Liver microbiome in healthy rats: the hidden inhabitants of hepatocytes

The tumor and tissue microbiota of human beings have recently been investigated. Gut permeability is known as a possible resource for the positive detection of tissue bacteria. Herein, we report that microbiotawere detected in high abundance in the hepatocytes of healthy rats, and that they were shared with the gut microbiota to an extent. We assessed male Sprague Dawley (SD) rats for the 16S ribosomal ribonucleic acid (rRNA) gene. After ratswere sacri�ced by blood drainage from the portal vein, we extracted total deoxyribonucleic acid (DNA) from their ileal and colonic contents and liver tissues. The V3–V4 region of the 16S rRNA gene was ampli�ed by polymerase chain reaction (PCR) and sequenced using an Illumina HiSeq 2500 platform. Sequences were assigned taxonomically by the SILVA database. We also detected bacterial lipopolysaccharide (LPS) and lipoteichoic acid (LTA) in situ using immuno�uorescence (IF) and western blotting and the 16S rRNA gene using �uorescent in situ hybridization (FISH). In the livers of six rats, we detected 54,867.50 ± 6450.03 effective tags of the 16S rRNA gene and clustered them into 1003 kinds of operational taxonomic units (OTUs; 805.67 ± 70.14, 729–893). Individuals showed conservation of bacterial richness, abundance, and evenness. LPS and the 16S rRNA gene were detected in the nuclei of hepatocytes. The main function composition of the genomes of annotated bacteria was correlated with metabolism (79.92 ± 0.24%). Gram negativity was about 1.6 times higher than gram positivity. The liver microbiome was shared with both the small and large intestines, but showed signi�cantly higher richness and evenness than the gut microbiome, and the β -diversity results showed that the liver microbiomeexhibited signi�cantly higher similarity than the small and large intestines (P < 0.05). Our results suggest that the bacteria in the liver microbiome are hidden intracellular


Introduction
With the help of high-throughput nucleotide sequencing and using the 16S ribosomal ribonucleic acid (rRNA) gene as a highly conserved marker in bacterial clades, gut microbiota-which are distributed throughout the whole gastrointestinal (GI) tract with diverse microbial communities-could be identi ed [1].
By collecting feces and analyzing the intestinal microbiota composition and its metabolites in health and disease, many human sample population studies have further proven the relationship between gut microbiome (mainly fecal microbiome) alterations and the occurrence of various diseases [2][3][4][5][6].These studies and reviews strongly suggested the gut microbiota mechanism of diseases [7,8].
Recently, researchers have found that a high-fat diet can elevate gut permeability and result in bacterial translocation from the intestine into tissues such as adipose and liver tissues [9][10][11][12].In our recent study, we also found that in feeding after-fasting or obesity rats, the villi of the small intestine formed a channel, providing transient and functional unrestricted channel absorption (data not shown).These results clearly indicated that gut leakage was the pathway by which gut microbiota resulted in diseases [13].
In addition, studies have also demonstrated bacteria in various tumor cells and proposed the concept of tumor type-speci c intracellular bacteria [14].Moreover, several studies have suggested the existence of tissue microbiota either in humans or in mice [15][16][17][18].The 16S rRNA gene signature was also demonstrated in the blood of healthy humans (primarily leukocyte and platelet fractions) [19] or patients [20,21] as well as in the circulation of animals [22][23][24].These key factors illustrate the crosstalk line between the gut microbiota, circulation, and tissues [25]; that is, the gut microbiota enter into the circulation through leakage and locate into tissues and change into tissue microbiota/tumor type-speci c intracellular bacteria.They then cause disease, although the mechanism of tissue microbiota development remains a hypothesis [26].
Based on this route, the liver is inevitably the frontier organ (the key gatekeeper) through which the gut microbiota enters into circulation.This relationship can be described as the "gut-liver axis" [27].The results of a study by Sookoian et al. support this possibility.Bacterial DNA was detected in non-alcoholic fatty liver disease (NAFLD) in obesity patients, and the LPS was detected in the portal tract by immunohistochemistry [26].It can also be deduced that the liver, as well as the mesenteric lymph nodes, plays immune ltering functions similar to the spleen and exports bacteria to the circulation as the transfer station.Regarding the complicated pathogenesis of diseases in visceral organs that are independent of gut and liver impairment, it is of great signi cance to investigate the liver bacteria in healthy individuals and clarify the relationship with the gut microbiota.
The objective of this study was to con rm the existence of the liver microbiome, its composition and location characteristics, and its relevance to the gut microbiome.The rationale of the study is as follows: (1) microbiota were detected with high-throughput 16S rRNA gene sequencing [1]; (2) the location of liver microbiota was con rmed by detecting LPS, LTA, or 16S rRNA gene in situ using immuno uorescence (IF) [14], uorescence in situ hybridization (FISH), and western blotting; and (3) the liver and gut microbiome relevance was con rmed by comparing their composition and diversity.Herein, we reported bacteria in the normal liver tissues and hepatocytes of rats without dietary interference or disease development.The comparison between the liver and gut microbiota suggests an inherent inhabitant of the liver microbiome in hepatocytes, and the "liver to gut" crosstalk of the microbiota can be deduced.

