Fusobacterium nucleatum Load Correlates with KRAS Mutation and Sessile Serrated Pathogenesis in Colorectal Adenocarcinoma

Colorectal cancer is the third most common cancer in incidence and is the second cause of cancer death worldwide. In 2020, it was estimated that 1.9 million new cases would be diagnosed, and 935,000 persons would experience death (1). Despite increased use of screening and improvements in treatment, the burden of this disease is still high (2). Attributable risk for colorectal cancer includes unmodifiable factors such as sex, age, race, and inherited gene mutations, and modifiable factors such as smoking, unhealthy diet, excess alcohol intake, obesity, and lack of physical activity (3). In the United States, more than half of all colorectal cancer cases are attributed to modifiable factors (2).

Because the discovery of abundant presence of Fusobacterium nucleatum (Fn) in colorectal cancers in 2012, evidence that the makeup of the gut microbiome contributes as a risk factor for colorectal cancer and influences the initiation and progression of colorectal cancer has been accumulating (4–6). Several bacterial species including Fn, Enterotoxigenic Bacteroides fragilis (ETBF), and colibactin-producing Escherichia coli are associated with colorectal cancer in epidemiologic studies, found to be enriched in colorectal cancer tissues, and facilitate colorectal tumors in preclinical models (6, 7). Among these species, Fn has been detected in most colorectal adenoma/carcinoma tissues examined by studies using 16s rRNA sequencing or metagenomic sequencing (8). Particularly, Fn infection is associated with a subgroup of colorectal cancers that are located on the right side of the colon, exhibit the CpG island methylation phenotype (CIMP)-high, and exhibit high levels of microsatellite instability (MSI-H; refs. 9, 10). Recently, by analyzing the sequence datasets from The Cancer Genome Atlas (TCGA) and the European Genome-Phenome Archive (11), Ternes and colleagues reported that Fn infection is associated with colorectal cancer belonging to consensus molecular subtypes (CMS) 1 and 3. While the data reported by Ternes and colleagues agree with previous results where Fn infection was associated with MSI-H colorectal cancer that consists of 74% of CMS1 colorectal cancer in TCGA cohort, the association of Fn infection with CMS3 colorectal cancer has not been reported previously (11–13). CMS3 colorectal cancers are characterized as being metabolically deregulated and enriched in KRAS mutations (70% of CMS3 in TCGA cohort; ref. 12). Although several studies demonstrated that MSI-H, BRAF mutation, MLH1 hypermethylation, or CIMP-high, characteristics shared with CMS1 colorectal cancer were associated with high levels of Fn infection, none of these studies showed any association between Fn infection and KRAS mutations (14–20). On the other hand, other studies have shown that Fn infection is associated with KRAS mutations in colorectal cancer (21–24). These discrepancies among previous studies could be due to the difference in methods to detect and quantitate Fn, and/or difference in cut-off levels to discriminate Fn-positive and Fn-negative (21).

We previously reported that Fn directly caused DNA damage in infected tissues and its infection was enriched in two subgroups of colorectal cancers, one exhibiting MSI-H and another exhibiting low levels of microsatellite instability (MSI-L) and elevated microsatellite alterations in selected tetranucleotide repeats (EMAST; L/E; ref. 10). We have shown that L/E exhibited by colorectal cancer is induced by dysfunction of the mismatch repair protein, MSH3, and is associated with tissue inflammation (25, 26). These results suggest that Fn infection might contribute to colorectal carcinogenesis not only by affecting the cellular DNA but also the tumor microenvironment. Regarding the status of BRAF mutations, MLH1 hypermethylation and KRAS mutations and their relationship to Fn infection, there are conflicting results exists among previous studies (14–22). Because clinical treatment of KRAS-mutated colorectal cancer is different from KRAS wild-type colorectal cancer, it is important to determine whether this group of colorectal cancers is associated with Fn infection that may negatively affect prognosis and the efficacy of chemotherapy (23, 24, 27). In this study, we aimed to determine whether and how KRAS mutations are associated with Fn infection compared with the association of BRAF mutation, MLH1 hypermethylation and MSI-H with Fn using a previously characterized colorectal cancer cohort. We also aimed to determine whether KRAS, BRAF mutations, and/or MLH1 hypermethylation is associated with Fn infection in colon adenomas.

