Fusobacterium nucleatum in the Colorectum and Its Association with Cancer Risk and Survival: A Systematic Review and Meta-analysis

The human gut has been estimated to contain over 3 trillion microbial cells which is approximately as many as the number of eukaryotic cells that comprise the human body (1), and may include as many as a 1,000 distinct bacterial species (2).

In recent years, there has been an increased understanding of the functional role of gut microbiota in the initiation of colorectal cancer (2). Cancer incidence in the large intestine is estimated to be 12-fold higher than that in the small intestine, which has been partially attributed to a larger bacterial density in the large intestine (3). Evidence suggests that an imbalance of normal intestinal microbiota can exacerbate chronic inflammatory conditions that lead to the production of carcinogenic metabolites, which promote the onset of colorectal cancer (3). A number of bacterial species have been implicated with colorectal cancer onset, including Helicobacter pylori, Escherichia coli, Bacteroides fragilis, Salmonella enterica, and Fusobacterium nucleatum (3).

F. nucleatum is a species of anaerobic gram-negative, non–spore forming, nonmotile bacteria of the genus Fusobacterium (4). F. nucleatum is an opportunistic commensal pathogen normally resident in the oral cavity (4), but may use saliva to survive passage through the stomach and into the lower gastrointestinal tract (5). In support of this, a recent study demonstrated that patients with colorectal cancer have identical F. nucleatum strains in their oral cavity and tumors (6). There is increasing evidence that F. nucleatum can adhere to and invade colorectal tumor tissue (7, 8) and may create a tumor permissive, proinflammatory microenvironment (TME; refs. 8–11).

Despite increasing evidence suggesting a potential role for F. nucleatum in colorectal cancer development and prognosis (12), to our knowledge, no systematic reviews with meta-analyses have fully evaluated the existing evidence. An existing systematic review (13) added to our understanding but did not conduct a meta-analyses to allow us to quantify the true association after accounting for differences in study results. This systematic review of the published scientific literature aimed to assess: (i) the prevalence/abundance of F. nucleatum colonization in biological samples from individuals with colorectal cancer or colorectal polyps compared with healthy controls or adjacent, normal tissue; (ii) the association of F. nucleatum with subsequent colorectal cancer development in individuals who were initially cancer-free; and (iii) the association between F. nucleatum and survival in individuals with colorectal cancer. To investigate these in sufficient depth, we were unable to review other bacterial species potentially linked with colorectal cancer.

An a priori systematic search of the literature was performed in accordance with PRISMA guidelines (Supplementary Table S1; ref. 14) using Medline (1947–), EMBASE (1974–), and Web of Science (1980–), from date of database inception to January 21, 2019 (after an update from October 2017), to identify all primary studies relating to the prevalence of F. nucleatum colonization and colorectal cancer risk and survival. A specific search strategy was devised to include at least one keyword or Medical Subject Heading (MeSH) term from each of the following: (i) colorectal cancer(s) or colon cancer(s) or rectal cancer(s) or cancer of the colorectum or cancer of the colon or cancer of the rectum or colorectal neoplasm(s) or colon neoplasm(s) or rectal neoplasm(s) or colorectal tumor(s) or colon tumor(s) or rectal tumor(s) or colorectal carcinoma(s) or colon carcinoma(s) or rectal carcinoma(s); and (ii) fusobacterium nucleatum or fusobacteria or fusobacterium or microbiome.

The search strategy was limited to studies carried out on humans. Both published full texts and conference abstracts were included. Review articles and individual case studies were removed as they would not provide new information to assess prevalence or abundance; however, to ensure relevance internationally, we did not restrict to English language papers.

Study selection was carried out independently, in duplicate, by two reviewers (from C. Gethings-Behncke H.G. Coleman, H.W. Jordao, L.J. Murray, and A.T. Kunzmann, with discrepancies settled by a third reviewer, using two levels of study screening.

