T cell phenotypes in peripheral blood associated with MSM and HIV

Previous studies have reported higher blood CD4+ and CD8+ T cell activation and higher levels of inflammatory cytokine producing colonic CD8+ T cells in HIV-negative MSM compared with HIV-negative MSW [3, 5, 6]. To confirm these findings in an independent cohort, we analyzed blood T cell phenotypes in 18 HIV-negative MSW and 19 HIV-negative MSM, and included 13 HIV-positive untreated MSM as controls since immune activation is well known to be increased with HIV infection [13, 22, 23] (S1A Table). Increased frequencies of CD38 HLADR expressing CD8+ T cells, but not CD4+ T cells, were found in the blood of HIV-negative MSM compared with HIV-negative MSW (Fig 1A–1C). As expected, HIV-positive MSM had the highest levels of blood T cell activation. Expression of CD69 on peripheral blood T cells was also measured (S1A and S1B Fig), and frequencies of CD69+ CD4+ T cells trended higher for HIV-negative MSM and HIV-positive MSM than HIV-negative MSW, though this did not reach statistical significance. Stimulated T cells were analyzed for intracellular TNF-α and IFN-γ expression, and significantly higher frequencies of inflammatory cytokine producing T cells were observed in HIV-positive, but not HIV-negative, MSM when compared with HIV-negative MSW (S1C and S1D Fig). Finally, a lower CD4/CD8 T cell ratio in the blood has been previously reported for high-risk HIV-negative MSM [6], but was not reproduced in our cohort (S1F Fig). The CD4/CD8 T cell ratio was expectedly decreased in HIV-positive MSM who were not receiving antiretroviral treatment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. HIV-negative and HIV-positive untreated MSM have higher frequencies of activated and mucosal homing T cells in the blood.

PBMC from HIV-negative MSW (n = 18), HIV-negative MSM (n = 19), and HIV-positive MSM (n = 13) were stained for T cell activation and mucosal homing markers and analyzed by flow cytometry. (A) T cell activation gating strategy. (B–E) Comparison of frequencies of CD38+ HLADR+ (B) CD8+ and (C) CD4+ T cells, and frequencies of CD103+ (D) CD8+ and (E) CD4+ T cells between HIV-negative MSW, HIV-negative MSM, and HIV-positive MSM. Each data point represents an individual, and lines represent median values. Statistical analyses were performed using t-tests to compare groups if data from both groups had normal distributions and Mann-Whitney tests to compare groups if data from at least one group had a non-parametric distribution. **** = p <0.0001, *** = p<0.001, ** = p<0.01, * = p<0.05. (F–G) Spearman correlations between (F) CD8+ T cell activation, plotted on a log scale axis but correlated using linear data, and CD103+ CD8+ T cell frequency, and between (G) CD103+ CD4+ T cell frequency and CCR5+ CD4+ T cell frequency. Dotted lines represent non-linear semi-log (F) and linear (G) regression analysis.

https://doi.org/10.1371/journal.ppat.1007611.g001

Gut mucosal T cell activation plays an important role in HIV replication and disease [24]. We therefore evaluated expression of the mucosal homing marker CD103 (also known as integrin αEβ7) [25] on blood T cells. Frequencies of CD103+ CD4+ T cells were significantly higher and frequencies of CD103+ CD8+ T cells trended higher in HIV-negative MSM compared with HIV-negative MSW (Fig 1D and 1E). CD103+ CD8+ T cell, but not CD4+ T cell frequencies were higher in HIV-positive MSM than in HIV-negative MSW. No significant differences were observed in frequencies of CD103+ T cells between HIV-negative and HIV-positive MSM. Frequencies of CD103+ CD8+ T cells significantly correlated with frequencies of CD38+ HLADR+ CD8+ T cells (Fig 1F), while expression of HIV coreceptor CCR5 [26]–which was not significantly different across groups (S1E Fig)–significantly correlated with frequencies of CD103+ CD4+ T cells (Fig 1G). Taken together, previous findings of higher immune activation in HIV-negative MSM compared with HIV-negative MSW are confirmed here in our cohort, and our results additionally show an association between blood T cell activation and mucosal homing in MSW and MSM.

Microbiome differences between MSW and MSM are transferred to mice through fecal transplant

Microbiome differences have been previously associated with MSM and HIV [2–4, 8], but the extent to which these differences drive immune activation in vivo is unclear. To directly assess the in vivo immunological effect of the gut microbiome from HIV-negative and HIV-positive MSM, we transplanted feces from human donors to gnotobiotic mice. Each mouse received a single gavage of feces from a unique human donor. Stool donors were 16 HIV-negative MSW, 19 HIV-negative MSM, and 12 HIV-positive ART-naïve MSM randomly selected from a previously described cohort [4]. Of these donors, 13 HIV-negative MSW, 17 HIV-negative MSM, and 9 HIV-positive MSM are represented in the above immune data (Fig 1, S1 Table). HIV-positive MSM donors had a median viral load of 101400 copies/ml with a median CD4 T cell count of 574 cells/μl (S1 Table). Each donor was tested in a single mouse recipient, with the exception of 4 HIV-negative MSW, 8 HIV-negative MSM, and 4 HIV-positive MSM which were randomly chosen to be tested in 2–3 replicate mice. A subset of donors was also tested for fecal bacterial load (8 HIV-negative MSW, 12 HIV-negative MSM, 11 HIV-positive MSM), which was not found to be significantly different across donor groups (Fig 2B), indicating that each recipient group did not receive a significantly higher or lower amount of bacteria in the fecal transplant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Microbiome composition differences between MSW and MSM are transferred to mouse recipients by fecal transplant.

