Phenotypic properties of transmitted founder HIV-1

IMCs of TF and CC Viruses.

Although a limited number of TF and CC IMCs has previously been reported (29, 31–33), available clones, especially from chronically infected individuals, were too few to conduct meaningful phenotypic comparisons (29). To create a more balanced panel, with respect to both the number of TF and CC IMCs and their subtype representation, we cloned additional viral genomes from individuals enrolled in acute and chronic HIV-1 infection cohorts. Using previously reported methods (31–33), we inferred 12 additional TF genomes (Fig. S1) representing mucosally transmitted viruses from single (n = 8) and multivariant (n = 4) infections. Together with existing constructs, these clones comprised a panel of 27 TF IMCs, with equal representation of subtypes B (n = 13) and C (n = 14) viruses (Table S1).

Chronically HIV-1–infected individuals harbor complex quasispecies of genetically diverse HIV-1 variants. Because it is impossible to predict based on sequence inspection alone which variants are biologically active and which variants are functionally impaired, we amplified between 20 and 40 env genes or 3′ half-genomes from chronic infection plasmas and searched for clusters of nearly identical sequences as indicators of recent clonal expansions (Fig. S2). We reasoned that the inferred common ancestors of these clusters must encode persistently replicating viruses and thus, represent relevant controls for biological comparisons with TF viruses. Consistent with this interpretation, we found that all chronic IMCs generated from such expansion rakes produced viruses that grew to high titers in CD4+ T cells. However, not all chronic infection plasmas were suitable for IMC construction. Analyzing specimens from over 60 individuals, we identified only 14, including 4 reported previously (29), that exhibited clonal expansion rakes in both 3′ and 5′ halves of their viral genomes (Fig. S2). These chronic plasma samples were used to construct CC IMCs representing both subtypes B (n = 5) and C (n = 9) infections (Table S1).

To determine the coreceptor use of 22 newly derived IMCs, we infected CCR5- and CXC chemokine receptor type 4 (CXCR4)-expressing reporter cells in both the presence and absence of their respective inhibitors (16). Consistent with previous analyses of Env pseudotypes (16, 28, 29, 34), the results indicated that all IMCs were CCR5 tropic, except for one TF and two CC viruses that were dual tropic for CCR and CXCR4 (Table S1).

Phenotypic Studies.

All biological experiments were performed using viral stocks that were CD4+ T cell-derived, sucrose-purified, and depleted of CD45+ microvesicles. Virus was quantified by measuring reverse transcriptase (RT) activity, viral RNA copy number, and Gag p24 antigen content. Comparing these values, we noticed that subtype C stocks seemed to contain about fivefold less p24 antigen per unit of RT activity than subtype B stocks (Fig. S3A). This discrepancy was also seen when p24 antigen was normalized using RNA copy numbers (Fig. S3B). However, no such difference was observed when virion RT activity was compared with RNA copy number (Fig. S3 C and D). Similar results were obtained when viral stocks were tested in two additional p24 detection assays (Fig. S3 E and F). Thus, three commonly used p24 capture assays bound subtype C core proteins considerably less efficiently than subtype B core proteins, leading to a systematic underestimation of the number of viral particles in subtype C viral stocks. We, therefore, used RT activity to normalize virus input and measure virus replication in all subsequent experiments.

To determine whether TF and CC viruses differed in their phenotypic properties, we used two complementary statistical approaches. First, we used a conservative nonparametric permutation test (perm test) to address the central question of whether TF and CC viruses exhibited reproducible phenotypic differences. Second, we used a generalized linear model (GLM) to test for interactions between different parameters, such as subtype and cell donors in replicate experiments, and test for the independence of viruses from the same subjects. Because both virus status (TF or CC) and subtype (B or C) influenced the biological outcome, the GLM test allowed us to interpret the data with these variables taken into account.

TF Viruses Exhibit Enhanced Infectivity.

