leader RNA acts as a riboswitch by alternating between different conformations, thus regulating multiple processes like genomic RNA dimerization, packaging, and translation (38–45). The 5′ leader of HIV-1 is thought to act as a riboswitch by folding either the branched multiple hairpin (BMH) or the long-distance interaction (LDI) conformation (Fig. 1B). The DIS hairpin is exposed in the BMH conformation, thereby enabling kissing-loop RNA dimerization. The alternative LDI conformation is not dimerization competent, as the DIS hairpin is not exposed (Fig. 1B) (46–49). Slightly different BMH-like RNA secondary-structure models have recently been published (1, 36, 50, 51). Also, an alternative dimerization-incompetent LDI conformation in which the loop of the DIS palindrome base-pairs with a U5 sequence was suggested (36, 52).

In the BMH conformation, the exposed hairpins are connected by short single-stranded regions (Fig. 1B). Through mutational analysis of the hairpins, much is known about their role in the HIV-1 replication cycle, but not many studies have focused on the single-stranded purine-rich regions flanking the DIS, SD, and Ψ hairpins: region 1 (²³⁷GCAGGA²⁴²), region 2 (²⁷⁸GGGA²⁸¹), and region 3 (³⁰¹AAAAA³⁰⁵). We probed the importance of these three regions by the in vivo SELEX (systematic evolution of ligands by exponential enrichment) approach (53, 54). DNA libraries of the full-length HIV-1 genome were designed with short randomized nucleotide stretches (Fig. 1C). These HIV-1 libraries were transfected into HIV-1-susceptible SupT1 T cells, and the viral progeny was passaged to select for the best replicators. Additional changes may be selected in the viral genome during the in vivo SELEX procedure, as HIV-1 replication triggers spontaneous evolution driven by the error-prone reverse transcriptase enzyme. These studies allowed us to probe the sequence space in regions 1, 2, and 3 that is compatible with efficient virus replication. We analyzed the impact of the selected and nonselected sequences on the RNA secondary structure in the context of the proposed LDI-BMH equilibrium.

MATERIALS AND METHODS

Cell culture.

SupT1, PM1, and 174×CEM cells were grown in advanced RPMI 1640 medium (Gibco) supplemented with 1% (vol/vol) heat-inactivated fetal bovine serum (FBS; Gibco), 2 mM l-glutamine (Gibco), 15 U/ml of penicillin, and 15 μg/ml of streptomycin at 37°C and 5% CO₂. human embryonic kidney (HEK) 293T cells were grown in Dulbecco's modified Eagle medium (DMEM; Gibco) supplemented with 10% (vol/vol) heat-inactivated FBS, 1× minimum essential medium nonessential amino acids (MEM NEAA; Gibco), 40 U/ml of penicillin, and 40 μg/ml of streptomycin at 37°C and 5% CO₂. CD4⁺ peripheral blood mononuclear cells (PBMCs) were obtained by isolation of PBMCs from buffy coats using a Ficoll density gradient, followed by the depletion of CD8⁺ lymphocytes with CD8⁺ magnetic beads (Invitrogen). CD4⁺ PBMCs were cultured in RPMI medium supplemented with 10% (vol/vol) heat-inactivated FBS, 100 U/ml of interleukin 2 (IL-2), 15 U/ml of penicillin, and 15 μg/ml of streptomycin.

Construction of HIV-1 SELEX libraries.

HIV-1 libraries were generated by PCR using oligonucleotide primers with a short randomized sequence motif. The generated PCR products were digested and inserted into a stuffer plasmid to prevent wild-type input sequences in the library plasmids. The pLAI-BX stuffer was generated by inserting the XbaI-BssHII fragment of pCMV-rtTA-F86Y-A209T (55) into the infectious pLAI molecular clone. The pLAI-BC stuffer was generated by replacing the BssHII-ClaI fragment of pLAI with the fragment of pSEAP2 control (Clontech). The region 1 library was created using forward primer pKp5′seq (56) and reverse primer NB5 (TGCCGTGCGCGCTTCAGCAAGCCGAGNNNNNNGTCGAGAGAGCTCCTCTGGT; randomized nucleotides are underlined). The PCR products were isolated from agarose gel, and the XbaI-BssHII fragment was inserted into the pLAI-BX stuffer. Region 2 and 3 libraries were created using, respectively, primers NB6 (GCTGAAGCGCGCACGGCAAGAGGCGAGNNNNGGCGACTGGTGAGTACGCCAA) and NB8 (GCTGAAGCGCGCACGGCAAGAGGCGAGGGGAGGCGACTGGTGAGTACGCCNNNNNTTTTGACTAGCGGAGGCTAGAAGGAGAGAG) as forward primers, combined with AD-Gag (57) as a reverse primer. After purification from agarose gel, the BssHII-ClaI fragment was inserted into the pLAI-BC stuffer. The ligation products of each library were transformed into DH5α-E bacteria (Invitrogen) by electroporation, and the DNA was isolated using the QiaFilter plasmid maxikit (Qiagen) according to the manufacturer's protocol. To verify the variation in each library, the insert of each library was sequenced, which yielded multiple overlapping peaks at the randomized region surrounded by clear single nucleotide peaks. In addition, random clones were isolated and sequenced.

