Metformin abrogates Fusobacterium nucleatum-induced chemoresistance in colorectal cancer by inhibiting miR-361-5p/sonic hedgehog signaling-regulated stemness

Background Chemotherapy resistance is the major cause of recurrence in patients with colorectal cancer (CRC). A previous study found that Fusobacterium (F.) nucleatum promoted CRC chemoresistance. Additionally, metformin rescued F. nucleatum-induced tumorigenicity of CRC. Here, we aimed to investigate whether metformin could revert F. nucleatum-induced chemoresistance and explore the mechanism. Methods The role of metformin in F. nucleatum-infected CRC cells was confirmed using cell counting kit 8 assays and CRC xenograft mice. Stemness was identified by tumorsphere formation. Bioinformatic analyses were used to explore the regulatory molecules involved in metformin and F. nucleatum-mediated regulation of the sonic hedgehog pathway. Results We found that metformin abrogated F. nucleatum-promoted CRC resistance to chemotherapy. Furthermore, metformin attenuated F. nucleatum-stimulated stemness by inhibiting sonic hedgehog signaling. Mechanistically, metformin diminished sonic hedgehog signaling proteins by targeting the MYC/miR-361-5p cascade to reverse F. nucleatum-induced stemness, thereby rescuing F. nucleatum-triggered chemoresistance in CRC. Conclusions Metformin acts on F. nucleatum-infected CRC via the MYC/miR-361-5p/sonic hedgehog pathway cascade, subsequently reversing stemness and abolishing F. nucleatum-triggered chemoresistance. Our results identified metformin intervention as a potential clinical treatment for patients with chemoresistant CRC with high amounts of F. nucleatum.

BACKGROUND Colorectal cancer (CRC) incidence and mortality remain high, constituting a major public health burden [1]. Chemotherapy is the most common treatment to reduce tumor growth and inhibit tumor metastasis. 5-fluorouracil (5-Fu) and oxaliplatin are the firstline chemotherapeutic drugs in the treatment of patients with advanced CRC [2,3]. However, CRC cells usually become resistant to chemotherapeutic agents, leading to tumor recurrence. It has been reported that more than 30% patients receiving surgery for stage II and III colon cancer are resistant to 5-Fu-based adjuvant chemotherapy within 8 years of follow-up [4]. Furthermore, about 50% of patients who were diagnosed with metastatic CRC were resistant to 5-Fu-based chemotherapy and their 5-year survival rate was lower than 15% [5]. Thus, it is necessary to determine the molecular mechanisms of CRC chemoresistance and develop new and effective strategies to improve chemosensitivity.
Chemoresistance of CRC is a complex process resulting from the interplay between intrinsic and extrinsic factors, including gene regulation, epigenetic modification, hypoxia, and the gut microbiota [6,7]. Accumulating evidence indicates that the gut microbiota profoundly influences the effect of chemotherapy with three main outcomes: facilitation of drug efficacy; restraint and compromise of anticancer effects; and mediation of toxicity effects [8]. Some bacteria, such as Lactobacillus johnsonii, L. murinus, and Bacteroidetes, promote chemotherapy efficacy [9,10]. However, some bacteria, such as F. nucleatum, a well-known CRC pathogenic factor, drive high chemoresistance [11][12][13][14]. Our previous work demonstrated that an increased abundance of F. nucleatum contributes to post-chemotherapy recurrence in patients with CRC [15]. As such, this is an important clinical consideration when evaluating conventional chemotherapeutic treatment of patients with CRC with a high amount of F. nucleatum. Therefore, a strategy capable of modulating the effects of F. nucleatum in the management of CRC is highly desirable. However, to date, the problem of CRC chemoresistance caused by F. nucleatum has not been solved. Our previous work pinpointed that metformin could elicit an antitumor effect by restoring the disordered intestinal microbiota and alleviating F. nucleatuminduced colorectal tumorigenesis [16]. Metformin, which has been used widely in type 2 diabetes (T2D) for more than 60 years, has attracted interest in the prevention and treatment of various cancers recently [17][18][19][20]. A phase II clinical trial reported a modest efficacy profile of a combination of metformin and 5-Fu in patients with refractory CRC [21]. These raised the possibility that metformin might play a role in treating chemoresistance in patients with CRC with high abundance of F. nucleatum. In the current work, we aimed to determine whether and how metformin affects F. nucleatum-induced CRC chemoresistance. We found that metformin plays a critical role in abrogating F. nucleatum-induced CRC chemoresistance in response to 5-Fu and oxaliplatin through the MYC/miR-361-5p cascade, and subsequently suppression of sonic hedgehog signaling and CRC cell stemness. Our approach might inspire a future combined treatment comprising metformin with chemotherapeutic drugs to managing recurrence in patients with CRC with high amounts of F. nucleatum.