Animals
Male adult Sprague-Dawley (SD) rats, 22 weeks old with a body weight of 270-360 g (ordinary rats) or 8 weeks old with a body weight of 250-280 g (clean rats), were used for 16S rRNA gene sequencing or location detection.All animals received humane care, and the study protocols, complied with the Laboratory animal-Guideline for ethical review of animal welfare (GB/T 35892-2018) and were approved by the Medical Ethics Committee of Lanzhou University (jcyxy20190302).This study also conformed to the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines and Replacement, Re nement and Reduction of Animals in Research (NC3Rs).

Sample collection and contamination avoidance
Animals were anesthetized with pentobarbital sodium (5 mg/100 g body mass) to avoid their suffering at each stage of the experiment, followed by three skin sterilizations using 75% ethanol and iodophor.To avoid environmental contamination, all operations were conducted under a laminar ow hood following aseptic surgery protocols, including the use of a surgical towel, and avoided contact between the skin and subcutaneous tissue.After the skin was opened, we discarded the original set of all surgical instruments and used another set.To eliminate possible blood contamination, animals were sacri ced by portal vein blood drainage.To prevent possible fecal contamination, we began by removing liver tissues, then those of the ileum (small intestine [SI]) and colon (large intestine [LI]).Subsequently, SI and LI contents were pushed out from the outside of the gut using another set of tweezers.Livers were separated and sampled without perfusion.All samples were collected in DNA-free clean Eppendorf tubes and stored at −80°C after a quick freeze in liquid nitrogen.To avoid possible contamination during 16S rRNA gene analysis, we transported samples on dry ice to a sequencing company (BMK Co., Beijing, China), which provided an integrated service in DNA extraction, polymerase chain reaction (PCR) ampli cation, and sequencing via procedures that strictly avoided contamination.

Bacterial DNA extraction
We extracted total DNA from frozen liver and fecal samples (0.25-0.5 g) using a Magnetic Soil and Stool DNA Kit (#DP812; TianGen Corp., Beijing, China, https://www.tiangen.com/)per the kit protocol.All procedures were performed with sterile and disposable materials to avoid cross-contamination; in addition, beads and DNA extraction blank controls were used during this process.The volume of total DNA solution was 40.0 μl with OD 260 /OD 280 values of 1.86-1.97,and the nanodrop concentration ranged from 42.2-176.3ng/μl.For the few sample concentrations lower than 5 ng/μl and with OD 260 /OD 280 exceeding the detecting range of the instruments (i.e., out of the range of 1.8-2.0),we determined bacterial DNA, as well as its integrity, with preampli ed PCR products using 1.8% agar gel.Brie y, the 16S rRNA gene's variable region 3-4 (V3-V4) was ampli ed using PCR (1 μl original solution, Q enzyme, 30 cycles), and the products were checked using 1.8% agar gel (120 V, 40 min).Then we quali ed them by analyzing the gel images via ImageJ software (National Institutes of Health, Bethesda, MD, USA).The positive bands proved the existence of bacterial DNA in samples, meeting the criteria for further sequencing.
We detected the products using 1.8% agar gel and quali ed them by analyzing the gel images via Image J software.Next, we mixed 150 ng of each sample (samples were 1.5-14.0μl) and puri ed it using an E.Z.N.A. Cycle-Pure Kit (Omega Bio-Tek, Inc., Norcross, GA, USA).The products were recovered using Monarch DNA Gel Extraction (NEB, T1020L) in 1.8% agar gel.

