Enrichment of gut-derived Fusobacterium is associated with suboptimal immune recovery in HIV-infected individuals

We explored the gut microbiota profile among HIV-infected individuals with diverse immune recovery profiles following long-term suppressive ART and investigated the relationship between the altered bacteria with markers of immune dysfunction. The microbiota profile of rectal swabs from 26 HIV-infected individuals and 20 HIV-uninfected controls were examined. Patients were classified as suboptimal responders, sIR (n = 10, CD4 T-cell <350 cells/ul) and optimal responders, oIR (n = 16, CD4 T-cell >500 cells/ul) after a minimum of 2 years on suppressive ART. Canonical correlation analysis(CCA) and multiple regression modelling were used to explore the association between fecal bacterial taxa abundance and immunological profiles in optimal and suboptimal responders. We found Fusobacterium was significantly enriched among the HIV-infected and the sIR group. CCA results showed that Fusobacterium abundance was negatively correlated with CD4 T-cell counts, but positively correlated with CD4 T-cell activation and CD4 Tregs. Multiple linear regression analysis adjusted for age, baseline CD4 T-cell count, antibiotic exposure and MSM status indicated that higher Fusobacterium relative abundance was independently associated with poorer CD4 T-cell recovery following ART. Enrichment of Fusobacterium was associated with reduced immune recovery and persistent immune dysfunction following ART. Modulating the abundance of this bacterial taxa in the gut may be a viable intervention to improve immune reconstitution in our setting.

In addition to microbial translocation-mediated immune activation, the composition of intestinal bacteria and its metabolites may also play a key role in HIV immunopathogenesis [13][14][15] . The diversity and composition of gut commensals are altered in men who have sex with men (MSM) 16 with HIV-infection and also following the commencement of ART 17,18 . This HIV-associated dysbiosis, assessed from fecal and mucosal-associated samples, has been shown to correlate with markers of immune activation including blood and colonic CD4 and CD8 T-cell activation markers 15,19,20 , levels of hsCRP 15 , sCD14 15,[17][18][19]21,22 , TNFalpha 21,22 and IL-6 21 . Interventional studies in HIV and SIV-infection have suggested that modulation of gut microbiota using dietary pro/prebiotics may be a viable approach to reduce HIV-related immune activation [23][24][25] . On the other hand, a recent meta-analysis on the impact of probiotics on CD4 T-cell counts reported limited benefits 26 . More recently, fecal microbial transplant (FMT) was used as a strategy to modulate gut microbiota in HIV 27 . In that study, only limited engraftment was found in ART-suppressed individuals following FMT, while no change was seen in systemic immune activation levels 8 weeks post-FMT. This result contradicts an earlier report of favorable FMT outcome in SIV-infected macaques 28 . The mixed results suggest gaps in our understanding of the complex interaction between commensal gut bacteria, HIV infection and its pathogenesis.
More than 95% of the global HIV epidemic occurs in developing countries where socio-economic status and environmental factors such as diet and endemic parasitic infections may significantly influence the composition of gut commensals 29,30 . Yet to date, only a few studies outside of the USA and Europe have examined how differences in gut microbiota may influence HIV immunopathogenesis [31][32][33] . To this end, our objectives were to explore the changes in gut microbiota profile of individuals in Malaysia (a middle income country) with diverse immune recovery outcomes following long-term suppressive ART, and explore the influence of altered bacterial taxa on markers of immune activation and CD4 T-cell composition following ART.

