The Two Human CXCR4 Isoforms Display Different HIV Receptor Activities: Consequences for the Emergence of X4 Strains

The chemokine receptor CXCR4 is vital for embryonic development and plays key roles in myelopoiesis, homeostasis, maintenance of the immune system, and cancer progression (1). Mutations in CXCR4 that increase its responsiveness to the chemokine ligand CXCL12 result in an immunodeficiency called WHIM (warts, hypogammaglobulinemia, infections, myelokathexis) syndrome (2). One consequence of increased agonist-induced CXCR4 signaling in this syndrome is an impaired egress of leukocytes, particularly neutrophils, from the bone marrow, causing neutropenia and, possibly, leukopenia.

Together with the chemokine receptor CCR5, used by HIV-1 R5 strains, CXCR4 is the coreceptor for dual tropic HIV-1 R5X4 (CCR5- and CXCR4-using) and X4 (CXCR4-using) strains, in addition to the main receptor CD4 (3). HIV-1 R5 strains predominate in the early stages of disease, and the proportion of CXCR4-using strains increases with the duration of infection (4). Of note, the emergence of CXCR4-using strains correlated with an acceleration of disease progression (3). One factor that could favor this R5 to X4 switch is the increase in CD4⁺ T cell CXCR4 surface density observed in about one third of HIV-1–infected individuals with a CD4 count < 400 cells/μl (5). This is demonstrated when CXCR4 is overexpressed at the surface of HIV-1 R5–infected cells, facilitating the emergence of CXCR4-using HIV-1 strains (6, 7). Moreover, patients with a high CD4⁺ T cell surface CXCR4 density harbor (R5)X4 strains more frequently than do patients with physiological levels of CXCR4 expression (5). CXCR4 overexpression observed in advanced disease could be mediated, in part, by IL-7, which is overproduced as a consequence of CD4⁺ T cell lymphopenia (6, 8).

Although CCR5 antagonists have been successfully developed as anti-HIV drugs (9), the long-term use of CXCR4 antagonists is limited by the physiological roles played by CXCR4 (10). CCR5 and CXCR4 antagonists usually inhibit the binding of their natural ligands to CCR5 and CXCR4; however, the role of CCR5 is dispensable, whereas the role of CXCR4 is not, over the long term.

Two human isoforms of CXCR4 have been described. They are coexpressed in PBMCs (11). The main isoform, CXCR4-B, is coded by a spliced mRNA (Fig. 1). The long variant (CXCR4-A) is coded by an unspliced mRNA containing a distinct 5′ untranslated region and a coding region initiated from an alternative start codon (Fig. 1). The CXCR4-A receptor has a longer NH2-terminal extremity (N-term) that is different from the CXCR4-B isoform by the first 9 aa residues (Fig. 1).

Because CXCR4 N-term was shown to be involved in HIV-1 X4 binding (12–14), we compared the capability of the both the A and B isoforms to support HIV-1 infection and chemotaxis.

CD4⁺CCR5⁺ HOS, HEK-293T, CD4⁻CCR5⁻ HeLa, and CD4⁺CXCR4⁺ HeLa-P4 cells were cultivated in DMEM supplemented with 10% FCS, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (Lonza) at 37°C and 5% CO₂. CD4⁺CCR5⁻CXCR4⁺ CEM cells, CD4⁺CCR5⁻CXCR4⁺ MT2 cells, and human PBMCs were cultivated in RPMI 1640 supplemented the same as DMEM and under the same conditions. Cell viability was quantified after staining with 0.4% trypan blue (Sigma) with an automated cell counter (MACSQuant; Miltenyi Biotec) using cell-counting chamber slides (Invitrogen).