For the prevalence/abundance studies (aim 1), participants had to be ages over 16 (to focus on adults but reduce risk of excluding population-based studies) and the study had to report the prevalence (proportion of samples with F. nucleatum DNA above detection threshold) or abundance (amount of F. nucleatum DNA in sample) of F. nucleatum status in any biological sample (including fecal, tissue, or dental): from individuals with colorectal cancer compared with individuals with colorectal polyp or healthy controls; from colorectal tumor tissue compared with adjacent normal tissue or; from individuals with colorectal polyps compared with healthy controls.

For the colorectal cancer development studies (aim 2), participants had to be ages over 16 and free from colorectal cancer at baseline and had to report the association between a measure of F. nucleatum status in any biological sample and subsequent colorectal cancer development.

For survival studies (aim 3), participants had to be had to be ages over 16 and diagnosed with colorectal cancer and the study had to report the association between a measure of F. nucleatum status in any biological sample and a measure of survival of the individuals with colorectal cancer.

Data extraction was conducted independently, in duplicate, by two reviewers (from C. Gethings-Behncke and A.T. Kunzmann), with discrepancies settled by a third reviewer (H.G. Coleman and L.J. Murray), and focused on authors; year of publication; study location; study design; F. nucleatum detection method; sample type (tissue, fecal, or dental); participant status (colorectal cancer, colorectal polyps, or healthy control); number of samples; prevalence of F. nucleatum in each sample/participant type; comparison reported; measure of colorectal cancer risk (aim 2) or prognosis (overall survival, colorectal cancer–specific survival or disease-free survival; aim 3); follow up years (aims 2 and 3); and adjustment for potential confounders. Study quality was assessed using the Newcastle Ottawa scale (15).

For prevalence of F. nucleatum (aim 1), random-effects meta-analyses were used to examine the pooled odds ratios (OR) and corresponding 95% confidence intervals (CI) of F. nucleatum prevalence in tissue and fecal samples, respectively, using published ORs, proportions, or numbers from:

• (i) Individuals with colorectal cancer compared with healthy controls.

• (ii) Individuals with colorectal cancer compared with individuals with colorectal polyps.

• (iii) Individuals with colorectal polyps compared with healthy controls.

• (iv) Colorectal tumor tissue compared with adjacent, normal tissue.

Too few studies had assessed the association between F. nucleatum prevalence or abundance and colorectal cancer development for a meta-analysis to be conducted (aim 2). Random-effects meta-analyses, with continuity correction for zero values, were used to examine the pooled hazard ratios (HR) and corresponding 95% CI for the association between the presence of F. nucleatum DNA and overall survival and disease-free survival (separately) in patients with colorectal cancer (aim 3). For all sets of meta-analyses, a X² test for heterogeneity was calculated and the I² statistic determined to estimate the proportion of variation between study results attributable to heterogeneity rather than chance (16). Heterogeneity was considered high if I² statistic was above 50% (16). Sensitivity analyses to examine the causes of heterogeneity were carried out by removing each study individually in turn. Because of low study numbers identified for each comparison, publication bias could not be formally assessed (17). Studies only reporting results for subgroups within cohorts were not included in meta-analyses due to the potential for reporting bias. Stata 14 software (Stata Corporation) was used for data analysis.

As shown in Fig. 1, 2,063 unique titles and abstracts were screened for potential eligibility, from which 184 full-texts were screened for eligibility. Of these, 45 relevant articles were identified (13, 18–62), some of which addressed multiple aims of the systematic review.

Eleven studies assessed the prevalence and/or abundance of F. nucleatum in colorectal tissue samples from separate individuals (Supplementary Table S2) with colorectal cancer compared with healthy colorectal tissue (six prevalence and four abundance studies) or colorectal adenoma (five prevalence and three abundance studies), and/or colorectal adenoma compared with healthy colorectal tissue (three prevalence and abundance studies).

