|Home | About | Journals | Submit | Contact Us | Français|
HIV-1 replication is concentrated within CD4+ T cells in B-cell follicles of secondary lymphoid tissues during asymptomatic disease. Limited data suggest that a subset of T follicular helper cells (TFH) within germinal centers (GC) is highly permissive to HIV-1. Whether GC TFH are the major HIV-1 virus-producing cells in vivo has not been established. Here, we investigated TFH permissivity to HIV-1 ex vivo by spinoculating and culturing tonsil cells with HIV-1 GFP reporter viruses. Using flow cytometry, higher percentages of GC TFH (CXCR5highPD-1high) and CXCR5+PD-1low cells were GFP+ than non-GC TFH (CXCR5+PD-1intermediate) or extrafollicular (CXCR5-) cells. When sorted prior to spinoculation, however, GC TFH were substantially more permissive than CXCR5+PD-1low or extrafollicular cells, suggesting that many GC TFH transition to a CXCR5+PD-1low phenotype during productive infection. In situ hybridization on inguinal lymph node sections from untreated HIV-1-infected individuals without AIDS revealed higher frequencies of HIV-1 RNA+ cells in GC than non-GC regions of follicle or extrafollicular regions. Superinfection of HIV-1-infected individuals’ lymph node cells with GFP reporter virus confirmed the permissivity of follicular cells ex vivo. Lymph node immunostaining revealed 96% of CXCR5+CD4+ cells were located in follicles. Within sorted lymph node cells from four HIV-infected individuals, CXCR5+ subsets harbored 11- to 66-fold more HIV-1 RNA than CXCR5-subsets, as determined by RT PCR. Thus, GC TFH are highly permissive to HIV-1, but downregulate PD-1 and to a lesser extent CXCR5 during HIV-1 replication. These data further implicate GC TFH as the major HIV-1-producing cells in chronic asymptomatic HIV-1 infection.
HIV-1 replication is concentrated within CD4+ T cells located in B-cell follicles of secondary lymphoid tissues during chronic asymptomatic disease prior to onset of AIDS (1–6). CD4+ T follicular helper cells (TFH) are the major CD4+ T cell subset located within B cell follicles and presumed to be the major source of virus replication. A subset of TFH migrate into the light zone of the germinal center (GC), interact with follicular dendritic cells (FDC) and B cells, and promote B cell affinity maturation. GC TFH are characterized by high levels of expression of both PD-1 and the follicular homing molecule CXCR5, and have a unique cytokine profile including production of IL-21 and IL-4. HIV-1 infection promotes the expansion of GC TFH, contributing to the GC hyperplasia and lymphadenopathy typical of chronic, asymptomatic HIV-1 infection (7). Impairments in TFH function have been reported in HIV-1 infection (8) and linked to diminished broadly-neutralizing antibodies in SHIV-infected rhesus macaques (9). Whether GC TFH are the major virus-producing cells in vivo has not been clearly established. In situ hybridization for HIV-1 and SIV RNA has localized virus-producing cells to B cell follicles (1–3), but whether they are primarily located in GC has not been determined. Furthermore, existing data based on measurement of HIV-1 and SIV RNA and DNA in lymphoid tissue cells sorted on the basis of GC TFH phenotypic markers are sparse and inconsistent (10–12).
Increasing evidence implicates B cell follicles as immune privileged sites due to the failure of virus-specific CTL to accumulate within follicles in large numbers (1, 2, 10). Nevertheless, it is possible that other factors may contribute to heightened virus replication at those sites. Limited data suggest that GC TFH may be more permissive than other cells to HIV-1. Thacker et. al. demonstrated that tonsillar T cells that expressed CD57, a marker for some GC TFH, produced four- to six-fold more p24 antigen than other tonsil cells (13). Previously, this group had demonstrated that HIV-1 virions bound to FDC are potently infectious to human CD4+ T cells (14). As GC TFH are in close proximity to FDC, this further supports the notion that GC TFH may be specifically vulnerable to HIV-1 infection.
To address these questions, we first investigated the permissiveness of GC TFH to HIV-1 in a series of experiments using tonsil cells from individuals at low risk for HIV-1 that were infected ex vivo with HIV-1 GFP reporter viruses. These studies suggested that GC TFH were highly permissive to HIV-1 ex vivo, but that some downregulate PD-1 and to a lesser extent CXCR5 in the context of productive infection. We next evaluated the location of HIV-1 RNA+ cells in vivo by performing in situ hybridization for HIV-1 RNA on lymph node tissue sections from chronically infected, asymptomatic HIV-1-infected individuals who were not receiving antiretroviral therapy. These studies revealed significantly higher concentrations of HIV-1 RNA+ cells in GC than in non-GC regions of follicle or extrafollicular regions. Finally, we measured the concentration of HIV-1 RNA in sorted lymph node cells from HIV-1-infected individuals and found that follicular (CXCR5+) subsets harbored 11- to 66-fold more HIV-1 RNA than extrafollicular (CXCR5-) subsets of CD3+CD8- cells. These data demonstrate that GC TFH are highly permissive to HIV-1, but downregulate PD-1 and to a lesser extent CXCR5 during HIV-1 replication. They further implicate GC TFH as the major HIV-1-producing cells in chronic asymptomatic HIV-1 infection.
Tonsils were obtained from children at low risk for HIV infection who had undergone routine tonsillectomy. Tonsil cells were isolated by mincing tonsil tissues in phosphate buffered saline (PBS, Mediatech, Manassas, VA) and filtering the cell through a 70 μm mesh filter. 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 (15), and consequently, informed consent was not required.