(A) Gnotobiotic mice were gavaged with feces from HIV-negative MSW (n = 16), HIV-negative MSM (n = 19), and HIV-positive MSM (n = 12) donors and sacrificed 21 days post-gavage for immunological analysis and 16S rRNA gene sequencing. (B) Bacterial load in donor stool. Isolated stool bacteria from a subset of donors (8 HIV-negative MSW, 12 HIV-negative MSM, 11 HIV-positive MSM) were stained and counted by flow cytometry. Each data point represents an individual donor, and lines represent medians. Mann-Whitney tests were performed to compare groups, and no statistically significant differences were detected. (C) Unweighted UniFrac clustering of donor (large dots) and recipient (small dots) microbiome compositions. Green dots represent HIV-negative MSW donor/recipient pairs, red dots represent HIV-negative MSM donor/recipient pairs, and blue dots represent HIV-positive MSM donor/recipient pairs. Lines connect donors to recipients and recipient replicates when applicable. (D) Taxa bar chart showing relative abundance of bacterial families in each donor and recipient. Legend identifies the top 20 most abundant families across all samples. P. copri refers to mice colonized with pure culture of Prevotella copri, and B. uniformis refers to mice colonized with pure culture of Bacteroides uniformis. f__ = unclassified bacterial family.

https://doi.org/10.1371/journal.ppat.1007611.g002

We performed 16S rRNA gene sequencing on donor fecal samples and fecal pellets from mouse recipients at 7, 14, and 21 days post-gavage. Since there was little variation in composition of the mice over time (S2 Fig), we focused our analysis on the terminal timepoint. Consistent with previous reports [2, 4, 8], unweighted UniFrac analysis showed that microbiome composition in the human donors (large circles) clustered distinctly by MSM and not HIV status (Fig 2C). Although the overall composition shifted along PC1 following fecal transfer, the distinct clustering of MSW and MSM along the PC2 axis was replicated in the mouse recipients (small circles), which are connected to their respective donors by lines (Fig 2C). Six 16S rRNA sequence variants were identified to be significantly different (with an FDR corrected p-value<0.1) between HIV-negative MSW and HIV-negative MSM recipients, 3 of which increased (Desulfovibrio sp., Holdemanella biformis, Howardella ureilytica) and 3 of which decreased (Clostridium sp., Bacteroides uniformis, Flavonifractor sp.) with MSM (S2 Table). Five of six of these variants were also found to be significantly different between HIV-negative MSW and HIV-negative MSM donors (S2 Table). Holdemanella biformis (formerly Eubacterium biforme) and Bacteroides uniformis have also previously been described to be different with MSM [4, 8]. In contrast, no statistically significant differences (FDR p<0.1) in variant abundance were detected between HIV-negative and HIV-positive MSM donors or recipients, and previously reported bacteria differing with HIV in analyses of larger cohorts [2, 4] were not observed in this smaller donor sample size (S3 Table). These data suggest that MSM-associated microbiome differences are readily transferred to mice through fecal transplant, while transfer of subtler HIV-associated differences may require an increased sample of donors.