Cell culture and treatment
Human CRC cell lines with different degrees of differentiation (HT-29, well differentiated; HCT 116, moderately differentiated) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in McCoy'5 A medium (GIBCO, Carlsbad, CA, USA) with 10% (vol/ vol) fetal bovine serum (FBS) (GIBCO) at 37°C in a humidified 5% CO 2 atmosphere. 1 mM metformin (Sigma-Aldrich, St. Louis, MO, USA) was used to treat CRC cells for 24-72 h. F. nucleatum at a multiplicity of infection (MOI) of 100 was used to treat CRC cells for 4 h. The miR-361-5p mimics and miR-361-5p inhibitor were transfected into CRC cells using the DharmaFECT transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA). The lentivirus control and lentiviral vectors expressing miR-361-5p inhibitors were purchased from Shanghai OBiO medical biotechnology company (Shanghai, China).

F. nucleatum culture
The F. nucleatum strain (ATCC 25586) was obtained from the ATCC. F. nucleatum was cultured for 24 h at 37°C under anaerobic conditions in brain heart infusion (BHI) broth containing hemin, Vitamin K1, K 2 HPO 4 , and L-Cysteine.

Mouse models
For the xenograft experiments, 4-5-week-old male Balb/c nude mice were purchased from the Shanghai Model Organisms Center (Shanghai, China) and housed in specific pathogen-free (SPF) conditions. HCT 116 cells (5 × 10 6 or 2.5 × 10 6 ) with different treatments were suspended in 100 μl of phosphate-buffered saline (PBS) and inoculated subcutaneously into the right axilla of the mice to establish the CRC xenograft model. Six days after subcutaneous injection, mice with xenograft tumors and no significant differences in tumor load were divided into different groups (n = 5). F. nucleatum (MOI = 100, 100 μl per mouse) were given by multipoint intratumoral injection, twice per week for 2 weeks. Metformin (125 mg/Kg) was administered via gavage every day for two weeks. Chemotherapeutic agents (FOX: 5-Fu 5 mg/Kg + Oxaliplatin 6 mg/Kg) were administered by intraperitoneal injection, twice per week for 2 weeks.
To explore the role of metformin in F. nucleatum-induced chemoresistance in vivo, we designed seven groups: (i) Control group; (ii) Metformin group; (iii) F. nucleatum group; (iv) FOX group; (v) FOX and metformin group; (vi) FOX and F. nucleatum group; (vii) FOX, F. nucleatum, and metformin group.
The length and width of the tumors were measured every 2-3 days. Tumor volume (mm 3 ) was calculated as (length × width 2 ) /2. If the tumor volume exceeded 2000 mm 3 , the mouse was killed for welfare reasons. After 2 weeks, the mice were killed humanely and subcutaneous tumors were collected for subsequent analysis of sonic hedgehog (SHH), GLI family zinc finger 1 (GLI1), SRY box transcription factor 2 (SOX2) and Nanog homeobox (NANOG) expression. Mouse experiments were conducted according to the guidelines approved by the Institutional Animal Care and Use Committee of Renji Hospital, School of Medicine, Shanghai Jiaotong University.