Inguinal lymph nodes were obtained as previously described (16, 17) from individuals who had documented HIV-1 infection for at least 6 months, were not receiving antiretroviral therapy, and had CD4+ T-cells ≥ 300/mm3. None of these subjects had an opportunistic infection, malignancy or an acute illness at the time of lymph node excision. Peripheral blood was obtained at the same time as the lymph node specimens. Informed consent was obtained from all subjects and the study was approved by the Colorado Multiple Institutional Review Board. One half of each inguinal lymph node was snap frozen in OCT, and the remainder was disaggregated as previously described (16) and cells were cryopreserved and stored in liquid nitrogen. Peripheral blood CD4+ T cell counts were determined by flow cytometry and plasma HIV-1 RNA concentration was measured by Roche COBAS Taqman 96 HIV-1 test (Indianapolis, IN).
The HIV-1 NL4-3-based CXCR4-tropic (X4) GFP reporter virus NLENG1-IRES and CCR5-tropic (R5) GFP reporter virus NLYUV3-GFP have been described elsewhere (18, 19). Virus stocks were generated by transfection of 293T cells using Effectene (Qiagen, Valencia, CA). Tonsil cells (5 × 106) were infected using spinoculation by resuspending them in 0.25 mL of either R5 GFP reporter virus stock or X4 GFP reporter virus stock or media only (mock infected) and centrifuging at 1200xg for 2 hours at room temperature. In a subset of experiments, to evaluate the effects of spinoculation on HIV-1 co-receptor expression, tonsil cells were spinoculated without addition of reporter virus, and compared to cells that were rested for the same period of time at room temperature. After spinoculation, cells were washed with PBS, diluted to 1 × 106 cells/ml in RMPI 1640 (Mediatech) with 10% fetal bovine serum (FBS, Atlanta Biologicals, Flowery Branch, GA) and 1% penicillin/streptavidin/L-glutamine (Sigma, St. Louis, MO) (R10 media) and incubated at 37°C in 5% CO2 for 48 to 60 hours.
All antibodies and their respective dyes and clones used are listed in Supplemental Table 1. Briefly, cells were incubated with 1% normal goat serum to block nonspecific staining followed by staining for 30 min. with labeled antibodies. Viability dye (GHOST 510, Tonbo Biosciences, San Diego, CA) was included in all phenotyping panels. Cells were washed in PBS containing 1% bovine serum albumin (Sigma) (wash buffer) and fixed in 2% paraformaldehyde (Sigma). Data were acquired using an LSRII flow cytometer (BD Immunocytometry Systems, San Jose, CA) and analyzed using FlowJo (Tree Star, Ashland, OR). Fluorescent minus one (FMO) was used to define negative populations where clear population separations were not obvious and mock infected cells determined the negative population for GFP+ cells.
To determine TFH subsets, after blocking with normal goat serum, cells were treated with biotinylated antibodies to PD-1 for 30 min at room temperature and washed in wash buffer before staining with streptavidin and remaining antibodies. Cells were analyzed by first gating for live cells and then for single lymphocytes using forward and side scatter. Cells were then gated on CD3+ followed by CD8- and were further divided into CXCR5- (EF), CXCR5+ (F), CXCR5+PD-1high (GC TFH), CXCR5+PD-1intermediate (non-GC TFH) and CXCR5+PD-1low as shown in Figure 1A and and2A.2A. Chemokine receptor expression was determined using PE CCR5 or PE CXCR4 (manufactured by BD Biosciences with known 1:1 PE:antibody ratio), and QuantiBRITE beads (BD Biosciences) were used to determine the mean number of CCR5 or CXCR4 molecules on the surfaces of lymphocyte subsets, as we have previously described (19). Memory subsets were defined as CD95+, central memory (CM) cells as CD95+CD28+CCR7+, and effector memory cells (EM) as CD95+CD28-CCR7-. Activated cells were defined as HLA-DR+CD38+ (DR+38+).
In some instances, tonsil cell subsets were isolated prior to spinoculation with GFP reporter viruses. To accomplish this, CD4+ T cells were first selected using Dynabeads Untouched CD4 T Cell Kit (Invitrogen, Grand Island, NY). The isolated cells were then stained with antibodies to PD-1 and CXCR5 as described above but not fixed and sorted on a FACSAria (BD Immunocytometry Systems, San Jose, CA) into the following subsets: CXCR5−, CXCR5+PD-1low, and CXCR5highPD-1high. Depending upon cell yields after the sorts, between 1.5 and 3 × 106 cells were spinoculated with reporter viruses and then cultured for 2.5 days, as described above.