The CXCR4 cDNA was obtained via the AIDS Research Reagent Program (Rockville, MD) from Dr. Nathaniel Landau (New York University School of Medicine, New York, NY). It was amplified to give CXCR4-A (corresponding to PubMed sequence NM_001008540 from nucleotide 297 to 1375) or CXCR4-B (corresponding to PubMed sequence NM_003467 from nucleotide 36 to 1154). Each cDNA was inserted downstream of the CMV promoter in the lentiviral transfer vector pHR-BX (15). These CXCR4 transfer vectors were cotransfected with the HIV packaging plasmid pAX2 and the plasmid pMD2G, which codes for the vesicular stomatitis virus envelope glycoprotein G, in HEK-293T cells by the calcium phosphate method. Supernatants were collected at day 2 posttransfection and concentrated on sucrose (Sigma) by ultracentrifugation at 17,000 × g for 1.5 h at 4°C. HOS or HeLa cells were plated in 24-well culture plates and exposed to the same amount of the p24 equivalent of HIV vectors harboring the CXCR4-B or CXCR4-A cDNA in 8 μg/ml Polybrene (Sigma) and centrifuged at 200 × g for 90 min at 30°C. Cells were washed 24 h posttransduction, amplified, and controlled for CXCR4 membrane expression by flow cytometry.

To produce replication-defective HIV virions, pNL4.3-Luc.R⁻.E⁻ transfer plasmid harboring a defective viral env gene with a firefly luciferase gene inserted in the viral nef gene was cotransfected with the R5 envelope plasmid pCMV-Ad8-Env, the X4 envelope plasmid VB34, or the vesicular stomatitis virus envelope glycoprotein G envelope plasmid pDM2G (AIDS Research Program) into HEK-293T cells at a molecular ratio of 2:1. The virions produced were collected in the culture supernatant 48 h posttransfection, filtrated at 0.45 μm, and concentrated on sucrose by centrifugation. For the single-round infection assay, HOS cells were cultivated in triplicates in 96-well culture plates at a density of 0.25 × 10⁶ cells/ml and infected with the various virions. In some experiments, cells were incubated for 1 h with CXCL12 (PeproTech) at various concentrations before infection. Cells were washed twice in PBS 24 h postinfection and kept in culture for 72 h. Cells were washed once in PBS and lysed with 50 μl lysis buffer, and the firefly luciferase activity was measured on a luminometer using a commercial kit (Luciferase Assay System; Promega).

The HIV-1 R5 Ad8 strain (AIDS Reagent Program) was expanded in CD4⁺CCR5⁺ HOS cells. The HIV-1 X4 NL4.3 DNA (AIDS Reagent Program) was transfected into HEK-293T cells by the calcium phosphate method, and the collected virions were amplified in the human CD4⁺CXCR4⁺ CEM cell line. HOS cells were plated in 24-well culture plates at a density of 0.04 × 10⁶ cells/ml. Cells were infected in triplicates with 10 ng/ml the p24 equivalent of Ad8 or NL4.3 overnight, washed with PBS, and kept in culture; cell number was adjusted twice a week. Virus production was monitored by measuring p24 gag concentration in the cell supernatant using a commercial ELISA (INNOTEST HIV Ag mAb; Ingen).

Reverse transcripts were quantified as previously described (16).

mRNA from 5 × 10⁶ PBMCs was extracted using a commercial kit (High Pure RNA Isolation Kit; Roche). Reverse transcription was carried out using random primers (Promega) and Superscript III reverse transcriptase (Invitrogen) at 50°C for 30 min, 55°C for 30 min, and 70°C for 15 min. cDNA was amplified with the appropriate primers—5′-CTTGCTGAATTGGAAGTGAATG-3′ and 5′-GGTGGGCAGGAAGATTTTATTG-3′ for CXCR4-A and 5′-CAGCAGGTAGCAAAGTGACG-3′ and 5′-ATGGAGTCATAGTCCCCTGAGC-3′ for CXCR4-B—in a LightCycler 480 (Roche) with SYBR Green, following the manufacturer’s recommendations. Cellular genomic S14 (5′-ACCAGTCACACGGCAGATG-3′ and 5′-GGGGAAGGAAAAGAAGGAAGAA-3′) was used for normalization. A standard curve was established by analyzing serial dilutions of a positive-control plasmid. The PCR cycle at which the amplification signal entered the exponential range was used to quantify the cellular DNA. CXCR4-A and S14 annealing were realized at 64°C, and CXCR4-B annealing occurred at 69°C. To quantify mRNA in infected cells, 5 × 10⁶ PBMCs, at a density of 2 × 10⁶ cells/ml, were exposed for 18 h to 350 ng p24 equivalent of the R5 strain Ad8 or the R5X4 strain 89.6, washed twice in PBS, and cultivated for 4 d.

