E. coli NF73-1 Isolated From NASH Patients Aggravates NAFLD in Mice by Translocating Into the Liver and Stimulating M1 Polarization

Objective The gut microbiota is associated with nonalcoholic fatty liver disease (NAFLD). We isolated the Escherichia coli strain NF73-1 from the intestines of a NASH patient and then investigated its effect and underlying mechanism. Methods 16S ribosomal RNA (16S rRNA) amplicon sequencing was used to detect bacterial profiles in healthy controls, NAFLD patients and NASH patients. Highly enriched E. coli strains were cultured and isolated from NASH patients. Whole-genome sequencing and comparative genomics were performed to investigate gene expression. Depending on the diet, male C57BL/6J mice were further grouped in normal diet (ND) and high-fat diet (HFD) groups. To avoid disturbing the bacterial microbiota, some of the ND and HFD mice were grouped as “bacteria-depleted” mice and treated with a cocktail of broad-spectrum antibiotic complex (ABX) from the 8th to 10th week. Then, E. coli NF73-1, the bacterial strain isolated from NASH patients, was administered transgastrically for 6 weeks to investigate its effect and mechanism in the pathogenic progression of NAFLD. Results The relative abundance of Escherichia increased significantly in the mucosa of NAFLD patients, especially NASH patients. The results from whole-genome sequencing and comparative genomics showed a specific gene expression profile in E. coli strain NF73-1, which was isolated from the intestinal mucosa of NASH patients. E. coli NF73-1 accelerates NAFLD independently. Only in the HFD-NF73-1 and HFD-ABX-NF73-1 groups were EGFP-labeled E. coli NF73-1 detected in the liver and intestine. Subsequently, translocation of E. coli NF73-1 into the liver led to an increase in hepatic M1 macrophages via the TLR2/NLRP3 pathway. Hepatic M1 macrophages induced by E. coli NF73-1 activated mTOR-S6K1-SREBP-1/PPAR-α signaling, causing a metabolic switch from triglyceride oxidation toward triglyceride synthesis in NAFLD mice. Conclusions E. coli NF73-1 is a critical trigger in the progression of NAFLD. E. coli NF73-1 might be a specific strain for NAFLD patients.


INTRODUCTION
The gut-liver axis is the passageway by which the gut microbiota and gut-derived hepatotoxic bacterial products can easily enter the liver (Wiest et al., 2017;Chu et al., 2019). Interplay between the gut microbiota and liver injury through the gut-liver axis is known to play a pivotal role in the development and progression of nonalcoholic fatty liver disease (NAFLD) (Schnabl and Brenner, 2014;Chu et al., 2019). Emerging evidence has demonstrated that the gut microbial profile in NAFLD and/or nonalcoholic steatohepatitis (NASH) patients is strikingly different from that of healthy subjects (Safari and Gerard, 2019;Kim et al., 2019;Kolodziejczyk et al., 2019). NAFLD patients generally exhibit decreased microbial diversity compared with healthy controls (Del Chierico et al., 2017). In addition, Proteobacteria, which includes Escherichia, is the most common differentially abundant taxon between obese patients with NASH and those without NASH (Boursier et al., 2016). Many studies have found that germ-free or antibiotic-treated (bacteria-depleted) mice exhibited less lipid accumulation and inflammation than conventional mice fed a high-fat diet (HFD), which highlights the pivotal role of the gut microbiota in the natural progression of NAFLD (Backhed et al., 2004;Brandt et al., 2017;Wang et al., 2018). Moreover, after being administered the microbiota of NASH patients orally by gavage, HFD-fed germ-free mice showed an exacerbated NASH phenotype, manifesting as increased liver steatosis and inflammation (Chiu et al., 2017). The gut microbiota, in brief, has been indicated to play a critical role in the development of NAFLD.
Until now, most studies have determined the bacterial composition of intestinal feces at the genus level but have failed to identify bacteria at the species or strain level. In previous studies, a common finding in NAFLD patients was an increase in Escherichia at the genus level (Zhu et al., 2013;Mouzaki et al., 2013;Jiao et al., 2018). Furthermore, it has been reported that Escherichia is a unique genus that is significantly elevated in NASH patients compared with healthy controls and obese patients . Another study confirmed that Escherichia was the predominant bacterium in NAFLD patients with small intestinal bacterial overgrowth (SIBO) (Kapil et al., 2016). Additionally, based on our previous research, a higher abundance of fecal Escherichia was observed in the Chinese population with NAFLD than in healthy subjects; of note, the abundance of Escherichia increased gradually in the progression of NAFLD to NASH (Jiang et al., 2015). These facts suggest that Escherichia can be a risk factor in the disease progression from NAFLD to NASH.
Of note, several studies have shown that multiple bacterial genera are associated with various diseases, and different strains of the same genus might have different pathogenic mechanisms (Kostic et al., 2013;Tsoi et al., 2017;Yu et al., 2017). Escherichia coli is an opportunistic pathogen that exists in NAFLD patients and healthy controls and is generally considered safe under normal health conditions. Increased bacterial colonization by E. coli or increased expression of virulence-related genes of E. coli may be inherently pathogenic (Kaper et al., 2004;Zargar et al., 2015). In addition, translocation of gut microbes was reported to trigger the progression of liver disease. Enterococcus gallinarum was reported to translocate into the liver in autoimmune hepatitis (AIH) (Manfredo Vieira et al., 2018). Previous studies have also shown that NAFLD is influenced by pathological bacterial translocation (PBT) from the gut . Therefore, we became interested in the effect and mechanism of gut microbes, especially E. coli, in the pathogenic progression of NAFLD.
Few studies have focused on the mucosal microbiota, especially bacterial strains, from the intestinal mucosa of clinical patients. In this study, to determine the relative bacterial abundance in mucosal samples at the genus level among healthy, NAFLD and NASH groups. We performed 16S ribosomal RNA (16S rRNA) amplicon sequencing. The data showed that, compared with the healthy group, the relative abundance of E. coli was increased in the NASH and NAFLD groups, especially in the NASH group, which was consistent with the changes in stool samples. These results further confirmed the accumulation of E. coli in the pathogenic progression from NAFLD to NASH. Bacterial cultivation, isolation and fulllength 16S rRNA sequencing were performed to cluster the strains of E. coli enriched in NASH patients. We named the most abundant E. coli strain E. coli NF73-1 (authorized patent in China: No. 201610591293.4). In addition to bacterial abundance, the expression of virulence-related genes was also reported to be inherently pathogenic (Kaper et al., 2004;Zargar et al., 2015); thus, we performed whole-genome sequencing of E. coli NF73-1. The data showed enrichment of adherence genes and pathogenic genes, such as fimH (Chen et al., 2009), ybt (Ellermann et al., 2019), etc. in the genome of E. coli NF73-1 (Table S1). Based on its highest abundance in NASH patients and its genetic pattern, we inferred that E. coli NF73-1 played a pivotal role in the progression of NAFLD.
The overall aim of the current study was to investigate the effect of E. coli NF73-1 in the progression of NAFLD, especially in liver inflammation and lipid metabolism.