In vivo SELEX.

The HIV-1 DNA libraries were transfected into SupT1 T cells by electroporation. For each library, 15 × 10⁶ cells were washed in transfection medium (advanced RPMI medium supplemented with 20% heat-inactivated FBS and 1.25% dimethyl sulfoxide [DMSO; Sigma-Aldrich]) and resuspended in 750 μl of transfection medium. Cells were divided over three 4-mm cuvettes, and each aliquot was mixed with 40 μg of DNA, incubated on ice for 10 min, and electroporated at 250 V and 975 μF. After 10 min of incubation at room temperature, the 3 individual electroporation products were pooled in 24 ml of culture medium and divided into 24 independent cultures, each supplemented with 0.5 × 10⁵ fresh SupT1 T cells. After the appearance of widespread syncytia, cell-free culture supernatant was passaged onto fresh SupT1 cells. Virus passaging was continued until a homogenous virus population was present, as verified by population sequencing. Highly infected cells were pelleted and washed once with phosphate-buffered saline at 1,500 × g for 4 min. The cells were resuspended in 10 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 0.5% Tween 20, and 200 μg/ml of proteinase K at 56°C for 60 min. Proteinase K was inactivated by a 10-min incubation step at 95°C. The proviral DNA was amplified by PCR with primers LAI5′X (57) and AD-Gag. The PCR product was sequenced using the BigDye Terminator cycle sequencing kit (Applied Biosystems).

Construction of HIV-1 variants.

Different in vivo-selected region 1, 2, and 3 sequences were introduced into pLAI to test their replication capacities. The isolated proviral DNA was PCR amplified with primers AD-Gag and U3XbaNotAD (56). The PCR product was isolated from agarose gel and digested with XbaI and ClaI for region 1 constructs and with BssHII and ClaI for region 2 and 3 constructs.

Virus production.

HEK293T cells were transfected using the calcium phosphate method in a 24-well format (58). Cells were transfected with 1 μg of DNA. The virus-containing culture supernatant was harvested 48 h after transfection, and the CA-p24 level was measured by enzyme-linked immunosorbent assay (ELISA).

CA-p24 ELISA.

The culture supernatant was heat inactivated for 30 min at 56°C in the presence of 0.05% Empigen-BB (Calbiochem, La Jolla, CA). The CA-p24 concentration was determined by a twin-site ELISA with D7320 (Biochrom, Berlin, Germany) as the capture antibody. Alkaline phosphatase-conjugated anti-p24 monoclonal antibody (EH12-AP) was used as the detection antibody. Detection was done with a Lumiphos Plus system (Lumigen, MI) in a LUMIstar Galaxy (BMG Labtechnologies, Offenburg, Germany) luminescence reader. Recombinant CA-p24 expressed in a baculovirus system was used as the reference standard.

Virus replication.

SupT1 T cells (1 × 10⁶ per ml) were infected with virus equivalent to 1 ng of CA-p24. Replication capacity of the viruses was assessed via measuring the CA-p24 levels in the culture medium every 1 to 2 days after infection.

Virus competition experiments, proviral DNA isolation, and sequencing.

To test the replication capacity of the different selected virus variants versus the wild-type (wt) LAI virus, virus competition experiments were conducted in SupT1 T cells. The competition experiments were started in duplicate by transfecting an equimolar DNA mixture of the wt and mutant constructs into SupT1 T cells, as described previously (59). All DNA mixes were verified by sequencing. The virus mixture was passaged onto fresh cells when massive syncytia were present. The 5′ leader sequence of the integrated proviral DNA was analyzed to determine the ratio of wt to variant virus over time. The sequences were aligned and the wt/variant virus ratio was estimated using the CodonCode Aligner software. The competition was stopped after 4 weeks or earlier, when a homogenous virus mixture was present.

RNA secondary-structure prediction.

Computer-assisted RNA secondary-structure predictions were performed using the RNAstructure algorithm, version 5.3, offered by the Mathews Lab (http://rna.urmc.rochester.edu/) (60). Folding of the complete HIV-1 5′ leader RNA (nt 1 to 368) and the DIS (nt 237 to 281), SD (nt 278 to 305), and Ψ hairpins (nt 301 to 333) was performed using standard settings (37°C). The thermodynamically most stable leader structure was chosen. The structure was listed as BMH if the polyA and DIS hairpins were formed, whereas LDI was listed when the polyA and DIS hairpins were absent.

RESULTS

Experimental design.