Plasma virus collected during ramp-up stages of acute simian immunodeficiency virus (SIV) infection has been shown to be significantly more infectious than virus collected from later plasma samples (35). To examine whether this observation was also true for HIV-1, we used a single-round infection assay to determine whether TF virions were more infectious on a per-particle basis than CC virions. Serial dilutions of CD4+ T cell-derived viral stocks were used to infect TZM-bl cells (which express luciferase under the control of an HIV-1 promoter), and the resulting relative light units were expressed as a function of the input RT activity. The results showed that TF viruses as a group were 1.7-fold more infectious than CC viruses (Fig. 1A), although this difference was only marginally significant (P = 0.049 by perm test; P = 0.062 by GLM). Similar results were obtained when subtypes B and C viruses were considered separately (there was no subtype dependence of infectivity) (Fig. 1B). To enhance virus infectivity, we repeated these experiments in the presence of diethylaminoethyl (DEAE) dextran, which is known to increase virus attachment to _target cells (36). As expected, DEAE dextran increased the per-particle infectivity of all viruses by up to three orders of magnitude (Fig. 1C), with a significantly greater effect on subtype C than subtype B virions (P = 0.015 by GLM) (Fig. 1D). However, in the presence of DEAE dextran, the marginally significant difference between TF and CC viruses shifted to nonsignificance (P = 0.068 by perm test; 0.083 by GLM). Taken together, these data indicate that TF viruses are, on average, nearly two times as infectious as viruses that persist during chronic infection. However, when normal barriers to cell-free infection are mitigated by DEAE dextran, this difference is no longer significant (although a trend was still evident).

TF Virions Have a Higher Env Content.

Previous computational analyses of TF and CC env gene sequences revealed genetic signatures in TF viruses that increased Env expression and particle incorporation when tested in the context of pseudoviruses (37). We, thus, examined whether TF virions packaged more Env per particle than CC virions. To detect both subtypes B and C Env proteins with comparable efficiency, we developed an ELISA that used antibodies previously shown to bind genetically highly diverse envelope (gp120) proteins. To capture Env, we used CD4-218.3-E51, a chimeric antibody that contains the first and second domains of human CD4 linked to the CD4-induced monoclonal E51 (38). To detect bound Env, we selected affinity-purified anti-gp120–specific polyclonal antibodies that were isolated from human plasma (Advanced Bioscience Laboratories, Inc.). Using this ELISA, we found that TF viruses contained 1.9 times more Env per unit of RT activity than CC viruses (P = 0.048 by perm test; P = 0.057 by GLM) (Fig. 1E). Subtype B viruses appeared to package 2.4-fold more Env per particle than subtype C viruses (Fig. 1F), but this result was largely caused by a binding preference of the capture antibody, which recognized subtype B Env glycoproteins two times more efficiently than subtype C Env glycoproteins. Nonetheless, this subtype bias did not affect the differential between TF and CC viruses, which was seen for both subtypes B and C, and it was controlled for in the statistical analyses (Fig. 1F). As expected, there also was a significant correlation between Env content and particle infectivity (r = 0.33; P = 0.036).

TF Viruses Bind DCs More Efficiently.

DCs have been proposed to play an important role in HIV-1 transmission, because they are located in the mucosa (39, 40), capture viruses using lectins (41, 42) and glycosphingolipid receptors (43), and efficiently transmit infectious particles to CD4+ T cells (44–46). To examine whether TF and CC viruses differ in their ability to bind DCs, we pulsed immature monocyte-derived DCs (moDCs) with equal amounts of virus (normalized by RT activity), washed the cells extensively to remove cell-free virions, and then lysed the cells to quantify the amount of cell-associated virus. Using cells from three different donors (Fig. 2A), we found that moDCs captured TF viruses 1.6 times more efficiently than CC viruses (P = 0.040 by perm test; P = 0.060 by GLM). This increase was observed for both subtypes B and C TF viruses (Fig. 2B), although subtype B viruses were captured 3.4 times more efficiently than subtype C viruses (P = 4.6 × 10⁻⁶ by GLM) (Fig. 2B).

