Inside Out: HIV, the Gut Microbiome, and the Mucosal Immune System

Concerted work during the last three decades has demonstrated that the trillions of microbes (bacterial, viral, and fungal) that call the human body home are not just inert passengers, but are active participants in the immune processes affecting our overall health. Our microbial tenants have a profound influence on host biology as well as on disease processes, impacting many fields of medicine, including infectious disease, gastroenterology, pulmonology, immunology, rheumatology, cardiology, endocrinology, geriatrics, and even oncology. This review focuses on alterations of the gut microbiome (termed “dysbiosis”) that occur with HIV infection and the associated disruption of gut homeostasis and subsequent clinical consequences.

The study of commensal bacteria traditionally has been limited to culturable organisms. However, the revolution in molecular phylogenetics initiated in the 1970s by Woese and Fox (1, 2) coupled with Norman Pace’s application (3) in the 1980s of then-new molecular biological techniques such as PCR and DNA/RNA sequencing to studies of microbial ecology ushered in the current era of culture-independent microbiology (3–9). High-throughput sequencing of marker genes (most notably the 16S and 18S rRNA genes of bacteria/archaea and eukaryotes, respectively), deep shotgun sequencing of mixed community DNA/RNA, metabolomics, and other ’omics technologies now allow the study of complex microbial ecosystems associated with the human body in much greater detail and without the need to first propagate axenic cultures (10–12).

The goal of rRNA gene analysis is centered on identifying and enumerating the types of organisms occurring in a microbial community (“who is there”), whereas the other ’omics technologies seek to understand the functional capacity and molecular outputs of the community (“what are they doing”) (10, 13). Small subunit (SSU) gene analyses were generally applied first and most frequently to questions of dysbiosis and HIV. Several highly conserved regions are found scattered across the SSU gene, permitting the design of broad-range PCR primer sets with universal (i.e., three domain) or pan-bacterial specificities. These primers allow one to PCR amplify the sequence genes from a broad range of microorganisms. Following SSU gene sequencing, significant bioinformatics processes are required to generate interpretable data. Although insertion into a well-supported phylogenetic tree is the gold standard for inferring the type of organism from which an SSU sequence was obtained, the sheer volume of sequences generated by next-generation sequencing platforms raises significant computational challenges in this regard. Therefore, most modern SSU analysis pipelines employ heuristic approaches to assign taxonomic names to sequences or closely related groups of sequences. In any case, the size of the reference database used, as well as the length and position of the SSU fragment sequenced (most next-generation sequencers cannot generate full-length sequences, but instead focus on particular variable regions), will place fundamental limitations on the accuracy of the sequence classification process: some regions of the SSU gene may not encode sufficient information to infer species-level classifications. Furthermore, microbial species names assigned on the basis of classical biochemical and morphological measurements may not map precisely onto the tree of life, which is derived from SSU sequences. Consequently, molecular sequence data often are used to define relatedness groups termed operational taxonomic units (OTUs) on the basis of sequence dissimilarity–based clustering (numerical OTUs) or taxonomic binning (taxonomic OTUs).

In addition to enumerating OTUs, sequencing data can be used to describe the biodiversity of one or more microbial communities (14). Alpha diversity describes the biodiversity within an individual microbial community (i.e., a single specimen) and can be measured by different indices: richness simply counts the number of OTUs in a community, whereas its overall complexity can be quantified by Shannon or Simpson diversity indices. In contrast, β diversity measures describe the overlap in distributions of OTUs between two or more communities and are commonly presented as the dissimilarity between the microbiomes of two samples. Indices such as Bray-Curtis, Jaccard, and UniFrac can all be used to infer β diversity (14, 15). UniFrac, for instance, measures the distances between communities based on lineages they contain; it can be used as a distance metric and can be visualized on a principal coordinate plot. However, each index differs with respect to its weighting of rare versus abundant taxa, presence/absence versus abundance measures, and consideration of phylogenetic differences among taxa. Furthermore, β diversity indices form the basis of statistical tests, such as permutation-based ANOVA (PERMANOVA) tests, that assess whether a potential predictor variable (e.g., disease occurrence) is statistically significant (16). More exploratory techniques not described in this review involving ordination (e.g., nonmetric multidimensional scaling), hierarchical clustering, and a variety of other machine learning algorithms also can be applied to OTU data, as well as to other ’omics datasets.