As shown in Fig. 2, a meta-analysis assessing F. nucleatum prevalence indicated that the odds of F. nucleatum DNA being detected were higher in colorectal tumor tissue compared with healthy tissue from controls (n = 6 studies, pooled OR = 10.06; 95% CI, 4.48–22.58; I² = 0.0%). Smaller increased odds of F. nucleatum DNA being detected were apparent in colorectal tumor tissue compared with polyp tissue (n = 5 studies, pooled OR = 1.83; 95% CI, 1.07–3.16; I² = 66.9%) and in colorectal polyp tissue compared with healthy tissue from controls (n = 3 studies, pooled OR = 2.51; 95% CI, 1.20–5.27; I² = 0.0%). When limited to studies rated as ≥4 on the Newcastle–Ottawa scale (Supplementary Table S3), similar results were apparent for colorectal cancer compared with healthy tissue (n = 4 studies; HR = 9.73; 95% CI, 4.02–23.54) or colorectal polyp tissue (n = 3 studies, no exclusions), and/or colorectal polyp tissue compared with healthy tissue (n = 2 studies, HR = 2.56; 95% CI, 1.20–5.48).

The high heterogeneity (I² = 66.9%) when comparing colorectal tumor tissue samples to polyp tissue from controls was reduced after exclusion of a study by Ito and colleagues (2015), although the association was no longer statistically significant (n = 4 studies, pooled OR = 1.41; 95% CI, 0.96–2.07; I² = 0.0%). The association between F. nucleatum in colorectal tumor tissue compared with colorectal adenoma tissue remained in a post hoc sensitivity analysis excluding a study by Fu and colleagues (55) and by Yu and colleagues (35), which used FISH, rather than qRT-PCR to detect F. nucleatum (n = 4 studies, pooled OR = 2.86; 95% CI, 2.22–3.69; I² = 0.0%). Neither study using FISH found a difference in F. nucleatum in colorectal tumor tissue compared with colorectal adenoma tissue.

The variation in method of assessment for abundance prevented pooling of study results, although each study indicated a higher abundance in tissue at a more advanced diseased stage. All four studies assessing the abundance of F. nucleatum DNA in colorectal tumor tissue compared with healthy colorectal tissue from controls, found a higher abundance in colorectal tumor tissue (4, 23, 40, 42). Both studies comparing the abundance of F. nucleatum in colorectal tumor tissue compared with colorectal polyp tissue, and found a higher abundance in colorectal tumor tissue (42, 54). Similarly, both studies comparing the abundance of F. nucleatum in colorectal polyp tissue compared with healthy colorectal tissue from controls, and found a higher abundance in colorectal polyp tissue (22, 42).

Thirteen studies assessed the prevalence and/or abundance of F. nucleatum in colorectal cancer tissue compared with healthy adjacent colorectal tissue (7 prevalence and 13 abundance) and/or colorectal adenoma tissue compared with healthy adjacent tissues (two prevalence and two abundance; Supplementary Table S4).

As shown in Fig. 3, meta-analyses assessing F. nucleatum prevalence indicated that the odds of F. nucleatum being detected were higher in colorectal cancer tissue compared with adjacent, normal tissue (n = 7 studies, pooled OR = 2.42; 95% CI, 1.62–3.61; I² = 50.0%), with all studies were rated as ≥4 on the Newcastle–Ottawa scale. Neither of two studies (4, 54) found differences in F. nucleatum positivity in colorectal polyp tissue compared with adjacent, normal colorectal tissue.

A higher F. nucleatum abundance was observed in colorectal tumor tissue compared with adjacent normal tissue in 11 of the 12 studies reporting this comparison; however, the use of different detection methods prevented amalgamation using meta-analyses. Only two studies, assessed F. nucleatum abundance in colorectal polyp tissue compared with adjacent, normal colorectal tissue but neither Flanagan and colleagues (54) or Proenca and colleagues (4), identified a statistically significant difference.