hybridization combined with immunohistochemical staining was performed on a minimum of three 6 μm thick frozen tissue cross sections each between 20 and 60 μm apart for each seropositive subject using combined riboprobes to HIV env, gag, and nef, as previously described (16). As a negative control, sections from a subset of subjects were hybridized with digoxigenin-labeled sense probes for env, gag, and nef using the same procedure. Tissue sections were treated with Antigen Retrieval Buffer-High pH (DAKO, Carpinteria, CA), as described by the manufacturer, then with 1% hydrogen peroxide to block endogenous peroxide activity for 20 min., and further blocked with 2.5% horse serum (Vector Laboratories, Burlingame, CA) for 20 min. Next, sections were incubated with rabbit antihuman IgD (Novus Biologicals, Littleton, CO) diluted in Tris-buffered saline (TBS) (1:100, 0.05M Tris hydrochloride, 0.15M NaCl, pH 7.6) with 1% BSA and 0.1% sodium azide for 1 hr. Next sections were incubated with ImmPRESS peroxidase labeled anti-rabbit Ig (Vector Laboratories) for 30 min. and color was developed using NovaRed (Vector Laboratories) as a substrate. For CD20 staining, the process was repeated using mouse antihuman CD20 (AbD Serotec, Raleigh, NC), and ImmPRESS Alkaline phosphatase labeled anti-mouse Ig (Vector Laboratories), and VectorRed alkaline phosphatase (Vector Laboratories) as a substrate. The number of HIV-1 RNA+ cells was determined by visual inspection and manually counting the RNA+ cells on each section and categorized by location as GC (determined morphologically by CD20+ staining and the absence of IgD staining), non-GC regions of follicle (CD20+ and IgD+), and EF (defined by the absence of clusters of CD20+ cells and IgD staining). The sum of the total number of virus producing cells was divided by the total area of the tissue sections inspected, and quantified by computerized image analysis (Leica Q5001W Image Analysis, Leica, Cambridge, UK) to determine the number of HIV-1 RNA+ cells per mm2 tissue and compartment for each subject.
Fresh cut 6 μm thick frozen tissue sections were fixed in 1% paraformaldehyde for 20 min., washed in TBS and blocked 20 min. with 1% hydrogen peroxide, and then washed in TBS followed by blocking with 2.5% normal horse serum (Vector labs). Sections were stained with mouse anti p24 (DAKO, Carpintera, CA) for 1 hour, washed and protein detected using ImmPRESS peroxidase kit for mouse (Vector Labs) and Nova Red Substrate (Vector Labs) as described above. Sections were blocked a second time with 2.5% horse serum and stained with mouse anti human CD20 (DAKO) and detected using ImmPRESS peroxidase kit for mouse and Vector SG (Vector Labs) as the substrate. Slides were briefly counterstained with 50% hematoxylin and coverslipped. The percent of p24 staining in the follicle was determined by quantitative image analysis using Qwin Pro (Leica v 3.4.0) on a Leica DMR brightfield microscope by determing the area of the follicle positive for p24 and dividing by the total area of the follicle.
Frozen lymph node sections 4 μm thick were fixed in 1% paraformaldehyde (Sigma, St. Louis, MO), stained for 1 hour with rat anti-CXCR5 (RF8B2, BD Bioscience), mouse anti-CD4 (RPA-T4, BD Bioscience) and rabbit anti-CD20 (Abcam, Cambridge, MA), then treated for 30 min with secondary antibodies AF488 donkey anti-rat, AF647 chicken anti-rabbit and AF594 goat anti-mouse (Invitrogen, Grand Island, NY). Stained slides were viewed on a Leica DM5000B fluorescent microscope and 10 to 15 randomly selected areas were imaged using Qwin Leica FW4000 software (Leica). Follicular and extrafollicular tissue areas within the images were defined morphologically by CD20 staining, and percentages of CXCR5+, CD4+, and CXCR5+CD4+ tissue area determined by quantitative image analysis (QWin Pro; v.3.4.0, Leica).
Cryopreserved, disaggregated lymph node cells from 3 HIV-1-infected individuals were thawed, placed in R10 media, and incubated overnight at 37°C in 5% CO2. Cells were spinoculated with X4-tropic HIV-1 GFP reporter virus, as described above for tonsil cells, and incubated for 3 days at 37°C in 5% CO2, followed by staining for flow cytometry as described above.
Frequencies of CD4+ cells within lymphoid tissue were estimated as previously described (2). Briefly, three full cross-sections of lymph node were double stained for CD4 and CD20. Ten randomly selected fields, each 0.74 mm2, were evaluated on each tissue section. Using quantitative image analysis, the area of the follicular and extrafollicular regions was determined as well as the area of each of these regions that stained positively for CD4. The percentage of the tissue area that stained positively for CD4 was divided by the average area of a CD4+ cell in the tissue to estimate the number of CD4+ cells per mm2. Frequencies of CD4+ cells that were producing HIV-1 RNA were calculated by dividing the frequency of HIV-1 RNA-producing cells per mm2 by the estimated number of CD4+ cells per mm2.
Thirty million cryopreserved disaggregated inguinal lymph node cells were thawed, and stained as described above but left unfixed. Stained cells were sorted into CD3+CD8-CXCR5+ and CD3+CD8-CXCR5- populations on a FACSAria cell sorter (BD). A minimum of 200,000 cells was collected for each population. Sorted cells were stored at −70°C in dry pellets of 100,000 cells. Cellular RNA was extracted from dry pellets using RNA Blood Extraction Kit (Qiagen, Gaithersburg, MD). HIV-1 RNA was quantified by a nested PCR amplification of the HIV-1 long terminal repeat (LTR) as previously described (19). HIV-1 DNA concentration in the RNA preparations was determined by PCR amplification without reverse transcription. The HIV-1 DNA concentration was subtracted from the measured HIV-1 RNA concentration to obtain a corrected HIV-1 RNA concentration.
Non-parametric statistical tests were used due to small sample sizes. Wilcoxon-signed rank two tailed t-tests were used to evaluate paired observations, and the Mann-Whitney U test for unpaired observations. For comparisons of 3 or more groups one-way ANOVA with multiple comparisons was used. For determining correlations, Spearman’s correlation was used. A p value <0.05 was considered statistically significant. Data were analyzed using Graphpad Prism (La Jolla, CA).
