Codelivery of Envelope Protein in Alum with MVA Vaccine Induces CXCR3-Biased CXCR5⁺ and CXCR5⁻ CD4 T Cell Responses in Rhesus Macaques

The induction of robust and long-lived Ab responses forms the basis of protective immunity elicited by most vaccines (1). Ab dynamics following immunization result from activation of Ag-specific B cells and their subsequent commitment into two distinct cell fates—extrafollicular plasmablasts or germinal center (GC) B cells. Plasmablasts are rapidly proliferating Ab-secreting cells (ASCs) within secondary lymphoid organs that mainly contribute to peak Ab titers within the first few weeks after immunization (1, 2). Long-lived serological memory is established by GC-derived bone marrow–resident plasma cells. GCs arise within B cell follicles typically 2–4 wk after immunization and are composed of Ag-specific B cell clones of varying affinity that result from rapid B cell proliferation and receptor diversification. High-affinity clones that successfully engage TCRs on CD4⁺ T follicular helper (T_FH) cells within the GCs receive T_FH cell help in the form of cytokines such as IL-21, IL-2, and IL-4 and costimulatory signals such as ICOS and CD40L resulting in their survival and differentiation to plasma cells or memory B cells (3, 4).

The vital role of T_FH cells in the induction of humoral immunity makes them attractive vaccine targets, and characterizing vaccine elicited T_FH cell responses associated with broad and robust Ab titers will provide valuable information for vaccine design. Until recently, tracking vaccine-elicited T_FH cells in humans represented a challenge because of the belief that T_FH cells are exclusively localized to the GCs of secondary lymphoid organs. However, there is some evidence that T_FH cells circulate transiently as CXCR5⁺CD4⁺ T cells in blood and, based on the expression profile of activation markers, are predictive of Ab responses. For instance, HIV⁺ individuals that respond to H1N1 vaccine show expansion of CXCR5⁺CD4⁺ T cells in the blood (peripheral [p] T_FH), and the frequency of ICOS⁺ pT_FH cells correlates with concurrent H1N1 titers (5). Likewise, CCR7^loPD-1⁺ cells within the pT_FH cell pool are induced after influenza vaccination; this subset is overrepresented in patients with autoimmune syndromes and highly correlates with anti-dsDNA Abs and disease severity (6). Taken together, these studies indicate that vaccine-elicited T_FH cells circulate during the effector response to vaccination, these pT_FH cells demonstrate an activated phenotype, and their magnitude correlates with vaccine-specific Ab titers generated within a month after vaccination. What is less understood is whether pT_FH cells are predictive of long-term Ab titers and quality and how they compare with lymph node (LN) T_FH cell responses.

Recent studies have underscored the expression of chemokine receptors as a key functional attribute of T_FH cells (7). Blood CXCR5⁺CD4⁺ T cells in humans are composed of CXCR3⁺ and CXCR3⁻ subsets, which show heterogeneity in B cell helper potential. For instance, induction of CXCR3⁺ICOS⁺CXCR5⁺ pT_FH cells at day 7 after influenza vaccination predicts increase in Ab titers at day 28 postimmunization (8). In contrast, in HIV-infected individuals, frequency of CXCR3⁻PD-1⁺CXCR5⁺ cells is associated with the development of neutralizing Abs (9). These data indicate that phenotypic characteristics of pT_FH cells may be specific to the vaccine/infectious agent and the resulting inflammatory response. In the context of HIV infection, the data suggest that CXCR3^- T_FH cells may be favorable for induction of Abs. However, this paradigm has not been explored in the context of HIV vaccination.

In the current study, we examined the role of blood and LN CXCR5⁺CD4⁺ Th cells in the development of Env-specific Ab responses in the context of a DNA prime, recombinant modified vaccinia virus Ankara (MVA) boost (DNA/MVA) SIV vaccine in the presence and absence of a gp140 protein boost adjuvanted with aluminum hydroxide (alum) in rhesus macaques. We found that booster immunization with MVA induced Ki-67⁺CXCR5⁺CD4⁺ T cells, which circulated in blood (albeit at low levels) and were heterogeneous for expression of CXCR3. Blood Ki-67⁺CXCR5⁺CD4⁺ T cells at the peak effector phase were representative of GC T_FH responses, and CXCR3 expression on GC T_FH cells was a determinant of persistence, neutralization potential, and avidity of anti-Env Ab response. These data offer an important parameter with which to identify, track, and characterize CD4⁺ T_FH cells in vaccine studies to gain an understanding of the types of T_FH cell responses that lead to protective Abs.

All animal protocols were approved by the Emory University Institutional Animal Care and Use Committee protocol YER-2002343. All experiments were conducted in strict accordance with U.S. Department of Agriculture regulations and the recommendations for conducting experiments in accord with the highest scientific, humane, and ethical principles as stated in the Guide for the Care and Use of Laboratory Animals. Animals were housed in pairs in standard nonhuman primate cages. Animals received standard primate feed as well as fresh fruit and enrichment daily and had free access to water. Upon infection, animals were housed singly. Immunizations, infections, blood draws, and biopsy procedures were performed under anesthesia by trained research staff. All efforts were made to schedule samples on paired animals concurrently so as to minimize potential distress.