The colorectal cancer cohort used in this study has been described previously (10) and consists of 91 unselected patients with cancers of rectal, sigmoid, or rectosigmoid junction who participated in the North Carolina Colon Cancer Study-Phase II [(NCCCS II) NC Rectal Cancer Study (IRB#99-0933)]. Patients were 40 to 79 years of age, resided in central North Carolina and diagnosed between 2001 and 2006 (28). An additional 213 patients with unselected sporadic colorectal cancer were obtained from the North Carolina site for the Cancer Care Outcomes Research and Surveillance consortium (CanCORS). CanCORS was a population-based prospective, case-only, multisite observational study of patients with colorectal and lung cancer (29). Those patients were at least 21 years of age, residents of central North Carolina, and diagnosed between 2003 and 2006. The study was approved by the University of North Carolina's Institutional Review Board (IRB# 04-0860). All data associated with the 304 total cases of colorectal cancer are presented in Supplementary Table S1. Thirty-two adenoma cases including sessile serrated adenoma (SSA, 10 cases) and tubular adenoma (TA)/tubular-villous adenoma (TVA; 22 cases) were used in this study and have been described previously (30). Fresh adenomas were collected during colonoscopies performed between 2014 and 2017 at Cremona Hospital (Italy) or Zurich Triemli Hospital (Switzerland) with approval of both hospitals’ research ethics committees, and DNA was extracted with AllPrep Mini Kit (QIAGEN). Tissues were histologically classified according to World Health Organization criteria. All data associated with the 32 cases of adenoma/polyp are shown in Supplementary Table S2. All tissues in each of the above cohorts were obtained via written informed consent from patients under IRB approval and all studies were conducted in accordance within recognized international ethical guidelines.

Paired normal and tumor genomic DNA from all colorectal cancers were prepared from formalin-fixed paraffin-embedded (FFPE) tissues using the QIAamp DNA FFPE Tissue Kit (QIAGEN). The MSI/EMAST assay was carried out as described previously (31). Briefly, the assay consists of 14 microsatellite markers in four reactions capable of determining MSI-H, MSI-L, EMAST, and microsatellite stable (MSS) simultaneously. The markers include two mononucleotide (BAT25 and BAT26), five dinucleotide (D2S123, D5S346, D17S250, D18S64, and D18S69), and seven tetranucleotide microsatellite sequences (D9S242, D20S82, D20S85, D19S394, D8S321, MYCL1, and RBM47). We defined MSI-H, MSI-L/EMAST (L/E), and MSS as described previously (32).

The method for detection and quantification of Fn DNA in colorectal cancer has been described before (10). Briefly, the absolute quantity of Fn DNA and that of tumor DNA in each sample was determined separately by the SYBER green–based standard curve method. The PCR primer set specific to Fn was designed to target the nusG gene of Fn (5) and the primer set specific to the human genome targets the non-gene coding region of human chromosome 9p24 (chr9:242200-2242300; GRCH38/hg38). Each reaction contained 2.5 ng of DNA, 1X Power SYBER Green Master MIX (Applied Biosystems), and 250 nmol/L of each primer and all samples were assayed in duplicate in 10 µL reactions. Amplification and detection of DNA were carried out using the ABI 7900HT Sequence Detection System (Applied Biosystems). The primer sequences for Fn and human genome at 9p24 were described before (10). Specificity of PCR products for Fn and human 9p24 region was monitored by comparing melting curves generated from reference Fn DNA (Fn strain VPI4355, ATCC) and genomic DNA from the human colon cancer cell line, DLD1. We also used the TaqMan-based standard curve method for detection and quantification of Fn DNA in adenoma/polyp tissues. The PCR primer and FAM probe sequences were as follows: Fn forward primer, 5′-GCTTGAAATGGAAGCTACAAGAG; Fn reverse primer, 5′-GGATCA GAACCAACTCCTACAA; Fn-FAM probe: 5′-AGTAGACCCTCGTGTATG. PCR conditions were 95°C for 10 minutes, followed by 45 cycles of 95°C for 15 seconds and 54°C for 1 minute. The reaction mixture consisted of TaqMan Universal Master Mix (Applied Biosystems), 300 nmol/L forward, 900 nmol/L reverse primers, 250 nmol/L Fn-FAM probe and approximately 50 ng of template DNA. For quantifying tumor DNA in adenoma samples, we used the SYBER green–based standard curve method as used for colorectal cancer samples.

Sensitivity of SYBER green–based standard curve method and TaqMan-based standard curve method to detect Fn DNA was examined as follows: We generated a standard curve using the same amount of template Fn genomic DNA (4-fold dilution starting from 1,000 ng/reaction) isolated from the Fn strain VPI4355 (ATCC) for each assay (Supplementary Fig. S1). In SYBER green–based assay, the slope of the reaction was −3.272 and amplification efficiency was 102% whereas the slope and amplification efficiency of the TaqMan-based assay was −3.62 and 89%, respectively. There was a about 2-fold difference (average: 2.29) in CT value at each DNA dilution point between SYBER green–based and TaqMan-based assay, indicating that TaqMan assay is approximately four times less sensitive for detecting Fn DNA. To compensate the different sensitivity, we included 10 times more adenoma DNA (50 ng) in TaqMan PCR reaction compared with colorectal cancer DNA (5 ng) in SYBER green assay. Fn DNA content in each sample was determined by the following formula: absolute Fn DNA in picogram/absolute tumor genomic DNA in nanograms. On the basis of the fact that the total Fn genome contains 2,170 kilobase pairs (33), the number of copies of Fn per nanogram of tumor DNA was calculated from Fn DNA content.

