Bacteriophage targeting microbiota alleviates non-alcoholic fatty liver disease induced by high alcohol-producing Klebsiella pneumoniae – Nature Communications

Genomic and biological characteristics of a lytic phage against HiAlc Kpn

HiAlc Kpn was defined as the causative agent of NASH accompanied with recurrent pancreatitis (RP) of a male (50–60 years old, with no alcoholism) at Beijing Chaoyang Hospital, Capital Medical University. Interestingly, metagenomic analysis showed that the abundance of K. pneumoniae in his intestinal microbiota was approximately inversely proportional to the abundance of Klebsiella phages (Fig. S1a, b). Furthermore, the quantity of K. pneumoniae in the patient’s microbiome was significantly increased at the inflammatory phase of the patient with NASH in the absence of antibiotic treatment, while the quantity of Klebsiella phage remained relatively low. In contrast, the opposite trend was observed in recovery periods of the patient.

To explore whether the presence of HiAlc Kpn and phages was associated with the clinical characteristic, 16 subjects from our previous cohort, who were with NASH but without alcoholism, were further investigated and followed-up. HiAlc and MedAlc (medium alcohol-producing, which produces 20-30 mmol/L alcohol) Kpn strains were isolated from fecal samples of these patients before weight loss using metagenomic analysis (Supplementary Data 1). The data revealed that the quantity of K. pneumoniae strains was significantly reduced in these NASH followed-up patients compared to the same patients before weight loss, while the quantity of Klebsiella phages was increased in gut microbiota of a total of 6 patients (Fig. S1c). Taken together, these results suggest that there should be an equilibrium in the abundances between K. pneumoniae and Klebsiella phages in vivo. These data also raised a possibility that the phage might alleviate inflammation via its self-regulation, which might be a potential natural medication for treatment of endo-AFLD.

To investigate whether Klebsiella phage specifically targets HiAlc Kpn but not affects other microbiota in vivo, we screened Klebsiella phages by using 86 strains of HiAlc Kpn isolated from clinical NAFLD patients as host bacteria. Results showed that a lytic phage phiW14 (CGMCC No. 23085, GenBank No. OK655936, which specifically targets HiAlc Kpn HK1) had the strongest cleavage ability and the most extensive host spectrum. Thus, the phage phiW14 and HiAlc Kpn HK1 were chosen for further experiments.

HiAlc Kpn HK1 (ST1536, which produces 49.7 mmol/L alcohol) was initially isolated from a NASH/RP patient, having similar biological characteristics to HiAlc Kpn TH1 (ST1536, which produces 60.8 mmol/L alcohol, GCA_001676825.1), presenting multidrug resistant but imipenem susceptible, and causing NAFLD through high-producing alcohol⁷. Furthermore, the phage phiW14 formed clear and large plaques on a lawn of strain HK1 (Fig. 1a). Transmission electron microscopy (TEM) indicated that phiW14 possesses a 50-nm diameter icosahedral head and a 20-nm non-contractile tail and belongs to the Podoviridae family viruses (Fig. 1a). The one-step growth curve showed that the latent period of phiW14 was 10 min, the burst time was 80 min, and the average burst size was 129 plaque-forming units (PFUs)/ cell (Fig. 1b). The optimal multiplicity of infection (MOI) was 0.001 of phiW14 vs HK1. At MOI 10 to 0.0001, phiW14 completely inhibited HiAlc Kpn HK1 (Fig. 1c). Phage phiW14 was able to infect 18 of 86 HiAlc Kpn strains at MOI 0.001, which belong to ST25, ST65, ST86, ST193, ST375, ST1536, and three new STs (Supplementary Data 2). Also, phiW14 was most stable at pH 5 to 10 (Fig. 1d) and at 4 °C to 40 °C (Fig. 1e), which could be tolerant of the environmental stresses. Whole genome sequence showed that phiW14 is a circular double stranded DNA of 50,247 bp with 79 putative open reading frames, but without genes of resistance-drug or virulence (Fig. 1f and Supplementary Data 3). In addition, a total of 52 proteins of phiW14 were identified (Supplementary Data 4), and 26 proteins were classified into different biological functions by Cluster of Orthologous Groups of Proteins (COG) analysis, including biosynthesis, transcription, mobilome, transport and metabolism-related proteins (Fig. 1g). Phylogenetic clustering of large terminase proteins clearly distinguished phiW14 from the other phage families (Fig. 1h). Taken together, these results suggest that phiW14 might be a potential candidate for treatment of endo-AFLD.

