Liver bacterial dysbiosis occurs in SIV-infected macaques and persists during antiretroviral therapy

Background: Liver disease remains a signicant contributor to morbidity and mortality in HIV-infected individuals, even during successful treatment with combination antiretroviral therapy (cART). In nonhuman primates, SIV infection is associated with gut microbiome dysbiosis as well as bacterial translocation into the colonic lamina propria and liver via the portal vein. Here the liver microbiome was evaluated in rhesus macaques to discern the inuence of SIV infection alone (SIV+) and during cART administration (SIV+cART) on liver bacterial dysbiosis and neutrophil inltration. Results: Dysbiosis in liver bacterial composition was observed, encompassing changes in a number of genera, during SIV infection in the absence and presence of cART. The most striking nding was an increase in the level of Mycobacterium, which while barely detectable in the uninfected macaques, was the most abundant genus observed in the livers of a majority SIV+ and SIV+cART macaques. Multi-gene sequencing analyses identied a species of environmental mycobacteria similar to the opportunistic pathogen M. smegmatis. The effect of M. smegmatis on host gene expression in primary hepatocytes was evaluated in vitro utilizing PILAM, a glycolipid cell wall component found in atypical Mycobacteria. PILAM induced an upregulation of inammatory responses, including an increase in the chemokines associated with neutrophil chemotaxis (CXCL1, CXCL5, and CXCL6). Assessment of the macaque livers by microscopy determined that neutrophil levels were reduced in SIV+cART macaques, suggesting that the SIV infection and/or cART treatment inuence the liver-associated neutrophil response. Conclusions: A of liver bacteria genera were altered following SIV infection even in the context of cART, possibly as a consequence of reduced neutrophil recruitment. Mycobacteria major component of the SIV infected macaque liver microbiome, possibility that bacteria of this genus disease

virus alone does not account for increased liver disease prevalence, as even persons virally suppressed with cART continue to experience high rates of liver disease [13]. Hepatotoxicity can contribute to liver disease during cART, and the use of certain nucleoside reverse transcriptase inhibitors (NRTIs) is associated with advanced liver disease [14,15]. However, liver disease continues to affect HIV + patients despite the reduced hepatotoxicity of modern cART regimens. Taken together, this suggests that additional factors may be contributing to liver dysfunction during HIV infection.
The liver is the primary site of bacterial clearance from blood exiting the gut, prior to entering systemic circulation. In healthy individuals, this process is one of immune tolerance. However, during disease, in which the liver may experience either elevated bacterial load or exposure to dysbiotic bacteria, the liver responds in an in ammatory nature [16]. Central to this in ammatory response are Kupffer cells, resident liver macrophages that are primarily responsible for clearance of microbial products from portal blood [17]. Upon engagement of innate receptors (e.g. toll-like receptors (TLRs)) on these cells by microbial products, in ammatory and pro brotic mediators are produced, such as TNF-α IL-12, IL-6 and TGF-β [18].
Neutrophils are recruited to the liver and aid in the clearance of bacterial products through the release of reactive oxygen species and pro-in ammatory cytokines [16,19]. Hepatocytes are impacted by bacterial stimulation and experience altered gene expression indicative of metabolism imbalance and production of in ammatory mediators [20,21], key features of NAFLD. Consistent with these ndings, several clinical studies have established the role of microbial translocation in the context of alcoholic liver disease and in NAFLD [22]. Given that bacterial translocation is associated with HIV disease progression and systemic immune activation, and NAFLD is one of the key hepatic disruptions observed during HIV infection [23], it is conceivable that bacteria-associated liver in ammation may play a role in HIV-associated liver disease.