The natural gut microbiome in mice is compositionally different from that in humans [27], therefore changes in relative abundance of microbes less or better adapted to the mouse gut were expected to occur following fecal transfer. To characterize microbiome shifts following colonization of mice, we identified bacterial families that significantly increased or decreased in relative abundance from donors to recipients. When analyzing all groups together (HIV-negative MSW, HIV-negative MSM, HIV-positive MSM), 25 families significantly changed (FDR p<0.1) as determined by nonparametric t-test comparing donors to recipients (S4 Table). These consisted of 19 families that decreased and 7 families that increased in relative abundance in recipients compared with donors. When stratifying the analyses by group, differences between donors and recipients for 20/25 families maintained consistent trends within each group, and differences for 11/25 families maintained statistical significance within each group (S4 Table). A striking donor-to-recipient change was the relative abundance of Prevotellaceae, which was significantly higher in MSM compared with MSW (Fig 2C) [2–4], and decreased in abundance for all groups following transfer (S4 Table). Taxa bar charts showing donors and recipients grouped separately (Fig 2D), or donors and their respective recipients side-by-side (S3 Fig), demonstrate the overall loss of Prevotellaceae in mice. Though Prevotellaceae was undetectable in almost all mouse recipients of HIV-negative and HIV-positive MSM donors, it maintained a strong presence in the two mouse recipients of Prevotellaceae-rich HIV-negative MSW donors (Fig 2D, S3 Fig). One reason that there may be colonization differences in MSW versus MSM recipients is the presence of different species or strains of Prevotella. We thus sought to explore whether different sequence variants within the Prevotella genus showed different colonization patterns. In the two mice that Prevotella successfully colonized, we identified only three Prevotella sequence variants to be present and all were classified as Prevotella copri (S4A Fig). Two out of three P. copri variants were of the highest abundance in MSW donors (S4B Fig), and increased in relative abundance from donors to mice. The third variant was the dominant and most abundant Prevotella sequence in both MSW and MSM donors, but only colonized recipients of MSW. Six other P. copri sequence variants that were present in at least 20% of MSM donors were identified, and only three of these could be found in Prevotella-rich MSW. None of these other six sequence variants were detectable in mouse recipients (S4 Fig). To investigate the ability of Prevotella to colonize mice in the absence of competing microbes, we gavaged gnotobiotic mice (n = 3) with a monoculture of P. copri (DSMZ 18205) or with a monoculture B. uniformis (ATCC 8492) for comparison (Fig 2D). By 16S gene sequence, the P. copri monoculture was determined to be the same variant as P. copri 3 found in fecal donors and their recipients. After 21 days of colonization, sequencing showed that P. copri was still present in feces of 2/3 mice, while B. uniformis was present in 3/3 mice (Fig 2D). Thus, three P. copri variants were able to colonize mice, with two of these variants being at a higher relative abundance in Prevotella-rich MSW than in MSM. The third variant, which was the dominant variant in both MSW and MSM, could only successfully colonize mice in pure culture or in the context of an MSW microbiome.

Since MSM have more Prevotella than MSW, we next investigated whether the inability of Prevotella to colonize mouse recipients of MSM resulted in a larger magnitude donor-to-recipient compositional shift for MSM. Unweighted Unifrac distance, a measurement of relative compositional difference, between each donor and their recipient was calculated. For donors with replicate mouse recipients, a representative mouse was randomly selected. Donor-recipient distances were not found to be significantly different across groups, indicating the overall composition of each donor group was altered on average by the same magnitude after transfer (S5A Fig). The percentages of unique sequence variants identified in each donor that were present in their mouse recipient were also not significantly different across groups, indicating equal colonization fidelity (S5B Fig). Despite the loss of taxa that differed with MSM following transfer, differences in specific microbes were recapitulated in mouse recipients to maintain distinct MSW and MSM compositions. Any immunological differences between recipient groups would therefore be associated with these compositional differences.

Immune activation and barrier integrity in mouse recipients

To investigate the immunological effect of the MSM and HIV-associated microbiome on mouse recipients, mice were sacrificed 21 days post gavage, and ileum, colon, and mesenteric lymph nodes were evaluated for markers of T cell activation. Due to lower cell recovery, only colons from a subset of mice (8 HIV-negative MSW, 10 HIV-negative MSM, 9 HIV-positive MSM) yielded enough cells for analysis by flow cytometry. Ileum tissues from all mice were analyzed and are highly relevant to HIV infection due to the small intestine’s abundance of Th17 cells [28], which are targets of HIV infection and play an important role in disease [18]. CD69 and CD103 were used to measure gut T cell activation, and though lowly expressed in the blood [29], these are relevant markers in the gut, and are used to define tissue-resident memory T cells retained in gut tissue [30, 31]. For stools that were tested in replicate mice, immune data from one replicate was used for each stool. In the ileum, frequencies of CD69+ CD8+ T cells were found to be significantly higher in mouse recipients of HIV-negative MSM compared with mouse recipients of HIV-negative MSW (Fig 3B). Frequencies of CD103+ T cells were significantly elevated for both CD4+ and CD8+ T cells in the ileum of recipients of HIV-negative MSM (Fig 3C and 3D). Interestingly, ileum frequencies of CD69+ or CD103+ T cells were not significantly different between recipients of HIV-negative and HIV-positive MSM, and only CD103+ CD4+ T cells were significantly higher in recipients of HIV-positive MSM compared with recipients of HIV-negative MSW (Fig 3D). Finally, stool recipients that were successfully colonized with Prevotella did not display noticeably different levels of immune activation compared to the MSW group median (S7B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Gut T cell activation and homing in mouse recipients.

Ileum and colon tissues were collected from mouse recipients at 21 days post gavage and analyzed for frequencies of CD69+ and CD103+ T cells. (A) Gating strategy for CD69+ T cells. Representative plots from ileum tissues are shown. (B-C) Frequencies of (B) CD69+ CD8+ T cells, (C) CD103+ CD8+ T cells, and (D) CD103+ CD4+ T cells in the ileum. (E) Frequencies of CD69+ CD4+ T cells in the colon. Each data point represents a single mouse gavaged with an individual donor’s feces, and a representative mouse was used for stools tested in replicate mice. Lines represent medians. Statistical analyses were performed using t-tests to compare groups if data from both groups had normal distributions and Mann-Whitney tests to compare groups if data from at least one group had a non-parametric distribution. ** = p<0.01, * = p<0.05.

