Loss of Circulating CD4 T Cells with B Cell Helper Function during Chronic HIV Infection

The interaction between follicular T helper cells (TFH) and B cells in the lymph nodes and spleen has a major impact on the development of antigen-specific B cell responses during infection or vaccination. Recent studies described a functional equivalent of these cells among circulating CD4 T cells, referred to as peripheral TFH cells. Here, we characterize the phenotype and in vitro B cell helper activity of peripheral TFH populations, as well as the effect of HIV infection on these populations. In co-culture experiments we confirmed CXCR5+ cells from HIV-uninfected donors provide help to B cells and more specifically, we identified a CCR7highCXCR5highCCR6highPD-1high CD4 T cell population that secretes IL-21 and enhances isotype-switched immunoglobulin production. This population is significantly decreased in treatment-naïve, HIV-infected individuals and can be recovered after anti-retroviral therapy. We found impaired immunoglobulin production in co-cultures from HIV-infected individuals and found no correlation between the frequency of peripheral TFH cells and memory B cells, or with neutralization activity in untreated HIV infection in our cohort. Furthermore, we found that within the peripheral TFH population, the expression level of TFH-associated genes more closely resembles a memory, non-TFH population, as opposed to a TFH population. Overall, our data identify a heterogeneous population of circulating CD4 T cells that provides in vitro help to B cells, and challenges the origin of these cells as memory TFH cells.


Introduction
Follicular helper CD4 T cells (T FH ) are crucial for the development of antigen-specific B cells within germinal centers (GC). T FH cells interact through co-stimulatory receptors and provide essential soluble factors (i.e. IL-4, IL-21) to promote the survival, isotype switching and selection of high affinity memory B cells [1]. Phenotypic and gene signature analysis has revealed a highly conserved molecular profile of T FH cells in humans, nonhuman primates (NHP) and mice, which is characterized by increased expression of Bcl-6, CXCR5, PD-1, ICOS and decreased expression of CCR7 [2][3][4]. Human T FH cells exhibit a polarized cytokine profile characterized by compromised production of T H1 cytokines and increased secretion of IL-4, IL-10 and IL-21 [5]. Although IL-21 is characterized as a ''hallmark'' cytokine of T FH cells, other T Helper subsets produce this cytokine [6].
The origin and differentiation of T FH is unclear, as previous studies found T FH cells can derive from T H1 or T H2 cells, or independently of other CD4 lineages [7][8][9]. However, it is well established that the transcription factor Bcl-6 regulates several molecules involved in T FH development (i.e. PD-1, IL-21R, CXCR5) [10,11]. Similarly, the fate of T FH , particularly those in the germinal center (GC-T FH ), following the effector phase of the immune response is unclear. We have recently shown that NHP GC-T FH display compromised in vivo cell cycling and are prone to in vitro cell death [4]. Other studies have shown that T FH can form a memory pool found in anatomical sites outside the lymph nodes [12]. Hence, T FH cells may adopt a ''central memory'' phenotype or undergo cell death after the effector phase [13]. In humans, a circulating CD4 T cell population characterized by high CXCR5 expression can provide in vitro help for B cell isotype switching and shares functional characteristics with T FH cells [14]. It was proposed that these circulating cells, termed ''peripheral T FH '' (pT FH ) could represent the memory counterparts of T FH outside the lymphoid organs. Further investigation is needed to establish a direct relationship between T FH cells and pT FH cells.
It is becoming increasingly important to understand the interplay between CD4 T cells and B cells during HIV infection, specifically with relation to the generation of broadly neutralizing antibodies. Chronic HIV/SIV infection results in profound changes in CD4 T cell dynamics in lymph nodes characterized by T FH accumulation and increased ability of non-T FH to egress the lymph node [4,15]. How this impacts upon the dynamics of pT FH is unknown. Elucidating the biology and dynamics of pT FH , and their ability to provide B cell help may be important for our understanding of T FH memory formation during chronic infection, as well as the establishment of immune correlates reflecting the interactions between CD4 T cells and B cells within secondary lymphoid organs. This is of particular interest for monitoring clinical studies where the B cell arm of the immune system is under investigation [16].
Here we define, detect, quantify and characterize peripheral CD4 T cell populations that support B cell differentiation. We show that particular circulating CD4 T cell populations with distinct cytokine profiles have the capacity to help B cells in vitro. We further show that the frequencies of pT FH populations are significantly compromised during chronic HIV infection but can recover with antiretroviral treatment (ART), although in vitro immunoglobulin production from HIV-infected subjects both on and off ART is reduced compared to healthy subjects. Furthermore, gene expression analysis of pT FH cells and CD4 T cells in tonsil tissue suggest pT FH cells are most closely related to a non-T FH memory population within secondary lymphoid organs. Overall, our data challenge the relationship between pT FH cells and T FH memory cells.