Understanding of the gut microbiome and immune changes during simian immunode ciency virus (SIV) infection have advanced the understanding of gut-driven systemic immune activation and HIV disease progression [24]. The microbiome is in uenced by numerous factors such as age, environment, food and antibiotics [25][26][27][28][29]. With regard to the liver during SIV infection, prior studies have identi ed elevated bacterial loads during SIV infection [30,31], as well as in the context of cART [32], and have shown that bacteria may increase immune cell in ltration [30]. During cART-treated SIV infection, the liver microbiome has been shown to be enriched for in ammatory Proteobacteria that preferentially translocate out of the gut and into the colonic lamina propria [24]. Altogether, these studies demonstrate a role for both bacterial load and bacterial composition in liver dysfunction during SIV infection, even in the context of successful cART therapy. Importantly, characterization of hepatic bacteria to lower taxonomic levels, such as species identi cation, has not been reported. Here SIV-infected (SIV+) macaques are characterized with regard to their liver microbiomes using 16S rRNA gene sequencing to identify prevalent bacteria, and the results are linked with changes in immune cell subsets within dysbiotic SIV + livers. These studies help advance the knowledge of bacteria-associated liver in ammation during SIV infection by providing information regarding an association between hepatic in ammation and the presence of dysbiotic microbes. incompatible behaviors were managed by the Behavioral Management staff and managed accordingly. SIV + macaques were kept in individual, adjoining cages allowing for social interactions with primate health observed daily by trained staff. All efforts were made to minimize suffering using minimally invasive procedures, anesthetics, and analgesics when deemed appropriate by veterinary staff. Animals were painlessly euthanized by sedation with ketamine hydrochloride injection followed by intravenous barbiturate overdose following the recommendations of the panel of euthanasia of the American Veterinary Medical Association. These macaques have been described previously [32]. Liver Tissue Collection Liver tissue was collected at necropsy from uninfected (N = 4), SIV+ (N = 6) and SIV + cART (N = 6) adult Indian rhesus macaques (Macaca mulatta). Control samples from uninfected macaques were acquired from the Tissue Donor Program at WaNPRC. SIV + macaques were infected intrarectally with SIVmac239x [32]. Macaques receiving cART were administered subcutaneous tenofovir (20 mg/kg body weight) and emtricitabine (30 mg/kg) and oral raltegravir (50 mg twice daily) starting 120 days postinfection and continuing for 35-36 weeks prior to euthanasia [32]. Tissue was formalin-xed, para nembedded for microscopy or ash-frozen in liquid nitrogen and then stored at -80 °C for nucleic acid extraction.

Immuno uorescence Staining
Liver tissue was obtained at necropsy (14-55 weeks post infection), xed in 10% formalin, and para nembedded. Slides were prepared from 5µ m tissue sections and dewaxed with xylene and rehydrated with a gradient of ethanol baths. Antigen retrieval was performed with a 1% citrate buffer (Antigen Unmasking Solution, Vector Laboratories, Burlingame, CA) in a decloaking chamber at 90 °C for 30 minutes with a 10minute cooldown. Slides were washed in a series of TBST (0.025% Triton X-100 in 1X TBS) followed by a 2-hour block step (0.1% BSA, 1% goat serum in TBST). Tissues were stained overnight at 4 °C for mouse anti-human myeloperoxidase (MPO, polyclonal, 1:2000, Dako) and rabbit anti-human CD68 (clone KP1, 1:250, Santa Cruz Biotechnology, Dallas, TX). Slides were then washed in another series of TBST washes and incubated with secondary antibody for one hour in the dark using AlexaFluor 488 goat anti-mouse (1:500, Life Technologies, Carlsbad, CA) to detect MPO + neutrophils and AlexaFluor 594 goat anti-rabbit (1:500, Life Technologies) to detect CD68 + macrophages. After incubation, slides were washed in a series of TBST washes and mounted with Vectashield Hard set DAPI (Vector Technologies) and allowed to set. For each liver section, eight random elds were imaged at 200x magni cation. Cells were counted using ImageJ.

Liver Tissue Disruption By Pulverization
Flash-frozen liver tissue was pulverized into a ne powder by ball milling with stainless balls under cryogenic conditions with liquid nitrogen (Retsch Planetary Ball Mill, Retsch Laboratory Equipment, Haan, Germany). Each sample was subjected to three cycles at 300 rpm for two minutes each. Following pulverization, the liver powder was collected and stored at -80 °C until DNA extraction.

