Follicular regulatory T cells impair follicular T helper cells in HIV and SIV infection

Human and simian immunodeficiency viruses (HIV and SIV) exploit follicular lymphoid regions by establishing high levels of viral replication and dysregulating humoral immunity. Follicular regulatory T cells (TFR) are a recently characterized subset of lymphocytes that influence the germinal centre response through interactions with follicular helper T cells (TFH). Here, utilizing both human and rhesus macaque models, we show the impact of HIV and SIV infection on TFR number and function. We find that TFR proportionately and numerically expand during infection through mechanisms involving viral entry and replication, TGF-β signalling, low apoptosis rates and the presence of regulatory dendritic cells. Further, TFR exhibit elevated regulatory phenotypes and impair TFH functions during HIV infection. Thus, TFR contribute to inefficient germinal centre responses and inhibit HIV and SIV clearance.

H IV rapidly establishes a productive viral infection in secondary lymphoid tissues, which is maintained during chronic stages of disease 1,2 . During chronic HIV infection, viral replication is highly concentrated within B-cell follicles in follicular T helper cells (T FH ) [3][4][5] . T FH are crucial initiators of the germinal centre (GC) response 6,7 . T FH have a distinct developmental pathway characterized by Bcl-6 expression, which is dependent on inducible T-cell costimulator (ICOS) expression 8 , and produce interleukin (IL)-21 and IL-4 that together optimally drive B-cell affinity maturation and antibody specificity 9,10 . ICOS expression on T FH is crucial for both T FH differentiation and immune function 8 . An expansion of T FH cells has been observed in HIV infection 11 and simian immunodeficiency virus (SIV) infection 12 , yet this expansion does not correlate with improved GC responses. Rather, it has been shown that T FH exhibit impaired activity, partly due to PD-1 ligation, manifested by reduced ICOS expression and inadequate production of IL-21 during HIV infection 13 . It remains unclear whether additional factors may drive the dysregulation of T FH during HIV and SIV infection.
It has recently come to light that B-cell follicles contain a novel subset of regulatory T cell (Treg), termed follicular regulatory T cells (T FR ) [14][15][16] . T FR display a unique transcriptional pattern overlapping that of both T FH and Treg, notably with combined expression of Bcl-6, Foxp3 and Blimp-1. T FR originate from Treg precursors, express CXCR5 and regulate GC responses through interactions with T FH [14][15][16] . These studies were performed in mouse models, however, and the presence or function of T FR have not yet been described in HIV or SIV infection. Some [17][18][19][20][21] , but not all [22][23][24][25] studies suggest proportional, not numerical, Treg increases in the peripheral blood of HIV-infected individuals. Studies in lymph nodes (LNs) and the spleen consistently suggest proportional increases of Treg in the context of HIV or SIV infection [26][27][28] , although absolute numbers have not been determined. The impact of Treg on HIV infection is controversial with some studies suggesting that Treg exert a beneficial effect by limiting autoimmunity, HIV replication and CD4 þ T-cell depletion 17,18,24,25 , whereas others suggest that Treg have a detrimental effect by inhibiting HIV-specific immune responses and causing disease progression 20,21,28,29 . Although it is reported that Treg from HIV-infected individuals have lower suppressive capacity than those from uninfected individuals 30 , it has also been reported that HIV binding to Tregs enhances their suppressive activity and lymphoid homing 31 . Thus, understanding the role of Treg in HIV infection is still evolving 32 , and virtually nothing is known about T FR number and function in HIV infection. Here, we provide evidence for HIV-mediated T FR expansion and the role of T FR in T FH dysregulation during HIV and SIV infection.
Through analyses of secondary lymphoid tissues from chronically HIV-infected humans and chronically SIV-infected rhesus macaques, as well as HIV infection ex vivo of human tonsils, we find that T FR are expanded both proportionally and numerically during infection. This expansion is due to a combination of factors, including viral entry and replication, Treg acquisition of CXCR5, transforming growth factor (TGF)-b signalling, T FR proliferation, low apoptosis rates and increased regulatory dendritic cell (DC) activity. In addition, we demonstrate that T FR suppress T FH activity during infection by inhibiting T FH proliferation, IL-21 and IL-4 production and downregulating T FH ICOS expression. The identification of this potent regulator of GC dynamics provides a new therapeutic target for enhancement of anti-viral humoral immunity and vaccine efficacy to promote clearance of HIV.

Results
T FR are increased in chronic HIV and SIV Infections. To determine if T FR were present in human lymphoid tissues, we immunofluorescently labelled LN tissue cross-sections from HIV uninfected and HIV-infected individuals with antibodies to CD4, Foxp3, CD20 and IgD. CD4 þ Foxp3 þ cells were readily detected throughout the LNs including follicular and GC regions, as shown in representative images ( Fig. 1a and Supplementary  Fig. 1a). Next, we quantified the number of CD4 þ Foxp3 þ cells in total LN, follicular (CD20 þ ) and GC (CD20 þ IgD À ) regions using two to three tissue cross-sections per subject. In total LN, the frequencies of CD4 þ Foxp3 þ cells per mm 2 were significantly higher in HIV-infected individuals compared with uninfected controls, but frequencies of CD4 þ Foxp3 þ in follicular and GC regions did not differ significantly (Fig. 1b). LN enlargement is a hallmark of HIV infection, and therefore frequencies of CD4 þ Foxp3 þ cells per mm 2 underestimate the true differences in absolute numbers of CD4 þ Foxp3 þ cells. The average LN, follicular and GC areas in LN cross-sections, as determined by quantitative image analysis, were significantly larger in HIV-infected individuals than seronegative individuals (Fig. 1c). We then calculated the number of CD4 þ Foxp3 þ cells per average LN tissue cross-section as well as within each compartment. Absolute numbers of CD4 þ Foxp3 þ cells per average cross-section were significantly increased in HIV-infected individuals compared with HIV-uninfected individuals in all regions (Fig. 1d). The percentage of tissue area that stained positively for CD4 did not differ between HIV seropositive and seronegative individuals, indicating that these individuals had not yet experienced profound CD4 depletion in the LN ( Supplementary Fig. 1b). In HIV-infected subjects, the percentage of activated (HLA-DR þ CD38 þ ) LN CD4 þ T cells, as previously determined by flow cytometry 33 , did not correlate with number of Treg or T FR and was inversely correlated with GC T FR in the average tissue cross-section ( Supplementary Fig. 1c).
