N- Linked Glycosylation of the HIV Type-1 gp120 Envelope Glycoprotein as a Major Determinant of CCR5 and CXCR4 Coreceptor Utilization*

The variable V1V2 and V3 regions of the human immunodeficiency virus type-1 (HIV-1) envelope glycoprotein (gp120) can influence viral coreceptor usage. To substantiate this we generated isogenic HIV-1 molecularly cloned viruses that were composed of the HxB2 envelope backbone containing the V1V2 and V3 regions from viruses isolated from a patient progressing to disease. We show that the V3 amino acid charge per se had little influence on altering the virus coreceptor phenotype. The V1V2 region and its N -linked glycosylation degree were shown to confer CXCR4 usage and provide the virus with rapid replication kinetics. Loss of an N-linked glycosylation site within the V3 region had a major influence on the virus switching from the R5 to X4 phenotype in a V3 charge-dependent manner. The loss of this V3 N- linked glycosylation site was also linked with the broadening of the coreceptor repertoire to incorpo-rate CCR3. By comparing the amino acid sequences of primary HIV-1 isolates, we identified a strong association between high V3 charge and the loss of this V3 N- linked glycosylation site. These results demonstrate that the N- linked glycosylation pattern of the HIV-1 envelope particles were pelleted (18,000 3 g for 2 h at 4 °C) from 0.5 ml of culture supernatant harvested from transfected C33A cells. The virus pellet was resuspended in 40 m l of of lysis buffer (50 m M Tris-HCl, 100 m M 2-mercaptoethanol, 2% SDS, 0.1% bromphenol blue, 10% glycerol) and heated to 95 °C for 5 min. The denatured proteins underwent SDS-polyacrylamide gel electrophoresis (PAGE) on 5% gels and transferred to nitrocellulose membrane. After overnight incubation at 4 °C with rabbit-anti-gp120 serum followed by a 1-h incubation with a goat anti-rabbit antibody conjugated to alkaline phosphatase, the position of the gp120 glycoprotein on the molecular weight scale was detected by nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phos- phate color development.

Many seven-transmembrane chemokine receptors, in conjunction with the CD4 molecule, have been shown to facilitate entry of human immunodeficiency virus type-1 (HIV-1) 1 into the various cell types that it infects (1)(2)(3)(4)(5)(6)(7)(8)(9). The CC-chemokine receptor CCR5 and the CXC chemokine receptor CXCR4 are the most significant coreceptors with regard to HIV-1 transmission and pathogenesis (10 -15). Viruses utilizing CCR5 (R5), also referred to as nonsyncytia-inducing or macrophage tropic (M-tropic) viruses, are those preferentially transmitted, whereas those using CXCR4 (X4), also termed syncytia-inducing (SI) or T cell line tropic (T-tropic) viruses, are those more associated with later stage disease and disease progression (16 -19). The full molecular events leading to alterations in coreceptor activity or to receptor switching in vivo are not fully understood.
The most striking association to date between viral envelope variation and biologic phenotype has been the overall amino acid charge of the V3 region, with a higher positive charge associated with the SI phenotype and utilization of the CXCR4 coreceptor (20 -34). The V3 region is not the sole determinant of biological phenotype or coreceptor utilization, and other envelope regions, namely the V1 and V2, have been implicated as important (32)(33)(34)(35)(36)(37)(38)(39)(40)(41). The V4 and V5 regions, in conjunction with the V1 and V3, have also been shown to influence coreceptor usage of a dual-tropic R5X4 virus (42). Additionally, N-linked carbohydrate moieties have been shown to have an influence on certain biological properties of both HIV-1 and HIV-2 viruses (43)(44)(45)(46)(47), some of which would be predicted to alter coreceptor binding and utilization (44 -47).
Several chemokine and chemokine receptor genotypes have been associated with HIV-1 transmission and disease progression, strongly suggesting that these factors play a significant role in controlling viral replication in vivo (12), with their efficacy dependant on the interaction of the virus with the relevant coreceptors (13-15, 48 -52). The natural ligands for the CCR5 receptor, RANTES (regulated on activation normal T cell expressed and secreted), macrophage inflammatory protein (MIP)-1␣, and MIP-1␤, have been shown to inhibit viral entry through either competing for the CCR5 receptor or by downregulating its surface expression (53)(54)(55). A virus with altered coreceptor activity is therefore likely to be selected during disease progression as a consequence of chemokine inhibition pressure. Indeed many studies show that minor alterations within the V3 region of the HIV-1 envelope can render a virus resistant to the blocking effects of the CC-chemokines (54 -56). A better knowledge of the molecular events contributing to viral evolution, specifically the switch from the R5 to X4 phenotype will be important in broadening our understanding of HIV-1 pathogenesis as well as providing information relevant to both HIV-1 therapy design and vaccine development.