Twenty-eight male Indian rhesus macaques were included in this study. Animals ranged in age from 2 to 4 y and in weight from 3 to 6 kg at the start of the study. Animals were STLV⁻ and SIV⁻ at study commencement. None of the animals expressed Mamu class 1 alleles B08 and B17, and each experimental group had a total of three Mamu A*O1 animals. Animals from both experimental groups were randomized into four sampling groups for immunization and sampling of blood and tissue biopsies. Animals were monitored throughout the study period and, based on clinical parameters and complete blood counts, were reported to be healthy and immunocompetent throughout. All animals were housed at the Yerkes National Primate Research Center (Atlanta, GA) and treated in accordance with the Yerkes National Primate Research Center Institutional Animal Care and Use Committee regulations.

The vaccine study consisted of two experimental groups; all 28 animals received two CD40L-adjuvanted DNA primes (0 and 8 wk), followed by two MVA boosts (16 and 32 wk; DNA DNA MVA MVA [DDMM] regimen). To determine whether inclusion of protein boost along with second MVA resulted in augmented Ab responses, 14 animals were randomized to receive gp140 Env in alum along with the second MVA (DDMM-Pro). The DNA immunogen expressed SIV239 Gag-Pol, Env, Tat, Rev, and coexpressed macaque CD40L and was delivered at 3 mg/dose (10). The MVA immunogen expressed SIV239 Gag, Pol, and Env and was delivered at a 10⁸ PFU/dose (11, 12). Protein vaccination consisted of 100 μg endotoxin-free SIV239 gp140 (Immune Technology), which was premixed with 500 μg alum (2% alhydrogel; InvivoGen) prior to vaccination and was delivered in the thigh contralateral to MVA immunization. Briefly, protein and alum were combined in a 1:1 volume ratio and were mixed by rocking at room temperature for 10 min. Prepared inoculum was stored on ice until the time of inoculation. All immunizations were delivered in PBS i.m. in a single shot in the outer thigh.

At 6 mo after vaccination, all vaccinated animals were challenged with SIVmac251 intrarectally on a weekly basis for a maximum of 5 wk or until detection of plasma viremia above 250 copies/ml for two consecutive weeks. Infection with SIVmac251 (day 8, 7.9.10 virus stock from N. Miller [National Institutes of Health, Bethesda, MD]) was performed using a 1-cc slip tip syringe containing 1 ml SIVmac251 at 200 50% tissue culture-infective dose. A syringe was inserted gently ∼4 cm into the rectum, the plunger was depressed to instill the virus, and the animal was returned to the cage in a prone position.

PBMCs were isolated from whole blood collected in sodium citrate tubes and isolated by density gradient centrifugation, according to standard procedures as described previously (11). PBMCs were isolated, counted, and used for various assays within 6 h of blood collection. LN biopsies were collected in RPMI 1640 medium and 10% FBS. To obtain single-cell suspensions, LNs were manually processed. Briefly, subsequent to removal of fat and connective tissue, biopsies were incised using a scalpel, and tissue segments were manually disrupted over a 100-μm cell strainer using a 10-cc plunger. Cells were washed and resuspended in complete media Biopsies and resulting suspensions were placed on ice at all times to preserve cell viability. After determination of cell counts, cells were used immediately or cryopreserved using standard techniques. Cell viability in both PBMCs and LN suspensions was determined to be >90%.

ELISPOT assays were performed using fresh PBMCs using a standard ELISPOT assay (13, 14). Prior to use, PBMCs were washed up to five times in media to remove traces of bound Abs that could contaminate the assay. ELISPOT plates (Millipore, Billerica, MA) were coated with SIV239 gp140 at 1 μg/ml for detecting Ag-specific ASCs or with IgG at 10 μg/ml (Rockland Immunochemicals, Rockland, PA) for detecting total IgG ASCs. After coating the plates overnight, PBMCs were incubated for up to 8 h and detected using biotinylated goat anti-monkey IgG (Rockland Immunochemicals), followed by alkaline phosphatase–conjugated streptavidin (Mabtech). Spots were detected using one-step NBT substrate (Thermo Scientific).

Binding Ab titers against SIV239 gp140 were performed on frozen sera that were thawed and heat-inactivated as described previously (10). A NaSCN displacement ELISA as previously described was used for determining avidity against SIV239 gp140 (10). SIV-specific neutralization Ab was measured as a function of reduction in luciferase reporter gene expression after a single round of infection in TZM-bl cells as described previously (15).

The SIV copy number in plasma was determined using a quantitative real-time PCR as described previously (10). For viral load determinations in LN, total DNA was extracted from ∼50,000 sorted CD4⁺ subsets, and data were normalized to albumin or GAPDH as depicted. Sorting experiments were performed on cryopreserved LN samples. All PCRs were performed in duplicates with a limit of detection of 60 copies/reaction.

Staining on whole blood was done at room temperature, whereas PBMCs and LN cell (10⁶ cells) suspensions were stained in PBS containing 2% FBS (FACS buffer) for 30 min at 4°C. Cells were stained with fluorochrome-conjugated Abs specific for CD3 (SP34-2), CD4 (OKT4), CD8 (SK1), CD20 (2H7), CD95 (DX2), CXCR3 (1C6), CCR5 (3A9), Bcl-6 (K112-91), Ki-67 (B56), IFN-γ (B27), TNF-α (MAb11), and IL-2 (MQ1-17H12).