The gut microbiota plays a critical role in gut homeostasis, which in turn influences a wide variety of clinical outcomes. In general, gut bacteria help in the processing of nutrients either through digestion of insoluble fibers, breaking down of nutrients, or by providing important metabolites that influence nutrition. On a micromolecular level, some of the microbial organisms of the gut produce important precursors to small molecules such as trimethylamine-N-oxide (TMAO), which has been associated with cardiovascular disease (17), and short-chain fatty acids (SCFAs), which are not only a source of energy for enterocytes but also may have important immunomodulatory effects on the gut mucosal immune system (18). The taxonomic composition of the microbiome is important in and of itself, as commensal organisms have a role in regulating the development and activation of the lamina propria T cell compartments (19–22). Additionally, certain components of the Bacteroidetes and Proteobacteria phyla can be considered as pathobionts (defined as commensal bacteria that have pathogenic potential in the appropriate clinical context) (23). On a clinical level, some cross-sectional studies have demonstrated differences in the gut microbiota in obese individuals compared with lean individuals, whereas a case control study identified differences at the phylum level between individuals who developed diabetes compared with healthy controls (reviewed in Ref. 24). In older populations, cross-sectional studies have associated differences in the gut microbiome with early onset frailty, physical debility, and neurocognitive dysfunction (25). Finally, the human gut is a critical site of host-microbe interaction, and any process that perturbs either side of that equation will have profound clinical and gut homeostatic implications. HIV infection may affect both sides by altering the components of the gut microbiome and by changing the host immune response to gut microbes.

The human gut bacterial microbiome consists primarily of four phyla: Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria (26). Composition of the gut microbiome varies significantly depending on socioeconomic factors, age, geography, and diet (27, 28). For example, in children from Burkina Faso, Africa, a diet high in fiber (normal for that region) appears to enrich the microbiome at the phylum level, with Bacteroidetes enriched at the expense of Firmicutes and Enterobacteriaceae compared with European children (27). Enterotype, or the general makeup of the gut microbiome, appears to be shared among people sharing the same dwelling (29). At the genus level there is a relatively robust association between the plant- and carbohydrate-based diet of agrarian cultures and an enterotype with a low Bacteroides/Prevotella ratio (28, 30). Conversely, an enterotype with a relatively high Bacteroides/Prevotella ratio is associated with the animal product–based Western diet, which is high in protein and saturated fat.

The extent to which HIV infection and resultant immune compromise alter the gut microbiome remains controversial. Controlling for the myriad factors that influence gut microbiota, including the factors mentioned above and other likely confounders such as antibiotic use, antiretroviral drugs themselves, and sexual practices, is quite difficult given the relatively small sample sizes generally available for study of the human microbiome. No well-controlled studies to date have evaluated the gut microbiota of the same individuals before and after infection with HIV. Studies using SIV-infected nonhuman primates have far more control over these factors and are able to obtain samples before and after experimental infection with SIV. Wild chimpanzees, which develop progressive immunodeficiency with SIV infection, develop instability of their gut microbiota following SIV infection (31). This is in contrast to wild gorillas, which maintain a stable gut microbiome following SIV infection, although it is unknown whether gorillas develop progressive CD4 depletion from SIV. Therefore, this discrepancy may be related to loss of intestinal immune surveillance in chimpanzees but not gorillas (32). However, rhesus macaques have repeatedly been shown to have an enteric microbiome that is not substantially altered by SIV infection, except temporarily during acute infection (33–35). Despite long-term stability of the gastrointestinal microbiome in macaques, microbial translocation in advanced disease occurs with the phylum Proteobacteria, which was found in lymph nodes and liver and was overrepresented in the stool in the acute phase of SIV infection (levels diminish with chronic infection) (36).