https://doi.org/10.1371/journal.ppat.1007611.g003

In colon tissue, frequencies of CD69+ CD4+ T cells were significantly higher for recipients of HIV-negative MSM than recipients of HIV-negative MSW (Fig 3E). Recipients of HIV-positive MSM did not statistically differ from recipients of HIV-negative MSM or HIV-negative MSW in colonic T cell activation, but did trend higher for frequencies of CD69+ CD8+ and CD4+ T cells, and CD103+ CD8+ T cells, compared with either HIV-negative group (Fig 3E, S6 Fig). Statistical power in detecting differences in the colon may have been limited by a small sample size and a high variance of immune measures. We therefore examined mesenteric lymph nodes (mLN), which were analyzed for all mouse recipients, as a proxy of gut immune phenotypes since the mLN drain ileum and colon tissues [32]. In the mLN, mouse recipients of HIV-positive MSM displayed significantly higher frequencies of effector memory (T_em) CD4+ T cells (CD62L-, CD44+) than recipients of HIV-negative MSW (Fig 4A). Furthermore, frequencies of CD4+ T_em in the mLN significantly correlated with frequencies of CD69+ CD4+ T cells in the colon (Fig 4B), indicating mLN phenotypes were representative of immune readouts in colon tissue. In mLN of monocolonized mice, which were also examined at 21 days post gavage, CD4+ T_em frequencies were within the range detected in stool recipients, and did not differ between mice gavaged with B. uniformis and mice gavaged with P. copri (S7A Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Frequencies of effector memory CD4+ T cells in the mesenteric lymph nodes of mouse recipients.

mLN from mouse recipients were collected and analyzed for effector memory CD4+ T cell frequency (CD44+ CD62L-) using flow cytometry. (A) Comparison of frequencies of CD44+ CD62- CD4+ T cells in the mLN across recipient groups. (B) Spearman correlations of effector memory CD4+ T cell (T_em) frequencies in the mLN with frequencies of CD69+ CD4+ T cells in the colon. Each data point represents a single mouse gavaged with an individual donor’s feces, and a representative mouse was used for stools tested in replicate mice. Lines represent medians. Statistical analyses were performed using t-tests to compare groups if data from both groups had normal distributions and Mann-Whitney tests to compare groups if data from at least one group had a non-parametric distribution. * = p<0.05.

https://doi.org/10.1371/journal.ppat.1007611.g004

Overall, mouse recipients displayed a significant MSM-associated effect in immune activation in the gut. Recipients of MSM also tended to have higher variation in T cell measures, with CD103+ CD4+ T cell frequencies in the ileum (Fig 3D) and CD69+ CD4+ T cell frequencies in the colon (Fig 3E) being significantly different in variance as determined by F test between HIV-negative MSM and HIV-negative MSW. Despite the lack of a significant HIV-associated effect, immunological differences between MSW and MSM recipients were enough to drive significant immune correlations between donors and recipients, including a correlation between frequencies of recipient ileum CD69+ CD8+ T cells and frequencies of donor blood CD38+ HLADR+ CD8+ T cells (S8A Fig), and a correlation between frequencies of recipient colon CD69+ CD4+ T cells and frequencies of HIV-negative donor blood CD103+ CD4+ T cells (S8B Fig). These correlations highlight immunological similarities between donors and their recipients.

Elevated markers of gut mucosal damage have been reported for MSM and correlated with MSM-associated microbiota [3, 6], and are well known to be associated with HIV infection [33]. To determine if bacteria from the fecal transfer directly caused intestinal damage, ileum and colon tissue samples from 5 HIV-negative MSW, 7 HIV-negative MSM, and 6 HIV-positive MSM mouse recipients were analyzed for histopathology. Tissue health and epithelial integrity was determined to be within normal limits for all mice examined (Fig 5A). Each tissue analyzed was graded with an overall inflammation score on a scale of 1–4 [34], and all tissues were assigned the same minimum inflammation rating of 1. In support of this, myeloperoxidase concentrations (measured in 12 HIV-negative MSW, 11 HIV-negative MSM, and 11 HIV-positive MSM recipients) were not different across groups in the colon (Fig 5B). Levels of soluble CD14 (sCD14) in the blood, a marker of monocyte activation in response to LPS [35] and bacterial translocation across the gut barrier [33], was measured in 11 HIV-negative MSW, 15 HIV-negative MSM, and 9 HIV-positive MSM recipients, and was also not found to be significantly different across groups (Fig 5C). Lack of signs of barrier damage suggest that transfer of microbiota from HIV-negative or HIV-positive men were not sufficient to cause gut tissue injury in mice (at least within 3 weeks of colonization), and that barrier breakdown was not necessary for microbiota from MSM to promote gut T cell activation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Absence of inflammation, epithelial damage, and bacterial translocation in mouse recipients.

