A Conserved Glycan in the C2 Domain of HIV-1 Envelope Acts as a Molecular Switch to Control X4 Utilization by Clonal Variants with Identical V3 Loops

Nearly all persons newly infected with HIV-1 harbor exclusively CCR5-using virus. CXCR4-using variants eventually arise in up to 50% of patients infected with subtypes B or D. This transition to efficient CXCR4 utilization is often co-incident with progression to AIDS. The basis for HIV-1’s initial dependence on CCR5, the selective force(s) that drive CXCR4-utilization, and the evolutionary pathways by which it occurs are incompletely understood. Greater knowledge of these processes will inform interventions at all stages, from vaccination to cure. The determinants of co-receptor use map primarily, though not exclusively, to the V3 loop of gp120. In this study, we describe five clonal variants with identical V3 loops but divergent CXCR4 use. Mutagenesis revealed two residues controlling this phenotypic switch: a rare polymorphism in C1 and a highly conserved N-glycan in C2. To our knowledge, this is the first description of co-receptor usage regulated by the N-glycan at position 262.


Introduction
The entry of HIV-1 into susceptible host cells is mediated by the viral envelope glycoprotein (Env). The functional envelope spike is comprised of three non-covalently associated gp41 transmembrane sub-units and three gp120 surface units. Each gp120 molecule is divided into five constant (C1-C5) and five variable (V1-V5) domains, and contains a binding site for the primary cell-surface receptor CD4. Following CD4-engagement, the envelope undergoes a conformational change that exposes or creates a binding site for its co-receptor, typically CCR5 (R5) or CXCR4 (X4). Binding to the co-receptor triggers conformational changes in gp41 that ultimately result in fusion of the virus and host-cell membranes [reviewed in [1].
Virtually all HIV-1 infections are established by exclusively R5-using virus, regardless of the presence of R5/X4 or obligate X4 virus in the index case [2,3] 1-2% of persons of Northern European descent are homozygous for a 32 base pair deletion in the CCR5 gene (CCR5Δ32) which alters CCR5 expression on the surface of their cells and renders them highly resistant to HIV. Persons heterozygous for CCR5Δ32 have lower cellsurface expression of CCR5, are partially resistant to infection, and tend to progress slower if infected [4,5]. This rigid constraint on an otherwise fluid and rapidly evolving virus has led to the development and testing of many interventions targeting CCR5 for prevention, treatment, and even cure [6][7][8][9][10]. Unfortunately, these efforts are at risk of failure should the virus successfully transition to efficient X4 utilization [11][12][13]. A better mechanistic understanding of the R5-to-X4 transition will allow scientists and clinicians to better predict, and potentially counter, such an escape strategy.
The determinants of co-receptor usage map primarily to the V3 loop; which makes extensive molecular contacts with the co-receptor [14][15][16]. It has been well-established that an overall shift towards positive charge, but especially positively charged substitutions at positions 11,24, and 25 in V3 are predictive of X4 utilization [17,18], as is the presence of an isoleucine at position 326 in the V3 stem [19,20]. Several algorithms have been developed to predict coreceptor usage based on V3 sequence, with accuracies estimated at 70-80% [21]. However, substitutions in other regions including the bridging sheet, C4, V1/V2, and gp41 have also been shown to influence co-receptor usage [22][23][24][25][26][27].
For this study, we screened a cohort of treatment-experienced, predominantly subtype-B infected subjects failing their current ARV regimens, using resistance to the CCR5-antagonist maraviroc (MVC) as an initial surrogate marker for efficient X4 utilization. We reasoned that this population would be more likely to harbor the X4 using or transitional variants of interest. Convenience sampling of ten subjects identified three with MVC-resistant virus. From one of these three, we isolated a series of closely related molecular envelope clones with identical V3 sequences but highly variable co-receptor usage. Further in vitro characterization revealed that X4 utilization was regulated by polymorphisms in C1 and C2. The C2 polymorphism disrupted a conserved potential N-linked glycosylation site (PNG) important for envelope function but not previously linked to co-receptor selectivity.

