Gut-liver Translocation of Pathogen Klebsiella Pne

nature microbiology
Article https://doi.org/10.1038/s41564-024-01890-9
Gut–liver translocation of pathogen

Klebsiella pneumoniae promotes
hepatocellular carcinoma in mice
Received: 29 May 2023 Xueliang Wang1,2,3,8, Yi Fang3,8, Wei Liang3,8, Yuhong Cai1,8, Chi Chun Wong2,
Junlin Wang3, Na Wang3, Harry Cheuk-Hay Lau 2, Ying Jiao2, Xingyu Zhou2,
Accepted: 15 November 2024
Liufang Ye2, Mengmiao Mo3, Tao Yang3, Miao Fan3, Lei Song3, Heming Zhou 2,
Published online: xx xx xxxx Qiang Zhao4, Eagle Siu-Hong Chu2, Meinong Liang3, Weixin Liu 2, Xin Liu 1,
Shuaiyin Zhang3, Haitao Shang3, Hong Wei3, Xiaoxing Li3, Lixia Xu5, Bing Liao6,
Check for updates
Joseph J. Y. Sung 2,7, Ming Kuang 1 & Jun Yu 2,3
Hepatocellular carcinoma (HCC) is accompanied by an altered gut

microbiota but whether the latter contributes to carcinogenesis is unclear.
Here we show that faecal microbiota transplantation (FMT) using stool
samples from patients with HCC spontaneously initiate liver inflammation,
fibrosis and dysplasia in wild-type mice, and accelerate disease progression
in a mouse model of HCC. We find that HCC-FMT results in gut barrier injury
and translocation of live bacteria to the liver. Metagenomic analyses and
bacterial culture of liver tissues reveal enrichment of the gut pathogen
Klebsiella pneumoniae in patients with HCC and mice transplanted with the
HCC microbiota. Moreover, K. pneumoniae monocolonization recapitulates
the effect of HCC-FMT in promoting liver inflammation and hepatocarci
nogenesis. Mechanistically, K. pneumoniae surface protein PBP1B interacts
with and activates TLR4 on HCC cells, leading to increased cell proliferation
and activation of oncogenic signalling. Targeting gut colonization using
K. oxytoca or TLR4 inhibition represses K. pneumoniae-induced HCC
progression. These findings indicate a role for an altered gut microbiota in
hepatocarcinogenesis.
Hepatocellular carcinoma (HCC) is a leading cause of cancer-associated transfer of gut microbes to the liver in disease states11. Intriguingly,
deaths1. HCC proceeds via a multistep cascade involving liver inflam- germ-free mice displayed delayed HCC progression due to reduced
mation, fibrosis, cirrhosis and HCC2. Beyond host factors, gut dysbiosis activation of Toll-like receptor 4 (TLR4)12, a sensor of bacterial infec-
plays important roles in HCC pathogenesis3. Recent studies have also tion. This highlights the possibility that gut dysbiosis may contribute
unravelled the existence of an intratumoural microbiome4 that pro- to hepatocarcinogenesis.
motes tumorigenesis and modifies the immune microenvironment5–8. Here we sought to establish the causality of HCC-associated gut
However, the function of gut and intrahepatic microbes in HCC patho- dysbiosis in promoting HCC. We performed faecal microbiota trans-
genesis remains largely unclear. plantation (FMT) of microbial consortia from patients with HCC and
The gut microbiome regulates liver homoeostasis via the gut– healthy donors to mice, revealing that HCC-FMT promotes inflam-
liver axis9,10. The gut barrier is the first line of defence against deleteri- mation, fibrosis, dysplasia and HCC. HCC-associated gut microbiota
ous gut pathogens and their products, and a leaky gut might enable disrupts gut barrier function and promotes translocation of live
A full list of affiliations appears at the end of the paper. e-mail: [email protected]
Nature Microbiology
Content courtesy of Springer Nature, terms of use apply. Rights reserved
bacteria to the liver, especially Klebsiella pneumoniae (K. pneumoniae). 16 mice) that were absent from HD-FMT and PBS groups (P < 0.0001)
K. pneumoniae monocolonization studies in mice validated its causa- (Extended Data Fig. 2b). Haematoxylin and eosin (H&E) staining con-
tive role in HCC. Mechanistically, HCC-FMT and K. pneumoniae activate firmed diagnosis of dysplasia in the HCC-FMT group (Extended Data
hepatic inflammatory and oncogenic signalling. Corroborating these Fig. 2b). Compared with HD-FMT, HCC-FMT increased hepatocyte
results, live K. pneumoniae is enriched in liver tissues of individuals proliferation (PCNA and Ki-67) (Extended Data Fig. 2c), hepatic inflam-
with HCC. Altogether, we revealed a causative role of gut dysbiosis in mation (TH1, TH17 and TH2) (Extended Data Fig. 2d) and liver fibrosis
hepatocarcinogenesis. markers (desmin, α-SMA and Sirius red) (Extended Data Fig. 2e).
Cancer pathway PCR array revealed that HCC-FMT upregulated
Results genes involved in cell proliferation (Cdc20 and Dkc1), angiogenesis
HCC stool accelerates DEN-induced liver tumorigenesis (Angpt2), epithelial–mesenchymal transition (Foxc2 and Snai2) and
We initially investigated the effect of polymicrobial consortia trans- anti-apoptosis (Nol3) (Extended Data Fig. 2f). Inflammatory response
plantation from stool of patients with HCC or liver cirrhosis (LC) and autoimmunity PCR array also showed that pro-inflammatory
and healthy donors on liver pathologies in germ-free and specific response (Tlr4, Myd88, Nfkb1 and Il1a) and lymphocyte chemotaxis
pathogen-free (SPF) mice. 16S ribosomal (r)RNA sequencing revealed (Ccl12, Ccl11 and Ccl7) were increased by HCC-FMT (Extended Data
a progressive decrease in microbial diversity in stool samples from Fig. 2g). Concordant results were obtained by FMT of HCC stool to SPF
healthy donors, LC to HCC, as determined by Shannon, Chao1 and mice (Extended Data Fig. 3a–f) and of independent stool consortia
Ace indices (Supplementary Fig. 1a). Klebsiella was highly enriched from individuals with HCC (N = 5) and healthy donors (N = 5) to germ-
in stool samples from individuals with HCC (Supplementary Fig. 1a). free mice (Extended Data Fig. 4a–f). Collectively, HCC-FMT spontane-
We observed accurate recapitulation of gut microbiome differences ously induces liver inflammation, fibrosis and dysplasia in mice.
of donor consortia in germ-free mice stool after FMT treatment (Sup-
plementary Fig. 1b,c), as evidenced by reduced microbial diversity HCC stool promotes gut leakage and pathogen translocation
and enrichment of Klebsiella in mice receiving HCC-FMT compared Gut dysbiosis modulates liver pathologies via the gut–liver axis13. We
with other groups. examined the effect of FMT treatment on gut barrier integrity and
Faecal microbiota consortia were transplanted from 15 demonstrated that HCC-FMT increased gut permeability to 500 kDa
HCC (GFD/HCC-FMT), 17 LC (GFD/LC-FMT) and 16 healthy donors FITC-dextran (without DEN: P < 0.0001; with DEN: P = 0.0024) com-
(GFD/HD-FMT) to diethylnitrosamine (DEN)-injected germ-free (GFD) pared with HD-FMT and PBS in germ-free mice (Extended Data
mice once per week (Fig. 1a), with PBS as blank control. Compared Fig. 5a,b). Transmission electron microscopy (TEM) confirmed abnor-
with the other groups, HCC-FMT increased maximum tumour diam- malities of colon intercellular junctions in HCC-FMT, with disappear-
eter (P = 0.007 vs LC-FMT; P < 0.0001 vs HD-FMT), tumour number ance of tight junctions and widening of adherens junctions (Fig. 2a
(P = 0.0001 vs HD-FMT) and tumour burden (P = 0.0012 vs LC-FMT; and Supplementary Fig. 4a). HCC-FMT reduced the thickness of
P < 0.0001 vs HD-FMT) (Fig. 1b and Supplementary Fig. 2a), indicating colonic mucus and downregulated the expression of tight junction
that HCC-associated microbes promote tumorigenesis. Immunohis- markers (claudin 3 (CLDN3), CLDN1 and E-cadherin (E-cad)), leading
tochemistry (IHC) showed that HCC-FMT induced hepatocyte prolif- to increased bacterial infiltration into mucosal lamina propria and
eration (PCNA, Ki-67) (Fig. 1c and Supplementary Fig. 2b) compared muscularis mucosa (Extended Data Fig. 5a,b, Fig. 2a and Supplemen-
with LC-FMT and HD-FMT. Meanwhile, HCC-FMT and LC-FMT similarly tary Fig. 4a). As a result, HCC-FMT induced inflammation in colon, as
induced liver inflammation, with increased pro-inflammatory T helper evidenced by PCR array and flow cytometry (Extended Data Fig. 5c,d).
1 (TH1) and TH17 cells, together with reduced anti-inflammatory TH2 These suggest that HCC-FMT disrupts the gut barrier and promotes
compared with HD-FMT (Supplementary Fig. 2c). Fibrosis markers bacterial infiltration into colonic mucosa.
desmin, α-smooth muscle actin (α-SMA) and Sirius red staining were We next asked whether disrupted gut barrier function might
also increased by HCC-FMT and LC-FMT (Supplementary Fig. 2d). enable translocation of gut microbes to the liver. Anaerobic and
Cancer pathway PCR array revealed that HCC-FMT upregulated genes aerobic culture of fresh liver homogenates from germ-free mice
involved in cell proliferation (G6pdx, Skp2, Pgf, and Sox10), angio- demonstrated that almost all liver samples from HCC-FMT displayed
genesis and embryonic development (Angpt2 and Gsc), epithelial– positive culture, while those from the HD-FMT group were largely
mesenchymal transition (Foxc2 and Snai1) and anti-apoptosis (Epo) negative (P < 0.0001) (Fig. 2b, Extended Data Fig. 5e and Supplemen-
compared with LC-FMT and HD-FMT (Fig. 1d). Inflammatory response tary Fig. 4b). Cultures of the PBS group and contamination controls
and autoimmunity PCR array showed that pro-inflammatory response all tested negative (Fig. 2b, Extended Data Fig. 5f and Supplementary
(Tlr4, Tnfsf14, Il6ra, Ccl4, andTlr2) and lymphocyte chemotaxis Fig. 4b). Fluorescence in situ hybridization (FISH) (universal 16S
(Ccr2 and Cxcr2) were increased by HCC-FMT compared with LC-FMT rRNA) confirmed that bacteria were readily detected in livers of the
and HD-FMT groups (Fig. 1e). HCC-FMT group but were absent in the HD-FMT (close to none) and
The protumorigenic effect of HCC-FMT compared with LC-FMT PBS groups (Extended Data Fig. 5g). TEM of colon and hepatic tissues
and HD-FMT were recapitulated in DEN-injected SPF mice (Fig. 1f–j confirmed the presence of bacteria in HCC-FMT-treated germ-free
and Supplementary Fig. 3a–d) and validated by another mixed stool mice (Fig. 2c). Corroborating these results, gut barrier disruption
consortium from HCC (N = 5) and healthy donors (N = 5) in DEN-injected and translocation of live bacteria to the liver was validated in SPF
germ-free mice (Extended Data Fig. 1a–d). Furthermore, individual mice with HCC-FMT (Extended Data Fig. 6a–c), germ-free mice trans-
FMT from healthy donors (N = 2) and patients with HCC (N = 2) to planted with another mixed stool consortium from individuals with
DEN-injected germ-free mice again showed that HCC-FMT acceler- HCC (N = 5) and healthy donors (N = 5) (Extended Data Fig. 7a–h),
ated HCC progression (Extended Data Fig. 1e–h). Together, faecal and individual HCC-FMT (N = 2) and HD-FMT (N = 2) (Extended Data
transplantation of HCC microbiota to mice accelerates DEN-induced Fig. 7i–k). Notably, LC-FMT evidently caused gut barrier dysfunction
HCC progression. and bacterial translocation to the liver, consistent with the involve-
ment of gut dysbiosis in liver pathologies.
HCC stool promotes inflammation, fibrosis and dysplasia To decipher the live bacterial composition in the liver, colonies
Next, we performed FMT of polymicrobial consortia from 16 patients were randomly selected from liver tissues for identification (Fig. 2d).
with HCC (GF/HCC-FMT) and 12 healthy donors (GF/HD-FMT) to K. pneumoniae was the top bacterium enriched in the HCC-FMT
germ-free mice without DEN injection (Extended Data Fig. 2a). Upon group compared with HD-FMT, followed by Enterococcus faecalis,
killing, mice receiving HCC-FMT displayed liver nodules (in 9 out of Proteus mirabilis, Enterobacter hormaechei and Bacteroides fragilis
Nature Microbiology
a d 15
Cancer pathway: 15
Cancer pathway:
14-day germ-free male BALB/c GFD/HCC-FMT vs GFD/HD-FMT GFD/HCC-FMT vs GFD/LC-FMT
DEN
Fold-regulation
Fold-regulation
0 6 29 weeks
Individual FMT once a week 10 10
GFD/PBS DEN + PBS gavage (n = 8)
GFD/HD-FMT DEN + healthy donor FMT (n = 16) 5 5
GFD/LC-FMT DEN + liver cirrhosis patient FMT (n = 17)

0 0
GFD/HCC-FMT DEN + HCC patient FMT (n = 15)
Angpt2
G6pdx
Skp2
Pgf
Foxc2
Car9
Ccl2
Cdc20
Epo
Gsc
Krt14
Lig4
Serpinb2
Snai1
Snai3
Tbx2
Bcl2l11
Fasl
Tinf2
Casp7
Tep1
Sox10
Ppp1r15a
Sirt1
Casp9
Casp2
Angpt2
Nol3
Pgf
Foxc2
Apaf1
Wee1
Birc3
G6pdx
Ppp1r15a
Aurka
Skp2
Ocln
Angpt1
Bcl2l11
Cdc20
Epo
Fasl
Gsc
Serpinb2
Snai1
Sox10
Casp7
b GFD/PBS GFD/HD-FMT GFD/LC-FMT GFD/HCC-FMT
e Inflammatory response and autoimmunity: Inflammatory response and autoimmunity:
GFD/HCC-FMT vs GFD/HD-FMT GFD/HCC-FMT vs GFD/LC-FMT
5 6
Fold-regulation
Fold-regulation
0 3
<0.0001 <0.0001
0.0001
10
Tumour number
<0.0001 60
Maximum tumour
8 0.0003
<0.0001
Tumour burden
diameter (mm)
0.0014 0.0007 0.0084 0.0012

0.0085
0.0039 0.0007
0.0033 (mm3) –5 0
Cxcl1
Ripk2
Bcl6
Tnfsf14
Tlr1
Ccr2
Cxcr2
Kng1
Ccr1
Ccl4
Il6ra
Ccl5
Tlr4
Ltb
Itgb2
Il1r1
Cd14
Il1b
Tlr2
Ccl19
Nos2
Ccl11
Ccr2
Ltb
Cxcr2
Tnfsf14
Myd88
Tlr2
Ccl4
Cd40
Tlr4
Tlr7
Il6ra
Il1a
4 5 30
0 0 0
GFD/PBS GFD/HD-FMT GFD/LCC-FMT GFD/HCC-FMT f 14-day SPF male BALB/c DEN

0 6 29 weeks
c <0.0001
<0.0001 GFD/PBS Individual FMT once a week
6 SPFD/PBS
80 <0.0001 <0.0001 GFD/HD-FMT DEN + PBS gavage (n = 20)
0.0017 0.0001
positive (%)
0.0004 0.0055
positive (%)
GFD/LC-FMT DEN + healthy donor FMT (n = 16)

0.0018
SPFD/HD-FMT
0.0006
PCNA
GFD/HCC-FMT
Ki-67
40 3 SPFD/LC-FMT DEN + liver cirrhosis patient FMT (n = 16)

SPFD/HCC-FMT DEN + HCC patient FMT (n = 16)
0 0
g SPFD/PBS SPFD/HD-FMT SPFD/LC-FMT SPFD/HCC-FMT

i 15
Cancer pathway:
SPFD/HCC-FMT vs SPFD/HD-FMT
15
Cancer pathway:
SPFD/HCC-FMT vs SPFD/LC-FMT
Fold-regulation
Fold-regulation
10 10
5 5
<0.0001 0.0026 <0.0001