RNA extraction and quantitative real-time PCR
Total RNA was extracted from HT-29 and HCT 116 cells using the Trizol reagent (Takara, Shiga, Japan). For the analysis of mRNA and primary microRNA (pri-miRNA), 1 μg of total RNA was reverse transcribed into firststrand cDNA using a PrimeScript RT Reagent Kit (Takara). For the miRNAs, 500 ng of total RNA was reverse transcribed using Mir-X TM miRNA First-Strand Synthesis Kit (Takara). The quantitative real-time PCR step was conducted using TB Green Premix Ex Taq™ II (Takara) on an ABI StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The relative gene expression was calculated using 2 −ΔΔCt method and expressed as the fold change [22]. ACTB (encoding β-actin) and U6 were used, respectively, as endogenous references for the mRNA, pri-miRNA, and miRNA assays. The mRNA primers used in real-time PCR are shown in Table 1. The miRNA primers and pri-miRNA primers were purchased from GeneCopoeia (Rockville, MD, USA).
Tumorsphere formation assay CRC cells were exposed to F. nucleatum (MOI = 100) for 4 h. Then, the medium containing F. nucleatum was replaced with McCoy'5 A medium supplemented with 10% FBS. After 24 h, HT-29 and HCT 116 cells were seeded in low-adherent 96-well plates at a density of 400 cells per well in serum-free tumorsphere medium with or without 1 mM metformin. The tumorsphere medium consisted of Dulbecco's Modified Eagle Medium/F12 (GIBCO) supplemented with necessary growth factors [23]. After a 4-day

Data acquisition
For the analysis of cancer stem cell pathways, the data and sample information of GSE102573, GSE90944, and GSE67342 were downloaded from the Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/ geo/). GSE102573 comprises the expression profile of Caco-2 cells infected with F. nucleatum or not. GSE90944 comprises RNA sequencing (RNA-seq) data of HT-29 cells with or without F. nucleatum treatment. GSE67342 comprises expression data of LoVo cells treated by metformin. The single-sample gene set enrichment analysis (ssGSEA) score was calculated using Gene Set Variation Analysis (GSVA) in the R package with the chemical and genetic perturbations gene set collection (MSigDB C2 CGP; 3358 gene sets available) [24]. Differential expression analysis was conducted using the R package Limma and an adjusted (Bonferroni-Holm method) two-tailed P < 0.05 was considered statistically significant.
For the analysis of metformin-regulated genes, we used the dataset GSE67342, which includes gene expression of LoVo cells treated by metformin. An adjusted P < 0.1 was considered statistically significant.

Dual luciferase assay
HT-29 and HCT 116 cells were co-transfected with 100 ng of the reporter plasmids and 10 ng of the pRL-TK-Renilla-luciferase plasmids using the FuGene transfection reagent (Promega, Madison, WI). At 6 h after transfection, the medium was replaced and cells were treated with metformin and/or F. nucleatum. At 72 h after transfection, dual luciferase activities were estimated using a Dual-Luciferase Assay Kit (Promega) according to the manufacturer's instructions. Each transfection was performed in four complex wells and repeated twice.

Statistical analysis
All data analyses were performed using GraphPad Prism version 7 (GraphPad Inc., La Jolla, CA, USA). Data are expressed as the mean ± standard deviation (SD) or the mean ± standard error of the mean (SEM). The differences between experimental groups were analyzed using an unpaired t test. In the figures, asterisks denote statistical significance (n.s., not significant, P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Metformin abolishes F. nucleatum-induced chemoresistance in vivo and in vitro
To explore the role of metformin in F. nucleatum-induced chemoresistance, we initially verified metformin's effect on CRC cell proliferation. After treating CRC cells with different concentration of metformin, we observed that 500 μM and 1 mM metformin did not affect the proliferation of HT-29 cells (Fig. S1a) or HCT 116 cells (Fig. S1b). However, 5 mM and 10 mM metformin inhibited the proliferation in both CRC cell lines (Fig. S1a, b). Importantly, 1 mM metformin is approximately the physiological concentration in the colon of patients taking this drug [25]. In addition, our previous work demonstrated that an MOI of 100 of F. nucleatum had no effect on CRC cell proliferation but could trigger CRC cell chemoresistance [15]. Therefore, we chose 100 MOI of F. nucleatum and 1 mM metformin for subsequent experiments. HT-29 and HCT 116 cells were pre-infected with F. nucleatum for 4 h and then treated with a gradient concentration of 5-Fu or oxaliplatin in the presence of 1 mM metformin. In line with our previous observations, F. nucleatum decreased the cytotoxicity of 5-Fu and oxaliplatin in HT-29 cells (Fig. 1a, b) and HCT 116 cells ( Fig. 1c, d), respectively, after 72 h. These results suggest that F. nucleatum promotes chemoresistance of CRC cells. Interestingly, metformin markedly enhanced the cytotoxicity of 5-Fu and oxaliplatin in F. nucleatum-infected HT-29 cells (Fig. 1a, b) and HCT 116 cells (Fig. 1c, d), but not in F. nucleatum-free HT-29 cells and HCT 116 cells. Thus, metformin could abolish F. nucleatuminduced CRC chemoresistance in vitro.
