www.nature.com/mi ARTICLE OPEN Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Yolanda Guillén1,2, Marc Noguera-Julian1,2,3, Javier Rivera1,3, Maria Casadellà1,2, Alexander S. Zevin4, Muntsa Rocafort1,2, Mariona Parera1, Cristina Rodríguez1, Marçal Arumí1, Jorge Carrillo1,2, Beatriz Mothe1,3,5, Carla Estany5, Josep Coll1,5, Isabel Bravo5, Cristina Herrero5, Jorge Saz6, Guillem Sirera5, Ariadna Torrella7, Jordi Navarro2,7, Manuel Crespo8, Eugènia Negredo2,3,5, Christian Brander1,2,3,9, Julià Blanco1,2,3, Maria Luz Calle3, Nichole R. Klatt4, Bonaventura Clotet1,2,3,5 and Roger Paredes1,2,3,5 Human immunodeficiency virus (HIV)-1 infection causes severe gut and systemic immune damage, but its effects on the gut microbiome remain unclear. Previous shotgun metagenomic studies in HIV-negative subjects linked low-microbial gene counts (LGC) to gut dysbiosis in diseases featuring intestinal inflammation. Using a similar approach in 156 subjects with different HIV-1 phenotypes, we found a strong, independent, dose–effect association between nadir CD4+ T-cell counts and LGC. As in other diseases involving intestinal inflammation, the gut microbiomes of subjects with LGC were enriched in gram-negative Bacteroides, acetogenic bacteria and Proteobacteria, which are able to metabolize reactive oxygen and nitrogen species; and were depleted in oxygen-sensitive methanogenic archaea and sulfate-reducing bacteria. Interestingly, subjects with LGC also showed increased butyrate levels in direct fecal measurements, consistent with enrichment in Roseburia intestinalis despite reductions in other butyrate producers. The microbiomes of subjects with LGC were also enriched in bacterial virulence factors, as well as in genes associated with beta-lactam, lincosamide, tetracycline, and macrolide resistance. Thus, low nadir CD4+ T-cell counts, rather than HIV-1 serostatus per se, predict the presence of gut dysbiosis in HIV-1 infected subjects. Such dysbiosis does not display obvious HIV-specific features; instead, it shares many similarities with other diseases featuring gut inflammation. Mucosal Immunology (2019) 12:232–246; https://doi.org/10.1038/s41385-018-0083-7 INTRODUCTION translocation of bacterial products to regional lymph nodes and In settings with adequate access to antiretroviral therapy (ART), systemic circulation.22,23 Infection also induces inflammation of the main clinical problems of people living with human the mesenteric fat and promotes extensive deregulation of local immunodeficiency virus (HIV)-1 (PLWH) derive either from late immune responses.24 In all, this contributes to chronic inflamma- HIV-1 diagnosis1,2 or from premature aging.3,4 Up to one-third of tion and immune activation, which are only partially restored by all HIV-1-related deaths in Western countries can be attributed to ART.25 Immune activation hinders immune reconstitution,26,27 late HIV-1 diagnosis.5 Low nadir CD4+ T-cell counts are a major diminishes the efficacy of HIV-1 preventive and curative strate- risk factor for developing AIDS- and non-AIDS-defining illnesses,6–8 gies,21 fosters immune senescence28–30, and at the mucosal level, bacterial infections,9 ART failure,5 and suboptimal immune is a major risk factor for HIV-1 transmission.21 reconstitution with ART.10,11 Premature aging involves the However, the effects of HIV-1 infection on the gut microbiome precocious development of type 2 diabetes, dyslipidemia, are not well understood and, to date, a consistent pattern of gut cardiovascular diseases, osteoporosis, and frailty syndrome.4 In dysbiosis has not been identified. Unlike the vaginal microbiome, PLWH, such illnesses have been linked to chronic inflammation, where compositional shifts from homeostasis are often recogniz- immune activation, and endotoxemia.4,12–14 In non-HIV-infected able as dysbiotic community types,31,32 dysbiosis in the gut is subjects, they have been related to shifts in the gut microbiome more difficult to detect and even to define. As a matter of fact, the associated with reduced microbial gene richness.15–20 debate on what characterizes a healthy gut microbiome is still The gut microbiome is believed to play a central role in HIV-1 ongoing33, and over 20% of the interperson microbiome immunopathogenesis as well as in HIV-associated chronic variability is associated with factors related to diet, drugs, and complications. HIV-1 infection damages the intestinal mucosal anthropometric measurements,34 which may confound associa- barrier and gut-associated lymphoid tissues14,21 enabling tions between gut microbiome shifts and diseases. 1 irsiCaixa AIDS Research Institute, Ctra de Canyet s/n, Badalona 08916 Catalonia, Spain; 2Universitat Autònoma de Barcelona, Bellaterra 08193 Catalonia, Spain; 3Universitat de Vic- Universitat Central de Catalunya, C. Sagrada Família 7, Vic 08500 Catalonia, Spain; 4University of Washington, 3018 Western Avenue, Seattle, WA 98121, USA; 5Infectious Diseases Service & Lluita contra la SIDA Foundation, Hospital Universitari Germans Trias i Pujol, Ctra de Canyet s/n, Badalona 08916 Catalonia, Spain; 6BCN Checkpoint, Carrer del Comte Borrell, 164, Barcelona 08015 Catalonia, Spain; 7Infectious Diseases Unit, Hospital Universitari Vall d’Hebrón, Passeig de la Vall d’Hebrón, 119-129, Barcelona 08035 Catalonia, Spain; 8Infectious Diseases Unit, Internal Medicine Department, Complexo Hospitalario Universitario, Vigo. IIS Galicia Sur, Estrada de Clara Campoamor, 341, Vigo 36312 Pontevedra, Spain and 9Institució Catalana de Recerca i Estudis Avançats (ICREA), Pg. Lluís Companys 23, Barcelona 08010 Catalonia, Spain Correspondence: Roger Paredes (rparedes@irsicaixa.es) These authors contributed equally: Yolanda Guillén, Marc Noguera-Julian. Received: 17 May 2018 Revised: 3 August 2018 Accepted: 8 August 2018 Published online: 31 August 2018 © Society for Mucosal Immunology 2018 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 233 Table 1. Study cohort characteristics All subjects Low-gene counts High-gene counts p Value (n = 156) (n = 103) (n = 53) Age, years, median (IQR) 43 (35; 51) 46 (36; 53) 38 (35; 46) 0.024 Gender, n (%) Men 124 (79.5%) 77 (74.8%) 47 (88.7%) 0.058 Women 31 (19.9%) 26 (25.2%) 5 (9.4%) 0.020 Transgender women 1 (0.6%) 0 1 (1.9%) 0.340 Ethnicity, n (%) Asiatic 1 (0.6%) 0 1 (1.9%) 0.340 Caucasian 124 (79.5%) 88 (85.4%) 36 (67.9%) 0.013 Hispanic-latin 28 (17.9%) 12 (11.7%) 16 (30.2%) 0.007 Other 3 (1.9%) 3 (2.9%) 0 0.551 HIV-1 risk group, n (%) MSM 100 (64.1%) 54 (52.4%) 46 (86.8%) <0.001 Non-MSM 56 (35.9%) 49 (47.6%) 7 (13.2%) HIV-1 status, n (%) HIV-1-negative 27 (17.3%) 11 (10.7%) 16 (30.2%) 0.004 HIV-1-positive 129 (82.7%) 92 (89.3%) 37 (69.8%) HIV-1 phenotype, n (%) Late presenter 11 (7.1%) 9 (8.7%) 2 (3.8%) 0.335 Discordant 18 (11.5%) 15 (14.6%) 3 (5.7%) 0.118 Concordant 53 (34.0%) 42 (40.8%) 11 (20.8%) 0.277 Early-treated 13 (8.3%) 8 (7.8%) 5 (9.4%) 0.277 ART-naive 15 (9.6%) 8 (7.8%) 7 (13.2%) 0.783 1234567890();,: Viremic controller 11 (7.1%) 5 (4.9%) 6 (11.3%) 0.033 Elite controller 8 (5.1%) 5 (4.9%) 3 (5.7%) 0.421 HIV-1-negative 27 (17.3%) 11 (10.7%) 16 (30.2%) 0.004 BMI, kg/m2, median (IQR) 23.8 (22.0; 26.1) 23.7 (21.8; 25.5) 24.4 (22.3; 26.3) 0.215 History of allergy, n (%) Yes 30 (19.2%) 21 (20.4%) 9 (17.0%) 0.673 No 122 (78.2%) 78 (75.7%) 44 (83.2%) 0.413 NA 4 (2.6%) 4 (3.9%) 0 0.300 Antibiotic intake, previous 3 months, Yes 2 (1.3%) 2 (1.9%) 0 0.549 n (%) No 154 (98.7%) 101 (98.1%) 53 (100.0%) Antibiotic intake, previous 6 months, Yes 35 (22.4%) 26 (25.2%) 9 (17.0%) 0.312 n (%) No 121 (77.6%) 77 (74.8%) 44 (83.0%) Fecal consistency, n (%) Hard 56 (35.9%) 33 (32.0%) 23 (43.4%) 0.217 Soft 91 (58.3%) 63 (61.2%) 28 (52.8%) 0.392 Liquid 5 (3.2%) 4 (3.9%) 1 (1.9%) 0.662 NA 4 (2.6%) 3 (2.9%) 1 (1.9%) 1.000 Abdominal transit alterations, n (%) Yes 23 (14.7%) 14 (13.6%) 9 (17.0%) 0.636 No 127 (81.4%) 85 (82.5%) 42 (79.2%) 0.666 NA 6 (3.8%) 4 (3.9%) 2 (3.8%) 1.000 CD4+ T-cell counts, median (IQR) Current, cells/mm3 700.0 (462.0; 860.0) 635.5 (331.0; 851.2) 750.0 (558.0; 930.0) 0.126 Current, % 33.0 (24.2; 40.2) 32.8 (23.6; 40.2) 33.8 (25.6; 40.3) 0.474 Nadir, cells/ mm3 337.0 (140.0; 528.5) 280.0 (113.0; 491.5) 443.0 (339.0; 601.0) 0.002 CD8+ T-cell counts, median (IQR) Current, cells/ mm3 776.5 (576.0; 1012.0) 791.5 (558.8; 991.2) 748.5 (604.2; 1158.0) 0.294 Current, % 39.4 (33.0; 48.92) 38.2 (31.6; 49.7) 41.6 (34.9; 47.1) 0.607 CD4+/CD8+ ratio 0.9 (0.5; 1.2) 0.9 (0.5; 1.3) 0.8 (0.5;1.2) 0.943 HIV-1 RNA, copies/mL, median (IQR) <40.0 (40.0; 711.5) <40.0 (40.0; 79.5) <40.0 (40.0; 1339.0) 0.074 MSM men-who-have-sex-with-men; ART antiretroviral therapy; BMI body mass index; SD standard deviation; IQR 25–75% interquartile range. Initial HIV gut microbiome studies using 16S rRNA gene “enterotype”-like microbiome clustering by HIV-infected vs. sequencing suggested a shift from Bacteroides to Prevotella uninfected individuals. Considering the large biomass, ecological dominance in HIV-1 infection.22,35–40 More recently, our group complexity and functional redundancy of the gut microbiome, it is was able to demonstrate that such changes were confounded by conceivable that altered microbiome states might be subtler, and sexual preference—men-who-have-sex-with-men (MSM) were possibly imply shifts in discrete groups of microbial species or enriched in Prevotella—and, after controlling for HIV-1 risk group, changes in functional features which might not necessarily be there was no consistent microbial dysbiosis pattern discernible by evident in genus-level ordination analyses. Moreover, gut 16S rRNA gene sequencing.41 One