Studies involving HIV-infected human cohorts, despite their inherent limitations, have found a number of reproducible differences in the gut microbiome between HIV-infected and uninfected individuals. Note that the literature is heterogeneous with regard to the type of subjects studied (treated versus untreated), type of sampling method used (rectal sponge, stool swab, stool, tissue site biopsy), and the level of bacterial classification examined. The microbiome differences observed using various sampling methods have been investigated by several groups. Early work by Zoetendal et al. (37) in HIV-negative individuals found that the bacteria associated with mucosal biopsy samples differed greatly from those found in feces from the same individual. Dillon et al. (38) likewise observed differences in HIV-associated changes in taxa abundance depending on the sampling method. Specifically, an increased abundance of the phylum Proteobacteria in HIV-infected individuals was only noted with mucosal samples, suggesting that luminal samples may miss potentially pathogenic alterations. However, HIV-associated changes in the relative abundance of the genera and families in the phylum Bacteroidetes were generally similar across all sampling platforms. Furthermore, abundance of certain taxa (e.g., Prevotella sp.) in mucosal samples strongly associated with markers of mucosal and systemic inflammation, whereas Prevotella abundance in fecal samples did not. This suggests that the mucosa-associated microbiome may more closely interact with and potentially influence the immune system. The gut microbiome also seems to vary depending on segment of the gut, with samples from the terminal ileum and right colon best differentiating between HIV-positive subjects and HIV-negative controls (39). Unfortunately, there does not appear to be consensus regarding which type of sampling should be used routinely in studies (see Table I), and this limits the ability to compare findings across studies.

In developed nations, where the great majority of human microbiome studies have been performed, studies have variously reported that HIV-infected individuals have significantly greater abundance of Prevotella and/or significantly less Bacteroides at the genus level than do uninfected controls (38–44). This finding was observed in cohorts of treated and untreated HIV-infected persons and in both stool and mucosal biopsy samples. However, not all microbiome studies of treated and untreated HIV-infected individuals observed this phenomenon in rectal sponge specimens or stool (45–47). Although Nowak et al. (47) did not find overabundance of the genus Prevotella in their Swedish HIV-positive cohort at baseline, they did note that Prevotella decreased following antiretroviral therapy (ART). In the largest study to date examining the gut microbiome of HIV-infected individuals, Noguera-Julian et al. (48) found, at the genus level, that a high Prevotella/low Bacteroides enterotype in stool specimens was highly associated with men who have sex with men (MSM) behavior regardless of HIV-1 infection status, which may explain the perceived association between this enterotype and HIV infection status in prior studies that did not control for sexual behavior. A study of stool samples from Uganda, where HIV transmission is predominantly heterosexual, seems to support the hypothesis that a Prevotella-predominant enterotype is driven by sexual behavior, as it did not find Prevotella predominance to be dependent on HIV-1 serostatus or level of immunosuppression (49). However, note that the HIV-1–seronegative Ugandans had Prevotella-predominant stool microbiota at baseline, possibly due to dietary differences, so the study may have been underpowered to detect a small HIV-associated increase in Prevotella abundance.

However, enrichment of the phylum Proteobacteria and reciprocal diminishment of the phylum Firmicutes in HIV-infected subjects, particularly in mucosal samples, appears to be more consistently reported in the literature (39, 43, 44, 50). In the study by Dinh et al. (45), Proteobacteria and several subtaxa, including Enterobacteriaceae, which contains many common pathogens, were overrepresented in HIV-positive individuals and were associated with immune activation. This association between HIV infection and increased abundance of Proteobacteria, particularly in mucosal samples, may be more meaningful than the Prevotella/Bacteroides shift given the tendency of Protoebacteria to translocate in the nonhuman primate model (36).

Supporting the role of HIV-1 in driving dysbiosis is the finding that individuals on virally suppressive ART tend to have a microbiome shifted closer to that of uninfected controls compared with individuals with untreated HIV infection, but ART was not associated with complete normalization of the microbiome in stool and rectal sponge samples (40, 46). The microbiome findings in treated versus untreated HIV-infected individuals suggest that ongoing mucosal inflammation driven by unchecked viral replication may largely drive the observed microbiome changes. Conversely, the partial “normalization” of the microbiome observed with virus suppression on ART would tend to suggest that as the mucosal immune system recovers, it is better able to manage the gut environment and allow for a more normal microbiome. However, not all studies have found a stool microbiome shift toward HIV-negative controls following ART (47). Although published data are limited, antiretrovirals themselves may have a direct effect on commensal bacteria and/or the phage viruses that infect them, which may in part explain persistent microbiome differences between long-term–treated HIV-infected individuals versus seronegative controls (36).