Characterization of peripheral T follicular helper (pT FH ) cells
Previous studies defined a population of circulating CD4 T cells that express CXCR5, promote the differentiation of naïve B cells and induce immunoglobulin secretion in vitro [14,17]. We further defined CXCR5 high CD4 T cells from blood, analyzed their cytokine production and determined their ability to promote B cell differentiation in vitro. CXCR5 high CD4 T cells were found predominantly within the CD27 high CD45RO high CD4 T cell population (hereafter referred to as central memory (CM)), in agreement with previous studies [17]. The majority of the CXCR5 high CD4 T cell population also expressed CCR7 and we found the CCR7 high CXCR5 high population represented 6.5+/22.8% (mean+/2S.D.) of total CD4 T cells in healthy subjects ( Figure 1A). The majority of CXCR5 high cells expressed CD150. We further analyzed these cells based on expression of CCR6, which was previously used in combination with CXCR3 to define a pT FH subset that promotes IgG and IgA production [14], and PD-1. CCR7 high CXCR5 high CCR6 high cells represented 1.2+/20.9% of total CD4 T cells and a minority of these cells were PD-1 high .
To analyze the ability of these populations to promote B cell differentiation, naïve and CM CD4 T cells from HIV-uninfected individuals were sorted based on expression of CCR7, CXCR5, CD150, CCR6 and PD-1 ( Figure 1A), and cultured with autologous naïve B cells (CD19 + CD27 2 IgD + ) as previously described [14,18] in the presence of staphylococcal enterotoxin B (SEB). Notably, our sorted naïve B cell population did not express isotype-switched immunoglobulin ( Figure S1A) and culture conditions that lacked SEB did not induce immunoglobulin production (data not shown). Naïve and CM CCR7 low CD4 T cells failed to promote B cell differentiation and immunoglobulin production whereas CM CCR7 high CXCR5 low cells induced limited production of IgM, IgG1 and IgG3 compared to the CCR7 high CXCR5 high populations ( Figure  1B). The CCR7 high CXCR5 high CCR6 high PD-1 high population induced the greatest production of IgG1, IgG3 and IgA compared to the CXCR5 low population. Prior studies defined pT FH cells based on surface expression of CXCR5, CCR6 and the lack of CXCR3 expression [14]. We found that the greatest help for immunoglobulin production was from CXCR5 high CCR6 high cell populations and, within those, from the PD-1 high cells. We did not eliminate a small population of CXCR3 + cells in order to avoid removing a larger population of CXCR5 high CCR6 high cells that induce B cell differentiation ( Figure S1B).
The cytokine profile of pT FH populations shared characteristics with other T helper subsets, including T H1 , T H17 and T reg cells. Supernatant from the CXCR5 high CCR6 high PD-1 low coculture contained the greatest quantities of TNF-a, IL-2, and IL-17 compared to the CXCR5 high CCR6 low PD-1 high coculture ( Figure 1C). Notably, the CXCR5 high CCR6 high PD-1 high population, which promoted the greatest production of IgG1, IgG3 and IgA, showed the greatest IL-21 production, although at low levels.
Overall, CXCR5 high CD4 T cell populations induced B cell immunoglobulin production, although the CXCR5 high CCR6 high PD-1 high population did so most efficiently. However, this population is not characteristic of a T FH population found in secondary lymphoid organs, as coculture supernatants included a broad array of cytokines characteristic of T FH cells and multiple other T helper subsets.

Author Summary
Follicular T helper cells (T FH ) interact with B cells within germinal centers of lymphoid tissue to promote the survival, isotype switching and generation of high affinity memory B cells and plasma cells. Recently, a population of circulating CD4 T cells that shares phenotypic and functional characteristics with T FH cells, named peripheral T FH cells, has been identified. The relationship between peripheral T FH cells in the blood and T FH cells within the lymphoid tissue remains unclear, and whether or not peripheral T FH cells can provide insight into T cell and B cell dynamics within lymphoid tissue during infection or vaccination is not understood. Here we characterize peripheral T FH cells and show that unlike T FH cells, peripheral T FH cells secrete a diverse array of cytokines and decrease, rather than increase, during chronic HIV infection. Furthermore, we did not observe a relationship between peripheral T FH cells and memory B cells, or with the production of neutralizing antibodies to HIV. Overall, our data indicate that while peripheral T FH cells share some characteristics with T FH cells, they may not represent a good surrogate to study T cell and B cell dynamics within lymphoid tissue. , CM CCR7 high CXCR5 low (orange), CM CCR7 high CXCR5 high CCR6 low PD-1 high (green), CM CCR7 high CXCR5 high CCR6 high PD-1 low (blue) and CCR7 high CXCR5 high CCR6 high PD-1 high (red). Before gating on CCR6 and PD-1, cells were first gated on CD150 high . Right: Scatter plot indicating the frequency of each