To define the molecular events involved with the coreceptor switching in vivo, we created and studied a panel of molecularly cloned viruses based upon the V1V2 and V3 envelope regions of viral isolates derived from an HIV-1-infected individual who has progressed to disease. The V1V2 region was shown to be significant in providing the virus with R5X4 dual tropism but not in affecting the level of CCR5 utilization. Here we demonstrate that the loss of an N-linked glycosylation event within the V3 region of the envelope was significant in determining strong CXCR4 utilization and in the virus switching from the R5 to X4 phenotype. In addition, we have compared the V3 sequences from a large number of primary isolates and found a strong association between an increase in V3 charge and the loss of the V3 N-linked glycosylation event, highlighting its in vivo significance. These results demonstrate that the N-linked glycosylation pattern of the gp120 envelope can contribute to viral coreceptor utilization and switching.

EXPERIMENTAL PROCEDURES
Molecular Cloning and Sequencing of Viruses-The initial panel of molecularly cloned HIV-1 chimeric viruses were composed of the LAI viral backbone carrying the HxB2 envelope where the V3 loop was replaced with the corresponding region amplified from a patient (ACH168) progressed in the disease course. These viruses, termed X, are shown in Fig. 1A and were created as previously described (22,27). With the X viruses, we have replaced the V1V2 region of the envelope glycoprotein with the corresponding region of a T-tropic isolate, 168.10, originating from the same patient (ACH168) and from a late time point in disease. The new set of viruses, termed X.10, of which the V1 region is shown in Fig. 1B, were created as follows. The BlueScirpt plasmid was modified for our requirements through replacing the KpnI site with a NcoI site by inserting the linker 5Ј-TTC CAT GGA AGT AC-3Ј at the KpnI site, resulting in the pBSn plasmid. The NcoI-BamHI fragment of the X virus HIV-1 envelope (HxB2 nucleotides 5675-8475) was subcloned into the NcoI-BamHI-cloning sites of the pBSn plasmid, resulting in the pAG plasmid. The KpnI-StuI fragment of the HXB2 envelope (nucleotides 6347-6832) was replaced by the KpnI-StuI fragment of the 168.10 envelope, resulting in the pAGX.10 plasmids. Finally, the envelope constructs were transferred into the pLAI backbone by NcoI-BamHI digestion and ligation, resulting in the X.10 infectious molecular clone plasmids.
For all the mutagenesis reactions the QuikChange site-directed mutagenesis kit (Stratagene) was used with the pAG plasmid series as the starting material. All procedures were performed according to the manufacturer's specifications. The modification of the N-linked glycosylation site in the V1 loop was achieved with the primers 5Ј-TGC CAC TAA  TGG TAG CTG GGA AAA GAT GGA AAA AGG-3Ј and 5Ј-CCT TTT  TCC ATC TTT TCC CAG CTA CCA TTA GTG GCA-3Ј, which altered the amino acid sequence of the V1 loop from NSTTNATIGSWE to NSTTYATIGSWE. The modification of the N-linked glycosylation site in the V3 loop was achieved with the primers 5Ј-GAC ATT AAT TGT ACA AGA CCC AAC AAC AAT ATA AGA AAA AGG-3Ј and 5Ј-CCT TTT TCT TAT ATT GTT GTT GGG TCT TGT ACA ATT AAT GTC-3Ј. The amino acid sequence of the V3 loop changed from CTRPNNNTRK to CTRPNNNIRK, and both amino acid sequences were found in primary isolates derived from individual ACH168. The construction of all molecularly cloned viruses and the production of viral stocks were monitored throughout using the technique of standard automated sequencing.