To study the role of the single-stranded regions 1, 2, and 3 in the BMH conformation of the HIV-1 5′ leader RNA (Fig. 1A and B), we created full-length HIV-1 DNA genome libraries in which one of these three regions was randomized (Fig. 1C). The DNA libraries were made in the infectious molecular clone of the LAI isolate, a primary CXCR4-using virus isolate (61). Theoretically, 4⁶ (region 1), 4⁴ (region 2), and 4⁵ (region 3) virus variants should be present in the respective libraries to cover the complete sequence spectrum. To ensure that we started with a large and diverse collection, the bacterial transformation of the sequence-randomized LAI plasmids was scaled up and a high number of transformants was pooled for DNA isolation and purification. The quality of each plasmid library was checked before transfection into SupT1 T cells. To that end, the proviral DNA library was assessed by population sequencing, showing the randomized segments (Fig. 1C). Furthermore, multiple individual clones per library were sequenced to confirm diversity and the absence of a predominant individual sequence (see Table S1 in the supplemental material). Each HIV-1 DNA library was transfected into SupT1 T cells, and the culture was split immediately to yield 24 independent cell cultures. This was done as bottlenecks during the SELEX experiment (e.g., transfection efficiency and passaging of virus supernatant) could affect the outcome of this selection experiment. This setup allowed us to select multiple sequences that represent efficiently replicating viruses. The transfected T cells were cultured and split 1:5 twice a week until the presence of HIV-induced syncytia became apparent, at which time the cell-free culture supernatant was passaged onto fresh SupT1 T cells. Virus passaging was continued for 3 to 24 weeks, depending on the randomized region and the particular cell culture, until a homogenous virus population was present as determined by population sequencing of the individual cell cultures.

Nucleotides upstream of the DIS hairpin (region 1) undergo strong selection pressure.

Region 1 consists of six nucleotides (²³⁷GCAGGA²⁴²) immediately upstream of the DIS hairpin (Fig. 1B). The in vivo selection of the HIV-1 library with a randomized region 1 lasted, on average, 3 to 5 weeks, with some exceptions of up to 11 weeks, when outgrowth of a single virus variant was apparent based on the population sequence. Of each culture, the randomized genome segment was PCR amplified from HIV-1 DNA integrated in the genomes of the infected cells and subsequently sequenced. In 4 out of 24 cultures, the wt GCAGGA sequence was selected. Strikingly, the variant motif with a single point mutation, GCAGUA (nucleotide differing from the wt underlined), was selected more frequently (11/24). Another point mutant (GCAAGA) was selected twice, and seven other sequence variants (a point mutant, four double mutants, and two triple mutants) were selected only once (Table 1).

TABLE 1.

In vivo SELEX of region 1

Sequencea	Frequencyb	ΔGDISc	ΔGLDIc	ΔGBMHc	ΔΔGLDI-BMHd
GCAGUA	11	−12.9	−141.9	−133.8	−8.1
GCAGGA (wt)	4	−12.9	−139.8	−134.2	−5.6
GCAAGA	2	−12.9	−138.3	−134.6	−3.7
GCAGAA	1	−12.9	−138.3	−135.0	−3.3
GCCAUA	1	−13.6	−141.1	−135.3	−5.8
GCCGGCe	1	−16.4	−142.3	−141.8	−0.5
GCGGAA	1	−12.9	−142.3	−134.8	−7.5
GCGUUA	1	−12.9	−140.3	−137.3	−3.0
GCUGGCe	1	−17.0	−142.3	−139.9	−2.4
GUAGUA	1	−12.9	−141.7	−135.6	−6.1

^aNon-wt nucleotides are in bold and underlined.

^bObserved frequency in 24 cultures.

^cΔG, in kcal/mol, of the DIS hairpin and the 5′ RNA leader in LDI and BMH conformation.

^dΔΔG in kcal/mol as a measure of the LDI-BMH equilibrium.

^eExtends the DIS hairpin.

The strong selection pressure operating at all 6 nucleotide positions is illustrated in Fig. 2. The circle diagrams show per position which nucleotides were selected (output) in this in vivo SELEX experiment compared to the nucleotide variation present in the library (input) and the natural variation among HIV-1 and simian immunodeficiency virus cpz (SIVcpz) strains. Only the wt G is selected at position 237, whereas at the other positions non-wt nucleotides are allowed to a very small extent. In general, the selected nucleotides were consistent with the most prevalent nucleotide in the natural isolates. Position 241 forms an exception, as U was selected more frequently than the wt G that is strongly preferred in nearly all natural virus isolates. The relatively short duration of the experiment and the small set of selected virus variants with a sequence similar to the wt indicate that a very strong selection pressure exists for region 1.

FIG 2.

In vivo SELEX of region 1. The nucleotide sequence in the wt HIV-1 LAI strain is indicated at the top. The SELEX input was analyzed by sequencing 33 randomly picked clones from the region 1 library. The actual sequences are presented in Table S1 in the supplemental material. The SELEX output summarizes the sequences of the HIV-1 variants selected in 24 independent cultures. The natural variation summarizes the 201 HIV-1 and SIVcpz sequences present in HIV Sequence Compendium 2010 (79).