Seventeen studies assessed the prevalence and/or abundance of F. nucleatum in fecal samples (Supplementary Table S5) from separate individuals with colorectal cancer compared with healthy colorectum (7 prevalence and 11 abundance studies); colorectal cancer compared with colorectal adenoma (three prevalence and four abundance studies); and/or colorectal adenoma compared with healthy colorectal tissue (three prevalence and seven abundance studies).

As shown in Fig. 4, meta-analyses assessing F. nucleatum prevalence indicated that the odds of F. nucleatum DNA being detected were higher in fecal samples from patients with colorectal cancer compared with healthy controls (n = 7 studies, pooled OR = 9.01; 95% CI, 3.39–23.95; I² = 72.6%). Smaller increases in the odds of F. nucleatum DNA being detected were apparent in individuals with patients with colorectal cancer compared with individuals with colorectal polyps (n = 3 studies, pooled OR = 3.31; 1.29–8.45; I² = 78.4%) and in individuals with colorectal polyps compared with healthy controls (n = 3 studies, pooled OR = 1.41; 95% CI, 0.53–3.74; I² = 81.7%). When limited to studies rated as ≥4 on the Newcastle–Ottawa scale (Supplementary Table S6), similar results were apparent. The high heterogeneity was reduced for all three comparisons after exclusion of individual studies by Amitay and colleagues (n = 6 studies, pooled OR = 11.59; 95% CI, 6.39–21.01; I² = 7.1%; ref. 29); Eklof and colleagues (n = 2 studies, pooled OR = 2.09; 95% CI, 1.25–3.50; I² = 0.0%; ref. 48); Suehiro and colleagues (n = 2 studies, pooled OR = 0.82; 95% CI, 0.56–1.20; I² = 0.0%; ref. 25), respectively.

A higher abundance of F. nucleatum DNA in fecal samples was found in patients with colorectal cancer compared with healthy controls in 12 of 13 studies; patients with colorectal cancer compared with polyp patients in three out of four studies, and; patients with colorectal polyp compared with healthy controls in four out of seven independent studies.

One study by Rajendran and colleagues (24) found a similar abundance of F. nucleatum DNA in dental plaque samples from patients with colorectal cancer compared with healthy controls (P = 0.86).

Three studies examined the association between F. nucleatum in biological samples and subsequent colorectal cancer development (Supplementary Table S7), although the samples used differed which meant a meta-analysis was not warranted. Mai and colleagues (21) did not find a statistically significant association between F. nucleatum in subgingival plaque samples and subsequent colorectal cancer risk (HR = 0.78; 95% CI, 0.18–3.43) in an analysis of a cohort of 1,252 individuals with 11.8 years of follow-up (mean). Yang and colleagues (59) found F. nucleatum in prediagnostic saliva samples from 99.6% of 231 individuals who subsequently developed colorectal cancer (cases) and in 99.6% of 461 controls (matched on age, ethnicity, smoking, season of enrolment, and recruitment method) who did not develop colorectal cancer, with no difference in prevalence (HR = 1.12; 95% CI, 1.70–27.90). Kwong and colleagues (49) found that 79 individuals with blood culture tests positive for F. nucleatum were at a higher subsequent risk of colorectal cancer development than 395 propensity matched controls (based on age, sex, and comorbidities) with blood culture tests negative for F. nucleatum (HR = 6.89; 95% CI, 1.70–27.90). All studies were rated as ≥4 on the Newcastle–Ottawa scale (Supplementary Table S8), although sample sizes were all small.

Seven studies assessed the association between F. nucleatum DNA in tumor tissue and survival in patients with colorectal cancer (Supplementary Table S9).

As shown in Fig. 5, a meta-analysis found that F. nucleatum positivity in tumor tissue was associated with poorer overall survival among patients with colorectal cancer (n = 5 studies, HR = 1.87; 95% CI, 1.12–3.11; I² = 60.6%). All five studies included in the meta-analysis were rated as ≥4 on the Newcastle–Ottawa scale (Supplementary Table S8). The high heterogeneity (I² = 60.6%) was reduced after exclusion of the largest study by Mima and colleagues (12) and the association remained statistically significant (n = 4 studies, HR = 2.25; 95% CI, 1.50–3.37; I² = 0.0%).