Patients and Mucosa Collection
This cross-sectional study was approved by the Conjoint Health Research Ethics Board of Peking University People's Hospital, and informed consent forms were signed by all the subjects prior to participation in this study. The diagnosis of NAFLD was established according to the Gastroenterology Organization Guidelines (Chalasani et al., 2012).
A total of 19 NAFLD patients were recruited at Peking University People's Hospital. All patients were confirmed to exhibit NAFLD based on either imaging or histological evidence. Additionally, 10 healthy subjects were invited to participate as controls ( Table 1). The 29 subjects agreed to undergo gastroscopy for physical examination and signed the appropriate informed consent form. In all cases, the mucosa samples were collected prior to endoscopy. On the day prior to the gastroscopic procedure, a blood sample was collected to measure the metabolic and hepatic parameters.

Mucosal Bacterial 16S rRNA Sequencing and Analysis
Microbial genomic DNA was extracted according to the manufacturer's instructions, with minor modifications (Qiagen, Hilden, Germany). After DNA extraction, bacterial 16S rDNA was amplified. Briefly, the V3-V4 region of 16S rDNA was amplified using paired primers (357F/806R). After library preparation and quantification (Thermo Scientific, USA), paired-end sequencing (2 × 150 bp) was performed on an Illumina HiSeq 2500 sequencer at the Center for Molecular Immunology of Chinese Academy of Sciences (Beijing, China). In addition, raw fastq files were quality-filtered using the fastq_quality_filter (-p 90 -q 25 -Q33) in FASTX-Toolkit (v.0.0.14) and merged by Fast Length Adjustment of SHort reads (FLASH) (Caporaso et al., 2010;Edgar et al., 2011;Magoc and Salzberg, 2011). The OTUs were aligned utilizing the UCLUST algorithm with 97% identity and taxonomically classified using the SILVA database, v128, released on 29/09/ 2016. Additionally, the relative abundances of the various phyla, classes, orders, families, and genera in each sample were computed and compared among all groups. The detailed methods have been described in our previous studies Jun et al., 2019).