HeLa cells were cultured in a 24-well plate at a density of 0.025 × 10⁶ cells/ml and transfected with 2 nM the anti–CXCR4-B small interfering RNA (siRNA) 5′-GCAGCAGGUAGCAAAGUGA-3′ or a nontargeting pool of siRNA (Thermo-Scientific) with INTERFERin (Polyplus Transfection). Three days posttransfection, cells were infected with 10 ng/ml p24 equivalent of the X4 strain NL4.3 for 24 h, washed in PBS, and cultured; cell number was adjusted twice a week. siRNA transfection was repeated once a week.

For CXCR4 staining, 2 × 10⁵ cells were incubated in PBS supplemented with 0.2% BSA (Sigma) and labeled with the anti-human CXCR4 mAb 12G5 or an isotype control (BD Biosciences) for 1 h on ice at a final concentration of 10 μg/ml. After washing, cells were incubated with a 1:100 dilution of Goat F(ab′)₂ fragment anti-mouse IgG (H+L)-FITC (Beckman Coulter) for another hour on ice. Cells were then washed, fixed in BD Cell Fix, and analyzed on a FACSCalibur flow cytometer (BD Biosciences).

Chemotaxis of HOS cells in response to CXCL12 was measured across 8-μm-pore polycarbonate filters in 24-well Transwell chambers, as previously described (17). HOS–CXCR4-A or HOS–CXCR4-B cells were added to the upper chamber of each Transwell (0.1 × 10⁶ cells in 100 μl 1% FCS RPMI 1640). CXCL12, at various concentrations, was added to the lower chamber under a volume of 600 μl. The plates were incubated for 4 h at 37°C in 5% CO₂. Cells that had migrated into the lower chamber were collected and enumerated with a cell counter (MACSQuant).

PBMCs from healthy donors (2 × 10⁶ cells/ml) were activated with 1 μg/ml PHA-M (Sigma) and 100 IU/ml IL-2 (PeproTech) for 3 d in culture medium, exposed to the plasma of an HIV-1 R5–infected subject, and cultured in 100 IU/ml IL-2. Then, we performed a coculture between these infected PBMCs and HOS–CXCR4-A or HOS–CXCR4-B cells. HOS cell number was adjusted twice a week by the addition of uninfected HOS cells. Viral production was monitored by measurement of gag p24 in cell supernatant twice a week. R5 to X4 switch was evaluated using MT2 cells cultured at a density of 0.1 × 10⁶ cells/ml in infected HOS–CXCR4-A and HOS–CXCR4-B cell supernatants, as previously described (7).

All data are representative of at least three different experiments. Values are expressed as the mean ± SD. Differences in infection, chemotaxis, LTR expression, and CXCR4 isoforms mRNA ratio were analyzed with the unpaired Student t test. Spearman rank correlations were used to evaluate the link between CXCR4-B/CXCR4-A mRNA ratio and viremia or false-positive rate (FPR).

The observational study involving patients was approved by the institutional review board Sud Méditerrannée IV. Written informed consent was obtained from all participants.

To study the characteristics of the two isoforms of CXCR4 (Fig. 1), we cloned the cDNA corresponding to each isoform in an HIV-1 vector and transduced CD4⁺CCR5⁺ HOS cells. Thus, we obtained two cell lines, HOS-CXCR4-A and HOS-CXCR4-B, which expressed the same cell surface density in each isoform (mean fluorescence intensity of 20 and 24, respectively, Fig. 2A–C). This remained stable over time. Thereafter, we evaluated the ability of both isoforms to support HIV-1 infection. With the aim of comparing one-round X4 infection in the two cell lines, we first exposed them to HIV-1 virions harboring a genome with a deletion in the env gene, a luciferase gene fused to the nef gene, and pseudotyped with an X4 envelope. Fig. 2D shows that, after a single cycle, HIV-1 expression, as quantified by luciferase activity, was higher in HOS–CXCR4-B cells than in HOS–CXCR4-A cells (luciferase activity of 93,559 ± 20,314 and 21,418 ± 7,568 arbitrary units, respectively [p = 0.029], for a viral input of 20 ng; luciferase activity of 177,473 ± 10,453 and 88,405 ± 15,607 arbitrary units, respectively [p = 0.009], for a viral input of 50 ng; and luciferase activity of 481,896 ± 35,250 and 152,031 ± 11,503 arbitrary units, respectively [p < 0.001], for a viral input of 100 ng). When we performed the same experiment with virions pseudotyped with an HIV-1 R5 envelope (Fig. 2E) or the G protein of the vesicular stomatitis virus (Fig. 2F) instead of an HIV-1 X4 envelope, no difference in viral production was observed between the two cell lines. This difference was accentuated when we exposed the two cell lines to the replicative CXCR4-using strain NL4.3 (Fig. 1G), to three other laboratory X4 strains from different subtypes, or to a primary X4 strain (Fig. 2I). In contrast, a wild-type R5 strain replicated as efficiently in HOS–CXCR4-B cells as in HOS–CXCR4-A cells (Fig. 2H).