Tissue DNA Extraction from Liver Powder
Liver powder (10-30 mg) was placed into a sterile, pre-chilled microcentrifuge tube. Genomic DNA was extracted using the NucleoSpin Tissue DNA extraction kit (Takara, Mountain View, CA) per the manufacturer's instructions, where samples were pre-lysed and allowed to incubate at 56°C for at least 1-3 hours vortexing occasionally. Samples were then lysed with provided buffer, vortexed vigorously, and incubated at 70°C for 10 minutes. Ethanol was added and samples were centrifuged in NucleoSpin Tissue Columns into a collection tube at 11,000 x g for 1 minute. After a series of washes, samples were eluted with elution buffer and collected. Following concentration determination with a NanoDrop 2000 Spectrophotometer (Thermo Scienti c, Waltham, MA), isolated genomic DNA was stored at -80 °C until use.
16s rRNA Gene Sequencing and Microbiome Analysis Genomic DNA extracted from the liver (20 µL) was used for 16 s rRNA gene sequencing through Illumina according to the EMP method. In brief, a 460-bp amplicon was generated targeting the V3-V4 region of the 16 s rRNA gene. PCR amplicons were cleaned with 0.8x AMPure XP beads (Beckman Coulter, Brea, CA) before the addition of Nextera XT dual index adaptors (Illumina Inc., San Diego, CA). Indexed amplicons were cleaned using 1.1 × AMPure XP beads (Beckman Coulter), quanti ed using a Qubit DNA high-sensitivity assay kit (Life Technologies), and multiplexed using an equal molar ratio of DNA for each sample. 16S rRNA gene libraries were loaded on a 300-cycle MiSeq kit and sequenced using Nextera sequencing read and index primers (all from Illumina Inc.). Paired-end demultiplexed FASTQ les from the Illumina base space were imported into the QIIME2 pipeline (QIIME 2 Core 2019.10) to create a demultiplexed QIIME2 object. These objects were matched to identi ed amplicon sequence variants (ASVs) using the dada2 algorithm which worked to detect and correct Illumina amplicon sequence data and denoise by trimming to 145 bases to remove low-quality regions. A rooted phylogenetic tree was constructed using the Mafft multiple sequencing alignment program and taxonomy was assigned using the Greengenes database speci c to the V3-V4 region. After taxonomy was determined, results were exported from the pipeline for downstream analysis in R using the phyloseq package.
Quanti cation of Mycobacterial DNA in the Liver by qPCR All liver DNA samples were diluted in nuclease-free water. Each sample (5 µL) was prepared in duplicate in a 20 µL volume reaction with the PowerUp SYBR Green Master Mix kit (Applied Biosystems, Waltham, MA) and Mycobacterium-speci c primers (MycoARB210: TTT GCG GTG TGG GAT GGGC and MycoARB585: CGA ACA ACG CGA CAA ACCA). A 'No Template' Negative Control was included to control for contamination and non-speci c ampli cation. A standard curve was generated by serially diluting pure M. bovis (BCG) DNA 10-fold, ranging from 10-0.001 ng/µL (R 2 > 0.95). PCR reactions ran one cycle at 50 °C for 2 minutes then increasing to 95 °C for 2 minutes followed by 45 cycles of 94 °C for 15 seconds, annealing at 61 °C for 30 seconds, and extending at 72 °C for 30 seconds with a nal extension step at 72 °C for 7 minutes. Following the qPCR cycles, PCR reactions were subjected to a melt curve analysis to examine products formed. The concentration of Mycobacterium per sample was determined through a non-linear regression on the standard curve and converted to copy number based on BCG molecular weight (5.63 × 10 12 mg/mole). The weight was then converted to 4.277 × 10 7 molecules/mole and the standard curve was plotted based on molecules where copy number was equal to 4.277 × 10 7 molecules/mole * log (CT) where the standard curve equation was extrapolated (y = -0.032ln(x) + 1.8377). Copy number of the liver Mycobacterium DNA was then calculated from the standard curve equation (copy number = e ((log (Ct) -1.8377)/-0.032)). Duplicates were averaged for each animal. qPCR to detect a conserved region of the 16S rRNA gene was performed as reported previously [32].