Study Population
All subjects were treatment experienced and screened for, but unable to enroll in, IMPAACT protocol P1020a [28]. Written informed consent was obtained by study candidate, parent or legal guardian prior to screening and recorded per protocol. The study was approved by the Institutional Review Boards at each investigator site (see listing in acknowledgements and manuscript PMID 25232777) and registered with ClinicalTrials.gov, Identifier NCT00006604. No subjects had prior exposure to entry inhibitors (including Maraviroc).
Envelope clones were sequenced bi-directionally, and the contigs assembled and edited using Sequencher (Gene Codes, Ann Arbor, MI). Alignments were constructed using MacVector (MacVector Inc, Cary, NC) and the LANL website tools (http://www.hiv.lanl.gov/content/ sequence/HIV/HIVTools.html). All amino acid residues are numbered based on the HXB2 reference sequence. Point mutations were introduced by site-directed mutagenesis (QuikChange multi-site directed mutagenesis kit; Stratagene, La Jolla, CA) according to the manufacturer's instructions and sequenced to confirm integrity. For the remainder of the manuscript, unmutated envelopes derived from subject plasma will be referred to as 'parental'.
All cell entry assays (no primary cell data were included) were performed using pseudotype virus in a 96-well plate format as previously described [48]. Briefly, using titers obtained by Xgal staining,~2000 units of pseudovirus in 16 μg/mL DEAE Dextran were added to each well containing target cells. Luminescence was quantified after a 48-hour incubation using a Promega Luciferase Kit (Promega, Madison, WI) and a FLUOstar luminometer (BMG Labtech, Cary, NC) according to manufacturer instructions.
Phenotypic determination of co-receptor utilization was performed by infecting TZM-bl cells in the presence or absence of AMD3100 (X4 antagonist) or TAK-779 (R5 antagonist), and by infection of GHOST cells (expressing CD4 and one or both of the co-receptors) with pNL4-3.Luc.R -Ederived pseudoviruses.
Drug sensitivity was determined by pre-treatment of TZM-bl cells for 1 hour at 37°C with increasing concentrations of the CCR5 antagonists MVC, TAK-779, AD101, the X4 inhibitor AMD3100, the fusion inhibitor T20, or the lectins HHA, GNA, or UDA prior to addition of pseudovirus.

Data Analysis
Drug sensitivity was quantified by calculating a mean 50% inhibitory concentration (IC 50 ) for each envelope/drug combination with the Prism 4 software (GraphPad, San Diego, CA) using data from at least two experimental replicates. Co-receptor usage, based on TZM or GHOST cell methods was expressed as a percentage normalized to entry in the absence of inhibitors in cells expressing both co-receptors (GHOST-R3/X4/R5 for that platform).

Clinical Characteristics of Study Subjects
The clinical characteristics of our cohort are listed in Table 1. All subjects were experiencing virologic failure (median plasma HIV RNA of 106,808 copies/ml) on their current ARV regimen, with at least moderate resistance to two or more drug classes as determined by genotypic analysis using the Stanford HIV drug resistance database. Of the ten subjects included in our initial round of screening, nine were infected with subtype B and one (subject 1002) was infected with subtype C HIV-1.