Mutational hot spots on KRAS codons 12 and 13, and on BRAF codon 600 involving valine to glutamic acid were investigated by PCR-direct sequencing. PCR reactions on extracted DNA were performed with the Q5 High-Fidelity master mix (New England Biolaboratories) in a 96-well thermal cycler (Applied Biosystems). The primers used were: KRAS forward primer: 5′-GGTACTGGTGGAGTATTTGATAGTG-3′, KRAS reverse primer: 5′-ACCTCTATTGTTGGATCATATTCGT-3′, BRAF forward primer: 5′-TGCTTGCTCTGATAGGAAAATG-3′, and BRAF-reverse primer: 5′-AGTAACTCAGCAGCATCTCAGG-3′.

The PCR products were purified by ExoSap-IT (Applied Biosystems) and sent to an outsource vender (Eurofins) for cycling sequencing.

The DNA isolated from tumor tissues or cell lines (∼500 ng) was modified with sodium bisulfite using EZ DNA Methylation-Gold Kits (D5005, Zymo Research). Methylation-specific PCR (MSP) was used for detecting promoter methylation of the hMLH1 locus (34). The sequences of the methylated-specific and unmethylated-specific primer pair and PCR cycling conditions were the same as described previously (34), except that AmpliTaq Gold 360 master mix (Applied Biosystems) was used for amplification. The PCR products were separated on 3% MetaPhor agarose (Lonza), stained with GelRed Nucleic Acid Stain (Biotium), and then visualized with UV illumination using a digital imaging system (ImageQuant LAS 4000, GE Healthcare). As a positive control for the MLH1 promoter hypermethylation, bisulfite DNA from the human colon cancer cell line RKO was used. As a negative control, bisulfite DNA from SW480 cells was used. If the band intensity of the methylation-specific PCR products was equal to or greater than that of the unmethylated PCR products on the gel image, the sample was defined as methylation-positive. To confirm the results obtained by MSP analysis, MLH1 IHC staining was performed on the samples that were positive for MLH1 MSP. Paraffin-embedded tissues were deparaffinized and rehydrated. After antigen retrieval (121°C for 15 minutes in 0.01 mol/L citrate buffer, (pH 6.0), the tissues were treated overnight at 4°C with anti-human MLH1 mouse antibody (1:400, catalog no.: 550838, BD Biosciences), followed by visualization of the MLH1 signal with ImmPRESS Universal PLUS Polymer Kit (MP-7800, Vector Laboratories)

All statistical analyses were carried out using the software XLSTAT (Addinsoft). The association between Fn infection and each genomic variable, and other variables, including sex, age, tumor location and stage, in colorectal cancer cohort was tested using a logistic regression model with Firth bias correction. The association between Fn infection and various variables in adenoma/polyp cohort was tested using Fisher exact test. The Mann–Whitney test and Kruskal–Wallis test were used to determine the difference in the amounts of Fn DNA present in different groups of colorectal cancers. The Kaplan–Meier test was used to determine whether Fn infection affected patients’ 5-year overall survival (OS) rate. When P values were less than 0.05, the difference or association was determined significant.

The data generated in this study are available upon request from the corresponding author.

We previously examined 304 cases of unselected colorectal cancer derived from North Carolina for determining the relationship between Fn infection and molecular subtypes of colorectal cancer including MSI-H, L/E, and MSS (10). Among the 304 cases previously used, five cases (including one case with MSI-H, two cases with L/E, and two cases with MSS) had depleted template DNA. Therefore, these cases were replaced by new cases of MSI-H (one case), L/E (two cases), and MSS (two cases) from the CanCORS cohort. In this revised cohort, 45.1% (137/304) were MSS, 42.4% (129/304) were L/E, 12.5% (38/304) were MSI-H, and 87.5% (266/304) of colorectal cancer non-MSI-H (Table 1; Supplementary Table S1).

Twenty-five of 304 cases (8.2%) exhibited BRAFV600E mutation. MLH1 hypermethylation was also detected in 8.2% (25/304) of colorectal cancer. A total of 101 of 304 cases (33.2%) showed mutations at KRAS G12/G13 (Table 1; Supplementary Table S1). Furthermore, the cohort exhibited the following clinicopathologic characteristics: tumor site (71 rectum and 233 colon cancers); tumor stage (85 localized and 173 regional/distant cancers); patients’ age (150 cases were ≤65 years and 152 cases were >65y); sex (143 female and 161 male); and race (235 White and 63 Black; Table 1).

Among 304 colorectal cancer cases, Fn was detected in 109 cases (Supplementary Fig. S2; Table 1). The minimum amounts of Fn DNA detected in our colorectal cancer cohort were 0.02 pg (equivalent to one copy of Fn genome) per one nanogram of tumor DNA and the maximum amounts were 291 pg (equivalent to 124,467 copies of Fn) per one nanogram of tumor DNA, respectively (Supplementary Fig. S2; Supplementary Table S1).