Progressive loss of pT FH cells in HIV infection
To determine the impact of HIV on pT FH populations, we compared pT FH cells from HIV-uninfected subjects and treatment-naïve HIV-infected subjects (Table S1) as a frequency of total CD4 cells. Irrespective of how pT FH cells were defined, there was a significant decrease in the pT FH population from HIVinfected subjects compared to HIV-uninfected subjects (Figure 2A). Subjects with CD4 counts greater than 200 had significantly lower pT FH populations, while subjects with CD4 counts less than 200 had the lowest frequency of all phenotypically defined pT FH populations. However, when we defined the CCR6 high PD-1 high population as a subset of the CXCR5 high population, the frequency of the CCR6 high PD-1 high population increased in subjects with CD4 counts less than 200 ( Figure S2A). The increase in PD-1 high cells was likely due to immune activation in HIV infection, as we observed increases in the frequency of both PD-1 high and ICOS high cells within the CXCR5 high population, with the greatest increases seen in samples with CD4 counts less than 200 ( Figure S2A). We also observed a positive trend between CXCR5 high PD-1 high cells and serum concentrations of soluble CD14. ( Figure S2A). For 10 HIV-infected individuals on whom we had longitudinal samples, we observed a loss of pT FH populations as a frequency of total CD4 T cells over 36 to 48 months ( Figure 2B). However, the frequency of PD-1 high , ICOS high and CCR6 high PD-1 high cells within the CXCR5 high population remained stable ( Figure S2B).
Next, we investigated the impact of ART on the frequency of pT FH within total CD4 T cells. Longitudinal analysis on samples from before and after 24 and 48 weeks of ART revealed a recovery of pT FH populations ( Figure 2C). However, the frequency of PD-1 high , ICOS high and CCR6 high PD-1 high cells remained stable within the CXCR5 high population ( Figure S2C). Overall, HIV infection causes a loss of pT FH cells and ART promotes the recovery of these populations.

Impaired B cell help by pT FH cells in HIV infection
To investigate the impact of HIV on the ability of pT FH cells to support B cell differentiation, we performed co-culture experiments with pT FH cells from HIV-infected subjects. We focused on the CXCR5 high CCR6 high population that included both PD-1 high and PD-1 low cells due to limited cell numbers in HIV-infected subjects. Similar to previous results, the CXCR5 high CCR6 high population from HIV-uninfected subjects supported significantly more immunoglobulin production compared to the CXCR5 low population. ( Figure 3A). However, for HIV-infected subjects we observed less overall immunoglobulin production when CXCR5 high CCR6 high CD4 T cells were co-cultivated with naïve B cells. Furthermore, in viremic subjects, we observed increased IgM and IgG1 production in co-culture supernatants from the CXCR5 low population, compared to HIV-uninfected subjects. Similar to HIV-uninfected subjects, we found that pT FH cells from HIV-infected subjects produced a broad spectrum of cytokines ( Figure S3A).
Our data raise the possibility that some pT FH cells exhibit a CXCR5 low phenotype in HIV infection. This phenotype could be due the down regulation of CXCR5 on pT FH cells, or indicate the existence of a unique CXCR5 low pT FH population in chronic HIV infection. In order to distinguish these two possibilities, we investigated whether CXCL-13 impacts CXCR5 expression on CD4 T cells. We found that incubation of HIV-uninfected PBMC with CXCL-13 led to a decrease in frequency of CXCR5-positive CD4 T cells, presumably due to the internalization of CXCR5 ( Figure 3B). Furthermore, in HIV infection we found that viral load positively correlated with CXCL-13 levels and negatively correlated with the frequency of CXCR5-positive CD4 T cells ( Figure 3C). However, we did not observe a direct correlation between CXCL13 levels and the frequency of CXCR5-positive CD4 T cells. Importantly, we also found that in vitro infection of CXCR5-expressing CD4 T cells did not impact CXCR5 surface expression ( Figure S3B). Therefore, our data support the possibility that in untreated HIV infected individuals, increased levels of CXCL-13 could effect CXCR5 surface expression on pT FH cells.