Identification of a Subject with both MVC-Resistant and MVC-Sensitive Virus
Full-length molecular envelope clones were generated from the plasma of all ten subjects by RT-PCR and fully sequenced. V3 loops were analyzed using the web-based algorithms Geno2-Pheno, PSSM, and 11/24/25. Pseudotype viruses were generated from all envelopes and tested for resistance to MVC, which served as a surrogate marker for X4 utilization at the screening stage. Data are presented in Table 2. Viral envelopes isolated from seven of the ten patients screened were predicted in silico to be exclusively R5-using and were phenotypically sensitive to MVC, thus no further studies were performed on them. Seven clones isolated from subject 1009 were predicted to use X4 by both Geno2Pheno and PSSM and were highly resistant to MVC. Two clones were isolated from subject 1005 with highly divergent V3 loop sequences, one of which was predicted to use X4 by Geno2Pheno and PSSM and was MVC-resistant, while the other was predicted to use X4 by Geno2Pheno but R5 by PSSM, and was sensitive to MVC. Finally, five clones were isolated from subject 1010, all with identical V3 loop sequences that were predicted to use X4 by Gen-o2Pheno but R5 by PSSM. Two of those clones were sensitive to MVC (IC 50~2 nM) while the other three were highly resistant (IC 50 >200nM). We chose to focus on the isolates from this last subject for further examination. These data are summarized in Table 2.

Identification of Non-V3 Loop Co-receptor Use Determinants in Subject 1010
The five env clones from subject 1010, which had identical V3 sequences, differed at nine positions throughout the remainder of gp160 (five residues in gp120 and four in gp41; Fig 1). Only two positions correlated with MVC sensitivity. The first was the highly conserved bridging sheet residue 123, at which the MVC-sensitive isolates had an unusual isoleucine substitution for the conserved threonine (T123I). Polymorphisms at this site were present in only 6 of 1501 complete subtype B sequences retrieved from the 2012 LANL database (two T123I and four T123A). The second position was the highly conserved serine 264 in the C2 domain, which   was mutated to a glycine in the MVC-sensitive isolates (S264G). Polymorphisms at this site were similarly rare, present in only 7 of 1501 subtype B LANL sequences (one S264G, one S264N, three S264T, and two nonsense mutations). For simplicity, envelopes will be referred to by the amino acids at positions 123 and 264 (i.e. I+G or T+S) for the remainder of the manuscript (as shown in Fig 1). It is important to note here that serine 264 is part of a PNG sequon, [NX(S/T)] including the conserved asparagine at 262, which is disrupted by the S264G polymorphism. This glycan has been localized to the outer domain of gp120, where it forms a glycan cluster with N295, N332, and N448 [most recently in the structure described in [49]].
To verify that these isolates were indeed using X4 for entry and not simply able to productively utilize MVC-bound CCR5 as described in [50], they were tested on cells pre-treated with two additional CCR5 inhibitors, TAK779 and AD101, to which they were also resistant. Envelopes were also tested for entry into GHOST cells expressing CD4 and either CCR5 or CXCR4 alone, with results concordant with those obtained from the TZM-based inhibitor assays (summarized in Fig 1). These results confirmed that we identified clonal envelope variants with highly divergent X4 usage regulated by polymorphisms outside the V3 loop.