Identi cation of Mycobacterium Species by Multi-Gene Sequencing
Genomic DNA (extracted as described above) was diluted in nuclease-free water and ampli ed by nested PCR per the conditions outlined in Additional File 1. For each rst-round PCR reaction, 500 ng (5 µL) of gDNA was added into a 50 µL reaction and ampli ed using the Platinum Taq DNA Polymerase reaction kit (Invitrogen, Carlsbad, CA). For nested PCR reactions, 1 µL of the rst-round PCR product was added to a 50 µL reaction containing the nested primers and Platinum Taq DNA Polymerase. Each round of PCR contained a positive control of BCG DNA and a negative no template control. Following nested PCR, each reaction was examined on a 1% agarose gel. Each PCR amplicon showing the correct size was cleaned up using a Nucleospin PCR Clean-up Kit (Takara) and eluted into 30 µL of EB buffer. Puri ed PCR amplicons (20 ng) were sent for Sanger sequencing using both forward and reverse nested primers in separate reactions. Following sequencing, DNA sequence quality was examined in 4Peaks software and low-quality reads from the 5' and 3' ends removed. Consensus sequences generated through the alignment of forward and reverse reads were analyzed using BLAST analysis.

Culture and Stimulation of Human Hepatocytes with Mycobacteria and Mycobacterial Antigens
Human HepaCure Hepatocytes on Matrigel overlay (350,000 hepatocytes/well) were acquired from Yecuris (Tualatin, OR) in 24-well dishes. HepaCure human hepatocytes are produced by the immunization of humanized FRG®KO mice with cadaver-derived human hepatocytes. Upon receipt, the media was immediately replenished with 500 µL InVitro GRO Hi Medium (BioIVT, Westbury, NY) supplemented with Torpedo Antibiotic Mix (BioIVT). Cultures were incubated at 37 °C, 5% CO 2 overnight. To determine the hepatocyte response to mycobacterial pathogen associated molecular patterns (PAMPs), M. smegmatis puri ed lipoarabinomannan (PILAM, 0.1 and 10 ug/mL, BEI Resources, Manassas, VA), or M. tuberculosis, Strain H37Rv, puri ed lipoarabinomannan (ManLAM, 0.1 and 10 ug/mL, BEI Resources), were added to hepatocytes. Each stimulation condition was conducted in triplicate. Plates were incubated at 37 °C, 5% CO 2 for 24 hours. For live mycobacteria stimulations, M. smegmatis bacteria (strain MC 2 155) were grown to exponential phase, washed with PBS and resuspended in InvitroGRO Hi Medium without antibiotics at 350,000 bacteria/µL. Hepatocytes were stimulated with M. smegmatis (MOI 10) in duplicate for 24 hours at 37 °C, 5% CO 2 . For all stimulations, conditioned media was collected and stored at -80 °C. The hepatocyte monolayer was then washed with 500 µL pre-warmed PBS and then lysed in 300 µL RA1 buffer containing beta-mercaptoethanol. RNA was isolated from the cell lysate following protocols from the NucleoSpin RNA isolation kit (Macherey-Nagel, Bethlehem, PA).
Transcriptomic Analysis of HepaCure Hepatocytes by Nanostring RNA was diluted to 20 ng/µL in nuclease-free water and used for transcriptomic analysis using a Nanostring In ammation Panel (Human v2) (Nanostring, Seattle, WA). Probe set-target RNA hybridization reactions were performed according to the manufacturer's protocol using 100 ng (5 µL) of total RNA. Puri ed probe set-targets were processed and immobilized on nCounter Cartridges using a nCounter MAX prep station. Transcripts of interest were quanti ed on the Digital Analyzer for each sample. For data analysis, nCounter RCC les were imported in nSolver Analysis Software 4.0 and checked for quality control. Determination of differentially expressed genes, pathways analysis, and cell pro ling was conducted using the Nanostring Advanced Analysis software per the manufacturer's instructions. For each stimulation condition, differentially expressed genes were determined by comparing the normalized count data between stimulated hepatocytes and unstimulated control hepatocytes. Heatmaps were generated in Prism version 5.0f software (GraphPad Software, Inc., San Diego, CA), showing fold change of each gene in the panel. Volcano plots were assessed using the python matplotlib package for signi cant genes using a threshold of 1.5-fold change (log2(1.5) = 0.585) and 0.05 adjusted p-value.

