We investigated how parental and VVC-resistant HIV-1 variants enter primary CD4+ T cells and U87-CD4-CCR5 cells via naturally expressed and stably transfected CCR5, respectively. We used activated CD4+ T cells because, unlike unstimulated cells, they are highly permissive to HIV-1 replication. We first used a panel of 10 MAbs to different CCR5 epitopes and four viruses that recognize CCR5 in different ways and then conducted additional analytical studies with a subset of the more diagnostically useful MAbs. Our principal conclusion is that CCR5 is present on the cell surface in multiple antigenic forms that VVC-sensitive and -resistant clones use for entry with a range of efficiencies. The compositions of these CCR5 subpopulations differ between primary CD4+ T cells and cell lines (U87-CD4-CCR5 and HOS-CCR5-GFP) and, presumably, also vary on other cell types. We discuss here what these subpopulations might be, where they may be located, and how they may be used differently by VVC-sensitive and -resistant viruses.
In the flow cytometry studies, the plateau level of binding did not reflect the affinity of a MAb for CCR5. For example, PA10 has a low affinity for CD4+
T cells but a high maximal binding level; conversely CTC5 binds with high affinity but to only a low level. In general, the CCR5 affinities measured on CD4+
T cells and U87-CD4-CCR5 cells correlated well, but the maximal binding levels did not. These findings are compatible with a model in which some MAbs preferentially detect distinct CCR5 subpopulations that have the same antigenic determinants on the two cell types but different absolute and relative abundances. Genetic and related host factors seem, however, to have a limited influence on the prevalences of the various antigenic forms of CCR5, as only moderate differences in MAb staining levels were found among four CD4+
T cell donors (see Table S1 in the supplemental material). MAb 2D7 is commonly used for CCR5 detection and quantification, but it had only intermediate reactivity with CCR5 on both cell types. Using 2D7 might underestimate the total CCR5 cell surface expression level under some conditions. For example, some apparently CCR5-negative cells, including primary lymphocytes, do express sufficient CCR5 to be targets for HIV-1 infection (11
PA11, PA14, and 2D7 were active against more than one VVC-sensitive and -resistant virus in one or both cell types, while 45531 and CTC5 were more selectively active, each inhibiting only one VVC-resistant virus strongly in a cell-type-specific manner. Overall, although affinity is likely to be important for the antiviral activity of a CCR5 MAb, it was not the only determinant, and the extent of cell surface binding did not predict whether a MAb was inhibitory. For example, CTC5 and 45531 have higher or comparable affinities for epitopes contiguous to those for PA11 and 2D7, respectively, but they clearly have a narrower ability to inhibit infection. Both the VVC-resistant clones were more sensitive than the parental viruses to PA11, PA14, and 2D7 in CD4+ T cells and to PA14 and 2D7 in U87-CD4-CCR5 cells. Overall, the emergence of mutant viruses that are panresistant to small-molecule CCR5 inhibitors but hypersensitive to CCR5 MAbs against various epitopes is compatible with the escape phenotype involving a reduction in the efficiency of the Env-CCR5 interaction. To compensate, there may be a requirement for a greater number of MAb-unencumbered CCR5 molecules for triggering the fusion reaction. This scenario would be consistent with the lower MAb inhibitory concentrations, and therefore lower minimal MAb-CCR5 occupancies, that we have observed with VVC-resistant viruses compared to their parents (; see Table S2 in the supplemental material). Additional studies will be required to test this hypothesis.
As noted, some MAbs inhibited infection in a cell-type-dependent manner. For example, Res-4V3 was highly sensitive to PA11 in CD4+
T cells but only minimally inhibited in U87-CD4-CCR5 cells. As sulfated tyrosines at residues 10 and 14 are critical both for the binding of some NT MAbs (e.g., PA10 and PA11, but not PA8) and for HIV-1 entry, it is possible that cell-type-dependent variation in the expression of partially and fully sulfated forms of CCR5 might influence PA10 or PA11 staining (12
). If, for example, CCR5 overexpression in U87-CD4-CCR5 cells saturates the cellular sulfotransferases, there may be proportionately more partially sulfated or nonsulfated CCR5 variants on the surfaces of these cells than on CD4+
T cells (15
). Another consideration is that, according to our preliminary observations, PA11 can induce CCR5 internalization into U87-CD4-CCR5 cells, but not CD4+
T cells (whereas PA10 downmodulates CCR5 in both), thereby reducing the number of coreceptors available for HIV-1 entry.