Defective cytokine production of pT FH in HIV infection
T FH -dependent B cell differentiation requires IL-21. To characterize directly cytokine production from pT FH cells from HIV-uninfected and HIV-infected subjects, we performed intracellular cytokine staining after ex vivo SEB stimulation. In addition to surface markers used to define pT FH cells, we detected CD154, IFN-c, IL-2, IL-17 and IL-21 ( Figure 4A). In HIV-uninfected individuals, a minority of CD154-positive, cytokine-positive cells express a CCR7 high phenotype (10.1% of IFN-c positive cells; 28% of IL-2-positive cells; 19.4% of IL-17-positive cells and 17.9% of IL-21-positive cells), while a gradual reduction of cytokine production was found in further differentiated cells based on CXCR5 and CCR6 expression ( Figure 4B). However, for all of the cytokines detected, we observed a population of cells that were CCR7 high CXCR5 high CCR6 high , including IL-21-producing cells. Overall, we determined that a mean of 4.5% of CD154-positive IL-21-positive cells were CCR7 high CXCR5 high CCR6 high ( Figure 4B). However, this pT FH population also produced IFNc, IL-2 and IL-17 (0.8% of IFN-c positive cells; 9.0% of IL-2positive cells and 7.1% of IL-17-positive cells).
Next, we analyzed cytokine production from HIV-infected subjects off-treatment. Overall, we observed a loss of cytokineproducing cells from the CCR7 high population and a general shift towards the CXCR5 low CCR6 low population ( Figure 4A). Thus, we observed a loss of CCR7 high CXCR5 high CCR6 high pT FH cells that produce IL-2, IL-17 and IL-21 ( Figure 3B; IL-2: 9.0% for HIVnegative vs 2.0% for HIV-positive; IL-17: 7.1% for HIV-negative vs 2.2% for HIV-positive; IL-21: 4.5% for HIV-negative vs 1.1% for HIV-positive).
To analyze HIV-specific cells, PBMC were stimulated with Gag peptide pools and analyzed for cytokine expression. Very few IL-2positive and IL-17-positive cells were detected within the CM compartment (data not shown). Gag-specific IFN-c and IL-21producing cells were detected, however, compared to SEBstimulation fewer HIV-specific cells expressed CCR7 (4.4% vs 10.7% of IFN-c positive cells; 3.5% vs 11.9% of IL-21-positive cells for Gag and SEB stimulation, respectively). A majority of HIV-specific cells were not CCR7 high CXCR5 high CCR6 high ( Figure 4C; 0.4% of IFN-c positive cells and 0.9% of IL-21positive cells were CCR7 high CXCR5 high CCR6 high ). population in HIV-uninfected subjects (n = 13). Cells were not gated on CD150 for phenotypic analysis. (B) Indicated CD4 T cell populations were cultured with autologous naïve B cells (CD19 high CD27 low IgD 2 ) in the presence of SEB for 12 days and Ig concentrations were measured from supernatants (n = 6). (C) Indicated CD4 T cell populations were cultured with autologous naïve B cells in the presence of SEB for 2 days and cytokine concentrations were measured from supernatants (n = 6). Horizontal lines indicate limit of detection. Significant differences were determined using the Friedman test with Dunn's multiple comparison post-test. *p,0.05; **p,0.01. doi:10.1371/journal.ppat.1003853.g001 Overall, we observed IL-21 production from the CCR7 high CXCR5 high CCR6 high pT FH population, although we detected the most IL-21 in non-pT FH cells, which were CCR7 low and CXCR5 low . In addition to IL-21, the CCR7 high CXCR5 high CCR6 high pT FH population produced IL-2 and IL-17, cytokines characteristic of T H1 and T H17 cells, respectively. However, from HIV-infected individuals we observed a loss of CCR7 high CXCR5 high CCR6 high cells making IL-2, IL-17 and IL-21.

No relationship between pT FH cells and neutralization activity
Previous studies have described a relationship between the frequency of peripheral CXCR5 high cells and memory B cells and antibody titers with vaccination [16]. Therefore, we analyzed the relationship between the frequency of pT FH cells and IgG-positive memory B cells in PBMC from HIV-infected individuals. We found no significant correlation between the frequency of pT FH cells and IgG-positive B cells ( Figure 5A). Similarly, we failed to detect a relationship between the frequency of pT FH and HIV-1 Env-specific antibody titers or total plasma IgG levels (data not shown).
It has also been reported that PD-1 high CD4 T cells in blood are associated with cross-clade neutralizing antibody responses during HIV infection [19] and these PD-1 high CD4 T cells may represent a population of pT FH cells. Thus, the relationship between pT FH cells and neutralization activity was analyzed using HIV-infected samples classified as good neutralizers (median ID50.100) or poor neutralizers (median ID50,100) [20]. Irrespective of how pT FH cells were defined, we failed to find any relationship between neutralization activity and pT FH cells ( Figure 5B).

