Ex vivo rectal explant model reveals potential opposing roles of Natural Killer cells and Marginal Zone-like B cells in HIV-1 infection

Our understanding of innate immune responses in human rectal mucosal tissues (RM) and their contributions to promoting or restricting HIV transmission is limited. We defined the RM composition of innate and innate-like cell subsets, including plasmacytoid dendritic cells; CD1c + myeloid DCs; neutrophils; macrophages; natural killer cells (NK); Marginal Zone-like B cells (MZB); γδ T cells; and mucosal-associated invariant T cells in RM from 69 HIV-negative men by flow cytometry. Associations between these cell subsets and HIV-1 replication in ex vivo RM explant challenge experiments revealed an inverse correlation between RM-NK and p24 production, in contrast to a positive association between RM-MZB and HIV replication. Comparison of RM and blood-derived MZB and NK illustrated qualitative and quantitative differences between tissue compartments. Additionally, 22 soluble molecules were measured in a subset of explant cultures (n = 26). Higher production of IL-17A, IFN-γ, IL-10, IP-10, GM-CSF, sFasL, Granzyme A, Granzyme B, Granulysin, and Perforin following infection positively correlated with HIV replication. These data show novel associations between MZB and NK cells and p24 production in RM and underscore the importance of inflammatory cytokines in mucosal HIV infection, demonstrating the likely critical role these innate immune responses play in early mucosal HIV replication in humans.


Pre-infection MZB and NK cells are associated with HIV replication in the rectal explant model.
In parallel with flow cytometric analysis of the RM, three biopsies from the same individuals collected at the same time were each challenged with HIV-1 laboratory variant, BaL, as an ex vivo model of HIV-1 infection of human RM tissue. Supernatants were collected at Day 3, 7, 10, 14, and 18 and assessed for p24 production normalized to biopsy weight. Median p24 concentrations varied substantially between individuals (n = 86, Fig. 2A). To determine whether individual variation in innate or innate-like subsets ( Fig. 1) could be contributing to this observed range of HIV-1 replication, Spearman rank correlations were calculated for each cell subset at the baseline vs. the median log Area Under the Curve (AUC) of p24 for each individual. This analysis revealed a positive correlation between the percentage of RM MZB cells and p24 production in parallel RM biopsies (n = 60, p < 0.0001, r = 0.51), and a negative correlation between NK cell abundance and p24 production (n = 60, p = 0.002, r = -0.39) (Fig. 2B-D). No other quantified subset correlated with p24 production (Fig. 2B), thus RM MZB and NK were the focus of further characterization.
MZB and NK cells within rectal mucosal tissues are quantitatively and qualitatively distinct from their circulating counterparts within the peripheral blood compartment. Both MZB cells and NK cells can be found circulating within the human peripheral blood compartment, in addition to their presence in the RM. To compare circulating cells to their RM-residing counterparts, an identical gating strategy ( Fig. 1) was utilized with PBMC collected from the same participant at the time rectal biopsies were obtained. In contrast to RM MZB and NK percentages (Fig. 2B-D www.nature.com/scientificreports/ cell subsets and p24 production within parallel rectal explants ( Fig. 3A-B). Direct comparison of blood and RM compartments first revealed distinct quantitative differences between both MZB and NK in these subsets. As a percentage of total CD45 + cells, B cells as a whole were more abundant within the RM vs. blood (RM mean = 26%, IQR 14-38%; Blood mean = 12%, IQR 8-14%; two-way ANOVA, p < 0.0001, all comparisons), with MZB cells proportionally more abundant within the RM (RM mean 12%, IQR 7-16%; blood mean 6%, IQR 4-8%; two-way ANOVA, p < 0.0001) (Fig. 3C). Compartmental differences were also seen within the NK population (Fig. 3D). NKs were more abundant within the blood vs. the RM (RM mean 5%, IQR 1-6%; Blood mean 13%, IQR 8-16%; two-way ANOVA, p < 0.0001, all comparisons), and cell surface markers were highly divergent and differentially distributed between CD56 + , CD16 + , and dual-positive NK between the two compartments. This differential was not due to differences in processing protocols between blood and RM cells, as blood cells carried through the RM processing protocol (with Collagenase IV and DNase, see Methods) were identical to unprocessed cells (dns), and tissue-specific variation in CD16 expression on NK cells is expected 27 .