Loss of fecal microbial diversity has been noted in the settings of Clostridium difficile colitis and inflammatory bowel disease (51, 52). Several authors have found a similar loss of diversity associated with HIV infection status (39, 46–48, 53). However, an association between HIV infection and decreased microbial diversity has not been consistently observed; in fact, several studies reported no loss of diversity, and one found increased diversity in varying HIV-infected patient cohorts (38, 40, 43, 45). Importantly, MSM sexual behavior itself was associated with increased microbial diversity, a finding that suggests that failure to control for sexual behaviors could confound the diversity analyses in some studies.

Anelloviruses, which are ssDNA viruses, are highly prevalent in stool and blood in the general population but have not been implicated as causative agents of disease (54). Early virome studies found that plasma levels of anellovirus DNA increase in AIDS and other immunocompromised conditions and then decline again following immunologic recovery (54). More recently, the potential role of these viruses in the pathogenesis of immune activation has been investigated. Anelloviruses such as torque teno virus have been associated with progression of liver disease in HIV/hepatitis C virus coinfection but do not appear to be driving chronic T cell activation in HIV (55, 56). In a nonhuman primate model, Handley et al.(34) found a >10-fold increase in stool viral sequences following pathogenic SIV infection including adenoviruses and picornaviruses. Adenoviruses in particular were associated with enteritis and progressive immune dysfunction (34). Nonpathogenic SIV infection in African green monkeys did not result in expansion of the stool virome (34). A human study from Uganda found that both anelloviruses and adenoviruses were increased in those with low CD4 counts (57, 58). Future studies should investigate what role, if any, these viruses play in disrupting the epithelial barrier integrity of the gut, thereby contributing to microbial translocation.

HIV infection is associated with a chronic inflammatory state as represented by increased circulating soluble TNFRs 1 and 2, IL-6 (59), and markers of T cell activation (CD38 and HLA-DR expression) (60). Additionally, HIV infection is associated with increased plasma markers for microbial translocation/monocyte activation (LPS and soluble CD14 [sCD14]) and epithelial barrier damage (e.g., intestinal fatty acid–binding protein) (61). Immune activation and gut barrier disruption are highest in acute infection and fall with chronic infection, but ART decreases them further (61). Some authors, however, have noted that these plasma markers of inflammation remain higher in treated, virally suppressed HIV patients (62, 63) or spontaneous viremic suppression compared with HIV-negative controls (64, 65). However, some studies have demonstrated resolution of microbial translocation with long-term ART (66) and have reported normal levels of monocyte activation in long-term nonprogressors (67).

On a clinical level, elevated systemic markers for inflammation, microbial translocation, and epithelial barrier damage have been linked to poor clinical outcomes even in treated populations (62). Sandler et al. (68) showed in 2011 that sCD14 was an independent predictor of mortality in HIV infection. Timmons et al. (69) analyzed levels of LPS and sCD14 in study participants with and without HIV infection and correlated them to body mass index, lipid panels, and insulin resistance. sCD14 in particular negatively correlated with high-density lipoprotein, especially in those HIV-infected individuals on ART (69). Note, however, that negative correlation did not account for class of ART, which is known to affect lipid levels as well. Tenorio et al. (70) showed that higher soluble markers of inflammation (soluble TNFR1, IL-6) correlated with non-AIDS–defining events such as stroke, cardiovascular disease, and non-AIDS cancers. Finally, Hunt et al. (63) showed that epithelial barrier dysfunction as measured by peripheral blood levels of intestinal fatty acid–binding protein and zonulin-1 predicted mortality in HIV infection, even after adjustment for CD4 count.

From a pathophysiological perspective, there have been multiple explanations proposed for this phenomenon, including a direct effect of viral replication and immune activation caused by coinfections (such as CMV). However, one theory focuses on the effect of HIV infection and resultant dysbiosis on the gut mucosal barrier and immune system. The implications of a leaky gut barrier are 2-fold: systemic inflammation is increased due to circulation of microbial components in the bloodstream (71), and there is an increase in the exposure of the resident gut mucosal T cell population to new Ags. The etiology of this gut barrier dysfunction in HIV infection may originate in the gut lamina propria and its resident CD4 T cells. Lamina propria T cells may be more susceptible to HIV infection due to high levels of activation and expression of HIV receptors such as CCR5 (72). Experiments in SIV nonhuman primate models revealed rapid and massive depletion of lamina propia CD4 cells as soon as 7 d after infection (73, 74). In humans, >50% of lamina propria CD4 cells are depleted in early and acute HIV infection (75), with a selective targeting of Th CD4 cells that produce IL-17 and IL-22. The mechanism of this depletion is likely cell death of productively infected cells via apoptosis as well as of bystander cells via pyroptosis (76, 77). The density of mucosal CD4 cells does not always fully recover despite effective ART and viral suppression (78, 79). Poor CD4 cell recovery leads to important functional consequences for the gut mucosal barrier, especially failure to protect against invading pathogens as well as loss of cytokines necessary to support normal barrier function (80, 81). Poor CD4 reconstitution has also been found to correlate with tight junction dysfunction between the epithelial cells of the mucosal barrier (82). Additionally, immunohistochemistry of gut biopsies in HIV-infected immune nonresponders with poor CD4 recovery despite ART revealed decreased epithelial cell proliferation and increased neutrophil infiltration (a surrogate for epithelial barrier dysfunction) (83). These changes were correlated with increased circulating sCD14, a systemic marker of LPS-induced monocyte activation and indirectly of microbial translocation.