Generating Infectious HIV-1 Viral Stocks-Infectious molecularly cloned HIV-1 viral stocks were generated by transfection of the relevant HIV-expressing plasmid into the human cervical carcinoma cell line C33A. Transfections were performed with 10 g of plasmid DNA using the CaCl 2 precipitation method. All plasmid DNA used for transfection was prepared using Qiagen kits in accordance with the manufacturer's instructions. The DNA precipitate was split between two wells of C33A cells plated 24 h earlier at 1.5 ϫ 10 6 cells/well in a 6-well tissue culture plate. Each transfection was performed in a final volume of 8 ml of Dulbecco's modified Eagle's medium supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 10% fetal calf serum. The following day, the cells were washed with phosphate-buffered saline and given fresh culture media, and the viral stocks were harvested on day 3 of culture in multiple freezing vials after a 0.2 M filtration step (Millipore). All culture supernatants were assayed at the time of freezing for levels of the viral p24 antigen with a standard p24 enzyme-linked immunosorbent assay. For all viruses produced, the levels of p24 were within the range 10 5 -10 6 pg/ml.
Determining Viral Coreceptor Utilization-The coreceptor utilization phenotype was determined by measuring viral replication on the U87.CD4 ϩ cell line expressing an array of different chemokine receptors (CCR1, CCR2b, CCR3, CCR5, and CXCR4), a gift from Dr. Dan Littman, Skirbill Institute, New York. These cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum plus the antibiotics puromycin (1 g/ml) and neomycin (300 g/ml). Coreceptor utilization was determined by adding 200 l of virus stock (between 10 5 and 10 6 p24 pg/ml) to 3.0 ϫ 10 4 U87.CD4 ϩ cells (plated 20 -24 h previously in a 96-well flat-bottomed culture plate) expressing the specific receptor under analysis. Cells were infected for 18 h before being washed twice with phosphate-buffered saline and fed with 200 l of fresh media. On day 10 of the culture, cells were scored for syncytia formation, and the p24 levels in the culture supernatants were determined using a standard enzyme-linked immunosorbent assay.
Determining Viral Infectivity and Replication Phenotype on CD4 ϩ Lymphocytes-All viral stocks were assayed for tissue culture infectious dose (TCID 50 ) on CD4 ϩ -enriched lymphocytes isolated from individuals who did not carry the ⌬32CCR5 allele, screened for by standard polymerase chain reaction technique. PBMCs were isolated from fresh buffy coats Central Laboratory Blood Bank, Amsterdam) by standard Ficoll-Hypaque density centrifugation. PBMCs were frozen in multiple vials at a high concentration and, when required, PBMCs were thawed and activated with 5 g/ml phytohemagglutinin and cultured in RPMI medium containing 10% fetal calf serum, penicillin (100 units/ml), and streptomycin (100 units/ml) with recombinant interleukin-2 (100 units/ ml). On day 4 of culture, the cells underwent CD4 ϩ enrichment by incubating with CD8 immunomagnetic beads (Dynal Oslo, Norway) and separating out the CD8 ϩ lymphocytes. CD4 ϩ -enriched lymphocytes were plated at 2 ϫ 10 5 cells/well in 96-well plates with 5-fold serial dilutions of the virus. The cells were fed on day 7 with fresh media and scored on day 14 for p24 levels, and the number of positive wells were identified. This figure was used to determine the TCID 50 value for each virus. PBMCs from the same donor were preferentially used throughout the study to minimize any potential experimental errors resulting from variation between donor cells. Where the same donor PBMC samples were not available, then cells behaving similarly in their replication capacity for R5 viruses were selected.
The replication kinetics of each virus was measured by infecting CD4 ϩ -enriched lymphocytes from either CCR5 ϩ/ϩ or CCR5 Ϫ/Ϫ individuals at 1,000 -3,000 TCID 50 . The kinetics of virus replication were monitored by measuring p24 antigen levels in the culture supernatants on days 4, 7, and 10 of infection. All experimental results with regard to replication kinetics, unless otherwise stated, are representative of three independently performed experiments.
