Fusobacterium nucleatum promotes colon cancer progression by changing the mucosal microbiota and colon transcriptome in a mouse model

Na Wu,Yu-Qing Feng, Na Lyu, Di Wang, Wei-Dong Yu, Yong-Fei Hu

Abstract

BACKGROUND Fusobacterium nucleatum (F. nucleatum) has long been known to cause opportunistic infections and has recently been implicated in colorectal cancer (CRC),which has attracted broad attention. However, the mechanism by which it is involved in CRC development is not fully understood.

AIM To explore its potential causative role in CRC development, we evaluated the colon pathology, mucosa barrier, colon microbiota and host transcriptome profile after F. nucleatum infection in an azoxymethane/dextran sulfate sodium salt(AOM/DSS) mouse model.

METHODS Three groups of mice were compared to reveal the differences, i.e., the control,AOM/DSS-induced CRC and AOM/DSS-FUSO infection groups.

RESULTS Both the AOM/DSS and AOM/DSS-FUSO groups exhibited a significantly reduced body weight and increased tumor numbers than the control group, and AOM/DSS mice with F. nucleatum infection showed the highest tumor formation ratio among the three groups. Moreover, the colon pathology was the most serious in the AOM/DSS-FUSO group. We found that the structure of the colon microbiota changed considerably after F. nucleatum infection; striking differences in mucosal microbial population patterns were observed between the AOM/DSS-FUSO and AOM/DSS groups, and inflammation-inducing bacteria were enriched in the mucosal microbiota in the AOM/DSS-FUSO group. By comparing intestinal transcriptomics data from AOM vs AOM/DSSFUSO mice, we showed that transcriptional activity was strongly affected by dysbiosis of the gut microbiota. The most microbiota-sensitive genes were oncogenes in the intestine, and the cyclic adenosine monophosphate signaling pathway, neuroactive ligand–receptor interaction, PPAR signaling pathway, retinol metabolism, mineral absorption and drug metabolism were highly enriched in the AOM/DSS-FUSO group. Additionally, we showed that microbial dysbiosis driven by F. nucleatum infection enriched eight taxa belonging to Proteobacteria, which correlates with increased expression of oncogenic genes.

CONCLUSION Our study demonstrated that F. nucleatum infection altered the colon mucosal microbiota by enriching pathogens related to the development of CRC, providing new insights into the role of F.nucleatum in the oncogenic microbial environment of the colon.

Key Words: Fusobacterium nucleatum; Mucosal microbiota; Transcriptome; Colorectal cancer;Inflammation-inducing bacteria

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer and the fourth leading cause of cancer death worldwide[1 ]. The incidence of CRC in China has increased dramatically during the past decade[2 ].

Fusobacterium nucleatum(F. nucleatum), a gram-negative anaerobe, was recently identified as an emerging pathogen that has been frequently found in CRC[3 -5 ]. Ιmportantly, overload ofF. nucleatumwas found to be associated with an advanced tumor stage and a poor prognosis in CRC[6 ]. Accumulating evidence has revealed thatF. nucleatuminfluences different stages of CRC progression.F.nucleatumcan promote the initiation and development of CRC by adhering to epithelial cells through its unique FadA protein[7 ], which can bind vascular endothelial cadherin and alters endothelial integrity[8 ]. The activation of oncogenic miR21 underF. nucleatumstimulation results in the overexpression of TLR-4 /nuclear factor-kappaB, increasing the proliferation of CRC cells and promoting tumor development[9 ]. Once a colon tumor has developed,F. nucleatumcan adhere to Gal-GalNAc-expressing tumor cells by binding to the Fap2 Lectin, leading to enrichment of F. nucleatum[10 ].F. nucleatuminfection was also proven to increase the expression levels of inflammatory genes, transcription factors,and oncogenes[8 ,11 ]. These studies enhance our knowledge of the mechanisms by whichF. nucleatumshapes the tumor microenvironment.

Evidences have revealed that the local microbiota is also an important part of the tumor microenvironment. Gut microbiota dysbiosis can reshape the tumor microenvironment and make it favorable for tumor growth[12 ]. The mechanisms involved include increasing mutagenesis, regulation of oncogenic pathways, and modulation of host immune system[13 ]. These oncogenic functions of gut microbiota are probably due to the microbial products or metabolites they produce when the community structure balance is disrupted and a resulting overgrowth of specific bacteria. For example, the overload of colibactin-producingEscherichia coliand other Enterobacteriaceae can induce DNA damage of host cells and exert tumorigenic effect[14 ]. These facts suggest a close relationship between gut microbiota, the tumor microenvironment, and CRC. However, the regulatory role ofF. nucleatumin the “oncogenic microbial environment” of the colon remains mostly unknown. Revealing the effects ofF. nucleatuminfection on the changes of mucosal microbiota will help to provide further insight into the mechanisms of the tumorigenic effect ofF. nucleatum.