Because in the above experiment, we had normalized virus input by RT activity but measured virion capture using the more sensitive p24 assay, we considered the possibility that the subtype-specific differences were an artifact of the p24 antigen capture assay. To explore this possibility, we repeated the binding experiment with cells from three additional donors, but this time, we normalized virus input by p24 content (Fig. 2 C and D). Despite adding an estimated fivefold excess of subtype C virions, TF viruses were again captured 1.8 times more efficiently than CC viruses (P = 0.030 by perm test; P = 0.014 by GLM). Moreover, there was still a subtype-specific difference, with subtype B viruses being captured 1.9 times more efficiently than subtype C viruses (P = 0.0029 by GLM). As a control, we treated an aliquot of the pulsed moDCs with 0.25% trypsin-EDTA and showed that this treatment removed all detectable cell-associated virus, confirming that most of the DC-associated virus was surface-exposed (47). We also cultured virus-exposed moDCs and showed that these cells were not productively infected. Finally, we asked whether the percentage of captured virus for each strain correlated between different donors. This association was indeed observed regardless of whether virus input was normalized by RT activity or p24 content, thus validating the DC binding assay (Fig. S4A).

Although the combined data from all six donors indicated that DCs captured TF viruses 1.7-fold more efficiently than CC viruses (P = 0.035 by perm test; P = 0.005 by GLM), there was considerable donor variability, with moDCs from two donors (A and F in Fig. 2) failing to yield significant binding differences (Fig. S4). To examine potential reasons, we compared surface expression levels of DC-specific intracellular adhesion molecule 3-grabbing nonintegrin (DC-SIGN) and macrophage mannose receptor (MMR) by flow cytometry. Interestingly, moDCs of donors A and F expressed lower levels of these molecules (Fig. S4 B and C), suggesting that the number of Env-specific binding sites on their cells was limited. For the remaining four donors, moDC virus capture tended to correlate with particle Env content (Fig. S4D). These data suggest that virion Env content represents one important determinant in moDCs binding, although expression of lectins and other factors clearly also play a role (48).

To examine the efficiency of virus transfer from moDC to CD4+ T cells, we pulsed moDCs with equivalent amounts of TF and CC viruses, cocultured these cells with autologous CD4+ T cells, and then measured RT activity in culture supernatants as an indicator of virus replication. Analysis of cells from two different donors showed that TF viruses replicated to slightly higher titers than CC viruses each day over the course of 10 d (one representative donor is shown in Fig. 3). Although these differences were not significant in the perm test (P = 0.104), significant differences in TF and CC virus titers were observed in the GLM analysis (P = 0.025). To explore the reason for this apparent discordance, we ran the perm test on subtypes B and C viruses separately (Fig. 3B), and we found that subtype B TF viruses (P = 0.004), but not subtype C TF viruses (P = 0.23), grew to significantly higher titers than their respective chronic controls. Averaging data from both donors and across all time points, this difference was 11.2-fold for subtype B viruses but not significant for subtype C viruses. Because subtype B virions bound moDCs 3.5-fold more efficiently than subtype C viruses (Fig. 2 B and D), we interpret these findings to indicate that virions that bind moDCs more efficiently are also transferred to CD4+ T cells more efficiently.

TF Viruses Are Relatively More Resistant to IFN-α.

CD4+ T cells represent the predominant _target cells in gut-associated lymphoid tissues and the first cell type to become productively infected after transmission (49). To determine the growth potential of TF and CC viruses in this cell type, we infected activated CD4+ T cells from three healthy donors with equivalent amounts of virus and monitored viral growth by measuring RT activity in culture supernatants (Fig. 4 A, C, and E, gray lines). Although TF viruses replicated to slightly higher titers, especially at the earliest time point postinfection, this difference was not statistically significant (P = 0.16 by perm test; P = 0.186 by GLM). Thus, in activated CD4+ T cells, TF and CC viruses replicated with comparable efficiency.

IFN-α is produced early in infection by plasmacytoid DCs (50), which are among the first cells to be recruited to the mucosal site of virus entry (4). This cytokine stimulates expression of hundreds of host genes [IFN-stimulated genes (ISGs)] (51), many of which have anti–HIV-1 activity (52). IFN-α has also been shown to limit Env incorporation into virus particles that are released from infected CD4+ T cells (53). To determine the sensitivity of TF and CC viruses to the antiviral activity of this cytokine, we analyzed their replication kinetics in CD4+ T cells in the presence of 500 U/mL IFN-α. Determining RT activity in culture supernatants as a measure of viral replication, we found significant differences in the growth rate of TF and CC viruses (Fig. 4 A, C, and E, black lines). However, this effect was highly subtype-specific (Fig. S5). Averaging replication data from all donors and across all time points, we found that subtype B TF viruses grew to 62-fold higher titers than subtype B CC viruses (P = 0.000013 by GLM). In contrast, subtype C TF viruses grew only 1.7-fold more efficiently than their respective CC viruses, which was not significant (P = 0.53 by GLM). Importantly, this difference was not because of a lack of IFN-α resistance of the subtype C TF viruses but an increased IFN-α resistance of the respective chronic controls (Fig. 4 B, D, and F). Inclusion of sex and risk group information into the statistical analyses indicated that neither was significantly associated with IFN-α resistance; however, the study was not sufficiently powered to examine the impact of sex and risk group, and a contribution of these factors cannot, therefore, be excluded (Fig. S5).