Three additional studies, which were not eligible for inclusion in the meta-analysis because either they only reported subgroup results [based on Pseudomonas fragi positivity (57) or palliative status (51)] or did not report HRs (53), also found poorer survival in individuals with F. nucleatum identified in their colorectal tumor tissue.

Mima and colleagues (12) found an association between F. nucleatum positivity and colorectal cancer–related survival (HR = 1.58; 95% CI, 1.04–2.39); as did Yan and colleagues (2017; HR = 2.22; 95% CI 1.48–3.33).

As shown in Fig. 6, a meta-analysis found that F. nucleatum positivity in tumor tissue was not associated with disease-free survival among patients with colorectal cancer (n = 3 studies, HR = 1.48; 95% CI, 0.84–2.59; I² = 88.5%). The high heterogeneity (I² = 60.6%) was reduced after exclusion of the lowest quality study by Oh and colleagues (50) and the association became statistically significant (n = 2 studies, HR = 1.95; 95% CI, 1.44–2.66; I² = 0.0%).

In this first comprehensive systematic review with meta-analyses of the published literature on F. nucleatum in colorectal cancer, we found consistent evidence that F. nucleatum was more prevalent in fecal and tissue samples from patients with colorectal cancer compared with adjacent tissue, or samples from comparative healthy controls or individuals with colorectal polyps. The presence of F. nucleatum in colorectal tumor tissue was also associated with worse overall survival in patients with colorectal cancer.

The substantial evidence that F. nucleatum is more likely to be present and more abundant in colorectal cancer samples than healthy samples and increases along the normal-adenoma-carcinoma tissue pathway may indicate a potential causal role in colorectal cancer development. F. nucleatum has been demonstrated to adhere to and invade colorectal tumor cells, using specific surface proteins to bind to tumors: Fap2 binds tumor Gal-GalNAc; and the adhesin FadA binds E-cadherin which may stimulate expression of inflammatory genes, a known risk factor for colorectal cancer (7, 8). F. nucleatum has also been demonstrated to promote immune evasion in colorectal cancer: F. nucleatum specifically binds the inhibitory receptor TIGIT expressed on natural killer (NK) cells through its Fap2 protein, thereby inhibiting the cytotoxic activity of the NK cells (8–11). Therefore, there is emerging evidence that F. nucleatum colonization creates a tumor permissive, proinflammatory microenvironment (TME; refs. 10, 61). However, it is also possible that F. nucleatum may merely exploit and replicate effectively in the hypoxic tumor microenvironment (29), or that its presence may reflect procarcinogenic diets (37). The increased prevalence seems robust to methodologic changes; whereas most studies used qRT-PCR, one study found similar results when using 16S rRNA FISH (35).

The higher prevalence in fecal samples from patients with colorectal cancer indicates F. nucleatum DNA could potentially be useful as an adjunct biomarker in colorectal cancer screening. Two recent smaller systematic reviews found that measuring F. nucleatum DNA in fecal samples could add predictive value for the presence of colorectal cancer (62, 63). Similar results have also been observed for serum samples (64). However, the utility of biomarkers for colorectal screening may also rely on detecting premalignant polyps. Our meta-analysis of three studies found no difference in the odds of F. nucleatum positivity between fecal samples from individuals with colorectal polyps than from healthy controls, which casts doubt on the utility of using F. nucleatum in the early detection of colorectal polyps. Abundance however may give a more comprehensive picture, although studies tended to use different reporting methods which did not allow for meta-analyses of abundance to be conducted. Therefore, further studies should attempt to investigate whether measurement of fecal or serum F. nucleatum prevalence or abundance adds predictive value to current fecal screening methods.