Monoclonal Abs against IL-21 (3A3-N2.1) were obtained from BD Pharmingen (San Jose, CA); IL-4 (7A3-3) was from Miltenyi Biotec (San Diego, CA); signaling lymphocyte activation marker (SLAM; A12), programmed death (PD)-1 (EH12.2H7), T-bet (4B10), and ICOS (C398.4A) were from BioLegend (San Diego, CA); CXCR5 (MU5UBEE) was from eBioscience (San Diego, CA); CD127 (R34-34) was from Tonbo Biosciences; and CCR7 (FAB197A) was from R&D Systems. Dead cells were excluded from analysis based on staining for Live/Dead Near-IR dead cell stain from Molecular Probes, Invitrogen (Grand Island, NY). Bcl-6, T-bet, and Ki-67 stains were performed after cells were stained for surface Ags, followed by permeabilization/fixation using the Foxp3 kit, and protocol, followed by intracellular staining. Prior to intracellular staining for cytokines, fresh PBMCs were stimulated with peptide pools of Gag and Env for 5 h in the presence of brefeldin A (GolgiPlug; BD Biosciences, San Jose, CA) and then fixed and permeabilized with CytoFix/CytoPerm (BD Biosciences), according to the manufacturer’s instructions. Unstimulated cells from each animal served as negative control, and PBMCs stimulated with phorbol myristic acetetate/ionomycin served as positive controls. Samples were acquired on an LSR Fortessa (BD Biosciences), and 500,000 total events were collected for each sample. Cell sorting was performed using the Aria II (BD Biosciences). Data were analyzed using FlowJo software version X.0.7 (Tree Star, Ashland, OR). Data were analyzed after gating out dead cells and subsequently gating on singlet cell populations. CD3⁺,CD20⁻ cell subsets were gated, and all analysis with CD4⁺ T cells was performed on parent CD4⁺,CD8⁻ T cell subsets. The minimum number of events required to score a response as positive was set at 20 events after subtraction from background.

Statistical analysis was performed using Graph Pad Prism version 5.0. A two-tailed nonparametric t test was used for all comparisons unless otherwise specified. Spearman correlation was used to determine associations between variables. Statistical significance was set at p < 0.05.

Twenty-eight rhesus macaques were vaccinated i.m. with DNA/SIV vaccine (10) at 0 and 8 wk and with MVA/SIV vaccine (10) at 16 and 32 wk (DDMM group) (Fig. 1A). From this group, 14 macaques also received i.m. SIV gp140 protein boost in alum concurrently with the second MVA (DDMM-Pro). Blood was sampled at baseline and peak effector and memory time points after each immunization. A single LN biopsy was obtained at 2 wk after the second MVA immunization.

To determine whether DNA/MVA vaccine–elicited CD4⁺ T cells in blood have the potential to migrate to B cell follicles, we examined Ki-67⁺ (marks proliferating cells) CD4⁺ fraction for expression of the chemokine receptor CXCR5, a defining T_FH cell marker that facilitates homing to B cell follicles/GCs (Fig. 1B) (16). We also monitored the CXCR5⁻Ki-67⁺CD4⁺ T cell fraction (presumably non-T_FH fraction). About 0.1–1.7% (mean of 1%) of CD4⁺ T cells prior to boost were CXCR5⁺Ki-67⁺, and this fraction increased by 2-fold at day 7 following the first MVA boost and returned to near baseline levels at 8 wk postimmunization. The frequency of CXCR5⁻Ki-67⁺CD4⁺ T cells was ∼3- to 4-fold higher compared with CXCR5⁺ counterparts and expanded with similar kinetics. Thus, CXCR5⁺ CD4 T cells constituted a small proportion (∼15–25%) of total vaccine-elicited CD4 T cell responses in circulation. A similar phenotypic distribution was also observed within the Ki-67⁻ memory compartment in blood (Supplemental Fig. 1A). This is consistent with data in mice demonstrating that ∼25% of circulating OT-II effectors represent T_FH cells (6) and with data in humans showing that only a small fraction (5–20%) of total tetanus–specific memory cells in blood is CXCR5⁺ (9).

Next, we examined SIV-specific CD4⁺ T cell responses to study how kinetics and phenotype of Ag-specific CD4⁺ T cells compared with that of Ki-67⁺CD4⁺ T cells. To enumerate vaccine-induced SIV-specific CD4 responses, PBMCs were stimulated with overlapping peptide pools (15-mer overlapping by 11 aa) derived from SIVmac239 Gag and Env sequences and examined for the production of IFN-γ and IL-21 using flow cytometry. The majority of IL-21–producing CD4 T cells coexpressed IFN-γ (data not shown) and peaked at week 1 post-MVA boost (Fig. 1C). The frequency of SIV-specific IFN-γ⁺IL-21⁺CD4 T cells showed a positive association with the frequency of total Ki-67⁺CD4⁺ T cells at day 7 following MVA boost (Fig. 1C, right panel), suggesting that proliferating cells were a reasonable indicator of vaccine-induced CD4⁺ T cells at the peak of the effector response.

To more formally identify Ag-specific CXCR5⁺CD4⁺ T cells in blood, we set out to examine CXCR5 expression on responding cells after stimulation. Downregulation of CXCR5 after 6 h of culture at 37°C necessitated staining of cells with CXCR5 prior to stimulation. With this strategy, we observed that CXCR5⁺CD4⁺ T cells made robust amounts of IFN-γ and IL-21 in response to PMA/ionomycin stimulation (Fig. 1D). IFN-γ and IL-21 responses in CXCR5⁺ cells after Gag peptide stimulation were modest but discernable. The frequency of Gag-specific CXCR5⁺ CD4 T cells expressing either IFN-γ or IL-21 ranged from 0.004 to 0.15% of total CD4s (Fig. 1E) and correlated positively with the frequency of Ki-67⁺CXCR5⁺ CD4 T cells (r = 0.42; p < 0.05) (data not shown). About 0.5 to 20% of IFN-γ⁺ or IL-21⁺ CD4 cells expressed CXCR5 (data not shown).