Statistics
Statistical analyses were performed using Prism version 5.0f software (GraphPad Software, Inc.). A nonparametric Mann-Whitney U test was used to compare the SIV + and SIV + cART groups to the uninfected controls. Linear regression and Spearman correlation analyses were performed. Analysis of gene expression panels was completed using nSolver (Nanostring, version 4.0.62).

Page 8/18
The Liver Microbiome during SIV Infection The liver microbiome was evaluated in macaques that were uninfected, SIV+, and SIV-infected-cART treated (SIV + cART) through 16S rRNA gene sequencing and analysis [32]. While macaques displayed variation in the liver microbiome, some overarching similarities were observed, including the presence of Pseudomonas, Bacillus, Stenotrophomonas, Massilia, and Delftia in the livers of many of the macaques (Fig. 1A-C). In uninfected macaques, the highest percentage of sequences that could be classi ed were Stenotrophomonas, a diverse genus with a wide range of species. However, this genus constituted less than 25% of sequences, and the majority of sequences in the uninfected macaques fell into the 'Other' classi cation, which includes all genera outside of the nineteen most abundant. Therefore, the uninfected macaques had a diverse liver microbiome with no dominant pervasive genus (Fig. 1A). In contrast, in the SIV + group, there was an increase in the proportion of sequences in the Mycobacterium genus (Fig. 1B). In addition to Mycobacterium in the SIV + livers, many bacterial genera that are common in the macaque intestinal microbiome, including Acinetobacter, Prevotella, Lactobacillus, and Bacillus [24,33], were also found to be in abundance (Figs. 1 & 2). Following cART treatment, Mycobacterium still remained as one of the most prevalent genera identi ed within the livers of the macaques (Fig. 1C). Other notable differences observed between the groups included a decreased relative abundance of the Lactobacillus and Blautia genera in SIV + macaques that persisted with cART, and an increased abundance of Pseudomonas in SIV + cART macaques (Figs. 1 & 2). However, Mycobacterium stands out, as when evaluating relative abundance, Mycobacterium was the most abundant bacterial genus found in the liver in SIV + and SIV + cART macaques, while it was present at extremely low levels in the uninfected macaques (Fig. 2). Importantly, although the percentage of bacteria attributed to the Mycobacterium genus varied within the SIV + and SIV + cART groups, Mycobacterium DNA was identi ed within every liver sample tested in each of these groups (with A14050 (SIV+), Z09068 (SIV+) and A13275 (SIV + cART) having the highest percentage of Mycobacterium sequences present) ( Fig. 1B-C).