Sequencing
After quality testing on a Qsep-400 (BiOptic, Inc., New Taipei City, Taiwan, ROC) and preparation of a ow cell chip, we subjected 500 ng PCR products to paired-end (PE) sequenced on an Illumina HiSeq 2500 platform (Illumina, Inc., San Diego, CA, USA) at Biomarker Technologies Co, Ltd. (Beijing, China) according to standard protocols.The sequencing length was 350-450 bp.Original image data les were transformed into raw data (PE reads) via base calling analysis.The negative control was not sequenced because it was bandless and sequencing would have been meaningless.

Quality assessment and data processing
According to the overlapping relationship, PE reads were merged with Fast Length Adjustment of SHort reads (FLASH) software v.1.2.7 (Johns Hopkins University Center for Computational Biology, Baltimore, MD, USA; raw tags) [29].We discarded tags with > 6 mismatches.The minimum overlap length was 10 bp, and the maximum mismatch ratio allowed in the overlap region was 0.2 (default).Raw tags with an average quality score < 20 in a 50-bp sliding window were then ltered using Trimmomatic software v.0.33 (USADELLAB.org)[30] and those shorter than 350 bp were removed (clean tags).We further removed possible chimeras using UCHIME v.4.2 (http://drive5.com/usearch/manual/uchime_algo.html)[31] to obtain effective tags.

Operational taxonomic unit analysis
Effective tags were clustered at a 97% similarity level to obtain OTUs using USEARCH software v.10.0 [32].We evaluated the α-diversity index of each sample using Mothur software v.1.30(http://mothur.org/)and the β-diversity index using Quantitative Insights Into Microbial Ecology (QIIME) software v2.2.Degrees of similarity in species diversity between different samples were compared.

Taxonomic analysis
We compared representative OTU sequences against the Silva microbial reference database (release 128; http://www.arb-silva.de)[33].The classi cation information of each OTU was obtained by comparison, and the OTU was annotated using RDPClassi er (v2.2;QIIME) [34].Next, we counted the community composition of each sample at phylum, class, order, family, genus, and species levels.Species richness at different taxonomic levels was assessed using QIIME, and the community structure diagram of each taxonomic level was drawn using R software v3.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

Equipment and settings
Images were captured using the software NIS-Element D using a Nikon-ECLIPSE 80i upright uorescence microscope out tted with a Nikon DS-Ri2 camera, Plan Fluor 40x DIC M N2 objective (objective numerical aperture: 0.75, refractive index: 1.000).The images were acquired at 96 dpi in the x-and y-axis with pixel dimensions of 4908 × 3264, and the image bit depth was 24.

16S rRNA uorescent in situ hybridization
We detected the 16S rRNA in tissues according to the instructions of a EUB338 FISH Probe Kit (#20 μM; FBPC-10; Creative Bioarray, Shirley, NY, USA).Brie y, frozen tissue slides (4 μm) were xed with 4% paraformaldehyde for 30 min and at 60°C for 10 min, followed by incubation with 0.3% Triton X-100 for 30 min.We then incubated the slides in lysozyme at 37°C for 15 min.As a positive control, a bacteria smear was incubated in 0.01 M HCl at RT for 20 min.Slides were xed with 4% paraformaldehyde for another 15 min at RT, after which we treated them with diethyl pyrocarbonate (DEPC) for 10 min and incubated them with BSA (3%) for 2 h.BSA was then discarded, and uorescein isothiocyanate (FITC)labeled probes (EUB338 GCTGCCTCCCGTAGGAGT) and a nonspeci c complement probe (nEUB338 CGACGGAGGG CATCCTCA; #FBPC-13; Creative Bioarray) were hybridized in a pre-warmed humidi ed hybridization chamber and incubated overnight at 39˚C (range, 38-42°C).We diluted FISH probe EUB338 or nEUB338 with 35% hybridization buffer (1:100), denatured it at 84°C for 5 min, and then incubated at 37°C for 3 min.After hybridization, we carefully removed the sealing lm by soaking the slides in wash solution (WS; 2 × saline-sodium citrate [SSC]/0.1% Tween 20) at RT for 15 min to loosen the coverslips.Slides were then rinsed twice in WS for 15 min each time, immersed in 75% and 100% ethanol for 2 min, and then air-dried for 20 min.We counterstained the slides with DAPI antifade solution (using EUB338 FISH Probe Kit) for 10 min and examined them under the Nikon uorescence microscope.