Results
Characteristics of study population. Table 1 summarises the demographic and clinical characteristics of the 26 HIV-infected and 20 HIV uninfected controls who participated in this study. Among the HIV-infected participants, 62% (16/26) were classified oIR and 38% (10/26) sIR. Participants in the sIR arm were older than oIR and uninfected controls (median (IQR): 49 (41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52) years, 40 (34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45) years and 31 (28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46) years, respectively). All participants were males. There was no significant difference in the proportion of MSM in the HIVinfected (optimal and suboptimal) and uninfected control groups (p > 0.05 for all paired comparisons). Apart from expected significant differences in baseline and current CD4 T-cell counts as well as CD4:CD8 ratio between the oIR and sIR groups, all other HIV-related clinical variables and hepatitis B/C co-infection status were comparable in the two groups. Different microbial profile in HIV optimal, suboptimal and uninfected controls. Illumina Miseq sequencing at the V4 gene region of bacterial 16S rRNA resulted in 2,756,344 high quality reads with an average of 59,920 ± 106,599 reads per subject. The profile of the most dominant taxa was similar in both the uninfected and the HIV optimal groups, with an abundance of Firmicutes (median relative abundance: 65.3% vs. 80.3%, p = 0.886), followed by Bacteroidetes (29.2% vs. 16.4%, p = 0.364) and Actinobacteria (3.5% vs. 2.4%, p = 0.494) (Fig. 1a). However, in the sIR arm, there was a dominance of Fusobacteria (11.0%), which was not obvious in the other two arms. Further analysis of this bacterial taxa in the entire cohort showed a significant increase in the relative abundance of Fusobacteria in participants from the sIR compared to oIR and uninfected controls ( Fig. 1b; Kruskal-Wallis test (KW): p = 0.003 with Dunn's test: p = 0.002), implying the potential importance of this taxa in the pathogenesis of suboptimal CD4 T-cell reconstitution following ART in this cohort. We note that there was a wide variability in the relative abundance of Fusobacterium within the suboptimal group. However, we could not find any obvious association between any of the measured clinical features or biomarkers in our cohort for this sub-structuring. The relative abundance of phyla and genus in each subject is shown in Supplementary Fig. S2.
Although the microbial composition between the HIV-infected and uninfected groups differed significantly as indicated by beta diversity using unweighted UniFrac distance metric (p < 0.05, pseudo-F statistic = 1.55; Supplementary Fig. S3a and b), alpha diversity, which measures species richness and evenness, was however similar between these groups (p > 0.05). When assessed by immune recovery status, no significant differences in both beta and alpha diversity were found (p > 0.05 for both; Supplementary Fig. S3c and d, Supplementary Fig. S4a-c).

Alteration of specific bacterial taxa in relation to HIV-infection and immune recovery status.
The LEfSe analysis identified several differentially abundant microbial taxa (Supplementary Fig. S5a and b; Supplementary Table S1) in the HIV-infected and uninfected groups which included the phylum Fusobacteria and its associated taxonomic levels: class Fusobacteria, order Fusobacteriales, family Fusobacteriaceae and genus Fusobacterium. There was also a notable increase in the abundance of the family Veilonellaceae 14,34 and a depletion of the genus Eubacterium 21,35 , as previously described in other HIV studies.
We further analysed the differences in the abundance of microbiota between oIR and sIR individuals. The phylum Fusobacteria and its descendant genus Fusobacterium were again enriched in the sIR group (Fig. 1c,d; Supplementary Table S3) while the bacterial taxa belonging to the family Corynebacteriaceae and its genus Corynebacterium were significantly depleted in the sIR arm together with the order Lactobacillales. We also compared the relative abundance of Fusobacterium in the different immune recovery groups against the overall mean of the relative abundance of Fusobacterium measured in the whole cohort (overall mean relative abundance = 0.0458). In this analysis, we found that the majority of sIR (70.0%; 7/10) had a higher relative abundance of Fusobacterium, whilst only 25.0% (4/16) and 10.0% (2/20) of the oIR and uninfected groups, respectively displayed an increase in the relative abundance of Fusobacterium.
To better understand the relative importance of the different bacterial taxa in modulating CD4 T-cell recovery, we restricted the subsequent analysis to bacteria which were found altered at the lowest taxonomic level and  Table S3). Lactobacillales and Fusobacterium had the strongest correlations with CD4 T-cell counts (CL > 0.50 for both taxa). CD4 T-cell count was negatively correlated with Fusobacterium relative abundance (p = 0.002, R = −0.58) (Supplementary Fig. S5c) but not with Corynebacterium and Lactobacillales relative abundance (data not shown), thus implying a more complex relationship between CD4 T-cell count and these bacterial taxa. We further analysed the association between the relative abundance of Lactobacillales and Fusobacterium on CD4 T-cell counts using multiple linear regression models while adjusting for selected confounders including baseline CD4 T-cell counts, age, antibiotics exposure and MSM status. We found that Fusobacterium (βˆ = −0.64, p = 0.001) was significantly associated with poorer CD4 T-cell recovery following ART (Table 2).