Identification and classification of E. coli in NAFLD/ NASH
Based on the 16S rRNA sequencing analysis, the significantly enriched E. coli were chosen for further study. First, E. colispecific culturing was performed to pick pathogenic strains. Approximately 100 strains of E. coli were cultured and subjected to full-length 16S rRNA sequencing. The data showed 99% similarity with E. coli in the NCBI database. To analyze the evolutionary relationship among the strains, a phylogenetic tree was established with kSNP3 software (Gardner et al., 2015) using the neighbor-joining (NJ) method. Based on the similarity of genomic sequences, three types of E. coli subspecies were identified, i.e., NF73, NF4 and NF23.
E. coli-specific Comparative Genomic Analysis E. coli NF73-1 was used for full-length genomic sequencing and further comparative genomic analysis. The genome of E. coli NF73-1 was sequenced using an Illumina HiSeq 4000 system (Illumina, San Diego, CA, USA) at the Beijing Genomics Institute (Shenzhen, China). The sequenced reads were assembled using SOAPdenovo v1.05 software.
For phylogenetic analysis based on full-length 16S rRNA, the evolutionary history was inferred by using the maximum likelihood method based on the Tamura-Nei model. Evolutionary analyses were conducted in MEGA7. A phylogenetic tree based on the whole-genome sequence was constructed by TreeBeST using the NJ method.
Gene prediction was performed on the E. coli NF73-1 genome assembly by glimmer3 with hidden Markov models. tRNA, rRNA and sRNA recognition was conducted with tRNAscan-SE (Lowe and Eddy, 1997), RNAmmer, and the Rfam database. Tandem repeat annotation was obtained using the Tandem Repeat Finder (http://tandem.bu.edu/trf/trf.html), and the minisatellite DNA and microsatellite DNA were selected based on the number and length of repeat units. The Genomic Island Suite of Tools (GIST) was used for genomic island analysis (http://www5.esu.edu/cpsc/ bioinfo/software/GIST/) with the IslandPath-DIOMB, SIGI-HMM, and IslandPicker methods. Prophage regions were predicted using the PHAge Search Tool (PHAST) web server (http://phast.wishartlab.com/), and CRISPR identification was performed using CRISPR Finder. The best hit was abstracted using the BLAST alignment tool for functional annotation. Database, namely Kyoto Encyclopedia of Genes and Genomes (KEGG) were used for general function annotation.
According to the dispensable gene in different strains, a heat map showed the cluster analysis among E.coli strains. A detailed description of genomes can be found in Table S1. Diamond (V0.9.18.119) software was used to compare the sequences of 2 strains in the KEGG database, and the annotation information for KEGG was obtained (Kanehisa et al., 2014). The key genes were selected to compare and display the heatmap.
At the time indicated, animals fasted for 12 h were anesthetized, and whole blood was withdrawn by cardiac puncture. Serum and tissues were collected, including the liver and intestine.

Stool DNA Extraction
For optimal isolation of bacterial DNA, mucosal biopsies were disrupted by bead-beating after digestion with an enzymatic cocktail of mutanolysin and lysozyme (Sigma-Aldrich, USA). Then, extraction and purification were performed using the QIAamp DNA Mini Kit. Stool DNA was quantified using a Nanodrop-1000 (Thermo Scientific, USA).

Serum biochemical measurements
Serum was collected from blood and assayed for levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), total triglyceride (TG) and cholesterol (TC) using an automatic biochemical detector (Labospect 008 AS, Japan).

Oil Red O Staining and H&E Staining
For Oil Red O staining, optimal cutting temperature (OCT)embedded frozen tissue was sectioned at 5 to 7 mm. After washing with PBS, dried slides were subsequently incubated with 100% propylene glycol and Oil Red O solution (ab150678; Sigma, USA). Stained slides were differentiated with 85% propylene glycol solution and distilled water. Then, stained slides were incubated in hematoxylin and mounted with glycerin.
For hematoxylin and eosin (H&E) staining, liver tissue was fixed in 10% neutral buffered formalin (Sericebio; China). Paraffin-embedded liver tissues were used. The NAFLD activity score was assessed by two blinded investigators.

Immunochemical Staining
Liver tissues were removed at sacrifice and then fixed in formalin, embedded in paraffin, and prepared for histology. The liver tissue slides were blocked with 10% goat serum for 30 min, followed by incubation for 1 h with anti-F4/80 antibody (1:200; ab16911, Abcam, China) and anti-pS6 antibody (1:200; 2211, CST, USA) at 4°C overnight. After incubation with a poly-peroxidaseconjugated goat anti-rat IgG or goat anti-rabbit IgG (Zhongshan Golden Bridge, China) at 37°C for 30 min, the slides were counterstained with 3,3′-diaminobenzidine (DAB) and hematoxylin and eventually mounted with glycerin.

RT-qPCR
Total RNA from the liver and macrophages was isolated TRIzol reagent (15596; Thermo Fisher Scientific) and reverse transcribed to complementary DNA by the qScript Reverse Transcriptase Kit (Thermo Fisher Scientific, USA). Gene expression profiles were quantified by quantitative PCR using the ABI 7500 Real-Time PCR System (ABI) with SYBR FAST qPCR Kits (Transgen, China). mRNA amounts were calculated by the comparative cycle threshold method (DDCt method) with normalization to actin mRNA. Primers were purchased from Sangon Company (Shanghai) and are listed in Table 2.