LPS and LTA western blotting
We separated cytoplasmic from nuclear components using a Minute TM Cytoplasmic and Nuclear Fractionation Kit for Cells (#sc-003; Invent, Beijing, China).The liver tissue was washed with precooled sterile PBS, and 60 mg of fresh (or frozen) soft liver tissue was added to a 1.5-ml sterile microcentrifuge tube.We then added 200 μl of precooled sterile PBS to the same tube and ground the tissue with a clean plastic grinding rod for 2-3 min on ice until no solid tissue was visible, after which the frozen liver tissue sample was thawed completely on ice and then ground again.After incubation on ice for 5 min, we carefully transferred the supernatant into another precooled sterile 1.5-ml microcentrifuge tube.Cells were harvested from the suspension by low-speed centrifugation (500 × g) at 4°C for 3 min.We added 200 μl cytoplasmic extraction buffer for every 20 μl of cell volume, vortexed the tube vigorously for 15 s, and incubated it on ice for 15 min.Then, we centrifuged it for 5 min at 16000 × g and 4°C in a microcentrifuge.The supernatant (cytosol fraction) was transferred to a fresh prechilled 1.5-ml tube; the pellet was washed in 0.5 ml cold PBS, centrifuged at 8000 × g for 5 min to reduce contamination of cytosolic proteins and frozen at −80°C.We added 100 μl nuclear extraction buffer to the pellet (ratio of cytoplasmic to nuclear extraction buffer, 2:1), vortexed the mixture vigorously for 30 s, and incubated the tube on ice for 2 min; this sequence of steps was repeated ve times.Immediately afterward, we transferred the nuclear extract to a prechilled lter cartridge via a collection tube and centrifuged it at 16000 × g in the microcentrifuge for 30 s at 4°C.We discarded the lter cartridge and stored the nuclear extract at −80°C.Measurements were repeated three times for every sample, and blanks without liver samples were established.The bicinchoninic acid (BCA) method was used to measure protein concentrations.Protein samples were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; 12% separating gel), transferred onto a polyvinylidene di uoride (PVDF) membrane (MilliporeSigma, Burlington, MA, USA), and probed with the above-indicated primary antibodies at 4°C overnight, followed by the appropriate secondary horseradish peroxide (HRP)conjugated IgG antibody at RT for 1 h.The following antibodies were used: primary, LPS (1:400; HM6011; Hycult), LTA (1:1000; #HM2048; Hycult), Lamin-B1 (1:2000, #ab16048; Abcam, Cambridge, UK), glyceraldehyde 3-phosphate dehydrogenase (GAPDH; 1:4000; #YM3215; Proteintech, Chicago, IL, USA); secondary, HRP-labeled goat anti-rabbit (1:5000, #RS0002; ImmunoWay Biotechnology Co., Plano, TX, USA) and HRP-labeled goat anti-mouse (1:5000; #RS0001; ImmunoWay).Protein bands were visualized using an electrochemiluminescence (ECL) kit (Super ECL Detection Reagent, Yeasen Biotechnology Co., Ltd., Shanghai, China) and ImageJ software v6.0.

Statistical analysis
All data were expressed as mean ± standard deviation.We conducted all statistical analyses using SPSS software v19.0 (IBM Corp., Armonk, NY, USA).Kruskal-Wallis H rank test was used for comparisons between the liver, SI, and LI samples, and a one-way analysis of variance (ANOVA) was used for mean comparison of the distance between the liver, SI, and LI samples.Differences were considered statistically signi cant at P < 0.05.

Liver microbiota detected by 16S rRNA gene sequencing
We sequenced the V3-V4 region of the 16S rRNA gene in liver tissues, obtaining 54,867.50± 6,450.03effective tags (Table S1).These tags were further clustered into 1003 kinds of OTUs (805.67 ± 70.14, 729-893), of which six rats shared 598 kinds of OTUs.Annotated species numbers at different levels are summarized in Table S2.Individuals showed obvious conservation of bacterial abundance (tags), richness (OTU number), and evenness (Table 1).The coe cient of variation (CV) of the α-diversity index was low.At the phylum level, liver microbiota were annotated and clustered into 19 species.No exclusive species was observed among individuals.Firmicutes, Proteobacteria, Bacteroidetes, Acidobacteria, Actinobacteria, Chloro exi, Verrucomicrobia, Spirochaetes, Cyanobacteria, and Gemmatimonadetes were the top 10 dominant bacteria in these rats (Fig. 1).CV ranged from 0.21 to 0.42 (Table S3).Microbiota annotation at the genus level is provided in Table S4.