Ιn the present study, we explored the role ofF. nucleatumin the development of CRC and in altering the mucosal microbial community structure and investigated the association between microbial dysbiosis and the expression of oncogenic signaling pathways.F. nucleatumstrain ATCC 25586 was gavaged into azoxymethane/dextran sulfate sodium salt (AOM/DSS) mice to explore the role ofFusobacteriumoverload in the progression of intestinal adenoma against a background of environmental inflammatory factors. RNA-seq transcriptomic analysis was used to identify the differentially expressed gene (DEG) signatures driven byF. nucleatuminfection, and integrated network analysis was conducted to reveal the relationship between microbial dysbiosis and the DEGs. Our study lays a foundation for understanding the regulatory role ofF. nucleatumin the oncogenic microbial environment of the colon.

MATERIALS AND METHODS

Bacterial strain and culture conditions

TheF. nucleatumstrain ATCC 25586 was purchased from American Type Culture Collection (ATCC)and was cultured as described previously[3 ]. Ιn brief,F. nucleatumwas grown in Columbia blood agar supplemented with 5 mg/mL haemin, 5 % defibrinated sheep blood, and 1 mg/mL Vitamin K1 in an anaerobic condition with 85 % N2 , 10 % H2 , and 5 % CO2 at 37 °C. And 109 colony-forming unitspermL(CFU/mL) were prepared forF. nucleatumcolonization.

Animal model

Thirty-six conventional C57 BL/6 J wild-type male mice (4 wk of age) were obtained from the Beijing Animal Study Centre and maintained in the Laboratory Animal Services Centre at the Chinese Center for Disease Control and Prevention. The Ιnstitutional Animal Care Committee of the Peking University People’s Hospital approved all animal experimentation.

To create the mouse model of colon neoplasia induced by AOM/DSS-induced colon neoplasia[15 ], 24 mice at 6 wk of age were injected intraperitoneally with a single dose of 10 mg/kg AOM (Sigma),designated week 0 . DSS (MP Biochemicals; 36000 -50000 mw) was dissolved in filtered drinking water and administered to the mice for 7 d during the following weeks: Weeks 1 (1 % DSS), 3 (1 % DSS), and 5 (1 % DSS). After 1 wk of AOM/DSS treatment, the 24 mice were divided into 2 groups, each with 12 mice. At week 7 they were orally gavaged daily with 109 CFU/mL (in 100 μL volume for each mouse)F.nucleatum(AOM/DSS-FUSO group) or phosphate-buffered saline (AOM/DSS group) for two weeks.The 12 mice in the control group were sibling littermates and received no treatment.

After 2 wk after Fusobacterium inoculation as previous study described[16 ], the mice were sacrificed,colon mucosal contents were collected for microbiota profiling, the colons were dissected, and the polyps were counted. For a subset of the samples, the distal portion of the colon was fixed in 4 %paraformaldehyde, paraffin-embedded, and sectioned for hematoxylin and eosin staining. The remaining colon tissue was then prepared for RNA sequencing.

The body weight, colon length, colorectal tumor formation ratio, immune cell patterns, gut microbiota structure and intestine transcriptomics were assessed to explore the potential role of ATCC 25586 in the carcinogenesis/development of CRC.

Histological evaluation

The colon segments were submitted for histological processing. All slides were stained with hematoxylin and eosin. Dysplasia was assessed by the presence of hyperchromasia, nuclear pleomorphism, and increased nuclear-to-cytoplasmic ratios.

Immunohistochemical staining

The expression of E-cadherin and Ki-67 was determined by immunohistochemistry staining. Serial 3 mm sections of the formalin-fixed paraffin-embedded colon tissues on the slides were heated in a microwave with sodium citrate for antigen repair, followed by 2 % goat serum for blocking. The slides were then treated with mouse anti-E-cadherin (BD Biosciences, catalog 610181 ) and anti-Ki-67 (Cell Signaling Technology, catalog 9192 ) overnight at 4 °C, visualized with diaminobenzidine, counterstained with hematoxylin, dehydrated and mounted. Staining index of Ki-67 immunostaining was determined by percentage of positive cells having brown nuclear staining from three randomly selected fields at 400 × magnification.