Th22 and Th17 cells along with a subset of innate lymphocyte cells (i.e., type 3 innate lymphoid cells) are responsible for production of IL-22, a critical cytokine for epithelial barrier maintenance that induces stem cell–mediated epithelial cell proliferation in the gut (84). Disruption of IL-22 production may worsen epithelial barrier dysfunction and increase microbial translocation. Indeed, compared with age-matched controls, multiple groups have shown depletion of Th22 cells in gut mucosa of HIV-infected individuals, and one group showed an association of Th22 cell depletion with markers of microbial translocation (50, 80, 81, 85). Note that not all investigators endorse the loss of Th22 cells as the primary cause of epithelial barrier dysfunction, instead pointing to type 3 innate lymphoid cells as most important for maintenance of epithelial barrier integrity during HIV infection (80). To address the effect of HIV-associated dysbiosis on Th22 expression, a recent small pilot study of fecal microbiota transplants (FMTs) from healthy macaques to SIV-infected macaques demonstrated increased frequencies of peripheral Th22 cells after transplantation and decreased gut mucosal inflammatory markers, although it did not comment on gut mucosal Th22 cell frequencies (86).

The components of the microbiome can directly contribute to CD4 cell depletion in HIV infection. Several recent publications have reported on the use of in vitro modeling to characterize interactions between commensal microbes and primary human lamina propria immune cells with regard to how they influence HIV-1 replication and depletion of lamina propria CD4 T cells. Exposure to heat-killed Escherichia coli (a pathobiont) resulted in enhanced activation, proliferation, and HIV-1 infection of CD4⁺ T cells within lamina propia mononuclear cell (LPMC) cultures (87). In subsequent studies, Dillon et al. (88) were able to show that mucosal bacteria that were altered in abundance in colon biopsies of HIV-infected individuals also enhanced HIV replication when cocultured with LPMCs in vitro. Increased CD4 infection following exposure to commensal bacteria appeared to be due in part to increased expression of the HIV-1 coreceptor CCR5, induced predominantly by exposure to Gram-negative bacteria. When highly expressed Gram-negative (LPS) and Gram-positive (lipoteichoic acid) cell wall components were tested, only LPS enhanced CD4 expression of CCR5 (88). The mechanisms of death of gut CD4 cells by HIV and the impact of bacterial exposure on cell death have also been investigated. Steele et al. (76) reported that pyroptosis was an important HIV-associated mechanism of bystander (non–productively infected) CD4 T cell killing in LPMCs in addition to the apoptosis caused by direct viral infection. Furthermore, apoptotic death of CD4 T cells was increased in HIV-infected LPMC cultures following exposure to pathobiont bacteria. These findings may explain the enhanced inflammatory signature seen in HIV gut enteropathy, as pyroptosis ultimately results in release of cytoplasmic contents, including IL-1β, which causes increased permeability in intestinal epithelial cell tight junctions (89) and may ultimately contribute to epithelial barrier dysfunction. Increased mucosal HIV-1 replication and apoptotic CD4 T cell death following bacterial exposure may help to explain how microbial translocation contributes to viral pathogenesis in the early stages of HIV infection.