Liver bacteria detected in situ
To prove the existence and location of bacteria in the liver, we further performed histological bacteria detection in situ.We conducted IF using antibodies against bacterial LPS and LTA to detect Gramnegative and -positive bacteria, respectively [9].We also used RNA FISH with a universal probe against the bacterial 16S rRNA (EUB338) to detect bacterial RNA in the liver [9].To control for nonspeci c staining, IF-negative control (no primary antibody) and FISH-negative control (nEUB338) were also applied to the samples.Bacterial LPS and 16S rRNA were positive in the liver and located in the nuclei of hepatocytes.LTA was not detected (Fig. 2).We then separated the cytoplasmic and nuclear components of liver tissue and detected the expression of LPS and LTA.The results con rmed the nuclear location of LPS (Fig. 3).

Functioning analysis of the genome composition of liver microbiota
According to a Kyoto Encyclopedia of Genes and Genomes (KEGG) database comparison using the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2) algorithm and the tax4fun package in R, the main function composition of the genomes of annotated liver bacteria was related to metabolism (79.92 ± 0.24%).Genetic information processing (6.31 ± 0.10%), environmental information processing (6.29 ± 0.12%), and cellular processes (3.49 ± 0.08%) were less dominant and showed no obvious harm to the body (2.58 ± 0.06%).Moreover, the organismal system function (1.41 ± 0.02%) was low (Table S5).
According to BugBase analysis, liver bacteria possessed diverse phenotypes.The Gram-negative phenotype was about 1.6 times more common than the Gram-positive one, and the aerobic and anaerobic phenotypes were nearly equally common.Moreover, liver microbiota also showed high expression of mobile elements, bio lm formation, and stress tolerance as well as potentially pathogenic phenotypes (Table S6).According to faprotax analysis, the liver bacteria were mainly heterotrophic with chemoheterotrophy (36.21 ± 0.88%), fermentation (20.33 ± 1.61%), and aerobic chemoheterotrophy (15.41 ± 0.93%; Table S7) functions.

Relationship between liver and gut microbiomes
The relationship between the microbiome of the liver and gut was extremely interesting.Therefore, we compared the 16S rRNA gene sequencing data of the liver microbiome with those of the fecal and ileal microbiomes.The OTU number of the liver microbiome (805.67 ± 70.14) was signi cantly higher than those of the SI (490.00 ± 268.20) and LI (430.50 ± 31.97), and a Chi-square of 7.520, P < 0.05 (Fig. 4A).
The liver microbiome shared a large percentage of bacteria with those of the ileum and feces (Fig. 4B).At the phylum level, Proteobacteria and Acidobacteria were dominant in the liver microbiome, while the Firmicutes were lower than in the gut microbiome (Fig. 4C).The heatmap showed an obvious difference between the distributions of the dominant bacteria in the liver and gut microbiomes (Fig. 4D).Except for the abundance distinction, the phylum and species distributions showed no difference (Table S8).At the genus level, among the 337 microbiota, only two bacteria (Psychrobacter and Leptolyngbya VRUC 135) were liver-exclusive, and three (uncultured bacterium o Rhodospirillales, Muribacter, and uncultured bacterium f Rs-E47 termite group) were exclusive to the SI and LI (Table S9), indicating remarkable homogeneity of the liver and gut microbiome.
As shown by α-diversity analysis, richness and evenness in the livers were signi cantly higher than in the SI or LI (Chi-square: ace 9.509, Chao1 8.187, Shannon 11.415, Simpson 10.900, P < 0.05; Fig. 5A-B, Table S10).According to the β-diversity analysis, the liver microbiome exhibited signi cantly higher similarity than the intestinal microbiome (P < 0.05; Fig. 5C-F).The visceral bacteria are widely considered to originate from the gut microbiome via gut leakage of the colonic mucosa (the "gut-to-liver axis").However, the similarity of the liver microbiomes among different individuals being greater than the distance between gut and liver (Fig. 5C-F) seemed to throw this concept into doubt.We then compared the bacteria of the liver with those of the intestine.Liver bacteria showed signi cantly higher richness than intestinal bacteria from phylum to species level (P < 0.05; Table S11).We compared liver bacteria individually with gut microbiota.In each animal, a large number of liver bacteria were exclusive to the liver in comparison with the animal's ileal-and fecal-exclusive bacteria (genus level, liver: 113.67 ± 83.32 vs. ileum: 4.67 ± 2.58 or feces: 12.00 ± 8.15; Chi-square: 8.172; P = 0.017).Moreover, liver-ileum and liver-ileum-feces shared bacteria (107.67 ± 83.73 and 71.33 ± 11.78, respectively) were signi cantly more numerous than liver-feces shared bacteria (14.00 ± 9.44; Chi-square: 17.430; P = 0.001; Fig. S1, Table S12).We also calculated the distance between liver and intestinal samples by the binary Jaccard method.The liver was signi cantly closer to the ileum than to feces (liver to liver: 0.09 ± 0.02; liver to ileum: 0.44 ± 0.25; liver to feces: 0.74 ± 0.03; P < 0.05; Fig. 6, Table S13).These relatively independent phenomena challenged the concept of the "gut-to-liver axis," suggesting instead an upstream "liver-to-gut axis" and the possibility of bacterial translocation from the liver to the gut (Fig. S2).