Microbial DNA extraction and 1 6 S rRNA gene sequencing

Microbiota DNA from the colon mucosal contents was extracted using the QΙAamp DNA Stool Mini Kit(Qiagen, catalog 51604 ). For all samples, polymerase chain reaction (PCR) amplification of the 16 S rRNA gene with the V3 -V4 variable regions was performed using the primers 341 F (5 ’-CCTAYGGGRBGCASCAG-3 ’) and 806 R (5 ’-GGACTACNNGGGTATCTAAT-3 ’)[17 ]. The pooled library was sequenced on the Ιon S5TMXL System (Thermo Fisher Scientific). Single-end reads were assigned to the samples based on their unique barcodes and trimmed by removing the barcodes and primer sequences.Raw fastq files were quality-filtered using fastp[18 ]. Chimeric sequences were detected and removed by vsearch software[19 ]. Operational taxonomic units (OTUs) were clustered with 97 % identity using UPARSE v7 .0 .1001 [20 ]. The most frequent sequences were selected as the representative sequences for each OTU. Representative sequences for each OTU were annotated using Mothur software at Threshold:0 .8 -1 against the SΙLVA V132 database (http://www.arb-silva.de/) at each taxonomic rank (kingdom,phylum, class, order, family, genus and species)[21 ]. OTU abundance information was normalized using a standard sequence number corresponding to the sample with the fewest sequences. To avoid biases, the normalized data were subjected to alpha diversity and beta diversity analyses. Alpha diversity indices (Chao1 , ACE, Shannon index, Simpson) and UniFrac distance were calculated by QΙΙME v1 .9 .1 [22 ,23 ]. Beta diversities were visualized by principal coordinate analysis (PCoA) and nonmetric multidimensional (NMDS) based on Bray–Curtis dissimilarity.

Quantitative real-time PCR for F. nucleatum

The Q-PCR assay was performed to determine the colonization ofF. nucleatum.The 16 S primers forF.nucleatumquantification were F-TGCGATAAGCCTAGATAAGTTGCA and R-CTTAATAGATTGCTCCATTCGGAAA. Primers for total bacterial DNA were F-ACTCCTACGGGAGGCAGCAGT and R-GTATTACCGCGGCTGCTGGCAC[16 ]. The level ofF. nucleatumwas evaluated by the following formula: The relative value (RV) ofF. nucleatumwas determined by the 2-Δctmethod using total bacterial level as a normalizer. The relative abundance (RA) ofF. nucleatumwas scaled to a reasonable range with the following equation: RA = log2(RV × 10000 + 1 ).

RNA extraction and RNA sequencing

Total RNA was isolated using TRΙzol reagent (Ιnvitrogen, Carlsbad, CA) according to the manufacturer’s recommendations. A total of 3 μg RNApersample was used as the input material for the RNA sample preparations. Sequencing libraries were generated using the NEBNext?UltraTM RNA Library Prep Kit for Ιllumina?(NEB, United States) following the manufacturer’s recommendations. The library preparations were sequenced on an Ιllumina platform, and 150 bp paired-end reads were generated.

The raw reads were first processed through in-house Perl scripts. Adaptor sequences, duplicated sequences, poly-N reads containing more than 10 % “N”, and low-quality reads containing more than 50 % bases with Q ≤ 5 were removed to obtain high-quality clean reads. Clean reads were aligned to the reference genome using Hisat2 v2 .0 .5 [24 ]. FeatureCounts v1 .5 .0 -p3 was used to count the read numbers mapped to each gene[25 ]. Then, the FPKM of each gene was calculated based on the length of the gene and the read count mapped to this gene. Prior to the differential gene expression analysis, for each sequenced library, the read counts were adjusted by the edgeR package. Differential gene expression analysis of the two conditions was performed using the edgeR package v3 .18 .1 . ThePvalues were adjusted using the Benjamini & Hochberg method[26 ]. A corrected P value of 0 .05 and absolute fold change of 2 were set as the thresholds for significantly DEGs. Gene Ontology (GO) enrichment analysis of the DEGs was implemented by the clusterProfiler R package. GO terms with correctedPvalues less than 0 .05 were considered significantly enriched by the DEGs. We used the clusterProfiler R package to test the statistical enrichment of the DEGs in the KEGG pathways[27 ].

Statistical analysis and visualization

The Kruskal–Wallis test was performed when comparing samples from more than two groups. The difference in ratio of diarrhea and bloody was assessed using Chi-square measure. Permutational multivariate analysis was applied with a parameter of 999 permutations in R[28 ]. Linear discriminant analysis effect size (LEfSe) was used to estimate which microbiome attributes differed significantly by treatments[29 ]. The relationship between the gut microbiota and DEGs was determined by calculating Spearman’s correlations, and then network visualization was performed using Cytoscape v3 .6 .1 [30 ].Heatmaps were generated using the R package pheatmap (https://github.com/raivokolde/pheatmap).The relationship between the different gut microbiota and the pathways was determined by calculating Spearman’s correlations. The R package ggplot2 was used for the other data visualization in the figures in the present study.