implication of our findings is dysbiosis might not necessarily be present in all HIV-1-infected that we should no longer expect “community-type structures” or subjects, but instead, be more frequently found in subjects with Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 234 Fig. 1 Low-microbial gene richness is linked to HIV-1-mediated immune suppression. a Probability density function showing a bimodal distribution of the study population according to gut microbial gene richness, i.e., the number of different microbial genes observed in each individual at 10 million sequence downsampling size. The local minimum between the two modes (621,808 genes in this analysis) was used as a threshold to define people with high (HGCs) and low-gene counts (LGCs). b The bimodal distribution became even more aparent when KEGG enzyme richness, i.e., number of different enzymes found in the KEGG database was used. c HIV-1-infected subjects (n = 156) were more likely to be classified as LGCs; in comparison, most HGCs were HIV-1 negative. d The association between LGC and HIV-1 infection remained when only men-who-have-sex-with-men (MSM) were analyzed (n = 100). e HIV-1 phenotypes associated with immune deterioration were enriched in LGCs. f Subjects with lower nadir CD4+ counts were gradually more likely to be classified as LGCs. Late presenter, subjects with CD4+ T-cell counts <200 cells/mm3 at diagnosis and no ART exposure; discordant, with HIV-1 RNA <50 copies/mL and CD4+ T-cell counts <300 cells/mm3; concordant, with HIV-1 RNA levels <50 copies/mL and achieving CD4+ T-cell counts >500 cells/mm3; early treated, HIV-1- infected subjects who initiated ART during the first 6 months after the infection, achieving HIV-1 RNA levels <50 copies/mL during at least 3 months and with no HIV-1 RNA blips after achieving HIV-1 RNA <50 copies/mL; ART naïve, with HIV-1 RNA >10,000 copies/mL, nadir CD4+ T- cell counts >500 cells/mm3 and no ART exposure; viremic controller, with HIV-1 RNA between 50 and 2000 copies/mL during at least 2 years in the absence of ART; elite controller, subjects with HIV-1 RNA <50 copies/mL during at least 2 years in the absence of ART higher degree of immune deficiency, longer duration of HIV-1 (HGC) or low-gene counts (LGC). Such studies have shown that infection or late presentation, as previously suggested by using subjects with LGC consistently display changes in their gut markers of immune activation.41–43 Thus, it may be oversimplistic microbiomes that most authors would accept as “dysbiosis”, rendering HIV-1 vs. non-HIV-1 comparisons. including increases in Proteobacteria and other microbial species A highly consistent feature of the gut microbiomes of people able to thrive in the presence of gut inflammation. In the present with different gut inflammatory diseases in studies using shotgun study, we used fecal whole metagenome shotgun sequencing and metagenomics is a reduction in microbial gene richness,15–20 direct short-chain fatty acid (SCFA) measurements to characterize which allows classifying individuals as having high-gene count the nature and possible mechanisms driving gut microbiome Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 235 Table 2. Factors associated with low-microbial gene counts a Univariate Multivariate OR 95% CI p Value OR 95% CI p Value Age Per each additional year 1.03 [1–1.07] 0.074 – – – Gender Female 1 Male 0.32 [0.10–0.82] 0.028 – – – Transgender woman – – – Ethnic Group Caucasian 1 1 Hispanic-Latin 0.27 [0.11–0.64] 0.003 0.26 [0.10–0.67] 0.006 Asiatic and othersb – – – – – – HIV-1 risk group Non-MSM 1 1 MSM 0.17 [0.06–0.39] <0.001 0.20 [0.07–0.51] 0.002 HIV-1 status Negative 1 Positive 3.68 [1.57–8.89] 0.003 – – – Nadir CD4+ T-cell count, cells/mm3 HIV-1 negative 1 1 >500 2.04 [0.75–5.74] 0.169 2.13 [0.73–6.45] 0.173 200–500 3.09 [1.19–8.37] 0.006 2.92 [1.03–8.62] 0.047 100–200 6.55 [1.86–27.71] 0.023 5.55 [1.40–26.15] 0.020 <100 26.18 [4.40–506.6] 0.003 14.00 [2.02–288.71] 0.023 Full dataset analysis, n = 156 subjects a b Analysis does not apply because all subjects are included in the same response group MSM men-who-have-sex-with-men shifts in HIV-1 infection, identify clinical factors associated with were more likely to be older (p = 0.024), Caucasian (p = 0.013), them, and explore their potential consequences on microbial non-MSM (p < 0.001) and women (p = 0.020) than HGC individuals pathogenesis and human health. (Table 1). Average microbial gene richness was also significantly lower in HIV-1-infected subjects, particularly in those with more advanced disease, immune discordant response, and lower nadir RESULTS CD4+ T-cell counts (Supplementary Figs. 3 and 4, p < 0.05 in all Study subjects cases). Neither current CD4+ T-cell counts nor CD4% were The study included 156 individuals: 129 (82.7%) HIV-1-infected associated with microbial richness after stratifying for nadir CD4 with different immune phenotypes and 27 (17.3%) not infected + T-cell counts (Supplementary Fig. 5). Similarly, we did not with HIV-1 (Table 1). Participants were mostly MSM (64%) of observe an effect of ART or previous antibiotic exposure on gene Caucasian ethnicity (79%), with median 43 years of age, and richness after stratifying for nadir CD4+ T-cell counts (Supple- median body mass index (BMI) of 23.8 kg/m2. HIV-1-infected mentary Figs. 6 and 7). subjects were recruited from two tertiary HIV-1 clinics in Barcelona, Spain. Most HIV-1-negative controls were enrolled Dose-dependent association between LGC and nadir CD4+ T-cell from a prospective cohort of HIV-negative MSM who attend counts quarterly medical and counseling visits at a community-based In multivariate regression (Table 2) there was a dose–effect center in Barcelona.44,45 relationship between nadir CD4+ T-cell counts and microbial gene richness. Subjects with nadir CD4+ T-cell counts between Microbial gene richness 200 and 500 cells/mm3 (OR = 2.9 [95 CI: 1.1–8.6], p = 0.047), 100 Microbial gene richness values obtained by whole fecal metagen- and 200 cells/mm3 (OR = 5.5 [95 CI: 1.4–26.1], p = 0.020) and <100 ome shotgun sequencing followed a bimodal distribution across cells/mm3 (OR = 14.0 [95 CI: 2.02–288.7], p = 0.023) were more the study cohort (Fig. 1a). Using 10 million (M)-sequence likely to be LGC than HIV-negative controls. Additional factors downsampling, the local intermodal minimum was 621,808 genes. independently associated with LGC were HIV-1 risk group (OR for This threshold was used to classify participants as having HGC (n MSM vs. non-MSM = 0.2 [95 CI: 0.1–0.5], p = 0.002) or ethnicity (OR = 53) or LGC (n = 103), as in ref. 18. Subjects in the HGC group for Hispanic vs. Caucasian = 0.3 [95 CI: 0.1–0.7], p = 0.006). As also had higher enzyme and metabolic pathway richness than expected, time between the study baseline and dates of HIV-1 LGCs (Fig. 1b, Supplementary Table 1). Gene richness measure- diagnosis and nadir CD4+ T-cell counts were also longer in ments were not biased by sequencing yield (Supplementary Fig. 1) subjects with lower nadir CD4+ T-cell counts (Supplementary or gene catalog used (Supplementary Fig. 2). Fig. 8). Factors associated with microbial gene richness Changes in microbial species composition HIV-1-infected subjects were more likely to be LGCs than HIV-1- Fecal microbiomes clustered by gene richness category rather negative controls, regardless of whether all subjects (71.3% vs. than by HIV-1 serostatus (Fig. 2a, Supplementary Fig. 9). Micro- 40.7%, p = 0.001, Fig. 1c) or only MSM (59.7% vs. 34.8%, p = 0.055, biomes from HGC subjects were more similar among themselves Fig. 1d) were considered.41 Subjects with LGCs were more than those from LGC individuals, which displayed many different frequently observed in individuals with more advanced HIV-1 configurations. disease stages (p = 0.009, Fig. 1e) and in gradually lower nadir CD4 The microbial species composition of HGC and LGC indivi- + T-cell count ranks (p = 0.002, Fig. 1f). In addition, LGC subjects duals was remarkably different (Fig. 2b, c, Supplementary Figs. Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 236 Fig. 2 Microbial species associated with gut microbial gene richness. a Nonmetric multidimensional scaling plot based on Bray–Curtis microbiome distances showing more different microbiome configurations in subjects with LGCs than in HGCs. Dot sizes are proportional to microbiome gene richness. Ellipses represent 95% confidence intervals. Microbial species associated to an environmental vector of more than 0.35 NMDS coordinate length are shown. b Microbial species positively (green) and negatively (red) correlated with gene richness (Benjamini–Hochberg corrected Spearman Rank correlation p value < 0.05) in MSM (n = 100) and non-MSM (n = 56) subjects. Most microbial species positively correlated with richness were GRAM-positive, whereas most species negatively correlated with richness were GRAM- negative (V is for “variable” GRAM staining). Blue squares on the left handside summarize some of the main microbial metabolism products of each species, including the short-chain fatty acids (SCFAs) acetate (Ace), butyrate (But), formate (For), and propionate (Pro); the acids lactate (Lac) and succinate (Suc); the gas methane (CH4), and the alcohols ethanol (Eth) and methanol (Met); * is for unknown. c Random forest analysis showing the contribution of each microbial species to overall gene richness. In green, red, and gray, respectively, microbial species enriched in HGCs, LGCs and not significantly enriched in either category. AUC area under the curve 10–14). Gut microbiomes