Despite the substantial evidence regarding an increased prevalence of F. nucleatum, the evidence for an association between F. nucleatum and future risk of colorectal cancer is limited with only three studies eligible for inclusion in the review. Two studies (21, 59) found no association between the presence of F. nucleatum and subsequent colorectal cancer risk, which could be in part due to the use of oral samples that may not accurately reflect the fecal microbiome as F. nucleatum is commensal to the oral cavity (33). Similarly, statistical power was limited as few cases or controls (<0.4%) had no F. nucleatum in the study by Yang and colleagues (59) and the cohort used by Mai and colleagues (21) was small (n = 1,252) with only 17 individuals developing colorectal cancer. Interestingly, a study by Kwong and colleagues found a higher subsequent risk of colorectal cancer in individuals previously diagnosed with bacteremia from F. nucleatum (49). However, the no lag time was built into the analysis for the association between bacteremia episode and colorectal cancer, reverse causality is possible. The results were not materially changed when excluding lower quality studies rated as ≥4 on the Newcastle–Ottawa scale, suggesting the results are unlikely due to low study quality. This review highlights the need for further population-based cohort or nested case–control studies measuring the fecal microbiome and subsequent colorectal cancer risk to examine whether F. nucleatum precedes colorectal cancer development.

The observation of an association between F. nucleatum DNA prevalence and overall survival could highlight a role in cancer progression or a role as a passive prognostic biomarker. Although the strength of the association between F. nucleatum and overall survival was modest and heterogeneous, which may limit the clinical utility as a prognostic biomarker, Mima and colleagues (12) found a stronger association with cancer-specific survival. The weaker association between F. nucleatum and overall survival in the study by Mima and colleagues (12), compared with other studies, could reflect the large, high-quality nature of the study design but could also reflect use of formalin-fixed paraffin-embedded tissue rather than fresh frozen tissue used for other studies reporting results for overall survival. Recent evidence also hints at a similarity in the microbiome between primary colorectal lesions and metastatic lesions (65), which may improve the prognostic benefits of targeted treatments. Therefore, studies should investigate whether F. nucleatum in primary or metastatic lesions could be used a prognostic indicator for cancer-specific survival to aid treatment decisions, or using mechanistic designs to assess whether F. nucleatum could be a therapeutic target.

The heterogeneity was high (I² over 50%) when comparing colorectal tumor tissue samples to polyp tissue from controls, for the analyses in fecal samples and for the overall survival analyses, which were lowered after removing individual studies. Often these studies were the larger, higher quality studies. Therefore, replication of the results in large high-quality analyses within large sample sizes may further increase confidence in future pooled estimates.

A strength of this systematic review is that it offers a comprehensive assessment of the existing literature on the role of F. nucleatum in colorectal cancer identification, development, and prognosis, which better quantifies the differences in prevalence and better highlights gaps in the literature than existing reviews (66). A comprehensive assessment of other bacterial species linked with colorectal cancer was not feasible within the scope of this review but future systematic reviews on these are recommended. However, a limitation is that the low study numbers prevented a formal assessment of publication/reporting bias (17), and may limit robustness of some meta-analyses with only three studies. Whole microbiome studies using shotgun metabolomics methods for assessing microbiome could help determine whether the increased prevalence in colorectal cancer tissue is unique to Fusobacteria or whether other bacterial genera are increased. However, the nature of whole microbiome studies may make reporting null results at a species level difficult which increases the risk of publication/reporting bias. Therefore, estimates should be interpreted cautiously, and future whole microbiome studies should use careful reporting of null results to reduce the potential for reporting bias.

There is consistent evidence indicating an increased prevalence and/or abundance of F. nucleatum DNA in fecal and tissue samples from patients with colorectal cancer. Further prospective studies investigating the role of fecal F. nucleatum as a causal factor or a predictive biomarker for colorectal cancer development and prognosis are required.

No potential conflicts of interest were disclosed.

We express our appreciation to the late Professor Liam Murray, whose contribution to this work and to the field of cancer epidemiology was of great significance. We would like to thank Miss Peipei Liu and Dr. Qing Wen for translating non-English language papers to assist the review.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.