Relationship between pT FH cells and T FH cells in human tonsil
While pT FH cells induce B cell differentiation and immunoglobulin secretion in vitro, the relationship between pT FH and T FH cells in secondary lymphoid organs remains unclear. Our in vitro coculture studies indicated the greatest isotype-switched immunoglobulin production was elicited from B cells cocultivated with CXCR5 high CCR6 high pT FH cells ( Figure 1B). Therefore, we investigated the expression of CCR6 on T FH (CXCR5 high PD-1 high ) and non-T FH (CXCR5 low PD-1 low ) tonsil cells to determine if the CXCR5 high CCR6 high pT FH population is related to T FH cells within secondary lymphoid organs ( Figure 6A). The lowest frequency of CCR6 high cells was found within the CXCR5 high PD-1 high compartment (1.5% of CXCR5 high PD-1 high cells) and the greatest frequency of CCR6 high cells within the non-T FH compartment (9% of CXCR5 low PD-1 low cells; Figure 6B).
Similarly, RNA sequence data from the CXCR5 high CCR6 high PD-1 high pT FH population more closely resembles a memory, non-T FH CD4 T cell population from the tonsil (CM CD57 low PD-1 dim CCR7 high CCR5 low CXCR4 low ) as compared to the non-germinal center T FH population (CM CD57 low PD-1 high CCR7 low CXCR5 high ) or the GC-T FH population (CM PD-1 high CD57 high ; Figure 6C). In agreement with previous reports [5,17], tonsil T FH populations expressed higher levels of BCL6, IL-21, and CXCL13, and lower levels of PRDM1 and S1PR1 compared to the non-T FH memory population. The pT FH population from HIV-uninfected individuals expressed compara-ble levels of S1PR1 and PRDM1 to the non-T FH memory population in the tonsil ( Figure 6). We also observed lower transcript levels of MAF, BCL6, IL-21, and CXCL-13 in the pT FH population compared to tonsillar T FH populations. Importantly, MAF protein expression was highest in the CCR6 high PD-1 high pT FH population compared to other peripheral populations, although still lower than tonsillar T FH cells. (Figure 6D). For many of the selected genes, pT FH cells from HIV-infected subjects were comparable to pT FH from HIV-uninfected individuals, however, we observed greater transcript levels of activation molecules such as ICOS and CD69. Additionally, the levels of IL-21 were decreased in pT FH cells from HIV-infected individuals, supporting earlier results ( Figure 4B). Collectively, our data suggest the pT FH population characterized as CXCR5 high CCR6 high most closely resembles a non-T FH memory population in the tonsil.

