Pks‐positive Escherichia coli in tumor tissue and surrounding normal mucosal tissue of colorectal cancer patients

Abstract

A significant association exists between the gut microbiome and colorectal carcinogenesis, as well as cancer progression. It has been reported that Escherichia coli (E. coli) containing polyketide synthetase (pks) island contribute to colorectal carcinogenesis by producing colibactin, a polyketide‐peptide genotoxin. However, the functions of pks ⁺ E. coli in initiation, proliferation, and metastasis of colorectal cancer (CRC) remain unclear. We investigated the clinical significance of pks ⁺ E. coli to clarify its functions in CRC. This study included 413 patients with CRC. Pks ⁺ E. coli of tumor tissue and normal mucosal tissue were quantified using droplet digital PCR. Pks ⁺ E. coli was more abundant in Stages 0–I tumor tissue than in normal mucosal tissue or in Stages II–IV tumor tissue. High abundance of pks ⁺ E. coli in tumor tissue was significantly associated with shallower tumor depth (hazard ratio [HR] = 5.0, 95% confidence interval [CI] = 2.3–11.3, p < 0.001) and absence of lymph node metastasis (HR = 3.0, 95% CI = 1.8–5.1, p < 0.001) in multivariable logistic analyses. Pks ⁺ E. coli‐low and ‐negative groups were significantly associated with shorter CRC‐specific survival (HR = 6.4, 95% CI = 1.7–25.6, p = 0.005) and shorter relapse‐free survival (HR = 3.1, 95% CI = 1.3–7.3, p = 0.01) compared to the pks ⁺ E. coli‐high group. Pks ⁺ E. coli was abundant in Stages 0–I CRC and associated with CRC prognosis. These results suggest that pks ⁺ E. coli might contribute to carcinogenesis of CRC but might not be associated with tumor progression.

Polyketide synthetase (pks)‐positive Escherichia coli (pks ⁺ E. coli), which has been reported to contribute to colorectal carcinogenesis, is more abundant in Stage I tumors than normal mucosal tissue and Stages II–IV tumors. Furthermore, a high level of pks ⁺ E. coli in tumor tissue is associated with a favorable prognosis of prognosis in multivariable Cox regression analysis.

Abbreviations

CI

confidence intervals

CRC

colorectal cancer

CSS

cancer‐specific survival

E. coli

Escherichia coli

HR

hazard ratio

IQR

interquartile range

pks

polyketide synthetase

RFS

relapse‐free survival

UICC

the Union for International Cancer Control

1. INTRODUCTION

Colorectal cancer (CRC) is the third most common and the second most deadly cancer in the world. In 2020, almost 2 million people were diagnosed with CRC and almost 1 million deaths occurred. The incidence of CRC continues to rise worldwide. ¹ , ² A diet centered on meat, along with alcoholic beverages and obesity, increases the risk of CRC. These factors also affect the gut microbiome, altering its structure and function, which in turn affects metabolic and immune pathways. ³ However, the underlying mechanism has not been identified.

Intestinal bacteria comprise approximately 30 trillion microorganisms. ⁴ Disruptions in the gut microbiota and its metabolites are pivotal in development of numerous diseases, including carcinomas, gastrointestinal disorders, metabolic diseases, and neurodegenerative conditions. ⁵ Accumulating evidence suggests a significant association between the gut microbiome and colorectal carcinogenesis, as well as cancer progression. This association operates through mechanisms such as DNA damage, inflammation, and effects on CRC‐related molecular pathways. ⁶ , ⁷ One strain of Escherichia coli (E. coli) carries a 54‐kilobase polyketide synthetase (pks) island (pks ⁺ E. coli). This island produces enzymes for colibactin, a polyketide‐peptide genotoxin. Colibactin alkylates DNA and induces double‐stranded breaks at locations corresponding to somatic mutation hotspots in CRC genomes. ⁸ , ⁹ Yet, the role of pks ⁺ E. coli in the initiation, proliferation, and metastasis of CRC remains unclear. Some research indicates that its prevalence in CRC and colorectal adenoma groups is significantly elevated compared to the healthy cohort. In contrast, other studies have found no notable differences. ¹⁰ , ¹¹ , ¹² , ¹³ Additionally, pks ⁺ E. coli is more prevalent in the mucosal microbiota of CRC patients than in healthy controls. ¹⁴ , ¹⁵