Res-4V3 was previously shown to be more dependent than its parent on specific residues in the CCR5 NT (7
). We now show that it is the virus most sensitive to CTC5, which blocks an NT epitope. Other CCR5 inhibitor-resistant viruses are also known to become highly sensitive to CTC5, which may be a corollary of a weakened interaction between gp120 and ECL2 (28
). PA11 also recognizes an NT epitope, and while it inhibited all four of the test viruses to different extents in CD4+
T cells, it was most active against Res-4V3. Using a competition binding assay, we found that preincubation of U87-CD4-CCR5 or CD4+
T cells with a saturating concentration of PA11 had no effect on CTC5 binding, implying that the two MAbs bind either nonoverlapping epitopes within the same CCR5 protein or nonoverlapping antigenic forms of CCR5 (see Fig. S4B in the supplemental material). CTC5 binding to the NT seems to be less dependent than PA11 on tyrosine sulfation but more dependent on CCR5 having a native conformation. Thus, PA11 binds to a linear peptide spanning amino acids 1 to 22 of the CCR5 NT, and more strongly if the peptide is sulfated on residues Tyr-10 and -14 (12
) (see Fig. S3 in the supplemental material). In contrast, CTC5 does not bind the NT peptides but is highly dependent on residue Asp-2 in the context of native CCR5 (see Fig. S3 in the supplemental material). Whether PA11 and CTC5 recognize the same or different CCR5 antigenic variants is now under investigation.
We used microscopy to identify where CCR5 conformational variants were located on the cell surface. PA14-reactive CCR5 molecules were homogeneously distributed over most of the cell surface, while 45531 staining was concentrated in distinct clusters that were not generally costained by PA14. In addition, alone among the test MAbs, 45531 binding was significantly decreased by cholesterol depletion. Cumulatively, these observations suggest that 45531 recognizes CCR5 variants that are concentrated in cholesterol-rich membrane subdomains or lipid rafts. Palmitoylation of the C-terminal tail reportedly influences whether CCR5 is located within lipid rafts, which are considered to be sites where signaling components are also enriched (10
). The interactions of CCR5 intracellular domains with various signaling or endocytic adaptor proteins, in a cell-type-dependent manner, could affect the conformation of external domains and hence their recognition by MAbs and HIV-1.
G-protein-coupled-receptors, including CCR5, can acquire multiple conformations as they transit from the signaling-active, agonist-bound conformation to the more stable, antagonist-bound, and signaling-inactive conformation. Between these two ends of the spectrum lie multiple intermediate conformations corresponding to various levels of activation and stability (58
). Most of the structural information on CCR5 is based on the rhodopsin crystal structure. Rhodopsin transitions from the inactive to the active state as a result of a tilt in transmembrane helix 6 that leads to a rearrangement of the overall structure (1
). Whether MAb 45531 binds to a signaling-active conformation of CCR5 remains to be determined, although there is some indirect support for this scenario. The binding of small-molecule inhibitors interferes with the binding of only a few CCR5 MAbs, a prominent example being 45531 (19
). We found that the CCR5 binding of 45531, but not other MAbs, such as 2D7 and PA14, was significantly reduced when VVC was added (see Fig. S4A in the supplemental material). VVC and other small-molecule inhibitors all bind within the hydrophobic cavity formed by the transmembrane domains of CCR5, but in subtly different ways (17
). Moreover, all these inhibitors stabilize distinct conformations of CCR5 with different abilities to activate G proteins. In other words, some inhibitors are more effective than others at stabilizing the G-protein-uncoupled, inactive states of CCR5 (17
). Nonetheless, irrespective of precisely how the various inhibitors bind to CCR5, they all seem to stabilize the receptor in a conformation that is not efficiently recognized by 45531 but is still visible to other MAbs, such as 2D7, PA14, and CTC5 (19
). Whether 45531 binds to signaling-active CCR5 molecules is now being directly studied using pharmacological agents that block various signaling pathways downstream of CCR5, as well as a panel of CCR5 mutants with altered signaling properties.
We propose the existence of a CCR5 conformational gradient in which inactive conformations bind PA14, signaling-active conformations bind 45531, and intermediate conformations bind both MAbs, as well as 2D7, to different extents (). U87-CD4-CCR5 cells overexpress transfected CCR5, which may be in excess over G proteins, but they do not produce chemokine ligands for CCR5 (9
). A considerably larger pool of CCR5 uncoupled from G proteins, and with a higher affinity for small-molecule inhibitors (17
), may therefore exist in U87-CD4-CCR5 cells than in activated CD4+
T cells (). The extent to which CCR5 is incorporated into lipid rafts is likely to correlate with the degree of signaling activity; both processes would conceivably be less common in U87-CD4-CCR5 cells than in CD4+
Fig. 8. Model summarizing the functional and structural properties of some CCR5 variants and their usage for HIV-1 entry. Three CCR5 subsets recognized by MAbs PA14 (gray), 2D7 (solid black lines), and 45531 (dashed lines) are represented by a Venn diagram. The (more ...)