To further verify this conclusion, HCT 116 cells were inoculated into Balb/c nude mice, followed by treatment with metformin (125 mg/Kg/d), F. nucleatum, chemotherapeutic agents (FOX: 5-Fu 5 mg/Kg + Oxaliplatin 6 mg/Kg) and other manipulations. We observed comparable tumor growth (Fig. 1e), tumor volumes (Fig. 1f), and tumor weights (Fig. 1g) in the control, metformin, and F. nucleatum groups. As expected, tumor growth (Fig. 1e), tumor volumes (Fig. 1f), and tumor weights (Fig. 1g) were significantly decreased by FOX treatment, and this decrease was blocked by F. nucleatum infection. These data further demonstrated that F. nucleatum participated in CRC chemoresistance. Remarkably, Balb/ c nude mice treated with F. nucleatum plus 1 mM metformin exhibited a better response to FOX, with impaired tumor growth, reduced tumor volumes, and lower tumor weights, when compared with the F. nucleatum plus FOX group ( Fig. 1e-g). Collectively, these data supported the view that metformin reverses F. nucleatum-induced CRC chemoresistance.

Metformin attenuates F. nucleatum-induced chemoresistance by inhibiting CRC cell stemness
We next explored the detailed mechanisms by which metformin abolished F. nucleatum-induced CRC chemoresistance. Given the crucial role of stemness in cancer chemoresistance [26], we hypothesized that metformin might rescue F. nucleatum-induced chemoresistance through inhibition of cancer stemness. To test this assumption, we performed bioinformatic analysis to predict the regulation of metformin and F. nucleatum in cancer stem cell pathways. We collected the gene expression data for CRC cells treated with metformin or F. nucleatum from GEO. Subsequently, ssGSEA showed that F. nucleatum upregulated cancer stem cell pathways, whereas metformin downregulated cancer stem cell pathways (Fig. 2a). Based on this data, we speculated that metformin might inhibit F. nucleatum-stimulated CRC stemness. Consistently, the tumorsphere formation assay showed that F. nucleatum increased the number and diameter of tumorspheres in both HT-29 cells (Fig. 2b) and HCT 116 cells (Fig. 2c). By contrast, metformin significantly inhibited the number and diameter of tumorspheres in F. nuleatum-stimulated HT-29 cells (Fig. 2b) and HCT 116 cells (Fig. 2c), but not in F. nuleatum-free cells.
In support of this data, we detected the expression of stemness markers after metformin and F. nucleatum treatment. NANOG and SOX2 are two important markers of stemness. F. nucleatum enhanced the mRNA and protein expression levels of SOX2 and NANOG in HT-29 and HCT 116 cells ( Fig. 2d-f). Metformin abolished F. nucleatum-induced upregulation of SOX2 and NANOG at both the mRNA (Fig. 2d, e) and protein (Fig. 2f) levels. In line with the above results, western blotting analysis also showed that metformin downregulated F. nucleatum-stimulated SOX2 and NANOG expression in FOX-treated subcutaneous xenograft mice (Fig. 2g). Taken together, our data indicated that metformin attenuates F. nucleatum-induced chemoresistance by inhibiting CRC cell stemness.