To determine if T FR are likewise increased during SIV infection, we analysed lymphoid cells from uninfected or SIVmac239-infected rhesus macaques by flow cytometry. As SIV infection reduces cell surface CD4, we gated on viable CD3 þ CD8 À cells, as previously described in humans 33 . The surface phenotype of CD25 hi CD127 À has been shown in humans to encompass Tregs and allow for live cell sorting [34][35][36] . We validated that this phenotype accurately detects Tregs in SIV-uninfected and SIV-infected rhesus macaques (Fig. 2a). Here, Tregs were defined as viable CD3 þ CD8-CD25 hi CD127 À , whereas T FR were defined as the subpopulation that are CXCR5 þ , and GC T FR as the subpopulation that are CXCR5 hi and PD-1 hi . In chronic SIV infection, percentages of Tregs, and particularly T FR and GC T FR , were significantly increased compared with SIV-uninfected animals (Fig. 2b). We next determined the ratio of regulatory cell populations to their nonregulatory counterpart and found the ratios of T FR to T FH and GC T FR to GC T FH were significantly higher than those from SIVuninfected animals, whereas the overall ratio of Treg/non-Treg CD3 þ CD8 À cells did not differ between infected and uninfected animals (Fig. 2c). Treg, T FR and GC T FR subsets also expressed significantly higher levels of CTLA-4 in SIV-infected compared with SIV-uninfected animals (Fig. 2d).
T FR expansion is mediated by HIV infection. To characterize mechanisms underlying T FR expansion in HIV/SIV infection, we utilized an ex vivo model of HIV-1 infection in human tonsil cells. Disaggregated tonsil cells were spinoculated with NL4-3 based X4-or R5-tropic green fluorescent protein (GFP) reporter viruses and cultured for 2 days. Productive infection was confirmed by the presence of GFP expression after 2 days and percentages of GFP þ T FR ranged from 1 to 15%. Using the flow gating strategy described above for rhesus macaques, we demonstrated that CD25 hi CD127 À cells encompassed the majority of Foxp3 þ cells (Fig. 3a). Compared with mock-spinoculated tonsil cells, a 24-h stimulation of mock-spinoculated tonsil cells with phorbol 12myristate 13-acetate (PMA) and ionomycin did not significantly increase T FR , whereas the positive control of exogenous TGF-b (ref. 37) did (Fig. 3b). A more marked expansion of T FR was observed with either X4-or R5-spinoculation (Fig. 3b). Pretreatment of cells before spinoculation with chemokine co-receptor antagonists, bicyclam (AMD) for X4 and maraviroc (MVC) for R5, inhibited on average 80 and 86% of productive HIV infection, respectively. Both AMD and MVC prevented T FR expansion (Fig. 3b). To determine if T FR expansion represented true numerical increases, mock-and X4-spinoculated cultures were analysed using flow cytometry counting beads. Total CD3 þ CD8 À cells were decreased and T FH cells were slightly lower 2 days after X4-spinoculation compared with  mock-spinoculated cells (Fig. 3c). Conversely, the T FR population was approximately threefold higher in X4-spinoculated cells compared with mock-spinoculated cells (Fig. 3c). We further confirmed the phenotype of T FR by determining expression of the transcription factors Bcl-6 and Blimp-1, and evaluated the effects of ex vivo HIV-1 infection on their expression. Bcl-6 expression was significantly higher in both T FH and T FR compared with non-follicular (CXCR5 À ) populations in mock-spinoculated cells at day 2 ( Fig. 3d). With X4 or R5 spinoculation, T FH had significantly lower Bcl-6 expression compared with mockspinoculated cells, whereas Bcl-6 expression in T FR was not significantly altered by HIV spinoculation (Fig. 3d). Blimp-1 expression was significantly higher in non-follicular and T FR populations compared with T FH in mock-spinoculated cells at day 2, but was not significantly altered after X4 or R5 spinoculation (Fig. 3e).  Of viable CD3 þ CD8 À cells, follicular subsets were defined as CXCR5 þ cells (F) and germinal centre subsets were defined as CXCR5 hi PD-1 hi cells (GC). Of these subsets, regulatory cells were defined as CD25 hi CD127 À . T FR (CXCR5 þ CD25 hi CD127 À ) were Foxp3 þ , whereas T FH (CXCR5 þ CD25 lo/ À ) were IDO and TGF-b production contribute to T FR expansion. To further address viral life cycle contribution to T FR expansion, tonsil cells were treated with the integrase inhibitor raltegravir (RAL) after spinoculation. RAL blocked on average 91% of productive HIV infection and significantly decreased, but did not fully suppress, T FR expansion (Fig. 4a). In addition, TGF-b is crucial for stable Foxp3 expression in Tregs, so culture supernatants were analysed for TGF-b production. Stimulation of mock-spinoculated cells with PMA and ionomycin did not lead to increased TGF-b production, whereas the presence of either X4  From viable CD3 þ CD8 À cells, T FR are defined as CXCR5 þ and CD25 hi CD127 -. T FR cells contain Foxp3 þ cells, whereas remaining T FH (CXCR5 þ CD25 lo/ À ) cells are Foxp3 À . (b) Percentages of T FR determined by gating strategies in a are shown. Experimental conditions include mock-spinoculated cells cultured with PMA (50 ng ml À 1 ) and ionomycin (1 mg ml À 1 ) or exogenous TGF-b (100 ng ml À 1 ) for 24 h and cells pretreated to block CXCR4 (AMD, 200 mM) and CCR5 (MVC, 2 mM). (c) Using flow cytometry counting beads, the number of cells per ml were determined for total (CD3 þ CD8 À ), T