As was true for prior results, more colorectal cancer with L/E (L/E-CRC) was infected with Fn than colorectal cancer with MSS (MSS-CRC) by univariate logistic regression analysis (OR: 1.9, P = 0.01; Table 1). Significantly more cases of colorectal cancer with MSI-H (MSI-H-CRC) were infected with Fn compared with MSS-CRC (OR: 4.34, P < 0.001) or L/E-CRC (OR: 2.28, P = 0.03; Table 1). When all cases were categorized into MSI-H-CRC and colorectal cancer with non-MSI-H (non-MSI-H-CRC), a higher fraction of the MSI-H cases (60.5%: 23/38) were infected with Fn compared with that of the non-MSI-H-CRC cases (32.3%: 86/266, OR: 3.12, P = 0.002; Table 1). Regarding BRAF mutations, 56% (14/25) of colorectal cancer with BRAF V660E mutation (BRAFV600E-CRC) were infected with Fn whereas significantly fewer of the colorectal cancers with wild-type BRAF (non-BRAFV600E-CRC; 34%: 95/279) were infected with Fn (OR: 2.39, P = 0.04; Table 1). In regards to MLH1 promoter hypermethylation, 25 cases of colorectal cancer exhibited hypermethylation. We performed MLH1 IHC staining for 20 out of 25 cases that exhibited MLH1 promoter hypermethylation and eight cases that were negative for MLH1 hypermethylation. All 20 colorectal cancer cases with MLH1 hypermethylation (MLH1 hypermethylated-CRC) lost MLH1 expression, whereas the eight colorectal cancer cases with unmethylated MLH1 (non-MLH1 hypermethylated-CRC) expressed MLH1 (Supplementary Table S1), indicating that MLH1 hypermethylation determined by our MSP assay accurately reflects silencing of the MLH1 expression. Sixty percent of cases with MLH1 hypermethylation (15/25) were infected with Fn while a significantly lower percentage (33.7%, 94/279) of colorectal cancer with unmethylated MLH1 were infected with Fn (OR: 2.84, P = 0.016; Table 1). Regarding KRAS G12/G13 mutations, 44.6% (45/101) of colorectal cancer with KRAS mutations (KRAS mutated-CRC) were infected with Fn. Compared with colorectal cancer without KRAS mutations (non-KRAS mutated-CRC), a significantly higher percentage of cases were infected with Fn (OR: 1.74, P = 0.02; Tables 1 and 2). Comparing colon with rectum, the colon was more significantly infected with Fn (OR: 2.5, P = 0.002; Table 1). Other variables including tumor stage, patients’ age, sex, and racial status were not associated with Fn infection (Table 1).

We next determined independent factor(s) associated with Fn infection using multivariate logistic regression modeling (Table 2). Variables including MSI-H, MLH1 hypermethylation, and BRAF mutation revealed severe multicollinearity when measured by variance inflation factor (VIF; VIF for MSI-H was 2.69, 6.41 for MLH1 hypermethylation, and 4.55 for BRAF mutation); we constructed four independent models where these three variables did not overlap. In Model 1, MSI-H was associated with Fn infection compared with MSS [OR: 4.6, 95% confidence interval (CI): 2.0–10.5, P < 0.001] and with L/E (OR: 1.7, 95% CI: 1.01–2.95, P = 0.045). KRAS mutations (OR: 1.8, 95% CI: 1.1–3.1, P = 0.03) and regional/advanced stage (OR: 1.8, 95% CI: 1.03–3.3, P = 0.04) were also associated with Fn infection, when compared with non-KRAS mutation and local stage, respectively (Table 2). Importantly, in all four models, KRAS mutation was independently associated with Fn infection (Table 2). Finally, Fn infection was associated with MSI-H (vs. non-MSI-H, OR: 3.4, 95% CI: 1.6–7.3, P = 0.002 in Model 2), MLH1 hypermethylation (OR: 3.5, 95% CI: 1.4–8.8, P = 0.008 in Model 3) or BRAF mutation (OR: 3.2, 95% CI: 1.2–8.2, P = 0.016 in Model 4; Table 2). Taken together, these results showed that MSI-H, BRAF mutations, MLH1 hypermethylation, and KRAS mutations are independently associated with Fn infection in colorectal cancer.

To see any difference in Fn loads among subgroups of colorectal cancers, we compared the copy number of Fn among colorectal cancers with L/E, MSS, MSI-H, non-MSI-H, MLH1 hypermethylation, non-MLH1 hypermethylation, BRAFV600E, non-BRAFV600E, KRAS mutation, and non-KRAS mutations using the Kruskal–Wallis test (Fig. 1A). The number of Fn copies in each sample was converted to log value before comparison. As shown in Fig. 1A, Fn loads were the highest in MLH1 hypermethylated-CRC. MSI-H-CRC and BRAFA600E-CRC had the second and third highest Fn loads, respectively. KRAS mutated-CRC and L/E colorectal cancer had the fourth highest Fn loads, and MSS-CRC had the lowest Fn loads. As reported previously (10), MSI-H-CRC had higher Fn loads than L/E-CRC (P = 0.029) and MSS-CRC (P < 0.0001), and L/E-CRC had higher Fn loads than MSS-CRC (P < 0.0001). Here, MSI-H-CRC, MLH1 hypermutated-CRC, BRAFV600E-CRC and KRAS mutated-CRC had higher Fn loads compared with non-MSI-H-CRC (P < 0.0001), non-MLH1 hypermutated-CRC (P = 0.001), non-BRAFV600E-CRC (P = 0.022) and non-KRAS mutated-CRC (P = 0.037), respectively (Fig. 1A). Also, the difference in Fn loads between MLH1 hypermethylated-CRC and KRAS mutated-CRC (P = 0.042) and between MSI-H-CRC and KRAS mutated-CRC (P = 0.029) was significant.