GFP expression was evaluated in tonsil cells two days after spincoculation with R5- or X4-tropic GFP reporter viruses or mock-spinoculation. Cells were gated on the CD3+CD8- population as CD4 is downregulated on HIV-1 producing cells both in vivo (19) and during ex vivo infection with these reporter viruses [(19, 20) Fig. 1A]. The CXCR5- population was defined as the extrafollicular (EF) population of cells, whereas the CXCR5+ population was defined as follicular (F) cells, as shown in representative flow cytometry plots in Fig. 1A. F cells were significantly more likely to express GFP and a higher geometric mean fluorescence intensity (gMFI) of GFP than EF cells in both R5- and X4-spinoculated tonsil cells (Fig. 1B). Because memory cells are known to be more permissive to HIV-1 than naïve cells (21), we further examined the influence of memory cell phenotype (CD95+) on GFP expression (Fig. 1B). Memory cells were more likely to be productively infected and to express a higher gMFI of GFP compared to naïve (CD95-) cells. Effector memory cells were slightly more permissive than central memory cells in X4-, but not R5-spinoculated cultures, and expressed a significantly higher gMFI of GFP than central memory cells in both cultures. Even after controlling for memory phenotype, memory F cells were substantially more likely to express GFP and have a higher GFP gMFI than memory EF populations. Thus, memory phenotype did not fully account for the elevated GFP expression seen in F cells.
To further investigate the permissiveness of distinct follicular subsets of cells, GFP expression was determined in additional tonsil cell cultures. Three distinct subpopulations of follicular cells were identified: a CXCR5highPD-1high population, which was defined as GC TFH, a CXCR5+PD-1intermediate population, which was defined as non-GC TFH, and a CXCR5+PD-1low population, as previously described by others (22) and shown in representative flow cytometry plots in Fig. 2A. In both R5- and X4-spinoculated cultures, GC TFH harbored significantly higher percentages of GFP+ cells than either non-GC TFH or EF cells (Fig. 2B). Percentages of productively infected cells, however, did not differ between GC TFH and CXCR5+PD-1low populations. gMFI of GFP, previously shown to correlate with the magnitude of HIV-1 production (18, 23), was significantly lower in the GC TFH population compared to all other populations and was highest in the CXCR5+PD-1low population. Interestingly, the CXCR5+PD-1low population largely lacked GFPlow cells that were apparent in the other subsets, as shown in the representative histograms in Fig. 2A.
As the CXCR5+PD-1low and GC TFH populations harbored the largest percentages of GFP+ cells, we next examined whether there were changes in the overall fraction of these cell subsets in the context of infection, which might be expected due to increased infection and cell death. Compared to mock-spinoculated cultures, percentages of GC TFH were lower whereas percentages of CXCR5+PD-1low cells were higher in both R5- and X4-spinoculated cultures (Fig. 2C). There were no significant differences in percentages of non-GC TFH, and percentages of EF were significantly higher in X4-spinoculated cultures only (Supplemental Fig. 1). The unexpected increase in percentages of CXCR5+PD-1low cells suggested that both PD-1, and to a lesser extent CXCR5, may be downregulated in the context of HIV-1 replication. Indeed, compared to mock-spinoculated cells, gMFI of PD-1 was significantly lower on CXCR5+ cells in both R5- and X4-spinoculated cultures, and CXCR5 was lower on all CD3+CD8- cells in X4-spinoculated cultures (Fig. 2D).
To further evaluate the permissiveness of GC TFH and CXCR5+PD-1low cells, tonsil cells from 6 donors were sorted into subsets of GC TFH, CXCR5+PD-1low, and EF populations, spinoculated with R5- or X4-tropic HIV-1 GFP reporter virus and analyzed for GFP expression by flow cytometry after 2.5 days (Fig. 2E). GC TFH were the most permissive cell subset with percentages of productively infected cells ranging from 2.3% to 4.5% in R5-spinoculated cultures and 6.3% to 22% in X4-spinoculated cultures. In contrast, CXCR5+PD-1low cells were substantially less permissive with percentages of GFP+ cells ranging from 0.07% to 0.2% in R5-spinoculated cultures and 0.2% to 2.7% in X4-spinoculated cultures. EF were the least permissive subset with percentages of GFP+ cells ranging from 0.01% to 0.05% in R5-spinoculated cultures, and 0.4% to 0.7% in X4-spinoculated cultures. In contrast to unsorted cells (Fig. 2B), gMFI of GFP was highest in the GC TFH subset (Fig. 2E). Due to significant downregulation of PD-1 and upregulation of CXCR5 induced by the sorting process itself (Supplemental Fig. 2), PD-1 and CXCR5 expression could not be assessed in sorted cell subsets. Collectively, these data suggest that GC TFH are substantially more permissive to HIV-1 than CXCR5+PD-1low populations, but that many GC TFH downregulate PD-1 and to a lesser extent CXCR5 during the course of productive infection and assume a CXCR5+PD-1low phenotype.