Electrophoretic Mobility of Virion-associated gp120 Proteins-We tested for the induced N-linked glycosylation pattern variations by following the electrophoretic mobility of virion-associated gp120 proteins. Infectious viral particles were pelleted (18,000 ϫ g for 2 h at 4°C) from 0.5 ml of culture supernatant harvested from transfected C33A cells. The virus pellet was resuspended in 40 l of of lysis buffer (50 mM Tris-HCl, 100 mM 2-mercaptoethanol, 2% SDS, 0.1% bromphenol blue, 10% glycerol) and heated to 95°C for 5 min. The denatured proteins underwent SDS-polyacrylamide gel electrophoresis (PAGE) on 5% gels and transferred to nitrocellulose membrane. After overnight incubation at 4°C with rabbit-anti-gp120 serum followed by a 1-h incubation with a goat anti-rabbit antibody conjugated to alkaline phosphatase, the position of the gp120 glycoprotein on the molecular weight scale was detected by nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate color development.

RESULTS
Effect of V3 Charge on Coreceptor Utilization-Our starting material for this study was a panel of molecularly cloned chimeric viruses that consisted of the HxB2 envelope in the LAI viral backbone. The V3 region of the HxB2 envelope was replaced by the equivalent V3 region from a M-tropic viral isolate (168.1), obtained from a patient early in the disease course (patient ACH168 of the Amsterdam Cohorts Studies) (27). Subsequently, the V3 region of this virus was altered by sitedirected mutagenesis to resemble the amino acid alterations that confer the positive charge changes seen in the V3 region of a late stage disease, T-tropic viral isolate (168.10). Fig. 1A depicts the V3 amino acid sequence differences between the early and late isolates (168.1 and 168.10) and the sequences of the sequentially altered V3 regions, which confer differences in overall positive charge. An additional virus (G) was generated that provided a further ϩ3-charged V3 envelope for this study.
Upon transfection of these viral plasmids into either MT-2 or Sup-T1 cells, the viral phenotype changed from that of M-tropic to T-tropic, as determined by the appearance of syncytia (Fig.  1A). This result suggested that as the positive charge of the V3 envelope increased, the viruses began to utilize the CXCR4 receptor as a coreceptor for viral entry. To test this hypothesis, we measured the infectivity of viral isolates generated by the transfection of the human C33A cell line on U87.CD4 ϩ cells expressing either the CCR5 or CXCR4 receptors. Surprisingly, we found that all viruses were strong users of the CCR5 receptor, and none were capable of utilizing CXCR4 (Fig. 1A). The full-length 168.10 viral envelope cloned into the LAI backbone produced a virus that showed strong CXCR4 usage but that could not utilize CCR5 (Fig. 1A). These results demonstrated that the V3 charge alterations, which could confer syncytia formation upon transfection of MT-2 and Sup-T1 cell lines, could not predict successful CXCR4 utilization and thereby suggested that some other region or factors within the envelope were responsible.
Influence of V1V2 Region on Coreceptor Utilization-Since we were unable to confer CXCR4 utilization onto our virus panel by altering V3 charge alone, we tested what effect the V1V2 region had on viral coreceptor utilization. The differences in the V1V2 region between the early isolate (168.1) and the late virus isolate (168.10) as well as the HxB2 virus can be seen in Fig. 1B. Sequence comparison revealed a strong homology between the V1V2 region of the HxB2 and the 168.1 envelopes, whereas both differed from the 168.10 isolate, with the main difference being the addition of a potential N-linked glycosylation site in the latter. We generated a panel of viruses where we replaced the V1V2 region of our original panel of viruses (termed X) with the V1V2 region of the late isolate, 168.10. We produced 10 new viruses that encompassed the array of different V3 charges (ϩ3 to ϩ6) (termed X.10). The new viruses along with the original (X) viruses were tested for replication on U87.CD4 ϩ cells expressing either CCR5 or CXCR4 and on CD4 ϩ lymphocytes isolated from individuals homozygous for the ⌬32 CCR5 allele (CCR5 Ϫ/Ϫ ) or from those who did not carry this allele (CCR5 ϩ/ϩ ). In this experiment non-diluted viral stocks were used to infect the U87.CD4 ϩ cells, and a 3,000 TCID 50 was used to infect the CD4 ϩ lymphocytes. With the original X viruses (top panel, Fig. 2), no replication was seen on the CCR5 Ϫ/Ϫ CD4 ϩ lymphocytes, indicating no CXCR4 utilization. With some of the V3 higher charged viruses we did observe slight CXCR4 utilization but at less than 0.1% of that seen for CCR5. Surprisingly, the X.10 viruses, irrespective of V3 charge, all demonstrated replication on U87.CD4 ϩ cells expressing CXCR4 or CCR5 and also replicated on CCR5 Ϫ/Ϫ CD4 ϩ lymphocytes (bottom panel, Fig. 2). Interestingly, there was no reduction in the replication of the X.10 viruses on the U87.CD4 ϩ cells expressing CCR5, and actually all showed higher levels of replication opposed to the U87.CD4 ϩ cells expressing the CXCR4 receptor. We therefore conclude that the V1V2 region from the late isolate, 168.10, allows the virus to utilize CXCR4 while not diminishing its ability to use CCR5.