(A) Formalin-fixed ileum and colon sections from mouse recipients (5 HIV-negative MSW, 7 HIV-negative MSM, 6 HIV-positive MSM) were paraffin-embedded and H&E stained, and then analyzed by a blinded pathologist for inflammation and epithelial integrity. Representative images of ileum and colon tissue are shown. Arrows point to the epithelium. All tissues analyzed were scored with the minimal inflammation rating of 1, on a scale of 1–4. (B) Myeloperoxidase concentrations measured by ELISA in colon tissues (12 HIV-negative MSW, 11 HIV-negative MSM, 11 HIV-positive MSM). (C) sCD14 levels measured by ELISA in plasma (11 HIV-negative MSW, 15 HIV-negative MSM, 9 HIV-positive MSM). For (B) and (C), each data point represents an individual mouse, lines represent medians, dotted lines represent the limit of detection. Statistical analyses were performed using t-tests to compare groups if data from both groups had normal distributions and Mann-Whitney tests to compare groups if data from at least one group had a non-parametric distribution. No statistically significant differences were detected.

https://doi.org/10.1371/journal.ppat.1007611.g005

Recapitulation of microbial and immune differences in mouse recipients supports a link between microbiome composition and immune activation in MSM. Microbial-immune correlations consistent across donors and recipients would further strengthen this link. We therefore correlated the relative abundance of gut microbes in donors and recipients with their respective immune data (S5 and S6 Tables). Only immune data significantly different with MSM in either donors or recipients were selected for analyses, since only MSM-associated microbiome differences were detectable in this cohort. Frequencies of blood CD103+ CD8+ T cells in donors, though not significantly different with MSM, were additionally selected for analysis due to their relevance to the gut. Correlations were computed with both the full donor group, as well as with HIV-negative donors only, since HIV may independently drive immune activation in HIV-positive MSM and confound microbial correlations. All correlations that were and were not significant in donors and in mice are shown in tables S5 (donors) and S6 (recipients). After comparing significant correlations across these datasets, we identified 8 microbes that correlated with at least one donor and one recipient immune measurement (Table 1), with 3 of these being negative and 5 of these being positive correlates. Out of these 8 microbes, 5 significantly correlated using the full donor cohort, while 3 significantly correlated using only HIV-negative donors. Two microbes, Bacteroides uniformis and Howardella ureilytica, significantly differed in relative abundance with MSM in both donors and recipients (S2 Table), and also significantly correlated with immune measurements in both donors and recipients. Though few of the shared microbe-immune correlates were statistically significant with FDR correction, consistency of correlations across humans and mice suggest biological significance, and provide further evidence that immune activation in MSM is influenced by the gut microbiome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Microbial correlates with donor/recipient T cell activation and gut homing, and in vitro HIV infection.

https://doi.org/10.1371/journal.ppat.1007611.t001

Microbiota from HIV-negative and HIV-positive MSM stimulate in vitro HIV infection

T cell activation has been correlated with HIV viral load [13], and associated with HIV transmission in women [36]. We therefore investigated whether microbiota from MSM could promote HIV infection in vitro. Fecal bacterial communities (FBCs) were isolated from stools of 12 HIV-negative MSW, 15 HIV-negative MSM, and 9 HIV-positive MSM by separating whole intact bacteria in stool from debris using Histodenz columns as previously described [8]. All donors of stool used for FBC isolation are represented in the PBMC immune phenotyping data (Fig 1), while FBCs from 7 HIV-negative MSW, 15 HIV-negative MSM, and 5 HIV-positive MSM were also used as donors in the fecal transplant experiments. HIV-positive individuals used for FBC isolation had a median viral load of 47650 copies/ml and a median CD4 T cell count of 577 cells/μl (S1 Table). Primary human lamina propria cells (LPCs) were stimulated by isolated FBCs, infected with HIV_bal, and intracellular HIVgag antigen expression was measured after five days. We found that LPCs stimulated with FBCs from HIV-negative and HIV-positive MSM had higher frequencies of HIV-infected cells compared with LPCs stimulated with microbiota from HIV-negative MSW (Fig 6A and 6B). LPCs stimulated with FBCs from HIV-negative and HIV-positive MSM displayed significantly elevated CD8+ T cell activation at the end of infection (Fig 6C), and frequencies of HIV-infected cells significantly correlated with both CD4+ and CD8+ T cell activation of stimulated LPCs (Fig 6D and 6E). FBCs from 7 HIV-negative MSW, 11 HIV-negative MSM, and 8 HIV-positive MSM used in this infection assay were also previously used to stimulate PBMC in vitro [8], and frequencies of HIV-infected cells induced by these FBCs significantly correlated with activation levels of peripheral blood CD4+ and CD8+ T cells stimulated by these same FBCs found previously (S8B and S8C Fig). 16S rRNA gene sequencing of FBCs showed that the compositional distinction between MSW and MSM was maintained following bacterial isolation (S9A and S9C Fig). Unweighted UniFrac distances between donor samples and their isolated FBCs were not significantly different across groups (S9B Fig), indicating the average magnitude of stool-FBC compositional shifts were consistent between groups. Specific bacteria that differed with MSM and with HIV in FBCs, and bacteria that significantly changed from stool to FBCs, were previously described [8], and are not presented here. Infection and T cell activation levels trended higher for cells stimulated with HIV-positive MSM FBCs than cells stimulated with HIV-negative MSM FBCs, though these differences were not statistically significant. Finally, six microbes that significantly correlated with in vitro HIV infection also correlated with fecal transplant donor and/or recipient immune measurements (Table 1). Interestingly, two sequence variants of H. biformis correlated with in vitro infection, with one significantly correlating with donor blood T cell activation and the other significantly correlating with recipient gut T cell activation. Significant and non-significant correlations between all 16S sequence variants in FBCs with in vitro HIV infection are shown in S7 Table. Consistencies were therefore observed between the in vivo transplant system and the in vitro assays. These data support the findings from the mouse model, and suggest that the gut microbiome in HIV-negative and HIV-positive MSM could impact HIV infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Fecal bacterial communities isolated from HIV-negative and HIV-positive MSM promote HIV infection of primary human lamina propria cells.