Site-Directed Mutagenesis Reveals a Novel Regulator of X4 Usage
We created a panel of mutant envelopes in which the residues at 123 and 264 were exchanged individually and in combination between the MVC-resistant/X4-using (T+S) and MVCsensitive/R5-using (I+G) variants. This panel was screened using MVC, TAK779, AD101, AMD3100, and assayed for entry into R5-and X4-expressing GHOST cells. Results are presented in Fig 1. The exchange of both residues (123 and 264) was sufficient to confer MVC-resistance and X4-utilization on the sensitive/R5 envelopes, and vice versa. The intermediate genotype T+G (which lacked a PNG at 262) was relatively sensitive to R5 inhibitors, but demonstrated elevated efficiency of X4-usage compared to the fully MVC-sensitive envelopes (I+G). In contrast, the intermediate genotype I+S (which possessed a PNG at 262) was relatively resistant to R5 inhibitors and had an X4-utilization efficiency much closer to the fully MVC-resistant (T+S) clones. These data indicate that exchange of residues at positions 123 and 264 is sufficient to confer full X4-tropism, and that position 264 is the dominant molecular switch.
Only one mutant, 547 I+G, was non-functional for pseudotyping in 293 cells. I+G mutants made using the other parental X4 envelopes (543 I+G and 544 I+G) were fully functional, as was the mutant 547 T+G. The most apparent difference between the parental 547 and 543/544 envelopes is the presence of a conserved PNG at position 392 in clone 547 that is absent in 543/ 544. Interestingly, this PNG was also absent in the parental R5 variants (542/550), both of  which have the I+G genotype. To determine whether a PNG at position 392 was functionally incompatible with the I+G genotype we constructed a second round of mutants, disrupting the PNG sequon with an S394N substitution in the non-functional mutant 547 I+G and adding a PNG site to clone 542 (a parental I+G) at position 392 via the reciprocal N394S substitution. Surprisingly mutant 547 I+G S394N remained non-functional for pseudotyping, while mutant 542 N394S retained function and was similarly sensitive to R5 inhibition as the parental 542 variant (data not shown). The functional and non-functional variants differed at only two additional residues, a glycine-to-aspartic acid at position 395, and a serine-to-glycine at position 668 in the gp41 membrane-proximal external region. Changes in T20 susceptibility, in the absence of gp41 mutations that directly affect binding, have been associated with altered fusion kinetics [51]. T20 binds to the fusion-intermediate conformation of gp41 and prevents final collapse of the gp41 trimer into the 6-helix bundle required for membrane fusion and entry [52]. To determine whether any of our parental or mutant envelopes had gross defects in their entry kinetics, we tested the full panel for susceptibility to T20. We did not observe any significant differences in T20 susceptibility between R5 and X4 variants, nor associated with any of our in vitro mutants (Fig 1).

Glycosylation of Asparagine 262, Rather Than the Specific Identity of the Residue at Position 264, Controls Co-Receptor Usage and Effects Neutralization by Mannose-, but Not Complex Glycan-, Dependent Soluble Lectins
To confirm that the S264G substitution was indeed acting by elimination of the glycan attached to asparagine 262, we constructed a set of mutants designated I+(Q)S and T+(Q)S, in which Columns have been color-coded to highlight the concentration of a given compound that is required to achieve the IC50. Expressing cut-offs for green, yellow, and red coding: for MVC and AD101:0-9.9, 10-24.9, >25nM; for TAK-779: 0-24.9, 25.0-49.9, >50nM; and for GNA and HHA: 0-0.49, 0.5-0.99, and >1.0 μg/ml. T20 and UDA were not color coded, as their IC50s did not correspond to the % X4 usage of the clone. Phenotypic data for reference viruses Bori, 89.6, and pNL4-3 are shown at the bottom of the figure. doi:10.1371/journal.pone.0128116.g001 the PNG sequon was disrupted by an N262Q substitution that left the serine at position 264 intact. The I+(Q)S and T+(Q)S mutants behaved identically to the I+G and T+G variants (all of which lack a PNG at 262). We constructed other mutants, designated I+T and T+T, in which the serine at position 264 was changed to a threonine, preserving the PNG sequon. The I+T and T+T mutants behaved like the I+S and T+S variants (all of which have a PNG at 262). These data confirm that it is the presence of a PNG at position 262, and not the specific identity of the residue at position 264, that regulates co-receptor usage in these clonal variants.
We also tested the parental and mutant envelopes with the soluble plant-derived lectins GNA, HHA, and UDA. GNA and HHA both bind high-mannose glycans, while UDA binds complex-type glycans [53]. All three lectins neutralized all tested envelopes (Fig 1). Activity of the two mannose-dependent lectins, GNA and HHA, correlated well with the activity of the R5 antagonists (R 2 0.68-0.79): the most lectin-resistant variants had the T+S genotype, the moderately resistant variants the I+S genotype, and the most sensitive had either the I+G or T+G genotypes. Resistance to the complex-glycan-dependent lectin UDA did not vary significantly among any of our isolates, nor correlated with the activity of any of the R5 antagonists (R 2 0.01-0.05). It is important to note that previously published biochemical analyses indicate that the glycan at position 262 is predominantly high mannose when produced in 293 cells [54,55]. Correlation plots are presented in Fig 2.