The replication kinetics of the X.10 viruses proved to be identical on both CCR5 ϩ/ϩ and CCR5 Ϫ/Ϫ CD4 ϩ lymphocytes irrespective of the V3 charge. However, the viruses did (closed circles, Fig. 3) show faster rates of replication than the corresponding panel of X viruses on CCR5 ϩ/ϩ lymphocytes (open circles, Fig. 3), indicating that the 168.10 V1V2 region can provide HIV-1 with a faster replicating phenotype irrespective of the charge in the V3 region. Again, the original X viruses showed no replication on CD4 ϩ lymphocytes from individuals lacking a functional CCR5 receptor.
Contribution of V1 and V3 N-Linked Glycosylation Events to Coreceptor Utilization-The panel of viruses with the 168.10 V1V2 region (X.10) were still capable of using CCR5 strongly, even with the high V3 charges, whereas the 168.10 full envelope was not, suggesting that other features within the envelope of the 168.10 isolate were contributing to its solo CXCR4 utilization. We therefore became interested in the V3 amino acid difference between the early 168.1 and late 168.10 isolates that resulted in the loss of a potential N-linked glycosylation site (Fig. 1A). We were also interested in whether the additional N-linked glycosylation site in the V1 region of the 168.10 envelope could contribute to coreceptor utilization since this site has previously been predicted to play a role in determining viral tropism (57). To test the contribution made by these two N-linked glycosylation events to viral coreceptor utilization and replication phenotype, we generated a panel of viruses where these sites were altered by site-directed mutagenesis (⌬gV1 or ⌬gV3). We chose a panel of viruses (G, N, RN, and RTQN together with the equivalent G.10, N.10, RN.10, and RTQN.10 viruses) that covered the array of V3 charges (ϩ3 to ϩ6).
Initially, to test whether the N-linked glycosylation sites were indeed glycosylated or not, we pelleted the viruses and determined the gp120 envelope glycoprotein molecular weight by SDS-PAGE gel electrophoresis and Western blot analysis using a gp120-specific polyclonal antibody. In Fig. 4 we show the result for the N series of viruses, where the envelopes lacking one N-linked glycosylation site are seen to migrate Ϫ, no syncytia) and SupT1 on day 7 (ϩϩϩϩ, very high; ϩϩϩ, high; ϩϩ, intermediate; ϩ, low; Ϫ, nil). Virus stocks generated by transfection of C33A cells were tested for infection on U87/CD4 ϩ cells expressing different coreceptors and on CD4 ϩ lymphocytes isolated from CCR5 ϩ/ϩ and CCR5 Ϫ/Ϫ individuals. Infection was monitored by measuring p24 on day 10 of infection (ϩϩϩϩ Ͼ 10 6 pg/ml; ϩϩϩ, 10 5 -10 6 pg/ml; ϩϩ, 10 4 -10 5 pg/ml; ϩ, 10 3 -10 4 pg/ml; and Ϫ, Ͻ10 3 pg/ml). B, amino acid sequence comparison of the V1 regions for the HxB2 virus and the early (168.1) and late (168.10) isolates from donor ACH168. The additional N-linked glycosylation site in the isolate 168.10 is boxed.
faster. Since we utilized molecular-cloned viruses altered by site-directed mutagenesis, the variation in molecular size shown in Fig. 4 should reflect the alterations to the N-linked glycosylation pattern.