Notably, CXCR5⁻CD4⁺ T cells were the major producers of IFN-γ and IL-21 in response to PMA/ionomycin and Gag. Although this could be due in part to our inability to capture all CXCR5-expressing cells, these data reveal that akin to human CD4 T cells IL-21 production is not an exclusive feature of CXCR5⁺ CD4 T cells (17). Taken together, the data point to the conclusion that DNA/MVA vaccine–induced Ag-specific CXCR5⁺CD4⁺ T cells circulate, albeit at very low frequencies, in peripheral blood during the peak effector response.

To determine whether these circulating CXCR5⁺CD4⁺ T cells possessed B cell helper potential, we cocultured FACS-sorted CXCR5⁺ and CXCR5⁻ memory CD4⁺ T cells along with autologous memory B cells, and cell supernatants were collected at days 3 and 7 to examine secreted IgG (Fig. 1F). At day 7, cells were also harvested and examined for B cell proliferation. Both CXCR5⁺ and CXCR5⁻CD4⁺ T cells efficiently induced IgG secretion and B cell proliferation compared with naive CD4⁺ T cells indicating that circulating CXCR5⁺ and CXCR5⁻CD4⁺ T cells possess B cell helper potential in vitro. This helper potential is consistent with the expression of ICOS by these cells (Fig. 2C). Taken together, these data indicated that the first MVA booster immunization resulted in the transient accumulation of Ki-67⁺ CD4 T cells in blood at the peak of the effector response. These cells were phenotypically heterogeneous for expression of CXCR5, but both CXCR5⁻ and CXCR5⁺ cells demonstrated B cell help functionality in vitro.

The cytokine profile of SIV-specific CD4⁺ T cells indicated the induction of Th1-like CXCR5⁺CD4⁺ T cells. Therefore, we wanted to determine the CXCR3 expression profile of Ki-67⁺ cells. Flow plots in Fig. 2A show distribution of CXCR3 (X3) and CXCR5 on Ki-67⁺CD4⁺ T cells at baseline and at 1 wk after the first MVA. Prior to MVA boost, X3⁺ cells were present in both CXCR5⁻ (X3 single positive, blue population) and CXCR5⁺ (double positive, green population) fractions, and the frequency of X3⁺ cells were lower compared with X3⁻ cells within the respective CXCR5⁺ or CXCR5⁻ subsets. However, at 1 wk following the MVA boost, the frequency of X3⁺ cells increased significantly in both CXCR5⁻ and CXCR5⁺ subsets indicating generation of Th1 biased non-T_FH and T_FH cell subsets (Fig. 2B). Noncycling (Ki-67⁻) memory CD4⁺ T cells showed a variable composition and phenotype of these subsets likely a result of diverse antigenic history and quiescent phenotype (Supplemental Fig. 1B).

To better understand the nature of cycling blood CXCR5⁺ and CXCR5⁻CD4⁺ T cells in the context of CXCR3, we performed a phenotypic characterization of these cells with respect to markers associated with LN-resident T_FH cells (Fig. 2C). A majority of blood Ki-67⁺CD4⁺ T cells within both the CXCR5⁻ and CXCR5⁺ compartments were positive for PD-1 with CXCR5⁺CD4⁺ T cells expressing relatively higher levels of PD-1. Interestingly, both CXCR5⁺ and CXCR5⁻ CD4 T cells expressed ICOS, and ICOS expression was higher on CXCR3⁺ cells compared with CXCR3⁻ cells. CXCR5⁻CXCR3⁺ cells expressed the highest levels of ICOS. This phenotype contrasts with blood CD4⁺ effectors in humans where CXCR5⁻ CD4 T cells do not express ICOS (8). Expression of the activation receptor associated with IFN-γ production, SLAM was lower on CXCR5⁺Ki-67⁺CD4⁺ T cells relative to CXCR5⁻ counterparts, reflective of the SLAM^lo phenotype of LN CXCR5⁺ cells (18, 19). Similar to ICOS expression, SLAM was expressed at higher levels on X3⁺ cells. Unlike LN-resident T_FH cells, which downregulate CCR7 to exit the T cell zone and gain access to the GCs (20), blood CXCR5⁺CD4⁺ T cells were positive for CCR7 irrespective of X3 expression. In contrast, a majority of X3⁻CXCR5⁻CD4⁺ T cells had downregulated CCR7 (20).

To further corroborate the Th1-like functionality of double positive CD4⁺ cells in blood, we wanted to examine cytokine production upon stimulation. Because dynamic changes in expression of CXCR3 and CXCR5 occur upon stimulation, we sorted CXCR5⁺ and CXCR5⁻ cells based on expression CXCR3 to accurately discern differences in cytokine profile among these subsets. Sorted subsets were stimulated with PMA/ionomycin and examined for expression of CD40L and IFN-γ. Consistent with their Th1 phenotype, X3⁺ cells showed higher frequency of CD40L⁺IFN-γ⁺ cells within both CXCR5⁻ and CXCR5⁺ compartments (Fig. 2D). In terms of median fluorescence intensity (MFI), CXCR5⁺ cells had a higher MFI of CD40L, whereas CXCR5⁻,X3⁺ subset showed highest expression of IFN-γ. Taken together, the data demonstrate that DNA/MVA-induced blood CXCR5⁺CD4⁺ T cells show phenotypic and functional heterogeneity based on CXCR3 expression, and vaccination significantly enhances the frequency of Th1-like CXCR5⁺ and CXCR5⁻ cells during the effector phase in blood.