Using CCA, we also explored the association between Lactobacillales and Fusobacterium relative abundance with various markers of immune dysfunction found to be significantly different in the optimal and suboptimal

Discussion
The relationship between gut microbiota and immune recovery following long-term suppressive ART is still poorly understood. This is the first study to explore this association in a developing country setting, where environmental factors that differ from those in high income countries may influence the gut commensal community composition differently. In this study, we showed that specific alterations in bacterial taxa composition occur in individuals with HIV-infection, and in individuals who experience suboptimal immune recovery (sIR) following ART. Poor immune reconstitution was most strongly associated with an increased abundance of Fusobacterium, even after adjusting for clinical co-variates previously shown to independently influence immune reconstitution and gut microbiota composition. Additionally, increased relative abundance of Fusobacterium was also significantly correlated with reduced naïve CD4 T-cells, increased CD4 T-cell activation and increased regulatory T-cells measured in peripheral blood, all markers of persistent immune dysfunction in HIV infection. These data suggest that changes in gut microbiota composition specifically the enrichment of Fusobacterium, may contribute to the immunopathogenesis of poor CD4 T-cell reconstitution following ART in our setting. Consistent with prior reports of altered gut microbiota profile in HIV-infected individuals 15,21,32 , we also found an increase in the abundance of bacteria from the families of Fusobacteriaceae, Succinivibrionaceae and Micoplasmataceae in the HIV-infected compared to the uninfected group 14,15,21,[35][36][37] . When HIV-infected individuals were further divided according to their immune recovery profile following a minimum of 2 years of suppressive ART, there was an increased relative abundance of Fusobacterium, Gallicola and Bilophila, in the suboptimal group compared to the optimal recovery group. While microbial diversity has been reported to be lower in HIV-infected (vs. uninfected) 17,34,35 and suboptimal (vs. optimal) CD4 T-cell recovery 16,17 groups, it was not observed to be statistically significant in the present study. Beta diversity as measured using the unweighted UniFrac distance was significantly associated with HIV status, but not immune recovery. Thus, significant associations in gut microbiota composition were observed in relation to HIV infection and immune discordance following ART in our setting, consistent with previous studies 15,21,32 .
Notably, the genus Fusobacterium from the Fusobacteria phylum was significantly more abundant in individuals with suboptimal vs. optimal CD4 T-cell recovery as well as uninfected controls. The association between the relative abundance of this taxa and reduced CD4 T-cell counts remained statistically significant even after adjusting for the effects of age, baseline CD4 T-cell counts, MSM status and antibiotic usage. We also measured markers of immune dysfunction in the HIV-infected participants and explored the association of these cellular markers with the altered bacterial taxa in the optimal and suboptimal recovery groups. Consistent with prior studies, sub-optimal CD4 T-cell recovery in our cohort was associated with a reduction in naive T-cell subsets, increased T-cell activation and increased T regulatory cells 38 . In canonical correlation analysis, the genus Fusobacterium again had the strongest modulating effect on this immunological profile followed by the order Lactobacillales and the genus Corynebacterium. Previous cross-sectional studies reported an increased abundance of Fusobacteria in the rectal/gut and oral microbiota of individuals with untreated 35,39 and treated HIV-infection 21 , but here, we show that abundance is also associated with the patient's dysfunctional immune profile. In a longitudinal study among HIV-infected MSM, a significant increase in the abundance of Fusobacteria was noted as individuals developed more severe immune deficiency while on zidovudine  37 . Our findings corroborate previous observations and further suggest that Fusobacteria enrichment in individuals receiving stable ART in a developing country setting is clinically relevant. Future studies and identification of specific strains by culture isolation or qPCR are needed to establish cause-effect relationships between these bacteria and immune function in HIV-infected individuals.