Discussion
Gut bacteria were largely overlooked until they were found to be closely related with various gut or gutrelated disorders, such as irritable bowel syndrome [35][36][37], infectious diarrhea [38], ulcerative colitis, colorectal cancer, liver diseases, and obesity-related disorders [39,40].In addition, gut bacteria can be utilized as a medication strategy [41][42][43][44][45], including fecal microbiota transplantation (FMT) [46][47][48][49][50][51].Gut bacteria recently became an explosive topic, going through roughly three development stages: (1) Feces microbiome.Findings have shown that various disorders in nearly every organ and system are associated with changes in the gut microbiome (samples primarily from feces) [52][53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68], and this has established the connection between the gut microbiome and diseases.It was suspected that gut microbiota and their metabolites, such as LPS, enter circulation (such as in bacteremia and toxemia) and result in organ injury [69,70].However, whether the disease-associated changes in gut microbiota occur before or after disease development cannot be determined.(2) Tissue microbiota in disease.The bacterial 16S rRNA gene as well as LPS have been detected in various tissue lesions, including tumor cells [14], leukocytes and platelet fractions [19], liver tissues of NAFLD patients [26], as well as in adipose tissues [10].Although whether these bacteria were intact and alive remains unclear, these ndings demonstrate the intimate relationship between the gut microbiome and various diseases.The key question is where these diseases arise from.The most accepted answer to this question may involve the process of gut leakage.Another question is whether they reside in the tissue before or after disease.(3) Tissue microbiota in health.Using an optimized 16S metagenomic sequencing pipeline, the bacterial 16S rRNA gene was reported in normal tissues in mice, including the brain, muscle, fat, heart, and liver [15].This report suggested that tissue microbiota may exist in the tissue prior to the occurrence of disease.
In our study, the bacterial 16S rRNA gene was detected in the liver tissues of healthy rats.In addition, we also performed the in situ detection of EUB338, LPS, and LTA using FISH and IF, and LPS and LTA detection of nuclear and plasma extracts using western blotting.We proved that the LPS was located in the nucleus.These results strongly suggested the existence of bacteria in the liver microbiome as hidden inhabitants in the normal hepatocytes of rats.To our knowledge, in addition to the positive detection of the 16S rRNA gene, this is the rst report of intracellular bacteria in normal parenchymal cells of the liver.
This study switched the focus of microbiome research from the gut to the parenchymal cells.The special location and bacterial variation in the liver strongly suggested that the microbiome present in the liver was signi cantly different from that present in the gut.We thus named these hidden inhabitants of the liver microbiome in the normal hepatocytes of rats as the "liver microbiome." With the help of metagenomic sequencing, we identi ed the existence of bacteria in the liver tissues of SD rats.We repeated the 16S rRNA gene sequencing experiment three times on two sequencing platforms (Beijing Biomarker Technologies Co., Ltd., www.biomarker.com.cn,Project Number: BMK191202-X098-01, BMK200916-AC763-0101, and Shanghai Biopro le Technology Co., Ltd., Shanghai, China, www.biopro le.cn, Project Number: BP20123) in ordinary or clean rats (BMK200916-AC763-0101, data not shown) and repeated the in situ detection in clean rats more than three times.All results were consistent.Because the sequencing of the 16S rRNA gene's V3-V4 region is a mature technique and a strictly controlled process, and because the technique for annotation of bacteria has been tested and widely used, these results were credible.
Bacteria in liver tissues were high in richness and abundance.A total of 1003 kinds of OTU were identi ed in six rats with 54,867.50± 6450.03 effective tags.Their α-diversity was even higher than that of intestinal bacteria.The most striking features of liver bacteria might be the conservation of richness, abundance, and evenness and higher similarity among different individuals compared to gut bacteria.