Metformin rescues F. nucleatum-induced stemness by inhibiting the sonic hedgehog pathway We next examined how metformin rescued F. nucleatum-induced stemness in CRC cells. Notch, Wnt/β-catenin, and sonic hedgehog signaling pathways are known to play a decisive role in tumor stemness via regulating SOX2 and NANOG expression [27]. We hypothesized that Notch, Wnt/β-catenin, or sonic hedgehog signaling pathways were involved in the recovery effect of metformin in F. nucleatum-induced stemness. To test this hypothesis, we evaluated these three signaling pathways by testing the expression of critical transcription factors individually, including SHH (sonic hedgehog pathway), GLI1 (sonic hedgehog pathway), Transcription factor 1 (TCF1) (Wnt pathway), and Hes family BHLH transcription factor 1 (HES1) (Notch pathway). F. nucleatum enhanced the mRNA expression of SHH, GLI1, TCF, and HES1 in HT-29 cells (Fig. 3a and Fig. S2a) and HCT 116 cells ( Fig. 3b and Fig. S2b). This indicated F. nucleatum activates these three pathways. To our surprise, metformin reduced the mRNA expression of SHH and GLI1 in F. nucleatum-infected HT-29 cells (Fig. 3a) and HCT 116 cells (Fig. 3b), but not in F. nucleatum-free CRC cells. Consistently, western blotting also showed metformin reduced SHH and GLI1 protein levels (Fig. 3c). By contrast, metformin could not downregulate the expression of TCF1 and HES1 in F. nucleatum-infected CRC cells (Fig. S2a, b). These data implied that the sonic hedgehog pathway might be involved in the metformin-mediated downregulation of F. nucleatum-induced stemness. In accordance with these results, CRC xenograft mice treated with FOX also showed that metformin diminished F. nucleatum-stimulated SHH and GLI1 protein levels (Fig. 3d). These data indicated that metformin inhibits F. nucleatum-induced stemness by inhibiting the sonic hedgehog pathway.
hedgehog signaling. When the SHH ligand binds to Patched 1 (PTCH1) at the cell membrane, GLI1 is accumulated and activated. GLI1 is the final transcriptional effector and its expression reflects the activation of sonic hedgehog signaling [28]. Thus, we constructed the recombinant luciferase reporter plasmid pGL3-GLI1, containing the promoter region of GLI1 ( − 979 to 33 nts) [29]. Luciferase assays showed that both F. nucleatum and metformin had no effect on the transcriptional activity of pGL3-GLI1 in HCT 116 cells (Fig. S3a). This result suggests that the regulation of GLI1 by metformin and F. nucleatum is not dependent on the direct transcriptional modulation of the GLI1 promoter. Therefore, we sought to determine if metformin and F. nucleatum regulated the GLI1 expression at the posttranscriptional level. MicroRNAs (miRNAs) are important post-transcriptional regulators of gene expression [30]. We hypothesized that dysregulated miRNAs might be involved in metformin-and F. nucleatumregulated GLI1. To test this hypothesis, we identified F. nucleatumrelated miRNAs from two studies [15,31]. F. nucleatum appeared to regulate 262 miRNAs significantly (Fig. 4a). Next, we used the bioinformatic tool, TargetScan (http://www.targetscan.org) to identify potential miRNAs that might regulate GLI1. Seventy-four potential GLI1-regulatory miRNAs were found (Fig. 4a). After overlapping these potential GLI1-regulatory miRNAs with the 262 F. nucleatum-related miRNAs, we identified six miRNAs (miR-361-5p, miR-509-3-5p, miR-509-5p, miR-103b, miR-4496, and miR-616-3p) that might regulate GLI1 (Fig. 4a).
We validated whether these six miRNAs are regulated by metformin and F. nucleatum using real-time PCR. The results showed that 1 mM metformin alone had no regulatory effect on the six miRNAs (Fig. 4b, c and Fig. S3b, c). Interestingly, only miR-361-5p was significantly downregulated by F. nucleatum and this downregulation could be reverted by 1 mM metformin in both HT-29 (Fig. 4b) and HCT 116 cells (Fig. 4c). Although F. nucleatum inhibited miR-616-3p expression, the decreased miR-616-3p expression could not be rescued by 1 mM metformin in either CRC cell line (Fig. S3b, c). The other miRNAs were not significantly regulated by either F. nucleatum or metformin (Fig. S3b, c). Thus, these data implied that miR-361-5p might contribute to the downregulation of sonic hedgehog signaling in response to 1 mM metformin in F. nucleatum-infected CRC cells.