We then compared Fn loads only in Fn-infected colorectal cancers from each group. As shown in Fig. 1B, there was no significant difference in the copy number of Fn among MSI-H, L/E, and MSS-CRC. There was also no difference in the copy number of Fn between (i) MSI-H-CRC and non-MSI-H-CRC (P = 0.088); (ii) BRAFV600E-CRC and non-BRAFV600E-CRC (P = 0.655); and (iii) KRAS mutated-CRC and non-KRAS mutated-CRC (P = 0.692). In contrast, Fn-infected MLH1 hypermethylated-CRC contained more Fn DNA compared with Fn-infected non-MLH1 hypermutated-CRC (P = 0.033; Fig. 1B). Taken together, and as shown in Fig. 1A, Fn colonizes in different subgroups of colorectal cancers with differing efficiency but grows at a similar rate once it establishes colonization, except that Fn infects with and grows more efficiently in MLH1 hypermethylated-CRC than in other groups of colorectal cancers as shown in Fig. 1B. Note that Fn colonizes less efficiently in KRAS mutated-CRC compared with MLH1 hypermethylated-CRC or MSI-H-CRC; however, it colonizes more efficiently in KRAS mutated-CRC compared with non-KRAS mutated-CRC (Fig. 1A).We further examined the relationship between each genotype including MSI-H, MLH1 hypermethylation, BRAF mutations, and KRAS mutations (response variables) and copy number of Fn (explanatory variable) using logistic regression analysis. As shown in Table 3, increasing copy number of Fn is associated with a higher probability of colorectal cancer with MLH1 hypermethylation (OR: 1.56, P < 0.0001), MSI-H (OR: 1.5, P < 0.0001), and BRAFV600E (OR: 1.33, P = 0.027), but not with KRAS mutations (OR: 1.15, P = 0.113) in univariate logistic regression analysis. After adjusting for tumor site and stage, and patients’ age, sex, and race, a higher copy number of Fn is independently associated with MLH1 hypermethylation (OR: 1.46, P < 0.0001), MSI-H (OR: 1.5, P < 0.0001), slightly with BRAFV600E (OR: 1.21, P = 0.051) but not with KRAS mutations (OR: 1.15, P = 0.142; Table 3). These results suggest that interaction between Fn and MLH1 hypermethylated-CRC or MSI-H-CRC, and interaction between Fn and KRAS mutated-CRC, are biologically different. It could be that increasing the copy number of Fn in precursor adenomas with the BRAFV600E mutation contributes to promoting hypermethylation of the prompter region of the MLH1 locus, resulting in MSI-H, leading to transition of adenoma to carcinoma. On the other hand, Fn may not be directly involved in KRAS-driven carcinogenesis; it may merely colonize more efficiently in KRAS mutated-CRC. We also examined the relationship between copy number of Fn and MSI-H, L/E, and MSS. It was seen that an increasing number of Fn DNA was accompanied by an increased probability of being MSI-H from L/E (OR: 1.35, P = 0.004) or from MSS (OR: 1.81, P < 0.0001) and was also accompanied by an increased probability of being L/E from MSS (OR: 1.3 P = 0.012; Supplementary Table S3). These results suggest that Fn infection may contribute to not only inducing MSI-H-CRC through MLH1 hypermethylation but also inducing L/E-CRC through dysfunction of MSH3 (9). Alternatively, Fn more efficiently colonizes MSI-H-CRC than it does L/E-CRC or MSS-CRC, and colonizes better in L/E-CRC than MSS-CRC.

We then examined whether Fn infection has any effects on the 5-year OS rate of patients with colorectal cancer. Although patients with Fn-infected colorectal cancer exhibited a shorter 5-year OS rate compared with patients with non–Fn-infected colorectal cancer, the difference was not significant (P = 0.45, log-rank test; Supplementary Fig. S3A). We also found no significant difference in 5-year OS rate between infected and noninfected patients with colorectal cancer with MSI-H (P = 0.24), with MLH1 hypermethylation (P = 0.74), with BRAF mutations (P = 0.41) and with KRAS mutation (P = 0.9; Supplementary Fig. S3B–S3E). There results suggest that Fn infection might not be a prognostic factor for 5-year OS of patients with colorectal cancer.