We next determined the relationship between CXCR5⁺CD4⁺ T cells in blood and LN and examined how inclusion of alum-adjuvanted gp140 protein along with second MVA boost impacted the magnitude of these CD4⁺ T cell responses. Examination of blood Ki-67⁺CD4⁺ T cells at week 1 following the second MVA boost showed that MVA+gp140 immunization resulted in ∼2-fold higher frequencies of Ki-67⁺ CD4 T cells in circulation. The frequency of Ki-67⁺CD4⁺ T cells ranged from 5 to 13% of total CD4s in MVA only compared with 15–33% of CD4s in MVA+gp140 group (Fig. 3A). We also observed an increase for SIV Env–specific IFNγ⁺, IL-21⁺, and IL-4⁺CD4⁺ T cells (Fig. 3B) in the protein group, and as expected, this increase was not observed for Gag-specific CD4⁺ T cells (Fig. 3C). Taken together, the data showed that concurrent gp140 immunization augmented magnitude of Env-specific CD4⁺ T cell responses.

Having identified that DNA/MVA induced CXCR5⁺ cells circulate in blood during the effector response, we next examined T_FH cells in the LNs and compared them with blood CXCR5⁺ cells. We chose to biopsy the right LN, which was nondraining to the second MVA immunization so as not to interfere with Ag complexes sequestered within the draining inguinal LN after immunization. It is conceivable that virus-like particles derived from MVA-infected cells in vivo could engage memory cells in distal lymphoid organs including the right LN, which was the draining LN for the preceding DNA primes and for the first MVA boost.

Examination of LN sections by immunohistochemistry showed presence of distinct clusters of proliferating B cells within B cell follicles (data not shown). Single-cell suspensions stained for flow cytometry showed presence of three distinct populations of CXCR5⁺CD4⁺ T cells showing graded levels of PD-1 expression: GC T_FH cells (CXCR5⁺PD-1⁺⁺), CXCR5⁺PD-1⁺, and CXCR5⁺PD-1⁻ cell subsets (Fig. 3D). The frequency of GC T_FH cells correlated directly with GC B cells identified by expression of Bcl-6 as shown and with CXCR5⁺PD-1⁺ cells but not with CXCR5⁻PD-1⁺ cells or with CXCR5⁺PD-1⁻ cells (Supplemental Fig. 2A). Examination of the frequency of GC responses revealed that the gp140 boost did not alter either the frequency of total GC T_FH cells or GC B cells (Fig. 3E). This is not surprising because total GC T_FH cells and GC B cells are composed of polyclonal T and B cell responses directed against epitopes of different antigenic specificities including MVA and SIV Gag; a modest increase in Env-specific GC responses elicited by gp140 protein boost would be difficult to discern.

We next evaluated possible associations between blood CXCR5⁺ cells and PD-1–expressing GC T_FH cells and GC B cells within the LN and found a positive correlation between Ki-67⁺CXCR5⁺ cells in blood and all three LN subsets (Fig. 3F). We next asked whether the Ki-67⁺CXCR5⁻ CD4 T cell subset was also positively associated with GC responses. The data showed that CXCR5⁻ cells did not correlate with any of the measured GC responses (Supplemental Fig. 2B). Therefore, the magnitude of responding blood CXCR5⁺CD4⁺ T cells but not CXCR5⁻CD4⁺ T at peak postvaccination was predictive of subsequent GC responses. Studies showing that gene expression profile of blood PD-1⁺CXCR5⁺ cells but not of CXCR5⁻ cells closely matches that of GC T_FH cells are consistent with this possibility (9).

Next, we wanted to ascertain whether inclusion of gp140+Alum resulted in discernable effects on phenotype of CXCR5⁺ cells in blood and LN. Consistent with the first MVA boost, the majority of Ki-67⁺ cells after the second MVA boost were composed of X3 SP CD4⁺ T cells, and the inclusion of protein boost further enhanced the frequency of these cells (data not shown). The protein boost did not significantly increase the frequency of CXCR5⁺Ki-67⁺ cells beyond that induced by second MVA alone. Interestingly, however, gp140 protein boost resulted in a significant induction of X3 expression on GC T_FH cells and CXCR5⁺PD-1⁺ cells in the LN (Fig. 4A). Although the relative contribution of gp140 Ag versus alum or a combination thereof to this phenotype cannot be discerned, these results demonstrate that alum did not skew against a Th1 CD4⁺ helper recall response established by the preceding DNA/MVA immunizations. Interestingly, we observed a direct positive association between the magnitude of GC B cell responses and proportion CXCR3⁺ GC T_FH cells within LNs in the DDMM-Pro group (Fig. 4A, right panel). As anticipated, CXCR3⁻ GC T_FH cells showed an inverse correlation (Supplemental Fig. 2A).

The increase in X3 expression on CXCR5⁺ cells following MVA boost led us to examine phenotypic differences among X3⁻ and X3⁺ subsets in the LN to understand possible functional differences among these two subsets. We started by examining expression of the transcription factor Bcl-6, a unique and characteristic marker of GC cells (Fig. 4B). PD-1⁺⁺ GC T_FH cells exclusively expressed Bcl-6 albeit at lower levels compared with GC B cells (data not shown). Within GC T_FH cells, the CXCR3⁺ subset expressed lower levels of Bcl-6 on a per cell basis in most of the animals. As expected, decreased levels of Bcl-6 were associated with higher relative expression of T-bet in CXCR3⁺ GC T_FH cells. This dichotomy in T-bet expression was true for all CXCR5⁺ and CXCR5⁻PD-1⁺ subsets. These data suggest that X3 expression identifies Th1-like T_FH cells in GCs and extrafollicular regions.