Specially eradicating HiAlc Kpn in fecal microbiota prevented steatohepatitis in recipient mice

To clarify the effect of phage treatment on endo-AFLD, mice were gavaged with the HiAlc Kpn HK1 pretreated with either phage phiW14 or phikp15 (HK1/phiW14-fed mice or HK1/phikp15-fed mice) (Fig. 2a), in which phage phikp15 was used as negative control because it cannot lyse HK1 in vitro. In addition, using HiAlc Kpn strains HK2 (ST447, which produces 42.5 mmol/L alcohol) and HK3 (ST101, which produces 41.5 mmol/L alcohol) (both was isolated from a patient with NASH) as bacterial host, phages phikp16 (targeting HK2) and phikp17 (targeting HK3) were obtained from the patient’s feces. As for FMT mice, fecal samples from the patient mentioned above were also proceed with the same pretreatment subsequently conducted to be transplanted in FMT/Phage mice (Fig. 2b).

a Constructions of experimental mice fed with Pair, EtOH, HK1, HK1/phiW14 and HK1/phikp15. b Constructions of FMT and FMT/phage murine models. c Compositions of phylum-based intestinal microbiota in feces of experimental mice. d Abundance of Klebsiella in feces of experimental mice (n = 6, P < 0.0001). e Hematoxylin–eosin (HE) and Oil Red O (ORO) stainings of sections of liver tissues of experimental mice. f Concentrations of alanine transaminase (ALT), aspartate transaminase (AST), triglyceride (TG), diamine oxidase (DAO), and D-lactate (D-LA) in serum of experimental mice (n = 6, except the P value of DAO levels between the group HK1-fed and HK1/ phiW14-fed is 0.001, other P values < 0.0001). Values are expressed as the mean ± SD (n = 6 mice/group). One-way ANOVA, P value < 0.01 (**) or 0.001 (***). Source data are provided as a Source data file.

Compared with mice fed with chow diet, HK1 or EtOH-fed, FMT mice treated with fecal microbiota carrying HiAlc Kpn HK2 and HK3 had steatohepatitis and obviously pathological changes in the livers. The contents of K. pneumoniae in feces (Fig. 2c, d), and as serological indicators, including levels of alanine transaminase (ALT), aspartate transaminase (AST), triglyceride (TG), diamine oxidase (DAO) and D-lactate (D-LA) levels (Fig. 2f), were significantly elevated in mice fed with HiAlc Kpn, HK1/phikp15, FMT, and EtOH, accompanying hepatic steatosis (Fig. 2e). However, there were no abnormalities observed in the mice either fed with HK1/phiW14 or fed with FMT/phage, compared with pair-fed mice. In addition, the overall composition of the intestinal microbiome at the phylum level was not affected by the phage treatment (Fig. 2c). These suggest that treatment with phage could prevent steatohepatitis possibly through specially eradicating HiAlc Kpn in fecal microbiota of recipient mice.

Bacteriophage therapy reprogrammed the gut microbiota without obvious side effects

To explore whether there are some unexpected side effects, mice with steatohepatitis induced by HiAlc Kpn HK1 were treated with phage phiW14 at the range of 10⁴, 10⁵, and 10⁶ PFUs for 1, 4 and 7 days. The results showed that abundance of Klebsiella was effectively inhibited by all three concentrations of phage at all time points in vivo (Fig. 3a, b). However, in the pathological point of view, significantly alleviated hepatic steatosis was only observed in the mice treated with phage continuously for one week (Fig. 3c). In addition, serological indicators, including levels of ALT, AST, TG, DAO and D-LA, were also significantly decreased in mice of the groups with phage therapy for one week (Fig. 3d). It was noted that there was no significant effect in the mice with phage treatment for1 or 4 days in vivo.