We next determined whether Fn infection was associated with BRAF mutation, KRAS mutations or MLH1 hypermethylation in adenomas/polyps. In this preliminary small study, 32 adenomas examined were dissected from the colon (Supplementary Table S2). Among the 32 adenomas, Fn DNA was detected in 14 adenomas, ranging from 0.003 to 7.3 pg per 1 ng of tumor DNA (mean DNA content: 0.28 pg/ng of tumor DNA; Supplementary Table S2). Compared with the amount of Fn DNA detected in colorectal cancer tumor tissues (mean DNA content: 5.49 pg/ng of tumor DNA; Supplementary Table S1), the Fn DNA detected in adenomas was significantly less (P = 0.003) by the Mann–Whitney test. Four of 10 SSA (40%) and 10 of 22 TA/TVA (45%) were positive for Fn infection. There was no association between Fn infection and adenoma types and histologic dysplasia (Table 4). BRAF mutations were found in nine of 10 SSA and one in 10 TA and were not associated with Fn infection (Supplementary Table S2; Table 4). KRAS mutations were found in one in 10 SSA, four in 14 TA, and four in eight TVA, and there was no association between KRAS mutations and Fn infection (Supplementary Table S2; Table 4). There was one SSA that was positive for MLH1 promoter hypermethylation and had no association with Fn infection (Table 4). Finally, there was no association of Fn infection with age or sex (Table 4). These results suggest that an association between Fn infection and BRAF/KRAS mutations or MLH1 hypermethylation may not be established at the adenoma/polyp stage.

One of the main findings presented here is that incidence of Fn infection is significantly high in colorectal cancer exhibiting not only SSP phenotype including MSI-H, MLH1 hypermethylation or BRAF mutations but also KRAS mutations (Tables 1 and 2). A unique finding of our study is that the mode of Fn infection differs between colorectal cancers with the SSP phenotype and those with KRAS mutations. The quantity of Fn in colorectal cancer with MSI-H or MLH1 hypermethylation is higher than in colorectal cancer with KRAS mutations (Fig. 1A). Furthermore, increasing loads of Fn are associated with MSI and MLH1 hypermethylation, suggesting that Fn may directly or indirectly cause hypermethylation of the MLH1 locus, leading to MSI-H (Table 3). On the other hand, Fn has a stronger affinity to KRAS-mutated colorectal cancer than to non–KRAS-mutated colorectal cancer, but the increased load of Fn is not associated with KRAS mutations. This suggests that Fn may not play a role in generating mutations in the KRAS gene (Fig. 1A; Table 3). In contrast to many studies, our results did not detect any impact of Fn infection on the 5-year OS rate (Supplementary Fig. S3).

In previous studies, there have been conflicting results regarding the effect of Fn infection on patients’ prognoses (refs. 17, 19, 20, 21, 35–42; Supplementary Table S4) and on the association of Fn infection with various molecular characteristics of colorectal cancer (refs. 10, 14–22, 35–37, 40, 42–46; Supplementary Table S4). To compare our results with these studies, we summarized 22 studies that explored the relationship between Fn infection and prognosis or molecular characteristics of colorectal cancer in Supplementary Table S4. As shown in the table, factors including biospecimens [fresh frozen (FF), FFPE, and methacarn], Fn detection/quantification methods (TaqMan, SYBAR green, Sequencing), number of cases examined, and difference in comparison (Fn-high vs. Fn-low/negative or Fn-positive versus Fn-negative) may have influenced the outcome of each study.

We divided these studies into two groups, A and B. Eleven studies in Group A used the TaqMan-based qPCR assay originally described by Castellarin and colleagues (5, 14–19, 36, 42–45). In this original article, nucleotide sequences of the probe overlapped with that of the PCR forward primer. Later, it was reported by Repass and colleagues that the probe sequence from the original article was incorrect and a different and new correct probe was used in their reproducibility experiments (47). The results obtained in the studies using the incorrect Fn probe may deserve further investigation for reproducibility. In contrast, the other 11 studies in Group B used various regimens to detect and quantify Fn loads (refs. 10, 20–22, 35, 37–41, 46; Supplementary Table S4). Here, we compare our results further with those of the studies listed in Group B.

It was reported that differences in biospecimen, FF/methacarn-fixed versus FFPE, made a difference in the detection and quantification of Fn DNA by PCR. Lee and colleagues showed that Fn was detected in 41% of FFPE colorectal cancer samples whereas Fn was detected in 100% of matched methacarn-fixed tissues from the same patients. They also showed that Fn was detected in 10 of 10 FF tissues examined whereas Fn was detected in only 12% of FFPE tissues (40). In our previous study, Fn was detected in 75% of FF tissues while it was detected in 38% of FFPE samples (10). These results indicate that FF or methacarn-fixed tissues are superior to FFPE tissues for assessing Fn DNA present in the original tissues. As shown in Supplementary Table S4, eight studies used FF tissues, one study used methacarn-fixed tissues while our current study and two other studies used FFPE tissues. Regardless of the various detection methods used, all six studies that examined FF or methacarn-fixed colorectal cancer tissues found a significantly negative impact of high loads of Fn on OS and/or recurrence-free survival (RFS) of patients with colorectal cancer (refs. 20, 21, 35, 37, 38, 40; Supplementary Table S4). Although the study using FFPE tissues by Yan that showed a high rate of Fn detection and significantly worse impact of high loads of Fn on patients’ cancer-specific survival and RFS (39), our study and that of Bundgaard-Nielsen, which examined FFPE colorectal cancer samples, found no impact of Fn infection on OS of patients with colorectal cancer (ref. 41; Supplementary Fig. S3). These results agree with those of a study by Kim and colleagues (48) and suggest that tissues with high Fn loads that have an impact on the prognoses of patients can be accurately identified by studies using FF or methacarn-fixed tissues or in some studies using FFPE samples if the Fn DNA is effectively amplified. In contrast, tissues with high loads of Fn may fail to be identified in many studies using FFPE samples such as ours reported here.