Primary lamina propria cells (LPCs) were isolated from human jejunum tissue, stimulated with isolated fecal bacterial communities (FBCs) from HIV-negative MSW (n = 12), HIV-negative MSM (n = 15), and HIV-positive MSM (n = 9) donor feces, and infected with HIV_bal. LPCs were stained for intracellular HIV-1gag expression and T cell activation markers at 5 days post-infection using flow cytometry. (A) Gating strategy for HIVgag+ T cells. (B–C) Comparison of levels of (B) HIV-1gag+ cells and (C) CD38+ HLADR+ CD8+ T cells following infection. Each data point represents the median value of one FBC tested across 3 different individuals’ LPCs. Lines represent medians within groups. Statistical analyses were performed using t-tests to compare groups if data from both groups had normal distributions and Mann-Whitney tests to compare groups if data from at least one group had a non-parametric distribution. * = p<0.05. (D–E) Spearman correlations between frequencies of HIV-1gag+ cells and frequencies of (D) CD38+ HLADR+ CD4+ T cells, and (E) CD38+ HLADR+ CD8+ T cells. Dotted lines represent linear regression analyses.

https://doi.org/10.1371/journal.ppat.1007611.g006

Study subjects

Stool samples were obtained from HIV-negative MSW, high-risk HIV-negative MSM, and HIV-positive MSM not on antiretroviral therapy. Male HIV-positive individuals enrolled in the study were determined to be MSM using a behavioral questionnaire. High-risk MSM were recruited from a high-risk cohort assembled for a study of a candidate HIV-1 preventative vaccine [48]. Designation of high-risk was according to a number of different behaviors including 1) a history of unprotected anal intercourse with one or more male or male-to-female transgender partners 2) anal intercourse with two or more male or male-to-female transgender partners and 3) being in a sexual relationship with a person who has been diagnosed with HIV. All enrolled individuals live in metropolitan Denver. Other inclusion criteria used were: 18–70 years old, body mass index (BMI) between 21–29 mg/kg² and weight stable for at least 3 months; for HIV-positive individuals, <10 days of ART treatment at any time, or previously on ART but off treatment for the previous 6 months, prior to stool and blood collection; liver and kidney function tests within normal range. Exclusion criteria used were: weight <110 pounds; received antibiotics within the prior 90 days; active gastrointestinal disease; history of bowel resection. All subjects used in this study were previously characterized for diet and sexual behavior as part of a larger cohort; diet or any particular sexual behavior that we measured were not identified as driving factors of the most prominent microbiome differences between MSM and MSW [4].

Ethics statement

Written informed consent was obtained from healthy HIV-negative and HIV-positive individuals for use of stool. The study protocol was approved by the Colorado Multiple Institution Review Board (COMIRB No: 14–1595). All subjects were adults. Mice were handled in accordance with the recommendations in the National Institutes of Health Guide for the Care and Use of Laboratory Animals and protocols were approved by the University of Colorado Institutional Animal Care and Use Committee Permit Number 00097. Primary human lamina propria cells were collected from otherwise discarded tissues from gut resection surgery, and was determined to not be human subject research under protocol number 14–1184 approved by the University of Colorado Multiple Institutional Review Board. Tissues were obtained from anonymous adult patients who were not asked for consent.

PBMC immune phenotyping

PBMC were isolated from patient blood by Ficoll-Paque (GE), and cryopreserved and stored in liquid nitrogen before analysis. PBMC were thawed, and 5 x 10⁵ cells were stained for 30 min at 4° C with the following antibodies: CD3-APC-Cy7 (OKT3, Biolegend), CD4-PerCP-Cy5.5 (OKT4, Biolegend), CD8-APC (SK1, Biolegend), CD38-BV421 (HIT2, Biolegend), CD69-BV510 (FN50, Biolegend), HLADR-FITC (L243, Biolegend), CD103-PE (BerACT8, Biolegend), CCR5-PE-Cy7 (J418F2, Biolegend). Cells were then washed in FACS buffer and fixed in 1% paraformaldehyde before being acquired on a BD FACS Canto II. 1 x 10⁶ PBMC were stimulated with a PMA/ionomycin cell stimulation cocktail (eBioscience), and treated with a protein transport inhibitor cocktail (eBioscience) for 4 h at 37° C, 5% CO₂. Cells were then stained for 30 min at 4° C with the following antibodies: CD3-APC-Cy7 (OKT3, Biolegend), CD4-PerCP-Cy5.5 (OKT4, Biolegend), CD8-APC (SK1, Biolegend). Cells were washed and treated with fix/perm buffer (eBioscience) for 30 min at 4° C, before being stained for intracellular cytokines for 30 min at 4° C with the following antibodies: IFN-γ-PE (4SB3, Biolegend), TNFα-APC-Cy7 (MAb11, Biolegend). Cells were then washed in FACS buffer and fixed in 1% paraformaldehyde before being acquired on a BD FACS Canto II.