To gain further insight into these differences, Hierarchical Stochastic Neighbor Embedding (HSNE) analysis was performed on MZB and NK cells 28 . To prevent a single participant or tissue compartment from dominating or skewing results, for MZB cells, no more than 5000 CD45 + , CD3-, CD20 + , HLA-DR + , CD1c + events from each participant from each compartment were analyzed based on CD20, HLA-DR, and CD1c expression. Based on these markers, blood and RM MZB cells did not segregate into discrete populations (Fig. 4A). However, clustering of cells was polarized based on their origin. While all analyzed cells were positive for CD20, HLA-DR, and CD1c, heatmap analysis illustrated blood MZB clustered based on elevated CD20 and CD1c expression, while RM MZB expressed relatively higher levels of HLA-DR (Fig. 4B). Mean fluorescence intensity (MFI) from matched blood and RM-derived MZB also revealed differential expression of these three markers between blood and RM-residing MZB (Fig. 4C, Wilcoxon test, p < 0.0001, all markers). Similar analysis of blood and RM-residing NK cells was performed on CD45+ , Lineage-, CD16± , CD56 ± . In this analysis, two distinct populations emerged (Fig. 5A), based substantially on the near absence of CD16 on RM-residing NK (Fig. 5B). MFI from matched blood and RM-residing NK cells again revealed statistical differences in expression of both CD16 and CD56 within NKs from these two compartments (Fig. 5C, Wilcoxon test, p < 0.0001, all markers). . Mean and median longitudinal concentrations were at or below the assay limit of detection for eight molecules: IL-2, IL-4, TNFα, IFN-λ1, IL-12p70, IFN-α2, IFN-λ2/3, and IFN-β2. Concentrations of these cytokines were similarly low/undetectable in the mock infected biopsy (biopsies exposed to culture media rather than viral challenge media), suggesting they are not present at appreciable concentrations in the rectal explant model; however, it is possible that they were present at levels below the detection threshold, or were expressed and degraded prior to the Day 3 sampling time-point. The remaining 14 molecules were present within the supernatants of infected rectal explants: IL-1β, IL-6, IL-8, IL-10, IL-17A, IFN-γ, IP-10, GM-CSF, sFas, sFasL, Granzyme A (GZA), Granzyme B (GZB), Granulysin, and Perforin (Supplemental Figs. 1, 1). As expected, no assayed cytokine or effector molecule was detectable in appreciable concentrations within the culture media used to maintain the biopsies (Supplemental Figs. 1, 1). Several molecules were present at high concentrations within the viral challenge stock media (> 10 pg/ml, including IFN-γ, GZA, GZB, GM-CSF, IL-2), likely due to propagation of the virus in human PBMC (Supplemental Figs. 1, 1). However, the complete absence of IL-2 in all infected biopsies (despite > 250 pg/ml concentration in the viral stock), along with the longitudinal kinetics of IFN-γ, GZA, GZB, GM-CSF in the infected biopsies (including complete absence in some biopsies) (Supplemental Figs. 1, 1) suggests post-challenge washing of biopsies was sufficient to remove excess viral challenge media, and the detected concentrations of these molecules in the infected biopsies represent de novo production, rather than carry-over from the challenge stock. www.nature.com/scientificreports/ To determine the relationship between the 14 detectable molecules and HIV-1 replication, longitudinal soluble molecule concentrations were normalized to biopsy weight. The resulting AUC (from days 3-14) was calculated for each molecule, and correlated with the corresponding normalized p24 logAUC values (Spearman, Fig. 6). Significant correlations (controlled for multiple comparisons, threshold of significance reduced to p < 0.0076 by two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli) were observed with IL-17A (r = 0.75, p < 0.0001), IFN-γ (r = 0.77, p < 0.0001), IL-10 (r = 0.63, p = 0.0006), IP-10 (r = 0.72, p < 0.0001), GM-CSF (r = 0.58, p = 0.002), sFasL (r = 0.59, p = 0.002), GZA (r = 0.69, p < 0.0001), GZB (r = 0.59, p = 0.002), Granulysin (r = 0.51, p = 0.007), and Perforin (r = 0.75, p < 0.0001). IL-6, IL-8, sFas, and IL-1β did not correlate with p24 production. Of note, IL-6, IL-8, IL-1β, and GM-CSF were also elevated in the mock infected biopsy, thus are possibly secreted in response to the injury induced by the biopsy procedure and cell death, and potentially obfuscated our ability to fully elucidate the relationship between these inflammatory cytokines and HIV replication in this model.