As shown by our previous studies, MSI-H was associated with Fn-positive samples in two independent colorectal cancer cohorts, one cohort consisting of FF samples from Japan, and another consisting of FFPE samples from the United States (10). Here, MLH1 hypermethylation and BRAF mutations were also significantly associated with Fn-positive colorectal cancer samples even though template DNA for Fn amplification was isolated from FFPE samples (Tables 1 and 2). These results suggest colorectal cancer with MSI-H, MLH1 hypermethylation or BRAF mutations must be heavily and/or specifically infected with Fn, so that reduction in amplifiable Fn DNA by FFPE treatment may not change these associations. In accordance with the results obtained from our studies using FFPE samples, two studies using FF tissues showed that high loads of Fn are associated with MSI-H (refs. 20, 40; Supplementary Table S4). Although the study by Proença did not see an association between any levels of Fn and MSI-H, this could be due to the small size of the cohort (43 cases; ref. 22; Supplementary Table S4). In the study by Wei, high loads of Fn were associated with loss of MLH1 expression, likely caused by MLH1 hypermethylation (37). In the study by Yamaoka, although the association between high loads of Fn and MLH1 hypermethylation was not significant, this could be due to the small number of cases with MLH1 hypermethylation in this cohort (nine cases among the 100 cases examined; ref. 21). In the study by Shariati, BRAF mutations were not associated with Fn-positivity; however, the number of BRAF-mutated cases were only two cases out of a total of 30 cases examined, leading to statistical insignificance (46). In another study by Kunzmann, lack of association of BRAF or KRAS mutations with high loads of Fn was reported; however, this may be because more than a one-third of the total cases (68 cases among 190 cases) were not examined for BRAF and KRAS mutations (20). In a study by Flanagan, the Fn load between colorectal cancer with KRAS mutations and those with non-KRAS mutations, as well as colorectal cancer with BRAF mutations and those with non-KRAS mutations were compared. Although the results showed no difference in quantity of Fn between BRAF-mutant and non–BRAF-mutant colorectal cancer or between KRAS-mutant and non–KRAS-mutant colorectal cancer, the question of whether high loads of Fn were associated with BRAF or KRAS mutations was not determined (35). Thus, although our results are the first to show a significant association between Fn infection and MLH1 hypermethylation or BRAF mutations among studies in Group B, our results need to be confirmed by future studies.

Regarding KRAS mutations, our results suggest that Fn may specifically infect colorectal cancer with KRAS mutations at low levels compared with colorectal cancer with non-KRAS mutations. Therefore, reduction in amplifiable Fn DNA by FFPE treatment may not change these associations (Fig. 1A). In Group B, three of four studies showed a significant association between Fn infection and KRAS mutations (20–22, 46). This agreed with our results (Supplementary Table S4). As mentioned above, the study by Kunzmann showed no association between Fn infection and KRAS mutations. This may be because of its failure to determine the mutation status of a large portion (68 cases) of the cohort (190 cases; ref. 20).

By comparing our results with those of the 11 previous studies listed in Group B, we could conclude that patients with high loads of Fn may show a shorter OS rate. The studies using FFPE tissues for Fn detection by PCR may fail to show this observation because of loss of amplifiable Fn DNA. High levels of Fn may infect colorectal cancer with MSI-H and MLH1 hypermethylation and this association can be detected in FF as well as FFPE colorectal cancer samples. Low levels of Fn may specifically infect colorectal cancer with KRAS mutations as compared with colorectal cancer with non-KRAS mutations and this association can be detected in either FF or FFPE colorectal cancer samples. In support of the association of Fn with MSI-H or KRAS mutations, although indirect, Ternes and colleagues showed that Fn infection is significantly enriched in CMS1 where 74% are MSI-H and in CMS3 where 70% carry KRAS mutations (11).

Our small non-cancer study suggests that there is no association between BRAF/KRAS mutations or MLH1 promoter hypermethylation and Fn infection in adenoma/polyp in contrast to carcinoma. There was also no association between Fn infection and adenoma tissue type (SSA vs. TA/TVA). MLH1 promoter hypermethylation was detected in one BRAF-mutated adenoma in our cohort (Table 4). These observations suggest that Fn infection may occur regardless of KRAS/BRAF mutation status or type of adenomas. Together with the colorectal cancer data presented here, the association of genetic alterations including MSI-H, BRAF mutation, and KRAS mutations or MLH1 hypermethylation with Fn infection may be established during and/or after adenomacarcinoma transition.