Gavages

1.5 grams of frozen feces stored at -80° C for up to one year from each donor was thawed in an anaerobic chamber and homogenized in 3 ml of anaerobic PBS, using a syringe handle. Fecal solutions were then filtered through a 100 μm nylon filter into a 50 ml conical tube. Tubes were sealed before removing from the anaerobic chamber and transferred to the mouse facility within 1 hour of thawing. 200 μl of each fecal solution was used to gavage each mouse.

Bacterial load in donor stool

2 mg of stool from donors were homogenized in PBS, and bacteria were then isolated using two sequential rounds of sucrose density centrifugation. Isolated bacterial cells were stained with thiazol orange and propidium iodide, and acquired by a FACS Canto II (BD) flow cytometer in the presence of counting beads. Data were processed using Flowjo software, and the concentration of live bacterial cells per mg of stool was calculated.

Gnotobiotic mice

Germ-free C57/BL6 mice were purchased from Taconic and bred and maintained in flexible film isolator bubbles. Male mice between 6–8 weeks of age were gavaged with fecal solutions prepared from donor fecal samples and housed individually (to ensure mice would not contaminate each other) following gavage for 3 weeks in a Tecniplast iso-positive caging system, with each cage having HEPA filters and positive pressurization for bioexclusion. Feces were collected from mice at day 0 for 16S rRNA gene sequencing to confirm germ-free status. 16S sequencing of feces from day 7 and day 14 showed relatively little variation in composition from day 21 (S2 Fig). Mice were euthanized at 21 days post gavage using isoflurane overdose and all efforts were made to minimize suffering. Blood from euthanized animals was collected using cardiac puncture and cells were pelleted in K2-EDTA tubes; plasma was then aliquoted and stored at -80° C. Data was combined from 10 batches of mice analyzed independently, with at least 4 mice per batch, and at least two donor groups represented in each batch.

Mouse ileum and colon single cell preparation

Ileum and colon tissues from mice at 21 days post gavage were collected for immune phenotyping and histology. Tissues were dissected length-wise and feces were washed away using PBS. 1–5 mm slices of ileum and colon closest to the cecum were taken from a subset of mice and fixed in 10% formalin and reserved for histology or ELISA. The remaining tissue was then washed in a PBS solution containing 1% EDTA for two 5 min intervals on a vortexer. Tissues were washed with PBS and strained through a 100 μm nylon filter in between intervals. Tissues were then washed for a third time in plain 1X PBS. Tissues were then minced using an octoMACs tissue dissociator (Miltenyi) and digested in 2.5 ml of a solution of complete RPMI (10% FBS, 1% PSG) containing 1 mg/ml collagenase D type I (Worthington Biotech) at 37° C, 5% CO₂. Released cells were quenched with cold complete media, filtered through a 70 μm nylon filter and washed with complete media before staining for flow cytometry.

Mouse immune phenotyping

0.5–1 x 10⁶ ileum and colon cells from each mouse were stained in FACS buffer for 30 min at 4° C with the following antibodies: CD3-PerCP-Cy5.5 (17A2, Biolegend), CD4-FITC (RM4-4, eBioscience), CD8-BV510 (53–6.7, Biolegend), CD69-PE-Cy7 (h1.2F3, eBioscience), CD103-PE-Dazzle-594 (2E7, Biolegend). Mesenteric lymph node cells were additionally stained with CD44-AlexaFluor-700 (IM7, eBioscience), and CD62L-APC-eFluor780 (MEL-14, eBioscience). Cells were then washed with FACS buffer and fixed in 1% paraformaldehyde before being acquired on a BD LSRii flow cytometer. Colon samples that had less than 1,000 T cells were excluded. Plasma samples were thawed and diluted 1:10 before being assayed for sCD14 by ELISA (R&D Systems) according to the manufacturer’s instructions. Ileum and colon tissues cultured for 24 h at 37° C, 5% CO₂ in complete RPMI were digested with CellLytic MT solution with proteinase inhibitor (SIGMA), and assayed for myeloperoxidase concentrations by ELISA (R&D Systems). ELISA plates were read using a Vmax Kinetic plate reader (Molecular Devices). Tissue sections from the distal end of the ileum and proximal end of the colon were processed for histology and H&E stained. Scoring [34] was performed by a pathologist blinded to experimental conditions.