We analyzed the impact of the selected sequences on the RNA secondary structure of the untranslated leader RNA. This in silico analysis was performed with the RNAstructure algorithm (60). If the polyA (nt 58 to 104) and DIS (nt 248 to 270) hairpins were formed in the leader RNA, we scored the BMH conformation (Fig. 1B). When these two hairpins were absent due to a long-distance base-pairing interaction, the leader RNA was scored as adopting the LDI conformation. The predicted thermodynamic stability of the DIS hairpin (ΔG_DIS) was also calculated for each selected virus variant (Table 1). The majority of the selected region 1 sequences had no or only a slight effect on DIS hairpin stability. The variants GCUGGC and GCCGGC stabilized the DIS hairpin to −17.0 and −16.4 kcal/mol by formation of four and three additional base pairs, respectively. The thermodynamic stability of the 5′ leader RNA in the LDI (ΔG_LDI) and BMH (ΔG_BMH) conformation was calculated to determine the preference of the sequence to fold the LDI (ΔΔG_LDI-BMH ≤ 0) or BMH conformation (ΔΔG_LDI-BMH > 0). The ΔΔG_LDI-BMH of the wt was −5.6 kcal/mol, consistent with a preferential folding of the LDI conformation. In none of the selected virus variants was an LDI-to-BMH switch (positive ΔΔG_LDI-BMH) observed. Even the two DIS-stabilizing variants did not induce such a switch, but they approach the LDI-BMH equilibrium with a ΔΔG_LDI-BMH of −0.5 (GCCGGC) and −2.4 (GCUGGC) kcal/mol. Although many region 1 mutants can be designed that would trigger an LDI-to-BMH switch (46), we never observed sequences that led to an actual switch in this SELEX experiment.

Nucleotides in between the DIS and SD hairpins (region 2) are subject to moderate selection pressure.

Region 2 consists of four nucleotides (²⁷⁸GGGA²⁸¹) that are flanked by the upstream DIS hairpin and downstream SD hairpin (Fig. 1B). In vivo selection of randomized region 2 viruses lasted around 17 weeks, with a few exceptions of up to 24 weeks. A slightly larger number of virus variants were selected in the 24 cultures than for region 1 (Table 2). The wt GGGA sequence was selected in 3 cultures. Interestingly, two point mutants were also selected frequently: 6 times for GGGU and 4 times for UGGA. The other 11 sequences were selected only once; the majority represent double mutants with a preference for changes in the first and last positions, 278 and 281 (Table 2). Although the wt sequence was not selected most frequently, the actual consensus sequence of all 24 cultures is similar to the wt (Fig. 3). At the two central positions a G nucleotide is highly preferred, which is also the case in natural HIV-1 isolates, whereas the two outer positions of region 2 allow more sequence variation. The preferred nucleotide for position 278 is the wt G, but U also seems acceptable. Position 281 reveals an interesting pattern: U and A were the preferred nucleotides in the in vivo SELEX experiment, whereas C is present in the majority of natural virus isolates. In fact, the LAI strain is atypical in having an A at position 281, which thus reappeared in this in vivo selection experiment. This result may allow us to cautiously infer that there is a second-site determinant in the LAI genome that dictates the relatively unique selection of 281A. Compared to region 1, the extended selection time (17 to 24 weeks) together with the selection of different nucleotides at positions 278 and 281 indicates a moderately strong selection pressure for a specific nucleotide sequence for region 2, with a preference for GGGA (wt), UGGA, or GGGU.

TABLE 2.

In vivo SELEX of region 2

Sequencea	Frequencyb	ΔGDISc	ΔGSDc	ΔGLDIc	ΔGBMHc	ΔΔGLDI-BMHd
GGGU	6	−12.9	−10.3	−140.1	−136.3	−3.8
UGGAe	4	−14.2	−9.6	−139.8	−135.2	−4.6
GGGA (wt)	3	−12.9	−9.6	−139.8	−134.2	−5.6
AGGA	1	−12.4	−9.6	−140.1	−136.5	−3.6
AGGU	1	−12.4	−10.3	−141.0	−137.2	−3.8
AGUA	1	−12.4	−9.6	−138.0	−136.1	−1.9
GAAG	1	−12.9	−10.3	−139.1	−135.0	−4.1
GAGU	1	−12.9	−10.3	−139.8	−135.2	−4.6
GCGAe	1	−12.9	−9.6	−141.7	−137.6	−4.1
GGGC	1	−13.3	−9.6	−141.7	−137.6	−4.1
GGU–	1	−12.9	−10.3	−139.9	−135.9	−4.0
UGGC	1	−14.2	−9.6	−139.2	−138.2	−1.0
UGGG	1	−14.2	−9.6	−138.9	−136.6	−2.3
UGGU	1	−14.2	−10.3	−141.3	−137.0	−4.3

^aNon-wt nucleotides are in bold and underlined.