We next examined whether miR-361-5p could suppress sonic hedgehog signaling and stemness in CRC cells. Targetscan was used to predict the binding site of miR-361-5p in the 3′ untranslated region (UTR) of GLI1. Position163-184 of the GLI1 3′ UTR was identified (Fig. 4d). To verify whether the 3′ UTR of GLI1 mRNA was a functional target of miR-361-5p, we performed dual luciferase reporter gene assays. The 3′ UTR sequences of GLI1 (3′ UTR wild-type, WT) and the mutant 3′ UTR sequences of GLI1 (3′ UTR mutant, MUT) were cloned into a luciferase reporter vector, separately (Fig. 4d). The dual luciferase assay demonstrated that miR-361-5p mimics suppressed the luciferase activity and miR-361-5p inhibitors elevated the luciferase activity in HCT 116 cells transfected with the WT GLI1 reporter plasmid, but not with the MUT reporter plasmid (Fig. 4e). These findings indicated that GLI1 is the specific target of miR-361-5p.
Then, to assess the function of miR-361-5p on the sonic hedgehog pathway and stemness, we transfected miR-361-5p mimics and inhibitors into HT-29 and HCT 116 cells, separately. Overexpression of miR-361-5p decreased the expression levels of SHH, GLI1, and NANOG at both the mRNA and protein level. By contrast, inhibition of miR-361-5p increased the expression of SHH, GLI1, and NANOG in HT-29 cells (Fig. 4f, g) and HCT 116 cells (Fig. 4h, i). In addition, the number and size of tumorspheres were reduced in the miR-361-5p overexpressing HT-29 ( Fig. 4j and Fig.  S3d) and HCT 116 cells (Fig. 4k and Fig. S3e). Thus, these data support the view that miR-361-5p suppresses the sonic hedgehog pathway and subsequently inhibits CRC cell stemness.

Metformin reverts miR-361-5p in F. nucleatum-infected CRC cells via inhibiting MYC
We next investigated the potential mechanisms responsible for the upregulation of miR-361-5p by metformin in F. nucleatuminfected cells. MiRNA genes are transcribed into primary miRNA (pri-miRNA) transcripts; therefore, we first detected pri-miR-361 expression using real-time PCR. F. nucleatum inhibited the expression of pri-miR-361 in HT-29 and HCT 116 cells (Fig. 5a, b). Metformin increased the expression of pri-miR-361 in F.   Fig. 3 Metformin rescues F. nucleatum-induced stemness by inhibiting sonic hedgehog pathway. a, b Real-time PCR was performed to evaluate SHH and GLI1 expression in HT-29 (a) and HCT 116 (b) cells co-cultured with F. nucleatum or treated with 1 mM metformin for 24 h; unpaired t test. c SHH and GLI1 levels were detected by western blotting in HT-29 and HCT 116 cells. The cells were co-cultured with F. nucleatum or treated with 1 mM metformin for 24 h. d Western blotting was performed to detect sonic hedgehog pathway-related proteins in xenograft tumors after different treatments. n.s., P > 0.05; ***P < 0.001; ****P < 0.0001.
nucleatum-infected HT-29 and HCT 116 cells, but not in F. nucleatum-free cells (Fig. 5a, b). These data indicated that metformin and F. nucleatum might regulate the transcription of miR-361-5p. To identify potential transcription factors of miR-361-5p, we employed the TransmiR v2.0 database (http:// www.cuilab.cn/transmirwas) and predicted 48 transcription factors that might govern the expression of miR-361-5p directly (Fig. 5c)    were identified after overlapping the 48 potential transcription factors and the metformin-regulated genes (Fig. 5c). Among these four transcription factor genes, MYC was the most significantly downregulated gene by metformin and was well characterized to bind to miRNA promoters directly [32].
To determine whether MYC contributed to the recovery effect of metformin in F. nucleatum-inhibited miR-361-5p, we first detected the expression of MYC in CRC cells treated with metformin and F. nucleatum. The levels of the MYC transcript (Fig. 5d, e) and protein (Fig. 5f) were enhanced in response to F. nucleatum infection in HT-29 cells and HCT 116 cells. Metformin reduced the levels of MYC mRNA (Fig. 5d, e) and protein (Fig. 5f) in F. nucleatum-infected HT-29 cells and HCT 116 cells, but not in F. nucleatum-free CRC cells. Moreover, metformin failed to upregulate pri-miR-361 and miR-361-5p when MYC was overexpressed in F. nucleatum-infected HT-29 cells (Fig. 5g) and HCT 116 cells (Fig. 5h). Collectively, the data suggested that metformin decreases miR-361-5p transcription in F. nucleatum-infected cells by inhibiting MYC.