Discussion
The findings from this study demonstrate that innate immune cell subsets, as a whole, are a minority population of CD45 + cells within the RM, yet some are potentially influencing HIV production within the RM. Most of the subsets defined here were present at or below 1% of total CD45 + cells. A notable exception to this trend were the MZB cells, found at a median of > 10% of CD45 + cells. MZB cells are emerging as a cell type of interest in HIV-1 transmission and pathogenesis. Recent studies have reported decreased numbers of MZB in the blood and within the female reproductive tract of HIV-exposed but uninfected commercial sex workers 22,23 . This suggests that MZB cells could be contributing to HIV transmission within the female genital tract, although mucosal B cells at this site are relatively rare 29 . Though MZB have not been specifically quantified in the male reproductive tract, CD19 + B cells as a whole comprise 1-3% of CD45 + cells 30 . This is in contrast to the MZB distribution we observed within the RM, where we noted a substantial proportion of CD45 + cells are MZB cells. Also, a recent study by Liechti et al.noted an inverse correlation between MZB within the blood and CD4 + T cell counts, and a positive correlation between MZB and viral load in the setting of chronic HIV-1 infection 31 . Congruent with these previous findings, we also observed a positive correlation between MZB and p24 production after ex vivo challenge of rectal tissue with HIV. In contrast to the RM residing subsets (Fig. 2), there was no correlation with MZB (A) or NK (B) percentages within blood and p24 production in the rectal explant. (C) Within the RM, there are more B cells, and proportionally more MZB cells, than found within the blood (2way ANOVA, p < 0.0001). (D) There are fewer NK within the RM vs. blood, however substantially more demonstrate the CD16-CD56 + phenotype (2way ANOVA, p < 0.0001). www.nature.com/scientificreports/  www.nature.com/scientificreports/ The precise mechanisms by which MZB could contribute to HIV transmission and replication has yet to be elucidated. MZB may be directly involved in HIV-1 replication via trans infection of tissue-resident CD4 + T cells. Although macrophages and DCs are more commonly associated with trans infection, there are higher proportions of MZB than macrophages, mDC, or pDC in RM, and it has been shown previously that peripheral B cells mediate trans infection as efficiently as other APC subsets 14 . Although traditional HIV-binding receptors such as CD21 and DC-SIGN were not quantified here, studies have reported MZB cells are capable of binding directly to HIV gp120 23 . When combined with unique intrinsic features of the rectal mucosa ex vivo HIV challenge model (i.e. the absence of requirement for immune activating agents commonly necessary for infection of PBMC), it is possible that the three-dimensional structure of the RM is maintaining cell-to-cell contacts thus facilitating trans infection. It is also feasible that within the RM, MZB are part of an immunological ecosystem that promotes increased HIV susceptibility of target cells. Low levels of CD4 + T cell activation have been associated with various HIV-exposed but uninfected populations 32 . In the context of murine studies, where MZB are generally only found within follicles, these cells are potent activators of CD4 + T cells 33 . Thus, MZB could indirectly contribute to p24 accumulation by robustly activating local CD4 + T cells within RM tissues. Mechanistically defining the relationship between MZB cells and HIV replication could prove critical for better understanding RM transmission and in the development of novel biomedical HIV prevention interventions. This will necessitate further utilization of the human rectal explant model, as our data demonstrate that human MZB circulating in peripheral blood are quantifiably different from those residing in the RM, even in comparative analysis restricted to a modest group of surface receptor markers. However, prior to this, it will be beneficial to utilize exploratory approaches, such as single-cell RNAseq, to characterize these MZB-like cells residing within non-splenic tissue compartments of humans. Ongoing studies within the field would allow for direct comparison and quantification of transcriptomic similarities of these RM-residing B cells to traditional splenic MZB, as well as circulating MZB and other B cell subsets 34 . This will provide critical foundational information as to the natural function of these cells within the RM, as well as providing insight into how these cells could potentially participate in propagating HIV replication. In addition, use of the human explant model is necessitated by limitations in animal models considering that circulating and tissue-residing MZB are not found in mouse models and may be absent in NHP models as well 35 .