^bObserved frequency in 24 cultures.

^cΔG, in kcal/mol, of DIS and SD hairpins and leader conformations (LDI and BMH).

^dΔΔG in kcal/mol as a measure of the LDI-BMH equilibrium.

^eShows a rearranged base-pairing scheme of the DIS hairpin in some of the RNAstructure predictions. For such structures the nucleotides which appear single stranded in the DIS hairpin of the wt (positions 247, 255 to 263, and 271 to 273) were forced to be single stranded in the predictions.

FIG 3.

In vivo SELEX of region 2. See the legend to Fig. 2 for details. The SELEX input is based on the sequence of 23 randomly picked clones. The actual sequences are presented in Table S1 in the supplemental material.

The effect of the selected sequences on the RNA secondary structure of the individual DIS and SD hairpins and the overall leader RNA was analyzed. Eleven variants were selected that slightly stabilize either the upstream DIS hairpin (UGGA, GGGC, UGGC, UGGG, and UGGU) or the downstream SD hairpin (GGGU, AGGU, GAAG, GAGU, GGU-, and UGGU) by formation of one or two additional base pairs (Table 2). The effect on the folding of the complete leader RNA was also modest. For none of the sequences was a complete LDI-to-BMH switch observed, and only modest shifts toward BMH were seen upon DIS stabilization.

Minor selection pressure for nucleotides in between the SD and Ψ hairpins (region 3).

Single-stranded region 3 consists of a well-conserved stretch of five A nucleotides (³⁰¹AAAAA³⁰⁵) flanked by the SD and Ψ hairpins (Fig. 1B). After around 20 weeks, with a few exceptions of only 7 weeks, the 24 SELEX cultures yielded a homogenous virus population, comprising a large and diverse set of virus variants (Table 3). Surprisingly, the wt sequence AAAAA was never recovered. Five of the 24 variants analyzed (CUACG, CUCCA, GUAUU, GUAGC, and the deletion mutant A----) were selected twice, and 14 variants were selected once (Table 3). The in vivo selection output shows a mixed sequence composition at all positions (Fig. 4). All selected sequences differ to a large extent from the wt sequence, but there seems to be a modest tendency to restore the wt A at central position 303. In addition, U seems to be slightly preferred at position 302, which contrasts with the consensus A in natural isolates and our wt LAI virus. Other positions show at most a weak tendency to select non-wt nucleotides. Not only was the wt sequence never selected but we also did not detect any point mutant; we detected relatively few double (3×) and triple (3×) mutants, and mostly sequence stretches with four (9×) or all five positions (4×) deviating from the wt sequence. This significant nucleotide variation among the selected viruses and the outgrowth of a deletion mutant (A----) confirm the absence of a strong sequence requirement for this leader RNA region. These results are especially striking considering the fact that in all HIV-1 isolates a strong purifying selection toward an A stretch is present at all positions, especially for positions 301 to 304 (Fig. 4).

TABLE 3.

In vivo SELEX of region 3

Sequencea	Frequencyb	ΔGSDc	ΔGΨc	ΔGLDIc	ΔGBMHc	ΔΔGLDI-BMHd
A––––	2	−9.6	−7.6	−141.1	−136.3	−4.8
CUACG	2	−10.1	−7.6	−144.1	−136.6	−7.5
CUCCA	2	−15.1	−9.7	−143.3	−141.0	−2.3
GUAUU	2	−9.6	−9.1	−143.5	−136.8	−6.7
GUAGC	2	−9.6	−9.6	−145.5	−137.9	−7.6
AAAUU	1	−9.6	−9.1	−142.9	−135.8	−7.1
AAGAU	1	−9.6	−8.9	−141.7	−135.4	−6.3
ACAAU	1	−9.6	−8.9	−141.7	−136.2	−5.5
AUUGU	1	−9.6	−8.9	−144.6	−137.9	−6.7
CACGG	1	−8.9	−7.6	−142.3	−135.6	−6.7
CGUUG	1	−8.9	−7.6	−144.8	−138.1	−6.7
GGAGA	1	−9.6	−7.6	−140.1	−136.7	−3.4
GUAAC	1	−9.6	−9.4	−144.1	−137.8	−6.3
GUACA	1	−9.6	−7.6	−141.3	−135.2	−6.1
GUAGG	1	−9.6	−7.6	−141.7	−134.7	−7.0
GUAGU	1	−9.6	−8.9	−142.7	−135.5	−7.2
UCGGG	1	−14.1	−7.6	−142.0	−139.4	−2.6
UCUGU	1	−14.8	−8.9	−146.2	−141.8	−4.4
UGUUG	1	−11.4	−8.9	−146.2	−141.8	−4.4
AAAAA (wt)	0	−9.6	−7.6	−139.8	−134.2	−5.6

^aNon-wt nucleotides are in bold and underlined.