Liver TG and TC Levels
Hepatic lipids were extracted from 50 mg of liver tissue and assayed using a triglyceride assay kit and a cholesterol assay kit (E1013, E1015, Applygen Technologies, China) according to the manufacturer's recommended protocols. Protein concentrations were quantified by using a Pierce BCA Protein Assay Kit (23227, Thermo, USA).
Cell Coculture and Oil Red O Staining (RAW 264.7, L02, H2.35, E. coli NF73-1) RAW 264.7 cells (obtained from ATCC and free from mycoplasma) were stimulated with 1 mM palmitate and oleic acid conjugated to fatty acid-free bovine serum albumin (BSA) (A2000, Applygen Technologies, China) or only fatty acid-free BSA for 1 h and then cultured with or without NF73-1. Then, surface immunofluorescence antibody staining was performed quickly for identification of macrophages. L02 and H2.35 (obtained from ATCC and free from mycoplasma) were stimulated with 1 mM palmitate and oleic acid conjugated with fatty acid-free BSA or only fatty acid-free BSA for 20 h and then cocultured with or without NF73-1 for 4 h. After washing with PBS, the cells were incubated with 4% neutral buffered formalin and Oil Red O solution (ab150678; Sigma, USA). Stained slides were washed with 85% propylene glycol solution and distilled water and then mounted with glycerin.
Bacterial Culture E. coli NF73-1 and E. coli MG1655 were transformed with plasmid CV129-EGFP to induce the expression of enhanced green fluorescent protein (EGFP) and b-lactamase, conferring resistance to ampicillin, to follow bacterial translocation from the intestine. E. coli EGFP-NF73-1 and E. coli EGFP-MG1655 were cultured in LB under aerobic conditions. Primers of EGFP were purchased from Sangon Company (Shanghai) and are listed in Table 2.
All experiments were performed under strict sterile conditions. At the time indicated, livers (100 mg each) were removed aseptically, homogenized, plated on ampicillinsupplemented Luria Broth plates and cultured at 37°C overnight. The CFUs were then counted to quantify bacteria.
Coculture of E. coli NF73-1, RAW264.7, and Hepatocytes RAW264.7 macrophages (2 × 10 3 cells/ml) were cultured in DMEM with oleic acid and palmitic acid in the upper chamber, and L02 and H2.35 (2 × 10 5 cells/ml) were cocultured in the lower chamber of a 6-well Transwell plate (3412; Corning, USA) for 20 h; the chambers were separated by a 0.4-µm track-etched membrane. Then, E. coli NF73-1 or rapamycin was added to the coculture system. After 4 h of cultivation, hepatocytes were collected for Western blot analysis.
In another experiment, after 4 h of cultivation with E. coli NF73-1, RAW264.7 cells were collected and labeled with APC-Cy7-F4/80, CD11b-PerCP, BV421-CD86, and CD206-FITC antibodies and detected using a Gallios (Beckman, USA) flow cytometer. Moreover, RAW264.7 cells were harvested for RNA extraction and Western blot analysis. The gating strategy is illustrated in Figure S3E. The data were analyzed using Beckman software.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism (version 5.01, San Diego, CA, USA). The results are shown as the mean ± S.E.M. (standard error mean). The results were analyzed by one-way analysis of variance between multiple groups when appropriate, and by a two-tailed Student's t test for comparison of two groups. P < 0.05 was considered statistically significant.

Isolation and Identification of E. coli Strains From the Intestinal Mucosa of NASH Patients
To explore the key bacterial strains involved in the pathogenesis of NAFLD, especially mucosa-adherent bacterial strains, we analyzed the abundance of Escherichia in mucosal samples from NAFLD patients recruited at Peking University People's Hospital. An increased abundance of bacteria belonging to Escherichia within the phylum Proteobacteria was observed in the microbiota from the NASH and NAFLD groups compared with the healthy group. Seven genera, namely Bacteroides, Faecalibacterium, Prevotella, Blautia, Lachnospiracea, Clostridium, and Roseburia, were less abundant in both the NAFLD and NASH patients than the healthy subjects ( Figure  S1A). In addition, the relative abundance of Escherichia in the NASH group was higher than that in the NAFLD group ( Figure  1A). The changes in Escherichia in the mucosal flora of these groups were consistent with the fecal flora in our previous research. In our data, the average relative abundance of E. coli in the healthy control group was 0.14 ± 0.5. Therefore, we define an E. coli colonization abundance less than 0.14 as "lowabundance colonization" and greater than 0.14 as "highabundance colonization". It was found that 40% of the healthy controls were high-abundance colonized populations, while the proportions for NAFLD and NASH patients were 80% and 77.8%, respectively ( Figure 1B).
We selected and isolated E. coli strains from the colonic mucosa of the NASH patient who had the highest abundance of Escherichia (Table 3 and Figures S1B-E). The genotypic identification of the above E. coli strains was carried out by full-length 16S rRNA sequencing of all 100 strains. The sequences showed 99% similarity with E. coli in the NCBI database. To analyze the evolutionary relationships among these strains, a phylogenetic tree was established using the maximum-likelihood method. The results showed that the 100 strains grouped into 3 clusters, which were named E. coli NF73-1 to NF73-73, E. coli NF23-1 to NF23-23 and E. coli NF4-1 to NF4-4 ( Figure S1F). The 16S rRNA sequence of different E. coli NF strains have been uploaded to the NCBI GenBank database (MT649758-MT649857). Thus, E. coli NF73-1-NF73-73 were the dominant strains of Escherichia in the colonic mucosa of NASH patients. Therefore, E. coli NF73-1 was chosen for genome sequencing and verification of the pathogenic effect on NAFLD progression.