Mice with steatohepatitis (gavaged with HiAlc Kpn for 4 weeks) were treated with phage phiW14 at 10⁴ (low concentration, LC), 10⁵ (median concentration, MC), and 10⁶ (high concentration, HC) PFUs for 1, 4 and 7 days. a Compositions of phylum-based intestinal microbiota in feces of experimental mice. b Klebsiella concentrations in feces of experimental mice (n = 6, P < 0.0001). c HE and ORO stainings of sections of liver tissues of experimental mice. d Concentrations of ALT, AST, TG, DAO, and D-LA in serum of experimental mice (n = 6, except the P value of D-LA levels between the group HK1-fed and LC-7d-fed is 0.003, and the P value of D-LA levels between the group HK1-fed and MC-7d-fed is 0.001, other P values < 0.0001). e HE staining of sections of liver, kidney and intestine tissues of mice treated with 10⁴ PFUs of phiW14. f Concentrations of ALT, AST, total bilirubin (TBil), direct bilirubin (DBil), indirect bilirubin (IBil), total bile acid (TBA), TG and total cholesterol (TC) in serum of mice treated with 10⁴ PFUs of phiW14 (n = 6). Values are expressed as the mean ± SD (n = 6 mice/group). One-way ANOVA, P value < 0.01 (**) or 0.001 (***). Source data are provided as a Source data file.

To further evaluate the side effects of phage therapy in vivo, hepatic and renal function-related indicators and histopathological changes of liver, kidney, and intestine tissues were also examined. The data indicated that there was no obvious pathological injury such as cell degeneration in the phage-treated mice (Fig. 3e). Compared with pair-fed mice, there were no significant changes either in levels of ALT, AST, total cholesterol (TC), TG, total bilirubin (TBil), direct bilirubin (DBil), indirect bilirubin (IBil) or in level of total bile acid (TBA) of the experimental mice treated with phage. These suggest that treatment with phage phiW14 could reverse fatty liver by scavenging HiAlc Kpn (Fig. 3f). Thus, all mice with HiAlc Kpn-induced steatohepatitis were treated with 10⁴ PFU of phage for one week in subsequent experiments.

Bacteriophage targeting of HiAlc Kpn alleviated steatohepatitis in vivo

To determine the mechanism by which phage alleviates endo-ALFD, mice with steatohepatitis were constructed through gavaging HiAlc Kpn for 4 or 8 weeks. The experimental mice were then fed with phage phiW14 or imipenem for further one week (Fig. 4a). As expected, hepatic steatosis, injury, inflammation and the blood alcohol concentrations (BAC) were relieved in the 4- and 8-week models of mice treated either by the phage or the antibiotic (Fig. 4b–d). Phage and antibiotic therapies also reduced indexes of injured liver function, including levels of ALT, AST, DAO, D-LA and TG, although some indicators such as ALT showed no significantly difference between the mice with steatohepatitis and the mice treated with phage- or antibiotic (Fig. 4c). The expression of 22 cytokines such as IL-1β, IL-6, IL-10 and TNF-α (Fig. 4d), 5 lipogenic genes including Srebp-1c and Scd-1, and 5 lipolytic genes (such as Pparα and Cyp4a14) (Fig. 4e) were also regulated after treatment with phage or antibiotic. These data suggest that phage therapy might be effective for the treatment of HiAlc Kpn-caused NAFLD. In addition, both HiAlc Kpn and alcohol feeding resulted in weight loss in mouse models (P < 0.001). Interestingly, although there was no statistically significant difference in body weight between groups of HK1- and HK1/phiW14-fed mice, the average weight of group of mice fed with HK1/phiW14 (25.83 g in 4 weeks, 28.14 g in 8 weeks) was also higher than that of group of mice fed with HK1 (25.40 g in 4 weeks, 27.02 g in 8 weeks). These suggest that phage treatment might improve mice health to a certain extent.

a Constructions of Pair-, EtOH-, HK1-, HK1/phiW14- and HK1/IMP-fed murine models. Mice were gavaged with strain HK1, EtOH or YPD broth for 4 weeks or 8 weeks, and then treated with phage phiW14, antibiotic imipenem or PBS for a further week. After the construction of murine models, mice were euthanized for subsequent analysis. b HE and ORO stainings of sections of liver tissues of experimental mice. c Concentrations of blood alcohol, ALT, AST, TG, DAO, and D-LA in serum of experimental mice (n = 6, P < 0.0001). d Concentrations of cytokines in liver tissues of experimental mice. e Relative expression levels of lipogenic genes in liver tissues of experimental mice (n = 6, P < 0.0001). f Body weight of experimental mice (n = 6, P < 0.0001). g Compositions of phylum-based intestinal microbiota in feces of experimental mice. h Abundance of Klebsiella in feces of experimental mice (n = 6, P < 0.0001). Values are expressed as the mean ± SD (n = 6 mice/group). One-way ANOVA, P value < 0.001 (***). Source data are provided as a Source data file.