^bObserved frequency in 24 cultures.

^cΔG, in kcal/mol, of the SD and Ψ hairpins and leader conformations (LDI and BMH).

^dΔΔG in kcal/mol as a measure of the LDI-BMH equilibrium.

FIG 4.

In vivo SELEX of region 3. See the legend to Fig. 2 for details. The SELEX input is based on the sequence of 29 randomly picked clones. The actual sequences are presented in Table S1 in the supplemental material.

Given the diversity of sequences selected in region 3, a variety of RNA structural effects were scored in the in silico analyses, and no common trends were apparent. The majority of the selected sequences had no or only a small effect on the stability of the neighboring SD and Ψ hairpins, and consequently the LDI-BMH equilibrium was not significantly affected (Table 3).

Selected virus variants have a near-wild-type replication capacity.

Perhaps the most surprising result is the frequent selection of particular non-wt sequence motifs for regions 1 (GCAGUA and GCAAGA) and 2 (GGGU and UGGA). To test whether these virus variants replicate as efficiently as or even better than the wt, we generated LAI mutants with the specific mutations and tested their replication capacity on three different cell lines (SupT1, PM1, and 174×CEM) and primary PBMCs (Fig. 5). Cells were infected with equal amounts of virions (based on CA-p24), and viral spread was monitored by measuring the CA-p24 level in the supernatant. The selected virus variants indeed replicated as efficiently as the wt in all cells (Fig. 5). We also generated some control LAI constructs with nonselected region 1 and 2 motifs (Table 4). In region 1, we mutated either the first three nucleotides (mutant 1D), the last three (1C), or all six (1E). In region 2, the wt GGGA motif was changed to AAAA (mutant 2C). Testing of these four mutants in SupT1, PM1, and 174×CEM cells demonstrated that they all exhibit moderate to severe replication defects (Fig. 5). As expected, the 1E mutant with 6 mutations exhibited the most severe replication defect. These results confirm the importance of these single-stranded sequences.

FIG 5.

Replication capacity of mutant LAI viruses. Cells were infected with equal amounts of virus, based on the CA-p24 content of the virus stocks. Virus spread was assessed by monitoring CA-p24 levels in the culture supernatant for up to 25 days. Viral replication capacity of the wt LAI virus and region 1 (1A to 1E), region 2 (2A to 2C), and region 3 (3A to 3C) variants was tested on SupT1, PM1, and 174×CEM cells and CD4⁺ PBMCs.

TABLE 4.

RNA secondary structure of RNA leader mutants with compromised replication capacity

Mutant	Sequencea	Region(s)b	ΔGDISc	ΔGSDc	ΔGLDIc	ΔGBMHc	ΔΔGLDI-BMHd
wt	GCAGGA-GGGA	1 + 2	−12.9	−9.6	−139.8	−134.2	−5.6
1C	GCAAAG-GGGA	1	−12.7	−9.6	−138.8	−134.7	−4.1
1D	CUGGGA-GGGA	1	−12.9	−9.6	−140.7	−135.6	−5.1
1E	CUGAAG-GGGAe	1	−12.7	−9.6	−136.3	−135.4	−0.9
2C	GCAGGA-AAAA	2	−12.4	−9.6	−138.0	−134.5	−3.5

^aSequence for regions 1 and 2; non-wt nucleotides are in bold and underlined.

^bMutated region.

^cΔG, in kcal/mol, of DIS, SD, and Ψ hairpins and leader conformations (LDI and BMH).

^dΔΔG in kcal/mol as a measure of the LDI-BMH equilibrium.

^eShows a rearranged base-pairing scheme of the DIS hairpin in some of the RNAstructure predictions.

To delineate the replication capacity of the selected non-wt HIV-1 variants in further detail, the in vivo-selected region 1 and 2 variants were tested in direct competition with the wt virus (Table 5) (62, 63). Equal amounts of DNA of the wt and non-wt variant were transfected into SupT1 cells. After 2 and 4 weeks, the ratio of the two virus variants was assessed by population sequencing. Both region 1 variants (GCAGUA and GCAAGA) and the region 2 variant UGGA were outcompeted by the wt within 2 weeks, but the other region 2 variant (GGGU) was still present in the coculture after 4 weeks.

TABLE 5.

Competition experiments

Regiona	Virus	Sequenceb	Competition 1c	Competition 2c
1	LAI wt	GCAGGA
	Mutant A	GCAGUA	wt	wt
	Mutant B	GCAAGA	wt	wt
2	LAI wt	GGGA
	Mutant A	GGGU	1:1	Mutant
	Mutant B	UGGA	wt	wt
3	LAI wt	AAAAA
	Mutant A	GUAUU	1:1	1:1
	Mutant B	A----	Mutant	Mutant
	Mutant C	CUCCA	wt	wt

^aRegion for which the sequence was selected.

