Gut Microbiota Linked to Sexual Preference and HIV Infection

1The main clinical problems of people living with HIV (PLWH) in areas with adequate healthcare standards and continued antiretroviral therapy (ART) supply are increasingly related to premature aging. That is, a precocious development of type 2 diabetes, dislipidemia, cardiovascular diseases, osteoporosis and frailty syndrome. Such diseases have been related to structural or metabolic perturbations in the gut microbiota of non-HIV-infected subjects whereas, in PLWH, have been linked to chronic inflammation, immune activation and endotoxemia. Thus there is considerable interest in understanding the role of the human gut microbiome in HIV pathogenesis and, in particular, its ability to perpetuate chronic inflammation and foster immune senescence. This has immediate clinical implications because, in theory, it might be possible to gear the gut microbiota towards “healthier” equilibrium states with the host, which might allow, for example, to achieve faster immune reconstitution, improve vaccine responses or reduce HIV reservoirs.

2However, although expectations are high, the HIV microbiome science is still at its early stages, and much remains to be known. Simple questions such as whether there is a consistent HIV-specific dysbiosis pattern, or which factors are relevant in shaping the microbiome in PLWH remain unanswered. Initial cross-sectional studies in humans have provided contradictory associations between microbial richness and HIV serostatus, and suggested shifts from Bacteroides to Prevotella predominance following HIV-1 infection. Such shifts, however, have neither been found in animal models nor in studies matching for HIV-1 risk groups. Conversely, large international studies in healthy populations have shown that at least in resource-rich countries, the gut microbiome forms a composition landscape with density peaks that can stratify the human population into enterotypes dominated by Bacteroides, Prevotella and Ruminococcus, respectively. The origin and clinical significance of such enterotypes is uncertain, but they have been linked to genetic, as well as to lifestyle and environmental factors, including long-term dietary patterns and exercise. Thereby, associations between Prevotella or Bacteroides and HIV infection might be easily confounded by other factors. Obtaining reliable information at this level is critical to advance our understanding of HIV pathogenesis, as well as to define the specific targets of novel therapeutic interventions on the human gut microbiome.

3This was a cross-sectional study in two independent European cohorts of HIV-1-infected subjects and HIV-negative controls. The study included one test cohort, one internal validation cohort and one external validation cohort (Supplementary Fig. 1).

4The test cohort (BCN0) was enrolled in Barcelona, Catalonia, Spain, between January and December 2014. HIV-1 infected patients were recruited from HIV Clinics at the University Hospitals Germans Trias i Pujol and Vall d'Hebrón. HIV-1-negative controls were mainly recruited from an ongoing prospective cohort of HIV-negative MSM at risk of becoming infected by HIV-1, who attend quarterly medical and counseling visits including HIV-1 testing (Alere Determine™ HIV-1/2 Ag/Ab Combo, Orlando, FL) at a community-based center for MSM in Barcelona. Additional controls were HIV-1-negative partners from HIV-1-infected subjects attending the HIV clinics.

5The inclusion criteria were: age within 18 and 60 years and body mass index (BMI) within 18.5 and 30. Exclusion criteria were: (a) any gross dietary deviation from a regular diet, or any specific regular diet, i.e., vegetarian, low-carb, etc.; (b) antibiotic use during the previous 3 months (with the exception of late presenters, who could receive antibiotics to treat opportunistic infections); (c) pregnancy or willingness to become pregnant; (d) current drug consumption or alcohol abuse; (e) any chronic digestive disease such as peptic ulcer, Crohn's disease, ulcerative colitis or coeliac disease; (f) any surgical resection of the intestines except for appendectomy; (g) any autoimmune disease; and (h) any symptomatic chronic liver disease or presence of hepatic insufficiency defined as a Child–Pugh C score. In addition, HIV-infected subjects were classified as elite controllers, viremic controllers, ART-naïve, early treated, late presenters, immune concordant or immune discordant (Supplementary methods).

6The internal validation cohort (cohort BCN1) included individuals from BCN0 who provided a second fecal sample one month later.

7Observations in Barcelona were externally validated in an independent observational cohort recruited at the HIV outpatient clinic, Karolinska University Hospital, Stockholm, Sweden (cohort STK). All HIV-1-infected patients in cohort STK were at least 18 years old, had been diagnosed with HIV-1 between one and 25 years earlier and were ART-naïve at the time of fecal sampling. Controls were healthy HIV-1-negative individuals matched by sex and age. Neither patients nor controls had been prescribed antibiotics or probiotics, or had had infectious diarrhea during the preceding two months.