Of the potential innate negative regulators of HIV replication within the RM, we identified a negative correlation between NK percentages and p24 accumulation. That is, biopsies from participants with reduced p24 production also displayed a higher proportion of NK cells within the RM of parallel biopsies. In this study, we considered lineage negative, CD163− (RB) or CD14− (blood), CD56 + and/or CD16 + cells as a pooled population, and did not stratify NK into the archetypical categories of CD56 bright and CD16 + CD56 dim , due to the diversity of expression of these molecules on NK residing in numerous tissues vs. traditional NK populations within peripheral blood 36 . In concordance with studies characterizing NK residing in other tissues, we also observed that this population of cells within the RM does not express high levels of CD16. This is in contrast to approximately 40-60% NK cells that reside within the penile mucosal tissue, which do appear to express CD16 30 . And, within the female reproductive tract, CD16 is expressed by NK residing within the ectocervix, while uterine NK are notable for the absence of CD16 37,38 . In this study, utilizing RM tissue, the absence of both CD16 and ADCC-mediating antibodies (all participants were HIV seronegative), the potential mechanism of action by which RM NK could reduce HIV replication would likely be mediated via any number of NCR, KLR, and KIR, each of which can have a profound impact on NK-mediated killing of HIV infected cells 8,9 . For example, HIV has evolved mechanisms www.nature.com/scientificreports/ to reduce the presence of Class I Major Histocompatibility Complexes on the surface of infected cells to evade antigen-specific T-cell responses (via HIV proteins Vpu and Nef), however this renders infected cells more susceptible to recognition and elimination by NK. The absence of self-ligands on the surface of infected cells (i.e. 'missing self ') could lead to activation of NK-killing, due to the absence of inhibitory signals mediated by KLR NKG2A or inhibitory KIR. While we have not identified the exact cellular source of the cell-killing effector molecules detected in this study (sFasL, GZA, GZB, granulysin, and perforin), it is possible that NK responses are contributing to the production of these molecules. Furthermore, this conclusion would suggest that just as HIV-1 out-maneuvers the B cell and T cell responses, HIV is likewise capable of outpacing the relatively rapid NK response within the RM 39 , similar to what has been observed within the female genital tract of SIV-infected rhesus macaques 40 . Our findings do introduce the possibility that increasing the presence of NK within the RM, or optimizing the anti-viral capabilities of RM-residing NKs, could contribute to an immunological environment that is not conducive to HIV replication. Similarly, as both RM-NK and uterine NK lack expression of CD16, similar enhancement of ADCC-independent cytotoxic pathways of NK residing in both of these tissue compartments could provide additional protection against HIV infection. Because RM-resident NKs are distinct from those found in PBMC, detailed phenotypic and genetic characterization of these RM-residing NK cells will be necessary to elucidate the inhibitory capabilities of this subset against HIV infection. This avenue of research must be expanded to explore potentially distinctive features of additional subject-derived HIV isolates (T/F and non-T/F) of numerous subtypes, to mechanistically define the selective pressures exerted by NK and other innate and innate-like subsets 41 .
In addition to identifying associations between RM-residing MZB and NK cells and HIV-1 replication, the explant model utilized here allowed for the identification of associations between p24 production and a number of assayed cytokines and effector molecules potentially released early after HIV infection, before the adaptive immune response matures. In the absence of this RM-specific data, elevated cytokines have been described in the context of systemic levels of inflammatory and anti-viral cytokines present in blood during acute/early HIV-1 infection 42,43 . These systemic studies provide an informative point-of-comparison for our observations, and we note a number of cytokines typically elevated systemically are also associated with p24 production in the rectal explant model. For example, IFN-γ was one of the first cytokines identified as elevated in serum during acute HIV-1 infection 42,43 . IFN-γ is also upregulated within the RM during chronic HIV infection, persisting even after initiation of ART 44 . We also identified IFN-γ as critical to propagation of HIV-1 infection within the rectal mucosa. Additional concordant cytokines previously reported to elevated in blood in acute HIV infection and now seen in our study within the rectal explant model include IP-10, IL-10, and GM-CSF, reemphasizing their role in the pathogenesis of early HIV infection 45 .