^bSelected sequence, with the nucleotides differing from the wild-type sequence in bold and underlined.

^cResults of both viral competition experiments after 4 weeks. Shown is the winning virus variant; “1:1” indicates a heterogenous population after 4 weeks.

We also tested some of the sequence variants in region 3 that were selected as winners in the in vivo SELEX experiment. These selected variants also replicated efficiently (Fig. 5). When tested in the direct competition assay, one of the region 3 variants was outcompeted (CUCCA), one variant (A----) outcompeted the wt virus (AAAAA), both within 2 weeks, and one variant was still present in equimolar amounts with the wt after 4 weeks (GUAUU) (Table 5). These results confirm that many diverse sequences in region 3 are compatible with efficient HIV-1 replication.

DISCUSSION

The 5′ leader of the HIV-1 RNA genome is highly structured and exposes several hairpins that have distinct regulatory functions in the viral replication cycle (1). The roles of the conserved and purine-rich single-stranded regions flanking the DIS, SD, and Ψ hairpins have not been studied in much detail. We set out to investigate the nucleotide requirements of these three single-stranded regions. For this purpose, we used the in vivo SELEX approach, in which efficiently replicating viruses are selected from a library of virus variants with a randomized genome segment (53). In multiple independent in vivo SELEX experiments, a small set of virus variants, including the wt and particular point mutants, were selected in a relatively short time in regions 1 and 2 that flank the DIS hairpin. For region 1, the wt sequence is strongly preferred at all positions. Surprisingly, a point mutant with 241U at the fifth position was selected frequently and was shown to replicate as well as the wt virus with 241G. Our virus selection experiments demonstrate that for region 2, the sequence of the two central positions (²⁷⁹GG²⁸⁰) is especially important. Although the point mutants ²⁷⁸UGGA²⁸¹ and ²⁷⁸GGGU²⁸¹ were selected frequently, the combination (²⁷⁸UGGU²⁸¹) was observed only once. Overall, these results indicate the presence of a very strong selection pressure on the nucleotide sequence of regions 1 and 2. In contrast, no particular sequence variants were purified for region 3, located in between the SD and Ψ hairpins.

For region 3, the wt sequence was never selected in the 24 cultures tested, and most cultures needed 5 months to purify a particular virus variant that almost always had a different sequence, with 3 to 5 mutations compared to the wt. The presence of the highly conserved A stretch in most natural HIV-1 variants contrasts with the absence of any apparent selection pressure in the SELEX experiment and subsequent in vitro replication tests. This may indicate that this sequence motif does play a role only during in vivo HIV-1 replication, but there are few examples for such an exclusive in vivo effect of an RNA signal (30, 32). Alternatively, conservation of the A stretch in natural HIV-1 isolates may reflect the generic evolutionary bias of HIV-1 to acquire A-rich sequences in single-stranded domains of its RNA genome (64, 65). Consequently, the viral genome will become A rich over time by default, unless evolution is restricted, e.g., when a particular function is attributed to a specific sequence motif. Thus, simply looking at natural sequence conservation as a first predictor for the importance of a certain sequence motif may be misleading.

The sequences of 5′ leader regions 1 and 2 are strongly conserved and apparently allow little sequence variation. These sequence motifs may form part of one or several protein binding sites that are essential for efficient viral replication. Some studies probed protein binding to region 1, e.g., the viral chaperone nucleocapsid (NC) protein, which has many proposed roles in the HIV-1 replication cycle (66–71). The consensus for specific NC-binding sites is a G-rich single-stranded sequence flanked by double-stranded stretches, which was revealed in experiments with short RNA transcripts (67) and more recently using SHAPE technology (51). The latter study mapped exact NC binding sites, of which one (nt 239 to 244) overlaps region 1 (nt 237 to 242). Early experiments suggested that region 3 is also part of a NC binding site, as was shown for short RNA transcripts of the MAL isolate (72). However, the MAL isolate is relatively unique in having only 2 instead of the regular 5 A residues. Another viral protein that has been suggested to bind to region 1 is Rev. This binding event functions as an alternative for Rev binding to the Rev response element (RRE) (73, 74) and has been suggested to play a role in genomic RNA packaging. Cellular proteins may also interact with these sequences, e.g., splicing factors that control the affinity of the major 5′ splice site embedded within the SD hairpin (34, 75).

Regions 1 and 2 flank the DIS hairpin and have the potential to stabilize the DIS hairpin by formation of an additional base pair when a 242C/U or 278U is selected. If selection is random, 25% (6 out of 24) of the sequences are expected to have this property. However, 242C was selected only twice and 242U was never present in region 1, whereas 278U was selected 7 times in region 2. Variants that stabilize the DIS hairpin with 2 bp were never observed. Apparently the virus allows extension of the DIS hairpin with a single base pair, but this occurs mainly via mutations in region 2, confirming the higher selection pressure on the sequence of region 1.