In addition to their contribution to the selective targeting and depletion of CD4 cells by HIV virus, the altered gut microbiome may have further effects on gut Th subset–related immune homeostasis. This is because commensal bacteria have a role in regulating the normal balance between regulatory T cells (Tregs), which are anti-inflammatory in nature, and Th17 cells, which produce the proinflammatory cytokine IL-17 and regulate host responses to commensal and pathogenic microbiota. One study in germ-free mice demonstrated that segmented filamentous bacteria (a gut commensal most often found in the terminal ileum) were shown to be necessary for the development of Th17 cells (20). Additionally, Clostridium species likely have a role in the accumulation of Tregs in mouse gut (19, 21). Bacteroides fragilis, a prominent anaerobic commensal, is thought to inhibit CD4 differentiation into Th17 and increase differentiation into Tregs in mouse gut (22). The importance of HIV infection–related dysbiosis in depletion of Th17 cells was highlighted (albeit indirectly) by studies in SIV-infected macaques that showed improvement of the frequency of polyfunctional Th17 cells (defined in the study as Th17 cells capable of producing IL-17 and IFN-γ) with probiotic/prebiotic treatment plus IL-21 supplementation (90, 91). These findings are currently being investigated in human studies (AIDS Clinical Trials Group 5352), which will further elucidate this mechanism.

Finally, gut microbiota may be associated with improvement in the CD4 count, which continues to be an important prognostic indicator and predicts non-AIDS events and mortality in addition to AIDS-associated morbidity and mortality (57, 92–96). As described above, poor CD4 recovery is linked to microbial translocation, although it is unclear whether there is a causal association and, if so, in which direction (97). Several authors have investigated the association between various gut bacterial taxa and peripheral CD4 recovery. As with previously noted microbiome associations, findings have not always been confirmed, and, importantly, note the taxonomic level at which the association was found. Abundance of the genus Bacteroides in stool and colonic biopsies was associated with lower peripheral CD4 recovery whereas Lactobacillales abundance was associated with a higher peripheral CD4 percentage (38, 98). At broader taxonomic levels, the order Lactobacilliales and family Bacteroidaceae seemed to protect against CD4 activation and depletion (43, 98). Other members of the family Bacteroidaceae or species-specific metabolic differences within the genus Bacteroides may account for the discrepant effects of the genus Bacteroides versus family Bacteroidaceae. Studies using Lactobacillus-containing probiotics in HIV appear to support the role of Lactobacillus spp. in reducing CD4 activation and depletion, although similar to much of the literature regarding probiotics, results have been mixed (99–101). Conversely, Prevotella and two subtaxa of the phylum Proteobacteria (the order Enterobacteriales and the family Enterobacteriaceae) have been associated with CD4 T cell activation and depletion (38, 43, 88, 102). Note that the association between Prevotella abundance and mucosal T cell activation was only noted with microbiome analysis from colon biopsy samples, not stool or fecal aspirates, suggesting that mucosa-adherent bacteria may be more important in determining mucosal immune status than are luminal organisms (38). To our knowledge, only four works published to date have examined microbiome specimens from gastrointestinal mucosal biopsy specimens from HIV-infected individuals (38, 39, 43, 103). Future larger studies are needed to clarify whether microbiome analyses of mucosal or luminal samples better correlate with systemic markers of inflammation.

As mentioned above, soluble markers of gut permeability, microbial translocation, and immune activation are clearly linked to clinical outcomes of interest (63, 68, 70). Linking particular microbes to these markers has been more challenging. Vujkovic-Cvijin et al. (43) found several soluble markers, including IL-6, soluble TNFR2, kynurenine/tryptophan ratio, and IP-10, to be associated with a Proteobacteria-heavy cadre of HIV-associated microbes. Dillon et al. (38) found an association between overall mucosal microbiome changes and plasma LPS as well as with peripheral blood and mucosal T cell activation. Dinh et al. (45) found that abundances of the order Enterobacteriales and family Enterobacteriaceae positively correlated with levels of sCD14, IL-1β, and IFN-γ. The class Erysipelotrichi and genus Barnesiella were positively associated with IFN-α levels. We should be cautious when interpreting these results, as alterations in plasma markers may have been driven by HIV infection and the alterations in microbiota by behavioral differences. Variations in the subjects, sampling methods, and taxonomic levels investigated make it particularly difficult, at this stage, to draw any firm conclusions regarding which microbes are most responsible for immune activation. Future studies, including in vitro and ex vivo experiments and direct evaluation of the metabolites found in stool and mucosal samples, will likely improve our understanding of these interactions.