of HGC subjects were enriched in and butyrate-producing species from the Ruminococcaceae methanogenic archaea (Methanobrevibacter spp. and Methano- (Subdoligranulum spp.) and Lachnospiraceae (Butyrivibrio cross- sphaera stadtmanae), sulfate-reducing bacteria (Desulfovibrio otus, Coprococcus comes, Conus catus) families (Fig. 2). In piger and Desulfovibrio desulfuricans) (Supplementary Fig. 14), addition, subjects with HGCs showed enrichment in Eubacterium Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 237 siraeum, Eubacterium eligens and species from Dorea genera Conversely, subjects in the LGC group had microbiomes (Dorea formicigenerans and Dorea longicatena), as well as in enriched in Bacteroides (Bacteroides thetaiotaomicron, Bacteroides cellulolytic bacteria from the Ruminococcus genus (Ruminococ- clarus, Bacteroides stercoris), Parabacteroides (Parabacteroides cus flavefaciens). merdae) and acetogenic bacterial species (Blautia Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 238 Fig. 3 Functional unbalance and microbial adaptation to oxidative stress. a Microbial functions significantly correlated with gene richness (Benjamini–Hochberg-corrected Spearman Rank correlation p value < 0.05) in both MSM and non-MSM subjects. b Differential enrichment in microbial enzymes involved in metabolism of reactive oxygen species (ROS) in LGCs (upper horizontal bars) vs. HGCs (lower horizontal bars), and contribution of microbial species to such enrichment. The plot shows significant differences in ROS gene enrichment between LGC and HGC (diamond, Wilcoxon test statistic) using a false discovery rate p value threshold of 0.05. Microbes are colored by phyla, i.e., Bacteroidetes in purple; Firmicutes in blue; Proteobacteria in red, Others in gray. c Abundance of microbial enzymes involved in ROS metabolism by nadir CD4+ T-cell count categories. G6PHD, glucose-6-phosphate dehydrogenase; 6pgd, 6-phosphogluconate dehydrogenase; gr, glutatione reductase; gpx, glutathione peroxidase; ahpc, peroxiredoxin AhpC; bcp, peroxiredoxin BCP; tpx, atypical 2-Cys peroxiredoxin; cat, catalase; sod2, superoxide dismutase Fe–Mn; msra, methionine sulfoxide reductase A hydrogenotrophica, Blautia producta, Marvinbryantia formatexigens, microbiomes were enriched in genes involved in the metabolism and Eubacterium limosum) (Supplementary Fig. 14); as well as in of fructose, and a wide range of disaccharides (sucrose, maltose, Proteobacteria (Supplementary Fig. 13), including the Betaproteo- cellobiose, and lactose) and polysaccharides (pectin, galactan, and bacteria Sutterella wadsworthensis. amylopectin). LGCs were also characterized by an enrichment of Interestingly, some of the best-known butyrate-producing genes involved in biosynthesis of lipopolysaccharide (LPS), the microbial species46,47 were present at significant amounts across main component of the outer membrane of Gram-negative the whole gene richness continuum. Whereas Faecalibacterium bacteria. Tryptophan-degrading enzymes such as tryptophanase prausnitzii was more abundant with higher gene richness, and aromatic-L-amino-acid decarboxylase, which catalyze the Roseburia intestinalis and Roseburia inulinivorans were more degradation of tryptophan to indole and tryptamine, respectively, prevalent with lower gene counts, and Eubacterium rectale was were increased in LGCs. homogeneously distributed across gene richness. Contribution of microbial species to functional shifts Gene richness and inflammation In an integrative taxonomic and functional comparative analysis to Individuals with lower gene richness had higher levels of systemic estimate taxon-level contributions to functional shifts (FishTaco),48 inflammation (IL-6, p = 0.020) (Supplementary Fig. 15). HIV-1 LGCs were enriched in genes involved in reactive oxygen and infection and particularly low nadir CD4 T-cell counts were also nitrogen species (ROS/RNS) metabolism, encoded by different significantly associated with higher levels of systemic inflamma- Bacteroides and Proteobacteria species (Fig. 3b, Supplementary tion (IL-6 and IP-10), and enterocyte damage (intestinal fatty acid- Fig. 20). Moreover, ROS/RNS enzymes were significantly more binding protein (IFABP)) (Supplementary Fig. 16). abundant as microbial gene richness (Supplementary Figs. 19 and 21) and nadir CD4+ T-cell counts (Fig. 3c, Supplementary Fig. 22) Diet, richness, and microbial composition decreased, in a dose–effect relationship. Microbial gene richness directly correlated with the dietary intake Bacterial antioxidant enzymes enriched in LGCs included redox of monounsaturated fats, carotenoids, iron, fiber and Vitamin A enzymes such as thioredoxin reductase and peroxiredoxines from (Supplementary fig. 17a). Dirichlet-multinomial regression identi- the thioredoxin system; glutathione peroxidase and reductase; fied a positive association between iron intake and levels of and glucose 6-phosphate dehydrogenase and 6- Prevotella copri, E. eligens, Ruminococcus bromii and Ruminococcus phosphogluconate dehydrogenase, the latter involved in the callidus, and a negative association with Bacteroides ovatus. oxidative phase of the pentose phosphate pathway. Catalase and Ethanol consumption was negatively associated with Faecalibac- superoxide dismutase are ROS-scavenging enzymes that directly terium prausnitzii (Supplementary Fig. 17b). There was a positive eliminate non-radical and radical ROS, respectively. RNS metabo- correlation between fiber consumption and the overall abundance lism included enzymes involved in dissimilatory nitrate reduction, of methanogenic archaea, including species from both Methano- nitrite detoxification, regulation of nitrate/nitrite responses and brevibacter and Methanosphaera genera (Supplementary Fig. 18). nitrogen availability, molybdate transport and dimethyl sulfoxide However, in a sensitivity analysis only iron intake remained (DMSO), trimethylamine N-oxide (TMAO), and tetrathionate associated with LGCs in a multivariate regression model (Supple- respiration. RNS and their derivatives, as well as S- and N-oxides mentary Table 2). and tetrathionate, are used by Proteobacteria and other bacteria enriched in LGC as alternative electron acceptors in anaerobic Shifts in microbial functional pathways respiration.49 Consistent with the stark differences in microbial species LPS biosynthesis in LGC subjects was also mainly encoded by composition, key microbial functional pathways were also strongly Bacteroidetes and Proteobacteria (Supplementary Fig. 23). A correlated with gene richness, both in MSM and non-MSM (Fig. 3a). comprehensive list of all functional shifts identified by FishTaco Functional differences between HGC and LGC (Supplementary is available (Supplementary material 3). material 2) indicated that microbiomes of both groups were able to use different energy resources and might provide different Differences in fecal SCFA content metabolic end products to their hosts. We then measured the abundance of acetic, butyric, propionic, Microbiomes from HGC subjects were enriched in genes valeric, isobutyric, and isovaleric acids in stool samples. Overall, involved in methanogenesis; biosynthesis of indole derivatives the molar ratio of acetic:butyric:propionic acids was approximately and carotenoids; and metabolism of pyruvate, fatty acids, 3:1:1, consistent with previous studies.50 Concentrations of butyric glycerolipids, and glycerophospholipids, the latter needed for and propionic acids were significantly higher in LGCs, whereas the biosynthesis of lipid components of the bacterial cell isovaleric acid increased in HGCs (Fig. 4). The acetic/butyric ratio wall. They were also enriched in pathways related to microbial increased in HGC individuals, which could indicate that the growth, replication and protein production, i.e., DNA replication, contribution of acetate to butyrate formation is higher in LGCs. No aminoacyl-tRNA biosynthesis, ribosome, purine, and pyrimidine significant SCFA variations were found between nadir CD4+ T-cell metabolism, and mismatch, base and nucleotide excision categories. Butyric acid levels were significantly correlated with repair. abundance of Roseburia intestinalis and Roseburia inulinivorans, Whereas HGCs were enriched in genes coding for enzymes which are known to be able to generate large amounts of butyrate implicated in glucose, starch, and cellulose utilization, LGC from acetate consumption.47,51 Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 239 Bacterial virulence factors and antimicrobial resistance gut microbial resistome (Supplementary Figs. 24 and 25). By Microbial shifts associated with immune deficiency also implied FishTaco,48 subjects with LGCs were enriched in intracellular toxins increases in bacterial virulence factors as well as changes in the (tccC), and toxins able to damage cell membranes (hlyIII, slo, rtxA1, Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 240 Fig. 4 Fecal short-chain fatty acid (SCFA) levels. a Fecal butyric and propionic acid levels were significantly higher in subjects with low-gene counts (LGC), whereas isovaleric acid levels were higher in subjects with high gene counts (HGC); no significant differences in SCFAs were observed between different nadir CD4+ T-cell count categories. b The acetyc/butyric ratio was significantly higher in HGCs than in LGCs, but was not significantly different by nadir CD4+ T-cell counts. c Butyrate levels were positively correlated with abundance of Roseburia intestinalis and Roseburia inulinivorans, two well-known butyrate producers. Only statistically