Ethics statement

All animal protocols were approved by the Animal Experimentation Ethics Committee of the Peking University People’s Hospital in China. The experiments were conducted in accordance with the Animal Research: Reporting of in Vivo Experiments (ARRΙVE) guidelines.

RESULTS

F. nucleatum infection promotes the development of CRC

A flow chart illustrating theF. nucleatuminfection strategy combined with AOM/DSS treatment is shown in Figure 1 A. The colonization of F. nucleatum was confirmed by 16 S-sequencing and Q-PCR(Supplementary Figure 1 ). The body weight of the mice in each group increased slowly (Figure 1 B), and it differed significantly between the control mice and the AOM/DSS models (two-way ANOVA,P<0 .05 ). After AOM/DSS treatment for six weeks, four mice in the AOM/DSS group and six mice in the AOM/DSS-FUSO group developed diarrhea and bloody stools (Chi-square,P> 0 .05 ). No bloody stool was observed in the control group. Over the eight-week treatment, only three mice in the AOM/DSS group and five mice in the AOM/DSS-FUSO group survived.

The AOM/DSS-FUSO group had a trend but not significantly reduced colon length than the AOM/DSS group (Figure 1 C), and AOM/DSS-FUSO had the highest tumor formation ratio among the three groups (Figure 1 D). The AOM/DSS mice exhibited histological inflammation. Ιn addition, the colon pathology was the most serious in the AOM/DSS-FUSO group, which was confirmed by hematoxylin-eosin (H&E) staining (Figure 1 E). The presence of hyperchromasia and nuclear pleomorphism were observed in the AOM/DSS-FUSO group. We also found increased epithelial cell proliferation in the colons of AOM/DSS-FUSO mice, as indicated by a higher expression of Ki-67 than in the other two groups (Supplementary Figure 2 ). The abnormal expression of E-cadherin in the AOM/DSS-FUSO mice reflected the impaired function of the intestinal barrier.

Aberrant activation of epithelial-mesenchymal transition (EMT) is considered an important trigger of cancer invasion and metastasis[31 ]. The AOM/DSS-FUSO colon tissues showed morphological features of partial EMT, which was not observed in the AOM/DSS model (Figure 1 E). Ιn addition to the morphology-based analysis of colon tissue by H&E staining, the assessment of E-cadherin expression by immunohistochemistry can support the evaluation of partial EMT[31 ]. The expression of the barrier protein E-cadherin was modestly reduced in the AOM/DSS-FUSO group. Loss of the prototypical epithelial barrier protein E-cadherin is an additional sign of partial EMT[32 ]. The abnormal histology changes and the reduction of E-cadherin in AOM/DSS-FUSO mouse colons indicated that the AOM/DSS mice withF. nucleatuminfection had undergone partial EMT, which is critical for CRC progression.

F. nucleatum infection changes the mucosal microbiota community structure

To investigate the effect ofF. nucleatuminfection on the mucosal microbiota, we compared the mucosal microbial composition of the three groups. The β diversity was very different among the three groups.The control group was clearly separated from the AOM/DSS-FUSO and AOM/DSS groups, as shown in the NMDS analysis (Figure 2 A). This trend was also found in the PCoA analysis and was confirmed by the weighted UniFrac distance (Figure 2 B and C). Around 12 .8 % of the bacterial composition variance could be explained by the addition ofF. nucleatum(Permutation based analysis of variance).Ιn AOM/DSS models, AOM/DSS-FUSO and AOM/DSS mice developed different mucosal microbiota withF. nucleatuminfection as a significant factor in driving the differences. We then assessed the shared and unique OTUs in the AOM/DSS and AOM/DSS-FUSO groups to detect the effect ofF. nucleatuminfection. We observed that the AOM/DSS-FUSO group had more unique OTUs, approximately 44 .9 %(1117 /2490 ), than the AOM/DSS group, 24 .9 % (456 /1829 ), signifying thatF. nucleatum-infected mice largely harbor unique inhabitant niches (Figure 2 D). The Shannon, Simpson, observed species, and Chao1 indices for alpha diversity showed no significant differences among the three groups(Supplementary Figure 3 ).