Direct identification of translocating bacteria may also be useful in determining what defines a proinflammatory microbiome. Klase et al. (36) found in an SIV infection model that bacterial DNA of the phylum Proteobacteria was overrepresented in mesenteric lymph nodes and liver tissue of infected primates, suggesting that these organisms preferentially translocate. The order Enterobacteriales and the family Enterobacteriacea of the phylum Proteobacteria were significantly more prevalent in both stool and gut mucosal samples of HIV-infected individuals than in uninfected control participants in several studies (38, 41, 43, 45, 102). A small study from China found that DNA from the order Pseudomonadales (phylum Proteobacteria) and in particular the genera Moraxella and Psychrobacter was found in the blood of treatment-naive subjects with advanced immunosuppression in significantly higher abundance than in HIV-negative controls (104). As sequencing technology improves, species- and strain-specific data will be invaluable in determining which organisms, if any, preferentially translocate during HIV-1 infection.

Metagenomics can be used to identify the genes for a particular metabolic pathway and allow us to bypass difficulties related to the identification of microbiota at the species or strain level. Microbial translocation and resultant immune activation have been linked to microbial zeatin biosynthesis and the metabolism of tryptophan to kynurenine (42, 43). The plasma kynurenine/tryptophan ratio is elevated in HIV-infected persons and has been linked with poor CD4 recovery, intestinal barrier dysfunction, and mortality (63, 105–107). Organisms with the genes for tryptophan to kynurenine metabolism are overrepresented in HIV-infected individuals (43). Serrano-Villar et al. (108) demonstrated that 3-hydroxyanthranilate, a product of the kynurenine pathway, was elevated in stool bacteria from HIV-infected individuals but was undetectable in controls and those with lupus or C. difficile colitis. Note, however, that IDO1 is expressed by activated monocytes and dendritic cells, and thus a portion of the tryptophan metabolism and kynurenine production that occurs in HIV infection is derived from human cells and is likely driven by inflammation (109). Lactobacillus spp. appear to inhibit IDO1 and are selectively depleted in SIV-infected macaques, so this may be a mechanism by which Lactobacillus spp. prevent CD4 activation and depletion (110).

Microbial metabolites other than those of tryptophan catabolism have also been associated with end-organ disease. Tang and colleagues (17, 111, 112) showed that the production of TMAO, a small molecule that is closely linked to coronary artery disease, mortality, and renal insufficiency is dependent on gut microbes. Studies of TMAO in HIV-infected cohorts have had conflicting results regarding whether TMAO is associated with higher risk of coronary events in HIV, and it does not appear that levels of TMAO are particularly elevated in persons with HIV infection (113–116).

Recently, SCFAs have received some attention as important biomolecules produced by gut microbes. SCFAs such as butyrate are produced from bacterial fermentation of insoluble fibers in the diet by certain components of the Firmicutes phylum (117). Butyrate has proven to be critical in a number of ways to the human gut. It functions as a direct energy source for colonocytes, such that starvation of such in inflammatory bowel disease is theorized to cause apoptosis and enhance epithelial barrier damage (118). Butyrate not only nourishes colonocytes but also stimulates production of mucin, enhancing gut mucosal barrier integrity. Xiong et al. (119) showed in 2016 that butyrate directly upregulated host defense peptides in piglet gut epithelial cells and helped ameliorate the pathogenic consequences of E. coli 0157:H7 infection. Additionally, butyrate reportedly regulates and expands the colonic Treg population, promoting an anti-inflammatory phenotype (120, 121). Gut dysbiosis in aging populations is characterized by decreased Firmicutes phyla species abundance and decreased production of SCFAs in the stool, as well as increased epithelial barrier dysfunction (25). Importantly, in the context of HIV infection, butyrate-producing species have also been shown to be decreased in abundance in enteric mucosa and stool (38, 39, 41, 46, 47), and high-dose butyrate may reduce bacteria-induced CD4 T cell cytokine production, activation, and HIV infection levels in an ex vivo intestinal cell model (S.M. Dillon, J. Kibbie, E.J. Lee, K. Guo, M.L. Santiago, G.L. Austin, S. Gianella, A.L. Landay, A.M Donovan, D.N. Frank, M.D. McCarter, and C.C. Wilson, submitted for publication). Interestingly, although not yet shown in the gut, oral flora that produce SCFAs have been shown to enhance reactivation of latent virus, including HIV-1, through inhibition of histone deacetylase (122). This could ultimately benefit the host, as latently infected cells can be a source of productive viral replication and result in persistent life-long infection.