To evaluate whether heightened permissivity of GC TFH is related to chemokine co-receptor expression, CCR5 and CXCR4 expression on tonsil cells was determined prior to spinoculation by staining with antibodies and using QuantiBRITE beads (BD Biosciences) to determine the mean number of molecules per cell (19) (Fig. 3A). Similar percentages of CCR5+ cells were found in the EF, non-GC TFH, and GC TFH populations, whereas significantly lower percentages of CCR5+ cells were observed in the CXCR5+PD-1low population. There were no significant differences in the mean number of CCR5 molecules found on any of the tonsil subsets. The highest percentages of CXCR4+ cells were observed among the CXCR5+PD-1low and GC TFH subsets, and the lowest percentages observed in the EF subsets. GC TFH expressed significantly more CXCR4 molecules per cell than all other subsets.
We previously reported that the majority of HIV-1 RNA is harbored by activated HLA-DR+CD38+ (DR+38+) cells in lymphoid tissue T cells in vivo (19). To evaluate whether cellular activation accounts for the unique permissiveness of GC TFH, we examined DR+38+ cells in tonsil cell subsets at baseline and 2 days after spinoculation with HIV-1 reporter viruses. At baseline, more GC TFH were DR+38+ than any other subset (Fig. 3B), although differences were relatively modest and less than two-fold in all instances. Evaluation of GFP expression on day 2 revealed no evidence of preferential virus replication in DR+38+ cells within any of the subsets. Indeed, percentages and gMFI of GFP+ cells were significantly lower in DR+38+ cells compared to other cells in some subsets (Fig. 3C).
A previous study reported that spinoculation upregulates CD4 and CXCR4 expression on a T cell line (24). To determine whether spinoculation alters tonsil T cell subsets or upregulates HIV-1 co-receptors, expression of cell surface markers was determined on cells at baseline immediately prior to spinoculation, mock spinoculated cells, and cells that were rested at room temperature for 2 hours. Spinoculation had little influence on expression of CXCR5, PD-1, HLA-DR, or CD38 (Supplemental Fig. 3A). Neither percentages nor gMFI of CD4 or CCR5 appeared to be upregulated by spinoculation (Supplemental Fig. 3B). There was a slight upregulation of CXCR4 gMFI in spinoculated cells compared to those that were rested in all subsets, although values were not substantially different from those at baseline (Supplemental Fig. 3B). There was no evidence of an increase in percentages of CXCR4+ cells associated with spinoculation.
We next evaluated whether the heightened permissiveness of GC TFH to HIV-1 ex vivo is also observed in vivo. Previously, we reported that HIV-1 RNA+ cells are concentrated within B cell follicles in lymphoid tissues of HIV-1-infected individuals in vivo (2, 3). We hypothesized that if GC TFH are more permissive than other cells in vivo, that virus-producing cells would be more concentrated within the GC of B cell follicles compared to non-GC follicular regions. In situ hybridization combined with immunostaining for CD20 and IgD was used to identify HIV-1-producing cells in GC and non-GC areas of follicular and extrafollicular regions in inguinal lymph node tissue cross sections from 22 HIV-1-infected individuals who were not receiving antiretroviral therapy, as shown in representative images (Fig. 4A). Both GC and non-GC regions of follicle constituted a minority of lymphoid tissue (Fig. 4B). Frequencies of HIV-1 RNA+ cells were significantly higher in GC compared to non-GC regions of follicle, and in both instances levels were significantly higher than those found in extrafollicular regions (Fig. 4C). Furthermore, GC harbored the majority of virus-producing cells in most subjects’ lymph nodes (Fig. 4D). Control sections stained with sense probes were uniformly negative. Plasma HIV-1 RNA correlated significantly with frequencies of HIV-1 RNA+ cells within GC (r=0.6224; p=0.0020), non-GC F (r=0.6620; p=0.0010), and EF (r= 0.8291; p<0.0001).
It is well established that FDC-bound HIV-1 is highly concentrated within lymphoid tissues and potently infectious to CD4+ T cells ex vivo (14). To determine whether this extracellular virus might contribute to the susceptibility of GC TFH, p24 antigen staining, which identifies virions bound to FDC, but not intracellular virus production, was performed on lymph node tissue sections and analyzed by quantitative image analysis. P24 antigen was localized exclusively within F, as shown in a representative image in Fig. 4E. The percentage of p24 antigen staining ranged from 0.1% to 4.1% (median, 0.8%) of the follicular area, and did not correlate with either plasma HIV-1 RNA (p=0.13) or frequencies of HIV-1 RNA+ cells in GC (p=0.56).
As follicular cells in lymph nodes from HIV-1 seropositive individuals may differ in their permissivity from tonsil cells in seronegative individuals, we evaluated expression of GFP in cryopreserved, disaggregated lymph node cells from 3 untreated HIV-infected individuals after superinfection by spinoculation with the HIV-1 X4 GFP reporter virus ex vivo. As shown in Fig. 5A, both percentages of GFP+ cells and gMFI of GFP were higher in follicular compared to extrafollicular cells, similar to what was seen in ex vivo infection of tonsil cells (Fig. 1B). To address the possibility that follicular cells might be more resistant to CTL or NK cell clearance than extrafollicular cells, a portion of lymph node cells were depleted of CD8+ cells prior to spinoculation. Depletion of CD8+ cells resulted in higher percentages of GFP+ cells as well as higher gMFI of GFP in both follicular and extrafollicular subsets. The ratio of percentages and gMFI of GFP in depleted to undepleted cells was similar in the follicular and extrafollicular subsets (Fig. 5A), suggesting no major differences in susceptibility to CD8+ cell-mediated virus suppression between cells in these two compartments.