We wanted to know whether the response of the two isoforms of CXCR4 to their natural ligand, the chemokine CXCL12, was similar. First, we compared the ability of CXCL12 to inhibit X4 infection in both cell lines. Fig. 3A shows that preincubation of HOS–CXCR4-A and HOS–CXCR4-B cells with a suboptimal concentration of CXCL12 inhibited viral production with the same efficiency (inhibition of 53 ± 10% and 56 ± 6%, respectively, p = 0,810). Second, because CXCR4 activation was shown to induce HIV gene expression (18), we analyzed the effect of CXCL12 binding to each CXCR4 isoform on LTR activation. To this end, we transduced CD4⁻CCR5⁻ HeLa cells stably transfected with a reporter luciferase gene driven by the HIV-1 LTR with CXCR4-A or with CXCR4-B cDNA, exposed them to 1 μg/ml of CXCL12, and measured luciferase activity 24 h later. As shown in Fig. 3B, LTR activity was enhanced to a similar level upon CXCL12 exposure in both cell lines (3.33 ± 0.51-fold and 3.27 ± 0.8-fold, respectively, p = 0.953). Third, we measured the chemotactic effect of CXCL12 on both cell lines. HOS–CXCR4-A and HOS–CXCR4-B cells migrated toward increasing concentrations of the chemokine in a similar fashion (1448 ± 418 and 2266 ± 291 cells, respectively [p = 0.249], at a CXCL12 concentration of 25 ng/ml; 4527 ± 314 and 4323 ± 177 cells, respectively [p = 0.629], at a CXCL12 concentration of 100 ng/ml; and 6545 ± 389 cells and 6417 ± 891 cells, respectively [p = 0.907], at a CXCL12 concentration of 400 ng/ml, Fig. 3C). These findings indicate that, although CXCR4-A is used less efficiently than CXCR4-B by X4 virions to infect CD4⁺ cells, the responses of both isoforms to CXC12 are undistinguishable in terms of inhibition of HIV infection, LTR activation, and chemotaxis.

The CXCR4 N-term region is involved in X4 envelope binding to CXCR4 (12–14). Because the two CXCR4 isoforms do not have the same sequence at this extremity, we tested whether this resulted in a difference in X4 infectivity due to entry kinetics. To test this hypothesis, we exposed both cell lines to a wild-type X4 strain in the presence or absence of 1 μM of the CXCR4 antagonist AMD3100, which is known to inhibit the binding of X4 envelopes to CXCR4. As expected, 72 h postinfection the HOS–CXCR4-B cells produced more virus (72.7 ± 0.6 ng/ml) than the HOS–CXCR4-A cells (32.9 ± 0.5 ng/ml, p < 0.001), and cells preincubated with the CXCR4 antagonist produced no virions (Fig. 4A). At 17 h postinfection, we used quantitative PCR to measure the amount of early reverse transcripts harbored by the cells. As shown in Fig. 4B, nearly twice as many early reverse transcripts were detected in HOS–CXCR4-B cells (2.39 ± 0.17 arbitrary units) than in HOS–CXCR4-A cells (1.52 ± 0.05 arbitrary units, p = 0.007). We also quantified late reverse transcripts in both cell lines and noted that the difference persisted but did not increase (p = 0.039, Fig. 4C). Neither early nor late reverse transcripts were amplified in AMD3100-treated cells (Fig. 4B, 4C). We concluded that the CXCR4-B isoform allowed more efficient X4 virion entry into target cells than did the CXCR4-A isoform.