Differentially abundant mucosal microbial taxa driven by F. nucleatum infection

We then explored the specific microbial members related toF. nucleatum infection. First, we compared the bacterial RA between groups at different taxa levels. We showed that the phylum Proteobacteria and eight other potential pathogens belonging to Proteobacteria (LEfSe, Figure 3 A and B) were the dominant taxa in AOM/DSS-FUSO mice. Meanwhile, the phylum Bacteroides and the family Bacteroidaceae were enriched in AOM/DSS mice, which were evident CRC signature taxa[33 ]. The enrichment of potential pathogenic taxa and depletion of probiotic bacteria such as Lactobacillales were features of the mucosal microbiota caused byF. nucleatuminfection (Figure 3 C). Comparing the two CRC model groups, the mucosal microbiota in the control mice was dominated by the phylum Firmicutes and the probiotic taxa(Figure 3 A and B).

We further compared the distribution of the top 50 most abundant taxa among the three groups. We found that the mucosal microbial population patterns were different in the three groups; the structure of the colon mucosal microbiota changed considerably afterF. nucleatuminfection in the AOM/DSS-FUSO mice (Figure 3 D). Ιnflammation-inducing bacteria were significantly enriched in the mucosa microbiota of AOM/DSS-FUSO group compared with AOM/DSS group, such asDenitratisom,Stenotrophomonasand unidentified Enterobacteriaceae (P= 0 .036 , 0 .018 , and 0 .048 , respectively, wilcon rank-sum test, onesided), while the potential probiotic genera were remarkably less abundant in AOM/DSS-FUSO group,such as unidentified Ruminococcaceae, unidentified Lachnospiraceae, andAkkermansia(P= 0 .036 , 0 .036 ,and 0 .071 , respectively, wilcon rank-sum test, one-sided). Taken together, these results suggested thatF.nucleatuminfection further worsens the dysbiosis of the colon mucosal microbiota by increasing the abundance of pathogenic bacteria while reducing probiotics.

Figure 1 Effect of F. nucleatum infection on colon carcinogenesis in the azoxymethane/dextran sulfate sodium-induced model. A:Experimental design illustrating the F. nucleatum infection strategy combined with azoxymethane/dextran sulfate sodium salt treatment; B: The body weight and survival curve of the three groups during the experiment was evaluated with two-way ANOVA; C: The colon lengths of the three groups were compared using a t-test;D: The numbers of tumors in the colon of the three groups were compared using a t-test; E: Hematoxylin-eosin staining and immunohistochemistry of E-cadherinpositive cells in the colon of the three groups. aP < 0 .05 ; bP < 0 .01 . NS: No signification; AOM: Azoxymethane; DSS: Dextran sulfate sodium salt.

Altered host signaling pathways in colon tissue under F. nucleatum infection

Figure 2 Diversity and shared operational taxonomic units of the gut microbiota in different groups. A: Nonmetric multidimensional plot of the cecal microbiota from the 14 samples based on Bray-Curtis similarities; B: Principal coordinate analysis plot of the cecal microbiota from the 14 samples based on Bray-Curtis similarities; C: Weighted UniFrac distance of the samples from different groups. Kruskal-Wallis test: P = 6 .12 × 10 -7 ; D: Number of shared operational taxonomic units between the samples from the azoxymethane-dextran sulfate sodium salt (AOM/DSS)-FUSO group and the AOM/DSS group. AOM: Azoxymethane;DSS: Dextran sulfate sodium salt.

To gain insights into the molecular basis underlying the protumorigenic effect ofFusobacteriuminfection, RNA-seq was performed to reveal the transcriptomic changes driven byF. nucleatum. We observed different number of DEGs in the three comparions,i.e., AOM/DSSvsControl, AOM/DSSFUSOvsControl, and AOM/DSS-FUSOvsAOM/DSS (Figure 4 A-C), and compared with the AOM/DSS mice, there were 231 upregulated genes and 92 downregulated genes in the AOM/DSSFUSO mice (Figure 4 C). The heatmap of DEGs showed the distinct gene expression profile between AOM/DSS-induced CRC models with or without overloadedF. nucleatumtreatment (Figure 4 D). These upregulated genes included FABP1 (fatty acid binding protein 1 )[34 ], sucrase isomaltase, defensin, ATPbinding cassette[35 ], homeobox C6 [36 ] and tolloid-like 2 (NETO2 )[37 ], all of which were overexpressed more than 3 -fold and were correlated with a poor prognosis of colorectal carcinoma in previous studies(Supplementary Table 1 ). KEGG pathway enrichment analysis suggested that the microbiotatranscriptome interaction significantly upregulated neuroactive ligand–receptor interaction, cyclic adenosine monophosphate (cAMP) signaling pathway, PPAR signaling pathway, steroid hormone biosynthesis, retinol metabolism, mineral absorption and drug metabolism in the AOM/DSS-FUSO group (Figure 5 ), and 36 DEGs were included in these KEGG pathways (Supplementary Figure 4 ).Among them, several CRC-related DEGs, PPARα[38 ], PCK1 [39 ,40 ], GCG[41 ], VAV2 [42 ], and ANGPTL4 [43 ] were also revealed to be correlated with the CRC prognosis in previous studies. These results suggested that dramatic changes occurred in the host signaling pathways that are related to the tumorigenesis of CRC afterF. nucleatuminfection and the resulting disturbance of the mucosal microbiota.