FH (CXCR5 þ CD25 lo/ À ) and T FR (CXCR5 þ CD25 hi CD127 À ) subsets in mock-and X4-spinoculated samples (n ¼ 3). (d) Bcl-6 expression is shown in CXCR5 À (grey), T FH (blue) and T FR (red) populations after mock-, X4-or R5-spinoculation (n (c) Tonsil cells were cultured in the presence of TGF-b blocking antibodies (2 mg ml À 1 ) and the percentages of T FR were measured (n ¼ 4). Addition of anti-TGF-b antibodies did not influence cell viability. (d) Tonsil cells were cultured under the conditions shown and then analysed for the presence of immature myeloid DCs (CD11c þ CD1c þ HLA-DRlo) and mature myeloid DCs (CD11c þ CD1c þ HLA-DR þ CD83 þ ; n ¼ 6). (e) Samples in d were also analysed for activated plasmacytoid DCs (CD123 þ HLA-DR þ , n ¼ 6). (f) Cell culture supernatants and lysates from d,e were analysed by ELISA to quantitate IDO production (n ¼ 6). The horizontal bars of each graph indicate the median value and are listed where appropriate for clarity. Statistical analyses were performed using Wilcoxon matched-pairs tests (a,c) or Mann-Whitney ranked sums tests (b,d-f) and significance is denoted by asterisks where *Po0.05, **Po0.01 and ***Po0.001. or R5 virus, even with AMD or MVC pretreatments before spinoculation, led to significant increases in TGF-b (Fig. 4b).
When tonsil cells were cultured in the presence of TGF-bneutralizing antibodies, T FR expansion was inhibited in both X4-and R5-spinoculation (Fig. 4c).
Regulatory DC are described as myeloid and plasmacytoid DCs that have an immature or semi-mature phenotype and produce anti-inflammatory molecules such as indoleamine 2,3-dioxygenase (IDO) that generates Tregs 38 . It has previously been reported that regulatory DCs inhibit effector T-cell function by exhibiting a lack of co-stimulatory molecule expression 39 and can produce IDO in lymphoid tissues during HIV and SIV infection [40][41][42] . In tonsil cultures, there were significantly more immature myeloid DCs with both X4-and R5-spinoculation, but no significant increase in mature myeloid DCs when compared with mock-spinoculated cells (Fig. 4d). X4-and R5-spinoculation also significantly increased activated plasmacytoid DC compared with mock spinoculation (Fig. 4e). Culture supernatants and lysed cells were then measured for IDO production. There was a significant increase in IDO with either X4-or R5-spinoculation compared with mock-spinoculation (Fig. 4f). We further analysed these culture supernatants for interferon (IFN)-g production, as IFN signalling is upstream of IDO production. X4-and R5-spinoculated cultures tended to have higher levels of IFN-g, but this increase was not statistically significant ( Supplementary  Fig. 2a).
T FR have high proliferation and low apoptosis rates. In mice, T FR arise from conventional Tregs, rather than from T FH 15 . To investigate the origin of T FR in human cells, we evaluated whether tonsil T FH phenotypically converted to T FR or whether Tregs gain CXCR5 expression. T FH (CD3 þ CD8 À CXCR5 þ CD25 À ) and non-follicular (CD3 þ CD8 À CXCR5 À ) T-cell populations were sorted and cultured without HIV-spinoculation for 2 days. Interestingly, T FH did not acquire Foxp3 expression, even in the presence of exogenous TGF-b, whereas a small population of cells from the initial CXCR5 À population acquired CXCR5 expression (Fig. 5a). These cells, labelled 'new' T FR , expressed high levels of Foxp3 (Fig. 5a).
We further hypothesized that cell proliferation may contribute to T FR expansion. We utilized 5-bromodeoxyuridine (BrdU) labelling in tonsil cells to measure the relative rates of T FR and T FH proliferation. Tonsil cells were mock-, X4-or R5-spinoculated and BrdU DNA incorporation was measured after 2 days by flow cytometry. BrdU incorporation rates were low overall, but in the context of HIV infection, T FR incorporated significantly more BrdU than T FH (Fig. 5b).
To evaluate the duration of tonsil T FR increases, we extended the counting experiments in Fig. 3c for 5 days. Total CD3 þ CD8 À , T FH and T FR populations from mock-spinoculated samples remained numerically stable up to 5 days (Fig. 5c). In X4-spinoculated samples, total and T FH cells declined by day 5, whereas T FR were elevated at day 2 and remained so through day 5 (Fig. 5c). We also analysed cell death rates by Annexin-V and propidium iodide (PI) staining in mock-and X4-spinoculated tonsil cells. Total CD3 þ CD8 À cells and T FH cells displayed increased rates of late apoptosis (Annexin þ PI þ ) and necrosis (PI þ ) in X4-spinoculated samples compared with mock-spinoculated cells (Fig. 5d). The majority of T FR remained in a state of viability or early apoptosis (Annexin þ ) in X4-spinoculated tonsil cells, similar to what was seen with mock-spinoculated cells (Fig. 5d). To determine if this occurs during in vivo infection, we analysed Annexin-V staining in disaggregated rhesus macaque lymphoid cells. The levels of Annexin-V staining in total CD3 þ CD8 À and Treg populations were significantly higher in SIV-infected animals compared with SIV-uninfected animals (Fig. 5e). There was not a significant difference in the level of Annexin-V staining in T FH cells in SIV-infected compared with SIV-uninfected animals (Fig. 5e). In contrast, the levels of Annexin-V staining in T FR and GC T FR were significantly lower in SIV-infected animals compared with SIV-uninfected animals (Fig. 5e).