We postulate that PA14 binds the greatest proportion of CCR5 molecules that are permissive for HIV-1 entry (). This scenario agrees with the capacity, noted earlier, of PA14 to block entry of all four viruses, although it does so with distinctly varying potencies. MAb 45531, which mostly binds to cell surface clusters that are not stained by PA14, and which only partly colocalizes with PA14, is only a weak inhibitor of HIV-1 infection; its activity is strongest against Res-3FP but is detectable only with Env pseudovirus on U87-CD4-CCR5 cells. The increased sensitivity of Res-3FP to 45531 in these cells does not indicate that the virus has switched to using CCR5 in the lipid rafts, because it is even more potently inhibited by PA14 than the parental virus. Moreover, unlike Par-3FP, Res-3FP is effectively blocked by 2D7 in U87-CD4-CCR5 cells. Both mutants, but particularly Res-3FP, seem to have a lower inhibitory threshold of MAb occupancy on the subset of CCR5 molecules that are relevant to entry. Even if 45531 binds the same CCR5 subset as 2D7, its weaker inhibition can be explained by the different location of its epitope. For 45531, and also the more potent MAbs, inhibiting infection of U87-CD4-CCR5 cells requires MAb binding to substantial numbers of largely signaling-inactive CCR5 molecules.
In contrast, we predict that fewer CCR5 molecules are completely uncoupled from G proteins, or are signaling inactive, in CD4+ T cells, as depicted by sliding the circles to the right in . In the intermediate zone, the overlap between PA14 and 2D7 would therefore be greater, which could explain why, unlike on U87-CD4-CCR5 cells, 2D7 is as potent and effective an inhibitor as PA14. The failure of 45531 to inhibit even Res-3FP Env pseudovirus infection of CD4+ T cells is depicted by a smaller overlap in binding between 45531 and 2D7 or PA14 than occurs on U87-CD4-CCR5 cells; the ratio between signaling-active and -inactive or intermediate conformations is modeled to be higher in these cells ().
Overall, a model of intersecting, antigenically heterogeneous forms of CCR5 can account for the differential inhibitory effects of the MAbs on HIV-1 infection in the absence of VVC (). In this scenario, a pool of CCR5 molecules with a high affinity for VVC, which we propose is better recognized by PA14 than the other MAbs, may be preferentially used for entry by the different viruses. The VVC-resistant viruses appear to require a greater density of such receptors in the absence of VVC (3
). We suggest those receptors are mostly located outside cholesterol-rich subregions of the cell membrane. CCR5 forms located within the latter regions may, however, be highly relevant to how resistant viruses enter in the presence of VVC. We have modeled the different VVC inhibition patterns observed in primary cells and cell lines (3
). Consistent with our current results, a virus similar to Res-3FP was strongly resistant to VVC in PBMCs but significantly more sensitive in the TZM-bl line. Moreover, the VVC inhibition pattern was not only cell type dependent, but also PBMC donor dependent (3
). The model predicted two forms of CCR5, one with low affinity for VVC and the other with high affinity (3
). When the proportions of the two CCR5 populations and their affinities for VVC were varied, the model created infection inhibition curves that fit data sets from both cell types. We proposed that VVC-sensitive viruses can enter via one or both of the free forms of CCR5 to various extents but can use neither of the VVC-bound forms. In contrast, Res-3FP could only use the VVC complex of the low-VVC-affinity version of CCR5 while losing the capacity to use the unbound low-affinity form and retaining or increasing its ability to use the unbound high-affinity form. The VVC inhibition pattern in cell lines is consistent with the high-VVC-affinity form of CCR5 predominating over the low-VVC-affinity form (3
). It may be relevant that TAK-779 and, to a lesser extent, MVC was recently shown to have a higher affinity for G-protein-uncoupled CCR5 molecules than for G-protein-coupled forms (17
A synthesis of our previous and new models is that Res-3FP may be able to use the low-VVC-affinity form of CCR5 (when occupied by VVC), which we argue corresponds to a subset of 45531-stained CCR5 molecules (when not occupied by VVC). These coreceptors may be predominantly localized in cholesterol-rich membrane microdomains () (3
). However, we do not know the identities or lipid characteristics of the clusters on the surfaces of U87-CD4-CCR5 cells that are stained by 45531, nor do we know the sublocalizations of CCR5 molecules recognized by this MAb on CD4+
T cells. The isolate from which Res-3FP was cloned emerged under the selection pressure of VVC in PBMCs (33
). We propose that in doing so it became adapted to using the VVC complex of the low-VVC-affinity form of CCR5 in lipid rafts, a CCR5 subpopulation that is proportionately more abundant in activated CD4+
T cells than in U87-CD4-CCR5 cells (). Cholesterol is not required for wild-type HIV-1 Env-induced fusion if the receptor density is high (56
). Instead, the presence of cholesterol promotes fusion and entry by increasing the mobility of the coreceptor and thereby allowing local enrichment at the sites where fusion occurs (32
). The cholesterol requirements for the resistant virus in the presence of VVC could be different from what applies to wild-type viruses. We do not yet know whether the FP substitutions directly affect the lipid interactions of gp41 or how the fusion process is triggered by different forms of CCR5 (3
). One possibility is that the FP sequence changes permit fusion to be triggered by VVC-CCR5 complexes in a cholesterol-rich lipid environment. We are now investigating this hypothesis.