significant correlations are shown (p < 0.05). d Abundance of R. intestinalis and R. inulinivorans across microbial gene richness values (HGC in green, LGC in orange). *(p < 0.05) Fig. 5 Summary of enzymes and metabolic processes related with ROS/RNS clearance and enriched in subjects with LGCs (gray boxes). Bacteria enriched in LGC reduce ROS and RNS and generate less toxic metabolites that are eventually excreted from the body. Moreover, different Proteobacteria utilize S-oxides and N-oxides as electron acceptors for anaerobic DMSO and TMAO respiration. S-oxides derive from dietary organic sulfides, which are oxidized by ROS. The N-oxide trimethylamine N-oxide (TMAO) is generated by oxidation of diet-derived tertiary amines like trimethylamine (TMA), either by hepatic flavin-containing monooxigenases or by the direct effect of ROS. TMAO concentrations, however, are also high in fish, from which it can be directly absorbed. Plasma TMAO levels have been epidemiologically linked (dashed line) to severe cardiovascular disease and metabolic disorders, which are also prevalent in HIV-1-infected subjects. G6PDH glucose-6- phosphate dehydrogenase; 6PGDH 6-phosphogluconate dehydrogenase; GR glutathione reductase; GPx glutathione peroxidase; Trx thioredoxin reductase; Prx peroxiredoxin; SOD superoxide dismutase, Fe–Mn family; Bfr bacterioferritin; NO nitric oxide; TMA trimethylamine; TMAO trimethylamine oxide; DMSO dimethyl sulfoxide tlyC, and plcC) and the extracellular matrix (cloSI, hya, and NEU1). higher antimicrobial resistance gene copies were observed in Similarly, they were also enriched in a number of enzymes subjects with LGCs, as well as in individuals with progressively involved in antimicrobial resistance (Supplementary Fig. 24), lower nadir CD4+ T-cell counts (Supplementary Fig. 25). While including single antimicrobial resistance genes like Fsr, involved subjects with HGCs were enriched in genes involved in self- in biofilm formation,52 and genes encoding for multidrug efflux antibiotic resistance, which control the overgrowth of other pumps.53 The latter comprises ATP-binding cassette (ABC) bacterial species61 individuals with LGCs were enriched in genes transporters (ABC-2 type permease), and genes from the four associated with beta-lactam, lincosamide, tetracycline, and macro- families of secondary multidrug transporters, i.e., major facilitator lide resistance. superfamily (YebQ, Blt, Mef, and TetA), small multidrug resistance (SMR: TC.SMR, SugE), multidrug and toxic compound extrusion (MATE: TC.MATE), and resistance-nodulation-cell division (EmrAB- DISCUSSION TolC; AcrA-TolC, MdtA, and MacA). Gut dysbiosis was also Using shotgun metagenomics and direct measurements of fecal associated with increases in genes encoding for other bacterial SCFA levels, we found evidence of a strong, dose-dependent survival mechanisms like the multiple peptide resistance factor association between reductions in gut microbial gene rich- (MprF), which confers resistance to antimicrobial peptides,54 and ness,18,19 and nadir CD4+ T-cell counts, which reflected complex metal ion efflux systems like the CusCBA copper efflux complex.55 compositional and functional changes occurring in the gut In a targeted exploration of virulence and resistance factors microbiome alongside HIV-1-associated immune deficiency. The included, respectively, in the virulence factor database (VFDB) observed reductions in oxygen-sensitive, Gram-positive syntrophic and the comprehensive antimicrobial resistance database (CARD) gut microbes like methanogenic archaea, together with increases curated databases, subjects with nadir CD4+ T-cells <100 cells/ in ROS/RNS-resistant generalists, like Bacteroides, Parabacteroides, mm3 had higher counts of important bacterial virulence genes, and Proteobacteria, suggests adaptation of the gut microbial such as catalase, ClpC ATPase,56 Type IV pili,57 type III secretion ecosystem to oxidative stress as a putative, not necessarily system,58 HSI-I,59 and Flagella60 (Supplementary Fig. 26). Similarly, exclusive, driver of such changes. Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 241 HIV-1 infection is a well-established source of oxidative stress. It induced immune damage, being a major surrogate marker of triggers gut inflammation by damaging the intestinal epithelial excess mortality, systemic inflammation, and clinical complications lining and submucosal and lymphoid tissues, allowing enteric in chronically HIV-1-infected subjects.11,93 Increases in genes adenovirus expansion,35,62 NF-kβ activation,63 and release of implicated in LPS biosynthesis, bacterial virulence factors, and cytokines and chemokines,49 arachidonic acid64 and neopterin,65 changes in the gut bacterial resistome, including depletion in which induce migration and accumulation of neutrophils and genes involved in self-antibiotic resistance, and enrichment in macrophages in gut tissues. Antimicrobial ROS and RNS are genes associated with beta-lactam, lincosamide, tetracycline, and produced by the NADPH-oxidase in neutrophils and inducible NO- macrolide resistance in subjects with LGCs, could be related to the synthases in macrophages. Production of free radicals is further increased susceptibility to bacterial infections seen in subjects enhanced by the effect of HIV-1 proteins,66–68 microbial products with severe immune suppression.11,93 like LPS,69,70 and HIV-1-associated mitochondrial dysfunction.71,72 Importantly, the compositional and functional shifts associated Increased residency times and reduced apoptosis of macrophages with gene richness in our study were highly consistent across HIV- in the context of gut dysbiosis extend the production of ROS/RNS 1 transmission groups, ruling out confounding by HIV-1 risk over time.73 The persistent accumulation of free radicals ultimately group,41 and have been described in other diseases characterized overwhelms antioxidant defenses in host cells,74,75 and hampers by gut inflammation,18,19 such as obesity and inflammatory bowel the growth of microbes deficient in ROS/RNS-resistance disease. This suggests that gut microbiome shifts observed in our mechanisms.49,76,77 study are not necessarily unique to HIV-1 infection, but instead, An oxidative environment thus favors the growth of commen- share important characteristics with those seen in other diseases sals with antioxidant capabilities and higher metabolic versati- featuring gut inflammation. Moreover, the existence of gut lity.46,78,79 Bacteroides spp. or Proteobacteria, for example, are microbiome shifts in some but not all HIV-1-infected subjects, ROS/RNS-resistant bacteria able to use a wide range of final particularly in those with higher immune damage in the past, electron acceptors during anaerobic respiration, including nitrate, suggests that gut microbiome testing has the potential to become TMAO, DMSO and tetrathionate.80 In our study, genes encoding a useful clinical tool to identify which individuals are more likely to for enzymes implicated in the utilization of the latter compounds develop gut microbiome-related diseases, including those asso- were enriched in LGC individuals, and were mainly encoded by ciated with inflammaging and accelerated aging, impaired Bacteroides and Proteobacteria (Fig. 5). immune reconstitution, and severe bacterial infections. Prospec- Of note, alongside with pro-inflammatory Proteobacteria like tive cohort studies linking gut microbiome data with AIDS and Sutterella wadsworthensis,79 subjects with LGCs also had increased non-AIDS-related diseases are thus essential to validate and butyrate and propionate levels in direct fecal SCFA measurements. extend our findings. Such studies are critical to evaluate Bacteroides spp. are propionate producers,81 which may explain potentially unmeasured confounders, limit information recall the increased fecal propionate levels. However, butyrate results biases, and include comprehensive information on other drugs might seem more counterintuitive given the anti-inflammatory that might modify the microbiome and might have not been properties of butyrate.82 Total fecal SCFAs measurements must be evaluated herein. In addition, the use of colonic biopsies might interpreted with caution because they neither measure direct provide further insights into the factors that condition gut SCFA production nor colonocyte uptake.83 Overall, our findings microbial shifts during HIV-1 infection. In conclusion, low nadir support the notion that shifts in hydrogenotrophic microbes in CD4+ T-cell counts, rather than HIV-1 serostatus per se, predict the context of exogenous gut inflammation may modify SCFA the presence of gut dysbiosis in HIV-1 infected subjects. None of metabolism in vivo. the compositional and functional shifts associated with such Hydrogenotrophic microbes, i.e., methanogenic archaea, dysbiosis seem to be HIV-specific; instead, they are similar to those sulfate-reducing bacteria and acetogenic bacteria, compete for found in other diseases featuring gut inflammation, suggesting gut microbial hydrogen. Methanogenic archaea are promoted by adaptation to oxidative stress as an important, not necessary fiber consumption84 possibly because it leads to increased exclusive, driver of gut microbial shifts in HIV-1 infection. hydrogen production by primary fiber degraders. In our study, fiber consumption was indeed associated with abundance of methanogenic archaea. Methanogenic archaea, however, are also METHODS highly intolerant to oxygen85 being outcompeted by alternative Study design hydrogenotrophic microbes like