To assess whether activated lymph node cells in HIV-1 seropositive individuals were more permissive than other cells ex vivo, we evaluated GFP expression in DR+38+ cells and compared it to other subsets (Fig. 5B). Of note, percentages of DR+38+ cells did not differ between the F (range, 17%–24%) and EF subsets (range, 16%–24%) of these subjects. Percentages of GFP+ cells were similar in the DR+38+ subset compared to other cells, whereas gMFI of GFP was consistently lower in the DR+38+ subset compared to other cells. CD8 depletion resulted in small increases in percentages and gMFI of GFP in two subjects within these subsets, and large increases in percentages and gMFI of GFP in the activated subset in one subject. The ratio of percentages and gMFI of GFP in depleted compared to undepleted cells was higher in all instances in the DR+38+ subset compared to other cells, suggesting that activated CD4+ T cells may be more susceptible to CD8+ cell-mediated virus suppression than others.
We have previously demonstrated that both HIV-1 (2) and SIV RNA+ cells (1) are highly concentrated within B cell follicles during chronic, asymptomatic disease. Nevertheless, measurement of SIV RNA in sorted subsets of cells from rhesus macaques did not reveal differences between TFH and other populations, with the exception of elite controllers (10). This study used CD4 and PD-1, as well as additional markers to identify TFH. In light of prior findings that CD4 is downregulated on HIV-1-producing cells (19) and observations in the present study that PD-1 is downregulated on GC TFH, we hypothesized that by using less stringent criteria to define follicular cells, as well as by gating on CD3+CD8- cells, we would obtain results in cell sorting experiments more concordant with in situ hybridization studies.
First, we sought to determine the accuracy of CXCR5 as a marker for follicular CD4+ T cells in HIV-1-infected lymphoid tissues. Lymph node tissue sections from 9 untreated HIV-1 seropositive individuals were immunofluorescently stained with antibodies to CXCR5, CD20, and CD4, as shown in images from a representative tissue section (Fig. 6A–6D). It should be noted that although CD4 is downregulated on cell surfaces in the context of HIV-1 infection, intracellular CD4 is readily detectable by immunofluorescent staining within tissue sections. A median of 28% (range, 18–45%) of CD4+ cells in the lymph node sections expressed CXCR5, and a median of 96% (range, 92–99%) of CD4+CXCR5+ cells were located within B cell follicles. Of CD4+ cells located in B cell follicles, a median of 53% (range, 42% to 74%) expressed CXCR5.
We next performed in situ hybridization for HIV-1 RNA combined with immunostaining for CD20 in 4 untreated, asymptomatic, chronically HIV-1-infected individuals, whose clinical characteristics are shown in Table I. We determined the frequencies of CD4+ cells/mm2 by immunohistochemical staining as well as frequencies of HIV-1 RNA+ cells in follicular and extrafollicular compartments, and used these data to estimate numbers of HIV-1 producing cells/106 CD4+ cells within these compartments (Table I). The likelihood that a follicular CD4+ cell would be productively infected was 41- to 164-fold more than that of an extrafollicular CD4+ T cell. Finally, using quantitative RT PCR, we determined HIV-1 RNA levels in CD3+CD8-CXCR5+ and CD3+CD8-CXCR5- sorted lymph node cell subsets from the same subjects. CXCR5+ T cells consistently harbored more HIV-1 RNA than CXCR5- cells ranging from 11 to 66 times more HIV-1 RNA copies per 105 cells (Table I).
This is the first study to examine in detail the permissiveness of human TFH subsets to HIV-1 infection ex vivo. We found that GC TFH in human tonsils were highly permissive to HIV-1, but downregulated CD4, PD-1, and to a lesser extent CXCR5 during productive infection such that many virus-producing cells did not have a typical GC TFH phenotype after 2 days of culture. We further demonstrated for the first time that HIV-1 replication is highly concentrated within GC compared to non-GC regions of the follicle, indicating that GC TFH are highly permissive to HIV-1 in vivo as well. We confirmed the permissiveness of follicular lymph node cells from HIV-1 seropositive individuals by infection of these cells with HIV-1 ex vivo. Finally, we demonstrated that almost all CXCR5+ CD4+ cells are located within follicles and that viral RNA is highly concentrated within CXCR5+CD3+CD8- cell subsets. Collectively, these data indicate that TFH, and most likely cells that originally had a GC TFH phenotype, are the major HIV-1-producing cells in vivo, and that increased permissivity of GC TFH contributes to heightened levels of HIV-1 replication in B cell follicles and GC.
GC TFH were the most permissive cell subset to HIV-1 in sorted cell cultures. GFP expression in the non-GC TFH subset was not evaluated in sorting experiments and consequently firm conclusions about this subset’s permissiveness cannot be made. It is noteworthy, however, that non-GC TFH represented a minority of productively infected cells both ex vivo and in vivo. Increased permissivity of PD-1high expressing cells has been reported previously in PHA-blasted PBMC (25). Both PD-1 (25) and CXCR5 (26) are transiently upregulated upon T cell activation, although the resulting PD-1highCXCR5high cells are not necessarily GC TFH (26). A strength of the present study is that secondary lymphoid tissue cells were not stimulated ex vivo, and therefore their permissivity and phenotype were not altered by nonphysiologic stimulation. It should be noted, however, that spinoculation does not necessarily recapitulate infection in vivo. Although we determined it did not substantially alter HIV-1 co-receptor expression, with the exception of a subtle upregulation of gMFI of CXCR4, it could enhance permissivity of cells in other ways, such as by promoting integration (24).