T FR regulatory phenotypes in HIV infection ex vivo. We next sought to determine the regulatory phenotype of tonsil T FR during infection. CTLA-4 and LAG-3 are used to contact and inhibit effector cell function, GITR is expressed on activated and proliferating cells, whereas galectin receptors are an important family in promoting effector cell apoptosis 43 . Both X4-and R5-spinoculation caused T FR to significantly upregulate total CTLA-4 ( Fig. 6a) and LAG-3 (Fig. 6b), as well as GITR and galectin-3, whereas galectin-9 expression was not altered (Supplementary Fig. 2b). Regulation of T FH has previously shown to occur through PD-1 ligation 13 , however, PD-L1 was expressed on fewer than 3% of T FR following either mock or X4 spinoculation (n ¼ 3). Blockade of chemokine co-receptors partially inhibited CTLA-4 and GITR upregulation on T FR ( Fig. 6a Fig. 3c).
IL-10 regulates effector cell function, whereas TGF-b prevents inflammatory responses and promotes Foxp3 expression 43 . IL-10 and TGF-b (LAP-1) production by tonsil T FR was determined by intracellular cytokine staining. Mock-spinoculated cultures treated with exogenous TGF-b as a positive control showed a significant increase of IL-10 þ T FR (Fig. 6c). X4-and R5-spinoculation led to significant increases of IL-10 þ T FR , which tended to be higher than TGF-b-treated T FR (Fig. 6c). Similar trends were observed in regards to TGF-b þ T FR (Fig. 6d). In addition, analysis of culture supernatant showed that infection significantly upregulated IL-10 compared with mockspinoculated controls ( Supplementary Fig. 2c), whereas the presence of X4 or R5 virus did not significantly increase IL-35 production ( Supplementary Fig. 2d).
T FR dysregulate T FH activity. T FR have been shown to modulate T FH function and thereby control GC reactions in mice [14][15][16] . Given that T FH exhibit impaired function in chronic HIV infection 13 we investigated whether T FH were impaired by T FR in the tonsil model. We first compared the expression of ICOS, a key molecule for T FH maintenance and function, in X4-or R5-spinoculation to mock-spinoculation. With both X4-and R5-spinoculation, T FH demonstrated a significant reduction of ICOS expression (Fig. 7a). We evaluated the role of Tregs in T FH ICOS downregulation by depleting CD25 þ regulatory cells, which effectively removes all Foxp3 þ cells from tonsil cell culture, and analysing T FH after 2 days. ICOS expression on T FH was high in mock-spinoculated controls regardless of the presence or absence of Treg (Fig. 7b). With X4-or R5-spinoculation, however, ICOS expression was not downregulated when Tregs were absent (Fig. 7b). Although CD25 þ cell depletion effectively removes overall regulatory populations, we could not rule out that T FR may have functions differing from those of Tregs. We therefore examined the regulatory effects of T FR specifically by isolating tonsil T FH and adding T FR back to T FH at increasing ratios. In X4-and R5-spinoculated cultures, we observed that T FH ICOS levels were decreased as the ratio of T FR increased (Fig. 7c).
Previous studies using ex vivo suppression assays have shown that Treg inhibit CD4 þ T-cell proliferation and cytokine production in HIV infection 18,21 . First, we investigated the ability of T FR to suppress proliferation of T FH in the context of ex vivo HIV infection. Tonsil T FH and T FR were isolated, spinoculated with either X4 or R5 virus, and cultured either alone or with equal numbers of T FR for 4 days in the presence of anti-CD3/anti-CD28 antibodies and IL-2. T FR consistently inhibited proliferation of T FH in both X4-and R5-spinoculated cultures (Fig. 7d).