8Clinical and laboratory data from BCN0 and BCN1 were collected in a centralized database specifically designed for this study (OpenClinica™, © 2015 OpenClinica, LLC). The clinical evaluation was performed following a standardized questionnaire including: a checklist for fulfillment of inclusion and exclusion criteria, anthropometric data, age at study entry, age at HIV diagnosis, gender, ethnicity, city of residence, HIV risk group, history of allergies, antibiotic intake between 3 and 6 months before inclusion, frequency and consistency of feces, history of medical or surgical problems or interventions, present and previous ART, history of AIDS- and non-AIDS-related diseases, nadir and most recent CD4 + T-cell counts, HIV-1 RNA levels, history of sexually transmitted diseases and infection by the human papillomavirus (HPV), hepatitis B (HBV) or hepatitis C (HCV).

9HIV-1 risk categories in our study were mutually excluding: male study participants who reported being MSM or referred insertive or receptive anal intercourse with other men were included in the MSM category, even if they also reported intravenous drug use or sex with women. Females and males not included in the MSM category reporting past intravenous drug use were classified as PWID. Heterosexual males or females not included in any of the previous 2 categories were classified as HTS. None of our study participants belonged to any other HIV-1 transmission category.

10Study participants in Barcelona received a thorough dietary and nutritional assessment by a specialized dietitian/nutritionist using two standardized and validated questionnaires, i.e.: (a) a prospective dietary nutrient survey aimed at recording, as precisely as possible, any food, supplement or liquid intake during 3 to 5 consecutive days, including at least one weekend day, and (b) a recall of food portions taken per week, on average, during the last year.

11Participants also went through a proctology evaluation by a specialized HIV physician/proctologist. In addition to visual inspection for anal or perianal lesions, HPV-related or not, the physician performed a rectal swab to rule out Chlamydia trachomatis and Neisseria gonorrhoeae infection using real-time PCR and an anal cytology. If the anal cytology reported an abnormal result, such as ASCUS (atypical squamous cells of undetermined significance), LSIL (low-grade squamous intraepithelial lesion) or HSIL (high-grade squamous intraepithelial lesion), the subject was properly treated and PCR typing of HPV was performed. No cases of anal cancer were detected.

12In all study participants, we produced MiSeq™ 16S rRNA sequence data on fecal microbiomes and measured soluble plasma markers of enterocyte damage (intestinal fatty acid-binding protein, IFABP), microbial translocation [soluble CD14 (sCD14) and lipopolysaccharide binding protein (LBP)] and systemic inflammation [interleukin-6 (IL-6), C-reactive protein (CRP) and interferon-gamma-inducible protein-10 (IP-10)].

13Study participants collected fecal samples in sterile fecal collection tubes the same day or the day before their clinical appointment, before the proctology exam, and following instructions pre-specified on standard operating procedures. If required, samples were stored at 4 °C overnight until DNA extraction. All samples collected in Barcelona were immediately extracted upon arrival to the laboratory. Additional aliquots were cryopreserved at − 80 °C for future studies. Samples collected in Stockholm were cryopreserved at − 80 °C and shipped on dry ice in batch to the IrsiCaixa AIDS Research Institute, where they were extracted, amplified, sequenced and analyzed using the exact same procedures applied to the Barcelona samples. The lag times to freezing were always < 36 h and no particular chemical stabilizers were added to samples used for the analyses presented here. Fecal sample collection procedures were the same for cases and controls.

14Detailed descriptions of the wet-lab procedures and the ecological and statistical analyses of the microbiome, soluble plasma markers and the nutritional assessment are available in the Supplementary methods section.

15The study was reviewed and approved by the Institutional Review Boards of the Hospital Universitari Germans Trias i Pujol (reference PI-13-046) and the Hospital Vall d'Hebrón (reference PR(AG)109/2014). The Stockholm study cohort was approved by the Regional Ethical Committee (Stockholm, Sweden, Dnr 2009-1485-31-3). All participants provided written informed consent in accordance with the World Medical Association Declaration of Helsinki. The study concept, design, patient information and results were discussed with the IrsiCaixa's Community Advisory Committee, who also provided input on the presentation and dissemination of study results (Supplementary methods).

16Raw Illumina MiSeq sequences and study metadata were deposited in the National Center for Biotechnology Information — NCBI repository (Bioproject accession number: PRJNA307231, SRA accession number: SRP068240).