It is clear that HIV infection results in severe damage to the intestinal mucosal compartment with epithelial barrier damage and depletion of CD4 T cell subsets critical to maintenance of intestinal homeostasis. Although many early pilot studies reported HIV-associated changes in the enteric microbiome, both in composition and in diversity, more recent studies suggest that confounding factors such as sexual behavior may explain some of those original findings rather than HIV infection status per se. However, a reduction in overall microbial diversity is found in HIV-infected individuals even when accounting for other lifestyle factors. Additionally, HIV infection has been associated with increases in Proteobacteria, which contains several species that can be considered pathobionts in the appropriate clinical setting. By and large, despite the preponderance of early data gathered, firm conclusions on the exact nature of HIV-associated dysbiosis, including the impact of lifestyle, diet, comorbidities, and treatment effects across the spectrum of HIV-infected patient populations, await larger well-controlled studies that include ethnically and geographically diverse study populations. Furthermore, controlled, longitudinal studies of persons from acute infection (and perhaps before infection if already part of an existing HIV uninfected cohort) to chronic infection, although technically challenging, would shed significant light on the evolution of microbiome changes with HIV infection. Despite these lingering questions, there is solid evidence of associations between microbiome features (especially in the gut mucosa) and mucosal and systemic measures of inflammation and immune activation (summarized in Table I), although whether these associations are causal or simply correlative remains unclear. Intriguing mechanistic hypotheses regarding the direct and indirect effects of the altered microbiome on gut immune function have been proposed, yet due to the complexity of the microbial communities involved and our evolving understanding of the factors that influence them (in normal health, much less during HIV infection), specific studies drilling down to the exact mechanisms involved are still incomplete.

Interventions designed to modify the gut microbiome and thereby reduce inflammation-associated comorbidities are obviously of great interest to the medical and scientific communities. Most interventions studied to date, in several different disease settings, have focused on either directly repopulating the gut with beneficial commensals (probiotics or direct fecal transplant) or encouraging the growth of beneficial commensals with intake of substrate for their growth (prebiotics). Other interventions involve repletion of important biomolecules such as introducing intracolonic butyrate to ameliorate autoimmune disease (118) or active infection (in piglets) (119). Furthermore, still other preclinical study interventions are investigating directed exercise regimens to alter the gut microbiota for health benefit (123). Early results of a pilot FMT trial in HIV were reported at the conference on retroviruses and opportunistic infections in Boston in 2016 [NCT02256592 (124)]. Somsouk et al. found, in a study of six subjects with treated HIV infection, that with FMT the microbiomes of subjects shifted, transiently, toward those of their respective donors. They did not find any changes in the serum kynurenine/tryptophan ratio, a marker of immune dysfunction in HIV that has been linked to mortality in HIV. Larger studies with potentially more intensive interventions will be needed to determine whether FMT is effective in reducing microbial translocation and immune activation.

Studies of probiotics in HIV-infected individuals are extremely heterogeneous, including varying probiotic organisms, formulations, durations of treatment, and outcomes. The one common finding in these studies has been that probiotics are generally safe and well tolerated (125). Studies of Bifidobacterium and Lactobacillus probiotics have shown some improvement in CD4 count with supplementation, although changes are not always statistically significant (100, 101, 126, 127). Microbial translocation and immune activation have also been the endpoints of probiotic trials, and both yeast (Saccharomyces)– and bacterial (Bifidobacterium and Lactobacillus)–based probiotics have shown promise in reducing some inflammatory markers, but results have been mixed (99, 126) A probiotic/prebiotic (Lactobacillus, Bifidobacterium, Streptococcus plus inulin) mixture was shown to improve gut CD4 reconstitution in an SIV model and this multistrain probiotic is currently the subject of a randomized, placebo-controlled trial, that is, AIDS Clinical Trials Group A5350 (90, 128). Results from this large, well-designed, multicenter study will help define whether these organisms play a role in reconstitution of the gut immune system and prevention of microbial translocation. A better understanding of the mechanistic interactions between gut microbes and their metabolic products and host mucosal immune cells in the setting of HIV-associated homeostatic disruption will likely lead to novel interventions that may ultimately reduce gut-derived inflammation and its associated comorbidities.

The authors have no financial conflicts of interest.