Next, we investigated the role of T FR on T FH cytokine production. Initially, we observed that tonsil T FH display decreased IL-4 production in either X4-or R5-spinoculated cultures, whereas IL-21 production did not have a uniform trend (Supplementary Fig. 4a). Removal of CD25 þ Tregs led to significant increases in IL-4 (Fig. 8a) and IL-21 (Fig. 8b) production by tonsil T FH following mock-, X4-and R5-spinoculation and significant increases in IL-4 and IL-21 production by T FH in cells from both uninfected and chronically SIV-infected animals (Fig. 8c). We further validated these observations by co-culturing tonsil T FH with increasing ratios of T FR following spinoculation with X4 virus. The absence of T FR led to increased T FH production of IL-4 and IL-21 (Fig. 8d). The regulation of T FH was T FR dose-dependent, as an increasing ratio of T FR led to increasing inhibition of T FH IL-4 and IL-21 production (Fig. 8d). To determine if the suppressive   (a) T FH and CXCR5 À T-cell populations were sorted and cultured without stimulation, or T FH were cultured in the presence of exogenous TGF-b (100 ng ml À 1 ). CXCR5 expression was analysed on sorted CXCR5 À cells after 2 days and cells expressing CXCR5 are labelled as 'new T FR '. The Foxp3 expression levels of T FH , T FH cultured with TGF-b and 'new' T FR were determined (n ¼ 3). (b) T FH and T FR were mock-, X4-or R5-spinoculated and cell proliferation measured by BrdU incorporation after 2 days of culture (n ¼ 7). (c,d) A 5-day time course was performed to monitor T-cell population counts and rates of cell death with mock-or X4-spinoculated tonsil cells (n ¼ 3). (c) Average counts of total (CD3 þ CD8 À , circles), T FH (CXCR5 þ CD25 lo/ À , squares) and T FR (CXCR5 þ CD25 hi CD127 À , triangles) are shown for the duration of culture (n ¼ 3). (d) Stages of cell death are shown at day 5 and defined as early apoptosis (AnnexinV þ ), late or advanced apoptosis (AnnexinV þ PI þ ) and necrotic death (PI þ ; n ¼ 3). (e) Cell subsets from disaggregated lymphoid tissues of chronically SIV-infected (n ¼ 6) and uninfected rhesus macaques (n ¼ 8) were analysed for apoptosis by percent Annexin-V binding. Cell phenotypes were defined as total    ARTICLE capacity of T FR was due to the production of suppressive cytokines, we performed intracellular cytokine assays in the presence of both IL-10-and TGF-b-neutralizing antibodies. Previous reports show that IL-10 potently blocks IL-4 production 44 , so we specifically addressed IL-21 production by T FH . Neutralization of IL-10 and TGF-b restored some, but not the majority of IL-21 production by T FH in the presence of T FR (Fig. 8e).

Discussion
In the present study, we provide a comprehensive analysis of T FR in HIV and SIV infection. We show that T FR are expanded in  (a) Tonsil T FH were analysed for surface ICOS, following mock-, X4-, or R5-spinoculation (n ¼ 7). (b) CD25 þ regulatory cells were depleted from each culture condition and ICOS expression on T FH was analysed (n ¼ 3). (c) Tonsil T FH and T FR were isolated, spinoculated and co-cultured for 2 days at the indicated ratios. Representative images of ICOS expression on T FH surface are shown (n ¼ 2). (d) Tonsil T FH were isolated, labelled with eFluor 670 proliferation dye, and cultured alone or with an equal number of T FR . Proliferation was measured by dye dilution (n ¼ 4). The horizontal bars of each graph indicate the median value and are listed where appropriate for clarity. Statistical analyses were performed using Friedman nonparametric tests (a) and significance is denoted by asterisks where *Po0.05.  (a) Tonsil T FH were analysed by intracellular cytokine staining for IL-4 and IL-21 production following mock-, X4-or R5-spinoculation. (b) CD25 þ regulatory cells were depleted before spinoculation and IL-4 (n ¼ 3) and IL-21 (n ¼ 7) production by T FH cells were analysed after 2 days. (c) IL-4 and IL-21 production by T FH were analysed in disaggregated lymph node cells from rhesus macaques with and without CD25-depletion (n ¼ 3). (d) Sorted populations of tonsil T FH and T FR were spinoculated, T FR were added back to T FH at an increasing ratio, and T FH production of IL-4 and IL-21 were measured by intracellular cytokine assays at day 2 (n ¼ 4). (e) Sorted tonsil T FH and T FR were cultured for 2 days in the presence of IL-10-and TGF-b-neutralizing antibodies and T FH production of IL-21 was analysed (n ¼ 2). Statistical analyses were performed by Friedman nonparametric tests (a,b) and significance is denoted by asterisks where *Po0.05 and ***Po0.001. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9608 ARTICLE secondary lymphoid tissues during both chronic HIV and SIV infection. In addition, we utilize HIV infection ex vivo to show that cell proliferation, regulatory DCs, TGF-b signalling, Treg acquisition of CXCR5 and resistance to apoptosis play a role in T FR expansion. Finally, we demonstrate that T FR are able to inhibit proliferation, ICOS expression and IL-4 and IL-21 production by T FH . Collectively, these findings reveal that T FR play a crucial role in the impairment of T FH during HIV and SIV infection.

X4-T FH
Although Tregs are diminished in gut-associated lymphoid tissue during HIV infection 23 , prior studies indicate an increase in Tregs in LNs 21,26,45 . We found that absolute numbers of CD4 þ Foxp3 þ cells in follicles and GCs per average LN crosssection were increased in HIV-infected individuals compared with uninfected individuals. Furthermore, following ex vivo infection of human tonsil cells, we demonstrated that T FR increase both proportionally and numerically. As Foxp3 can be transiently upregulated after T-cell activation, we evaluated whether percentages of activated cells were related to Foxp3 expression. Nevertheless, we found no correlation between percentages of activated LN cells and Tregs and T FR , and numbers of GC T FR were inversely related to percentages of HLA-DR þ CD38 þ CD4 þ cells, which is the opposite of what would expect if immune activation was driving Foxp3 expression.
Blockade of the HIV chemokine co-receptors CXCR4 and CCR5 prevented T FR expansion in tonsil cells, thus demonstrating that viral entry was necessary to promote expansion. Intriguingly, however, treatment with the integrase inhibitor RAL, which blocks the virus replication cycle after cell entry, largely blunted but did not completely ablate T FR expansion. It has recently been demonstrated that many cells during ex vivo HIV infection are abortively infected and only a minority of infected cells are productively infected 46 . Viral sensing mechanisms are important for innate immune cell activation in HIV infection and similar mechanisms may also contribute to T FR activation and proliferation. HIV RNA sensing through internal TLR7 activates pDC 47 and TLR7 ligation has been shown to induce CD4 T-cell proliferation 48 . It is possible that HIV sensing by T FH or T FR contributes to T FR expansion and further studies concerning abortive and productive infections are warranted. It has been reported that Treg increases in untreated HIV-infected individuals are diminished after treatment with antiretroviral therapy (ART), but do not normalize 49 . A recent study demonstrated that addition of MVC to ART regimens led to further reductions in Tregs in individuals who were already virologically suppressed 50 . These data, in conjunction with our findings, suggest that in HIV-treated individuals who are fully suppressed on ART regimens that do not include MVC, ongoing nonproductive HIV infection or a low level of HIV replication is occurring resulting in persistent T FR expansions that may contribute to immune impairments.