Differences in microbiome alpha diversity were also apparent between the macaque groups ( Fig. 3A-B). Examination of microbial richness indicated that SIV+ (but not SIV + cART) macaques had high variation in the number of observed taxa compared to the uninfected macaques though mean richness was unchanged by SIV infection or cART (Fig. 3A). The SIV + group exhibited a larger range of relative abundance within the Mycobacterium genus when compared to the SIV + cART group (Fig. 2), and outside of mycobacteria, some SIV + macaques had a high number of observed taxa and others had a low number of observed taxa compared to the mean (Fig. 1, Fig. 3A). By contrast, SIV + cART macaques had low variation in richness, similar to the uninfected controls. However, cART treatment did not fully restore the baseline liver microbiome as the composition of the taxa in the livers of SIV + cART macaques resembled more closely the SIV + macaques, including the high prevalence of bacteria within the Mycobacterium genus (Fig. 1, Fig. 3A). Microbial evenness was also impacted by untreated and cARTtreated SIV infection (Fig. 3B). SIV + and SIV + cART macaques exhibited a larger range in evenness score similar to the larger range in richness score. Interestingly, while mean evenness in SIV + macaque livers was similar to that in the uninfected control group, this measure of the representation by each genus tended to decrease in SIV + cART macaques, indicating a redistribution of taxa especially in the setting of cART treated infection. Altogether, these ndings support that the liver microbiome is altered during SIV infection and does not fully recover during cART. Assessment Of Mycobacterial DNA Since 16S rRNA gene sequence abundance is a relative estimate that re ects the abundance of other bacteria, qPCR was conducted to con rm that Mycobacterium was indeed increased in the liver during SIV infection in both untreated and cART suppressed macaques. Based on previous methods [34], extracted liver DNA was assessed using Mycobacterium 16 s rRNA gene-speci c primers and Mycobacterium was quanti ed in the liver of each macaque. Mycobacterium were detected in all macaques; however, levels were signi cantly higher in both the SIV+ (p = 0.0048) and the SIV + cART macaques (p = 0.0095) when compared to uninfected macaques (Fig. 4). These data con rm the presence of Mycobacterium in the liver during SIV infection as seen in the 16S rRNA gene sequencing, and that drug therapy does not restore the liver microbiome to normal composition even during viral suppression.
To better understand which Mycobacterium are present in the liver during SIV infection, the Mycobacterium present in the liver were identi ed by multi-gene amplicon sequencing. High sequence homology in closely related mycobacteria necessitates the use of multiple genes to help discriminate at the species level. Thus, Mycobacterium-speci c primers for both the 16S rRNA gene and the rpoB gene were utilized to amplify variable regions of each gene, followed by sequencing. Identi cation of liver mycobacteria using the 16S rRNA gene indicated the presence of non-tuberculous mycobacteria (NTM) of a few possible species, including M. smegmatis, M. marinum, or M. goodii, with greater than 99% sequence match ( Table 1). Two of the macaques (Z09086, Z09096) yielded top BLAST hits exclusively for M. smegmatis. The rpoB gene has less sequence coverage in the BLAST database but is valuable in combination with the 16 s rRNA gene analysis. The rpoB gene sequencing analysis consistently yielded the identi cation of M. smegmatis in the liver of each macaque with > 99% identity matches for all liver DNA samples tested, with exception to Z09096 which did not have enough DNA template for this secondary PCR veri cation (Table 1). Taken together, these data suggest that M. smegmatis or a closely related relative is likely the speci c Mycobacterium species present in the livers of SIV + and SIV + cART macaques.

Discussion
Different bacterial species have the potential to be in ammatory or anti-in ammatory; thus the composition of the microbiome exerts important in uences on homeostasis [22]. Like HIV infection [37], SIV infection is associated with gut and liver microbiome dysbiosis, including an enrichment for in ammatory Proteobacteria [24]. Preferential translocation of bacteria originating in the gut into the colonic lumen during SIV infection [24] suggests a model whereby bacteria travel from the gut into the colonic lumen prior to entering the liver via the portal vein. This process lters the blood from the intestine, and these liver microbial products are associated with immune activation and liver damage [30].
Here we sought to identify the liver bacteria to lower taxonomic levels and mechanistically assess the relationship between the liver bacteria and hepatic immune activation. Using bacterial 16S rRNA gene sequencing together with qPCR, we characterized the liver microbiome to the genus level and con rmed elevated levels of the most proli c genus, Mycobacterium, in SIV + and SIV + cART-treated macaques. Multigene PCR sequencing identi ed M. smegmatis or a closely related Mycobacterium species as the predominant species in SIV + and SIV + cART-treated liver samples. We evaluated the impact of M. smegmatis on hepatocyte in ammation in vitro, nding that M. smegmatis PILAM induces an upregulation of neutrophil chemotactic mediators. Surprisingly, in vivo, elevated bacterial DNA in the livers of SIV + and SIV + cART-treated macaques was associated with reduced rather than elevated neutrophil counts. These ndings suggest that the normal physiological process of neutrophil recruitment to the liver in response to bacteria and host cellular chemotactic signals may be impaired during SIV infection, thereby promoting dysbiosis and the presence of Mycobacteria in the liver, which does not normalize during cART. These ndings provide key insights into understanding SIV-associated liver in ammation and the microbial composition that is altered during SIV infection and cART therapy.