Discussion
In this study, we analyzed a series of clonal HIV-1 envelope variants with identical V3 loops but varied co-receptor usage. We found that substitutions at conserved residues in the bridging sheet (position 123) and the C2 domain (PNG at position 262) were responsible for regulating X4 usage in these isolates. To our knowledge, neither of these positions has been previously identified as a regulator of co-receptor utilization.
It is noteworthy that in this case the atypical (I+G) variants are the more phenotypically 'normal' (MVC-sensitive and R5-using), while the genotypically normal (T+S) variants are phenotypically unusual (MVC-resistant with efficient X4-utilization). This limits our ability to draw broad conclusions about X4 evolution in general, but does point to a previously unappreciated role of the N-glycan at 262 in governing the conformation or exposure of the co-receptor binding site. Previous studies have identified this glycan as important for viral infectivity [56,57], and it will be of interest to determine what compensatory mutations have emerged in these naturally occurring variants.
The lack of effect resulting from exchange of the PNG at 392 between functional and nonfunctional I+G variants was surprising to us. The conserved N-glycan at position 392 is present in only one of the three X4-using clones and its absence does not affect any measure of co-receptor usage that we tested. Additional studies will be required to determine what functional relationship exists between PNG 392 and positions 395/668, and whether those interactions are strain-specific or generalizable.
One limitation of our study is the lack of longitudinal sampling. While it is very likely this subject was originally infected with an R5 virus, we cannot be certain without testing earlier samples. If this were indeed the case, it would be of interest to know whether the transmitted isolate possessed a PNG at 262 and, if not, how the acquisition of that PNG is temporally related to the emergence of a strong X4 phenotype. Moreover, if the transmitted isolate was indeed exclusively R5-using and evolved to utilize X4 while still possessing a PNG at 262, then dissecting the selective forces that favored some variants loosing their ability to use X4, as well as the conserved PNG at 262, may shed more light on the underlying biology of co-receptor selection.
It is also of interest that the X4 clones from this subject (both parental and mutant) were relatively sensitive to all three CCR5 antagonists (MVC, TAK779, and AD101), not achieving an IC50 >200nM until the strain had >50% predicted X4 usage (see Fig 1). Examination of TZM neutralization curves (exemplar shown in Fig 3) indicated that both R5 inhibitor-sensitive and R5 inhibitor-resistant variants had similar slopes but reached plateaus at the approximate maximum-percent-inhibition predicted by the GHOST cell assay (hence the inability to determine an IC 50 value for envelopes which used X4 very efficiently). Given that MVC, AD101, and TAK779 bind CCR5 differently and may act through different mechanisms [35,50,51,[58][59][60][61][62], these data suggest that the envelopes which utilized X4 efficiently maintained a conventional interaction with CCR5, rather than shifting their binding to a different part of the receptor or increasing affinity.
Considering that all variants isolated from this subject had identical V3 loops, genotypically predicted to use X4, we speculate that they represent a lineage of V3 sequences that are in fact highly adapted to using both co-receptors, but for which the polymorphism at 123 and the loss of the PNG at position 262 selectively compromise X4 binding. It is also possible, since we used assays with productive infection as their endpoint, that the described polymorphisms do not limit X4-binding per se, but instead compromise down-stream conformational changes which ultimately lead to more abortive entry events upon X4 binding, though we consider this alternative hypothesis unlikely.
In summary, we have isolated and described a series of closely related, naturally occurring clonal envelope variants with identical V3 loops that can utilize both CCR5 and CXCR4 efficiently. These variants have two molecular switches, a threonine at position 123 and an Nglycan at position 262, that act cooperatively to permit or constrain X4 utilization without affecting CCR5 usage. This study is, to the best of our knowledge, the first published evidence that these two sites have a direct and significant influence on co-receptor utilization by HIV-1 envelope.

Accession Numbers
Nucleotide sequences associated with this manuscript have been submitted to GenBank with accession numbers: KP693359-KP693388.