The results for the replication of these viruses on U87.CD4 ϩ cells expressing CCR3, CCR5, and CXCR4 are shown in Fig. 5. For the ϩ3V3 virus (G.10), there is no influence on coreceptor utilization with the presence or absence of V1 glycosylation, but CXCR4 usage does appear to be much weaker than for the CCR5 receptor. For the ϩ4V3 virus (N.10), there is a dramatic decrease in CXCR4 utilization (Ͼ99.5%) when the N-linked glycosylation event in V1 is missing, and furthermore, this virus also showed a reduced replication capacity on CD4 ϩ lymphocytes isolated from a CCR5 Ϫ/Ϫ individual (data not shown). The N.10⌬gV1 virus also gained the ability to use the CCR3 receptor expressed on U87.CD4 ϩ cells. Interestingly, the ϩ5V3 and ϩ6V3 virus, which were lacking the V1 glycosylation event (RN.10⌬g V1 and RTQN.10⌬gV1), showed no reduction in their ability to utilize CXCR4, with the RTQN.10⌬gV1 virus gaining CCR3 usage. These observations show that the degree of glycosylation in the V1V2 region can be critical for coreceptor utilization, but in a context restricted by the charge found in the V3 region.
More dramatic effects were observed with the loss of the V3 N-linked glycosylation event, shown in Fig. 5. The ϩ5V3 virus (RN⌬gV3) could replicate to equal levels on U87.CD4 ϩ cells expressing CCR5 or CXCR4 and had gained the ability to use CCR3, whereas the original RN virus was only capable of using CCR5. The ϩ5V3 virus with the 168.10V1V2 region (RN.10⌬gV3) showed high levels of replication on U87.CD4 ϩ cells expressing CXCR4 while at the same time showing a loss in the capacity to utilize CCR5 (Ͻ2.0% of the RN.10 virus). Remarkably, the loss of the V3 N-linked glycosylation site in the ϩ6V3 virus (RTQN⌬gV3) resulted in the abolition of CCR5 usage, the maintenance of its CXCR4 usage, and a gain in the ability to use CCR3 as a functional receptor for replication (Fig.  5, lower right panel). At the lower V3 charges (ϩ3 and ϩ4), no effect was seen with the alteration of the V3 N-linked glycosylation site (Fig. 5, top two panels). We can therefore conclude that for our panel of viruses, the N-linked glycosylation pattern of the V3 region plays a crucial role in directing coreceptor utilization and coreceptor switching. In accordance, the replication kinetics of these viruses (lacking the N-linked glycosylation site) on CCR5 Ϫ/Ϫ CD4 ϩ lymphocytes demonstrate that the loss in V3 N-linked glycosylation provides for a virus with increased replication when the V3 charge is highest (filled circles in lower panel, Fig. 6A).
We tested further the effect that the glycosylation modifications within the gp120 envelope glycoprotein had on viral phenotype by determining the TCID 50 of each virus separately on CCR5 Ϫ/Ϫ and CCR5 ϩ/ϩ CD4 ϩ lymphocytes and calculating this as a percentage (Fig. 6B). We confirmed that as the charge of the V3 increased, the lack of V3 glycosylation (⌬gV3) significantly enhanced the viral infectivity of CD4 ϩ lymphocytes from CCR5 Ϫ/Ϫ donors, presumably reflecting an enhanced infectivity via the CXCR4 coreceptor. We also observed that both the FIG. 2. Replication of X and X.10 viruses. Virus stocks of the X (without the V1V2 of the late disease isolate 168.10) and X.10 (with the V1V2 of the late disease isolate 168.10) viruses generated by transfection of C33A cells were tested for infection of U87/CD4 ϩ cells expressing either CCR5 or CXCR4 and on CD4 ϩ lymphocytes isolated from CCR5 ϩ/ϩ and CCR5 Ϫ/Ϫ individuals. The U87 cell replication results are shown with error bars based on the S.D. derived from three separate data points included in the same experiment. For replication on CD4 ϩ lymphocytes, the results are depicted without error bars but are representative of three separately performed experiments.