Discussion
The development and nature of human T FH memory cells following an effector immune response are not known. The ability to define a population of memory T FH cells in PBMC (pT FH ) would help inform our understanding of CD4 T cell dynamics within lymphoid tissue during vaccination or infection. Studies of chronic infection may be helpful in this regard [21]. Whether the accumulation of T FH cells during chronic infection [4,15] impacts the T FH memory population is of particular interest, especially if memory T FH cells migrate between lymphoid organs and peripheral tissues. Recent studies [14,16] have suggested that circulating CXCR5 high CD4 T cells may represent the peripheral counterparts of T FH cells. However, the relationship between pT FH and T FH cells within secondary lymphoid organs remains unclear. Therefore, it is of great relevance to determine if pT FH cells originate from GC-T FH cells and represent a memory T FH population, reflect a precursor population that differentiates into GC-T FH upon re-exposure to antigen, or both. Our studies begin to address these issues by further defining pT FH cells, comparing pT FH cells to tonsillar T FH cells, and analyzing the effect of HIV on these cells.
In concordance with previous studies, we showed that circulating CXCR5 high CD4 T cells support B cell differentiation in vitro [14,17]. A majority of the CXCR5 high cells expressed CD150, and while CD150 was used for gating in the co-culture assays, we found it did not impact the loss of pT FH cells or effect our results with respect to loss of pT FH cells, recovery with ART or lack of association with B cell or antibody responses (data not shown). However, within the CXCR5 high population the expression of CCR6 and PD-1 did further define pT FH populations with differential abilities for naïve B cell help and isotype switching. Thus, pT FH cell populations support both the activation and maturation of naïve B cells, and immunoglobulin isotype switching. Correspondingly, the individual pT FH populations produced cytokines associated with B cell maturation and survival, such as IL-21 [22], IL-2 [23] and IL-17 [24], in contrast to T FH cells within secondary lymphoid tissue, which display a limited cytokine profile that includes IL-4, IL-10 and IL-21, but compromised production of IL-2 and IL-17 [4]. Whether these pT FH populations represent different stages of T FH memory development or originate from separate CD4 T cell populations within lymphoid tissue [25] is still unclear. In order to better understand the relationship between T FH and pT FH cells, we compared gene expression levels between pT FH and tonsillar CD4 T cell populations and focused on genes important for T FH differentiation, migration, and function. We found that the pT FH population with the greatest B cell helper function most closely resembled a CM, non-T FH CD4 T cell subset within the tonsil. While our studies do not directly address the relationship between GC-T FH in lymph nodes and circulating CD4 T cells from the same patients, our data challenge whether pT FH are memory T FH cells. A recent study reported that germinal center T FH cells in mice migrate throughout the follicle, but generally do not leave the follicle to enter the blood [26]. While it is conceivable that pT FH cells represent a very minor population of T FH cells that exit the follicle, it is also possible that pT FH cells are reflective of a precursor T FH population that exits the lymphoid organ and enters the circulation before entering the follicle. However, while we find the CXCR5 high CCR6 high PD-1 high pT FH population does not resemble a memory T FH population, Locci and colleagues found a CXCR5+CXCR3-PD-1+ pT FH subset that functionally and transcriptionally resembles a memory T FH population [27]. A recent study in mice reported that memory T FH cells have reduced mRNA expression of T FH markers such as Bcl6, IL-21, ICOS and PD-1 compared to the effector T FH population [28], indicating the expression of these molecules may change depending on the phase of infection. Therefore, further investigation of pT FH subsets and their relationship to memory and effector populations at multiple stages of infection is needed.
pT FH and naïve B cell co-cultures from HIV-infected subjects produced fewer immunoglobulins compared to co-cultures from HIV-uninfected subjects. The observed defect in immunoglobulin production is likely due to impaired pT FH help to B cells instead of B cell dysfunction, as co-cultures included naïve B cells rather than memory B cells that exhibit abnormalities in HIV infection [29]. Furthermore, while co-culture supernatants from HIV-infected subjects demonstrated a heterogeneous cytokine profile, similar to HIV-uninfected subjects, intracellular cytokine staining showed that fewer CCR7 high CXCR5 high CCR6 high pT FH cells produced IL-2, IL-17 and IL-21 in chronic HIV infection compared to HIVuninfected individuals. Furthermore, gene expression analysis of HIV-infected pT FH revealed fewer IL-21 and IL-4 transcripts, although the overall levels of cytokine transcripts were low.
Recent studies have shown T FH cells within secondary lymphoid organs accumulate in some donors or animals during chronic HIV/SIV infection and that T FH accumulation is associated with GC B cell expansion and increased serum immunoglobulin concentrations [4,22,30]. In contrast to T FH cells, our studies revealed pT FH cells consistently decrease in chronic HIV infection, with disease progression resulting in a greater reduction of these compartments within the total CD4 T cell population. However, it should be noted that we were unable to analyze T FH cells within secondary lymphoid organs from these subjects and therefore we are unable to directly compare the frequency of pT FH cells and T FH cells from the same individual. The differences between the increase in T FH cells and decrease in pT FH cells may be due to differences in disease state (i.e. early vs late infection) or represent a steady state of T FH cells trafficking between the lymphoid tissue and the blood. The decreased frequency of pT FH in the blood may indicate impaired ability of T FH to exit the lymph node in chronic HIV infection where the tissue architecture is not intact. Alternatively, the decreased frequency of pT FH in the blood may be a result of pT FH trafficking to secondary lymphoid organs. In agreement with previous studies [14,17], we found a majority of CXCR5 high cells express CCR7, and it has previously been suggested that pT FH cells migrate to secondary lymphoid organs upon infection due their expression of CCR7 and CD62L [14].
A confounding factor with regard to how we interpret the decrease in pT FH cells is that we also found a reduction in the surface expression of CXCR5 on CD4 T cells in chronic HIV infection, which may result from increased sera levels of CXCL-13 [31,32]. Furthermore, our co-culture data indicate that CXCR5 low CD4 T cells from viremic subjects can induce some B cell differentiation. These data support the possibility that in chronic HIV infection, a subset of functional pT FH cells may be phenotypically defined as CXCR5 low . Additionally, it should be noted that analysis of cellular subsets within the CXCR5 high population in chronic HIV infection revealed the frequency of CCR6 high PD-1 high cells increased. These results are consistent with a state of generalized immune activation, as we also observed increased surface expression of ICOS on CXCR5 high and CXCR5 high PD-1 high cells, and a positive association between the frequency of PD-1 high cells within the CXCR5 high population and serum concentrations of soluble CD14 [33]. Similarly, gene expression analysis indicated increased transcript levels of activation markers, such as ICOS and CD69 within the pT FH population during HIV infection. Overall, these data emphasize the difficulty in defining pT FH cells in chronic HIV infection and understanding the relationship between pT FH cells and T FH cells.
The uncertain definition of pT FH cells in HIV infection may provide an explanation as to why we were unable to identify correlations between pT FH populations and circulating IgGpositive memory B cells, or between pT FH cells and HIV-specific IgG (data not shown). Furthermore, we found no correlation between the frequency of pT FH and the neutralization activity of a well-characterized cohort of HIV-infected donors [20]. However, the absence of a correlation between pT FH cells and circulating HIV Env-specific IgG may also be explained by the lack of a timedependent association (early vs. late infection) between T FH and pT FH cells, or indicate that the generation of IgG and broadly neutralizing antibodies is regulated by parameters other than pT FH , confounded by T-cell independent antibody production commonly observed in HIV infection [34] or generalized immune activation. Thus, our data challenge the application of the pT FH population as a surrogate of GC T FH -B cell interactions in chronic HIV infection. While our studies did not find a correlation between pT FH cells and neutralizing antibodies, several recent studies, each with a different definition of pT FH cells, have reported an association with antibody responses during vaccination, infection or autoimmune disease [27,[35][36][37]. Therefore, further studies are needed to establish the association between pT FH subsets and the generation of neutralizing antibodies, especially in HIV infection.
Overall, our data indicate that a range of circulating CD4 T cell populations can provide B cell help, possibly through differential secretion of soluble factors and/or cell-cell contact interactions  [17,35] and that HIV infection results in loss of these cells over time, but with relative increases within the CXCR5 high compartment which may be explained by immune activation. Furthermore, we did not find any association between pT FH and measures of B cell function such as HIV neutralization breadth/potency, HIV-specific IgG, or total IgG, suggesting application of this population as a surrogate of GC T FH -B cell interactions during HIV infection may be limited. A better understanding of the differentiation process and the developmental relationship between pT FH subsets and lymph node T FH cells is critical for the establishment of reliable peripheral blood CD4 T cell correlates for monitoring infection-or vaccine-associated B cell responses.