As a different way to analyze these data, we also calculated the ratio of virus production in the presence and absence of IFN-α for each virus strain (Fig. 4 B, D, and F). Comparing across all donors, we again found that TF viruses were significantly more resistant to inhibition by IFN-α than chronic viruses (P = 0.027 by perm test). However, when considering the two clades separately, this finding was significant only for subtype B (P = 0.0004 by one-sided Wilcoxon test) and not for subtype C (P = 0.84). Viruses that were highly resistant in one donor were also highly resistant in the other donors (Fig. S6), indicating that their IFN-α phenotype was consistent between donors.

To determine whether the higher titers of TF viruses were because of increased particle release or differences in viral spread, we used flow cytometry to measure the percent of infected cells in IFN-α–treated cultures from two donors (Fig. S7). In the presence of IFN-α, TF viruses infected a larger number of cells than CC viruses, although the overall effect was only marginally significant (P = 0.049 by perm test). However, GLM analysis again detected a significant interaction between subtype and TF/CC status (P = 0.041), indicating that the differences were primarily caused by subtype B viruses. Because the ratio of virus production in the presence and absence of IFN-α correlated well with the percent of Gag-positive cells (P < 0.0001) (Fig. S7 B and C), it is likely that IFN-α resistance represents an enhanced ability to spread between CD4+ T cells. Interestingly, this phenotype correlated only weakly with particle infectivity and Env content, indicating that the higher levels of particle-associated Env in TF viruses are not the main drivers of their relative IFN-α resistance.

IMC Construction.

The sequences of full-length TF and CC viruses were inferred and molecularly cloned as described (27, 29, 31–33). GenBank accession numbers are listed in Table S2.

Viral Stocks.

CD4+ T cells pooled from nine donors were infected with 293T cell-derived transfection supernatants. Viral stocks were harvested after 11 d and depleted of CD45+ microvesicles. Viral p24 core protein content was measured using the PerkinElmer High-Sensitivity AlphaLISA, the PerkinElmer Alliance ELISA, and the method described by Biancotto et al. (83). RT activity was measured using the Roche Reverse Transcriptase Assay. Viral RNA was quantified using the Roche COBAS AmpliPrep/COBAS Taqman HIV-1 v. 2.0 Test.

Phenotypic Analyses.

Coreceptor use was determined as described (33). Virion infectivity was measured using the TZM-bl reporter cell line. Particle-associated Env content was measured using a newly developed ELISA (SI Materials and Methods). To assess DC capture and CD4+ T-cell transfer, peripheral blood monocytes were differentiated to moDCs using 100 IU/mL granulocyte-macrophage colony-stimulating factor (GM-CSF) and 200 IU/mL IL-4. For moDC capture experiments, 1 × 10⁵ cells were incubated with viral stocks containing either 30 pg RT activity or 600 pg p24 protein. For CD4+ T-cell transfer experiments, viral stocks containing 50 pg RT activity were added, and moDCs were cultured at a 1:5 ratio with autologous CD4+ T cells. For CD4+ T-cell infection studies, 2 × 10⁶ negatively selected CD4+ T cells were infected overnight with viral stocks containing 30 pg RT activity. Cells were then propagated in media containing 30 U/mL IL-2 in the presence and absence of 500 U/mL IFN-α. Infected cells were identified by flow cytometry (SI Materials and Methods).

Statistical Analyses.

Data were analyzed using a nonparametric perm test that controlled for virus subtype, donor cells, and time of sampling as well as a GLM that tested for interactions between and the relative importance of these variables (SI Materials and Methods).