The level of physiological expression of the two CXCR4 isoforms was measured by quantitative RT-PCR in the PBMCs of healthy donors. Fig. 5A shows that the PBMCs of all of the volunteers expressed both isoforms, the proportions of CXCR4-A and CXCR4-B forms being 10–30% and 70–90%, respectively. Next, we studied the effect of R5 infection on this expression. For this purpose, we exposed the nonactivated PBMCs of a healthy donor to an R5 strain. At day 6, we detected viral production in the culture supernatant. At that time, R5 infection induced an increase in the CXCR4-B/CXCR4-A mRNA ratio in the PBMCs (Fig. 5B). Concurrently, we infected the same PBMCs with the R5X4 strain 89.6. R5X4 infection also induced an increase in the CXCR4-B/CXCR4-A mRNA ratio (Fig. 5B). We then quantified the CXCR4-B/CXCR4-A mRNA ratio in the PBMCs of 19 individuals infected with R5 strains, either under antiretroviral therapy or not. We observed a correlation between this ratio and viral load of these patients (r = +0.601, p = 0.054, Fig. 5C). Thus, R5 infection favors CXCR4-B expression over CXCR4-A expression in vitro, as well as in vivo.

CXCR4 overexpression facilitates X4 replication (7). Moreover, we showed that CXCR4 overexpression at the surface of target cells infected with an R5 strain favors the emergence of X4 strains (6, 7). In line with these observations, because X4 strains replicate more easily in CXCR4-B⁺ cells than in CXCR4-A⁺ cells, we questioned whether the presence of CXCR4-B at the surface of R5-infected cells could facilitate R5 to X4 switch compared with CXCR4-A. To answer this question, we exposed HOS–CXCR4-A and HOS–CXCR4-B cells to an R5 strain and monitored the appearance of CXCR4-using strains in the cell supernatant. As shown in Fig. 6A, X4 strains, identified by their ability to induce syncytia in CD4⁺CCR5⁻CXCR4⁺ MT2 cells, emerged more rapidly in the HOS–CXCR4-B culture than in the HOS–CXCR4-A culture. The overexpression of CXCR4-B in the course of R5 infection might increase the risk for emergence of CXCR4-using strains in vivo. A surrogate marker of the risk for R5 to X4 switch is the FPR score, calculated by the Geno2pheno algorithm using the sequence of the V3 region of the envelope. The FPR corresponds to the risk for falsely predicting as R5 a virus that is, in fact, X4. Recently, the FPR score was shown to be correlated with the percentage of X4 strains detected by ultradeep sequencing (19). Therefore, we measured the CXCR4-B/CXCR4-A ratio of mRNA in the PBMCs of 11 R5-infected individuals and compared it with their FPR score. As shown in Fig. 6B, we observed a strong inverse correlation between these two parameters (r = −0.633, p = 0.039).

Because both CXCR4 isoforms are expressed in CD4⁺ T cells and mediate chemotaxis toward CXCL12, whereas only the B isoform supports a highly productive HIV-1 X4 infection, the specific downregulation of CXCR4-B could be envisaged as an anti–HIV-1 X4 strategy. To test this possibility, we designed an siRNA complementary with a sequence in CXCR4-B that is absent in CXCR4-A and transfected it into CD4⁺ HeLa-P4 cells that coexpress both CXCR4 isoforms. The siRNA resulted in downregulation of the CXCR4-B mRNA (from 13.8 ± 0.1 to 5.2 ± 0.1 arbitrary units, p < 0.001, Fig. 7B), whereas CXCR4-A remained unaffected (Fig. 7A). Consequently, we observed a decrease in cell surface CXCR4 density (Fig. 7C–E). Specific inhibition of the CXCR4-B isoform resulted in a drastic decrease in X4 infectability of HeLa-P4 cells, as shown by the difference in viral production at day 7 postinfection (764 ± 109 versus 266 ± 44 ng/ml, p = 0.013) and day 11 postinfection (1448 ± 181 versus 283 ± 100 ng/ml, p = 0.005, Fig. 7F).