12 10 0.0011 240
<0.0001 0 0
Maximum tumor
<0.0001
Tumour number
Tumour burden
diameter (mm)
0.0293 0.0479
0.0283
Atrx
Ddb2
G6pdx
Lig4
Ercc3
Kdr
Cdh2
Tnks
Tbx2
Hmox1
Gpd2
Tnks2
Lpl
Ppp1r15a
Cdc20
Map2k3
Adm
Bcl2l11
Acsl4
Gadd45g
Arnt
Casp2
Pfkl
Casp9
Tep1
Tek
Birc3
Map2k1
Dkc1
Flt1
G6pdx
Gusb
Gadd45g
Lig4
Ddb2
Atrx
Hmox1
Lpl
Ccl2
Tinf2
Casp2
Mki67
Kdr
Tnks2
Cdh2
Flt1
Pfkl
Casp9
Stmn1
Ercc3
Ldha
Xiap
Tnks
Cdc20
Tbx2
Ppp1r15a
Cflar
Ccnd3
Map2k3
Tep1
Igfbp7
Acsl4
E2f4
B2m
0.0033 0.0025
0.0335
0.0054
(mm3)
0.0012
6 5 120
j
0.0033
Inflammatory response and autoimmunity: Inflammatory response and autoimmunity:

30
SPFD/HCC-FMT vs SPFD/HD-FMT SPFD/HCC-FMT vs SPFD/LC-FMT
0 0 0 30
Fold-regulation
Fold-regulation
SPFD/PBS SPFD/LC-FMT SPFD/HCC-FMT SPFD/HD-FMT 20

15
h 100
<0.0001 <0.0001
0.0002
10
0.0008 10 <0.0001 SPFD/PBS
<0.0001 0
positive (%)
positive (%)
0.0003 0.0063 SPFD/HD-FMT

0.0013 0.0034
PCNA
Ki-67
50 SPFD/LC-FMT
5 –10 0
SPFD/HCC-FMT
Nos2
Cxcr4
Ccl3
Hsp90ab1
Cxcl10
Ccl5
Myd88
Tlr4
Ccr3
Tlr9
Cd14
C3ar1
Il7
Nr3c1
Kng1
Tirap
Tlr5
Tollip
Cxcl9
Tlr2
Nfkb1
Ccl4
Tlr1
Itgb2
Ccr7
Cebpb
Tlr7
Il1rn
C4b
Crp
Ccl2
Fos
Hsp90ab1
Il7
Nos2
Tlr1
Cxcr4
C3ar1
Ccl3
Cxcl9
Gusb
Tlr3
Ccr7
Kng1
Myd88
Tlr7
Il1b
Sele
Tlr5
Cxcl10
B2m
Cd14
Ccr1
Tlr6
Ccl25
Ccl24
Tollip
Bcl6
Il1rn
0 0
Fig. 1 | Fetal microbiota from patient with HCC promotes HCC in both germ- SPFD/HD-FMT group n = 16; SPFD/LC-FMT group n = 16; SPFD/HCC-FMT
free and SPF mice with DEN treatment. a, Design of FMT experiment for group n = 16) and relevant results (g–j). g, Representative images of liver gross
DEN-treated germ-free mice (GFD/PBS group n = 8; GFD/HD-FMT group n = 16; morphology (yellow circles indicate tumour) with tumour incidence, tumour
GFD/LC-FMT group n = 17; GFD/HCC-FMT group n = 15) and relevant results number and tumour burden. n = 20 (SPFD/PBS), 16 (SPFD/HD-FMT), 16 (SPFD/LC-
(b–e). b, Representative images of liver gross morphology (dashed yellow FMT), 16 (SPFD/HCC-FMT). h, IHC staining for PCNA and Ki-67. n = 6 biologically
circles indicate tumour) with tumour incidence, tumour number and tumour independent samples. i, Mouse cancer pathway finder array of liver tissues. n = 3
burden. n = 8 (GFD/PBS), 16 (GFD/HD-FMT), 17 (GFD/LC-FMT) and 15 (GFD/HCC- biologically independent samples of each group were mixed for assay. j, Mouse
FMT) mice. c, Immunohistochemistry (IHC) staining for PCNA and Ki-67. n = 6 inflammatory response and autoimmunity PCR array of liver tissues. In i and j,
biologically independent samples. d, Mouse cancer pathway finder array of n = 3 independent experiments with similar results. In b,c,g,h, data are presented
liver tissues. e, Mouse inflammatory response and autoimmunity PCR array of as mean ± s.e.m. Each data point in bar plots represents one mouse. Statistical
liver tissues. In d and e, n = 3 independent experiments with similar results. significance was calculated using one-way ANOVA. Adjustments were made for
f, Design of FMT experiment for DEN-treated SPF mice (SPFD/PBS group n = 20; multiple comparisons.
(Fig. 2d). Absolute quantification by qPCR in human donors and Live K. pneumoniae is enriched in liver tissue of patients
recipient mouse samples demonstrated that faecal K. pneumoniae with HCC
was increased in individuals with HCC compared with those with We next sought to validate our discoveries in human patients. Universal
LC and healthy donors, and a consistent pattern was reproduced 16S rRNA FISH of tissue arrays revealed the enrichment of bacteria in
in the stool of FMT recipient mice (Fig. 2e). Correspondingly, liver tumorous compared with the paracancerous region in patients with
K. pneumoniae levels were higher in the HCC-FMT group compared HCC (P < 0.0001) (N = 45) (Supplementary Fig. 5a). Anaerobic and
with LC-FMT and HD-FMT groups (Fig. 2e), which was confirmed aerobic culture of fresh liver homogenates from 54 individuals with
by FISH using a K. pneumoniae-specific probe (Fig. 2f). Together, HCC and 15 with LC demonstrated a living microbiome in HCC and
HCC-associated gut dysbiosis disrupts gut barrier function, thereby LC tissues, which were confirmed by FISH, lipopolysaccharide (LPS)
promoting translocation of bacteria to the liver and enrichment of and lipoteichoic acid (LTA) staining (Fig. 3a,b). Mass spectrometry
intrahepatic K. pneumoniae. identification of randomly picked live bacterial colonies and 16S rRNA
Nature Microbiology
a <0.0001
Germ-free with DEN
<0.0001
number per slide

<0.0001 <0.0001 < 0.0001
< 0.0001
<0.0001 100 0.8 160 GFD/PBS GFD/LC-FMT
E-cad optical
disappear rate
< 0.0001
density (a.u.)
Tight junction
<0.0001
100
positive (%)
< 0.0001
Alcian blue
<0.0001 < 0.0001
< 0.0001
Bacteria
<0.0001 < 0.0001
< 0.0001
50 0.4 80 GFD/HD-FMT GFD/HCC-FMT
(%)
50
0 0 0 0
<0.0001
Germ-free without DEN

<0.0001
number per slide

100 0.0004 100 0.4 <0.0001
120
< 0.0001
GF/PBS
disappear rate
Tight junction
E-cad optical
density (a.u.)
positive (%) <0.0001
Alcian blue
<0.0001 < 0.0001
Bacteria
50
GF/HD-FMT
0.2
(%)
50 60
GF/HCC-FMT
0 0 0 0
b Germ-free with DEN Germ-free without DEN

<0.0001 <0.0001
<0.0001
<0.0001
<0.0001 40 20
<0.0001 <0.0001 <0.0001
c.f.u. mg−1 tissue
c.f.u. mg−1 tissue

<0.0001
<0.0001 <0.0001 100
Liver with living

100
Liver with living
GFD/PBS GFD/LC-FMT GF/PBS
bacteria (%)
bacteria (%)
<0.0001
20 10 GF/HD-FMT
50 50
GFD/HD-FMT GFD/HCC-FMT GF/HCC-FMT
0 0 0 0
Bacteria in hepatic Bacteria in parenchymal Bacteria in parenchymal

c Gut barrier dysfunction sinusoid tissue tissue
GFD/HCC-FMT
GFD/HCC-FMT
GFD/HCC-FMT
GF/HCC-FMT
Tight junction
mice colon
mice liver
mice liver
mice liver
Adherens junction
1 µm 2 µm 2 µm 1 µm
d Germ-free with DEN Germ-free without DEN

Live bacterial abundance
Live bacterial abundance
Bacteroides dorei
100 Klebsiella pneumoniae Ligilactobacillus faecis 100 Klebsiella pneumoniae
at species level (%)
at species level (%)
Ligilactobacillus animalis
Enterococcus faecalis Odoribacter splanchnicus Enterococcus faecalis
Lactobacillus johnsonii Enterococcus hirae
Escherichia coli Bacteroides thetaiotaomicron
Oscillospiraceae bacterium N12 Ligilactobacillus faecis
Proteus mirabilis Escherichia coli
Bacteroides salyersiae
Enterobacter hormaechei Streptococcus parasanguinis
50 Fusobacterium mortiferum 50 Citrobacter amalonaticus
Citrobacter amalonaticus Enterocloster bolteae Citrobacter farmeri
Eubacterium limosum Enterobacter hormaechei
Enterobacter ludwigii Citrobacter freundii
Hungatella effluvii Proteus mirabilis
Bacteroides thetaiotaomicron Phocaeicola vulgatus
Klebsiella oxytoca Enterobacter ludwigii
Bacteroides fragilis Lysinibacillus fusiformis Staphylococcus warneri
0 Clostridium innocuum 0 Enterococcus faecium
Bacteroides intestinalis Limosilactobacillus reuteri
GFD/HD GFD/LC GFD/HCC Bacteroides cellulosilyticus Bacteroides uniformis GF/HD GF/HCC- Enterobacter cloacae
Bacteroides ovatus
-FMT -FMT -FMT -FMT FMT
e
Absolute KP number per
×108
Absolute KP number per
×106 <0.0001
mg mouse liver tissues
<0.0001 4
4
Absolute KP number
<0.0001
per g human faeces
<0.0001 0.0091 4 0.0003 0.0149

<0.0001 0.0103
g mouse faeces
2 2 2
0 0 0
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
HD LC patients HCC patients GFD/HD-FMT GFD/LC-FMT GFD/HCC-FMT
GFD/HD-FMT GFD/LC-FMT GFD/HCC-FMT
f <0.0001
GFD/PBS GFD/HD-FMT GFD/LC-FMT GFD/HCC-FMT
KP number per slide
80 0.0174
<0.0001
0.0473 <0.0001 GFD/PBS GFD/LC-FMT
KP FISH
40 GFD/HD-FMT GFD/HCC-FMT
20 µm
Fig. 2 | Faecal microbiota from patients with HCC impairs gut barrier function bacterial strains as determined by microbial mass spectrometry identification.
and promotes bacteria translocation to the liver in germ-free mice with or e, Absolute number of K. pneumoniae (KP) in human faeces (left), mice faeces
without DEN treatment. a, Quantitative analysis of TEM, Alcian blue staining, (middle) and mice liver tissues (right) as detected by qPCR. n = 10 biologically
E-cad IHC staining and Cy3-conjugated EUB338 probe FISH of colon tissues. n = 6 independent samples. f, Representative images (left) and quantification (right)
biologically independent samples. b, Quantitative analysis of bacterial culture of of Cy3-conjugated K. pneumoniae probe FISH detection in mice liver tissues. n = 6
liver tissues under anaerobic and aerobic conditions with quantitative analysis. biologically independent samples. In a (excluding tight junction disappear rate),
n = 8 (GFD/PBS), 16 (GFD/HD-FMT), 17 (GFD/LC-FMT), 15 (GFD/HCC-FMT), 6 b (excluding liver with living bacteria) and f, data are presented as mean ± s.e.m.
(GF/PBS), 12 (GF/HD-FMT), 16 (GF/HCC-FMT) mice. c, Representative TEM images Each data point in bar plots represents one mouse. Tight junction disappearance
of colon tissues and liver tissues in GFD/HCC-FMT mice and representative TEM rate and liver with living bacteria were analysed using Fisher’s exact test. Unless
image of liver tissues in GF/HCC-FMT mice. n = 3 independent experiments otherwise stated, statistical significance was calculated using one-way ANOVA.
with similar results. d, Stacked bar plot of relative abundance of culturable Adjustments were made for multiple comparisons.
Nature Microbiology
sequencing (Fig. 3c,d) revealed that Klebsiella (K. pneumoniae) domi- asked whether K. pneumoniae might boost macrophage-derived MMP-
nates the liver microbiome of individuals with HCC compared with 2/-9. Supporting our notion, in vitro co-culture of THP-1-induced mac-
LC, which was supported by absolute quantification (P = 0.0002) and rophages with K. pneumoniae-conditioned medium induced MMP-2/-9
K. pneumoniae-specific FISH probe (P < 0.0001) (Fig. 3e,f). Finally, activity, as assessed by gelatin hydrolysis assay and gelatin zymog-
we demonstrated that upregulation of faecal K. pneumoniae in indi- raphy assays (Fig. 4d). To test this hypothesis in vivo, we depleted
viduals with HCC negatively correlated with overall survival and macrophages in germ-free mice using Clodrosome, followed by gav-
progression-free survival of HCC patients (Supplementary Fig. 5b). age of K. pneumoniae (Fig. 4e–h and Extended Data Fig. 8b–d). Before
These results implicate a living hepatic microbiome in individuals with collection, we gavaged these mice with Cy5.5-d-Lys-labelled K. pneu-
HCC with preferential enrichment of K. pneumoniae. moniae (Supplementary Fig. 9a) to trace its fate in vivo. While the
K. pneumoniae-only group showed positive fluorescence and culture in
K. pneumoniae initiates gut barrier disruption by gelatinase liver tissues with 100% incidence, Clodrosome pretreatment reduced
To identify the driver bacterium in promoting HCC, we next selected intrahepatic K. pneumoniae (2 out of 6) (Fig. 4f). K. pneumoniae-specific
5 top enriched Gram-negative bacteria (K. pneumoniae, Bacteroides FISH confirmed that K. pneumoniae levels were reduced in livers of
thetaiotaomicron (BT), Enterobacter hormaechei (EH), Escherichia macrophage-depleted mice (Extended Data Fig. 8c). Macrophage
coli (EC) and P. mirabilis (PM)) from liver tissues of HCC-FMT-treated depletion also alleviated K. pneumoniae-promoted gut barrier dys-
mice and gavaged them to germ-free mice for 10 weeks (Fig. 4a). Com- function, gelatinase activity (Fig. 4g), collagen degradation (Sup-
pared with PBS control, only K. pneumoniae elevated gut permeabi plementary Fig. 9b), disruption of colonic intercellular junctions and
lity (P < 0.0001) and translocated to the liver as live bacteria (Fig. 4b). bacterial infiltration into gut mucosa (Fig. 4h, Extended Data Fig. 8d
Moreover, K. pneumoniae disrupted colonic intercellular junctions, and Supplementary Fig. 9c). Consequently, Clodrosome relieved
reduced thickness of colonic mucus, downregulated expression of K. pneumoniae-induced liver pathologies (Supplementary Fig. 9d,e)
tight junction proteins (E-cad, CLDN3 and CLDN1) and increased bac- and oncogenic and pro-inflammatory signalling (Supplementary
terial infiltration into mucosal lamina propria and muscularis mucosa Fig. 9f,g). Overall, K. pneumoniae initiates gut barrier disruption
(Fig. 4c and Extended Data Fig. 8a), while other candidates had no by elevating macrophage-mediated gelatinase activity, which facili-
effect. This suggests that K. pneumoniae as a driver bacterium that tates its eventual translocation to the liver.
disrupts the gut barrier to enable its translocation to the liver.
Breakdown of extracellular matrix by gelatinase is the initial step K. pneumoniae promotes precancerous lesions and HCC in mice
in the invasion of pathogens into gut mucosa14. We thus assessed faecal To validate the function of K. pneumoniae in HCC, we gavaged
gelatinase activities in germ-free mice treated with individual bacte- K. pneumoniae to germ-free mice (GF/KP) for 36 weeks (Fig. 5a),
ria. Only faecal supernatants from the K. pneumoniae group lique- with E. coli MG1655 (GF/EC) and PBS as controls. Fluorescence
fied gelatin, suggesting gelatinase activity (Supplementary Fig. 6a). tracing of Cy5.5-d-Lys-labelled K. pneumoniae and E. coli identified
Gelatin zymography indicated Collagenase IV activity, as evidenced K. pneumoniae-derived signals in both the gut and liver, whereas
by clear bands at 72 kDa and 92 kDa (Supplementary Fig. 6b). Matrix E. coli signals were restricted to the gut (Supplementary Fig. 10a
metallopeptidase (MMP) activity assays validated the elevated MMP- and Fig. 5b). Bacterial culture and FISH showed the presence of live
2/-9 (type IV collagenases) activities in faecal supernatant from mice K. pneumoniae in GF/KP mouse liver (8 out of 10), whereas E. coli
treated with K. pneumoniae (Supplementary Fig. 6c). Consistent with and PBS were both negative (Fig. 5b). These results indicate that
in vitro data, Masson’s trichrome staining revealed collagen deple- monocolonization with K. pneumoniae is sufficient for its trans-
tion in colonic mucosa of K. pneumoniae-gavaged germ-free mice location to the liver. We confirmed previous observations that
compared with other bacteria or PBS, which was confirmed by IHC of K. pneumoniae compromised gut barrier function by inducing
collagen type IV (COLIV) and fibronectin (Supplementary Fig. 6d). In gelatinase activity (Fig. 5c–e, Extended Data Fig. 8e and Supple-
this connection, faecal supernatants from individuals with HCC and mentary Fig. 10b,c). Critically, K. pneumoniae-gavaged germ-free
LC exhibited increased gelatinase activity (Supplementary Fig. 7a), mice developed dysplasia that were absent in E. coli or PBS-gavaged
as did faecal supernatants from HCC-FMT- and LC-FMT-treated mice mice (P < 0.0001) (Fig. 5f). K. pneumoniae also induced hepato-
(Supplementary Fig. 7b). The direct effect of MMP-2/-9 in attenuating cyte proliferation (Fig. 5g and Extended Data Fig. 8f), fibrotic injury
gut barrier markers (COLIV, fibronectin, E-cad and ZO-1) was dem- (Fig. 5h and Extended Data Fig. 8g) and oncogenic and pro-
onstrated in normal NCM460 cells (Supplementary Fig. 7c). Hence, inflammatory signalling (Extended Data Fig. 8h). In addition, the
K. pneumoniae and HCC-FMT disrupt colonic barrier function by pro- protumorigenic effect of K. pneumoniae was further reproduced in
moting gelatinase activity. SPF mice (Fig. 5i–n and Supplementary Fig. 11a–d) and DEN-injected
Interestingly, K. pneumoniae-conditioned medium lacks gelati- SPF mice (Fig. 5o,p and Supplementary Fig. 12a–g). In these models,
nase activity (Supplementary Fig. 8a–c), implying an indirect effect K. pneumoniae disrupted gut barrier integrity and translocated to
of this bacterium. Macrophages are known to secrete MMP-2/-9 the liver. Our results implicate K. pneumoniae as a gut pathogen that
(refs. 15,16), which could damage gut barrier integrity17. We thus promotes precancerous lesions and HCC in mice.
Fig. 3 | Live bacteria, especially K. pneumoniae, are enriched in tumour probe FISH detection, LPS and LTA IHC staining of HCC patient liver tissues (n = 3
tissues of patients with HCC. a, Top: representative images of bacterial independent experiments with similar results). c, Stacked bar plot of relative
culture of liver tissues from patient with LC under anaerobic and aerobic abundance of culturable bacterial strains from liver tissues from individuals with
conditions. N = 15 biologically independent samples. From left to right and top LC or with HCC as determined by microbial mass spectrometry identification.
to bottom, the culture plates are blood agar plates, chocolate blood agar plates, d, Stacked bar plot of relative abundance at genus level of bacteria from liver
MacConkey agar plates and Columbia blood agar plates, respectively. Bottom: tissues from patients with LC or HCC as detected by 16S rRNA sequencing. N = 11
Cy3-conjugated EUB338 probe FISH detection, LPS and LTA IHC staining of (LC), 35 (HCC). e, Absolute number of K. pneumoniae in liver tissues from patient
liver tissues from patient with LC (n = 3 independent experiments with similar with LC and HCC as detected by qPCR. n = 10 biologically independent samples.
results). b, Top: representative images of bacterial culture of tumour tissues from f, Representative images of Cy3-conjugated K. pneumoniae probe FISH detection
patient with HCC under anaerobic and aerobic conditions. N = 54 biologically in liver tissues from patient with LC (top) and HCC (bottom). n = 10 biologically
independent samples. From left to right and top to bottom, the culture plates independent samples. In f, data are presented as mean ± s.e.m. Each data point in
are blood agar plates, chocolate blood agar plates, MacConkey agar plates and bar plots represents one participant. Statistical significance was calculated using
Columbia blood agar plates, respectively. Bottom: Cy3-conjugated EUB338 Student’s t-test.
Nature Microbiology
K. pneumoniae cell wall protein binds to and activates TLR4 assay demonstrated attachment of K. pneumoniae to Hep3B and
We next investigated the molecular mechanism of K. pneumoniae in Huh7 cells (Fig. 6a). Co-culture of pasteurized K. pneumoniae pro-
HCC. Because K. pneumoniae-conditioned medium (KPCM) had no moted growth of HCC cells and HCC patient-derived organoids by
effect on HCC cell viability (Supplementary Fig. 13a,b), we hypoth- inducing proliferation (Fig. 6b,c), suggesting that surface proteins of
esized that bacteria–HCC cells interaction is necessary. In line with K. pneumoniae might be involved. To unravel K. pneumoniae proteins
this, scanning electron microscopy (SEM), TEM and cell attachment that interact with HCC receptors, we performed biotin pulldown and
a Live bacteria in LC liver tissues : 12 out of 15