The DIS hairpin, with the palindromic sequence GCGCGC at the top of the apical loop, is the initiation site for genomic RNA dimerization. Deletion of this palindromic sequence decreases genomic RNA dimerization and affects virus replication capacity. Interestingly, deletion of 7 nucleotides immediately upstream of the DIS hairpin (nt 236 to 242, including the complete region 1) has an even larger effect (76), although this could be the result of an altered RNA secondary structure. Nucleotides flanking the DIS hairpin (regions 1 and 2) also have an effect on in vitro extended dimer formation, which is most likely a sequence-specific effect (46). When sequence motifs overlapping region 1 (nt 236 to 242) or region 3 (nt 302 to 309 or nt 300 to 312) were deleted, dramatic effects on genomic RNA dimerization were scored (77). The deletions could affect the RNA secondary structure, but protein binding sites could also be removed. These results reinforce our findings of important sequence information in regions 1 and 2. The large effects of deleting region 3, while we observed a low selection pressure for a particular sequence in our SELEX experiments, suggests that the sequence can be varied as long as the local and overall RNA secondary structures are not affected.

A counterselection against changes in the overall RNA secondary structure of the 5′ leader is evident from the SELEX data. A complication is that several base-pairing models have recently been proposed (2, 36, 40, 50–52). Region 3 is present as a single-stranded domain in all models, which is extended in some models by 6 single-stranded nucleotides on the 3′ side. Regions 1 and 2 are single stranded and flank the DIS hairpin in the BMH conformation, whereas the DIS hairpin is extended by 2 or 3 G-C base pairs in two other models (1, 51). Both DIS extensions create an internal bulge, and the 3-bp extension does in fact trigger shortening of the SD hairpin by a single base pair. A second report by one of these groups did not confirm the DIS extension (36). The primary LAI isolate is atypical in having 281A, which does not allow DIS extension, whereas C is present in the majority of HIV-1 isolates to allow one additional base pair. Interestingly, 281A reappeared unequivocally in the SELEX experiment with the LAI strain. These results demonstrate the importance of 281A for LAI replication and do not support the two HIV-1 leader RNA structure models with extended DIS base-pairing.

Another recent structure model proposes a pseudoknot interaction between region 1 sequences (nt 234 to 238) and the 5′ end of the Gag open reading frame (nt 348 to 352) (50). This pseudoknot model extends the previously identified U5-AUG long-distance contact (2) with 5 bp, which might enforce this long-distance interaction. Our results show that region 1 residues ²³⁷GC²³⁸ are strongly preferred, which is consistent with this base-pairing scheme. Region 2 is predicted to be single stranded in this model.

For HIV-1, it has been proposed that multiple 5′ leader functions, e.g., genomic RNA dimerization and packaging, are regulated or even coordinated by a riboswitch (40, 49, 52). We proposed that the 5′ leader region is in equilibrium between the LDI conformation in which the DIS hairpin is masked and the BMH conformation in which the DIS hairpin is exposed, thus enabling regulation of RNA dimerization. We showed previously that DIS mutants can modulate the LDI-BMH and monomer-dimer equilibrium in vitro (46). For very many 5′ leader mutants and revertants, obtained by spontaneous virus evolution, we described a nearly perfect correlation between the ΔG of the LDI and BMH conformations and the replication capacity of those viruses. An efficiently replicating virus preferentially folds the LDI structure (ΔΔG_LDI-BMH remains negative), and 5′ leader mutants with a positive ΔΔG_LDI-BMH value have severe defects in replication capacity. This correlation was confirmed in this study. We performed an in silico analysis of all selected virus variants to test for the effect of the acquired mutations on the stability of the RNA secondary structure and consequently the LDI-BMH equilibrium. Viruses with a sequence motif that significantly disturbs the LDI-BMH equilibrium were never selected and presumably counterselected, as such mutants were present in the constructed libraries (data not shown). Only small shifts were allowed as long as the ΔΔG_LDI-BMH remained negative, thus favoring the LDI conformation. The combined results indicate the importance of this riboswitch to impose control over the process of RNA dimerization. The findings also argue for a toxic effect of mutant sequences that establish such a switch, consistent with the previous analysis of mutant viruses (49).

In conclusion, this study showed that the strongly conserved regions 1 and 2 provide important sequence motifs for viral replication. For the sequence of region 3 a very low selection pressure is apparent, which is in sharp contrast with the strong conservation of this sequence in natural isolates. We propose that the A-rich stretch in region 3 was shaped by mutational bias over evolutionary times, a soft but decisive pressure that is visible throughout the HIV-1 RNA genome (64, 78). Thus, conservation of a particular sequence motif in natural HIV-1 isolates may not necessarily point to the importance of that sequence. For the DIS flanking regions, there is selection at the nucleotide sequence level, but there is also a counterselection to prevent large disturbances of the local and global RNA secondary structure. It will be of interest to unravel the biological function of these three single-stranded sequence motifs in further detail.