In the current study, we provide evidence that the B isoform of CXCR4 is used by HIV-1 strains to enter and productively infect target cells much more efficiently than the A isoform, whereas both isoforms are used with the same efficiency by the natural ligand CXCL12; HIV-1 R5 infection favors CXCR4-B expression over that of CXCR4-A, a phenomenon that facilitates the emergence of CXCR4-using HIV-1 strains; and it is possible to downregulate CXCR4-B mRNA without modifying CXCR4-A mRNA expression.

Our observation that the chemotactic response of CXCR4-A⁺ and CXCR4-B⁺ cells to CXCL12 is similar is in contradiction to the data reported by Gupta and Pillarisetti (11). One possible explanation for this discrepancy is a technical difference; they worked with stably transfected rat basophil leukemia cell lines, whereas we worked with human cells transduced with the open reading frames of each CXCR4 isoform. Stable transfectants need to be selected, and this selection may randomly isolate clones deficient for characteristics other than the one for which they are being selected. In Gupta and Pillarisetti’s study (11), no positive control verified that the chemotactic defect observed in the cells stably transfected with the CXCR4-A open reading frame was specific. In contrast, as shown in Fig. 2, all of the cells that we transduced expressed CXCR4, and we did not need to perform any selection to obtain cells expressing one or the other CXCR4 isoform. Moreover, we confirmed this result with two other assays in which we measured the ability of each CXCR4 isoform to mediate the effect of CXCL12 on HIV-1 infection and HIV-1 LTR activation.

The gp120 and CXCL12 binding sites overlap but remain distinct, and the involvement of the N-term in CXCR4 coreceptor activity is the present paradigm. Substitution of the CXCR4 N-term with the one derived from CCR5 impairs the fusion capability of X4 strains (12). Moreover, Tyr-7, Tyr-12, and Tyr-21 (13), as well as Glu-15 and Glu-32 (14), play a role in CXCR4-mediated viral entry. In contrast, Doranz et al. (20) reported that the first 27 residues of CXCR4 are not necessary for activation by CXCL12.

Interestingly, the induction of CXCR4-B expression over that of CXCR4-A as a consequence of HIV-1 R5 infection could be a factor that facilitates the emergence of CXCR4-using strains. Combined with the increase in CD4⁺ T cell surface CXCR4 density observed in one third of the patients during the course of the infection, this preferential expression of the CXCR4 isoform, which is the most efficient as an HIV-1 coreceptor, could favor the expansion of X4 strains. The increase in CD4⁺ T cell surface CXCR4 density and in CXCR4-B expression over that of CXCR4-A could promote X4 emergence via a common mechanism. This mechanism is probably a higher amplification of the rare HIV-1 strains able to use CXCR4 as a coreceptor, that randomly appear as a consequence of mutations in the envelope gene of R5 strains rather than via the induction of these mutations. Of note, the increase in CXCR4-B/CXCR4-A ratio under R5 infection is clearly not limited to infected cells. If this were the case, and because <1% of the circulating CD4⁺ T cells are infected in vivo, we could not have detected any correlation between this ratio and HIV RNA plasma level. The molecular mechanisms responsible for the variability in CXCR4-B/CXCR4-A ratio among the tissues, as well as for its increase under HIV-1 R5 infection, remain to be elucidated.

The discrepancy between the dual nature of CXCR4 as an HIV-1 coreceptor and its chemokine receptor functions opens a potential new therapeutic strategy. The use of CXCR4 antagonists in HIV-1 infection has been limited by the fact that they blocked CXCL12 function. Using a complementary siRNA, we were able to specifically downregulate CXCR4-B expression. By transducing hematopoietic stem cells with a transgene encoding such a short hairpin RNA, together with another transgene coding for CXCR4-A, it will be possible to generate cells that specifically express the CXCR4-A isoform, are partly resistant to HIV-1 X4 infection, and are able to respond to CXCL12 signals. Such a strategy could be combined with the gene-therapeutic approaches that are currently being tested to make target cells resistant to HIV-1 R5 infection.

We thank Martine Peeters and Eric Delaporte for the gifts of HIV strains MP577, MP578, and MP1417, Pierre Delobel for the gift of the R5X4 strain, and Brigitte Montès for the gift of the primary X4 strain. We are grateful to the individuals who volunteered for this study.

The authors have no financial conflicts of interest.