Unique to the rectal explant model, however, was the identification of the strong association of IL-17A production with p24 accumulation. Though it has been conceptually hypothesized that IL-17 could influence HIV replication (reviewed in 46 ), this relationship has been difficult to elucidate in humans. Serum levels of IL-17A are not elevated in most individuals during acute infection 47 , nor in NHP models utilizing pathogenic and nonpathogenic SIV variants 48 . It is logical that this relationship is observable directly in ex vivo challenged rectal explants, even though it does not emerge in systemic studies of acute HIV infection, as there are tissue-dependent roles for IL-17A in both healthy and disease states 49 . Within the gut mucosa, IL-17A plays a vital role in balancing interactions with the microbiome and in maintaining intestinal mucosal integrity. Overexpression of IL-17A is associated with inflammatory bowel disease, while attempts to suppress IL-17A production in these instances can lead to further intestinal epithelial injury and colitis 49 . While our observations have identified IL-17A as a facilitator of HIV replication and, as such, it could be an attractive novel target for prophylactic or therapeutic interventions, the importance of maintaining homeostatic levels of IL-17A in human RM may deem it a poor target for use in clinical interventions. Future studies should capitalize on the rectal explant model to determine whether targeting IL-17A production in a localized, temporary manner in RM tissues during HIV-1 exposure limits, or promotes, HIV transmission.
Also notable in the explant model utilized here are the cytokines elevated upon systemic HIV infection, such as Type 1 interferon, that were not elevated in the RM explant model 47 . Neither IFN-α nor IFN-β were present at substantial concentrations in supernatants at any point between Day 3 and Day 14 post infection. It is hypothesized that these cytokines could act as a strong selective pressure on transmitted/founder viruses, restricting the replication of the most susceptible genetic variants, contributing to the genetic bottleneck frequently observed during transmission 50 . Notably, in female rhesus macaques, pDCs within the vaginal mucosa are strongly positive for both IFN-α and -β 51 . It is possible that the absence of these cytokines in the rectal explant model are due to the dearth of pDCs in the RM and the inability to recruit additional cells to the site of infection in the explant model. Alternatively, the absence of substantial IFN-α in this study might be attributable to sex-specific differences in pDC function 52 . Finally, it is also possible that the laboratory HIV-1 variant, BaL, does not have the same capacity to induce Type 1 interferon as patient-derived isolates 53,54 . Again, further utilization of the rectal explant model with subject-derived T/F and non-T/F viruses will be critical for quantifying differences in i) cytokine induction and ii) susceptibility to the anti-viral cytokines we observed here, to further define the cytokines exerting selective pressure on the virus at the time of rectal transmission.