Ethics statement
Signed informed consent was obtained in accordance with the Declaration of Helsinki and approved by the appropriate Institutional Review Board. Tonsil cells were acquired from anonymized discarded pathologic specimens from Children's National Medical Center (CNMC) under the auspices of the Basic Science Core of the District of Columbia Developmental Center for AIDS Research. The CNMC Institutional Review Board determined that study of anonymized discarded tonsils did not constitute 'human subjects research.'

Subjects
Fresh HIV-uninfected peripheral blood mononuclear cells (PBMC) were obtained from individuals participating in the NIH research apheresis program. Fresh HIV-infected blood was obtained from the Vaccine Research Center Clinic or Drexel University College of Medicine. Frozen HIV-infected PBMC were obtained from three study populations (Table S1). For untreated HIV infection, cells were obtained from volunteers who participated in a therapeutic vaccination trial (no efficacy was observed) conducted in the 1990's prior to the advent of combination antiretroviral therapy (cART) [38]. The second study population consisted of donors from a cohort used to identify individuals with HIV broadly neutralizing antibodies [20]. To study the effect of cART, we obtained PBMC from HIV-infected donors participating in AIDS Clinical Trials Group study A5142 prior to initiation of cART and 24 and 48 weeks post-therapy [39,40]. PBMC and Intracellular cytokine staining: 3610 6 PBMC were incubated in 1 mL of medium containing brefeldin A (10 ug/mL) in the absence or presence of HIV-1 Gag-peptide pools (15mers overlapping by 11 residues; NIH AIDS Research and Reference Reagent Program) or 1 ug/mL SEB (Sigma) for 6 hours. Cells were surface stained, permeabilized (Cytofix/Cytoperm kit; BD Biosciences), and stained with anti-CD3, anti-IFN-c, anti-IL-2, anti-IL-17a, anti-IL-21 and anti-CD154. Events were collected on a modified LSRII flow cytometer (BD Immunocytometry Systems) and electronic compensation was performed with antibody capture beads (BD Biosciences). Data were analyzed using FlowJo Version 9.6 (TreeStar).