Our results raise the question of why and how Fn infection is enriched in colorectal cancer with SSP phenotypes including MSI-H, MLH1 hypermethylation, and BRAF mutations, and, independently, KRAS mutations. Compared with carcinoma, Fn infection in adenoma/polyps seems random, not selective to BRAF- or KRAS-mutated cases and is characterized by poor bacterial-enabling growth as indicated by a lower number of Fn copies. One possibility is that prolonged Fn infection, even at a low level in adenomas, may induce genome-wide hypermethylation of the CpG island promoter of tumor suppressor genes (TSG), leading to silencing of the expression of the TSGs. It has been shown that BRAF or KRAS mutations in initiated cells co-operate with silenced TSGs by promoter hypermethylation to progress toward carcinoma formation (49, 50). Thus, SSA with BRAF mutations and TA/TVA with KRAS mutations may have a selective advantage for progressing to carcinoma when their TSGs are hypermethylated and inactivated through Fn infection. Furthermore, if hypermethylation occurs at the MLH1 locus, the adenoma will result in MSI-H colorectal cancer (9). In support of this possibility, Fn infection has been associated with CIMP not only in colorectal cancer, but also in colon tissues from patients with ulcerative colitis (51). Another study showed that bacteria such as Fn and Hungatello hathewayi are frequently found in colorectal cancer, are associated with TSG promoter hypermethylation, and upregulate DNMT activity in infected cells (52). If this is the case, Fn infection must be associated with two sets of hypermethylated TSGs, one such as MLH1 is accompanied with BRAFV600E and another is accompanied with KRAS mutations (53, 54).

Another possibility is that an association of Fn infection with the SSP phenotype may be established after tumors progress to carcinoma. In this scenario, Fn may have greater affinity for colorectal cancers with SSP than with other colorectal cancers, so that Fn may efficiently colonize them. For instance, overexpression of galactose and/or GalNAc on the surface of tumor tissues has been a target of Fn infection and colonization through Fap2 (55). There is the possibility that the content of Gal-GalNAc or GalNac on the surface of colorectal cancers with the SSP phenotype is relatively higher than that of other subgroups of colorectal cancers. In support of this scenario, the expression of the Tn antigen (GalNca-Ser/Thr), a precursor of Gal-GalNc, is elevated in human cancer cells with a BRAF mutation or in colon tissues from BRAF^V600E−inducible mouse models (56). Also, loss of the GALNT6 protein associated with the SSP phenotype results in the expression of truncated O-glycan on the cell surface, leading to an increase in Tn antigen (57). This increased Gal-GalNac contents on cell surface of SSP tumors may not only attract Fn but also provide space for further proliferation of Fn. On the other hand, affinity of Fn to KRAS-mutated adenoma or carcinomas could be explained by KRAS mutated tissues’ requirement of a higher number of amino acids to survive and multiply in a nutrient-deficient environment (58). It has been shown that Fn produces various metabolites including formate, succinic acid, 2-hydroxybutyrate, and amino acids such as glutamic acid, aspartic acid, glycine, isoleucine, leucine, phenylalanine, and valine in vitro when contact with colon cancer tissues, or in vivo (11). Thus, presence of Fn may be advantageous for survival and growth of KRAS-mutated adenoma and/or carcinoma.

Although a small preliminary study, this is the first report showing that there is no association between Fn infection and BRAF or KRAS mutations in colon adenoma/polyp tissues in contrast to colorectal cancer, the small number of samples of adenoma/polyp tissues (32 cases) limited the statistical power of this work. Further study analyzing a cohort with a larger sample size is necessary. Another limitation of this study is that FFPE tissues were used to determine the effect of Fn infection on colorectal cancer patients’ OS rate. By comparing ours with other studies, we conclude that the ability to detect and quantify Fn in FFPE samples by PCR is limited. Therefore, our results showing a lack of association between Fn infection and patients’ OS rate needs to be reevaluated through future study.

In this study, we were not able to determine whether Fn infection is associated with a shorter RFS rate; some studies showed that Fn infection in tumor tissues is associated with shorter RFS of colorectal cancer patients’ postchemotherapy (27, 37, 39). This phenomenon could be explained by (i) acquired resistance of colon cancer cells to the toxic effects of 5-fluorouracil and/or oxaliplatin (27, 59); (ii) reduced antitumor immune response (60, 61); and/or (iii) increased stemness (11) by Fn infection. Our results show that Fn infection is enriched in clinically distinctive subgroups of colorectal cancers, MSI-H, and KRAS-mutated colorectal cancers. Therefore, it is important to determine whether Fn infection is a critical factor that modifies response to chemotherapy and the prognosis of colorectal cancers with MSI-H and KRAS mutations.

J.M. Carethers reports grants from NIH and from University of Michigan during the conduct of the study. No disclosures were reported by the other authors.

K. Takeda: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft. M. Koi: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. Y. Okita: Resources. S. Sajibu: Resources. T.O. Keku: Resources, data curation. J.M. Carethers: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, writing-original draft, project administration, writing-review and editing.

This work was supported by the United States Public Health Service (R01 CA258519 and P30 DK120515) and funds from the Department of Internal Medicine at the University of Michigan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article.