DISCUSSION
Chemotherapeutic agents, such as 5-Fu and oxaliplatin, remain the backbone of treatment for patients with CRC; however, the development of chemoresistance is the major cause for treatment failure. Cancer chemoresistance is a complex process and results from the interplay between intrinsic and extrinsic factors. Tumors with genetic and epigenetic alterations are critical for the CRC chemotherapeutic response [6]. Recent studies also showed that the gut microbiota controlled the response to chemotherapy by modulating the tumor microenvironment [33]. F. nucleatum abundance is associated with the properties of CRC, such as tumorigenesis, development, metastasis, and recurrence [34]. Through a combination of bioinformatic analyses, biological experiments, in vivo models, and clinical studies, we demonstrated that abnormal proliferation of F. nucleatum led to CRC chemoresistance and recurrence [15]. Given that broad spectrum antibiotics have a negative effect on the healthy intestinal microbiota, and no F. nucleatum-specific antimicrobial agent has been discovered, chemoresistance caused by F. nucleatum remains a thorny problem in the clinic [35]. Therefore, it is necessary to develop a safe and effective approach to eliminate chemoresistance caused by F. nucleatum.
Based on our previous work showing that metformin could attenuate F. nucleatum-induced tumorigenesis [16], we further explored the effect of metformin on F. nucleatum-induced chemoresistance. To our surprise, metformin abrogated F. nucleatum-induced CRC chemoresistance in CRC cells and xenograft mice. However, how metformin affects F. nucleatummediated chemoresistance was unknown. A distinct tumor cell subpopulation with stemness, known as the cancer stem cell population, exists in cancers, which mediates chemoresistance and metastatic progression [36,37]. The use of cancer stemness inhibitors, such as napabucasin, can overcome chemoresistance in Fig. 4 miR-361-5p inhibited the sonic hedgehog signaling pathway and stemness. a Schematic illustration of the target miRNA candidate screening process. b, c Expression of miR-361-5p was quantified by real-time PCR in HT-29 (b) and HCT 116 cells (c). The cells were co-cultured with F. nucleatum or treated with 1 mM metformin for 24 h; unpaired t test. d The predicted binding sequences for miR-361-5p within the human GLI1 3′ UTR. Seed sequences are highlighted. e Luciferase activity was measured in HCT 116 cells transfected with miR-316-5p mimics or control mimics, miR-316-5p inhibitors or control inhibitors for 72 h. The luciferase reporters expressing wild-type or mutant human GLI1 3′ UTRs were used; unpaired t test. f, g Real-time PCR (f) and western blotting (g) were performed in HT-29 cells to detect the expression levels of SHH, GLI1, and NANOG after transfection with miR-361-5p mimics or inhibitors for 48 h; unpaired t test. h, i Real-time PCR (h) and western blotting (i) were performed in HCT 116 cells to detect the expression levels of SHH, GLI1, and NANOG after transfection with miR-361-5p mimics or inhibitors for 48 h; unpaired t test. j, k Tumorspheres were observed by light microscopy in HT-29 (j) and HCT 116 (k) cells to detect the tumorsphere formation capability after transfection with miR-361-5p mimics; the number of the tumorspheres were quantified in the right panel. Scale bar, 400 μm; unpaired t test. n.s., P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.
human cancers [38][39][40]. Emerging evidence supports the view that bacterial infections, such as Enterococcus faecalis, nonpathogenic E. coli, and Enterotoxigenic Bacteroides fragilis, might stimulate cancer cell stemness via upregulated expression of stemness markers [41][42][43]. Our bioinformatic studies demonstrated that a stemness-related pathway is enriched in the modulation of metformin and F. nucleatum in CRC cells. Indeed, the effect of metformin depends on the specific regulation of stemness in F. nucleatum-infected CRC cells. Accordingly, metformin decreased the expression of stemness-related proteins, SOX2 and NANOG, and inhibited tumorsphere formation in CRC cells co-cultured with F. nucleatum. Thus, we concluded that metformin attenuates F. nucleatum-induced chemoresistance by inhibiting stemness in CRC cells. These data might explain why metformin recovers chemosensitivity in F. nucleatum-infected CRC cells and xenograft mice.