Contamination
control LC patient 1 LC patient 2 LC patient 3 LC patient 4 LC patient 5 LC patient 6 LC patient 7
Anaerobic
Aerobic
16S FISH
LTA
LPS
20 µm 50 µm 50 µm
b Contamination Live bacteria in HCC liver tissues: 28 out of 54

control HCC patient 1 HCC patinet 2 HCC patient 3 HCC patient 4 HCC patient 5 HCC patient 6 HCC patient 7
Anaerobic
Aerobic
16S FISH
LTA
LPS
20 µm 25 µm 50 µm
c Klebsiella pneumoniae d e
per mg human liver tissues
abundance at species level (%)
Enterococcus faecalis 100 Unclassified 6

100 Escherichia coli
Absolute KP number
Other 0.0002
Enterococcus casseliflavus
Relative abundance
Live bacteria relative
Staphylococcus
Staphylococcus warneri 75 Fusobacterium
at genus level
Actinomyces oris
Enterobacter bugandensis Streptococcus
3
Enterobacter cloacae 50 Blautia
50
Odoribacter splanchnicus Enterococcus
Staphylococcus epidermidis Escherichia
Staphylococcus hominis 25 Megamonas
Klebsiella variicola Prevotella 0
Phocaeicola vulgatus 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 910
Bilophila
0 Streptococcus mitis oralis 0
Bacteroides intestinalis Klebsiella LC patients HCC patients
LC HCC LC HCC
f LC1 LC2 LC3 LC4 LC5 LC6 LC7 LC8 LC9 LC10
200
KP number per slide
KP FISH
<0.0001
20 µm
100
HCC1 HCC2 HCC3 HCC4 HCC5 HCC6 HCC7 HCC8 HCC9 HCC10
KP FISH
20 µm 0
LC HCC
Nature Microbiology
a c e
<0.0001 <0.0001 0.001
8-week germ-free male BALB/c 1.0
disappear rate (%)

<0.0001 <0.0001 0.0036
120
E-cad optical
density (a.u.)
160 <0.0001
Tight junction
<0.0001 0.0005
positive (%)
<0.0001
Alcian blue
<0.0001 0.0005
0 10 weeks <0.0001
<0.0001 0.0139
8-week germ-free male BALB/c
80 60 0.5
Gavage once a week
0 10 weeks
PBS PBS gavage (n = 6) 0 Gavage once a week
0 0
BT Bacteroides thetaiotaomicron (BT) (n = 6) PBS PBS (n = 6)
<0.0001 <0.0001
PM Proteus mirabilis (PM) (n = 6) 0.6
<0.0001
1.0
number pe slide
160 KP Klebsiella pneumoniae (KP) (n = 6)
CLDN 3 optical
<0.0001
CLDN 1 optical
0.0004 0.0004
density (a.u.)
density (a.u.)
<0.0001
EC Escherichia coli (EC) (n = 6) 0.0034 0.0086
Bacteria
<0.0001
0.0012 0.0002
0.3 <0.0001 KPC Klebsiella pneumoniae + Clodrosome (KPC) (n = 6)
EH Enterobacter hormaechei (EH) (n = 6) 0.5 0.0007 0.0005 80
KP Klebsiella pneumoniae (KP) (n = 6)
0 0 0
PBS BT PM EC EH KP
f PBS KP KPC ×10
8
Epi-fluorescence
7
b d THP-1 differentiated into macrophage in vitro
(photons)
30
<0.0001
PBS 6
dextran (µg ml )
<0.0001 + +
CD14
−1
80 CD14 Phorbol
500 kDa FITC-
<0.0001 BT 60 5
<0.0001 8.1 ester 38.4
PM
Count
40
Count
15
<0.0001
40 4
EC 20
200 µm 200 µm
EH 0 0 7
101 102 103 104 105 101 102 103 104 105 ×10
0 KP 4.0
CD14-brilliant violet 711 CD14-brilliant violet 711
Epi-fluorescence
MMP activity of KP and marcrphage co-culture conditional medium 3.8
(photons)
PBS BT PM EC EH KP
Gelatin hydrolysis Gelatin zymography
0/6 0/6 0/6 0/6 0/6 6/6
3.6
EC KP
Culture
Ctrl 5% 10% 5% 10% 3.4

PC Ctrl EC KP
0.031
tissue
<0.0001 <0.0001
30
16S FISH
<0.0001 <0.0001
MMP-9 92 kDa
100
Liver with live

bacteria (%)
MMP-2 72 kDa
20 µm
15
−1
50
c.f.u. mg
0 0
g h
Gelatin hydrolysis Gelatin zymography 0.0022
0.0035
26 <0.0001 <0.0001 140 0.4 160
dextran (µg ml )
−1
PC PBS KP KPC
disappear rate (%)
500 kDa FITC-
Bacteria number
<0.0001 0.0021 0.0008 0.007 <0.0001 <0.0001
Tight junction
E-cad optical
density (a.u.)
MMP-9 92 kDa
per slide
13 MMP-2 72 kDa
70 0.2 80
0
PBS KP KPC
PBS KP KPC 0 0 0
MMP-2 MMP-9 80 <0.0001 0.001 0.4

Faecal supernatant
Faecal supernatant
MMP activity (RFU)
MMP activity (RFU)
80 PBS 80 PBS 0.001 0.007