Overall, in the livers of SIV + animals in this study, we observed a bloom of prevalent genera during infection, with Mycobacterium as the most abundant genus present. In HIV + patients, there is reduced diversity in gut microbiome composition that does not generally recover back to pre-HIV levels after cART treatment is initiated [37]. In the livers of SIV + macaques, we did not detect a signi cant drop in the alpha diversity of taxa present, but instead saw a wider variation in the number of observed taxa around the mean, suggesting variation in how individual animals respond to the infection. With the introduction of cART treatment, the bloom of bacterial populations did seem reduced, as the evenness across present genera decreased when compared to uninfected animals, yet Mycobacterium persisted and dominated. The nding that cART did not result in a substantial reduction in the levels of environmental mycobacteria in the liver of SIV + macaques could be due to the fact that cART treatment was still in an early stage, with treatment lasting 35 to 36 weeks before tissues were examined. Alternatively, clearance of mycobacteria may be di cult once colonization is established. Characterization of the liver microbiome over time during infection and at later time points following the introduction of cART will aid in our understanding of these ndings. The low detection of Mycobacterium in the uninfected macaques indicates that Mycobacterium is likely present in the normal liver microbiome, but SIV infection allows the bacteria to opportunistically thrive to a higher prevalence. Thus, SIV infection may provide the opportunity for speci c genera that are present in the liver microbiome to increase in prevalence, rather than allow for the introduction of new genera; these changes in microbial composition appear to be not reversed by cART.
In our previous publication, we evaluated changes in the liver macrophage populations that expand during SIV infection and correlate with both in ammatory (TNF-α, CCL3) and brosis (TGF-β) mediators [32]. Evaluating the CCL2-CCR2 chemokine network as an integral inducer of monocyte/macrophage in ltration into the liver, we observed an upregulation of both CCL2 and CCR2 in the liver in macaques during untreated SIV infection. This CCR2 expression positively correlated with the frequency of CD68 + macrophages, leading us to speculate that viral stimulation in the liver alters the immune environment through induction of CCL2, and possibly other chemokines, resulting in immune cell in ltration [32]. Liver resident macrophages (Kupffer cells) are exposed to translocated gut-derived bacterial products by portal circulation and function to sense and remove pathogens through pattern-recognition receptors (PRRs) such as TLRs. These TLRs recognize gut microbiota-derived bacterial products (e.g. LPS), triggering a response through TLR4 to produce in ammatory cytokines [17,38]. Here, we assessed a second key phagocytic cell population, neutrophils. Neutrophils are rapidly recruited to sites of acute in ammation, though the method of recruitment of these cells to the liver is not well known [19,38]. However, activated neutrophils can also promote disease progression via the secretion of pro-in ammatory cytokines [19]. In HIV + and HIV + cART patients, an increase in neutrophil frequency and survival was reported; the increased survival correlated inversely with the ratio of Lactobacillus to Prevotella in the gut and Lactobacillus was associated with a decrease in neutrophil survival [39]. Bacterial PAMPs affect neutrophil survival differently, and all other tested bacterial species except for Lactobacillus signi cantly increased neutrophil survival after incubation with whole blood [39]. In the present study, a decreased frequency of MPO + neutrophils in the livers of SIV + and SIV + cART macaques correlated with a higher level of 16S rRNA gene. We hypothesize that this decrease of neutrophils in the liver during SIV infection results in a loss of the ability to identify and clear bacteria and bacterial products, including environmental Mycobacterium species. Interestingly, within the liver microbiome assessment reported herein, there was a decrease of the Lactobacillus genus during SIV infection.