We dissected the mechanisms by which metformin regulates F. nucleatum-induced stemness in CRC cells. Aberrant activity of the sonic hedgehog pathway has been linked to stem cell self-renewal and chemoresistance in variety of solid neoplasms [44]. We demonstrated that sonic hedgehog signaling is a bridge connecting metformin-and F. nucleatum-modulated stemness and     addition, metformin reverses F. nucleatum-stimulated stemness, sonic hedgehog signaling and chemoresistance in a miR-361-5pdependent manner. In line with this notion, we found that metformin decreases miR-361-5p transcripts in F. nucleatuminfected cells by downregulating MYC expression. Our data suggested that metformin reverses F. nucleatum-induced stemness by inhibiting the MYC/miR-361-5p/sonic hedgehog signaling axis, and then biologically and mechanistically reverting CRC chemoresistance induced by F. nucleatum.
In addition to its biological importance, our work might be relevant in the clinical management of patients with CRC. The abundance of F. nucleatum is associated with the risk of CRC chemoresistance and recurrence; therefore, combining conventional chemotherapeutic regimens with metformin might be an effective strategy to reverse CRC chemoresistance in patients with high amounts of F. nucleatum. Our results highlighted the need for a clinical trial of metformin as a potential treatment in Fusobacterium-associated CRC chemoresistance. Furthermore, an important question raised by our data is whether metformin directly affects F. nucleatum. It has been reported that metformin exerts its hypoglycemic benefits partly through alteration of the abundance of certain members of the gut microbiota, such as Bacteroides fragilis and Akkermansia muciniphila [45][46][47][48]. Researchers also found that metformin could regulate microbial folate and methionine metabolism, which is required for bacterial growth [46,49]. These provide the possibility that patients with CRC might benefit from the direct modulation of metformin on F. nucleatum. In addition to the details of the mechanisms by which metformin affects F. nucleatum-induced host responses, we will explore the direct effect of metformin on F. nucleatum colonization and dissemination in a future study.
Our results provide a foundation for the metformin-gut microbiota-host response network in CRC. This network regulation mode might be a direction for meaningful basic and translational medicine research in the future.

DATA AVAILABILITY
Previously published data sets were available in GEO under accession code GSE102573, GSE90944, and GSE67342. Fig. 6 Metformin reverses F. nucleatum-induced stemness and chemoresistance via the miR-361-5p/sonic hedgehog axis. a, b HT-29 (a) and HCT 116 (b) cells were transfected with mimics or inhibitors of miR-361-5p. After culturing with F. nucleatum for 24 h, SHH, GLI1 and NANOG levels were detected by western blotting. c, d Tumorspheres were observed by light microscopy in HT-29 (c) and HCT 116 (d) cells transfected with miR-361-5p mimics, and then co-cultured with F. nucleatum. The number and size of the tumorspheres were quantified in the lower panel. Scale bar, 400 μm; unpaired t test. e, f Western blotting was performed to detect the levels of SHH, GLI1, and NANOG. After transfection with mimics or inhibitors of miR-361-5p, HT-29 (e) and HCT 116 (f) cells were treated with 1 mM metformin in the presence of F. nucleatum for 24 h. g, h After transfection with miR-361-5p mimics, HT-29 (g) and HCT 116 (h) cells were treated with 1 mM metformin in the presence of F. nucleatum. Tumorspheres were observed by light microscopy in HT-29 (g) and HCT 116 (h) cells. The number and size of the tumorspheres were quantified in the right panel. Scale bar, 400 μm; unpaired t test. i Representative data of the tumors in mice under different conditions. j, k Statistical analysis of tumor volumes (j) and weights (k) in the different groups, n = 5/group; unpaired t test. Values indicate mean ± standard error of the mean (SEM); unpaired t-test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.