17This study was mainly funded through philanthropy and private donations, which had no influence on its contents. Funds were obtained from a personal donation from Mr. Rafael Punter, the Gala contra la SIDA 2013 and 2014 editions, and the Nit per la Recerca a la Catalunya Central 2015 edition. M.R. is funded through a FI-DGR grant (FI-B00184) from Agència de Gestió d'Ajuts Universitaris i de Recerca (AGAUR) at the Secretaria d'Universitats i Recerca del Departament d'Economia i Coneixement de la Generalitat de Catalunya. Y.G. is supported through a post-doctoral grant from the Fundación Paideia Galiza. M.C. is funded through the Red de Investigación en SIDA, RD12/0017/0002 as part of the Plan Nacional R + D + I and cofinanced by the Instituto de Salud Carlos III (ISCIII)-Subdirección General de Evaluación y el Fondo Europeo de Desarrollo Regional (FEDER). J.R. is supported through a grant for doctoral studies from Noel Alimentaria to the University of Vic (UVic-UCC). B.M. is a Joan Rodés investigator from the ISCIII (JR13/00024), Madrid, Spain. M.L.C. is funded through the grant MTM2012-38067-C02-02, Spanish Ministry of Economy and Competitiveness, Spain. PB, FH and GZ are supported by the European Molecular Biology Laboratory. F.H. was funded from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement no. 600375. P.B. acknowledges the European Research Council grant Cancerbiome, reference 268985. The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all study data, and had final responsibility for the decision to submit for publication.

18The study included 240 individuals, 156 in Barcelona (Table 1) and 84 in Stockholm (Table 2). The test cohort BCN0 comprised 129 (82.7%) HIV-1-infected and 27 (17.3%) HIV-negative subjects. The internal validation cohort BCN1 included 110 individuals, 87 HIV-1-infected (79.1%) and 23 non-HIV-infected (20.9%). The external validation cohort STK had 77 HIV-1-infected (91.6%) and 7 non-HIV-infected individuals (8.4%). In Barcelona, the median age of study participants was 43 years and their median body mass index was 23.8 kg/m2. Eighty percent of subjects were men, mostly from Caucasian ethnicity. Sixty-four percent of all subjects were MSM, 26% HTS and 10% PWID. There were 8 (5.1%) elite controllers, 11 (7.1%) viremic controllers, 15 (9.6%) ART-naïve, 13 (8.3%) early-treated, 53 (34.1%) immune concordant, 18 (11.5%) immune discordant, and 11 (7.1%) late presenters. HIV-1-infected subjects were slightly older and were more likely to be HBV and HCV positive than HIV-negative controls. Groups were well balanced in all other factors. In Stockholm, 60% of subjects were men; 23% were MSM, 66% HTS and 11% PWID. Only half were nationals from Scandinavian countries; 62% individuals were Caucasian and 33% were Black.

19The fecal microbiota was significantly richer and more diverse in MSM than non-MSM individuals in both cities, also after correcting for multiple comparisons (Fig. 1, Supplementary Figs. 2 and 14). This indicated that the measurement of the effect of HIV-1 on gut microbial richness and diversity had to take HIV transmission group into account. After stratifying for MSM vs. non-MSM, HIV-1 infection remained consistently associated with reduced bacterial richness (15% to 30% reduction relative to HIV-negative individuals) in both groups and both cities (Fig. 1, Supplementary Figs. 2 and 14). In the Barcelona cohort, the lowest microbial richness and diversity was observed among HIV-1-infected individuals with an immune-virological discordant phenotype (Fig. 2). Subjects with an immune-virological concordant phenotype had higher microbial richness than immune discordant individuals, but, nevertheless, still showed reduced microbial richness relative to HIV-negative controls, suggesting that despite adequate immune recovery [median (IQR) CD4 + T-cell counts: 761 (640, 932) cells/mm3] at the time of testing, ART had not been able to fully normalize microbial richness.

20Clustering of the fecal microbiomes in BCN0 and STK using a partitioning around medoids (PAM) algorithm suggested the presence of at least 2 clusters of fecal microbiomes in both cities (Fig. 4c). Such clusters were enriched either in Bacteroides or Prevotella, and had a similar bacterial composition to the corresponding previously described enterotypes (Supplementary Fig. 3). As expected from previous work on gut enterotypes, there were strong positive correlations between the genus Bacteroides and Parabacteroides, Barnesiella, Alistipes and Odoribacter, as well as between Prevotella and Alloprevotella, Catenibacterium, Mitsuokella and Intestinimonas, among others (Fig. 3), highlighting that differences between the groups extended beyond a single genus. The genera correlating with Prevotella were negatively correlated with Bacteroides and vice versa. Moreover, the microbiomes of the Bacteroides and Prevotella clusters showed remarkably different functional profiles (Supplementary Figs. 4 and 5), also in agreement with previous enterotype descriptions.