Whole-Genome Sequencing Analysis of the E. coli NF73-1 Strain
The genome of E. coli NF73-1 (authorized patent in China: No. 201610591293.4) was sequenced using an Illumina HiSeq 4000 system (Illumina, San Diego, CA, USA) at the Beijing Genomics Institute (Shenzhen, China). All experiments were conducted by HUADA GENE Company (Project Code. F17FTSNCKF1064_ECOaxmR). The sequenced reads were assembled using SOAPdenovo v1.05 software, leading to a final assembly of 113 contigs with an N50 of 75,968 bp ( Table 4). The results of genomic component information analysis (Seq type scaffold) are presented in Table 5. The values for total length values were 4,626,248 (scaffold) and 4,612,049 (contig) depending on the seq type. The genomic sequence of E. coli NF73-1 has been uploaded to the SRA database (BioProject ID: PRJNA529543).
The phylogenetic tree constructed by TreeBeST showed that E. coli NF73-1 was distinguished from other strains and shared  Table S1. (E) Comparative genomics of E. coli NF73-1 and E. coli MG1655 in KEGG general functional annotations. Gene names are represented on Y-axis and the E. coli genomes are listed on the X-axis. The color key representing the degree of differences of the genomes based on the presence or absence of genes. Detail description of genomes can be found in Table S3. Data are calculated as the mean ± SEM, n = 9-10 per group. **p <0.01 (one-way analysis of variance between multiple groups).
relatively close evolutionary relationships with E. coli ETEC.H10407 and E. coli K12.MG1655 ( Figure 1C). These data suggested that E. coli NF73-1, which was isolated from the intestinal mucosa of NASH patients, possessed several critical and pathogenic genes and differed from known strains. The analysis is shown in the Table S1. The genomes were assembled and annotated, and cluster analysis was performed ( Figure 1D). A total of 4372 genes were obtained, of which 3241 were successfully annotated in KEGG databases ( Figure S2A and Table S2). Through comparative genomics analysis with E. coli MG1655, it was found that virulence genes related to adhesion and invasion were more abundant in E. coli NF73-1 than E. coli MG1655. Virulence genes related to secretion system, manganese/iron transport system etc. only existed in E. coli NF73-1 ( Figure 1E and Table S3).
In addition, the abundance of Escherichia was commonly increased in not only the stool samples but also the intestinal mucosa of NASH patients (Jiang et al., 2015). We speculated that the highly abundant E. coli NF73-1 may be an important pathogen causing NAFLD progression.

E. coli NF73-1 Aggravated Hepatosteatosis and Inflammatory Cell Infiltration in NAFLD Mice
Considering the enrichment of E. coli NF73-1 in the intestinal mucosa sample, we investigated the potential effect of this particular bacterial strain on NAFLD progression (Figure 2A). To further confirm whether E. coli NF73-1 independently caused the progression of NAFLD, we utilized flora-deficient mice in our gavage model to exclude the potential impact of other symbiotic gut microbes and repeat the experiments. None of the groups showed abnormal signs or symptoms. E. coli NF73-1 did not result in mouse death. E. coli NF73-1-treated HFD ( Table 6) mice showed significantly higher weight gain than nontreated mice ( Figure S2B) We did not observe significant differences in weight gain in ND mice harboring high E. coli NF73-1 abundance.
Both small and large lipid droplet vacuoles within hepatocytes were significantly increased in the HFD-NF73-1 group compared with the HFD-LB and HFD-MG1655 groups, which was confirmed by liver H&E and Oil Red O staining ( Figures 2B, C). Notably, only the HFD-NF73-1 group exhibited much higher hepatic triglyceride levels than paired-fed HFD mice ( Figure 2D).
After antibiotic cocktail treatment, bacterial abundance decreased significantly in stool and was maintained at a very low level ( Figure S3A), indicating the establishment of natural gut flora depletion. Consistent with the conventional mouse model, elevated hepatic inflammation levels, triglyceride     content and serum liver enzymes were also observed in the HFD-ABX-NF73-1 group compared with the paired-fed HFD-ABX mice ( Figures 2B-G).
Translocation of E. coli NF73-1 via the Gut-Liver Axis Is Observed in HFD but Not ND Mice To follow bacterial translocation, EGFP-labeled, ampicillinresistant E. coli EGFP-NF73-1 and E. coli EGFP-MG1655 were generated and used in animal experiments ( Figure S3B and Figure 3A).
Therefore, regardless of antibiotic cocktail treatment, E. coli EGFP-NF73-1 can adhere to the intestinal mucosa and translocate into the liver in NAFLD mice, but not in ND mice, via the gut-liver axis.