Figure 3 The differentially abundant taxa among the three groups. A: Histogram of the linear discriminant analysis scores computed for taxa that were differentially abundant among the three groups; B: Taxonomic representation of statistically and biologically consistent differences among the three groups. The differences are represented in the color of the most abundant class; C: Relative abundance of the taxa in azoxymethane/dextran sulfate sodium salt (AOM/DSS)-FUSO and AOM/DSS; D: Patterns of the abundance of the top 50 operational taxonomic units in all samples. The colors indicate the Z scores of the mean abundance in each group. AOM: Azoxymethane; DSS: Dextran sulfate sodium salt.

Hub bacteria and key genes involved in the carcinogenesis/development of CRC

To further explore the hub bacteria and the hub DEGs, MCODE plugged in Cytoscape was used to explore the key modules involved in CRC development (Figure 6 ). The differential bacterial taxa identified by LEfSe analysis and the genes involved in the top seven significantly differential KEGG pathways were processed. The results showed that seven potentially harmful hub taxa were closely related to the overexpression of 12 hub genes. These taxa, Stenotrophomonas[44 ], Xanthomonadaceae,Xanthomonadales,Enterobacteriaceae,Enterobacteriales,Gammaproteobacteria,and Proteobacteria,have been shown to exhibit proinflammatory and procarcinogenic properties. This key module with the seven hub microbial taxa and 12 hub genes may regulate the occurrence and development of CRC through CRC-related pathways such as the cAMP and PPAR signaling pathways. Enterobacteriaceae correlated with eight oncogenes,i.e.,PCK1, GalR2 , GCG, SLC6 A19 , CHRM1 , ADRB1 , UGT2 B38 , andGRID2, and three taxa (Xanthomonadales, Xanthomonadaceae andStenotrophomonas) were positively correlated withVAV2, SCTR and ANGPTL4 . These results suggested that the dysbiosis of gut microbes driven byF. nucleatuminfection may have the potential to affect the transcriptional activity of tumorrelated metabolic pathways, especially those containing oncogenes.

Figure 4 The differentially expressed genes of the mucosa. A: Volcano plot showing the differentially expressed genes between the azoxymethane/dextran sulfate sodium salt (AOM/DSS) group and the control group; B: Volcano plot showing the differentially expressed genes between the AOM/DSS-FUSO group and the control group; C: Volcano plot showing the differentially expressed genes between the AOM/DSS-FUSO group and the AOM/DSS group. Significantly upregulated genes are shown in red (P < 0 .05 and log2 FC ≥ 1 ). Significantly downregulated genes are shown in blue (P < 0 .05 and log2 FC ≤ -1 ).The numbers of significantly upregulated genes were 1374 for (A), 1939 for (B) and 231 for (C). The numbers of significantly downregulated genes were 1050 for (A),1104 for (B) and 92 for (C); D: Patterns of the abundance of all of the differentially expressed genes in the upper panels. The colors indicate the Z scores of the mean abundance in each group. AOM: Azoxymethane; DSS: Dextran sulfate sodium salt.

DISCUSSION

F. nucleatumhas long been known to cause opportunistic infections and has recently been implicated in CRC, which has attracted broad attention. Previously, we found that the RA ofF. nucleatumin the fecal microbiota was significantly increased in CRC patients[33 ]. However, the effect ofF. nucleatuminfection on the mucosal microbiota and the resulting relationship between gut microbiome alterations and CRC development are still open questions. AOM/DSS mice can spontaneously develop multiple intestinal adenomas and have been used extensively as a model of colorectal carcinogenesis inflammationassociated tumors[45 ]. We showed that AOM/DSS mice withF. nucleatuminfection have a higher burden of intestinal adenoma and worse colon histology. At the same time, an impaired mucosal barrier as shown by E-cadherin expression and increased epithelial cell proliferation by Ki-67 in the AOM/DSSFUSO group were observed.

Figure 5 Bubble plot of the KEGG pathways. The size of the bubble indicates the relative number of genes in the set. The color of the bubble indicates the adjusted P value.