Stool collection and FBC preparation

Stool samples were collected by the patient, both on a sterile swab and with a sterile scoop within 48 hours of a clinic visit. Samples were stored immediately in a cooler with -20°C freezer packs. After delivery to the clinic, the swab was subsequently stored at -80° C to await DNA extraction. Isolation of fecal bacterial communites (FBCs) from stool was described previously [8]. Briefly, two grams of feces were homogenized in sterile PBS and filtered through a 100uM filter. Homogenate was then subjected to two density gradient centrifugations with 80% Histodenz (Sigma). Bacterial layers were collected after each spin, visualized with a light microscope, and quantified by flow cytometry using a bead counting kit (BD Biosciences). FBCs were stored at -80° C before use in in vitro assays.

16S rRNA gene sequencing

16S rRNA targeted sequencing was conducted according to earth microbiome project standard protocols (http://www.earthmicrobiome.org). DNA was extracted from donor stool swabs, mouse fecal pellets, and from a 250 μL aliquot of FBC isolates using the standard PowerSoil protocol (QIAGEN). PCR amplification of the extracted DNA, along with water controls, was conducted with barcoded primers targeting the V4 region of the 16S rDNA gene (515F-806R; FWD:GTGCCAGCMGCCGCGGTAA; REV:GGACTACHVGGGTWTCTAAT). Amplified DNA was quantified using a PicoGreen assay (Invitrogen) so equal amounts of DNA from each sample could be pooled and cleaned using the UltraClean PCR Clean-up protocol (Qiagen). The final DNA pool was sequenced using the Illumina MiSeq platform (San Diego, CA) using the V2 2x250 kit.

Sequencing analysis

Raw sequences were assigned to samples based on their barcodes using QIIME 2.6 [49]. The libraries were denoised and grouped by sequence variants using dada2 1.2.2. Samples contained from 21,357 to 159,296 sequences; analyses that contained donor and mouse recipient feces were conducted on the standardized sequence number of 21,357, analyses that contained FBC isolates and their original stool were conducted on the standardized sequence number of 21,812, and analyses that contained mice monocolonized with P. copri and B. uniformis were conducted on the standardized sequence number of 31,812 –these numbers were determined as the minimum read number acquired for a single sample in each independent analysis. Sequence variants were classified taxonomically using the RDP classifier [50] trained on the greengenes 3_8 taxonomic database [51]. Principal Coordinates Analysis of unweighted UniFrac distance matrices, and rendering of taxa bar charts, were conducted using QIIME 2.6. For immune correlations, comparisons of relative abundance between groups, and P. copri variant analysis, sequence variants not observed in at least 20% of the samples in each analysis were removed.

LPC collection

Macroscopically healthy jejunum tissues from patients undergoing elective gut resection surgery were digested to release lamina propria cells (LPCs). Briefly, tissues were trimmed of fat and the muscularis layer was removed using scissors. Trimmed tissues were incubated on a rocker with a 1.6 mM dithiolthreitol (DTT) PBS solution for 45 min at 37° C, and then with a 1 mM EDTA PBS solution for 60 min at 37° C. Tissues were then washed with PBS on a vortexer, minced using scissors, and incubated for three 45 min intervals with a digestion solution of complete RPMI (10% HS, 1% PSG) containing 1 mg/ml collagenase D (Sigma-Aldrich) and 10 μg/ml DNase I (Sigma-Aldrich). Released cells were collected and filtered through a 70 μm nylon filter after each digestion interval. Cells from each interval were pooled together, RBC lysed, and cryopreserved.

In vitro LPC infection

LPCs were thawed in complete RPMI containing 10 μg/ml DNase I (Sigma-Aldrich), and washed again with complete RPMI before plating. 5 x 10⁵ LPCs were plated per well in a 96-well round bottom plate. Bacterial cells were added to specified wells at 5:1 bacteria:LPC ratio. HIV_bal passaged in Molt4 T cells and quantified by qPCR were then added to each well at 10⁷ HIV RNA copies/well. Cells were incubated at 37° C for 24 h to allow for infection, and then washed twice with complete RPMI to remove cell-free virus. Bacteria were added back to each well at a 5:1 ratio after washing, and cells were incubated for another 4 days at 37° C. LPCs were then stained with anti-CD3-APC-Cy7, CD4-PerCP-Cy5.5, CD8-APC, HLADR-FITC, CD38-BV421, CD69-BV510, and CCR5-PE-Cy7. Cells were then washed and treated with fix/perm buffer (eBioscience) for 30 min at 4° C, and stained with anti-HIVgag-PE (KC57, Fisher) for 1 h at 4° C. Cells were then fixed in 1% paraformaldehyde and acquired using a FACS Canto II flow cytometer (BD Biosciences). Unstimulated, uninfected cells were used as controls to gate on HIVgag+ cells.

Statistics

Immune data were compared between groups using t-tests if data from both groups passed the D’Agostino and Pearson normality test, and Mann-Whitney tests if data from at least one group did not pass the normality test. All immune data were analyzed using Prism 7 (Graphpad). Comparisons of bacterial relative abundances between groups were performed using non-parametric t-tests on non-transformed data. Spearman correlations were used for all correlation analyses. Replicate recipient mice were treated as independent observations for all microbiome comparisons and correlations. Corrected p-values were determined with the False Discovery Rate (FDR) technique of Benjamini and Hochberg. All statistical analyses involving 16S sequencing data were performed using QIIME 1.9.