Tonsil T FR increased their regulatory capacity during X4 or R5 HIV infection, as evidenced by increased CTLA-4, LAG-3, GITR and galectin-3 expression and production of IL-10 and TGF-b. In contrast to T FR expansions, however, the increased regulatory phenotype was only partially reduced by the addition of AMD and MVC, but fully reduced by CD4 blockade. This finding is consistent with the observation that HIV binding to CD4 molecules enhances their regulatory phenotype and suppressive capabilities 31 . In addition, we found that blockade of TGF-b signalling fully inhibited T FR regulatory phenotype enhancement during HIV infection. Importantly, we found almost no expression of PD-L1 by tonsil T FR , suggesting that this regulatory molecule does not mediate their effects on T FH . Dissecting out the distinct mechanisms responsible for expansion and enhanced regulatory profiles of T FR in HIV infection could identify potential therapeutic targets to suppress excessive T FR activity and could have relevance to host-pathogen interactions in other diseases as well.
We also investigated whether intrinsic characteristics of T FR might promote their expansion during HIV infection. We found significant increases in cell proliferation of tonsil T FR compared with tonsil T FH during both X4 and R5 HIV infection. It has been well-established that regulatory DCs induce Tregs in HIV infection, and exert this effect in part by low rates of maturation and co-stimulatory molecules 39 and in part by IDO production 40,51 . The lack of DC maturation observed during HIV infection could contribute to increased Treg and T FR 52 . In tonsil cells, we found that myeloid DCs were mostly immature, whereas activated plasmacytoid DCs were significantly increased in HIV infection, which recapitulates previous studies in HIV 39,53,54 and SIV 41,42 infection in vivo. Tonsil cell cultures also exhibited increased levels of IDO, the IDO upstream activator IFN-g and TGF-b during HIV infection. Type I IFNs were not specifically studied in this work, but they have also shown to be increased in tonsil cells infected with NL4-3 HIV 55 . These factors present in the follicular milieu likely contribute to enhanced proliferation and numerical increases of T FR .
HIV-infected individuals have high rates of overall CD4 þ T-cell apoptosis 56 and higher levels of Annexin-V binding and caspase-3 activation in Tregs compared with uninfected controls 29 . We similarly found that total CD3 þ CD8 À cells and Tregs had higher levels of Annexin-V binding in chronically SIV-infected rhesus macaques. Remarkably, however, we found that T FR , GC T FR and GC T FH had lower levels of Annexin-V binding in SIV-infected rhesus macaques compared with uninfected animals. Furthermore, T FR and GC T FR had the lowest levels of Annexin-V binding of any subset in SIV-infected macaques, but not in uninfected macaques. These findings imply that regulatory follicular cells may be more resistant to apoptosis than other subsets, which could promote their accumulation. This could explain our observation that lymphoid tissue ratios of T FR to T FH and GC T FR to GC T FH in chronically SIV-infected rhesus macaques were significantly higher than those from SIV-uninfected animals, but that this phenomenon was not observed in the Treg population as a whole compared with non-Treg CD3 þ CD8 À cells. We recently reported that tonsil follicular CD4 þ T cells have elevated Bcl-2 expression and Bcl-2 is further upregulated on cells that are productively infected with R5 virus 57 . However, these studies did not discriminate between T FR and T FH populations. Thus, elevated Bcl-2 expression on T FR could be a mechanism that protects T FR from apoptosis. The mechanisms of T FR apoptosis and how they contribute to T FR number warrant further investigation.
In characterizations of T FR in mice, it was shown that T FR arise from non-T FH precursors and express both Blimp-1 and Bcl-6 (refs 14,15). To assess this in humans, we isolated both T FH and non-follicular (CXCR5 À ) subsets from tonsils and looked for alterations in their phenotype. T FH did not acquire a T FR phenotype, even when provided with exogenous TGF-b. A small population of CXCR5 À cells, however, acquired CXCR5 expression and expressed Foxp3. Despite being numerically small, this new T FR population could contribute substantially to increase the T FR population. Intriguingly, it was recently shown that T FH differentiation and maintenance was optimized by TGF-b in combination with T FH cytokines IL-12 and IL-23, but naïve CD4 þ T cells cultured with TGF-b alone acquired Treg expression profiles while promoting CXCR5 expression 58 . This further supports the notion that the expansions of T FR observed in our study likely originate from non-T FH precursors. We also find that human T FR express Bcl-6 and Blimp-1 and neither was altered after HIV spinoculation. However, expression of Bcl-6 by T FH was significantly reduced after HIV spinoculation. A more detailed analysis would be valuable to reveal the effects of HIV on both T FH and T FR transcriptional profiles.