E. coli NF73-1 Promoted M1 Macrophage Polarization in NAFLD Mice
We subsequently attempted to identify the liver immune cell subtypes of NAFLD mice by flow cytometry analysis after E. coli NF73-1 administration.
Collectively, E. coli NF73-1 might promote the progression of steatohepatitis by acting on hepatic M1 macrophage polarization.

Depletion of Macrophages Attenuates E. coli NF73-1-Induced Hepatic Inflammation and Hepatosteatosis
We treated HFD and flora-deficient HFD mice with liposomeencapsulated clodronate (CLOD) to deplete macrophages or liposome-encapsulated PBS as a negative control ( Figure 5A). Macrophages were significantly depleted by injection of CLOD in the HFD-NF73-1 and HFD-ABX-NF73-1 groups ( Figure 5B). Liver H&E staining revealed that the depletion of polarized macrophages could significantly reduce hepatic steatosis and inflammation ( Figure 5C). Liver enzymes and NAS were dramatically alleviated in the HFD-NF73-1 and HFD-ABX-NF73-1 groups treated with CLOD ( Figures 5D, E). These results suggested that removing polarized M1 macrophages significantly ameliorated the liver damage and lipid abnormalities caused by E. coli NF73-1 in NAFLD mice.

E. coli NF73-1 Induces the Disruption of Triglyceride Synthesis and Oxidation in NAFLD Mice
An excess of fat accumulation was also accompanied by upregulation of sterol regulatory element binding protein-1 (SREBP-1) and downregulation of peroxisome proliferatoractivated receptor alpha (PPAR-a) . E. coli NF73-1-mediated upregulation of SREBP-1 and its target genes, including acetyl-coenzyme A carboxylase (ACC1) and fatty acid synthase (FAS), was significantly higher than in the HFD-LB group ( Figure 7A). Moreover, Western blot analysis showed that E. coli NF73-1 exposure caused significant impairment in fatty acid oxidation, as evidenced by decreased expression of the nuclear receptor PPAR-a ( Figure 7A).
Furthermore, immunohistochemical analysis revealed that elevated positive signals of S6 phosphorylation were colocalized primarily in the cytoplasm of lipid droplet-rich hepatocytes of the HFD-NF73-1 group ( Figure 7B). Constitutively activated mTOR greatly elevates de novo lipid synthesis (Duvel et al., 2010;Chen et al., 2018). Phosphorylation of mammalian target of rapamycin (mTOR) at Ser2448 was increased in the HFD-NF73-1 group  compared with the HFD-LB group ( Figure 7C). Thr389 phosphorylation of S6K1, which is modulated by mTOR, was increased. The phosphorylation of the S40 ribosomal protein S6 at Ser235/Ser236, a well-characterized substrate of S6K1 (S6 kinase 1), was increased ( Figure 7C).
In addition, regardless of whether antibiotics were used, the expression of the above genes presented a consistent trend. Collectively, E. coli NF73-1-mediated disruption of mTOR-S6K1-SREBP-1/PPAR-a signaling caused the disruption of triglyceride synthesis and oxidation in NAFLD mice.  We investigated whether E. coli NF73-1 is involved in regulating triglyceride accumulation in hepatocytes following translocation ( Figure S4A). We proposed that E. coli NF73-1 is a promoting factor of triglyceride accumulation in hepatocytes. However, lipid accumulation in oleic acid-and palmitic acid-induced L02 and H2.35 cells treated with or without E. coli NF73-1 was similar ( Figure S4B). E. coli NF73-1 cannot restore the expression of lipolytic genes (PPAR-a, SREBP-1 and p-mTOR) in hepatocytes ( Figure S4C). Our results suggest that E. coli NF73-1 could not directly promote lipid accumulation in hepatocytes.
In vitro, E. coli NF73-1 enriched lipid accumulation in oleic acid-and palmitic acid-induced L02 and H2.35 cells via M1 polarized RAW264.7 macrophages ( Figure 8A), respectively, which was more significant than the accumulation observed in the non-E. coli NF73-1 coculture group. The expression of p-mTOR and its downstream genes was altered consistently between the 2 cell lines, leading to increased lipid synthesis and reduced lipid oxidation catabolism ( Figures 8B, C). These pathways coincide with those observed in NAFLD mice, consistent with lipid dysmetabolism. After adding rapamycin, an inhibitor of p-mTOR, the lipid accumulation and gene expression profiles were significantly reversed and showed consistent changes in the two cell lines (Figures 8B, C).