CLDN 1 optical
density (a.u.)
positive (%)
Alcian blue
KP KP PBS
KPC KPC
<0.0001
<0.0001
KP
<0.0001
40 40 40 0.2
<0.0001
KPC
0 0
0 60 120 180 240 min 0 60 120 180 240 min 0 0
Fig. 4 | K. pneumoniae initiates gut barrier dysfunction by elevating colon of body and liver after Cy5.5-d-Lys-labelled K. pneumoniae gavage (top), and
macrophage-mediated MMP-2/-9 activities. a, Design of 5 Gram-negative live bacterial culture of liver tissues on blood agar plates and quantification
bacteria gavage in germ-free mice without DEN (PBS group n = 6; BT group n = 6; (bottom). n = 6 biologically independent samples. g, Gut permeability assays
PM group n = 6; EC group n = 6; EH group n = 6; KP group n = 6) and related results using 500 kDa FITC-dextran (top left) (n = 6 biologically independent samples)
(b,c). b, Gut permeability assays using 500 kDa FITC-dextran (top), live bacteria and MMP-2/-9 activity in mouse faecal supernatant detected using gelatin
translocated into liver cultured in blood agar plates (middle) and Cy3-conjugated hydrolysis assay (top middle), gelatin zymography (top right) (n = 3 independent
EUB338 probe FISH assay of liver tissues (bottom). n = 6 biologically independent experiments with similar results) and dynamic activity assay (bottom) (n = 6
samples. c, Quantitative analysis of TEM, Alcian blue staining, E-cad, CLDN3, biologically independent samples). h, Quantitative analysis of TEM, Alcian
CLDN1 IHC staining, and Cy3-conjugated EUB338 probe FISH of colon tissues. blue staining, E-cad and CLDN1 IHC staining, bacteria number per slide using
n = 10 biologically independent samples. d, THP-1-induced macrophage using Cy3-conjugated EUB338 probe (PBS group) and K. pneumoniae probe (KP and
phorbol ester (top row) and MMP-2/-9 activity in K. pneumoniae and THP-induced KPC groups) FISH of colon tissues. n = 10 biologically independent samples. In
macrophage co-culture conditional medium using gelatin hydrolysis assay b, c (excluding tight junction disappear rate) and f–h (excluding liver with live
and gelatin zymography (bottom row). n = 3 independent experiments with bacteria and tight junction disappear rate) data are presented as mean ± s.e.m.
similar results. Ctrl: BHI and macrophage co-culture conditional medium; EC: Each data point in bar plots represents one mouse. Liver with living bacteria and
E. coli and macrophage co-culture conditional medium; KP: K. pneumoniae and tight junction disappearance rate were analysed using Fisher’s exact test. MMP
macrophage co-culture conditional medium. e, Design of colon macrophage activity was analysed using two-way ANOVA. Unless otherwise stated, statistical
depletion using Clodrosome and its effect on K. pneumoniae-mediated gut significance was calculated using one-way ANOVA. Adjustments were made for
barrier dysfunction in germ-free mice without DEN treatment (PBS group n = 6; multiple comparisons.
KP group n = 6; KPC group n = 6) and related results (f–h). f, In vivo imaging
identified putative candidates by mass spectrometry (Fig. 6d). One (Fig. 6e and Supplementary Fig. 13d). Surface plasmon resonance
candidate protein, penicillin-binding protein 1B (PBP1B), is reported (SPR) using purified proteins confirmed direct PBP1B–TLR4 inter-
to localize to the peptidoglycan-based cell wall of K. pneumonia. We action with an affinity of 0.269 nM (Fig. 6f and Supplementary
next constructed glutathione S-transferase (GST)-tagged PBP1B (Sup- Fig. 13e). In silico modelling showed that PBP1B docks into the
plementary Fig. 13c), followed by GST pulldown to identify its corre- TLR4 catalytic site with binding energy of −318 kcal mol−1 (Fig. 6g).
sponding HCC receptor (Fig. 6e and Supplementary Fig. 13d). Overlap Co-immunoprecipitation (Co-IP) assays confirmed the interplay
of common candidate proteins unravelled TLR4 as a receptor for PBP1B between GST–PBP1B and TLR4 in Hep3B and Huh7 cells (Fig. 6h,i).
Nature Microbiology
To investigate whether PBP1B promotes malignant behaviour of inhibited TLR4 activation, hepatocyte proliferation (PCNA, Ki-67),
HCC cells, we incubated recombinant PBP1B with HCC cells in vitro. fibrosis marker expression, and oncogenic and pro-inflammatory
PBP1B promoted growth of HCC cells and HCC patient-derived orga- signalling (Extended Data Fig. 9k and Supplementary Fig. 15b–d).
noids (Fig. 6j). PBP1B robustly increased TLR4, MyD88, p-P65 and TLR4-IN-C34 thus acts by blocking local effects of K. pneumoniae in
PCNA protein levels (Fig. 6k), suggesting TLR4 signalling activation. the liver, without impacting gut barrier dysfunction or its transloca-
TLR4 inhibitor (TLR4i; TLR4-IN-C34) abolished PBP1B-induced HCC tion to the liver. We also confirmed that TLR4-IN-C34 blocked the
cell growth (Fig. 6l,m), suggesting TLR4 as its functional target. Fur- HCC-promoting function of HCC-FMT in DEN-treated germ-free mice
thermore, K. pneumoniae gavage to germ-free and SPF mice elevated (Extended Data Fig. 10a,b), an effect independent of gut barrier dys-
TLR4 expression (Fig. 6n). Our data suggest that K. pneumoniae PBP1B function (Extended Data Fig. 10c,d) but is driven by TLR4 downregula-
binds to TLR4 in HCC cells, leading to TLR4-mediated tumorigenesis. tion (Extended Data Fig. 10e–g). Collectively, both impairment of gut
permeability and local activation of the PBP1B–TLR4 axis in the liver are
Anticolonization or TLR4i reverses K. pneumoniae-induced essential for the HCC-promoting effect of K. pneumoniae, which can be
HCC targeted to overcome K. pneumoniae-induced hepatocarcinogenesis.
Given that colonic infiltration of K. pneumoniae is obligatory for trans-
location to the liver, suppressing its colonization in the gut represents Discussion
a potential therapeutic strategy. Klebsiella oxytoca has been reported Here we provide evidence of the causality of HCC-associated gut
to promote colonization resistance against pathogens18,19. We thus gav- dysbiosis in contributing to hepatocarcinogenesis. Reconstitution
aged DEN-treated germ-free mice with K. oxytoca, K. pneumoniae and of HCC microbiota in both germ-free mice and SPF mice led to spon-
K. oxytoca plus K. pneumoniae (Extended Data Fig. 9a). Remarkably, taneous liver inflammation, fibrosis and dysplasia, and exacerbated
K. oxytoca co-treatment reduced tumour incidence (P = 0.0027), carcinogen-induced HCC. Intriguingly, HCC microbiota was found
tumour number (P < 0.0001) and tumour burden (P = 0.0001) com- to impair gut barrier integrity, thereby enabling the transfer of live
pared with K. pneumoniae alone (Extended Data Fig. 9b). K. oxytoca pathogenic bacteria to the liver and consequential activation of
co-treatment also restored gut barrier integrity (Extended Data pro-inflammatory and oncogenic signalling to perpetuate tumori-
Fig. 9c,d), thereby preventing infiltration of K. pneumoniae into genesis. Characterization of the liver microbiota identified enrichment
mucosal lamina propria and muscularis mucosa (Extended Data Fig. 9c) of pathogenic K. pneumoniae in mice given HCC-FMT and in human
and translocation to the liver (Extended Data Fig. 9e and Supplementary patients with HCC. Finally, K. pneumoniae monocolonization in mice
Fig. 14a). K. oxytoca also abrogated the effect of K. pneumoniae on TLR4 phenocopied the effect of HCC-FMT, corroborating K. pneumoniae as
activation, hepatocyte proliferation (PCNA, Ki-67) and oncogenic and an oncogenic pathogen in HCC.
pro-inflammatory signalling (Supplementary Fig. 14a,b). K. oxytoca Emerging evidence suggests that gut dysbiosis impacts liver dis-
itself had no effect on HCC development. These results indicate that eases via the gut–liver axis13. We investigated this proposition in HCC
K. oxytoca antagonizes K. pneumoniae-induced HCC by arresting its by transplantation of HCC patient stool to germ-free mice or SPF mice,
colonic intrusion and translocation to the liver. revealing spontaneous development of liver inflammation and fibro-
Next, we asked whether TLR4 blockade might overcome sis, characterized by induction of pro-inflammatory TH1/TH17 cells
HCC-promoting K. pneumoniae via the PBP1B–TLR4 axis. Supporting and suppression of anti-inflammatory TH2 cells. Chronic inflamma-
our hypothesis, the TLR4i TLR4-IN-C34 reversed the effect of K. pneu- tion and fibrosis are contributors to hepatocarcinogenesis20. Indeed,
moniae on tumour incidence (P = 0.0011), tumour number (P < 0.0001) half of HCC-FMT mice developed dysplasia. Furthermore, HCC-FMT
and tumour burden (P = 0.0004) in DEN-injected germ-free mice accelerated DEN-induced HCC, thus establishing a causative role of
(Extended Data Fig. 9f,g). Although TLR4-IN-C34 failed to restore HCC-associated dysbiosis in hepatocarcinogenesis.
the gut barrier or prevent translocation of K. pneumoniae to the liver Gut barrier integrity is fundamental to a functional gut–liver axis,
(Extended Data Fig. 9h–j and Supplementary Fig. 15a), it effectively as it prevents indiscriminate access of bacteria and their products
Fig. 5 | K. pneumoniae promotes precancerous lesions and HCC in both EC group n = 6; KP group n = 6) and relevant results (j–n). j, Gut permeability
germ-free and SPF mice. a, Design of K. pneumoniae gavage for promoting assays using 500 kDa FITC-dextran. n = 6 biologically independent samples.
precancerous lesions in germ-free mice without DEN treatment (GF/PBS group k, Live bacterial culture of liver tissues on blood agar plate (top), and Cy3-
n = 10; GF/EC group n = 8; GF/KP group n = 10) and related results (b–h). conjugated EUB338 probe (PBS group), E. coli probe (EC group) or K. pneumoniae
b, In vivo imaging of body and liver after Cy5.5-d-Lys-labelled K. pneumoniae probe (KP group) FISH detection in liver tissues (bottom). n = 6 biologically
gavage (left), live bacteria culture of liver tissues on blood agar plate (top right) independent samples. l, TEM (left) and quantitative analysis (right) of colon
and Cy3-conjugated EUB338 probe (GF/PBS group), E. coli probe (GF/EC group) tissues. n = 6 biologically independent samples. m, Representative images of
or K. pneumoniae probe (GF/KP group) FISH detection in liver tissues (bottom liver gross morphology (top) (dashed yellow circles indicate nodules), H&E
right). c, Gut permeability assays using 500 kDa FITC-dextran (left) (n = 10 staining of mice liver sections (middle), and nodule number and incidence of
(GF/PBS), 8 (GF/EC), 10 (GF/KP)) and quantitative analysis of TEM (middle) dysplasia (bottom). n = 6 biologically independent samples. n, Representative
(n = 6 biologically independent samples) and E-cad IHC staining (right) images of IHC staining for PCNA and Sirius red staining of liver sections (left) with
(n = 6 biologically independent samples). d, MMP-2/-9 activity in mouse faecal quantitative analysis (right). n = 6 biologically independent samples. o, Design
supernatant using gelatin hydrolysis assay (bottom left), gelatin zymography of K. pneumoniae gavage on HCC in SPF mice with DEN treatment (DPBS group
(top left) and dynamic activity assay (right). n = 3 biologically independent n = 7; DEC group n = 6; DKP group n = 7) and related results (p). p, Representative
samples. e, Masson’s trichrome staining (top left) and COLIV IHC staining images of liver gross morphology (left) (dashed yellow circles indicate tumour)
(top right) of colon tissues and COLIV protein level by western blot (bottom). and H&E staining of mice liver sections (middle) with tumour incidence and
n = 6 biologically independent samples. f, Representative images of liver gross tumour number quantitative analysis (right). n = 7 (DPBS), 6 (DEC), 7 (DKP).
morphology (top left) (dashed yellow circles indicate nodules), H&E staining of In c (excluding tight junction disappear rate), d, e–h (excluding incidence of
mice liver sections (bottom left), and nodule number and incidence of dysplasia dysplasia), j,k,m (excluding incidence of dysplasia), n and p (excluding tumour
(right). n = 10 (GF/PBS), 8 (GF/EC), 10 (GF/KP). g, IHC staining for PCNA and incidence), data are presented as mean ± s.e.m. Each data point in bar plots
Ki-67 (left) (n = 6 biologically independent samples), and western blot assay represents one mouse. Tight junction disappearance rate, incidence of
of PCNA (right) (n = 3 biologically independent samples). h, Desmin (left) and dysplasia and tumour incidence were analysed using Fisher’s exact test. MMP
α-SMA (middle) IHC staining, and Sirius red staining (right) of liver sections. activity was analysed using two-way ANOVA. Unless otherwise stated, statistical
n = 6 biologically independent samples. i, Design of K. pneumoniae gavage on significance was calculated using one-way ANOVA. Adjustments were made for
precancerous lesions in SPF mice without DEN treatment (PBS group n = 6; multiple comparisons.
Nature Microbiology
to portal circulation21. Here we observed that HCC-FMT increased with a previous study indicating that Klebsiella was increased in patients
colonic gelatinase activities. Excess gelatinase activities can degrade with early HCC26. K. pneumoniae has been reported to disrupt the intes-
extracellular matrix and impair epithelial integrity in the gut14,22–24. tinal barrier27–29; however, its molecular mechanism remains unclear.
Gelatinase activity thus has a crucial role in induction of mucosal dam- Here we showed that K. pneumoniae promoted macrophage-derived
age, loss of barrier function and disruption of intestinal tight junctions gelatinase activity in colon to induce gut barrier dysfunction. Together,
in HCC-FMT-treated mice. This is manifested by a marked increase in K. pneumoniae is a driver pathogen that impairs gut barrier function
gut permeability after HCC-FMT, exposing the liver to gut pathogens, and enables its own colonization in the liver.
pathogen-associated molecular patterns and metabolites25. K. pneumoniae is part of the human gut flora (~5% prevalence)30 and
Consequently, impaired gut barrier function led to colonization an opportunistic pathogen involved in hepatobiliary infections31, pyo-
of the liver by live and culturable gut microbes in HCC-FMT-treated genic liver abscess32 and non-alcoholic fatty liver disease28; however,
mice that were absent in mice given HD-FMT. Although recent studies its role in tumorigenesis is unknown. We revealed that K. pneumoniae
have described intratumoural microbiota in many cancers4, the origin exerts protumorigenic function by stimulating the growth of HCC cells
of these microbes remains enigmatic. Our results implicate the gut and HCC patient-derived organoids. In mice, intrahepatic coloniza-
as the origin of intrahepatic pathogens in HCC. Cross-validation in tion of K. pneumoniae robustly promotes inflammation, fibrosis and
patients with HCC and HCC-FMT-treated mice demonstrated increased oncogenic/inflammation signalling. This indicates K. pneumoniae as a
K. pneumoniae in both stool and liver microbiomes. This is consistent gut pathogen promoting hepatocarcinogenesis. In line with our data,
a b Live KP translocated from gut into liver c

8
8-week germ-free male BALB/c GF/ GF/ GF/ ×10 <0.0001 <0.0001 <0.0001
disappear rate (%)

GF/PBS GF/EC GF/KP 80 100 0.4
dextran (µg ml )
PBS EC KP <0.0001
−1
500 kDa FITC-
Tight junction
E-cad optical
<0.0001 <0.0001
density (a.u.)
3
Epi-fluorescence
0 36 weeks
Culture
Gavage once a week
0/10
8/10
0/8
2
(photons)
40 50 0.2
GF/PBS PBS (n = 10)
1
GF/EC Escherichia coli MG1655 (n = 8) ×10
7
1.2 0 0 0
FISH
GF/KP Klebsiella pneumoniae (n = 10)

30 µm
0.6
d GF/GF/GF/ 100
e 0.0179 0.0096 f GF/PBS GF/EC GF/KP 4
0.0004
80 0.0266 0.8 0.0009
MMP-2 activity
PC PBS EC KP GF/PBS COL IV relative COL IV optical

expression level density (a.u.)
number
Nodule
0.0115
positive (%)
MMP-9 GF/EC
Masson
GF/KP 2
<0.0001
<0.0001
(RFU)
92 kDa
MMP-2 50 40 0.4
72 kDa
0
<0.0001
0 0 0 100
dysplasia (%)
Incidence of
0 20 40 60 80 100 120 min <0.0001
14 0.0057
GF/ GF/ GF/
GF/ GF/ GF/ 400
0.0075 50
PBS EC KP
MMP-9 activity
PBS EC KP GF/PBS COL IV

GF/EC 7
H&E
GF/KP
161 kDa 0
(RFU)
<0.0001
200 GF/PBS
<0.0001
β-actin
42 kDa 0 GF/EC
200 µm
0 GF/PBS GF/EC GF/KP GF/KP
0 20 40 60 80 100 120 min
g 8
0.0012
0.0025
GF/PBS GF/EC GF/KP h i j
PCNA
positive (%)
0.4
<0.0001
0.4
<0.0001
8
<0.0001
8-week SPF male C57BL/6J 46 <0.0001
dextran (µg ml )
Desmin optical
29 kDa
−1
α-SMA optical
<0.0001 <0.0001 <0.0001
500 kDa FITC-

PCNA
density (a.u.)
density (a.u.)
<0.0001
positive (%)
4 β-actin 0 36 weeks
Sirius red
42 kDa 0.2 0.2 4 Gavage once a week

23
0 3.0 0.0017 PBS PBS (n = 6)
Relative PCNA
0.0004 0.0018
expression
4 0.0007 EC Escherichia coli (n = 6)

0 0 0
positive (%)
0
1.5 KP Klebsiella pneumoniae (n = 6) PBS EC KP
Ki-67
2 GF/PBS GF/EC GF/KP
0
0 GF/PBS GF/EC GF/KP
k PBS EC KP m PBS EC KP
n <0.0001
8
PBS EC KP <0.0001
Culture
positive (%)
0.0006
60
0/6
0/6
5/6
Bacteria number
0.0006
PCNA
PCNA
4
per slide
50 µm
30
0
FISH
0
30 µm 4
<0.0001
H&E
positive (%)
Sirius red
Sirius red
<0.0001
l PBS EC KP <0.0001 200 µm 250 µm

2
disappear rate (%)
100 <0.0001
Tight junction
0.0233 <0.0001 0
TEM
100
dysplasia (%)
Incidence of
4 0.0233 <0.0001 PBS

50 PBS KP EC
number
Nodule
0.5 µm EC
2 50
KP
0
Tight junction Adherens junction 0 0
o 14-day SPF male C57BL/6J p DPBS DEC DKP

DPBS DEC DKP 0.034
0.021 6 0.0114
Tumour incidence (%)
100
Tumour number
DEN 0.0014
0 6 29 weeks DPBS
Gavage twice a week 500 µm DEC
3
H&E
DPBS DEN + PBS (n = 7) 50 DKP
DEC DEN + Escherichia coli (n = 6)

DKP DEN + Klebsiella pneumoniae (n = 7) 0 0
Nature Microbiology
a b c
Hep 3B Huh 7 Ctrl Pasteurized KP
Cell proliferation
Cell proliferation
1.0 2 <0.0001
<0.0001
Ctrl Ctrl 3
PCNA
(OD = 570)
(OD = 570)
KP SEM of KP adhesion to HCC cells
Relative PCNA
<0.0001
Pasteurized KP Pasteurized KP
Hep 3B
29 kDa 2
1
protein
0.5
β-actin
0.0031
Ctrl KP 42 kDa 1
0 0
0 1 2 3 4 5 Day 0 1 2 3 4 5 Day Ctrl Pasteurized KP
Hep3B
0
KP Hep 3B Huh 7 PCNA Hep 3B Huh 7
29 kDa
Huh 7
2 µm 5 µm 0.5 µm Ctrl
Ctrl Pasteurized KP Ctrl Pasteurized KP β-actin Pasteurized KP
42 kDa
d
Huh7
KP – – + + - - + + KP
KP
2 µm 5 µm 0.5 µm - + - + Biotin- Huh 7
0.0155 0.0022
Biotin-Hep 3B – + – +
200 200
Colony formation
Colony formation
+ - + -
of Hep 3B (%)
Hep 3B + – + – Huh 7
of Huh 7 (%)
TEM of KP adhesion to HCC cells Hep 3B
Marker Marker
Live KP attached to HCC cells
Hep 3B + KP Huh 7 + KP 100 100
180 kDa 180 kDa

130 kDa 130 kDa
0 0
100 kDa PBP1B 100 kDa PBP1B
2 µm 2 µm Ctrl Pasteurized KP 95 kDa 95 kDa
4 <0.0001
70 kDa 70 kDa
HCC organoid
organoid (mm)
Huh 7
Diameterew of
55 kDa 55 kDa
2 40 kDa 40 kDa
2 mm
1 µm 1 µm 35 kDa 35 kDa
Ctrl Pasteurized KP 0
e f Kd (M) = 2.6910
−10
100 nM h i IgG + – –
NC GST GST-PBP1B 160 Hep 3B α-TLR4 – + +
50 nM
Response unit
+ + + + + Anti-GST beads 25 nM IP:GST Input GST-PBP1B + – +

GST-PBP1B GST 123 kDa
Hep 3B TLR4
GST/GST-PBP1B 80 12.5 nM – + – +
– + + + +
IP
Marker
Hep3B
6.25 nM Hep 3B + + + +
TLR4 96 kDa
+ – + – + Hep 3B membrane 3.125 nM GST 113 kDa
0
Input
GST 113 kDa
0 100 200 300 400 seconds TLR4 96 kDa
180 kDa g TLR4 96 kDa
130 kDa Binding energy V85 R289
H229
IgG + – –
E336 2.7 T175
100 kDa TLR4 = –317.60 kcal mol
−1 2.7
R92 2.4 Huh 7 α-TLR4 – + +
2.0
96 kDa
1.6
70 kDa D88
2.4
IP:GST Input GST-PBP1B + – +
D96
GST-PBP1B GST 123 kDa
2.8
T357 D432 – + – +
TLR4
55 kDa R300
Huh 7 + + + +
IP
TLR4 96 kDa
GST 113 kDa
1.8
40 kDa S244
GST 113 kDa
2.2
K477
Huh 7
Input
Y246
TLR4 96 kDa
35 kDa D550 2.9
Q384 TLR4 96 kDa
S545 2.1
j k n
Hep 3B Huh 7 PBS EC KP
Cell proliferation
Cell proliferation
1.0 1.8 Hep 3B Huh 7 Hep 3B Huh 7 In germ-free mice

expression level
expression level
Ctrl Ctrl 2 2 without DEN

<0.0001
<0.0001
PBP1B PBP1B 0.0009 0.019

Ctrl PBP1B Ctrl PBP1B
Relative
Relative
0.5 0.9
TLR4 96 kDa 1 0.0009 1 0.0064 50 µm
0 0 MyD88 33 kDa 0 0
0 1 2 3 4 5 day 0 1 2 3 4 5 day PBS EC KP
TLR4 MyD88 TLR4 MyD88
p-P65 65 kDa
without DEN
In SPF mice
Hep 3B Huh 7
expression level
expression level
2 0.0012 2
Ctrl PBP1B Ctrl PBP1B P65 65 kDa 0.0169
Relative
Relative
0.0392 50 µm
PCNA 29 kDa 1 1 0.044
PBS EC KP
β-actin 42 kDa
In SPF mice
0 0
with DEN
Colony formation
Colony formation
0.01
0.0195 p-P65/P65 PCNA p-P65/P65 PCNA
300 260
l m
of Hep 3B (%)
of Huh 7 (%)
Hep 3B Hep 3B Huh 7 50 µm

Cell proliferation Cell proliferation
150 130
2
Ctrl PBP1B + PBP1B +
(OD = 570)
PBP1B + TLR4i Ctrl TLR4i Ctrl TLR4i PBS BT PM