Furthermore, although the abundance of Klebsiella was decreased significantly in the mice treated by both the phage and the antibiotic, the microbial community structures were different between the two treatments (Fig. 4g, h). Obvious changes in the overall composition of the fecal microbiome were observed in mice treated with imipenem. In 4-week model of mice with HiAlc Kpn-induced steatohepatitis, treatment with imipenem reduced the abundance of Bacteroidetes and Proteobacteria, but increased the abundance of Tenericutes. In 8-week models, the absolute abundance of Firmicutes and Proteobacteria was decreased in the mice treated with imipenem (though the relative abundance was increased). Thus, long term treatment with imipenem caused changes in intestinal flora, and even resulted in dysbacteriosis although such treatment also cleared HiAlc Kpn and alleviated steatohepatitis.

Transcriptional and metabolic reprogramming induced by phage therapy in vivo

We further analyzed transcriptome and metabolome profiles in the livers of all groups. The results showed that mice with HiAlc Kpn-induced steatohepatitis had significantly changes after phage or antibiotic therapy. After phage treatment, muscle synthesis-associated pathways of mice with steatohepatitis were upregulated, while pathways of inflammation or apoptosis, such as the Wnt signaling pathway and extracellular signal-regulated protein kinase 1 cascades were downregulated (Fig 5a, b and Supplementary Data 5). In contrast, the treatment with imipenem upregulated biological process-associated pathways, but downregulated inflammation and apoptosis pathways, including leukocyte migration and mitogen-activated protein kinase (MAPK) cascades (Fig 5c, d and Supplementary Data 5).

a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in upregulated DEGs; b KEGG pathways enriched in downregulated DEGs comparing groups of mice fed with HK1 and HK1/phiW14. c KEGG pathways enriched in up-regulated DEGs; d KEGG pathways enriched in downregulated DEGs comparing groups of mice fed with HK1 and HK1/IMP.

In addition, amounts of the metabolites citraconic acid, α-tocopherol, and sebacic acid, which were enriched into glycerolipid metabolism pathways, as well as glycerol 3-phosphate of the glycerophospholipid metabolism pathways were increased, while L-lactic acid in the glycolysis and gluconeogenesis pathways were decreased in mice treated with HK1/phiW14-fed, as compared with mice with HiAlc Kpn-induced steatohepatitis (Fig. 6a–e and Supplementary Data 6). Furthermore, metabolites including 2,2,2-trichloroethanol, digalacturonic acid, and UDP-glucuronic acid, which were enriched in pentose and glucuronate interconversions, ascorbate and aldarate metabolism, and metabolism of xenobiotics by cytochrome p450 pathways were upregulated in the of mice fed with HK1/IMP (Fig. 6f–j and Supplementary Data 6). These data suggest that phage therapy has efficacy on mice with HiAlc Kpn-induced steatohepatitis, possibly through not only targeting HiAlc Kpn, but also regulating inflammation, lipid metabolism and carbohydrates metabolism.

a Orthogonal partial least-squares-discriminant analysis (OPLS-DA); b Volcano plot of differential metabolites; c KEGG pathways enriched in upregulated metabolites; d KEGG pathways enriched in downregulated metabolites; e differential metabolites between groups of mice fed with HK1 and HK1/phiW14. f OPLS-DA; g volcano plot of differential metabolites; h KEGG pathways enriched in upregulated metabolites; i KEGG pathways enriched in downregulated metabolites; j differential metabolites between groups of mice fed with HK1 and HK1/IMP. Values are expressed as the mean ± SD (n = 6 mice/group). Differential metabolites with P value < 0.05 and VIP value > 1 were chosen for Metabo Analyst-based KEGG pathway enrichment analysis.

Bacteriophage therapy efficiently relieved HiAlc Kpn-induced steatohepatitis in germfree mice

The effect of phage therapy was further verified in germfree mice (Fig. 7a) gavaged by intestinal flora as mentioned in Fig. 2b. Similar with the effect of phage in SPF mice, the intestinal flora from the patient with NASH caused steatohepatitis in germfree mice, and K. pneumoniae colonization in gut was observed by scanning electron microscope (SEM). After phage treatment, the related indicators of steatohepatitis such as 16S rRNA sequencing (Fig. 7b, c), pathological change (Fig. 7d) and serological indicators (Fig. 7e) were alleviated, while the K. pneumoniae strains were eliminated in the gut of germfree mice.