These experiments showed that E. coli NF73-1 could increase triglyceride synthesis and reduce the catabolism of hepatocytes via the mTOR pathway through the activation of M1 macrophages. A B C FIGURE 7 | Activation of mTOR signaling and promotion of NAFLD in E. coli NF73-1-fed mice. (A) Increasing nuclear translocation of SREBP-1(N) and expression of SREBP-1 (P), as well as its targets, such as ACC1 and FASN, after E. coli NF73-1 treatment and decreasing PPAR-a expression after E. coli NF73-1 treatment. (B) Positive immunostaining for phosphorylated S6 (brown color), the predominant location in lipid-rich hepatocytes (red arrows) of NAFLD mice. (C) Representative immunoblots of phosphorylation of mTOR, S6K1, and S6 in livers from three mice in each group. Total liver lysates of the indicated genotypes were subjected to western blot analysis and densitometry analysis. Data are means ± SEM (n = 3). NAFLD, non-alcoholic fatty liver disease; mTOR, mammalian target of rapamycin; N, cleaved nuclear (68 kDa) form of SREBP-1; P, precursor (125 kDa) form of SREBP-1; S6K1, S6 kinase 1; SREBP-1, sterol-regulatory element binding protein-1; ACC1, acetyl-coenzyme A carboxylase; FASN, fatty acid synthase; PPAR-a, peroxisome proliferator-activated receptor alpha.
investigation of NAFLD pathogenesis. Therefore, little data available from animal models showing that changes in gut bacterial strains contribute to NAFLD. More importantly, deeper insight into how the gut microbiota influences NAFLD at the strain level is still lacking.
Different strains of different bacterial species have different effects and clinical manifestations in human hosts with NAFLD. Klebsiella pneumoniae, a high-alcohol-producing strain, causes autobrewery syndrome (ABS) and hepatic mitochondrial dysfunction (Yuan et al., 2019). Enterobacter cloacae B29, an opportunistic pathogen, causes obesity in mice via insulin resistance (Fei and Zhao, 2013;Yan et al., 2016). Even different strains of the same bacterial species have different effects in the human host. As shown in the present study, liver injury was not induced by a nonpathogenic control bacterial strain (E. coli MG1655). Indeed, the mechanisms in NAFLD development, which are launched by specific strains of gut microbiota, need to be further explored.
In this study, we also found that the Escherichia abundance increased in NAFLD patients, not only in feces but also in the gut mucosa. E. coli NF73-1, the most highly enriched strain of Escherichia in NAFLD and NASH patients, was successfully isolated from the intestinal mucosa of NASH patients, and its effect on NAFLD progression was confirmed in animal experiments. Several liver diseases have been shown to be influenced by PBT from the gut (Fouts et al., 2012;Manfredo Vieira et al., 2018), such as autoimmune liver disease (Manfredo Vieira et al., 2018), cholestatic liver injury Liao et al., 2019), liver cirrhosis  and NAFLD . We revealed that E. coli NF73-1 translocated into the liver of bacteria-depleted NAFLD mice, which also supported the conclusion regarding PBT. Interestingly, E. coli NF73-1 translocated only into the liver of NAFLD mice. We inferred that this was based on damage to the intestinal mechanical barrier of NAFLD mice (Peterson and Artis, 2014), as well as the immune barrier (Bibbo et al., 2018), intestinal vascular barrier (Mouries et al., 2019) and even liver barrier (Balmer et al., 2014).
We further confirmed that the liver immune response, which was activated by E. coli NF73-1 exposure, was related to the progression from NAFLD to NASH. The data showed that liver macrophages played a critical role in the pathogenic mechanisms of NASH induced by E. coli NF73-1. Liver macrophages exhibit heterogeneity and plasticity in the tissue microenvironment, especially under pathological conditions such as steatohepatitis (Sica et al., 2014). Macrophage infiltration is an early event in the development of NASH and is associated with progressive disease (Gadd et al., 2014). Microbial detection requires the recognition of pathogen-associated molecular patterns (PAMPs) by pattern recognition receptors (PRRs) in innate immune cells (Himes and Smith, 2010;Csak et al., 2011;Jenne and Kubes, 2013;Miura et al., 2013;Mridha et al., 2017). Previous studies have found that macrophages activate local inflammatory responses and cytokine production, and the coactivation of TLR2 and free fatty acids leads to the progression of NASH (Spruss et al., 2009;Miura et al., 2013;Roh and Seki, 2013). Additionally, the NLRP3 inflammasome has been reported to be involved in bacterial recognition as well as macrophage activation. Complex lipids of bacterial origin trigger inflammation through the binding of NLRP3 to induce Caspase-1 cleavage in macrophages (Vandanmagsar et al., 2011). Taken together, the results indicated that E. coli NF73-1 induced the activation of the TLR2-NF-KB/NLRP3-Caspase-1 signaling pathway to mediate the polarization of macrophages.