1
0 0
In germ-free mice
0 50 µm
without DEN
Ctrl PBP1B 4
0 1 2 3 4 5 day
organoid (mm)
<0.0001
Diameterre of
HC corganoid
120
Colony formation
120
Colony formation
Huh 7 EC EH KP
of Hep 3B (%)
of Huh 7 (%)
2
2 Ctrl
(OD = 570)
PBP1B + TLR4i
60 60
2 mm
1
0
Ctrl Pasteurized KP
0 0 0
0 1 2 3 4 5 day Ctrl PBP1B + Ctrl PBP1B +
TLR4i TLR4i
tumour-resident microbiota have been shown to have contributing represent viable strategies to repress K. pneumoniae-induced HCC
roles in promoting cancer progression and immunosuppression7,33. progression. Alternatively, recent work showed that K. pneumoniae
Mechanistically, K. pneumoniae promotes HCC via its surface could also be targeted by bacteriophages34.
protein PBP1B, which interacts with and activates TLR4 signalling in In summary, in this study we established a causal link between
HCC cells. Indeed, TLR4 blockade reversed the protumorigenic effect HCC-associated gut dysbiosis and hepatocarcinogenesis. HCC dys-
of K. pneumoniae and HCC-FMT in mice. Importantly, the abilities of biosis disrupts gut barrier function and promotes translocation of live,
K. pneumoniae to disrupt gut barrier integrity and locally activate pathogenic bacteria via the gut–liver axis. We identified K. pneumoniae
TLR4 in the liver are both essential for the HCC-promoting effect as a gut pathogen that initiates gut barrier dysfunction and translocates
of this bacterium. Hence, targeting its gut colonization or TLR4 to the liver, where it exacerbates HCC development via the PBP1B–TLR4
Nature Microbiology
Fig. 6 | K. pneumoniae promotes HCC via its surface protein PBP1B binding to h, TLR4 from Hep3B and Huh7 cell lysates was pulled down by GST–PBP1B
and activating TLR4 in HCC cells. a, Representative SEM (top) and TEM (bottom) according to the GST pull-down assay. i, Immunoprecipitation of TLR4 from
images of Hep3B and Huh7 HCC cell lines after co-culture with K. pneumoniae Hep3B or Huh7 cells lysates validated the binding between TLR4 and GST–
(MOI = 10). After co-culture of HCC cells with K. pneumoniae, cell lysate was PBP1B. In h and i, n = 3 independent experiments with similar results. j, Effect of
spread on BHI agar for bacterial culture. n = 3 independent experiments with PBP1B (0.05 μM) on HCC cell proliferation (top) (n = 3 biologically independent
similar results. b, Effect of pasteurized K. pneumoniae (MOI = 10) on HCC cell samples), colony formation (middle) (n = 3 biologically independent samples)
proliferation (top) (n = 3 biologically independent samples), colony formation and growth of HCC patient-derived organoid (bottom) (n = 20 biologically
(middle) (n = 3 biologically independent samples) and growth of HCC patient- independent samples). n = 3 independent experiments with similar results.
derived organoid (bottom) (n = 20 biologically independent samples). c, Effect k, Effect of PBP1B (0.05 μM) on TLR4, MyD88, p-P65, P65 and PCNA protein
of pasteurized K. pneumoniae (MOI = 10) on PCNA expression in HCC cells. n = 3 expression in HCC cells, as determined by western blot. n = 3 biologically
(Ctrl), 6 (pasteurized KP). d, Screening of K. pneumoniae adhesins by biotin pull- independent samples. l,m, TLR4i (30 μM) abolished the effect of PBP1B on HCC
down assay. e, Hep3B membrane protein was incubated with GST or GST–PBP1B cell proliferation (l) and colony formation (m). n = 3 biologically independent
together with GST magnetic beads for GST pull-down assay. Corresponding samples. n, Effect of K. pneumoniae on TLR4 expression in liver tissues of both
bands in GST–PBP1B and Hep3B cell membrane protein groups were subjected germ-free and SPF mice with or without DEN treatment by IHC staining. In b,c,
to mass spectrometry analysis. In d and e, n = 3 independent experiments with j–m, data are presented as mean ± s.e.m. Each data point in bar plots represents
similar results. f, Binding affinity between PBP1B and TLR4 was detected using one mouse. Cell proliferation was analysed using two-way ANOVA. In b,c,j,k,m,
SPR. Kd, dissociation constant. g, Representative structure of PBP1B and TLR4 comparisons between two groups were analysed using Student’s t-test. Unless
after molecular docking (left) and the residues involved in the interaction otherwise stated, statistical significance was calculated using one-way ANOVA.
between PBP1B with TLR4 (right). The structure of TLR4 is coloured pink while Adjustments were made for multiple comparisons.
the PBP1B is coloured yellow. The light green dash represents hydrogen bond.
axis (Supplementary Fig. 17). Targeting K. pneumoniae represents a Mice