21We explored variables potentially influencing the composition of the fecal microbiomes, according to a univariate ADONIS test of ecological distance and found possible effects of HIV-1 risk group, gender, feces consistency, place of residency, ethnicity, HIV-1 serostatus and altered abdominal transit (Supplementary Table 1). However, only the HIV-1 risk group retained statistical significance in a multivariate ADONIS analysis with terms added sequentially (R2: 0.373, p < 0.001).

22Fecal microbiomes in BCN0, BCN1 and STK clustered by HIV transmission group rather than by HIV-1 serostatus, using either Bray–Curtis (Fig. 4b) or other ecological distances (Supplementary Figs. 6 to 8). Although a few individuals showed marked differences between the two time points, fecal microbiota ordination was highly concordant between BCN0 and BCN1 (Procrustes m2 = 0.3475, PROTEST p = 0.001) (Supplementary Fig. 9), indicating that differences in microbial ordination were not due to random variation. The fecal microbiota composition in both BCN0 and STK significantly differed by HIV transmission group, with MSM and non-MSM subjects mostly belonging to the Prevotella and Bacteroides clusters, respectively (Fig. 4a and 4d and Supplementary Figs. 10 to 14). Alpha and beta diversity and genus abundance analyses were reproducible using a different analysis pipeline (Supplementary Fig. 14 and Supplementary methods).

23In an analysis accounting for the potential interdependency of sexual preference and HIV-1 serostatus (LEfSe), there were consistent differences in both cities only by sexual preference group, with enrichment of Prevotella, Alloprevotella, Succinvibrio, Dorea, RC 9 gut group, Desulfovibrio, Phascolarctobacterium and unclassified Bacteroidales in MSM, and enrichment in Bacteroides, Odoribacter and Barnesiella in non-MSM individuals (Supplementary Fig. 15).

24To quantify the strength of the association between HIV transmission group, HIV serostatus and global fecal microbiota composition, we applied a previously validated global microbiota classification concept based on LASSO regression to our BCN0 dataset. Cross-validation accuracy was extraordinarily high for sexual preference group (mean AUC = 95%), confirming a different fecal microbiota composition in MSM and non-MSM individuals (Fig. 5). In contrast, HIV-1 status was not associated with consistent changes in the global fecal microbiota composition at the genus level, suggesting that the reduction in microbial richness observed in HIV-infected individuals was not genus-specific.

25Relative to non-MSM subjects, MSM were younger, were more likely to live in Barcelona City, reported softer fecal consistency, and were less likely to be infected with HBV and HCV (Supplementary Table 2). However, none of these factors among others were likely to confound the previous LASSO models (Supplementary Fig. 16). Although long-term dietary patterns have been linked to alternative enterotype states, the effect of diet on microbiota composition was limited in our setting (Fig. 6 and Supplementary Fig. 17) and none of the diet components was selected by multivariate LASSO regression as a consistent predictor of microbiota clustering.

26Markers of enterocyte damage, microbial translocation and systemic inflammation followed an overall predictable response across different HIV phenotypes, being generally higher in immune discordant and late presenters (Supplementary Figs. 18 and 19). However, they did not differ between the Bacteroides or Prevotella clusters or between MSM and non-MSM individuals.

27In two independent European cohorts with different ethnic and cultural background, the fecal microbiota of MSM was consistently richer and more diverse than that of non-MSM subjects, and was systematically enriched in genera from the Prevotella enterotype. The strength of such association was unusually high, reaching 95% accuracy in a microbial composition-based classifier. These findings have important implications for HIV microbiome science. To our knowledge, this is the first evidence that, in addition to genetic, lifestyle and environmental factors, factors related with sexual preference might also affect the gut microbiota composition.

28Based on our findings, previous associations between HIV infection and Prevotella might be explained by enrichment of HIV-infected groups by MSM relative to HIV-negative controls selected from hospital or research staff, gut biopsy donors, or college students. Contradictory associations between HIV infection and microbial richness could also be affected by unbalances in the proportion of MSM between groups. Of note, a selection bias as such could also affect the interpretation of in silico inferences on bacterial metabolism, or even direct metabolomic or metatranscriptomic measurements, which also rely on bacterial composition.