The different polarization states of macrophages are correlated with their functional diversity (Chawla et al., 2011). Macrophages are generally classified as M1 and M2 phenotypes. Evidence from both human and animal studies suggests that during the progression of NASH, excess metabolites and bacterial products induce macrophage M1 polarization (Itoh et al., 2013;Lotowska et al., 2013;Gadd et al., 2014). In our study, we demonstrated that the M1 polarization induced by E. coli NF73-1 was responsible for the progression of NAFLD. The above findings from animal experiments were further confirmed by in vitro assays. Hence, our data suggested that E. coli NF73-1 was involved in triggering of inflammation in macrophages. Abnormal metabolism of triglycerides is also an important factor in the pathogenesis of NAFLD (Kawano and Cohen, 2013). Previous studies have reported that the mTOR signaling pathway regulates lipid metabolism (Ai et al., 2012). With obesity, overnutrition and alcoholic liver disease (ALD), mTOR is hyperactivated, resulting in persistent activation of SREBP-1 in the liver (Ai et al., 2012). Based on our data, triglyceride metabolism disruption, which was induced by E. coli NF73-1, tended to be an important factor in NAFLD development. We discovered that the mTOR pathway was triggered by E. coli NF73-1 in mice with NAFLD, which is characterized by an increase in mTOR-mediated phosphorylation and activity of S6K1 and S6. We speculate that activation of the mTOR signaling pathway by E. coli NF73-1 might be the mechanism of NAFLD aggravation. Collectively, our data suggested that E. coli NF73-1 switched macrophage polarization to the M1 phenotype, inducing hepatocyte lipolysis.
Numerous studies have shown that gut dysbiosis is a pathogenetic factor that plays a significant role in the development and progression of liver diseases. Different pathogenic bacteria have been found to exert distinct effects in different liver diseases. It is necessary to identify pathogenic bacteria and functional bacteria, which might be a novel target for the prevention and diagnosis of liver diseases. However, most previous studies focused on the disturbance of the gut flora at only the genus and species levels; they did not illustrate how these pathogenic bacteria aggravate liver diseases at the strain level. Thus, more breakthroughs at the strain level are needed to obtain deeper insight into the prevention of liver diseases associated with the gut flora. In summary, the exploration of bacterial strains is a key point for investigating the mechanisms of liver diseases.
In this study, the pathogenic strain was successfully isolated from clinical intestinal mucosa samples. We demonstrated that translocation of intestinal E. coli NF73-1 into the liver was responsible for the high hepatic M1 population, which further aggravated liver injury, leading to disturbance of the hepatic triglyceride metabolism and, eventually, NAFLD progression. In conclusion, E. coli NF73-1 might be a critical trigger in the progression of NAFLD to NASH, and future clinical trials are needed to clarify the role of E. coli NF73-1 in a larger NAFLD population.

DATA AVAILABILITY STATEMENT
The 16S rRNA sequence of different E. coli NF strains have been uploaded to the NCBI GenBank database (MT649758-MT649857).

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Ethics Committee of Peking University People's Hospital. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by the Ethics Committee of Peking University People's Hospital.  The human liver normal cell line L02 and mouse liver normal cell line H2.35 were exposed to E. coli NF73-1 (multiplicity of infection = 1:10) for 4 h. (B) Lipid accumulation determined by Oil Red O staining of palmitic acid-and oleic acid-treated hepatocytes cocultured with or without E. coli NF73-1. (C) Protein synthesis of phosphorylated mTOR, S6K1, S6, SREBP-1, ACC1, FASN, and PPAR-a in palmitic acid-and oleic acid-treated hepatocytes cocultured with or without E. coli NF73-1. mTOR, mammalian target of rapamycin; S6K1, S6 kinase 1; SREBP-1, sterol-regulatory element binding proteins-1; ACC1, acetyl-coenzyme A carboxylase; FASN, fatty acid synthase; PPAR-a, peroxisome proliferator-activated receptor alpha; N, cleaved nuclear (68 kDa) form of SREBP-1; P, precursor (125 kDa) form of SREBP-1.