Frequency of the V3 N-Linked Glycosylation Pattern with Relation to V3 Charge-To address the biological relevance of our findings, we compared the amino acid sequences from primary viral isolates taken from the Los Alamos data base and compared the frequency of N-linked glycosylation events in the V1V2 and V3 regions with the V3 amino acid charge. Analyzing the sequences of a large number of viruses (n ϭ 2,562) from different HIV-1 subtypes, we found no correlation between any N-linked glycosylation events in the V1V2 regions and the charge of the V3 loop (data not shown). We did, however, find a strong association between high V3 charge and the lack of the N-linked glycosylation event within the V3 loop for the different subtypes of HIV-1 (A, B, D, and E), but not for subtype C, where the envelope charges remained relatively low, and no de-glycosylation was observed (Fig. 7). This finding is in accordance with the low frequency of SI viruses found in nature for subtype C viruses compared with the high frequency encountered for subtypes D and E. DISCUSSION We have shown for our panel of molecularly cloned viruses that the overall V3 charge is not the solo determinant of viral coreceptor utilization or replication phenotype in vitro but does play a significant role. With our viruses, the V3 charge is only important with regard to these two parameters when placed within the correct V1V2 amino acid context and/or V1 and V3 glycosylation environments. This result is somewhat surprising given the previous findings, where it was demonstrated that the V3 charge was the minimal requirement to be associ- FIG. 4. Electrophoretic mobility of virion associated gp120 envelope molecules during SDS-PAGE. Electrophoretic mobility of virion-associated gp120 molecules during SDS-PAGE. The preparation of virion-associated gp120 molecules from the chimeric RN, RN.10, and glycosylation mutant viruses and SDS-PAGE were performed as described under "Experimental Procedures." No attempt was made to correct for differences in the gp120 content of the samples. ated with the T-tropic phenotype (22,27). We show that the V1V2 region has the capacity to significantly alter coreceptor utilization and generate dual-tropic (R5X4) viruses as well as confer a rapid-growing phenotype on CD4 ϩ lymphocytes. The loss of an N-linked glycosylation event within the V3 region, in association with a high positive charge, can lead to the complete switch of the virus from the R5 to X4 phenotype. The same de-glycosylation event in the V3 region within or out of the context of the V1V2 region is able to render a virus capable of utilizing the CCR3 coreceptor and, again, with a restriction based around the V3 charge. These facts, combined, demonstrate that coreceptor utilization and switching can result from an association between different events in the V1V2 and V3 regions of the envelope and that N-linked glycosylation patterns can be highly significant with regard to coreceptor function.
The effect of the 168.10 V1V2 region was significant with respect to CXCR4 usage irrespective of the V3 charge, whereas de-glycosylation of the V3 region rendered the virus incapable of using CCR5 when the V3 charge was high. Since transmitted viruses are expected to have the lowest V3 amino acid charges, it would be predicted that the alterations in the V1V2 are more likely to be selected earlier in infection for both CXCR4 utilization and faster replication kinetics than the de-glycosylation event of the V3 region. The virus was substantially weakened for CXCR4 utilization when the 168.10 V1V2 region was linked with a lower V3 charge, in conjunction with V3 de-glycosylation, thereby suggesting that the V3 de-glycosylation event is only favorable for viral replication efficiency after the charge of the V3 has risen. The broadening in the coreceptor repertoire of the R5X4 dual-tropic chimeric viruses to include CCR3 by alterations in N-linked glycosylation of the envelope suggests that the viruses with the higher V3 region charges may be in a more structurally open conformation.
We emphasize that we have studied in depth the V1V2 and V3 regions and their interactions with the coreceptor of viral isolates established from a single patient. In other individuals the equilibrium of these interactions may be somewhat different. Nevertheless, our data demonstrate that the V1V2 and V3 regions play a crucial role. A previous study reports that the de-glycosylation of the V3, when associated with a specific V3 backbone sequence, can have a negative effect for CXCR4 utilization (58), and in our study the HxB2 envelope lost its infectivity with the loss of the V3 region N-linked glycosylation site (data not shown). This suggests that in certain cases the de-glycosylation of the V3 can be detrimental for the virus with respect to CXCR4 utilization, but it is worth noting that the ϩ9V3 charge of HxB2 is unusually high and infrequently found in nature.