We further con rmed the results by LPS/LTA and EUB338 detection in situ.We found that bacteria were located in the nuclei of hepatocytes and were well organized.Unlike LPS and EUB338, LTA was not detectable as previously reported [26].We hypothesized that LTA might be a less sensitive biomarker for the detection of Gram-positive bacteria in hepatocytes, or that such bacteria in hepatocytes might differ structurally from environmental Gram-positive bacteria (e.g., lacking expression of LTA), because functional analysis of the genome compositions showed that the liver microbiome was highly Grampositive and showed high percentage of chemoheterotropy, fermentation, and aerobic chemoheterotrophy.After we isolated liver tissues into the cytoplasm and nuclear components, western blotting also proved there was a strong expression of LPS in the latter, with negative in the former.LTA was not detected in both types of components.
After the identi cation of the liver bacterial locations, two key questions might arise: (1) where are the bacteria from, and ( 2) what functions do they have and by what mechanisms are they performed?The visceral bacteria have long been considered to come from the gut microbiome through the mechanism of gut leakage [13].In this study, the liver microbiome did share bacteria with the gut microbiome, with only a few exclusive bacteria.However, in each individual, a large proportion of the liver microbiome was exclusive when compared with its gut microbiome counterpart.Moreover, in this study, the liver bacteria showed signi cantly higher α-diversity than the intestinal bacteria.Even if the bacteria were from the gut microbiome, how and when did they enter the liver and locate themselves in the nucleus in a wellorganized manner in healthy individuals?This result challenges the gut-liver axis in which the gut is in the upstream position.One reasonable explanation is that the liver microbiota are originating from the continuing leakage of bacteria from the gut.The gut microbiota showed variation and dynamic changes caused by diet and environment [71][72][73][74][75][76], whereas the liver microbiota remained constant.Thus, this exclusivity can be acceptable to some extent.Based on the special location and difference of the gut microbiota, the liver-gut axis can be deduced, in which the liver microbiota may transport to the gut through bile secretion [26,77].The gut and liver may establish a microbiota translocation circulation (Fig. S2).
However, another question arises: why were the liver microbiome of different individuals signi cantly more similar than the gut microbiome?These ndings suggest a bold hypothesis and another possibility for the origin of the liver microbiota: that the intrinsic components of the hepatocytes are inherent.
Recently, we detected the intracellular microbiota in other visceral organs of SD rats with 16S rRNA gene sequencing, and also detected the cultured cell lines HepG2, Huh-7, Hepa1-6, and HSC-T6 using the western blotting method with isolated cytoplasm and nucleus extractions as described above.Furthermore, we obtained consistent results (Cellular Microbiota: An Inherent Inhabitant of Cells, under review by Cellular Microbiology).However, we still need to make further efforts to design experimental plans to prove the existence and composition of the liver microbiome in human healthy liver tissues.Our results provide a profound understanding and reasonable explanation for the positive detection of intrahepatic bacteria [26,27,59,77] and support the hypothesis that the liver microbiomes are inherent components of hepatocytes.
The roles of the liver microbiome are also unknown.According to the functional analysis, the functions of liver bacteria were primarily metabolic rather than those that would harm the host.This was consistent with the characteristics of native bacteria.In terms of whether these were parasitic or symbiotic, and whether these bacteria were involved in the regulation of gene expression or cell life activities, we found that the question of the origin of life may be involved.This should be further proven the way the existence of oncogenes was.Furthermore, as far as the harm to the body is concerned, are the liver bacteria associated with the development of various metabolic diseases, such as steatosis, insulin resistance, or hepatic cancer?

Conclusion
We found that the liver microbiome existed in hepatocytes, of which its bacterial species might be inherent inhabitants.The sources and functions (transcriptomic) of these bacteria in cell life activities and disease development should be investigated in the future.

Figure 5 Alpha
Figure 5

Table 1 .
Alpha diversity among individual liver samples CV, coefficient of variation, calculated as standard deviation (SD) divided by mean.Simpson index was calculated as 1 -∑pi 2 .