Because Bcl-6 positively regulates expression of PD-1 (9), we next examined PD-1 expression and found that X3⁺ cells also expressed lower amounts of PD-1 on a per cell basis. This difference was only observed within the GC T_FH subset. We asked whether the same was true for expression of another activation/costimulation marker, ICOS. Unlike PD-1, ICOS was expressed at higher levels by CXCR3⁺ GC T_FH cells reflecting expression profile of X3⁺ and X3⁻Ki-67⁺CXCR5⁺ cells in blood. Higher magnitude expression of ICOS on CXCR3⁺ GC T_FH cells is supportive of enhanced B cell helper potential. PD-1⁺ and PD-1⁻CXCR5⁺ cells expressed lower levels of ICOS compared with GC T_FH cells and demonstrated no difference in expression in the context of X3. Expression of an IFN-γ–related marker SLAM was consistently higher on X3⁺ subsets; X3⁻ GC T_FH cells did not express SLAM showing expression levels comparable to naive cells. X3⁺ subsets showed higher levels of CD95. Frequency of Ki-67 among X3⁺ and X3⁻CXCR5⁺PD-1⁺ subsets was comparable, indicating that the phenotypic differences were not due to cell cycle status of these subsets.

Levels of another memory marker, CD127 were low on GC T_FH cells but relatively higher in the X3 compartment and the same was true for the CXCR5⁺PD-1⁺ compartment, raising the possibility that X3⁺ GC T_FH cells represent memory precursors that differentiate into CXCR5⁺PD-1⁻ cells after Ag clearance.

To determine whether these phenotypic differences resulted in differential ability to provide B cell help, we sorted GC T_FH and CXCR5⁺PD-1⁺ cells from the LN based on X3 expression (Supplemental Fig. 2C). Sorted subsets were cocultured with autologous memory B cells in the presence of staphylococcal enterotoxin B. Determination of IgG in supernatant after 7 d revealed that both X3⁺ and X3⁻ GC T_FH subsets were able to provide B cell help. The relative ability was variable; two animals RGf14 and RVz12 showed 1.5- to 8-fold higher IgG in X3⁺ GC T_FH B cell cocultures compared with X3⁻ counterparts, whereas in two animals RPq13 and RWf114, the X3⁻ GC T_FH subset showed higher B cell helper potential. CXCR5⁺PD-1⁺ cell subsets also showed similar variability among X3⁺ and X3⁻ compartments. In all, the data indicate that LN CXCR3⁺ GC T_FH cells are phenotypically distinct from CXCR3⁻ GC T_FH cells, but both subsets demonstrate potential to help B cells in vitro.

Among the first B cell effectors are plasmablasts/ASCs, derived either from activation of naive B cells during a primary response or from memory B cells during Ag re-exposure (21, 22). After the second MVA boost, gp140-specific ASC responses were observed in 100% of immunized macaques (Fig. 5A). These responses ranged from 30 to 160 (mean 63) ASCs per million PBMCs in the DDMM group. Concurrent immunization of gp140 with the second MVA increased the magnitude of ASC response by 3-fold, which ranged from 30 to 400 (mean 200) ASCs per million PBMCs. MVA-specific ASC responses were comparable between the two groups and were not inhibited as a result of gp140 immunization. To explore the role of CD4⁺ T cells in fostering plasmablast responses, we examined peripheral Ki-67⁺CD4⁺ T cell responses at day 7 as a surrogate for T cell responses in the LN on or prior to day 5. After the second MVA boost, the magnitude of total Ki-67⁺CD4⁺ T cells (Fig. 5B) in the DDMM-Pro group correlated with higher gp140 ASC responses. Correspondingly, Env IFNγ⁺CD4⁺ T cells were also positively correlated with ASC responses (Supplemental Fig. 3A). The gp140 ASC response at day 5 was a key determinant of peak gp140 titers at 2 wk following the second MVA boost.

To delineate CD4 determinants of peak and memory anti-gp140 Ab responses, we first examined the kinetics of Ab by ELISA (Fig. 5C). Anti-gp140 Ab titers were below the limit of detection in 25 of 28 animals after the second DNA immunization. The first MVA immunization induced titers ranging from 5 to 200 μg/ml at week 2, which contracted by week 8 and were potently recalled after the second MVA resulting in higher magnitude gp140 titers at memory.

We next determined the neutralization potential of anti-Env Ab against tier 1 (easy to neutralize) and tier 2 (harder to neutralize) SIVE660 pseudovirus isolates. Although no detectable responses against tier 2 isolates was observed, coimmunization with alum in gp140 resulted in stronger neutralization titers against tier 1 isolates at weeks 2 and 20 following the second MVA boost (Fig. 5D). As a measure of Ab maturation, we monitored relative Ab avidity following first and second MVA immunizations. Our analyses using an ELISA against SIV239 gp140 with a 1.5 M sodium thiocyanate wash showed that gp140 avidity increased over time and reached a maximal response at week 20 after the second MVA boost (Fig. 5E). To ascertain whether inclusion of gp140 protein boost during the second MVA increased avidity of anti-Env Ab, we examined relative increase in avidity at week 20 versus week 2 and observed greater increase in avidity with inclusion of gp140 protein boost (Fig. 5F).