acetogenic bacteria in oxidative This was a cross-sectional study in HIV-1-infected subjects with environments. different virologic and immunologic phenotypes and HIV-negative In this study, subjects with LGCs had lower amounts of controls. Comprehensive details of the cohort design and methanogens and sulfate-reducing bacteria. As seen elsewhere,86 characteristics have been published elsewhere.41 Study partici- methanogenic archaea were, indeed, negatively correlated with pants were recruited in Barcelona, Catalonia, Spain, between fecal butyrate concentrations. In contrast, acetogenic bacteria January and December 2014. HIV-1 infected subjects were were more abundant in LGC subjects. Acetogens compete with recruited from the HIV Clinics of the tertiary care hospitals methanogens for hydrogen and are more tolerant to oxidative Germans Trias i Pujol and Vall d’Hebrón. HIV-1-negative controls stress and oxygen fluxes.85,87–89 They produce large quantities of were mainly recruited from a prospective cohort of HIV-negative acetate through acetogenesis,89 which is not only a source of MSM at risk of becoming infected by HIV-1,44 who attend quarterly energy for gut epithelial cells, but also contributes to butyrate medical and counseling visits including HIV-1 testing (Alere formation by butyrate-producing bacteria like Roseburia Determine HIV-1/2 Ag/Ab Combo, Orlando, FL) at, a community- spp.86,90,91 Therefore, we suggest that, in the absence of based center in Barcelona.45 Additional controls were HIV- methanogens, acetogenic bacteria increase their contribution to negative partners from HIV-1-infected subjects attending the HIV interspecific H2 transfer. This increases butyrate production by clinics. acetate-consuming butyrogenic bacteria like Roseburia spp., as The study included participants between 18 and 60 years and evidenced by lower acetate:butyrate ratios in LGCs, also observed BMI within 18.5 and 30. Exclusion criteria were: (a) any gross in previous studies.92 dietary deviation from a usual diet, or any specific regular diet, i.e., The strong association between intestinal microbial shifts and vegetarian, low-carbohydrates, etc.; (b) antibiotic use during the nadir CD4+ counts found in this study may be of importance in previous 3 months (with the exception of late presenters, who HIV medicine. Nadir CD4+ T-cell counts decrease with longer could receive antibiotics to treat opportunistic infections); (c) times since HIV-1 diagnosis and thus reflect the extent of HIV-1- pregnancy or willingness to become pregnant; (d) current drug Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 242 consumption or alcohol abuse; (e) any chronic digestive disease was constructed for each sample. Total fecal DNA was sequenced such as peptic ulcer, Crohn’s disease, ulcerative colitis, or coeliac in an Illumina Hi-Seq® platform at the Institut de Medicina Predictiva disease; (f) any surgical resection of the intestines except for del Cancer (IMPPC), Badalona, Spain. appendectomy; (g) any autoimmune disease; and (h) any symptomatic chronic liver disease or presence of hepatic Sequence quality assessment insufficiency defined as a Child-Pugh C score. Sequence quality was assessed using the FastQC software.97 Total HIV-1-infected subjects were classified into the following seven sequences were filtered by length and quality using Trimmo- mutually excluding categories based on their virological and matic98 ensuring a minimum base quality of Q30 for both leading immunological phenotype: (a) elite controllers: HIV-1 RNA < 50 and trailing bases, a minimum length of 75 basepairs and a copies/mL during at least 2 years in the absence of ART; (b) viremic minimum sequence quality average of Q20 for 30 basepair sliding controllers: HIV-1 RNA between 50 and 2000 copies/mL during at windows across sequences. Human contamination was removed least 2 years in the absence of ART; (c) early-treated: ART initiation by mapping filtered sequences against the human genome using during the first 6 months after HIV-1 infection, HIV-1 RNA levels the bwa99 software and discarding reads that uniquely aligned to <50 copies/mL during at least the 3 last months and with no HIV-1 the human reference with mapping quality higher than Q20. After RNA blips after achieving HIV-1 RNA <50 copies/mL; (d) ART-naïve: filtering, a mean (standard deviation) of 31.0 (9.3) million HIV-1 RNA >10,000 copies/mL, nadir CD4+ T-cell counts >500 sequences were obtained for each individual. cells/mm3 and no ART exposure; (e) immune concordant: HIV-1 RNA levels <50 copies/mL, and CD4+ T-cell counts >500 cells/ Taxonomical classification mm3 during at least 2 years; (f) immune discordant: HIV-1 RNA <50 The Metaphlan v2 software100 was used to infer the microbial copies/mL and CD4+ T-cell <300 cells/mm3 during at least 2 years composition of each stool sample. under ART; and (g) late presenters: CD4+ T-cell counts <200 cells/ mm3 at HIV-1 diagnosis and no ART exposure. Microbial gene richness assessment Filtered sequences were mapped against the integrated gene Data collection catalog (IGC)101 using the bwa software.99 Unique alignments with Clinical and laboratory data were collected in a centralized a minimum quality of Q20 were selected for subsequent analyses. database specifically designed for this study (OpenClinica, 2015 An average of 24.8 million sequences (~80 % of total filtered OpenClinica, LLC) using a standardized questionnaire.41 sequences) mapped against the IGC reference catalog. To be able to compare the microbial gene content across samples we used a Dietary evaluation downsampling size of 10 million aligned sequences (Supplemen- Study participants received a thorough dietary and nutritional tary Fig. 2). assessment by a specialized dietitian/nutritionist using two Gene richness was measured as the total number of different standardized and validated questionnaires: genes present in the sample regardless of their abundance and length, as described previously.18 A minimum of one filtered a. A prospective dietary nutrient survey aimed at recording, as mapped sequence was set to consider the presence of a gene. The precisely as possible, any food, supplement or liquid intake copy number of each gene was estimated by dividing the total during 3–5 consecutive days, including at least one week- reads mapping to a gene divided by the gene’s length. Gene end day. relative abundance was measured as its copy number divided by b. A second questionnaire was a recall of food portions taken the sum of the total gene copies in the sample. As in previous per week, on average, over the last year. To a great extent, studies,18 gene richness followed a bimodal distribution in a this questionnaire is equivalent to the one used by the probability density function. The local minimum of the density Health Professionals Follow-up Study of the Harvard School function for gene richness distribution between the two modes, was of Public Health (http://www.hpfstudy.org),94 but has been considered the threshold value to classify individuals in two groups adapted to include local dietary products and exclude items according to their microbial gene richness content: HGC and LGC. that are almost not consumed at all in our environment (e.g., To rule out possible biases due to the use of an external peanut butter, etc). reference catalog not necessarily enriched in microbiome sequences from PLWH, we also constructed our own microbial After completion of both questionnaires, the study subject gene catalog by de novo assembly and annotation of metage- reviewed them with the study dietitian. Individual nutritional data nomic sequences using the MOCAT2 pipeline.102 This catalog was was standardized by energy intake by fitting a linear model and used to confirm the gene richness categorizations obtained using taking residuals as new nutritional values (mean = 0 and SD = 1).95 the IGC catalog.101 This ensured that amounts of nutrients were comparable between subjects with different total energy intake. A Dirichlet-multinomial Microbial functional profile distribution was assumed for species counts to account for data Genes found in metagenomes downsampled at 10 M sequences overdispersion, as suggested previously.41,96 were associated to one or multiple KEGG categories according to the IGC reference database.101 To measure KEGG abundances, all Fecal DNA extraction genes associated to the same KEGG category were collapsed and Study participants collected fecal samples at home in sterile fecal their copy numbers were summed. If a gene was associated to collection tubes the same day or the day before their medical more than one KEGG category, all categories were considered appointment, before the proctology exam, and following pre- separately. The relative abundance of KEGG categories was specified standard operating procedures. If required, samples were measured as in the relative gene abundance approach detailed stored at 4 °C overnight until DNA extraction. Total DNA contained above. To analyze the metabolic pathways represented in all in aliquots of ~200 mg of each fecal sample was extracted using samples, we fed the HUMAnN pipeline103 with the KEGG copy the PowerSoil DNA Extraction Kit (MO BIO Laboratories, Carlsbad, number table. The HUMAnN output included the relative CA, USA) and then cryopreserved at −80 °C until sequencing. abundances of metabolic modules and pathways for each subject. KEGG functions that were not associated to prokaryotic or Microbial DNA sequencing archaeal metabolism according to KEGG catalog information104 Whole fecal DNA was fragmented with the Nextera-XT® Illumina were manually filtered out, to ensure that only microbial functions kit. One library of approximately 300-basepair-clone insert size were analyzed. Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 243 Contribution of microbial species to functional shifts approach based on ecological distance matrices calculated by Bidirectional imbalances in microbial enzyme abundance between Bray–Curtis dissimilarities, as implemented R (Vegan, metaMDS HGC and LGC groups were identified using the FishTaco and ggplot2 packages). NMDS ellipses were drawn based on a software,48 which also estimates the contribution of specific confidence interval (CI) of 0.95, and only species associated to an microbial species to such imbalances. The genomic content of the environmental vector with a length higher than 0.35 were member species was inferred de novo by coupling the initial represented. taxonomic and functional abundance profiles. Taxonomic profiles The association between baseline variables and the HGC and were generated considering only microbial species present in at LGC categories was evaluated using the two-tailed Fisher’s exact least 5% of all patient samples. A false discovery rate threshold q test or the Wilcoxon rank sum test, as needed. To determine < 0.05 was used to correct for multiple comparisons. Functional factors independently associated with LGC, we first fitted a unbalances of interest were plotted using the FishTacoPlot R univariate logistic regression model including each baseline package. variable showing a significantly different distribution between HGC and LGC. Variables significantly associated with LGC in Microbial virulence factors and antimicrobial resistance univariate regression were then included in a multivariate logistic To further characterize the presence of bacterial virulence factors regression analysis. If two variables were redundant or collinear and antimicrobial resistance genes in subjects with and without (e.g., MSM vs. non-MSM and gender) only one of them was dysbiosis, we mapped 10 M randomly selected, quality-filtered, included in the multivariate regression analysis. Odds ratios and sequence reads from each sample against the VFDB105 and the 95% CIs were used to evaluate the weight of each variable relative CARD,106 respectively. Gene copy number was measured as to its reference. described above. Correlations between gene richness and other continuous variables, including microbial species, metabolic functions, and Soluble markers of enterocyte integrity, bacterial translocation, specific microbial enzymes were evaluated with the Spearman and systemic inflammation rank correlation test. Differences between HGC and LGC groups in Plasma samples were collected and cryopreserved at −80 °C until abundance of microbial species, KEGG counts, antimicrobial quantification. Six different soluble plasma markers were mea- resistance and bacterial virulence genes were evaluated using a sured in all study participants. Plasma levels of markers of Wilcoxon rank sum test. Differences in microbial species, functions enterocyte damage (IFABP), microbial translocation (soluble CD14 and enzymes by nadir CD4+ T-cell count categories were (scCD14) and LPS binding protein (LBP)) and systemic inflamma- evaluated using a Kruskal–Wallis test with post hoc comparison tion (interleukin-6 (IL-6), C-reactive protein (CRP), and interferon- tests as required. In all previous evaluations the gamma-inducible protein-10 (IP-10)) were analyzed using com- Benjamini–Hochberg correction was applied to avoid type I error mercial ELISA kits in accordance with manufacturer’s instructions inflation due to multiple comparisons. A random forest analysis [IL-6: high sensitivity ELISA with signal amplification, BMS213HS, was used to determine the relative contribution of each microbial eBioscience (San Diego, CA); sCD14: 0014-27 (ng/mL), Diaclone species associated with richness to the classification of subjects as (Besançon, France); IFABP: DuoSet ELISA development system (pg/ HGC or LGC. Area-under-the-curve values were used to compare mL), R&D systems (Minneapolis, MN); CRP: KHA0031 (mg/mL), the classification accuracy of the different predictors in both Invitrogen (Carlsbad, CA); IP-10: 1P10-09 (pg/mL), Diaclone machine-learning approaches. (Besançon, France); LBP: KA0448 (ng/mL), Abnova (Taipei, As in previous publications from our group,41 nutrient Taiwan)]. If plasma levels were above the upper detection limit, consumption was standardized by energy intake by fitting a the concentration considered was the maximum quantifiable linear model and taking residuals as new nutritional values. value, i.e., the highest concentration included in the standard Dirichlet-multinomial regression107 was implemented to identify curve. the strongest relationships between nutritional standardized nutrient intake and species abundance and variance, penalizing Fecal short-chain fatty acid levels weak associations using a penalized likelihood approach.41 In Acetic, butyric, propionic, valeric, isobutyric, and isovaleric acids addition, nutrients independently correlated with gene richness concentrations were measured in 146 stool samples. Approxi- were identified using Spearman correlation or Wilcoxon mately ~100 mg of stool were suspended in sterile acidified water Mann–Whitney nonparametric tests, as needed. (pH 2.5) at a final concentration of 100 mg stool/mL and vigorously vortexed at room temperature for 5 min until the stool Ethics and community involvement was completely homogenized. The homogenate was then The study was reviewed and approved by the Institutional Review centrifuged at 13 K RPM for 3 min and transferred 500 µL of the Boards of the Hospital Universitari Germans Trias i Pujol (reference supernatant into a fresh microcentrifuge tube. Then 30 µL of 8 mM PI-13-046) and the Hospital Vall d’Hebrón (reference PR(AG)109/ 2-methylvaleric acid (Sigma Aldrich) was added in ethyl acetate as 2014). All participants provided written informed consent in an internal standard and 500 µL of ethyl acetate to the super- accordance with the World Medical Association Declaration of natant and vigorously vortexed this mixture for 10 min. These Helsinki. The study concept, design, patient information and extractions were centrifuged at 13 K RPM for 3 min and transferred results were discussed with the HIVACAT Community Advisory 300 µL of the upper organic phase to a glass autosampler vial. The Committee (CAC), who provided input on these aspects as well as SCFA were then analyzed on an HP6890 GC (Agilent) using a on the presentation and dissemination of study results. Stabilwax column (30 m x 0.32 mmID x 0.25 µm df, Restek) and 1 µL splitless injections. The oven temperature was as follows: 100 °C Sequence and data availability for 2 min then increased by 25 °C/min to 200 °C for 1 min followed Raw Illumina MiSeq sequences and study metadata were by 250 °C for 1 min. The detector was an HP5973 MSD (Agilent). deposited in the National Center for Biotechnology Information - We prepared standard solutions of acetic, propionic, isobutyric, NCBI repository (Bioproject accession number: PRJNA307231, SRA butyric, isovaleric, and valeric acids in acidified water and treated accession number: SRP068240). them as described above. Statistical analyses of the microbiome composition and function ACKNOWLEDGMENTS Microbiome samples were clustered according to their species We thank the members of the IGTP Genomics and Bioinformatics Core Facilities composition using a nonmetric multidimensional scaling (NMDS) (Maria Pilar Armengol, Lauro Sumoy and Iñaki Martínez) for their contribution to this Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 244 publication. This study was supported by philanthropic donations from Fundació 13. Douek, D. C. Disrupting T-cell homeostasis: how HIV-1 infection causes disease. Glòria Soler, Fundació Catalunya-La Pedrera, the Gala SIDA 2015 and 2016 editions, AIDS Rev. 5, 172–177 (2003). the Nit per la Recerca a la Catalunya Central 2015 edition, and People in Red – 14. Sandler, N. G. & Douek, D. C. Microbial translocation in HIV infection: causes, Barcelona 2016 edition. IrsiCaixa is supported by the RED de SIDA RD16/0025/0041 consequences and treatment opportunities. Nat. Rev. Microbiol. 10, 655–666 cofinanced by the ISCIII and the European Regional Development Fund (ERDF), (2012). “Investing in your future”. M.R. is funded through a FI-DGR grant (FI-B00184) from 15. Claesson, M. J. et al. Gut microbiota composition correlates with diet and health Agència de Gestió d’Ajuts Universitaris i de Recerca (AGAUR) at the Secretaria in the elderly. Nature 488, 178–184 (2012). d’Universitats i Recerca del Departament d’Economia i Coneixement de la Generalitat 16. Tang, W. H. W. et al. Intestinal microbial metabolism of phosphatidylcholine and de Catalunya. M.L.C. is funded through the grant MTM2015-64465-C2-1-R, Spanish cardiovascular risk. N. Engl. J. Med. 368, 1575–1584 (2013). Ministry of Economy and Competitiveness, Spain. Funding for this research 17. Koeth, R. A. et al. Intestinal microbiota metabolism of l-carnitine, a nutrient in (Alexander S. Zevin and Nichole R. Klatt) was provided by the Department of red meat, promotes atherosclerosis. Nat. Med. 19, 533–534 (2013). Pharmaceutics at the University of Washington. The sponsors of the study had no 18. Chatelier, E. Le et al. Richness of human gut microbiome correlates with role in study design, data collection, data analysis, data interpretation, or writing of metabolic markers. Nature 500, 541–546 (2013). the report. The corresponding author had full access to all study data, and had final 19. Cotillard, A. et al. Dietary intervention impact on gut microbial gene richness. responsibility for the decision to submit for publication. Nature 500, 585–588 (2013). 