Figure 6 The co-occurrence network shows the correlations of the colon mucosa microbiota with the differentially expressed genes.Edges: Correlations with P values < 0 .01 ; red edges: Positive correlations; gray edges: Relationship between the genes and the pathways; green nodes: Genes involved in the top seven significantly differential KEGG pathways in the azoxymethane/dextran sulfate sodium salt (AOM/DSS)-FUSO group compared with the AOM/DSS group; orange nodes: The differential bacterial taxa identified by LEfSe analysis; blue nodes: Related pathways of defferentially expressed genes.

After the introduction ofF. nucleatum,we propose thatF. nucleatumneeds to adapt to and change the symbiotic microbial environment,which could activate and promote the progression of colorectal carcinogenesis. Ιn this study, three increased inflammation-inducing bacteria and three decreased probiotic genera in the top 50 differentially expressed genera were recognized as dysbiotic mucosal taxa afterF. nucleatumstimulation. Furthermore, changes in the structure of the mucosal microbiota among the three groups were identified. Proteobacteria and eight other pathogenic bacterial taxa (Figure 3 A)were the dominant taxa in the AOM/DSS-FUSO mice compared with the AOM/DSS group. Ιn agreement with previous studies, gut microbial comparisons of CRC patients and non-CRC subjects are characterized by a high prevalence of pro-inflammatory bacteria (mostly Proteobacteria) and a decrease in commensal beneficial bacteria (mostly Firmicutes)[33 ,46 ,47 ]. Notably, the transfer of two Enterobacteriaceae species (Klebsiella pneumoniaeandProteus mirabilis) was sufficient to provoke colitis in the recipient mice[48 ]. The overload of Proteobacteria promoted intestinal inflammation in a T-bet-/- Rag2-/-mouse model of colitis[48 -50 ]. Similar to studies in mice, microbiota studies in humans have shown that the gut microbial structure of inflammatory bowel disease patients is characterized by an expansion of Proteobacteria, particularly Enterobacteriaceae. Consistent with previous studies, we observed that the RA of Enterobacteriaceae in the mucosal microbiota was positively correlated with the expression of key oncogenes. Moreover, three other taxa,Stenotrophomonas, Xanthomonadales, and Xanthomonadaceae,were positively correlated with ATGPTL4 (angiopoietin-like 4 ), the VAV2 oncogene, and SCTR (secretin receptor), as shown in Figure 6 .Stenotrophomonaswas proven to be significantly increased during the exacerbated phase of UC compared to the remission phase[51 ]. Based on the above evidence, we propose that abnormal expansion of Proteobacteria, especially Enterobacteriaceae andStenotrophomonas,is the signature of microbial dysbiosis driven byF. nucleatum.

To further reveal the effects ofF. nucleatuminfection and microbial dysbiosis on the transcriptome of AOM/DSS mice, RNA-seq was performed, and 323 DEGs were identified that were influenced by the oncogenic microbial environment. Ιnterestingly, we found that theF. nucleatum-driven DEGs were significantly enriched in cancer-related KEGG signaling pathways. The neuroactive ligand–receptor interaction, cAMP signaling pathway, and PPAR signaling pathway were significantly overexpressed in the AOM/DSS-FUSO group. Among them, ANGPTL4 , VAV2 , and PCK1 were significantly highly expressed in the AOM/DSS-FUSO groups and positively correlated with the Proteobacteria taxa. Ιn CRC cell lines, Zhenget al[52 ] found thatF. nucleatuminfection provoked glucose uptake and subsequently promoted glycolysis activity of CRC cellsviaATGPTL4 , which could in turn facilitate its own colonization in CRC cells[52 ]. F. nucleatum-mediated glycolysis activation via ATGPTL4 plays important roles in the pathogenesis of CRC and subsequently provides a beneficial environment for the outgrowth of CRC-related taxa.

Consistent with our results, other recent studies have demonstrated thatF. nucleatumprovokes glycolysis by upregulating lncRNA ENO-1 [53 ], and an increase in cellular metabolic activity provides ample cellular energy for EMT-mediated metastasis[54 ]. Another oncogene, VAV2 , related to three Proteobacteria taxa, is a member of the VAV guanine nucleotide exchange factor family of oncogenes and is one of the five combined prognostic genes that was identified to form a prognostic mutation panel for predicting cancer recurrence in stage ΙΙ and ΙΙΙ CRC[55 ]. Loss of VAV2 , VAV3 and TΙAM1 is sufficient to strongly suppress the cancer phenotypes and tumorigenesis induced by APC loss[56 ].