Next, we wanted to investigate how the GC T_FH response in the LN at week 2 after the second MVA boost influenced the magnitude and longevity of anti-Env Ab titers. The frequency of CXCR3⁺ GC T_FH cells showed a trend for correlation with the binding titers at week 20 (p = 0.05) or at week 10 (data not shown) and directly correlated with neutralizing activity and avidity of anti-Env Ab. We also observed an inverse correlation between the frequency of CXCR3⁺ GC T_FH and fold contraction in peak Ab titers (Fig. 5G). These results demonstrated that CXCR3⁺ GC T_FH cells contribute significantly for the generation of long-lived Ab responses.

Studies in humans have shown that blood X3⁺CCR6⁺CD4⁺ T cells express high levels of the HIV coreceptor CCR5 (23), raising the possibility that vaccine-mediated generation of CXCR3⁺CXCR5⁺ cells could favor viral replication upon infection. To address this possibility, we first examined CCR5 expression in LN and blood X3⁺ CD4⁺ T cells in SIV naive animals (Fig. 6A, 6B) and found that a significantly higher fraction of X3⁺CD4⁺ T cells express CCR5 compared with X3⁻ cells. This was true both in LN (Fig. 6A) and blood compartments (Fig. 6B) and within CXCR5⁺ and CXCR5⁻ subsets. Following five weekly intrarectal challenges with SIVmac251, all vaccinated animals became infected but showed a 2-log-fold decrease (p < 0.01) in peak viremia compared with unvaccinated controls (data not shown). The induction of CXCR3⁺CXCR5⁺ CD4 T_FH cells following vaccination and the higher expression of CCR5 on CXCR3⁺ CD4 T cells prompted us to test associations between CD4⁺ T cell response induced at peak after second MVA boost and peak viremia at 3 wk post-SIV challenge. We found a moderate but significant direct correlation with Ki-67⁺CXCR5⁺X3⁺CD4⁺ T cells and peak viremia (Fig. 6C, last panel). Interestingly, this association was not observed with either total Ki-67⁺CD4⁺ T cells or CXCR5⁺X3⁻CD4⁺ T cells.

To determine whether permissiveness of X3⁺ T_FH cell subsets to SIV infection facilitated seeding of reservoirs, we sorted LN CXCR5⁺ cells (PD-1⁺⁺ and PD-1⁺) based on X3 expression at 3 wk postinfection from several animals with a log-fold difference in viral RNA copies in plasma; the data showed that in five of the six animals sampled X3⁺PD-1^+/++ T_FH cells harbored significantly higher levels of proviral DNA compared with X3⁻ counterparts (p < 0.05) (Fig. 6D).

We do not, at present, fully understand the relationship between the SIV-specific CD8 T cell responses postinfection and the infection status of CXCR3⁺ T_FH cells. However, we observed that the frequency of Ki-67⁺ CD8 T cells associated with peak viral load and positively correlated with frequency of GC T_FH cells and proportion of CXCR3⁺ GC T_FH cells, consistent with Ag or inflammation-driven immune activation of T cells (Supplemental Fig. 4). More detailed studies are needed to investigate the dynamics of Ag-specific CD8 T cell responses in peripheral blood and lymphoid compartments in controlling frequency and infection status of CXCR3⁻ and CXCR3⁺ T_FH cell subsets.

In summary, our data raise the possibility that vaccination regimens that skew toward generation of X3⁺CXCR5⁺ cells, while favoring durable, higher-avidity Ab, potentially could facilitate viral replication and seeding of reservoirs by generating target cells that are efficiently recalled upon Ag re-exposure.

Post-RV144, HIV vaccine strategies to augment anti-Env Ab responses have generated immense interest and concurrent immunization of Env subunit proteins with recombinant viral vectors expressing HIV immunogens is emerging as a promising HIV vaccine modality (24). Although a great deal of research has focused on Ab responses following Env subunit immunization, relatively little is known about the CD4⁺ helper response and, in particular, the B cell helper CD4⁺ T_FH cell response resulting from these immunization regimens. A better understanding of CD4 helper correlates of Ab responses will, undoubtedly, yield useful insights for rational vaccine design (25).

Using the best available experimental model for HIV vaccines—the rhesus macaque—we conducted an in-depth characterization of the CD4⁺ T cell response and examined its effects on Env B cell responses in the context of a DNA/MVA⁺ gp140 vaccine regimen. Our data allow for three main conclusions; first, DNA/MVA vaccine induces CXCR5⁺CD4⁺ T cells with B cell helper potential, which circulate at the peak of immune response. Second, vaccine-induced CXCR5⁺CD4⁺ T cells are skewed toward expression of CXCR3, which is augmented by coimmunization with alum-adjuvanted gp140 protein. Third, CXCR3 expression on CXCR5⁺ LN-resident CD4⁺ T cells is strongly associated with Ab persistence and with Ab avidity. The data offer an important parameter with which to predict Env memory responses and suggest that strategies to target the CXCR5⁺CXCR3⁺ subset could provide a means to enhance Env titers. However, an association between vaccine-raised CXCR5⁺CXCR3⁺ cells and peak viral load, albeit modest, suggests that these cells could represent targets for viral replication and predispose to higher acute viremia upon infection. Taken together, these results provide a strong rationale to investigate and delineate protein-adjuvant driven CXCR5⁺CXCR3⁺CD4⁺ T cell responses in HIV vaccine studies to better understand CD4 helper correlates of vaccine efficacy and viral control.