We investigated several potential mechanisms to explain the unique permissiveness of GC TFH to HIV-1. Naïve CD4+ T cells harbor multiple restriction factors that impair HIV-1 integration and replication [reviewed in (21)], and indeed we found that memory cells were significantly more permissive than naïve ones in our tonsil culture system. Nevertheless, even after controlling for memory phenotype, follicular cells remained substantially more permissive than non-follicular cells to HIV-1. Chemokine receptor expression was not strongly related to tonsil cell subsets’ ability to replicate HIV-1, suggesting that viral entry into cells was not a major determinant of differences in permissivity. A recent study revealed that CXCR3+CXCR5+CD4+ T cells in PBMC are associated with higher levels of viremia in vaccinated rhesus macaques as well as CCR5 expression (27). Whether CXCR3 is associated with increased GC TFH permissivity would be an important question to address in future studies.
Although GC TFH had modestly elevated levels of HLA-DR and CD38 co-expression, which has been linked to HIV-1 viral loads (28) as well as disease progression (28, 29), activated cells did not demonstrate elevated GFP expression in any of the follicular subsets in tonsils from naïve hosts or lymph nodes from HIV-1-infected individuals superinfected with reporter virus. These findings are surprising, as HIV-1 RNA is highly concentrated within DR+38+ lymph node cells in vivo (19), and suggest that the association of HLA-DR and CD38 co-expression with virus replication in vivo may be due to correlation rather than causation. The mechanism underlying this correlation remains to be determined as it is not recapitulated during ex vivo infection of disaggregated cells. Thacker et. al. demonstrated that when FDC were added to PHA-blasted CD4+ T cell cultures, HIV-1 transcription doubled, virus production increased approximately four-fold, and these phenomena were mediated by FDC release of TNF-alpha and increased activated NF-κB (13). They further demonstrated that CD57+CD4+ cells isolated from tonsils expressed elevated levels of activated NF-κB. Even though FDC were no longer in intimate contact with cells in our cultures, the GC TFH probably retained elevated activated NF-κB similar to what Thacker et. al. observed, which likely mediated their increased permissivity in our studies. It is possible that when FDC interact with CD4+ T cells they induce expression of HLA-DR and CD38 as well as infect them with HIV-1. It would be worthwhile to explore this in future studies as a possible explanation of the association of HIV-1 replication with DR+38+ cells in vivo, but not in our ex vivo system.
Our studies revealed downregulation of CD4, PD-1, and to a lesser extent CXCR5 on virus-producing GC TFH infected ex vivo. Downregulation of CD4 by HIV-1 Nef has been well described, and confers enhanced HIV-1 infectivity by increasing virion incorporation of envelope (30–32). Downregulation of PD-1 expression on HIV-1 producing cells has been reported previously in PHA-blasted PBMC infected with X4-tropic HIV, and was shown to promote survival of those cells (25). In our study, CXCR5 was downregulated from CXCR5high to CXCR5+ levels on some virus-producing cells, and in X4-infected cultures percentages of CXCR5- cells increased overall, suggesting that some cells may completely lose CXCR5 expression in the context of HIV-1 infection. Nef-induced downregulation of CXCR5 has been reported in HIV-1-producing cell lines (33), but to our knowledge this is the first report of HIV-1-induced downregulation of CXCR5 in primary cells. It is well established that expression of CXCR5 is essential for migration of TFH into B cell follicles (34). Downregulation of CXCR5 on GC TFH likely fosters HIV-1 pathogenesis by promoting emigration of productively infected cells from follicles into non-GC regions of follicles. It is possible that CXCL13 expression in B cell follicles further induces CXCR5 downregulation and thereby promotes dissemination of virus-producing cells to other sites in the body.
We and others have previously reported that the majority of HIV-1 and SIV replication during chronic infection prior to onset of severe immunodeficiency occurs within B cell follicles in secondary lymphoid tissues (1–6). This is the first study to demonstrate quantitatively that virus-producing cells are most highly concentrated in GC of the follicles. A prior study showed that virions bound extracellularly to FDC in GC are potently infectious to CD4+ T cells in vitro (14). In the present study the percentage of virion-bound FDC in tissues did not correlate with either plasma viral load or frequencies of HIV-1 RNA+ cells within the tissues. These findings do not exclude a role for virion-bound FDC in HIV-1 infection of GC TFH, but suggest that the number of virions on FDC is not limiting to HIV-1 replication.
We have previously demonstrated that virus-specific CD8+ T cells fail to accumulate in large numbers in B cell follicles in both chronic HIV-1 and SIV infection, and posited that this could at least partially account for heightened virus replication at those sites (1, 2). We also previously reported that follicular CD4+ T cells have relatively elevated levels of Bcl-2 expression, suggesting that they could be relatively resistant to killing by either NK cells or CTL (20). Importantly, CD8 depletion studies showed no evidence that lymph node follicular cells from HIV-1-infected individuals were less susceptible to CD8+ cell-mediated killing than extrafollicular cells. Furthermore, DR+38+ cells appeared slightly more susceptible to CD8+ cell-mediated virus suppression, weighing against resistance to CTL killing as a potential mechanism for the fact that these cells harbor the majority of HIV-1 RNA in vivo. It should be noted, however, that the lymph node cells were superinfected with X4 virus. It would be important to validate in future studies that this is also true of R5 virus infections, which constitute the majority of infections prior to development of AIDS.