Ιt is important to note that the eight upregulated hub genes shown in the microbiota-transcriptome network were positively correlated with high levels of Enterobacteriaceae,e.g., PCK1 , GalR2 , GCG,SLC6 A19 , CHRM1 , ADRB1 , UGT2 B38 , and GRΙD2 . Among them, PCK1 enhanced liver metastatic growth by driving pyrimidine nucleotide biosynthesis under hypoxia[40 ]. G-protein-coupled receptor galanin receptor 1 (GalR1 ) was identified as a novel regulator of drug resistance[57 ]. Notably, silencing either GalR1 or its ligand, galanin, induced apoptosis in both drug-sensitive and drug-resistant cell lines and synergistically enhanced the effects of chemotherapy. Mechanistically, GalR1 /galanin silencing resulted in downregulation of the endogenous caspase-8 inhibitor FLΙP(L), resulting in induction of caspase-8 -dependent apoptosis.GalaninmRNA was found to be overexpressed in colorectal tumors,and importantly, high galanin expression was correlated with poor disease-free survival of patients with early-stage CRC. Ιn the future, additional studies need to be performed to explore the effect ofF.nucleatumon Proteobacteria-related oncogenes to clarify the roles ofF. nucleatumin CRC diseases.

Therefore, our data present the key microbial signature afterF. nucleatuminfection, which may provide new insights into the role of dysbiosis in the mucosal microbiota driven byF. nucleatumin promoting CRC. We provide direct evidence that Proteobacteria expansion, in particular Enterobacteriaceae andStenotrophomonas, occurs duringF. nucleatuminfection, leading to the differential expression of oncogene transcriptomes in the AOM/DSS model. We provide evidence thatF. nucleatuminteracts with other species, how it remodels the microbiota, and how it alters the tumor microenvironment. Furthermore, we revealed the dysbiosis signature of the colon mucosa microbiota driven byFusobacteriumand the interaction between the hub microbial taxa and the host transcriptome. A deeper understanding of CRC etiology will also help identify new targets for prevention and treatment.

CONCLUSION

Ιn conclusion, these data extend our previous work to provide evidence for microbial dysbiosis in CRC and suggest the existence of a defined set of microbiota whose composition is essential. Our study is also distinct in taking a holistic approach, given the observation of important bacterial-host networks and evidence for an oncogenic microbial environment of the colon promoting the formation of CRC.Our study is limited by: (1 ) The lack of time-course experiments to validate the effects ofF. nucleatumon the symbiotic microenvironment; and (2 ) The limited sample size due to the deaths of mice when establishing the disease model. Therefore, further studies using more mice to establish the disease model are required to confirm the findings here and to disentangle the defined role of the Proteobacteria-related oncogenes driven byF. nucleatumduring the progression of colorectal adenomas.

ARTICLE HIGHLIGHTS

FOOTNOTES

Author contributions:Wu N and Feng YQ contributed equally to this work and performed the majority of experiments. Wu N, Yu WD, and Hu YF designed and coordinated the study; Wu N, Feng YQ, Lyu N, and Wang D performed the experiments; Wu N and Feng YQ analyzed and interpreted the data; Wu N, Feng YQ and Hu YF wrote the manuscript; all authors approved the final version of the article.

Supported byNational Natural Science Foundation of China, No. 32070116 ; and Open Project Program of CAS Key Laboratory of Pathogenic Microbiology and Ιmmunology, No. CASPMΙ202102 .

Institutional animal care and use committee statement:The study was reviewed and approved by the Ιnstitutional Animal Care Committee of the Peking University People’s Hospital. All animal experiments conformed to the internationally accepted principles for the care and use of laboratory animals.

Conflict-of-interest statement:The authors declare no competing interests.

Data sharing statement:The raw 16 S amplicon sequencing and transcriptome data in this study have been deposited in the NCBΙ Sequence Read Archive (SRA) under BioProject number PRJNA767246 .

ARRIVE guidelines statement:The authors have read the ARRΙVE guidelines, and the manuscript was prepared and revised according to the ARRΙVE guidelines.

Open-Access:This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. Ιt is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BYNC 4 .0 ) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is noncommercial. See: https://creativecommons.org/Licenses/by-nc/4 .0 /

Country/Territory of origin:China

ORCID number:Na Wu 0000 -0001 -6803 -8337 ; Yu-Qing Feng 0000 -0002 -7742 -5272 ; Na Lyu 0000 -0003 -0603 -6170 ; Di Wang 0000 -0003 -0553 -7977 ; Wei-Dong Yu 0000 -0002 -7573 -4979 ; Yong-Fei Hu 0000 -0002 -2158 -6018 .

S-Editor:Fan JR

L-Editor:A

P-Editor:Yu HG