A hallmark of HIV infection is the production of high levels of HIV-specific antibodies that are unable to suppress HIV 27 . HIV-specific neutralizing antibodies are produced in many individuals during HIV infection 59 , however, the virus is able to rapidly mutate and avoid recognition 60,61 . Although broadly neutralizing antibodies (bNabs) are effective in suppressing HIV replication 62 , few people develop bNabs 63 . Whether T FR limit the ability of HIV-infected individuals to develop bNabs is an important question. BNabs are characterized by high-affinity maturation rates in GCs, and functional T FH are crucial for somatic hypermutation levels of maturing B cells in follicles 6 . The level of bNabs in a cohort of HIV seronegative individuals who received an HIV vaccine positively correlated to functional circulating resting memory T FH 64 . We demonstrate that T FR inhibit the ability of T FH to proliferate, to produce crucial B-cell help cytokines IL-4 and IL-21, and maintain ICOS expression. Whether these factors are key to developing bNabs is of critical importance to address. Apart from direct effects on T FH , T FR may inhibit bNab production through direct interactions with B cells, or indirectly by altering the follicle microenvironment or follicular DC network during HIV infection. Thus, understanding the impact of T FR on the GC reaction and development of bNabs is complex. These questions cannot be readily addressed within tissue cultures systems, but would be best answered in an interventional study in which T FR are manipulated.
The dynamics of GC formation in acutely SIV-infected rhesus macaques were recently characterized 65 . Interestingly, animals with the most impaired GC development experienced the most rapid disease progression. It would be important to evaluate whether a robust T FR response correlates with the greatest impairment in GC formation in these animals. Interestingly, Treg-depleted mice infected with influenza had limited GC size and compromised influenza-specific T FH responses 66 . One possible explanation is that T FR are critical for GC formation and fine-tuning of immune responses during acute infections, but may exert more suppressive responses in chronic disease 67 . It will be important to determine the precise role of T FR in GC development and maintenance during different stages of HIV, as well as in other chronic infections.
T FR are emerging as important drivers of GC dynamics. Our findings that T FR are expanded during HIV infection and impair T FH function have multiple important implications. First, they provide a potential mechanism for the impaired humoral responses observed in untreated HIV-infected individuals, particularly the inability to develop bNabs. Second, they provide a potential explanation for persistent impairments in humoral immunity seen in HIV-infected individuals, despite good virologic suppression on ART that does not contain a CCR5 receptor blocker. They further suggest that prophylactic HIV vaccines should be designed to minimize induction of T FR , as they could impair development of protective antibody responses. Finally, these studies have implications for diseases and vaccine strategies apart from HIV infection, as it is likely that T FR play a key role in immune responses to many infections. Thus, a better understanding of the role of T FR in the context of HIV infection can provide important insights into the dynamics of GC responses and humoral immunity during infection and adds a new dimension to the intricacies of vaccine development.

Methods
Human and rhesus macaque subjects and clinical specimens. Inguinal LNs were obtained by excisional biopsy in an outpatient procedure as previously described 4 from individuals with documented HIV-1 infection for at least 6 months, who were not receiving ART and had CD4 þ T-cell counts of Z300 per mm 3 . None of these subjects had an opportunistic infection, malignancy or acute illness at the time of LN excision. CD4 þ T-cell counts ranged from 214 to 1,103 cells per mm 3 (median, 478 cells per mm 3 ), and plasma viral load ranged from 2.87 to 5.88 log 10 copies per ml (median, 4.26 log 10 copies per ml). A total of 10 (53%) of subjects were females, ages ranged from 21 to 50 years (median, 34 years) and 10 (53%) were white, including 3 Hispanics, and 9 were black (47%). Inguinal LNs were also obtained from HIV-1-seronegative individuals while they underwent a non-emergent surgical procedure in the groin area. A total of 6 (75%) of seronegative subjects were female, ages ranged from 32 to 71 years (median, 50 years), and 7 (88%) were white, including 1 Hispanic, and 1 (12%) was black. Informed consent was obtained from all LN donors. Human tonsils were obtained from the Colorado Children's Hospital (Aurora, CO, USA) following routine tonsillectomy from individuals at low risk for HIV infection. Use of tonsil specimens for these studies was reviewed by the Colorado Multiple Institutional Review Board and determined to not constitute human subjects research, in accordance with guidelines issued by the Office of Human Research Protections (http://www.hhs.gov/ohrp/policy/checklists/decisioncharts.html), and consequently, informed consent was not required. All research involving human subjects conformed to the principles set forth in the Declaration of Helsinki and was approved by the Colorado Multiple Institutional Review Board. LNs and the spleen were obtained from SIVmac239-infected and uninfected Indian rhesus macaques (Macaca mulatta). Animals were infected either intravenously or intrarectally with SIVmac239, and had been infected from 12 to 241 weeks (median, 19.5 weeks) at the time that specimens were obtained. Plasma SIV RNA concentrations ranged from 3.77 to 6.76 log 10 copies per ml (median, 5.72 log 10 copies per ml), and CD4 þ T cell counts ranged from 142 to 569 cells per mm 3 (median, 348 cells per mm 3 ). Of SIV-infected animals, 4 (40%) were female and they ranged in age from 7 to 17 years (median, 9 years). Of SIV uninfected animals, 2 (22%) were female and they ranged in age from 3 to 23 years (median, 4 years). Animals were housed and cared for in accordance with the American Association for Accreditation of Laboratory Animal Care standards in accredited facilities, and all animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committees of the Wisconsin National Primate Research Center. Tissues were either shipped overnight on ice in cold RPMI 1640 and disaggregated or disaggregated at the Wisconsin National Primate Research Center and frozen cells shipped on liquid nitrogen to the University of Colorado. Sample sizes in this study were based on availability of specimens and prior experience of the necessary sample size to see anticipated effects.