Fusobacterium is a butyric acid-producing bacteria. Currently, indication of its pathogenic role in HIV has largely been confined to its increased abundance in the oral cavity, and its association with HIV-associated periodontitis 40 . In non-HIV related studies, increased gut Fusobacterium has been associated with the pathogenicity of inflammatory disorders, including inflammatory bowel disease [41][42][43] , acute appendicitis 44 and colorectal cancer 45,46 , suggesting a role in regulating local gut immunity. In mechanistic studies, Fusobacterium has been shown to be a strong promoter of NF-kb activation 46 and an inhibitor of histone deacetylases (HDAC), with the potential to reactivate HIV from latently infected cells in vivo 47 . Thus, enrichment of Fusobacterium in the gut could contribute to HIV-1 re-activation of latently infected cells in GALT, further compromising the integrity of the gut mucosal barrier. Fusobacterium also possesses independent virulent factors that may directly induce apoptotic cell death in polymorphonuclear cells (PMNs) 48 . These pathways could lead to increased local and systemic immune activation, subsequently increasing bystander CD4 T-cell death, thus potentially explaining the inability of these individuals to achieve normalised peripheral CD4 T-cell thresholds despite years of suppressive ART. We observed higher levels of CD8 and CD4 T-cell activation in individuals with suboptimal CD4 T-cell recovery, but only the latter was correlated with the abundance of Fusobacterium. A high turnover of CD4 T-cells could also lead to increased homeostatic proliferation of naïve CD4 T-cells in circulation, leading to a reduced proportion of this subset in favour of expanded memory CD4 T-cell subsets, as observed in our cohort.
We also found an increase in circulating Tregs in individuals with suboptimal immune recovery, which correlated with the observed alterations in microbiota composition in this population. Both mucosal and peripheral Tregs have been found to retain their suppressive function during chronic HIV infection 49 , and have been associated with reduced HIV-specific immunity especially in individuals with low CD4 T-cell counts 50 . No study to date has described an association between Fusobacterium and Tregs in HIV-infected individuals, though a prior study in SIV-infected rhesus macaques has linked the depletion in gut-resident Lactobacillales to an increase in IDO (indoleamine 2,3-dioxygenase) activity 20 , which is known to promote Treg differentiation in circulation and in rectosigmoidal tissues 51 . A previously reported reduction in the abundance of Lactobacillales 18,31 was also observed in the present cohort and associated with an increase in Tregs in canonical correlation analysis. However, reduced Lactobacillales was not correlated with increased plasma IDO activity in our study (data not shown). The lack of correlation could potentially be explained by the presence of other IDO mediators including IFN-α levels, which we recently showed to remain elevated despite years of suppressive ART in our setting 52 . The association between reduced Lactobacillales and CD4 T-cell counts was also weaker when compared to Fusobacterium in our cohort and was lost when both bacterial taxa were included in a multiple linear regression model. Multiple pathways may contribute to the loss of CD4 T-cells in treated HIV-infection and thus it is conceivable that several bacterial taxa may contribute to this process by different and/or inter-related mechanisms 13 . Future in vitro studies should further explore if there is a symbiotic relationship between the depletion of protective Lactobacillales and the enrichment of pathogenic Fusobacterium which could potentiate the loss of CD4 T-cells in HIV.
The strength of this study is in the characteristics of the cohort which included individuals with distinct immune recovery profiles despite years of suppressive ART. Additionally, all participants were men who received the same NNRTI-based regimen; MSM status, which has been shown to significantly affect the gut microbial composition, was recorded and accordingly adjusted for in our analysis. Several caveats necessarily accompany our interpretation of the present observational study. First, dietary patterns peculiar to the participants may have influenced the observed outcome, but no information about this aspect was obtained from the participants. Secondly, our choice of using rectal swabs with fecal material instead of biopsy samples for sampling gut microbiota composition was mainly motivated by logistical considerations, which nonetheless may potentially bias the result. Thirdly, the biological effects attributed to the microbiota found in this study are likely driven by their metabolites, the analysis of which is beyond the scope of this study. We suggest future work to include a component of metabolomics and meta-transcriptomics as data from these two lines of enquiry are essential to our understanding of the underlying mechanisms associated with altered microbiota in HIV.
In conclusion, we found the composition of the gut microbiota of individuals receiving suppressive ART to be significantly different from those without HIV-infection. Specific bacterial taxa were associated with impaired CD4 T-cell reconstitution and persistent immune dysfunction following ART. Interventions that modulate the composition of gut microbiome, particularly Fusobacterium, may be a useful approach to augment immune reconstitution in individuals with suboptimal CD4 T-cell recovery despite years of suppressive ART in our setting.