The nding of high levels of environmental mycobacteria in the livers of the SIV + and SIV + cART macaques was unexpected. NTM, such as M. smegmatis are ubiquitous and inhabit a range of environmental reservoirs, including natural and municipal water, soil, aerosols, food and dust, with water being the most common source of infection [40]. Overall, water treatment processes have been shown to e ciently remove mycobacteria, indicating that mycobacteria recovered from water systems most likely contaminate post-treatment [41].  [43]. It is important to note that detection of Mycobacterium DNA requires specialized lysis steps to rupture the cell wall [44,45]. Here, we utilized a ball mill to mechanically disrupt liver tissue prior to DNA extraction, which likely enhanced the recovery of Mycobacterium DNA. Remarkably, many opportunistic infections caused by Mycobacterium have been identi ed in HIV + patients, particularly with members of the M. avium complex (MAC) [46]. The pathogenic potential of mycobacteria has been described beyond the gut as well, such as in the case of Lady Windermere syndrome, a polymicrobial infection including MAC that affects the lungs [47]. Generally, environmental mycobacteria do not pose a health risk to healthy individuals, but these mycobacteria can cause disease in immunocompromised individuals. Early studies investigating the connection between HIV and NTM found that in patients with HIV there was a higher chance of isolating M. xenopi and M. kansasii from cultured respiratory secretions, in addition to M. fortuitum, M. terrae, and M. scrofulaceum from extrapulmonary sites [48]. In fact, M. kansasii has been shown to cause serious pulmonary infections in patients with late stage AIDS [49,50].
Through sequencing of the 16S and rpoB genes we were able to identify the Mycobacterium present in the livers of the SIV + macaques as being closely related to M. smegmatis. M. smegmatis is an environmental NTM that has the potential to be an opportunistic pathogen in immune-suppressed people [51]. The Mycobacterium genus comprises hundreds of species that range from pathogens with signi cant clinical importance, such as members of the M. tuberculosis complex, to environmental NTM that are prevalent in water and soil [52]. NTM are increasingly associated with opportunistic infections in immunocompromised hosts. In one case study, an immunocompromised patient with an inherited interferon-gamma receptor de ciency was diagnosed with a mycobacterial infection identi ed as M.
smegmatis, which proved fatal despite treatment [51]. Interestingly, M. smegmatis has been shown to be pathogenic in other laboratory models; gold sh M. smegmatis infection induced giant cell replication and recruitment to the liver and increased mortality [53]. To delineate the effect of M. smegmatis on hepatocytes, in vitro experiments were conducted using puri ed PILAM, which is a component of the M. smegmatis cell wall. M. smegmatis PILAM induced an upregulation of neutrophil chemotactic mediators, CXCL1, CXCL5, CXCL6, which is similar to results obtained with M. tuberculosis ManLAM stimulation. Interestingly, a reduction in liver neutrophils was observed in SIV + and SIV + cART macaques in this study, which correlated with increased bacterial DNA in the liver. Taken together, our data suggest that neutrophil de ciency may ultimately enable incomplete bacterial clearance from the liver during SIV infection, thereby allowing opportunistic pathogens to thrive.

Conclusions
Liver disease is currently a major contributor to the morbidity and mortality observed in HIV+ and HIV+cART patients. Here, we identi ed an altered microbiome within the livers of SIV+ rhesus macaques that includes an increase in the levels of environmental mycobacteria identi ed as M. smegmatis, or a close relative. Our data raise questions regarding the presence of Mycobacterium in HIV+ people, including those on cART. Obtaining critical specimens, such as stool, liver and other tissues from these patients followed by optimized DNA extraction techniques is critical for determining the extent to which environmental mycobacteria are part of the microbiome during HIV infection.

Declarations
Ethics. The ethics statement concerning non-human primates is provided in the rst sub-section of Methods.
Availability of data and materials. The datasets generated and analyzed during the current study are available in the SRA and Geo repositories.
Competing interests. The authors declare that they have no competing interests. Authors contributions.
BSF and DLS designed the study.
BSF, KAF, ATG, CF, and MPW carried out the experiments.
MG and ATG conducted the microbiome sequencing.
BSF and KAF conducted the Nanostring analysis.
JS coordinated and oversaw the animal work.
BSF, KAF, ND, and DLS wrote the paper.