Limitations of the current study include a single-sex cohort, and use of a single HIV laboratory variant; follow-up studies are being planned in our lab to address these limitations. The rectal explant challenges were also carried out with the use of cell-free viral stocks. Exposure of RM tissue to HIV infected cells and cell-associated virus found within the complex milieu of semen might result in different selective pressures exerted by different immune cells and cytokines within the RM, and should be explored in future studies as well 55,56 . Furthermore, it will be critical to include additional markers relevant to the identified cellular subsets of interest in future studies to fully elucidate potential mechanisms of action for our observed associations with HIV replication (e.g. quantification of CD21, DC-SIGN, IgM for MZB; KIRs, NCRs, KLRs for NK). Additionally, all participants Scientific Reports | (2020) 10:20154 | https://doi.org/10.1038/s41598-020-76976-5 www.nature.com/scientificreports/ included in this analysis were negative for bacterial STIs. It is possible that RM immuno-environments that include symptomatic or asymptomatic STI infections, or an immunological landscape created by any number of inflammatory bowel conditions, are radically different from the distribution of immune cell subsets observed in this study. In alternative 'inflammatory' environments, it is possible additional immune cell subsets and inflammatory cytokines are associated with HIV replication in the ex vivo model. The present study is also limited in that the observations are currently correlative. However, our findings are leading to focused, hypothesis-driven mechanistic studies within the explant rectal challenge system in future studies. For example, flow cytometry panels focused on defining features of MZB and NK cells can be utilized to interrogate biopsies over the course of ex vivo HIV infection. We anticipate NK cells might proliferate in response to HIV infection of RM tissue, while the expression of activation markers, such as HLA-DR and CD38, might increase on both MZB and NK cell subsets. Flow analysis after HIV infection would also allow for intracellular staining to identify the cellular sources of the cytokines and effector molecules identified in the explant supernantants in the current study. Furthermore, if NK cells within the RM are targeting the first HIV infected cells within this tissue in an ADCC-independent mechanism, this activity could be augmented in future studies by supplementing the explant supernatant with NK activators, such as IL-15, or check-point inhibiting antibodies that could enhance ADCC-independent killing pathways, such as αNKG2a or αKIR antibodies. It also might be possible to interrogate the role of MZB via antibodies without disrupting the three-dimensional structure of the RM biopsies via inhibition of potential B-cell mediators of trans infection, such as αCD21 or αDC-SIGN. Furthermore, within the current analysis, we cannot determine whether the positive association between cytokines and p24 is due to i) cytokine production emerging as a direct result of p24 production and inflammation, or, ii) the presence of these cytokines is further augmenting and enhancing HIV replication. In future analyses, explants can be pre-treated with exogenous cytokines or cytokine neutralizing antibodies to directly elucidate cause-and-effect relationships between the associations identified here.
In summary, we have identified a number of unique innate immune cell subsets and cytokines associated with the replication of HIV-1 in human rectal mucosal tissues after viral exposure. These observations would not have been possible without this novel use of human RM and the rectal explant model of HIV infection, due to the tissue-specific and potentially species-specific features of MZB cells, NK cells, and IL-17A. Future studies will likewise necessitate the use of human RM tissue to further characterize these cellular subsets, to identify the source of IL-17A, and to determine the mechanisms by which they are contributing to or restricting HIV replication in RM tissues. This further investigation could prove essential in pursuit of a better understanding of the process of HIV-1 transmission across RM tissue and also in developing strategic methods for reducing RM transmission events.

Methods
Study population. This study obtained approval from Emory University Institutional Review Board (IRB).
All procedures and experiments were performed in accordance with appropriate guidelines and regulations. Written informed consent was obtained from all participants. Participants were healthy, HIV-negative, aged 18-69 years (median age 36), from the Atlanta metropolitan area, and tested negative for rectal gonorrhea, rectal chlamydia, and syphilis at the time of sampling. Participants included in this analysis were recruited for a larger parent study focused on men who have sex with men that enrolled men who do and do not regularly engage in receptive anal intercourse; therefore, women were not eligible for enrollment. Individuals who were determined to be high risk for complications from rectal biopsy procedures or who intended to take pre-exposure prophylaxis medications during the study were not enrolled. Biopsies were collected 3 to 10 cm from the anal verge via rigid sigmoidoscopy with no prior bowel preparation. During the same visit, PBMC were collected via BD Vacutainer Cell Preparation (CPT) tubes.
Statistical analyses. All graphing, calculations, and analyses were performed in Prism 7.0. Kruskal-Wallis multiple comparisons analysis was used to evaluate frequencies of each innate cell type within the RM. Spearman correlations were utilized for associations between p24 logAUC and cell subsets, and cytokine AUC. For cytokine and effector molecule correlations, the threshold of significance was adjusted to p < 0.0076, after correction for multiple comparisons (Benjamini, Krieger and Yekutieli). Two-way ANOVA was used to compare B cells and NK composition of the blood vs. RM. Wilcoxon test was used to compare MFI of markers between matched blood and RM subsets. HSNE diagrams were generated with Cytosplore 28 . We also conducted sensitivity analyses to examine the influence of sexual behavior on our results. Of note, there was no statistical difference between HIV replication (Mann-Whitney, p > 0.05), MZB (p > 0.05), or NK (p > 0.05) when data were stratified between men who reported receptive anal intercourse and men who did not. Therefore, data for the combined total cohort are reported in this manuscript.