Our results are in agreement with the concept that the V1V2 region interacts with the V3 region of the envelope to determine coreceptor usage, and we show here that the N-linked glycosylation pattern can determine the effectiveness of this interaction toward coreceptor utilization. An interesting observation in this study has been the finding that upon transfection of the original panel of viruses (X, without the 168.10 V1V2) into either Sup-T1 or MT-2 cells, the SI is seen in relation to the positive charge of the envelope, but infectious virus capable of utilizing CXCR4 is not produced. However, when the viruses were altered by removing the N-glycosylation site in the V3 region, the virus was capable of using this coreceptor. This would indicate that as the charge of the V3 region of the virus becomes more positive, the affinity of the envelope for CXCR4 increases, but the V3 N-linked glycosylation event in the V3 FIG. 5. Replication of the X, X.10 and the glycosylation variant viruses. Virus stocks of the X, X.10, and de-glycosylated (X.10/⌬gV1, X/⌬gV3, and X.10/⌬gV3) viruses in either the V1 or V3 region were generated by transfection of C33A cells and tested for infection of U87/CD4 ϩ cells expressing either CCR3, CCR5, or CXCR4. The V3 charge and amino acid substitutions are indicated for each panel of viruses. The results are shown with error bars based on the S.D. derived from three separate data points included in the same experiment. Due to DNA cloning difficulties, the RTQN.10⌬gV3 virus was not included in this study. region inhibits the virus from being able to utilize this receptor. The maintenance of the N-linked glycosylation event also allows the virus at high charge to maintain CCR5 utilization. This is pertinent given the recent finding that the N-linked glycosylation pattern of the CXCR4 chemokine receptor itself can influence viral coreceptor utilization patterns (59). These results taken together strongly suggest that glycosylation patterns, either envelope-or receptor-specific, are capable of determining coreceptor utilization but not necessarily coreceptor binding and that the interaction of the V1V2 and V3 regions is a dynamic phenomenon in the course of the evolution of the virus.
The study of the kinetics of viral replication on CD4 ϩ -enriched lymphocytes provides for an interesting analysis. The addition of the late 168.10V1V2 region provides for rapidreplicating viruses irrespective of the charge in the V3 region, whereas alterations in V3 N-linked glycosylation results in a virus with a slower rate of replication, although CXCR4 utilization appears stronger. At higher V3 charges the N-linked glycosylation of the V3 seems more significant in determining replication kinetics than the V1V2 region.
A recent study has also suggested that an N-linked glycosylation event in the V2 region has a significant effect on CD4 ϩ / CCR5 receptor interactions and that this can influence the neutralization potential of this virus by specific monoclonal antibodies (60). Additionally, with simian immunodeficiency virus it has been shown that de-glycosylation events within the V1 region of the envelope result in a virus with increased antibody neutralization potential (61). With HIV-1, the N-linked glycosylation pattern has also been shown to have a varied effect on the induction of neutralizing antibodies (44,(62)(63)(64)(65)(66)(67). It is therefore possible that the selection of viruses with different coreceptor utilization patterns and replication phenotypes, based upon altered envelope glycosylation events, may be aided by the escape from a neutralizing antibody response. The emergence of CXCR4 viruses late in disease and the concurrent rapid rise in viral loads seen in a proportion of infected individuals may merely reflect viral escape from a controlling neutralizing antibody response influenced by the N-linked glycosylation pattern of the envelope. It is likely that a combination of events encompassing altered viral replication phenotype and immune evasion results in the rapid rise in viral load and progression to disease. Alterations in glycosylation events influencing coreceptor utilization will also allow the virus to escape the controlling responses of the CC chemokines in vivo, which are believed to play an effective role in inhibiting viral replication and slowing disease progression (68). The strong association observed between the V3 charge and the loss of the V3 N-linked glycosylation site for primary viral isolate sequences suggests that the V3 de-glycosylation event may be worth targeting to prevent the virus switching toward using the CXCR4 and CCR3 coreceptors and the patient ultimately progressing to disease.
Subtype C envelopes have low V3 charges and a high frequency of V3 glycosylation, which is in support of reports describing a low frequency of SI isolates among subtype C-infected individuals even when the patients had low CD4 ϩ cell counts and had progressed to AIDS (69). In accordance, HIV-1 subtypes D and E demonstrate the highest frequency of V3 de-glycosylation, even at the lower V3 charges, and both these subtypes reportedly have a high frequency of SI viruses among infected individuals (71,72). It will be interesting to address why subtype C viruses keep a low charge in the V3 region, whereas rates of disease progression do not appear to be different nor do the viral loads seem to be lower than for the other subtypes (70). We can speculate that the low V3 charge maintenance, hence CCR5-using phenotype, in combination with high viral load may help explain why subtype C viruses are spreading rapidly throughout the world. Structural constraints in the envelope or biological constraints on the virus may keep subtype C with the more optimal CCR5 phenotype for efficient transmission. The elucidation of such mechanisms will be important for the better understanding of HIV-1 disease pathogenesis.