20. Shoaie, S. et al. Quantifying diet-induced metabolic changes of the human gut microbiome. Cell Metab. 22, 320–331 (2015). AUTHOR CONTRIBUTIONS 21. Klatt, N. R., Chomont, N., Douek, D. C. & Deeks, S. G. Immune activation and HIV R.P., Y.G., M.N., J.B., N.K., and B.C. conceived and designed the study. R.P., I.B., B.M., persistence: implications for curative approaches to HIV Infection. Immunol. Rev. E.N., J.Co., J.S., A.T., J.N., C. B., M.Cr., and B.C recruited the study participants and 254, 326–342 (2013). performed their clinical evaluations. C.E. performed the dietary assessment. G.S. and 22. Lozupone, C. A. et al. HIV-induced alteration in gut microbiota: driving factors, J.Co. performed the proctology studies. C.H. coordinated the study logistics including consequences, and effects of antiretroviral therapy. Gut Microbes 5, 562–570 the fulfillment of all ethical and legal requirements of the study as member of the (2014). Contract Research Organization overseeing the study, in coordination with R.P. Fecal 23. Brenchley, J. M. et al. CD4+ T-cell depletion during all stages of HIV disease DNA was extracted, amplified and sequenced by M.P, M.Ca., and M.R. under the occurs predominantly in the gastrointestinal tract. J. Exp. Med. 200, 749–759 supervision of M.N. and R.P. Y.G, J.R., M.R., and C.R. performed the bioinformatic and (2004). statistical analyses, with the supervision of M.N., M.L.C., and R.P. M.Ca. performed the 24. Morais, D. et al. Microbiota-dependent sequelae of acute infection compromise inflammation analyses under the supervision of J.Ca., J.B, and R.P. A.Z. and N.K tissue-specific immunity article microbiota-dependent sequelae of acute infec- performed the SCFA determinations. R.P. and Y.G. wrote the paper, which was tion compromise tissue-specific immunity. Cell 163, 354–366 (2015). reviewed, edited and approved by all authors. 25. Althoff, K. N., Smit, M., Reiss, P. & Justice, A. C. HIV and ageing: improving quantity and quality of life. Curr. Opin. HIV AIDS (2016). https://doi.org/10.1097/ COH.0000000000000305 26. Corbeau, P. & Reynes, J. Review article Immune reconstitution under antiretroviral ADDITIONAL INFORMATION therapy: the new challenge in HIV-1 infection. Blood 117, 5582–5591 (2017). The online version of this article (https://doi.org/10.1038/s41385-018-0083-7) 27. Massanella, M., Negredo, E., Clotet, B. & Blanco, J. Immunodiscordant responses contains supplementary material, which is available to authorized users. to HAART—mechanisms and consequences. Expert Rev. Clin. Immunol. 9, 1135–1149 (2013). Competing interests: The authors declare no competing interests. 28. Kilbourne, A. M. et al. General medical and psychiatric comorbidity among HIV- infected veterans in the post-HAART era. J. Clin. Epidemiol. 54(Suppl 1), S22–S28 (2001). REFERENCES 29. Sudano, I. et al. Cardiovascular disease in HIV infection. Am. Heart J. 151, 1. Antinori, A. et al. Late presentation of HIV infection: a consensus definition. HIV 1147–1155 (2006). Med. 12, 61–64 (2011). 30. Guaraldi, G. et al. Premature age-related comorbidities among HIV-infected 2. Mocroft, A. et al. Risk factors and outcomes for late presentation for HIV-positive persons compared with the general population. Clin. Infect. Dis. 53, 1120–1126 persons in Europe: results from the Collaboration of Observational HIV Epide- (2011). miological Research Europe Study (COHERE). PLoS Med. 10, e1001510 (2013). 31. Klatt, N. R. et al. Vaginal bacteria modify HIV tenofovir microbicide efficacy in 3. Paiardini & Müller-Trutwin, M. HIV-associated chronic immune activation. African women. Science 356, 938–945 (2017). Immunol. Rev. 254, 78–101 (2013). 32. Gosmann, C. et al. Lactobacillus-deficient cervicovaginal bacterial communities 4. Deeks, S. G. HIV infection, inflammation, and aging. Annu. Rev. Med. 62, 141–155 are associated with increased HIV acquisition in young south african women (2011). report lactobacillus-deficient cervicovaginal bacterial communities are asso- 5. Battegay, M., Fluckiger, U., Hirschel, B. & Furrer, H. Late presentation of HIV- ciated with increased HIV acquisition. Immunity 46, 29–37 (2017). infected individuals. Antivir. Ther. 12, 841–851 (2007). 33. Costea, P. I. et al. Enterotypes in the landscape of gut microbial community 6. Ingle, S. M. et al. Impact of risk factors for specific causes of death in the first and composition. Nat. Microbiol. 3, 8–16 (2018). subsequent years of antiretroviral therapy among HIV-infected patients. Clin. 34. Rothschild, D. et al. Environment dominates over host genetics in shaping Infect. Dis. 59, 287–297 (2014). human gut microbiota. Nature 555, 210–215 (2018). 7. The INSIGHT START Study Group. Initiation of antiretroviral therapy in early 35. Monaco, C. L. et al. Altered virome and bacterial microbiome in human asymptomatic HIV infection. N. Engl. J. Med. 373, 795–807 (2015). immunodeficiency virus-associated acquired immunodeficiency syndrome. Cell 8. May, M. T. et al. Mortality according to CD4 count at start of combination Host Microbe 19, 311–322 (2016). antiretroviral therapy among HIV-infected patients followed for up to 15 years 36. Williams, B., Landay, A. & Presti, R. M. Microbiome alterations in HIV infection a after start of treatment: Collaborative Cohort Study. Clin. Infect. Dis. 62, review. Cell Microbiol. 18, 645–651 (2016). 1571–1577 (2016). 37. Williams, B. et al. A summary of the first HIV microbiome workshop 2015. AIDS 9. Connor, J. O. et al. Effect of immediate initiation of antiretroviral therapy on risk Res. Hum. Retroviruses 32, 935-941 (2016). of severe bacterial infections in HIV-positive people with CD4 cell counts of 38. Ling, Z. et al. Alterations in the fecal microbiota of patients with HIV-1 infection: more than 500 cells per μL: secondary outcome results from a randomised an observational study in a Chinese population. Sci. Rep. 6, 30673 (2016). controlled trial. Lancet HIV 3018, 4–11 (2017). 39. Vázquez-Castellanos, J. F. et al. Altered metabolism of gut microbiota con- 10. Robbins, G. K. et al. Incomplete reconstitution of T-cell subsets on combination tributes to chronic immune activation in HIV-infected individuals. Mucosal antiretroviral therapy in the AIDS Clinical Trials Group protocol 384. Clin. Infect. Immunol. 8, 760–772 (2015). Dis. 48, 350–361 (2009). 40. Mutlu, E. A. et al. A compositional look at the human gastrointestinal micro- 11. Negredo, E. et al. Nadir CD4 T-cell count as predictor and high CD4 T-cell biome and immune activation parameters in HIV-infected subjects. PLoS Pathog. intrinsic apoptosis as final mechanism of poor CD4 T-cell recovery in vir- 10, e1003829 (2014). ologically suppressed HIV-infected patients: clinical implications. Clin. Infect. Dis. 41. Noguera-Julian, M. et al. Gut microbiota linked to sexual preference and HIV 50, 1300–1308 (2010). infection. EBioMedicine 5, 135–146 (2016). 12. Brenchley, J. M. et al. Microbial translocation is a cause of systemic immune 42. Dillon, S. M., Frank, D. N. & Wilson, C. C. The gut microbiome and HIV-1 activation in chronic HIV infection. Nat. Med. 12, 1365–1371 (2006). pathogenesis: a two-way street. AIDS 30, 2737–2751 (2016). Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 245 43. Vujkovic-Cvijin, I. et al. Dysbiosis of the gut microbiota is associated with HIV 71. Arunagiri, C., Macreadie, I., Hewish, D. & Azad, A. A. C- terminal domain of HIV-1 disease progression and tryptophan catabolism. Sci. Transl. Med. 5, 193ra91 accessory protein Vpr is involved in penetration, mitochondrial dysfunction and (2014). apoptosis of human CD4+ lymphocytes. Apoptosis 2, 69–76 (1997). 44. Coll, J. et al. Early detection of HIV infection and of asymptomatic sexually 72. Ivanov, A. et al. Oxidative stress during HIV infection: mechanisms and con- transmitted infections among men who have sex with men. Clin Microbiol sequences. Oxid. Med. Cell Longev. 2016, 1–18 (2016). Infect. 24, 540-545 (2018). 73. Kanther, M. et al. Commensal microbiota stimulate systemic neutrophil migra- 45. Meulbroek, M. et al. BCN Checkpoint, a community-based centre for men-who- tion through induction of serum amyloid A. Cell Microbiol. 16, 1053–1067 (2015). have-sex-with-men in Barcelona, Catalonia, Spain, shows high efficiency in HIV 74. Flores, S. C. et al. Tat protein of human immunodeficiency virus type 1 represses detection and linkage to care. HIV Med. 14, 25–28 (2013). expression of manganese superoxide dismutase in HeLa cells. Proc. Natl Acad. 46. Barcenilla, A. et al. Phylogenetic relationships of butyrate-producing bacteria Sci. USA 90, 7632–7636 (1993). from the human gut. Appl. Environ. Microbiol. 66, 1654–1661 (2000). 75. Sandstrom, P. A., Roberts, B., Folks, T. M. & Buttke, T. M. HIV gene expression 47. Duncan, S. H., Barcenilla, A., Stewart, C. S., Pryde, S. E. & Flint, H. J. Acetate enhances T-cell susceptibility to hydrogen peroxide- induced apoptosis. AIDS utilization and butyryl coenzyme A (CoA): acetate-CoA transferase in butyrate- Res. Hum. Retrovir. 9, 1107–1113 (1993). producing bacteria from the human large intestine. Appl. Environ. Microbiol. 68, 76. Bäumler, A. J. & Sperandio, V. Interactions between the microbiota and patho- 5186–5190 (2002). genic bacteria in the gut. Nature 535, 85–93 (2016). 