T and B cell culture
Co-culture experiments were performed with freshly isolated PBMC. 5610 4 CD4 T cell populations were sorted based on expression of CCR7, CXCR5, CD150, CCR6 and PD-1 and cultured with 5610 4 autologous naïve B cells (1:1 ratio) in the presence of SEB (0.5 mg/ml). Supernatants harvested on Day 2 were analyzed for cytokines using Luminex technology (Milliplex MAP Kit, HTH17MAG-14K, Millipore). The lower limit of detection (LOD) was set at the lowest concentration on the standard curve and values below the LOD were counted as zero. Supernatants collected on Day 12 were analyzed for immunoglobulins (Milliplex MAP Kit, HGAMMAG-301K). Some supernatants exceeded the saturation limit of the standard curves for IgM and IgG3. These values were included in the analysis and quantified as being equivalent to the highest determined concentration.

ELISA
Soluble CD14 and CXCL-13 (R&D Systems) were measured in plasma or sera from HIV-infection patients according to the manufacturer's instructions.

CXCL-13 treatment
Freshly isolated PBMCs were incubated with recombinant human CXCL-13 (R&D Systems) at 37uC or 4uC and analyzed for CXCR5 surface expression by FACS.

Illumina deep sequencing of messenger RNA
CD4 T cell populations were sorted from uninfected PBMC (n = 5), HIV-infected PBMC (n = 5) and uninfected human tonsils (n = 4) based on expression of CCR7, CXCR5, CD150, CCR6 and PD-1 for PBMC and CD57, PD-1, CCR7, CXCR5, CCR5 and CXCR4 for tonsils. Total RNA was purified from sorted cell populations and treated with DNAse I (Ambion) to minimize genomic DNA contamination. Polyadenylated RNA was isolated using Oligo-dT Dynabeads (Life Technologies), chemically fragmented, and used to construct barcoded Illumina Truseq libraries. Libraries were size-selected, quantified, pooled, sizeselected and quantified again, and clustered on an Illumina Truseq Paired-End Flowcell v3. The flowcell was sequenced on an Illumina HiSeq 2000 in a 2675-base paired-end, indexed run. Adaptor sequence was trimmed from the raw sequencing reads using Trimmomatic. The trimmed sequencing reads were subsequently aligned to the human genome (hg19) using TopHat 2. Differential expression testing was done using Cufflinks 2 and visualization of differential expression was done using the R package cummerbund. Accession numbers of the selected genes are shown in Supporting Table S2.

Virus neutralization
Neutralization activity of patient sera was determined against 20 viral isolates using a TZM-bl neutralization assay as previously described [20].

In vitro infection
Freshly isolated PBMCs were stimulated with PHA (10 mg/ml). After 12 hours stimulation, CXCR5 high cells were sorted by FACS Aria based on surface molecule expression and infected by a multiplicity of infection (MOI) of 0.01 with either HIV NL-E or HIV NLAD8-E [41]. The infected cells were cultured in the presence of 50 U/ml recombinant human interleukin-2 (R&D) for 5 days and analyzed for CXCR5 expression by FACS.

Statistics
Experimental variables were analyzed using the nonparametric Mann-Whitney U test, the Wilcoxon matched-pairs signed rank test or the Friedman test with Dunn's multiple comparison posttest. Correlation analysis was performed using the nonparametric Spearman test. Error bars depict mean+SEM in all bar graphs shown. The GraphPad Prism statistical analysis program (Graph-Pad Software, version 5.0) was used throughout. showing the frequency (%) of indicated populations in CXCR5expressing cells from PBMC from HIV-uninfected subjects (open circles; n = 13) and HIV-infected subjects before (n = 14, week 0; black circles) and after ART (week 24, dark gray circles; week 48, light gray circles). (EPS) Figure S3 Characterization of pT FH cells in HIV infection. (A) CCR7 high CXCR5 low and CCR7 high CXCR5 high CCR6 high CD4 T cells isolated from PBMCs were cultured with autologous naïve B cells (CD19 high CD27 low IgD 2 ) in the presence of SEB for 2 days and cytokine concentrations were measured from supernatants (HIV-uninfected, n = 5; HIV-infected (non-viremic), n = 4, HIVinfected (viremic), n = 0-1). Due to limited cell numbers we were unable to collect CCR7 high CXCR5 high CCR6 high cells from viremic individuals. (B) Sorted CXCR5 high central memory cells isolated from blood do not down-regulate surface expression of CXCR5 upon X4 or R5 in vitro infection. (EPS) Table S1 CD4 count, viral load and neutralization activity of subjects studied.