a Constructions of FMT-, FMT/phage-, FMT/IMP-, EtOH- and Pair-fed murine models. Mice were treated with FMT, EtOH or YPD broth for 4 weeks or 8 weeks, and then gavaged with phage, antibiotic imipenem or PBS for a further week. After the construction of murine models, mice were euthanized for subsequent analysis. b Compositions of phylum-based intestinal microbiota in feces of experimental mice. c Abundance of Klebsiella in feces of experimental mice (n = 6, P < 0.0001). d HE and ORO stainings of sections of liver tissues of experimental mice, and scanning electron microscope (SEM) micrographs of the proximal colon showing the colonization status of HiAlc Kpn in vivo in germfree mice gavaged for 8 weeks. e Concentrations of ALT, AST, TG, DAO, and D-LA in serum of experimental mice (n = 6, except the P value of TG levels between the group FMT-fed and FMT/phage-fed in 4-week model is 0.006, the P value of TG levels between the group FMT-fed and FMT/IMP-fed in 4-week model is 0.021, the P value of DAO levels between the group FMT-fed and FMT/phage-fed in 8-week model is 0.001, the P value of D-LA levels between the group FMT-fed and FMT/phage-fed in 8-week model is 0.001, and the P value of D-LA levels between the group FMT-fed and FMT/IMP-fed in 8-week model is 0.004, other P values < 0.0001). Values are expressed as the mean ± SD (n = 6 mice/group). One-way ANOVA, P value < 0.05 (*), 0.01 (**), or 0.001 (***). Source data are provided as a Source data file.

The efficacy of bacteriophage therapy in murine model of NASH

To explore the efficacy of phage in general NAFLD model, murine model of NASH induced by MCD (methionine-choline deficient) diet was also included in the present study. As shown in Fig. 8a, mice were randomized into 5 groups, including MCD-fed, MCD/HK1-fed (MCD-fed and HK1 gavaged), MCD/HK1/phiW14-fed (MCD-fed, HK1 and phiW14 gavaged), MCD/phiW14-fed (MCD-fed and phiW14 gavaged) and MCS (fed with methionine-choline-supplemented, as negative control). From the results of 16S rRNA sequencing (Fig. 8b, c), pathological changes (Fig. 8d) and serological indicators (Fig. 8e), phage treatment alleviated steatohepatitis in mice with MCD diet plus HiAlc Kpn feeding, which might reflect the important role of HiAlc Kpn in the pathogenic process of NASH.

a Constructions of MCD/HK1-, MCD/HK1/phiW14-, MCD/phiW14, MCD- and Pair-fed murine models. Mice were fed with MCD or MCS diet, gavaged with HK1 or PBS for 4 weeks or 8 weeks, and then gavaged with phage or PBS for a further week. After the construction of murine models, mice were euthanized for subsequent analysis. b Compositions of phylum-based intestinal microbiota in feces of experimental mice. c Abundance of Klebsiella in feces of experimental mice (n = 6, P < 0.0001). d HE, ORO and Masson stainings of sections of liver tissues of experimental mice. e NAS (Non-alcoholic fatty liver disease activity scores) in liver tissues, and concentrations of ALT, AST, TG, DAO, and D-LA in serum of experimental mice (n = 6, except the P value of TG levels between the group MCD/HK1-fed and MCD/HK1/phiW14-fed in 4-week model is 0.004, and the P value of D-LA levels between the group MCD/HK1-fed and MCD/HK1/phiW14-fed in 8-week model is 0.001, other P values < 0.0001). The murine model was conducted with 3 independent experiments. Values are expressed as the mean ± SD (n = 6 mice/group). One-way ANOVA, P value < 0.05 (*), 0.01 (**), or 0.001 (***). Source data are provided as a Source data file.

In addition, we observed that the MCD/HK1 group had more severe pathological changes in hepatic steatosis and inflammation compared to the MCD group, and there was a noticeable occurrence of cellular phagocytosis in the MCD/HK1 group of 8-week model (Fig. 8d). The NAFLD activity scores (NAS) also showed that the MCD/HK1 group had a significantly higher score (Fig. 8e) than that of the MCD group in 8-weeks model (P < 0.05). However, there was no statistically significant difference in serological indicators between two groups. This is possibly due to that though the hepatic steatosis in both two groups has pathologically occurred, little change in the serum indicators can be observed until the disease progressed to a more severe stage. This, of course, needs to be verified in further research.

Phage treatment did not alleviated steatohepatitis in mice fed with MCD diet alone, this is probably pointing that there are diversities in the pathogenic mechanisms of NAFLD. On the other hand, these data also suggest that phage therapy might be highly specific, at least in our animal models of NASH.