strategy for HCC prevention and treatment. The DEN-induced HCC Animals used in experiments were gender and age matched. SPF BALB/c
mouse model used in this study is not a perfect tumour model and and C57BL/6J mice were purchased from GemPharmatech and raised
probably does not reflect human disease very well. at Jingtuo Biotechnology. Germ-free male BALB/c mice were bred and
housed at the Germ-Free Mouse Research Facility, The First Affiliated
Methods Hospital of Sun Yat-sen University or Shenzhen Jingtuo Biotechnology.
Ethics oversight Mice were raised in SPF or sterile facilities, following a cycle of 12 h
Written informed consent was obtained for all human samples and this light and 12 h dark conditions. Food (for SPF mice: 1010058, Jiangsu
study was approved by the Ethics Committee of The First Affiliated Xietong; for germ-free mice: 1019018, Jiangsu Xietong) and sterile
Hospital of Sun Yat-sen University (ID No. [2024]141 and [2020]361). water were provided ad libitum. Mice were randomly assigned to vari-
Corresponding compensation was provided to participants. ous experimental groups and experimental treatments. No statistical
All experiments were approved by the Animal Experimentation methods were used to predetermine sample sizes, but our sample sizes
Ethics Committee of The First Affiliated Hospital of Sun Yat-sen Uni- are similar to those reported in previous publications6. All mice were
versity (ID No. [2021]506, [2023]361, [2024]141) or Shenzhen Jingtuo euthanized by pentobarbital sodium and cervical dislocation.
Biotechnology (ID No. [2023]51), which required a maximum tumour
volume not exceeding 2,000 mm3, maximum diameter not exceed- Bacteria
ing 15 mm and tumour weight not exceeding 5% of body weight. The K. pneumoniae (isolated from tumour tissues of patients with HCC
maximum tumour size and tumour burden in this study did not exceed in this study), E. coli (used in Fig. 4a, isolated from tumour tissues of
this limit. patients with HCC in this study), B. thetaiotaomicron (isolated from
tumour tissues of HCC-FMT-treated mice in this study), P. mirabilis
Human samples collection (isolated from tumour tissues of HCC-FMT-treated mice in this
As shown in Supplementary Table 1, we collected stool from healthy study), Enterobacter hormaechei (isolated from tumour tissues of
donors (N = 20), individuals with LC (N = 20) and individuals with HCC HCC-FMT-treated mice in this study), E. coli MG1655 (700926, Ameri-
(N = 20). In addition, we also collected stool from another batch of can Type Culture Collection (ATCC)) and K. oxytoca (DSM5175, DSMZ)
healthy donors (N = 5) and individuals with HCC (N = 5) (Supplementary were all cultured in brain heart infusion (BHI) broth (024053, Huanan
Table 2) at The First Affiliated Hospital of Sun Yat-sen University. The Microbial) at 37 °C under aerobic/anaerobic conditions.
exclusion criteria were as follows: (1) use of antibiotics within the past
3 months, (2) had received previous cancer treatment, (3) had a history Cell lines
of other cancers and (4) presence of intrahepatic cholangiocarcinoma Both human HCC cell lines Hep3B (HB-8064, ATCC), Huh7 (01042712,
and a history of gastrointestinal surgeries. Fresh faecal samples were Sigma-Aldrich) and normal colonic epithelial cells NCM460
collected using sterile stool collection tubes, placed in anaerobic (NCM460D, Incell) cells were cultured in Dulbecco’s modified Eagle’s
conditions created by the AnaeroPack-Anaero (C-35, Mitsubishi Gas medium (DMEM) (11965092, Thermo Fisher) supplemented with 10%
Chemical) and transported to the laboratory on ice. Stool samples were fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 °C in a
suspended in sterile PBS in an anaerobic glove box (COY-8300400, Coy humidified atmosphere of 5% CO2.
Laboratory) and the suspension was filtered using a 40-μm cell strainer
and centrifuged at 100 g for 1 min to remove debris. Supernatant was Germ-free and SPF mouse models
then centrifuged at 10,000 g for 8 min, and the bacterial pellet was For germ-free mice without DEN treatment, 8-week-old male germ-free
suspended with 25% sterile glycerol and stored at −80 °C. We also col- BALB/c mice were selected for FMT or bacterial gavage. For SPF mice
lected stool samples from individuals with HCC (N = 24) with overall without DEN treatment, 8-week-old male SPF BALB/c or C57BL/6J mice
survival and progression-free survival data (Supplementary Table 3). were selected for FMT or bacterial gavage. The number of mice in each
Fresh LC tissues (N = 15) and HCC tumours tissues (N = 54) were group is shown in the figure legends.
collected for live bacterial culture, 16S rRNA sequencing and wax block For germ-free mice with DEN treatment, 14-day-old male
embedding (Supplementary Table 4) from patients who underwent germ-free BALB/c mice were selected and given 25 mg kg−1 DEN
surgery at The First Affiliated Hospital of Sun Yat-sen University. (N0756, Sigma-Aldrich) intraperitoneal injection. For SPF mice with
Nature Microbiology
DEN treatment, 14-day-old male SPF BALB/c or C57BL/6J mice were Operational taxonomic unit (OTU) analysis was performed using the
selected and given 25 mg kg−1 DEN intraperitoneal injection. The num- Uparse package (v.7.0.1001) with 97% sequence identity. Each OTU was
ber of mice in each group is shown in the figure legends. FMT or other taxonomically assigned to the silva database using the RDP classifier.
bacterial (as shown in figure legends) gavage only started when mice OTUs with relative abundance values >0.001% in at least one sample
reached adulthood at 8 weeks of age. were retained.
At the end point, mice were anaesthetized with sodium pento-
barbital intraperitoneally. Tumour number and size in the liver were Shotgun metagenomic sequencing
recorded. Tumour volume was measured by length (a) and width DNA of mouse faecal bacteria was extracted using the Quick-DNA Fae-
(b) using a digital caliper and calculated as tumour volume = ab2/2. cal/Soil Microbe Miniprep kit (D6010, Zymo). Shotgun metagenomic
Tumour burden was the summation of tumour volumes of each liver. sequencing of faecal DNA was performed on an Illumina NovaSeq 6000
Liver and colon tissues were collected in 4% paraformaldehyde, 2% platform by Gene Denovo. Taxonomy was assigned to metagenomic
glutaraldehyde or snap-frozen. For bacterial culture and 16S rRNA reads using k-mer-based algorithms implemented in the Kraken taxo-
sequencing, fresh liver tissues were collected in the biosafety cabinet. nomic annotation pipeline. A standard database comprising 13,844
Liver histology was assessed by H&E staining (BL735B, Biosharp) of bacterial genomes from NCBI was built using the Jellyfish programme
paraffin-embedded sections. Fresh liver and colon tissues were col- by counting distinct 31-mers. Each k-mer in a read was mapped to the
lected for flow cytometry. Liver histology was evaluated by a patho lowest common ancestor of all reference genomes with exact k-mer
logist, B.L., blinded to the treatment conditions. matches. Each query was thereafter classified to a taxon, with the
highest total hits of k-mers matched by pruning the general taxonomic
Bacterial culture from liver tissue and strain identification trees affiliated with mapped genomes. Final metagenomic read counts
Fresh liver tissues from mice, and patients with LC or HCC were col- were normalized with the cumulative sum scaling method using meta
lected in sterile tubes with BHI broth under aerobic and anaerobic con- genome Seq R/Bioconductor.
ditions. Tissues were washed six times with sterile phosphate-buffered
saline (PBS) and chopped into pieces. Tissue pieces were homogenized K. pneumoniae correlates with prognosis of patients with HCC
using TissueLyser II (85300, QIAGEN), and 100 μl tissue homogenate K. pneumoniae relative abundance in stool samples from individuals
was directly spread onto different agar plates including blood agar with HCC (N = 24) was detected using qPCR assay (Forward: 5′-CGATG
plates, chocolate agar plates, MacConkey agar plates and Columbia CTACTTATCCCGACA-3′; Reverse: 5′-AGCCGGTTGAGACGTAAAC-3′),
blood agar plates in both anaerobic and aerobic conditions at 37 °C. and correlation analysis between K. pneumoniae relative abundance
Colony-forming units (c.f.u.) per mg tissue was counted. Full-length and overall survival or progression-free survival was performed.
16S rRNA sequencing or microbial mass spectrometry (MALDI Biotyper
smart, Bruker) was used for strain identification. IHC
IHC on paraffin-embedded tissues was used for the detection of
Whole-genome sequencing and assemblies of K. pneumoniae PCNA (1:5,000), Ki-67 (1:400), TLR4 (1:5,000), E-cad (1:200), CLDN1
Genomic DNA was extracted from the isolated K. pneumoniae using (1:200), CLDN3 (1:100), α-SMA (1:1,000), desmin (1:2,000), fibronectin
the Tris-EDTA-NaCl buffer extraction method. The collected DNA (1:2,000), COLIV (1:500), LPS (1:200) and LTA (1:400) (antibodies listed
was detected using agarose gel electrophoresis and quantified using in Supplementary Table 5). Quantitative analysis was performed using
Qubit. After quality control, the high-quality genomic DNA was sub- ImageJ (v.1.51, National Institutes of Health (NIH)).
jected to sequencing library building and the genome of K. pneumo-
niae was sequenced using single-molecule, real-time technology Sirius red staining
(Novogene Bioinformatics). Low-quality reads were filtered using Sirius red staining was conducted using the commercial kit (BP-DL030,
SMRT Link v.8.0 and the filtered reads were assembled using the Nanjing SenBeiJia) according to manufacturer instructions to evalu-
software Canu to generate one contig without gaps. PacBio long-read ate collagen fibre content in liver tissues.Quantitative analysis was
sequencing of our K. pneumoniae isolate from HCC tissues revealed performed using ImageJ (v.1.51, NIH).
the absence of virulence factors associated with liver abscess (for
example, peg-344, iroB, iucA, rmpA and rmpA2 encoding proteins) Masson’s trichrome staining
(Supplementary Table 6), which might explain why it colonizes in Masson’s trichrome staining was conducted using the commercial kit
HCC without causing abscesses. (BP-DL022, Nanjing SenBeiJia) according to manufacturer instructions
to evaluate collagen fibre content in colon tissues. Quantitative analysis
16S rRNA sequencing of liver tissues and stool samples was performed using ImageJ (v.1.51, NIH).
Liver tissues from mice and humans were collected in a biosafety
cabinet for 16S rRNA sequencing. Liver tissues were disrupted by Western blot
bead beating after digestion with mutanolysin and lysozyme. DNA Total protein was extracted using Minute Total Protein Extraction Kit
extraction was performed with the QIAamp DNA mini kit (51304, for Animal Cultured Cells/Tissues (SD-001, Invent Biotechnologies).
QIAGEN). DNA of mice and human faecal bacteria were extracted Protein levels were measured using the BCA Protein Assay kit (ZJ101,
using the Quick-DNA Faecal/Soil Microbe Miniprep kit (D6010, Shanghai Epizyme). Proteins (30 μg) were separated using 12% gel an
Zymo). DNA library preparation and 16S rRNA gene sequencing d transferred onto PVDF membranes (SEQ85R, Merck Millipore). Mem-
were performed on an Illumina NovaSeq 6000 platform by Gene branes were incubated with primary antibody (1:1,000) overnight at
Denovo. The V3–V4 region of bacterial 16S rRNA was PCR amplified 4 °C and then secondary antibody (1:5,000) (HRP-linked anti-mouse
using 16S primers 341F (5′-CCTAYGGGRBGCASCAG-3′) and 806R IgG 7076, HRP-linked anti-rabbit IgG 7074, Cell Signaling Technology)
(5′-GGACTACNNGGGTATCTAAT-3′), with no template DNA reaction for 1 h. Proteins were visualized using a chemiluminescence reagent
as negative control. Sequencing libraries were generated using the (SQ201, Shanghai Epizyme). β-actin was used as the total protein load-
TruSeq DNA PCR-Free Sample Preparation kit (FC-121-3001, Illu- ing control. Antibodies used are listed in Supplementary Table 5.
mina) and sequenced on an Illumina NovaSeq platform. After qual-
ity control, paired-end reads were overlapped to assemble the final FITC-dextran permeability assay
sequences (overlap region >10 bp, mismatch ratio <0.2). Chimaera Mice were gavaged with 500 kDa FITC-dextran (FD500S, Sigma-Aldrich)
tags were filtered out using the Gold database by UCHIME (v.4.2.40). after overnight fasting. Blood samples were collected from the tail vein
Nature Microbiology
at 4 h post administration. FITC-dextran in the plasma was measured in (1:100; 107622, Biolegend) or Ly-6C (1:100; 560525, BD Biosciences).
a spectrophotometer (Varioskan LUX, Thermo Scientific). For intracellular markers, cells were stimulated for 8 h in DMEM with
leucocyte activation cocktail. Then, cells were fixed and permeabilized
Gelatinase activity assay with BD GolgiPlug kit (550583, BD Bioscience) and then stained with
The gelatin hydrolysis activity of faecal supernatant with various treat- intracellular cytokine IFNγ (1:100; 557649, BD Bioscience), IL-4 (1:100;
ments or KPCM were detected using a gelatin biochemical identifica- 564005, BD Bioscience) and IL-17 (1:100; 563354, BD Bioscience) anti-
tion tube (075310, HuanKai Microbial) and a gelatin zymography assay bodies. Flow cytometry was performed on an Attune NxT Acoustic
kit (RTD6143, Real-Times Biotechnology). MMP-2 and MMP-9 activities Focusing cytometer (ThermoFisher). Gating strategy for flow cytom-
were detected using the MMP Activity Assay kit (ab112146, Abcam). etry is shown in Supplementary Fig. 16a. Antibodies used are listed in
Supplementary Table 5.
Effects of MMP-2/9 protein on gut barrier markers in vitro
MMP-2 (HY-P70268, MCE) and MMP-9 (HY-P70145, MCE) were used THP-1-induced macrophage in vitro and co-culture experiment
at 0.01 μg ml−1 and co-cultured with NCM460 cells for 3 days. Then, THP-1 monocyte cells (TIB-202, ATCC) were cultured in RPMI 1640
western blot assay was conducted to evaluate the effects of MMP-2/9 medium (11875093, Gibco). THP-1 cells were differentiated into mac-
protein on gut barrier markers COLIV, fibronectin, E-cad and ZO-1. rophages by 3 days incubation with 150 nM phorbol ester (P8139,
Antibodies used are listed in Supplementary Table 5. Sigma), followed by morphological observation and flow cytometry
to evaluate CD14 (1:100; 367140, Biolegend) expression. THP-1-induced
FISH assay macrophages were then co-cultured with K. pneumoniae (multiplicity
Paraffin-embedded liver and colon sections were used for FISH assay of infection (MOI) = 10) for 2 h and subsequently cultured for another
using the Bacterial DNA FISH Assay kit (D-0015, FOCO Biology). Hybridi- 3 days. After that, the cell suspension was centrifuged at 10,000g for
zation was performed at 37 °C for 2 days with a universal 16S rRNA 10 min at 4 °C to collect the supernatant for gelatinase activity assay.
probe EUB338 (5′-GCTGCCTCCCGTAGGAGT-3′, with a Cy3 label), The gating strategy for flow cytometry is shown in Supplementary Fig.
E. coli probe (5′-GCATAAGCGTCGCTGCCG-3′, with a Cy3 label) or 16b. Antibodies used are listed in Supplementary Table 5.
K. pneumoniae probe (5′-CCTACACACCAGCGTGCC-3′, with a Cy3
label) at 0.25 μM in hybridization buffer. After washing three times Colon macrophage depletion mouse model
in wash buffer, slides were rinsed with sterile double-distilled water, We used the macrophage depletion reagent Clodrosome (CLD-8909,
air-dried in the dark and mounted in mounting medium with DAPI Encapsula NanoSciences) according to manufacturer instructions to
(ab104139, Abcam). Microscopic observations were performed using deplete colon macrophages. After killing, macrophages in colon were
a fluorescence microscope (BX63F, Olympus). detected by flow cytometry. Macrophages in colon were marked as
CD45+ MHC II+ CD11b+ Ly6C−. The gating strategy for flow cytometry
RT2 profiler qPCR array is shown in Supplementary Fig. 16c. Antibodies used are listed in Sup-
Total RNA was isolated from liver and colon tissues using TRIzol reagent plementary Table 5.
(15596026, Thermo Fisher) and reverse transcribed using RT2 First
Strand kit (330401, QIAGEN). RT2 Profiler PCR Array Mouse Cancer Live K. pneumoniae attachment to HCC cells assay
Pathway Finder (PAMM-033ZA-6, QIAGEN) or RT2 Profiler PCR Array Live K. pneumoniae (MOI = 10) was suspended in DMEM medium sup-
Mouse Inflammatory Response and Autoimmunity (PAMM-077Z, QIA- plemented with 10% FBS and co-cultured with Hep3B or Huh7 cells for
GEN) was used for qPCR analysis. 2 h in 6-well plates. Then, cells were washed with warm PBS six times
and scraped off from plates. The cells were collected and stripped in
HCC patient-derived organoid culture and experiment BHI agar for culture.
Organoids from patients with HCC (668T) were provided by N. Wong
(The Chinese University of Hong Kong), embedded into Matrigel SEM and TEM assay of K. pneumoniae and HCC cell attachment
(356231, Corning) and cultured in AIM-V medium (12055083, Thermo Hep3B or Huh7 cells (1 × 106) were seeded in sterile coverslips and
Fisher) supplemented with 10% FBS, 1% penicillin/streptomycin, co-cultured with K. pneumoniae (MOI = 10) for 2 h. Next, cells were
1% l-glutamine (25030081, Thermo Fisher), MEM non-essential washed with PBS three times and fixed with 2.5% glutaraldehyde
amino acids solution (11140050, Thermo Fisher) and insulin- fixative at 4 °C overnight. The fixed coverslips were washed with
transferrin-selenium (41400045, Thermo Fisher). Culture medium Sorensen’s phosphate buffer three times and fixed in 1% osmium
was changed every 3 days. After 10 days, HCC organoids were digested tetroxide for 2 h, followed by rinsing with double-distilled water.
using TrypLE Express (12604013, Thermo Fisher) and collected for After dehydration and coating with gold-palladium, specimens were
experiments. To investigate the effects of candidate bacteria or pro- examined using a Cold Field Scanning Electron Microscope (SU8010,
teins on organoid growth, images were captured and the diameter of Hitachi). Hep3B or Huh7 cells (1 × 106) were seeded in 6-well plates
organoids in each random field was measured using ImageJ (v.1.51, NIH) and co-cultured with K. pneumoniae for 4 h (MOI = 10), followed by
after treatment for 5 days. The effect of PBP1B (0.05 μM) on the growth fixation in 2.5% glutaraldehyde. Fixed samples were further fixed
of HCC patient-derived organoids was also investigated as stated above. in 2% osmium tetroxide and dehydrated with a gradient of ethanol
and propylene oxide. Samples were then embedded, sectioned and
Flow cytometry analysis of immune cells stained with 3% uranyl acetate and lead citrate. Pictures were taken
Multicolour flow cytometry was performed to study the immune cell using an HT7700 TEM (Hitachi).
types in liver and colon tissues. Single-cell suspensions of liver and
colon tissues were prepared with the Liver Dissociation kit (130-105- Cell membrane protein extraction
807, Miltenyi Biotec) or the Lamina Propria Dissociation kit (130-097- Hep3B or Huh7 cells were extracted with 0.5 ml ice-cold lysis buffer
410, Miltenyi Biotec), respectively. For surface staining, cells were (150 mM NaCl, 10 mM Tris-HCl (pH 7.4), 1 mM EDTA, 1 mM EGTA (pH
incubated with Fc block antibody purified rat anti-mouse CD16/CD32 8.0), 0.2 mM sodium orthovanadate, 0.2 mM phenylmethylsulfo-
(1:100) (553142, BD Biosciences) for 30 min. Then, cells were stained nylfluoride, 1% Triton X-100, 0.5% NP-40) for 1 h at 4 °C, followed by
with the following antibodies: CD45 (1:100; 157213, Biolegend), FVS510 sonication on ice for 15 min. The cell suspension was centrifuged at
(1:100; 564406, BD Biosciences), CD3 (1:100; 100330, Biolegend), CD4 10,000 g for 10 min at 4 °C to collect the supernatant, which was then
(1:100; 100434, Biolegend), CD11b (1:100; 101216, Biolegend), MHC II used as the cell membrane protein lysate.
Nature Microbiology
Biotin pull-down assay and silver staining regenerated completely with 10 mM glycine-HCl as injection buffer at a
Biotin pulldown was performed as previously described35. In flow rate of 30 μl min−1 for 30 s to remove the analyte, then the injection
brief, Hep3B or Huh7 cells were labelled with 1 mM EZ-Link and regeneration steps were repeated for the next concentration cycle
Sulfo-NHS-LC-Biotin (21338, Thermo Scientific) at 4 °C for 2 h, and of the analyte PBP1B. The Biacore T200 Evaluation Software v.2.0 was
then incubated with K. pneumoniae surface membrane proteins over- used to determine the binding affinity.
night at 4 °C, followed by incubation with streptavidin magnetic beads
for 6 h at 4 °C. As a negative control, streptavidin agarose resin was Molecular docking analysis
incubated with non-biotinylated Hep3B or Huh7 surface proteins. Molecular docking analysis was performed to determine recep-
Finally, the beads were washed with PBST (PBS with 0.1% Tween 20) tor–ligand interactions. The three-dimensional structures of PBP1B
three times, boiled with SDS–PAGE gel loading buffer and centrifuged. (accession: A0A086I8Y5) and TLR4 (O00206) were predicted using
Eluted proteins were subjected to SDS–PAGE and visualized by silver AlphaFold. Protein was pretreated using Discovery Studio software
staining using the Pierce Silver Stain for Mass Spectrometry kit (24600, to remove water molecules, hydrogenation and ions. The primary
Thermo Scientific). Corresponding bands of gels were identified by ligand was extracted from the structure and the treated protein
mass spectrometry. was visualized using Pymol. The 2D ligand–protein interaction dia-
grams and hydrophobic interactions were constructed using the
Expression and purification of PBP1B protein with or without software LigPlus.
GST tag
To generate recombinant GST–PBP1B protein, the non-transmembrane Alcian blue staining
domain of PBP1B (accession: A0A086I8Y5) was cloned into the pGEX- Paraffin-embedded colonic sections were stained using the Alcian
6P-1 vector and transformed into E. coli BL21 (DE3). Transformed Blue Stain kit (S0134, Bioss). Alcian blue staining-positive areas were
E. coli was grown in ampicillin-supplemented LB broth at 37 °C. analysed using ImageJ (v.1.51, NIH).
When optical density reached 1.0, 0.5 mM IPTG (HY-15921, MCE) was
added to induce protein synthesis at 20 °C overnight. Bacterial pel- TEM assay of colon and liver tissues
lets were collected and suspended in PBS-basic lysis buffer (50 mM Ultra-thin sections of colon and liver tissues were prepared and exam-
NaH2PO4, 150 mM NaCl, pH 7.2, 10 mg ml−1 lysozyme and proteinase ined using an HT7700 TEM (Hitachi), then photographed for analysis.
inhibitors). After centrifugation at 10,000 g for 30 min, GST–PBP1B
(113 kDa) was purified from the supernatant using GST-tag Protein K. pneumoniae quantification by qPCR
Purification kit (P2262, Beyotime). GST–PBP1B was incubated with Stool DNA was prepared using the Quick-DNA Faecal/Soil Microbe kit
PreScission protease (27084301, Cytiva) to obtain PBP1B without the (D6110, Zymo), while bacterial DNA from liver tissues was prepared
GST tag (87 kDa). using the QIAamp DNA Mini kit (51306, QIAGEN) combined with
RNase A (R6513, Sigma-Aldrich) and lysozyme (L6876, Sigma-Aldrich).
GST pull-down assay Quantification of K. pneumoniae was performed using the K. pneumo-
To perform the GST pulldown, Hep3B or Huh7 membrane proteins were niae-EASY Genesig kit (Z-Path-K. pneumoniae-standard, Primer Design)
incubated with or without recombinant GST–PBP1B protein overnight and Oasig lyophilised 2× qPCR Master Mix (Z-oasig-standard-150,
at 4 °C, followed by addition of glutathione magnetic agarose beads Primer Design) in an Applied Biosystems Real-Time PCR System (Quant-
(HY-K0234, MCE) for another 4 h at 4 °C. The beads were then washed Studio 5, Thermo Fisher).
and boiled with SDS–PAGE gel loading buffer. Eluted proteins were
separated by SDS–PAGE. Then, gel was analysed by silver staining, fol- Colony formation and cell proliferation
lowed by mass spectrometry analysis of target bands, or western blot KPCM was prepared by culture in BHI broth for 2 days, followed by
assay. Antibodies used are listed in Supplementary Table 5. bacterial removal by centrifugation and filtration with a 0.22-µm filter.
Pasteurized K. pneumoniae was prepared by incubating live K. pneumo-
Co-immunoprecipitation niae at 70 °C for 30 min. For colony formation, cells were seeded on
For co-immunoprecipitation of TLR4, anti-TLR4 antibody (1:100) and 6-well plates (1,000 cells per well). KPCM (5% or 10%) or pasteurized
Hep3B or Huh7 cell membrane protein were incubated with or without K. pneumoniae (MOI = 10) treatment was performed for 3 h day−1, then
recombinant GST-PBP1B overnight at 4 °C. Mouse IgG (1:100) served as the medium was replaced with MEM or DMEM containing 10% FBS, 1%
the negative control. Dynabeads Protein G magnetic beads (10004D, penicillin/streptomycin and 20 mg ml−1 gentamycin. The procedure
Invitrogen) were then added and incubated for another 4 h. The beads was repeated for 14 days and colonies were stained with 0.5% crystal
were washed and boiled with loading buffer. Boiled samples were violet. For the cell proliferation assay, cells were seeded into 96-well
subjected to SDS–PAGE and immunoblotted with anti-TLR4 (1:1,000) plates (1,000 cells per well), followed by live K. pneumoniae or culture
or anti-GST (1:5,000) (2624, Cell Signaling Technology). Antibodies supernatant treatment. At different time intervals, cell viability was
used are listed in Supplementary Table 5. determined using the MTT (475989, Sigma-Aldrich) assay and meas-
ured under an optical density of 570 nm using a spectrophotometer. In
SPR addition, PBP1B (0.05 μM) effects on HCC cell proliferation and colony
Binding affinity between PBP1B and TLR4 was evaluated by SPR using formation were also assessed as stated above.
Biacore T200 (GE Healthcare). TLR4 human recombinant protein
(HY-P73586, MCE) was immobilized on a Series S Sensor Chip CM5 Bacterial fluorescence tracing experiment
(29149603, GE Healthcare) using an amine coupling kit (BR-1000-50, GE To label K. pneumoniae and E. coli MG1655, bacteria were cultured in BHI
Healthcare) with a flow rate of 10 μl min−1, resulting in a final ligand cou- containing 100 μmol l−1 Cy5.5-d-Lys at 37 °C for 48 h. Bacterial suspen-
pling of ~180 response units. Different concentrations of PBP1B (100, sion was collected by centrifugation and washed with PBS six times.
50, 25, 12.5, 6.25, 3.125 and 0 nM) diluted using an analyte buffer were Labelling was determined using a fluorescence microscope (BX63F,
injected at a flow rate of 30 μl min−1 for an association phase of 120 s, Olympus). To trace bacteria in vivo, labelled bacteria were gavaged
followed by a 300-s dissociation. The association and dissociation to germ-free mice and fluorescence detection was performed before
processes were all handled in the analyte buffer. We repeated 7 cycles bacterial gavage and 8 h after gavage using the in vivo imaging system
of analysis according to analyte concentration in ascending order. (IVIS Lumina III, Perkin Elmer). Hepatic fluorescence detection was also
After each cycle of interaction analysis, the sensor chip surface was conducted after collecting the liver.
Nature Microbiology
TLR4i treatment in vitro and in vivo 3. Behary, J. et al. Gut microbiota impact on the peripheral
For in vitro treatment, the TLR4i TLR4-IN-C34 (HY-107575, MCE) was immune response in non-alcoholic fatty liver disease related
used at 30 μM, while mice received TLR4-IN-C34 (1 mg kg−1) gavaged hepatocellular carcinoma. Nat. Commun. 12, 187 (2021).
daily in vivo. 4. Nejman, D. et al. The human tumor microbiome is composed
of tumor type-specific intracellular bacteria. Science 368,
Colonization resistance challenge of K. pneumoniae in vivo 973–980 (2020).
K. oxytoca was used to cause colonization resistance to K. pneumoniae 5. Dejea, C. M. et al. Patients with familial adenomatous polyposis
in the gut of germ-free mice. Mice were gavaged with 1 × 108 c.f.u. of harbor colonic biofilms containing tumorigenic bacteria. Science
K. oxytoca diluted in 200 µl PBS. After 3 days of precolonization, 359, 592–597 (2018).
mice were gavaged with 1 × 108 c.f.u. of K. pneumoniae. This process 6. Iida, N. et al. Chronic liver disease enables gut Enterococcus
was conducted once a week. For mice receiving only K. oxytoca treat- faecalis colonization to promote liver carcinogenesis. Nat. Cancer
ment, mice were given 1 × 108 c.f.u. of K. oxytoca gavaged once a week. 2, 1039–1054 (2021).
7. Fu, A. et al. Tumor-resident intracellular microbiota promotes