Several prior studies have sorted lymphoid tissue cells from HIV-1-infected humans and SIV-infected rhesus macaques to evaluate the contribution of GC TFH to virus replication. Perreau et. al. found that lymph node cell TFH produced more virus following anti-CD3 and anti-CD28 stimulation, and also harbored the largest concentrations of HIV-1 DNA compared to other subsets, but did not report HIV-1 RNA levels in TFH (11). In the SIV-infected rhesus macaque model, Petrovas et. al. reported that TFH harbored significantly more SIV DNA during acute SIV infection, but not during chronic disease (12). They also did not report cell-associated SIV RNA measurements. Finally, Fukazawa et. al. reported that both SIV DNA and RNA was concentrated in the TFH of elite controllers, but not progressors in the SIV-infected rhesus macaque model (10). A fundamental limitation of all these studies is that cells were sorted on the basis of expression of CD4 and PD-1 as well as other markers, and consequently did not contain the GC TFH that had downregulated PD-1 and CD4. To address these problems, we sorted CD3+CD8- cells into CXCR5+ and CXCR5- populations to quantify follicular and extrafollicular HIV-1 RNA production, respectively, within human lymph node cells. These studies demonstrated between 11- and 66-fold more HIV-1 RNA per 105 cells in the CXCR5+ compared to the CXCR5- subset of lymph node cells from four HIV-1-infected individuals, none of whom was an elite controller. Interestingly, in 3 of 4 subjects the difference in frequencies of virus-producing cells in follicular and extrafollicular compartments was greater than that suggested by measurement of cellular HIV-1 RNA. One potential explanation is that extrafollicular cells produce more HIV-1 RNA than follicular cells, although this was not observed during ex vivo infection of either tonsils or human lymph nodes. Alternatively, or in addition, this could be explained by downregulation of CXCR5 on some of the follicular virus-producing cells. Regardless, these studies establish that follicular T cells harbor the majority of HIV-1 RNA during chronic untreated disease prior to the onset of immunodeficiency, and are largely concordant with in situ hybridization studies within intact tissue sections.
Our studies indicate that the majority of virus-producing cells originate in GC of B cell follicles in chronically HIV-1-infected individuals prior to AIDS. These findings suggest that GC TFH may form a substantial portion of the latent reservoir in antiretroviral-treated HIV-1-infected individuals. Fukazawa et. al. (10) reported that in SIV-infected rhesus macaques who were treated with antiretroviral therapy for at least 4 months, SIV DNA was not concentrated in CD4+ T cells that expressed PD-1 and CD200. Nevertheless, the plasticity of GC TFH during HIV-1 infection suggested by the present study indicates that one cannot conclude that the cells that harbor the latent reservoir were not originally GC TFH, only that they do not presently have that phenotype. Recent studies have validated the presence of memory TFH in humans [reviewed in (35)], and they have been characterized phenotypically as CXCR5+, but lack expression of most other typical markers of TFH. One study identified the phenotype of memory TFH in humans as CXCR5+CXCR3-PD-1low (36). It is noteworthy, however, that CXCR5 expression diminished or was completely lost on effector TFH that were adoptively transferred into a mouse with no antigen expression (37). Thus, determining the contribution of GC TFH to the latent HIV-1 reservoir in humans with virologic suppression in the context of antiretroviral therapy could be challenging.
In Fukazawa et. al.’s study of antiretroviral-treated rhesus macaques, the majority of SIV RNA was found in the PD-1+CD200+CD4+ cells (10). The central role of GC TFH in active virus replication during suppressive therapy further indicates that a better understanding of the role of FDC, which are in intimate contact with these cells and potently infectious due to extracellular virions (14), is urgently needed. In untreated HIV-1-infected individuals’ lymphoid tissues, the amount of HIV-1 RNA bound to FDC is approximately 10-fold higher than that inside of productively infected cells (38), and although it decays over time is nonetheless detectable after 6 months of ART (39). Furthermore, FDC-bound HIV-1 has been shown to remain infectious for at least 9 months within a non-permissive mouse model (40). Despite this, little attention has been directed at understanding the rate of decay and persistence of infectivity of virions bound to FDC in the context of ART treatment in humans. Substantial controversy exists as to whether there are active cycles of virus replication or simply reactivation of virus from latently infected T cells. The potential role of FDC in reseeding GC TFH should be carefully examined, particularly in light of the failure of bone marrow transplant patients to be cured of HIV-1 infection (41). A better understanding of the mechanisms that promote virus replication in GC TFH and particularly the role of FDC could provide critical insight into both the active and latent reservoirs of HIV-1, as well as suggest novel strategies to eliminate these reservoirs.
This work was supported by the University of Colorado School of Medicine Research Track and the University of Colorado Department of Medicine (S.K.), and Public Health Services Grants R01AI096966 (E.C.), R56AI080418 (EC), R01 AI078783A (DL), HL103286 (T.A.), 5T32AI007447 (B.M), and T32 AI007405 (S.M.). Cell sorting was performed by the University of Colorado Cancer Flow Cytometry Shared Resource, which is supported by the Cancer Center Support Grant P30CA046934 and the Skin Diseases Research Cores Grant P30AR057212.
These data were presented in part at the19th Conference on Retroviruses and Opportunistic Infections, Seattle, March 5-8, 2012 [abstract 203] and the 20th International AIDS conference, Melbourne, July 20-25, 2014 [abstract A-641-0017-02285].