The chemokine expression profile of CD4⁺ T cells typifies differentiation status and is regulated by the antigenic stimulus and the resulting inflammatory response (7). MVA is a potent inducer of IFN-γ production in DCs (26), and significant systemic induction of IFN-γ–induced protein 10, one of the ligands of CXCR3, is observed after primary and booster MVA immunization in rhesus macaques (27). Because CXCR3 is induced by IFNs, the generation of CXCR3 skewed CD4 response after MVA immunization is anticipated. Notably, this phenotype was augmented by addition of alum-adjuvanted gp140 and was accompanied by an increase in Env-specific IFNγ⁺,IL-4⁺ responses. Induction of a Th1 bias with alum, which is considered largely as a Th2 adjuvant, is surprising but is not without precedence. Studies in mice show that OVA in alum immunization results in CXCR3 expression on ASCs and GC B cells, which is mediated by IFNγ⁺ Ag-specific CD8⁺ T cells (28, 29). Our data show that Th1 and Th1-skewed CXCR5⁺ recall responses primed by preceding DNA/MVA immunizations were reacquired after the second MVA boost and were not inhibited by alum adjuvant. This phenotype, likely driven by vaccine-mediated induction of T-bet in CD4 T cells, is consistent with the proposed model of T_FH and Th1 CD4⁺ helper differentiation during recall responses (30).

The mechanisms by which CXCR3⁺ GC T_FH cells promote Ab longevity and avidity remain to be understood. Our in vitro assays indicate that both X3⁻ and X3⁺ subsets were effective at helping B cells. The dynamics of X3⁺ cells in vivo in fostering Ab responses could be due to differential localization and/or due to enhanced expression of B cell helper cytokines and costimulatory signals in the context of an IFN-γ–rich environment. With regards to HIV, there is evidence that CXCR3⁻CXCR5⁺ cells possess superior B cell help potential. Recently, Locci et al. (9) reported higher percentage of blood PD-1⁺CXCR3⁻CD4⁺ cells (within total CXCR5⁺ cells) at 4 and 40 mo postinfection in HIV⁺ individuals that develop broadly neutralizing Abs. Furthermore, CXCR3⁺ but not CXCR3^- cells were effective at providing B cell help in vitro. It is possible that phenotypic attributes of T_FH cells that foster Ab responses during vaccination and during infection are very different especially during chronic Ag stimulation where CD4⁺ T cells are viral targets. This is particularly relevant as CCR5 expression is largely confined to CXCR3⁺CD4⁺ subsets. In our study, we found that the magnitude of vaccine-induced X3⁺ CXCR5⁺CD4⁺ T cells was significantly associated with peak viral load postinfection. This would suggest that rapid recall of vaccine induced X3⁺CXCR5⁺CD4⁺ T cells upon Ag re-exposure results in generation of target cells for viral replication (31). These data indicate that in the context of a Th1-skewed immunization regimen, CXCR3⁺ T_FH cells favor Ab responses but could skew toward higher peak viremia.

Our findings underscore a pressing need to better understand the effect of adjuvanted Env subunit protein immunizations on CD4⁺ T cell helper responses in humans. Protein immunizations augment CD4⁺ responses and typically do not raise/boost CD8 responses (32). The generation of CD4 viral target cells that are efficiently recalled upon Ag re-exposure in absence of rapid CD8 recall responses could favor initial viral replication. There is evidence supporting this paradigm—a previous macaque study by Buge et al. (33) demonstrated that inclusion of alum-adjuvanted gp120 boost to a DDM vaccination regimen resulted higher peak viremia following a simian HIV challenge compared with DDM alone. Several studies have shown that a combination regimen of Env subunit proteins and viral vectored vaccines does not improve control of viremia compared with subunit protein immunization alone (32, 34). In light of the study by Buge et al. and the present findings, it would be important to include a viral vector alone vaccine arm to determine whether of addition of adjuvanted protein boost results in higher peak viremia. This is critical as extended boosting regimens with subunit proteins are being tested in RV144 vaccinees as a means to overcome waning humoral immune responses over time. Although multiple booster immunizations offer the advantage of augmenting magnitude and quality of Ab responses, their potential to engender target CD4⁺ T cells without boosting CD8 responses could tip the balance in favor of viral replication. In this context, characterizing and understanding CD4 responses in terms of CXCR5 and CXCR3 expression within genital and rectal mucosa, the major portals of HIV entry could provide significant insights into their roles in acquisition of infection.

In conclusion, our data present an integrated view of CD4 helper determinants of Env Ab responses in a multicomponent HIV vaccine in rhesus macaques. The study adds to our understanding of vaccine-induced immune responses; further research is needed to determine the application of our findings to other leading poxvirus vectors and to HIV vaccine outcomes in humans.

We thank the Yerkes Division of Research Resources, Stephanie Ehnert, Christopher Souder, Robert Sheffield, and all the animal care staff for immunizations, infections sample collections, and animal care. We thank Drs. Elizabeth Strobert, Sherrie Jean, and veterinary staff for veterinary counsel. We also thank the Emory Flow Cytometry Core, Barbara Cervasi, and Kiran Gill for cell sorting. We thank Nancy Miller for SIV251challenge stock; Patricia Earl and Jeffrey Americo for providing the MVA virus; the National Institutes of Health AIDS Research and Reference Reagent Program for the provision of peptides; the Emory University Center for AIDS Research Virology Core, Benton Lawson, and Melon T. Nega for assays on viral load; Shane Crotty and Colin Havenar-Daughton for critical input regarding CXCR5 staining in rhesus samples; Chris Ibegbu, Rama S. Akondy, and Vijayakumar Velu for suggestions on Abs for flow cytometry; Michael J. Sabula for technical help; and Reben Rahman for assistance with figures.

R.R.A. is the coinventor of DNA/MVA technology that has been licensed to Geovax, Inc. by Emory University. The other authors have no financial conflicts of interest.