Statistical analysis metastatic colonization in breast cancer. Cell 185, 1356–1372
No animals or data points were excluded from the analyses. Data col- (2022).
lection and analysis were not performed blind to the conditions of the 8. Dohlman, A. B. et al. A pan-cancer mycobiome analysis reveals
experiments. Data distribution was assumed to be normal but this was fungal involvement in gastrointestinal and lung tumors. Cell 185,
not formally tested. Measurements were taken from distinct samples. 3807–3822 (2022).
All in vitro experiments were repeated three times as three independ- 9. Pabst, O. et al. Gut–liver axis: barriers and functional circuits.
ent experiments unless otherwise stated. For in vitro experiments, all Nat. Rev. Gastroenterol. Hepatol. 20, 447–461 (2023).
attempts at replication were successful. For in vivo experiments, all 10. Schnabl, B. & Brenner, D. A. Interactions between the intestinal
experiments were successful without replication. Data are expressed microbiome and liver diseases. Gastroenterology 146, 1513–1524
as mean ± s.e.m. for both in vivo and in vitro experiments. Comparison (2014).
between two groups was performed using Student’s t-test. One-way 11. Manfredo Vieira, S. et al. Translocation of a gut pathobiont drives
ANOVA was used to compare differences among multiple groups. autoimmunity in mice and humans. Science 359, 1156–1161 (2018).
Differences in cell proliferation and MMP activity were determined 12. Dapito, D. H. et al. Promotion of hepatocellular carcinoma by the
using two-way ANOVA. Fisher’s exact test was used to evaluate the pro- intestinal microbiota and TLR4. Cancer Cell 21, 504–516 (2012).
portional difference in categorical variables between groups. Simple 13. Albillos, A., de Gottardi, A. & Rescigno, M. The gut-liver axis in
linear regression analysis was conducted to characterize the correla- liver disease: pathophysiological basis for therapy. J. Hepatol. 72,
tion between two variables. Correlation coefficient, r, was calculated 558–577 (2020).
using Goodness of Fit test. All statistical analyses were conducted 14. McCawley, L. J. & Matrisian, L. M. Matrix metalloproteinases:
using GraphPad Prism v.10.3.1. Flow cytometry data were analysed they’re not just for matrix anymore! Curr. Opin. Cell Biol. 13,
with the FlowJo v.10.9.0 software. P value < 0.05 indicates statistical 534–540 (2001).
significance. 15. Kirkegaard, T., Hansen, A., Bruun, E. & Brynskov, J. Expression
and localisation of matrix metalloproteinases and their natural
Reagent and resource sharing inhibitors in fistulae of patients with Crohn’s disease. Gut 53,
Further information and requests for resources, biological materials 701–709 (2004).
and reagents should be directed to and will be fulfilled by the lead 16. Tan, T. K. et al. Macrophage matrix metalloproteinase-9 mediates
contact, J.Y. epithelial-mesenchymal transition in vitro in murine renal tubular
cells. Am. J. Pathol. 176, 1256–1270 (2010).
Reporting summary 17. Bai, X. et al. Changes in MMP‑2, MMP‑9, inflammation, blood
Further information on research design is available in the Nature coagulation and intestinal mucosal permeability in patients with
Portfolio Reporting Summary linked to this article. active ulcerative colitis. Exp. Ther. Med. 20, 269–274 (2020).
18. Osbelt, L. et al. Klebsiella oxytoca causes colonization resistance
Data availability against multidrug-resistant K. pneumoniae in the gut via
All datasets generated and/or analysed during the study are avail- cooperative carbohydrate competition. Cell Host Microbe 29,
able. All related raw data have been deposited to public datasets with 1663–1679 (2021).
accession numbers. Whole-genome sequencing data of Klebsiella 19. Osbelt, L. et al. Klebsiella oxytoca inhibits Salmonella infection
pneumoniae reported in this study have been deposited at the Genome through multiple microbiota-context-dependent mechanisms.
Sequence Archive (GSA) of the China National Center for Bioinforma- Nat. Microbiol. 9, 1792–1811 (2024).
tion (CNCB) under accession number CRA019172. 16S rRNA sequencing 20. Okuda, K. Hepatocellular carcinoma: recent progress. Hepatology
data of liver tissues and stools from patients with LC and HCC have been 15, 948–963 (1992).
deposited at the GSA of CNCB under accession numbers CRA019234 21. Brescia, P. & Rescigno, M. The gut vascular barrier: a new player in
and CRA019232, respectively. 16S rRNA sequencing data and metagen- the gut-liver-brain axis. Trends Mol. Med. 27, 844–855 (2021).
omic sequencing sequencing data of mice faeces have been depos- 22. Zitta, K. et al. Serum from patients undergoing remote
ited at the GSA of CNCB under accession numbers CRA019231 and ischemic preconditioning protects cultured human intestinal
CRA019388, respectively. Source data are provided with this paper. cells from hypoxia-induced damage: involvement of
matrixmetalloproteinase-2 and-9. Mol. Med. 18, 29–37 (2012).
References 23. Jakubowska, K. et al. Expressions of matrix metalloproteinases
1. Sung, H. et al. Global cancer statistics 2020: GLOBOCAN (MMP-2, MMP-7, and MMP-9) and their inhibitors (TIMP-1, TIMP-2)
estimates of incidence and mortality worldwide for 36 cancers in in inflammatory bowel diseases. Gastroenterol. Res. Pract. 2016,
185 countries. CA Cancer J. Clin. 71, 209–249 (2021). 2456179 (2016).
2. Fujiwara, N., Friedman, S. L., Goossens, N. & Hoshida, Y. Risk 24. Rodrigues, D. M. et al. Matrix metalloproteinase 9 contributes to
factors and prevention of hepatocellular carcinoma in the era of gut microbe homeostasis in a model of infectious colitis. BMC
precision medicine. J. Hepatol. 68, 526–549 (2018). Microbiol. 12, 105 (2012).
Nature Microbiology
25. Spadoni, I. et al. A gut-vascular barrier controls the systemic X. Liu, S.Z. and X. Li conducted investigations. Q.Z., L.X., H.S., H.W.,
dissemination of bacteria. Science 350, 830–834 (2015). X. Li and M.K. acquired resources. X.W., L.S., M.F., J.W., N.W., T.Y., M.M.
26. Ren, Z. et al. Gut microbiome analysis as a tool towards targeted and B.L. performed visualization. L.X. and J.Y. acquired funding.
non-invasive biomarkers for early hepatocellular carcinoma. Gut J.Y. administered and supervised the project. X.W., C.C.W. and J.Y.
68, 1014–1023 (2019). wrote the original draft. X.W., C.C.W. and J.Y. reviewed and edited
27. Nakamoto, N. et al. Gut pathobionts underlie intestinal barrier the manuscript.
dysfunction and liver T helper 17 cell immune response in primary
sclerosing cholangitis. Nat. Microbiol. 4, 492–503 (2019). Competing interests
28. Yuan, J. et al. Fatty liver disease caused by The authors declare no competing interests.
high-alcohol-producing Klebsiella pneumoniae. Cell Metab. 30,
675–688 (2019). Additional information
29. Atarashi, K. et al. Ectopic colonization of oral bacteria in the Extended data is available for this paper at
intestine drives TH1 cell induction and inflammation. Science 358, https://doi.org/10.1038/s41564-024-01890-9.
359–365 (2017).
30. Ganaway, J. R. in The Biology of the Guinea Pig (eds Wagner. J. E. & Supplementary information The online version
Manning, P. J.) 121–135 (Elsevier, 1976). contains supplementary material available at
31. Wagner, K. A., Hartmann, F. A. & Trepanier, L. A. Bacterial culture https://doi.org/10.1038/s41564-024-01890-9.
results from liver, gallbladder, or bile in 248 dogs and cats
evaluated for hepatobiliary disease: 1998–2003. J. Vet. Intern. Correspondence and requests for materials should be addressed to
Med. 21, 417–424 (2007). Jun Yu.
32. Lederman, E. R. & Crum, N. F. Pyogenic liver abscess with a focus
on Klebsiella pneumoniae as a primary pathogen: an emerging Peer review information Nature Microbiology thanks Tim Greten and
disease with unique clinical characteristics. Am. J. Gastroenterol. the other, anonymous, reviewer(s) for their contribution to the peer
100, 322–331 (2005). review of this work.
33. Niño, J. L. G. et al. Effect of the intratumoral microbiota on spatial
and cellular heterogeneity in cancer. Nature 611, 810–817 (2022). Reprints and permissions information is available at
34. Federici, S. et al. Targeted suppression of human IBD-associated www.nature.com/reprints.
gut microbiota commensals by phage consortia for treatment of
intestinal inflammation. Cell 185, 2879–2898 (2022). Publisher’s note Springer Nature remains neutral with regard to
35. Long, X. et al. Peptostreptococcus anaerobius promotes jurisdictional claims in published maps and institutional affiliations.
colorectal carcinogenesis and modulates tumour immunity.
Nat. Microbiol. 4, 2319–2330 (2019). Open Access This article is licensed under a Creative Commons At
tribution-NonCommercial-NoDerivatives 4.0 International License,
Acknowledgements which permits any non-commercial use, sharing, distribution and
This project was funded by the National Science and Technology reproduction in any medium or format, as long as you give appropriate
Major Project of China (2023ZD0500200, J.Y.), the GuangDong Basic credit to the original author(s) and the source, provide a link to the
and Applied Basic Research Foundation (2022B1515120031, J.Y.), Creative Commons licence, and indicate if you modified the licensed
RGC-CRF Hong Kong (C4039-19GF and C4008-23W, J.Y.), the Strategic material. You do not have permission under this licence to share
Seed Funding Collaboration Research Scheme CUHK (3133344, J.Y.), adapted material derived from this article or parts of it. The images
the Strategic Impact Enhancement Fund CUHK (3135509, J.Y.), the or other third party material in this article are included in the article’s
Impact case for RAE CUHK (3134277, J.Y.), the National Natural Science Creative Commons licence, unless indicated otherwise in a credit
Foundation of China (82173191, L.X.), and the Guangdong Basic and line to the material. If material is not included in the article’s Creative
Applied Basic Research Foundation (2019B151502009, L.X.). Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain
Author contributions permission directly from the copyright holder. To view a copy of this
X.W. and J.Y. conceptualized the project. X.W., H.S., H.W., X. Li and licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
J.J.Y.S. developed the methodology. X.W., Y.F., W. Liang, Y.C., J.W., N.W.,
L.S., H.C.-H.L., Y.J., X.Z., L.Y., M.M., T.Y., M.F., H.Z., E.S.-H.C., M.L., W. Liu, © The Author(s) 2025
1
Department of Liver Surgery, Center of Hepato-Pancreato-Biliary Surgery, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China.
2
Institute of Digestive Disease and The Department of Medicine and Therapeutics, State Key Laboratory of Digestive Disease, Li Ka Shing Institute of
Health Sciences, CUHK Shenzhen Research Institute, The Chinese University of Hong Kong, Hong Kong, China. 3Institute of Precision Medicine, The First
Affiliated Hospital, Sun Yat-sen University, Guangzhou, China. 4Organ Transplant Center, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou,
China. 5Department of Oncology, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China. 6Department of Pathology, The First Affiliated
Hospital, Sun Yat-sen University, Guangzhou, China. 7Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore.
8
These authors contributed equally: Xueliang Wang, Yi Fang, Wei Liang, Yuhong Cai. e-mail: [email protected]
Nature Microbiology
Extended Data Fig. 1 | Mixed HCC fecal microbiota transplantation and fecal DEN+IHD group n=12; DEN+IHCC group n=9) and related results (f-h).
microbiota from two individual HCC patients promote hepatocarcinogenesis (f) Representative images of liver gross morphology (yellow circles indicate
in DEN-treated germ-free mice. (a) Design of mixed FMT experiment to DEN- tumors) and H&E staining. Tumor incidence, tumor number, and maximum
treated germ-free mice (DEN+PBS group n=6; DEN+HD group n=12; DEN+HCC tumor diameter per mouse in liver. n=9 (PBS), 10 (DEN+PBS), 12 (DEN+IHD),
group n=11) and related results (b-d). (b) Representative images of liver gross 9 (DEN+IHCC). (g) Ki-67 and PCNA IHC staining. n=6 biologically independent
morphology (yellow circles indicate tumors) and H&E staining of mouse liver samples. (h) α-SMA and Desmin IHC staining, and Sirius red staining of mouse
sections (yellow circles indicate tumor areas). Tumor incidence, tumor number liver sections. n=6 biologically independent samples. Data (excluding tumor
and the maximum tumor diameter per mice. n=6 (DEN+PBS), 12 (DEN+HD), 11 incidence) are presented as mean ± SEM. Each data point in bar plots represents
(DEN+HCC). (c) IHC staining of Ki-67 and PCNA. n=6 biologically independent one mouse. Tumor incidence was calculated using Fisher’s exact test. Unless
samples. (d) Desmin and α-SMA IHC staining, and Sirius red staining of liver otherwise stated, statistical significance was calculated using one-way ANOVA.
sections. n=6 biologically independent samples. (e) Design of FMT experiment Adjustments were made for multiple comparisons.
to germ-free mice with DEN treatment (PBS group n=9; DEN+PBS group n=10;
Nature Microbiology
Extended Data Fig. 2 | Individual HCC patient fecal microbiota initiates liver independent samples. (e) Desmin and α-SMA IHC staining, and Sirius red staining
precancerous lesions in germ-free mice without DEN treatment. (a) Design of of liver sections. n=6 biologically independent samples. (f) Mouse Cancer
FMT experiment to germ-free mice without DEN treatment (GF/PBS group n=6; Pathway Finder Array and (g) Inflammatory Response and Autoimmunity PCR
GF/HD-FMT group n=12; GF/HCC-FMT group n=16). (b) Representative images Array of liver tissues. f and g, n=3 independent experiments with similar results.
of liver gross morphology (yellow circles indicate nodules), and H&E staining of Data (excluding incidence of dysplasia and PCR array results) are presented
mice liver sections with incidence of dysplasia and nodules number statistical as mean ± SEM. Each data point in bar plots represents one mouse. Incidence
results. n=6 (GF/PBS), 12 (GF/HD-FMT), 16 (GF/HCC-FMT). (c) IHC staining for of dysplasia was calculated using Fisher’s exact test. Unless otherwise stated,
PCNA and Ki-67. n=6 biologically independent samples. (d) Hepatic infiltration statistical significance was calculated using one-way ANOVA. Adjustments were
of Th1, Th17, and Th2 cells was evaluated by flow cytometry. n=3 biologically made for multiple comparisons.
Nature Microbiology
Extended Data Fig. 3 | Individual HCC patient fecal microbiota initiates liver (e) Hepatic infiltration of Th1 (n=3 biologically independent samples), Th17
precancerous lesions in SPF mice without DEN treatment. (a) Design of FMT (n=4 biologically independent samples), and Th2 (n=3 biologically independent
experiment to SPF mice without DEN treatment (SPF/PBS group n=19; SPF/HD- samples) cells was evaluated by flow cytometry. (f) Mouse Cancer Pathway Finder
FMT group n=16; SPF/HCC-FMT group n=16). (b) Representative images of liver Array and Inflammatory Response and Autoimmunity PCR Array of liver tissues.
gross morphology (yellow circles indicate nodules), and H&E staining of mice n=3 independent experiments with similar results. Data (excluding incidence of
liver sections. Incidence of dysplasia and number of liver nodules. n=19 (SPF/ dysplasia and PCR array results) are presented as mean ± SEM. Each data point
PBS), 16 (SPF/HD-FMT), 16 (SPF/HCC-FMT). (c) IHC staining for PCNA and Ki-67. in bar plots represents one mouse. Incidence of dysplasia was calculated using
n=6 biologically independent samples. (d) Desmin and α-SMA IHC staining, and Fisher’s exact test. Unless otherwise stated, statistical significance was calculated
Sirius red staining of liver sections. n=6 biologically independent samples. using one-way ANOVA. Adjustments were made for multiple comparisons.
Nature Microbiology
Extended Data Fig. 4 | Five mixed HCC fecal microbiota transplantation Th17, and Th2 cells infiltration in liver was determined by flow cytometry. n=3
initiates hepatocarcinogenesis in germ-free mice without DEN treatment. biologically independent samples. (f) α-SMA and Desmin IHC staining, and
(a) Design of FMT experiment to germ-free mice (PBS group n=10; HD-FMT Sirius red staining of liver sections. n=6 biologically independent samples. Data
group n=12; HCC-FMT group n=9). (b) Representative images of liver gross (excluding incidence of dysplasia and PCR array results) are presented as mean
morphology (yellow circles indicate nodules) and H&E staining of mice liver ± SEM. Each data point in bar plots represents one mouse. Incidence of dysplasia
sections. n=10 (PBS), 12 (HD-FMT), 9 (HCC-FMT). (c) IHC staining for Ki-67 was calculated using Fisher’s exact test. Unless otherwise stated, statistical
and PCNA. n=6 biologically independent samples. (d) Mouse Cancer Pathway significance was calculated using one-way ANOVA. Adjustments were made for
Finder PCR Array and Inflammatory Response and Autoimmunity PCR Array multiple comparisons.
of liver tissues. n=3 independent experiments with similar results. (e) Th1,
Nature Microbiology
Extended Data Fig. 5 | See next page for caption.
Nature Microbiology
Extended Data Fig. 5 | HCC patient fecal microbiota impairs gut barrier and Th2 cells in germ-free mice without DEN was evaluated by flow cytometry.
function, induces intestinal inflammation, and promotes bacteria n=3 biologically independent samples. (e) Workflow of fresh liver homogenate
translocation into liver in germ-free mice with or without DEN treatment. preparation for bacteria culture. (f) PBS contamination control used in tissues
(a) Gut permeability assays using 500 kDa FITC-dextran, CLDN 3 and CLDN 1 IHC bacteria culture experiment. (g) Cy3-conjugated EUB338 probe FISH detection
staining of germ-free mice without DEN. n=6 biologically independent samples. of liver tissues. n=6 biologically independent samples. Data (excluding PCR array
(b) Gut permeability assays using 500 kDa FITC-dextran, CLDN 3 and CLDN 1 IHC result) are presented as mean ± SEM. Each data point in bar plots represents
staining of germ-free mice with DEN. n=6 biologically independent samples. one mouse. Statistical significance was calculated using one-way ANOVA.
(c) Mouse Inflammatory Response and Autoimmunity PCR Array of colon Adjustments were made for multiple comparisons.
tissues. n=3 independent experiments with similar results. (d) Colon Th1, Th17,
Nature Microbiology
Extended Data Fig. 6 | HCC patient fecal microbiota impairs gut barrier 20 (SPFD/PBS), 16 (SPFD/HD-FMT), 16 (SPFD/LC-FMT), 16 (SPFD/HCC-FMT).
function and promotes bacteria translocation into liver in SPF mice with or From left to right and top to bottom, the culture plates are blood agar plate,
without DEN treatment as shown in Fig. 1f and Extended Data Fig. 3a. (a) Gut chocolate blood agar plate, MacConkey agar plate, and Columbia blood agar
permeability assay using 500 kDa FITC-dextran. n=6 biologically independent plate, respectively. And, representative images of Cy3-conjugated EUB338 probe
samples. (b) Representative pictures of Alcian blue staining, E-Cad, CLDN 3 FISH detection in mice liver tissues. n=6 biologically independent samples. Data
IHC staining, and Cy3-conjugated EUB338 probe FISH of colon tissues with (excluding liver with live bacteria) are presented as mean ± SEM. Each data point
quantitative analysis. n=3 biologically independent samples. (c) Representative in bar plots represents one mouse. Liver with live bacteria was calculated using
bacterial culture of liver tissues under anaerobic and aerobic conditions with Fisher’s exact test. Unless otherwise stated, statistical significance was calculated
quantitative analysis. n= 19 (SPF/PBS), 16 (SPF/HD-FMT), 16 (SPF/HCC-FMT), using one-way ANOVA. Adjustments were made for multiple comparisons.
Nature Microbiology
Nature Microbiology
Extended Data Fig. 7 | Mixed HCC fecal microbiota transplantation and fecal 12 (HD-FMT), 9 (HCC-FMT). From left to right and top to bottom, the culture
microbiota from two individual HCC patients impair gut barrier function and plates are blood agar plate, chocolate blood agar plate, MacConkey agar plate,
promotes bacteria translocation into liver in germ-free mice with or without and Columbia blood agar plate, respectively. (g) Gut permeability assay using
DEN treatment. (a) Gut permeability assay using FITC-dextran. n=6 (DEN+PBS), FITC-dextran and (h) Cy3-conjugated EUB338 probe detection of liver tissue.
12 (DEN+HD), 11 (DEN+HCC). (b) Representative pictures of Alcian blue staining, n=6 biologically independent samples. e-h are related to Extended Data Fig. 4a.
CLDN 3 and E-Cad IHC staining of colon tissues with quantitative analysis. n=6 (i) Pictures of Alcian blue staining, CLDN 3 and E-Cad IHC staining of colon
biologically independent samples. (c) Representative bacterial culture of liver tissues with quantitative analysis. n=6 biologically independent samples. ( j) Gut
tissues under anaerobic and aerobic conditions with quantitative analysis. n=6 permeability assay using FITC-dextran. n=9 (PBS), 10 (DEN+PBS), 12 (DEN+IHD),
(DEN+PBS), 12 (DEN+HD), 11 (DEN+HCC). From left to right and top to bottom, 9 (DEN+IHCC). (k) Bacterial culture of liver tissues under anaerobic and aerobic
the culture plates are blood agar plate, chocolate blood agar plate, MacConkey conditions with quantitative analysis. n=9 (PBS), 10 (DEN+PBS), 12 (DEN+IHD),
agar plate, and Columbia blood agar plate, respectively. (d) Images of 9 (DEN+IHCC). i-k are related to Extended Data Fig. 1e. Data (excluding liver
Cy3-conjugated EUB338 probe detection. n=6 biologically independent samples. with live bacteria) are presented as mean ± SEM. Each data point in bar plots
a–d are related to Extended Data Fig. 1a. (e) Pictures of CLDN 3 and E-Cad IHC represents one mouse. Liver with living bacteria was calculated using Fisher’s
staining, and Alcian blue staining of colon tissues with quantitative analysis. exact test. Unless otherwise stated, statistical significance was calculated using
n=6 biologically independent samples. (f) Bacterial culture of liver tissues one-way ANOVA. Adjustments were made for multiple comparisons.
under anaerobic and aerobic conditions with quantitative analysis. n=10 (PBS),
Nature Microbiology
Extended Data Fig. 8 | K. pneumoniae initiates gut barrier dysfunction with quantitative analysis related to Fig. 4h. (e) Representative pictures
through elevating colon macrophage mediated MMP-2/-9 activities and of TEM and E-Cad IHC staining of liver tissues related to Fig. 5c. Masson’s
promotes precancerous lesions in germ-free mice. (a) Representative trichrome staining and COL IV IHC staining of colon tissues related to Fig. 5e.
pictures of TEM, Alcian blue staining, E-Cad, CLDN 3, CLDN 1 IHC staining, (f) Representative pictures of PCNA and Ki-67 IHC staining related to Fig. 5g.
and Cy3-conjugated EUB338 probe FISH of colon tissues related to Fig. 4c. (g) Representative pictures of Desmin and α-SMA IHC staining, and Sirius red
(b) Macrophage abundance in colon was evaluated by flow cytometry related to staining of liver sections related to Fig. 5h. (h) Mouse Cancer Pathway Finder
Fig. 4e. n = 6 biologically independent samples. (c) Cy3-conjugated conjugated Array and Mouse Inflammatory Response and Autoimmunity PCR Array of liver
EUB338 probe (PBS group), K. pneumoniae probe (KP and KPC groups) FISH tissues related to Fig. 5a. n = 3 independent experiments with similar results.
detection in liver tissue related to Fig. 4f. n = 6 biologically independent Data (excluding PCR array results) are presented as mean ± SEM. Each data point
samples. (d) Representative pictures of TEM, Alcian blue staining, E-Cad, CLDN 1 in bar plots represents one mouse. Statistical significance was calculated using
IHC staining, and Cy3-conjugated conjugated EUB338 probe (PBS group), one-way ANOVA. Adjustments were made for multiple comparisons.
K. pneumoniae probe (KP and KPC groups) FISH detection of colon tissues
Nature Microbiology
Nature Microbiology
Extended Data Fig. 9 | Antagonizing K. pneumoniae colonic colonization or tumor incidence, tumor number and tumor burden. n=6 (GFD/PBS), 6 (GFD/
TLR4 inhibition reverse K. pneumoniae induced-HCC development. (a) Design EC), 7 (GFD/KP), 7 (GFD/KPi). (h) E-Cad IHC and Cy3-conjugated EUB338 probe
of K. oxytoca on K. pneumoniae induced HCC promotion in germ-free mice with (GFD/PBS group), E. coli probe (GFD/EC), K. pneumoniae probe (GFD/KP) and
DEN treatment (GFD/PBS group n=6; GFD/EC group n=6; GFD/KO group n=6; GFD/KPi groups) detection of colon tissues with quantitative analysis. n=6
GFD/KP group n=7; GFD/KOP group n=8) and relevant results (b–e). (b) Images biologically independent samples. (i) Gut permeability assay using FITC-dextran.
of liver gross morphology (yellow circles indicate tumor area), and H&E with n=6 biologically independent samples. ( j) Live bacteria culture of liver tissues
tumor incidence, tumor number and tumor burden. n=6 (GFD/PBS), 6 (GFD/ on blood agar plates with quantitative analysis. n=6 (GFD/PBS), 6 (GFD/EC), 7
EC), 6 (GFD/KO) 7 (GFD/KP), 8 (GFD/KOP). (c) Alcian blue staining, E-Cad, CLDN (GFD/KP), 7 (GFD/KPi). (k) Cy3-conjugated conjugated EUB338 probe (GFD/
3, CLDN 1 IHC, and Cy3-conjugated EUB338 probe detection with quantitative PBS group), E. coli probe (GFD/EC), K. pneumoniae probe (GFD/KP) and GFD/KPi
analysis. n=6 biologically independent samples. (d) Gut permeability assay groups) FISH, TLR4, PCNA, Ki-67 IHC, and Sirius red staining of liver sections.
using FITC-dextran. n=6 (GFD/PBS), 6 (GFD/EC), 6 (GFD/KO) 7 (GFD/KP), 8 n=6 biologically independent samples. Data (excluding tumor incidence and
(GFD/KOP). (e) Live bacteria culture of liver tissues on blood agar plates with liver with live bacteria) are presented as mean ± SEM. Each data point in bar
quantitative analysis. n=6 (GFD/PBS), 6 (GFD/EC), 6 (GFD/KO) 7 (GFD/KP), 8 plots represents one mouse. Tumor incidence and Liver with live bacteria
(GFD/KOP). (f) Design of TLR4i on K. pneumoniae induced HCC promotion in were calculated using Fisher’s exact test. Unless otherwise stated, statistical
germ-free mice with DEN treatment (GFD/PBS group n=6; GFD/EC group n=6; significance was calculated using one-way ANOVA. Adjustments were made for
GFD/KP group n=7; GFD/KPi group n=7) and relevant results (g-k). (g) Images multiple comparisons.
of liver gross morphology (yellow circles indicate tumor area), and H&E with
Nature Microbiology
Nature Microbiology
Extended Data Fig. 10 | TLR4 inhibition reverses HCC-FMT induced-HCC and Cy3-conjugated EUB338 probe FISH detection of colon tissues with
development. (a) Design of TLR4i on HCC-FMT induced HCC promotion in quantitative analysis. n=6 biologically independent samples. (e) Cy3-conjugated
germ-free mice with DEN treatment (GFD/PBS group n=8; GFD/HCC-FMT group EUB338 probe FISH, Cy3-conjugated K. pneumoniae probe FISH, TLR4, PCNA,
n=15; GFD/HCC-FMTi group n=10) and relevant results (b–g). TLR4i was daily Ki-67 IHC of liver sections. n=6 biologically independent samples. (f) α-SMA,
gavaged in this experiment. (b) Representative images of liver gross morphology Desmin IHC staining, and Sirius red staining of liver sections. n=6 biologically
(yellow circles indicate tumor area), and H&E staining of mice liver sections independent samples. (g) Mouse Cancer Pathway Finder Array and Inflammatory
with tumor incidence, tumor number, and tumor burden. n=8 (GFD/PBS), 15 Response and Autoimmunity PCR Array of liver tissues. n=3 independent
(GFD/HCC-FMT), 10 (GFD/HCC-FMTi). (c) Representative bacterial culture of experiments with similar results. Data (excluding tumor incidence, liver with live
liver tissues under anaerobic and aerobic conditions with quantitative analysis. bacteria and PCR array results) are presented as mean ± SEM. Each data point in
n=8 (GFD/PBS), 15 (GFD/HCC-FMT), 10 (GFD/HCC-FMTi). From left to right and bar plots represents one mouse. Tumor incidence and Liver with live bacteria
top to bottom, the culture plates are blood agar plate, chocolate blood agar were calculated using Fisher’s exact test. Unless otherwise stated, statistical
plate, MacConkey agar plate, and Columbia blood agar plate, respectively. Gut significance was calculated using one-way ANOVA. Adjustments were made for
permeability assay using 500 kDa FITC-dextran. n=6 biologically independent multiple comparisons.
samples. (d) Representative pictures of Alcian blue staining, E-Cad IHC staining,
Nature Microbiology
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
[email protected]