48. Manor, O. & Borenstein, E. Systematic characterization and analysis of the 77. Lemaitre, B. & Hoffmann, J. The Host Defense of Drosophila melanogaster. Annu. taxonomic drivers of functional shifts in the human resource systematic char- Rev. Immunol. 25, 697–743 (2007). acterization and analysis of the taxonomic drivers of functional shifts in the 78. Smith, C. J., Rocha, E. R. & Paster, B. J. The medically important bacteroides spp. human microbiome. Cell Host Microbe. 21, 1–14 (2017). in health and disease. Prokaryotes 381–427 (2006). https://doi.org/10.1007/0- 49. Winter, S. E., Lopez, C. A. & Bäumler, A. J. The dynamics of gut-associated 387-30746-x microbial communities during inflammation. EMBO Rep. 14, 319–327 (2013). 79. Mukhopadhya, I. et al. A comprehensive evaluation of colonic mucosal isolates 50. Canfora, E. E., Jocken, J. W. & Blaak, E. E. Short-chain fatty acids in control of body of sutterella wadsworthensis from inflammatory bowel disease. PLoS ONE 6, weight and insulin sensitivity. Nat. Rev. Endocrinol. 11, 577–591 (2015). 1–10 (2011). 51. Bergey, D. H., Buchanan, R. E. & Gibbons, N. E. Bergey’s Manual of Systematic 80. Winter, S. E. & Bäumler, A. J. A breathtaking feat: to compete with the gut Bacteriology. (Springer-Verlag, New York, NY, 1974). microbiota, salmonella drives its host to provide a respiratory electron acceptor. 52. Ahmed, M., Lyass, L., Markham, P. N., Taylor, S. S. & Neyfakh, A. A. Two highly Gut Microbes 2 (2011). similar multidrug transporters of Bacillus subtilis whose expression is differen- 81. Reichardt, N. et al. Phylogenetic distribution of three pathways for propionate tially regulated. J. Bacteriol. 177, 3904–3910 (1995). production within the human gut microbiota. ISME J. 8, 1323–1335 (2014). 53. Fernández, L. & Hancock, R. E. W. Adaptive and mutational resistance: role of 82. Canani, R. B. et al. Potential beneficial effects of butyrate in intestinal and porins and efflux pumps in drug resistance. Clin. Microbiol. Rev. 25, 661–681 extraintestinal diseases. World J. Gastroenterol. 17, 1519–1528 (2011). (2012). 83. Cummings, J. & Macfarlane, G. The control and consequences of bacterial fer- 54. Andra, J., Goldmann, T., Ernst, C. M., Peschel, A. & Gutsmann, T. Multiple peptide mentation in the human colon. J. Appl. Bacteriol. 70, 443–459 (1991). resistance factor (MprF)-mediated resistance of Staphylococcus aureus against 84. Fernandes, J., Wolever, T. M. S. & Rao, A. V. Interrelationships between age, total antimicrobial peptides coincides with a modulated peptide interaction with dietary fiber intake and breath methane in humans. Nutr. Res. 20, 929–940 artificial membranes comprising lysyl-phosphatidylglycerol. J. Biol. Chem. 286, (2000). 18692–18700 (2011). 85. Jarrell, K. F. Extreme oxygen sensitivity in methanogenic archaebacteria. 55. Rensing, C. & Grass, G. Escherichia coli mechanisms of copper homeostasis in a Bioscience 35, 298–302 (1985). changing environment. FEMS Microbiol. Rev. 27, 197–213 (2003). 86. Abell, G. C. J., Conlon, M. A. & Mcorist, A. L. Methanogenic archaea in adult 56. Neckers, L. & Tatu, U. Review molecular chaperones in pathogen virulence: human faecal samples are inversely related to butyrate concentration. Microb. emerging new targets for therapy. Cell Host Microbe 4, 519–527 (2008). Ecol. Health Dis. 18, 154–160 (2006). 57. Craig, L., Pique, M. E. & Tainer, J. A. Type IV pilus structure and bacterial 87. Karnholz, A., Küsel, K., Gößner, A., Schramm, A. & Drake, H. L. Tolerance and pathogenicity. Nat. Rev. Microbiol. 2, 363–378 (2004). metabolic response of acetogenic bacteria toward oxygen. Appl. Environ. 58. Coburn, B., Sekirov, I. & Finlay, B. B. Type III secretion systems and disease. Clin. Microbiol. 68, 1005–1009 (2002). Microbiol. Rev. 20, 535–549 (2007). 88. Dolla, A., Fournier, M. & Dermoun, Z. Oxygen defense in sulfate-reducing bac- 59. Mougous, J. D. et al. A virulence locus of pseudomonas. Science 1526–1530 teria. J. Biotechnol. 126, 87–100 (2006). (2013). https://doi.org/10.1126/science.1128393 89. Drake, H. L. et al. Acetogenic bacteria: what are the in situ consequences of their 60. Rossez, Y., Wolfson, E. B., Holmes, A., Gally, D. L. & Holden, N. J. Bacterial diverse metabolic versatilities? Biofactors 6, 13–24 (1997). Flagella: twist and stick, or dodge across the kingdoms. PLoS Pathog. 11, 1–15 90. Duncan, S. H. et al. Contribution of acetate to butyrate formation by human (2015). faecal bacteria. Br. J. Nutr. 91, 915 (2004). 61. Hopwood, D. A. How do antibiotic-producing bacteria ensure their self- 91. Chassard, C. & Bernalier-Donadille, A. H2 and acetate transfers during xylan resistance before antibiotic biosynthesis incapacitates them? Mol. Microbiol. 63, fermentation between a butyrate-producing xylanolytic species and hydro- 937–940 (2007). genotrophic microorganisms from the human gut. FEMS Microbiol. Lett. 254, 62. Handley, S. A. et al. Pathogenic simian immunodeficiency virus infection is 116–122 (2006). associated with expansion of the enteric virome. Cell 151, 253–266 (2012). 92. Serrano-Villar, S. et al. The effects of prebiotics on microbial dysbiosis, butyrate 63. Fiume, G. et al. Human immunodeficiency virus-1 Tat activates NF-??B via production and immunity in HIV-infected subjects. Mucosal Immunol. 10, 1279- physical interaction with IκB-ɑ and p65. Nucleic Acids Res. 40, 3548–3562 (2012). 1293 (2017). 64. Serrano-Villar, S. et al. Gut bacteria metabolism impacts immune recovery in 93. Bray, S. et al. Predictive value of CD4 cell count nadir on long-term mortality in HIV-infected Individuals. EBioMedicine 8, 203–216 (2016). HIV-positive patients in Uganda. HIV AIDS 4, 135–140 (2012). 65. Baier-Bitterlich, G., Fuchs, D. & Wachter, H. Chronic immune stimulation, oxi- 94. Wu, G. D. et al. Linking long-term dietary patterns with gut microbial enter- dative stress, and apoptosis in HIV infection. Biochem. Pharmacol. 53, 755–763 otypes. Science 334, 105–108 (2011). (1997). 95. Willett, W., Howe, G. & Kushi, L. Adjustment for total energy intake in epide- 66. Miller, R. A. & Britigan, B. E. Role of oxidants in microbial pathophysiology. Clin. miologic studies. Am. J. Clin. Nutr. 4 Suppl, S1220–S1231 (1997). Microbiol. Rev. 10, 1–18 (1997). 96. la Rosa, P. S. et al. Hypothesis testing and power calculations for taxonomic- 67. Vilhardt, F. et al. The HIV-1 Nef protein and phagocyte NADPH-oxidase activa- based human microbiome data. PLoS ONE 7, 1–13 (2012). tion. J. Biol. Chem. 277, 42136–42143 (2002). 97. Andrews, S. FastQC: a quality control tool for high throughput sequence data. 68. Mangino, G. et al. HIV-1 myristoylated nef treatment of murine microglial cells (2010) http://www.bioinformatics.babraham.ac.uk/projects/fastqc activates inducible nitric oxide synthase, NO2 production and neurotoxic 98. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina activity. PLoS ONE 10, 1–21 (2015). sequence data. Bioinformatics 30, 2114–2120 (2014). 69. Cassatella, M. A. et al. Molecular basis of interferon-g and lipopolysaccharide 99. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows–Wheeler enhancement of phagocyte respiratory burst capability. J. Biol. Chem. 265, transform. Bioinformatics 25, 1754–1760 (2009). 20241–20246 (1990). 100. Truong, D. T. et al. MetaPhlAn2 for enhanced metagenomic taxonomic profiling. 70. Jacob, B. A. et al. HIV-1-induced pulmonary oxidative and nitrosative stress: Nat. Methods 12, 902–903 (2015). exacerbated response to endotoxin administration in HIV-1 transgenic mouse 101. Li, J. et al. An integrated catalog of reference genes in the human gut micro- model. Am. J. Physiol. Lung Cell. Mol. Physiol. 291, L811–L819 (2006). biome. Nat. Biotechnol. 32, 834–841 (2014). Mucosal Immunology (2019) 12:232 – 246 Low nadir CD4+ T-cell counts predict gut dysbiosis in HIV-1 infection Y Guillén et al. 246 102. Kultima, J. R. et al. Genome analysis MOCAT2: a metagenomic assembly, Open Access This article is licensed under a Creative Commons annotation and profiling framework. Bioinformatics 32, 2520–2523 (2016). Attribution 4.0 International License, which permits use, sharing, 103. Abubucker, S. et al. Metabolic reconstruction for metagenomic data and adaptation, distribution and reproduction in any medium or format, as long as you give its application to the human microbiome. PLoS Comput. Biol. 8, e1002358 appropriate credit to the original author(s) and the source, provide a link to the Creative (2012). Commons license, and indicate if changes were made. The images or other third party 104. Kanehisa, M. & Goto, S. KEGG: Kyoto encyclopedia of genes and genomes. material in this article are included in the article’s Creative Commons license, unless Nucleic Acids Res. 28, 27–30 (2000). indicated otherwise in a credit line to the material. If material is not included in the 105. Chen, L., Zheng, D., Liu, B., Yang, J. & Jin, Q. VFDB 2016: hierarchical and refined article’s Creative Commons license and your intended use is not permitted by statutory dataset for big data analysis—10 years on. Nucleic Acids Res. 44, 694–697 regulation or exceeds the permitted use, you will need to obtain permission directly (2016). from the copyright holder. To view a copy of this license, visit http://creativecommons. 106. Jia, B. et al. CARD 2017: expansion and model-centric curation of the compre- org/licenses/by/4.0/. hensive antibiotic resistance database. Nucleic Acids Res. 45, 566–573 (2017). 107. Chen, J. & Li, H. Variable selection for sparse Dirichlet-multinomial regression with an applicaton to microbiome data analysis. Ann. Appl. Stat. 7 (2013). © The Author(s) 2018 Mucosal Immunology (2019) 12:232 – 246