Methods
Study design and participants. Participants for this study were identified from a previous cross-sectional study established to study functional immune response following suppressive ART at University Malaya Medical Centre (UMMC) 53 . Briefly, individuals fulfilling the following inclusion criteria were recruited during their routine clinic appointments: (1) men >18 years old; (2) receiving suppressive ART (HIV RNA <50 copies/mL) for a minimum of 2 years; (3) CD4 T-cell counts persistently <350 cells/µL (suboptimal immune recovery, sIR) or >500 cells/µL (optimal immune recovery, oIR) on suppressive ART; and (4) no acute illness at recruitment. In addition, a total of 20 HIV uninfected controls were recruited from a local community-based clinic and individuals responding to study fliers.
All participants and controls provided informed consent and answered a standardised questionnaire while clinical parameters for the HIV-infected participants were retrieved from medical records. The study protocol was approved by the UMMC ethics review board (Permit Number: MEC 20161-1992). All procedures contributing to this research were conducted in accordance with the relevant guidelines and regulations.
SCientifiC REpoRtS | (2018) 8:14277 | DOI:10.1038/s41598-018-32585-x 16s rRNA sequencing of gut microbiota. DNA was extracted from rectal swabs as previously described 53 and was PCR-amplified targeting the V4 region of the 16S rRNA gene using a protocol modified from Caporaso et al. 54 , details of which have been previously described 55 . Amplicon sequencing was subsequently performed on the Illumina MiSeq platform with a 2 × 150 cycle run (Illumina, San Diego CA, USA).
Gut microbiota bioinformatic analysis. Quantitative Insights Into Microbial Ecology (QIIME) software version 1.9.1 was used for the pre-processing steps, and for the downstream analysis (i.e. taxa summary plot, alpha diversity and beta diversity) of gut microbiota sequencing data 56 (details in Supplementary Methods).
To identify bacterial taxa that are statistically different between the groups of interest, Geometric Mean of Pairwise ratios (GMPR) normalization 57 was used to overcome the sparse and compositional effects on the raw count data. Then, the LDA Effect Size (LEfSe) analysis was performed using the online interface Galaxy (https:// huttenhower.sph.harvard.edu/galaxy/) 58 . The threshold for statistical significance using the Kruskal-Wallis rank sum test was set at the 5% level of significance, a threshold of 3 was used for the logarithmic LDA score. Correction for multiple testing was done using the Benjamini-Hochberg procedure 59 .

Soluble activation markers. Whole blood was collected in Ethylenediaminetetraacetic acid (EDTA)
vacutainers and processed within 4 hours of collection. Soluble CD14 was measured in plasma using the Quantikine sCD14 kit (R&D Systems, USA) according to the manufacturer's instructions as previously described 60 . Kynurenine and tryptophan concentrations were measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS, Agilent, 6400 series) as previously described 61 . The HPLC conditions used were as follows: the column was a Synergi Polar RP column (75 mm × 4.6 mm) from Phenomenex (CA, USA) and the mobile phase was composed of 0.1% formic acid in ultrapure water and a mixture of methanol and acetonitrile (7% v/v) with gradient program. The flow rate was set at 1.0 ml/min and the run time for each sample was 6 min.
Statistical Analyses. The chi-squared test was used to test associations between categorical variables (i.e. between smoking status and control/HIV infected group). Where the result of chi-squared test was uncertain (expected cell count below 5), we used Fisher's exact test. The non-parametric Mann-Whitney test was used to test whether the median of the continuous predictor variables differed significantly between control/HIV subgroups. Correlation between the bacterial taxa of interest and markers of immune activation and CD4 T-cell maturational subsets in HIV-infected participants was evaluated using Spearman's rank correlation. Statistical significance was declared when p-value < 0.05. Canonical correlation analysis (CCA) was used to discover associations between two sets of variables, one consisting of CD4 T-cell count and CD4 T-cell subsets/markers of activation, and another consisting of selected bacterial taxa that differ significantly in relative abundance between the optimal and the suboptimal responders as obtained from LEfSe analysis. After identifying the correlated variables from CCA, multiple linear regression was used to retest the association between the relative abundance of the specific bacterial taxa and CD4 T-cell counts controlling for the effect of age, baseline CD4 T-cell counts, MSM status and antibiotic use. All statistical analyses were performed using R (Version 3.3.3) 62 .

Data Availability Statement
The datasets generated during and/or analysed during the current study are available in NCBI BioProject Repository (Accession Number: PRJNA489590).