Gut-liver Translocation of Pathogen Klebsiella Pne

Uploaded by

Copyright:

Available Formats

Gut-liver Translocation of Pathogen Klebsiella Pne

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gut-liver Translocation of Pathogen Klebsiella Pne

Uploaded by

Copyright:

Available Formats

nature microbiology

Gut–liver translocation of pathogen

Hepatocellular carcinoma (HCC) is accompanied by an altered gut

GFD/LC-FMT DEN + liver cirrhosis patient FMT (n = 17)

0.0014 0.0007 0.0084 0.0012

GFD/PBS GFD/HD-FMT GFD/LCC-FMT GFD/HCC-FMT f 14-day SPF male BALB/c DEN

GFD/LC-FMT DEN + healthy donor FMT (n = 16)

40 3 SPFD/LC-FMT DEN + liver cirrhosis patient FMT (n = 16)

g SPFD/PBS SPFD/HD-FMT SPFD/LC-FMT SPFD/HCC-FMT

<0.0001 0.0026 <0.0001

Inflammatory response and autoimmunity: Inflammatory response and autoimmunity:

SPFD/PBS SPFD/LC-FMT SPFD/HCC-FMT SPFD/HD-FMT 20

0.0003 0.0063 SPFD/HD-FMT

number per slide

Germ-free without DEN

number per slide

<0.0001 < 0.0001

b Germ-free with DEN Germ-free without DEN

c.f.u. mg−1 tissue

Liver with living

GFD/PBS GFD/LC-FMT GF/PBS

Bacteria in hepatic Bacteria in parenchymal Bacteria in parenchymal

d Germ-free with DEN Germ-free without DEN

<0.0001 0.0091 4 0.0003 0.0149

a Live bacteria in LC liver tissues : 12 out of 15

b Contamination Live bacteria in HCC liver tissues: 28 out of 54

Enterococcus faecalis 100 Unclassified 6

disappear rate (%)

Ctrl 5% 10% 5% 10% 3.4

Liver with live

MMP-2 MMP-9 80 <0.0001 0.001 0.4

MMP activity (RFU)

80 PBS 80 PBS 0.001 0.007

a b Live KP translocated from gut into liver c

disappear rate (%)

Gavage once a week

GF/KP Klebsiella pneumoniae (n = 10)

PC PBS EC KP GF/PBS COL IV relative COL IV optical

PBS EC KP GF/PBS COL IV

<0.0001 <0.0001 <0.0001

500 kDa FITC-

42 kDa 0.2 0.2 4 Gavage once a week

4 0.0007 EC Escherichia coli (n = 6)

2 GF/PBS GF/EC GF/KP

l PBS EC KP <0.0001 200 µm 250 µm

4 0.0233 <0.0001 PBS

o 14-day SPF male C57BL/6J p DPBS DEC DKP

DPBS DEN + PBS (n = 7) 50 DKP

DEC DEN + Escherichia coli (n = 6)

Hep 3B + KP Huh 7 + KP 100 100

180 kDa 180 kDa

+ + + + + Anti-GST beads 25 nM IP:GST Input GST-PBP1B + – +

1.0 1.8 Hep 3B Huh 7 Hep 3B Huh 7 In germ-free mice

Ctrl Ctrl 2 2 without DEN

PBP1B PBP1B 0.0009 0.019

Hep 3B Hep 3B Huh 7 50 µm

PBP1B + TLR4i Ctrl TLR4i Ctrl TLR4i PBS BT PM

axis (Supplementary Fig. 17). Targeting K. pneumoniae represents a Mice

Extended Data Fig. 5 | See next page for caption.

Extended Data Fig. 7 | See next page for caption.