Quantification of Foxp3 þ CD4 þ cells in lymphoid tissues. Six-micrometre frozen sections of LN were thaw mounted onto slides and fixed in 1% paraformaldehyde (Sigma) in PBS. Two or three sections, at least 60 mm apart, were stained and evaluated for each subject. Indirect immunofluorescent staining was performed by first staining with Rabbit anti-IgD (Novus, cat# NB120-17184, 1:50) and mouse anti-Foxp3 (clone 236A/E7; eBiosciences, cat#14-4777-82, 1:100) for 1 h followed by detection with Pacific Blue anti-Rabbit IgG (cat# P10994, 1:100) and AF488 anti-mouse IgG (cat#A21200, 1:200; both from Invitrogen) for 30 min. After washing, sections were incubated for an additional hour with Rabbit anti-CD20 (Abcam, cat# ab27093, 1:400) and Rat anti-CD4 (clone YNB46.1.8; AbD Serotec, cat# MCA484G, 1:100) followed by detection with AF647 anti-rabbit IgG (Invitrogen, cat#A21443, 1:200) and AF594 anti-rat IgG (Invitrogen, cat#A21209, 1:200). Slides were covered with coverslips using SlowFade Gold (Invitrogen). Images of full sections were generated at Â 60 using an Olympus VS120 scanner outfitted with an OrcaR2 camera (Olympus). Ten random images were extracted for each section from the main scans and Foxp3 þ CD4 þ cells per mm 2 of total, follicle (CD20 þ ) and GC (CD20 þ IgD À ) was determined using visual inspection and quantitative image analysis (Qwin Pro 3.4.0, Leica Microsystems). The areas of each LN cross-section, as well as the follicular and GC compartments, were determined by quantitative image analysis and averaged for each subject. The average number of Foxp3 þ CD4 þ cells per LN cross-section and compartment was calculated by multiplying the frequency of Foxp3 þ CD4 þ cells by the average area of the LN, follicle or GC in each subject.
T FR counts and cell death time course. For the 5-day time course, cells were prepared and cultured as above with the exception of addition of 10 U ml À 1 of IL-2 to culture. Cells were harvested and analysed each day for cell death and total counts. Cell counts were performed using CountBright absolute counting beads (Molecular Probes, C36950) according to the manufacturers' instructions. Briefly, all cell fractions were resuspended in 400 ml of buffer and 50 ml of beads were added. At least 1,000 bead events were collected for each sample and the number of cells per ml was calculated based on the fixed bead concentration supplied by the manufacturer. In addition, a separate analysis of the same cell fractions was performed using the Annexin-V PE apoptosis detection kit (eBioscience, 88-8102) according to the manufacturers' instructions. Each T-cell population was analysed based on early apoptosis (Annexin-V staining), late or advanced apoptosis (Annexin-V and PI) and necrosis (PI).
ELISAs. Cell culture supernatants were measured for TGF-b-1 secretion (R&D Systems, DB100B), IL-10 (R&D Systems, DB1000B), IFNy (R&D Systems, DIF50) and IDO (Cloud Clone SEB547Hu) according to the manufacturers' instructions. Absorbance values were converted to protein levels using a protein standard dilution. Cell lysates as well as culture supernatants were evaluated by IDO ELISA.
CD25-depletion cultures. CD25 þ cells were depleted immediately after tonsil cell disaggregation using a FITC-positive selection kit (StemCell, 18558) and CD25-FITC-BC96 (BioLegend, 302604). Cells were spinoculated and cultured as described above. Mock-and HIV-spinoculated cultures from non-depleted cells were run in parallel with each experiment. After 2 days of culture, cells were analysed for T FH and T FR phenotypes by flow cytometry as described above. A subset of CD25 þ -depleted cells was analysed to confirm the absence of Foxp3 þ cells.
BrdU ex vivo proliferation assay. Ex vivo BrdU labelling was performed using a flow cytometry kit (eBioscience, 8817-6600) according to the manufacturers' instructions. For BrdU labelling of primary cells in culture, a range of incubation times for BrdU uptake were tested and 24 h was found to be the optimal timepoint. Tonsil cells were spinoculated and cultured as above and labelled with 10 mM of BrdU for 24 h before analysis. After 2 days of culture, cells were collected, stained with surface antigens, fixed and permeabilized and treated with DNase I to allow BrdU antibody binding.
Cell sorting and T FR /T FH co-culture. Disaggregated tonsil cells were sorted using a MoFlo Astrios EQ. Cells were sorted into non-follicular (CD3 þ CD8-CXCR5 À ), T FH (CD3 þ CD8-CXCR5 þ CD25 À ) and T FR (CD3 þ CD8-CXCR5 þ CD25 hi ) populations. A subset of non-follicular and T FH cells were cultured either without treatment or with 100 ng ml À 1 of exogenous TGF-b to analyse whether T FH convert to T FR . After sorting, T FH and T FR were spinoculated with X4 or R5 HIV GFP reporter viruses. T FH were then seeded at a set number of 1 Â 10 5 cells in a 24-well plate and T FR were added at ratios of 1:1, 1:10 and 1:50. After 2 days, cells were stimulated to measure IL-4 and IL-21 by ICS or left unstimulated to measure ICOS expression. In select experiments, T FH and T FR were co-cultured in the presence of neutralizing antibodies to IL-10 (1 mg ml À 1 ; BioLegend, 501406) and TGF-b (1 mg ml À 1 , BioLegend, 521704).
Statistical analysis. Comparisons of infected and seronegative human LNs or rhesus macaque spleen or LNs were performed by non-parametric Mann-Whitney tests. Comparisons of tonsil cultures were performed by unpaired Mann-Whitney or Friedman nonparametric tests. In direct comparisons of paired data, a paired Wilcoxon ranked sums test was performed to compare the two group medians of interest. Significance is denoted in each figure by asterisks, as *Po0.05, **Po0.01 and ***Po0.001.