29In concordance with data from animal models and studies matching for HIV risk factors, we were unable to identify a consistent HIV-specific fecal dysbiosis pattern after stratifying for HIV transmission group. Yet, HIV-1 infection remained associated with reduced bacterial richness independently of sexual orientation, indicating that the most evident hallmark of HIV infection on the gut microbiome is, like in other intestinal inflammatory diseases, a reduction in bacterial richness. In line with previous observations linking bacterial richness with immune dysfunction, the lowest bacterial richness was found in immune discordant subjects, followed by immune concordant individuals with adequate immune recovery on ART. Conversely, bacterial richness was conserved in subjects initiating ART during the first 6 months of HIV infection, as well as in ART-naïve individuals with > 500 CD4 + counts/mm3, suggesting that early ART initiation might help to preserve gut microbial richness.

30The strong epidemiological association of fecal microbiota composition with sexual orientation in two independent cities is yet to be translated into specific mechanisms. We ruled out multiple confounders and only found a limited effect of diet in our setting. We did not collect information on exercise, but exercise has been linked to fecal microbiota composition in athletes and even in them diet plays an important role. A formal assessment of the socioeconomic status of our patients was out of the scope of this work, although based on our findings, rigorous studies assessing the role of socioeconomic status in the fecal microbiota composition are needed. Non-MSM subjects in our study were older and more likely to be co-infected with HBV and HCV than MSM, reflecting current trends of the HIV epidemic in Europe, i.e.: most new HIV-1 infections occur in young MSM who rarely use intravenous drugs. Fecal consistency was also softer in MSM than in non-MSM subjects, which, indirectly, might reflect better overall health habits, including a healthier diet, higher water consumption and physical activity. However, none of these factors, nor ethnicity, achieved a significant weight in LASSO models.

31Further studies are needed to evaluate the existence of ecological adaptations of commensal bacteria to changes in gut mucosa induced by sexual practices. Populations of commensal bacteria are controlled by substrate competition and glycan availability and several factors might affect distal colorectal mucosa, including hyperosmolar substances like semen or certain lubricants, colorectal cleansing or use of sexual toys. Longitudinal studies should also clarify if the observed association is stable over time, and if it varies according to the number of sexual partners (i.e., long-term single relationships versus frequent partner exchange) or by insertive versus receptive anal sex. It is also important to clarify if the observed association remains in heterosexual women who engage in receptive anal sex and if increased microbiota richness can be related to person-to-person transmission of commensal bacteria. Future studies should also investigate if the observed association has implications for transmission of infectious agents, including HIV-1. We did not find an association between fecal microbiome and HBV, HCV, syphillis or rectal HPV, C. trachomatis or N. gonorrhoeae infections, but did not evaluate HSV-2 infection. In our study, the observed association between sexual orientation and microbiota composition did not translate into gross differences in terms of systemic inflammation or microbial translocation. Shotgun metagenomic analyses of bacterial species and richness, as well as the virome and perhaps the mycobiome, in clinical trials balanced by HIV risk factors might provide novel clues as to the impact of HIV infection on the gut microbiome.

32In conclusion, the fecal microbiota of gay men in Europe is richer and has a distinct composition. However, HIV-1 infection remains independently associated with reduced bacterial richness. This offers new avenues for therapeutic interventions on the gut microbiome which might improve HIV-associated immune dysfunction.

33R.P., M.N., P.N., A.S., J.B. and B.C. conceived and designed the study. R.P., I.B., B.M., E.N., J.Co., J.S., A.T., J.N., C.B. and B.C in Barcelona and P.N., and A.S., in Stockholm, recruited the study participants and performed their clinical evaluations. C.E. performed the dietary assessment. G.S. and J.C. performed the proctology studies. C.H. coordinated the study logistics including the fulfillment of all ethical and legal requirements of the study as member of the Contract Research Organization overseeing the study, in coordination with R.P. Fecal 16S rDNA was extracted, amplified and sequenced by M.P, M.C., M.R. and R.B. under the supervision of M.N. and R.P. M.R., Y.G, J.R., C.R., F.H. and G.Z. performed the bioinformatic and statistical analyses of the 16S rDNA data, with the supervision of M.N., M.L.C., P.B., F.G. and R.P. M.C. performed the inflammation analyses under the supervision of J.Ca., J.B and R.P. J.R., did the statistical analyses of the relationship between the microbiota and inflammation and diet, under the supervision of M.N., M.L.C., F.G. and R.P. F.H and G.Z. performed the multivariate analysis of factors determining microbiota clusters and ran the confirmatory analyses with the independent sequence analysis pipeline LotuS, under supervision of P.B. R.P. wrote the paper, which was reviewed, edited and approved by all authors.

34The authors declare that they have no conflicts of interest.