While various results have been reported, and numerous questions persist regarding the relationship between pks ⁺ E. coli and CRC, our initial objective was to determine whether pks ⁺ E. coli contributes to the carcinogenesis of CRC. If pks ⁺ E. coli is significantly involved in carcinogenesis, it should be identifiable even in Stages 0–I CRC. To ensure that we captured even trace numbers of pks ⁺ E. coli, we opted for digital PCR over quantitative PCR due to its superior sensitivity. ¹⁶ The primary objective of this study was to determine whether the detection rate of pks ⁺ E. coli in Stages 0–I of CRC is greater than that in Stage II or more advanced cancers. Our secondary goal was to characterize clinicopathological characteristics of tumors in which pks ⁺ E. coli is found. Moreover, we explored differences in prognosis between pks ⁺ E. coli‐positive CRC and pks ⁺ E. coli ‐negative CRC.

2. METHODS

2.1. Patients and sample collection

This observational cohort study included consecutive CRC patients who underwent surgical treatment at Nippon Medical School Hospital between January 2017 and March 2021. To determine the positivity rate of pks ⁺ E. coli in early‐stage cancer, it would be ideal to include cases of endoscopic resection. However, these cases were not included in this study, as fresh sample acquisition can adversely affect clinical diagnosis. Registration continued until 75 Stages 0–I patients were enrolled. We excluded patients who received preoperative treatments such as chemotherapy or radiotherapy. Additionally, patients with hereditary CRC (e.g., Lynch syndrome or hereditary adenomatous polyposis) were excluded, as they can develop cancer despite the influence of intestinal bacteria. Cancer stages were classified according to the eighth edition of the Union for International Cancer Control. All procedures were carried out in compliance with the Declaration of Helsinki and received approval from the institutional review board (approval no. B‐2022‐506). Information regarding the right to opt out was provided on the websites of the main hospital of Nippon Medical School.

In patients who underwent surgery between January 2017 and December 2019, samples were taken only from their tumors. In contrast, those who had surgery after this period had samples collected from both the tumor and the normal mucosal tissue. For normal mucosal tissue, samples were obtained from the oral side, situated more than 30 mm and less than 50 mm from the tumor. All samples were preserved at −80°C until they were analyzed. DNA was extracted from both tumor and normal mucosal tissue using NucleoSpin Tissue Kits (Takara Bio, Kusatsu, Shiga, Japan). Quantities of DNA were determined using Qubit 2.0 Fluorometers (Thermo Fisher Scientific, Waltham, USA).

2.2. Droplet Digital PCR for pks ⁺ Escherichia coli absolute copy number analysis

We determined the absolute pks ⁺ E. coli copy number of DNA exacted from tumor tissue and normal mucosal tissue using the QX200 Droplet Digital PCR System (ddPCR, Bio‐Rad Laboratories, Hercules, California, USA) according to the manufacturer's protocols. The PCR reaction was performed in a 20‐μL volume containing 10 ng of genomic DNA, 10 μL 2X EvaGreen Supermix (Bio‐Rad Laboratories), and 250 nmol/L of each primer. Primer sequences for pks ⁺ E. coli were as follows: forward primer, 5′‐GCGCATCCTCAAGAGTAAATA‐3′; reverse primer 5′‐GCGCTCTATGCTCATCAACC‐3′. ¹⁷ Droplet generation was performed using a QX200 Droplet Generator (Bio‐Rad Laboratories). PCR conditions were 7 min at 95°C, 35 cycles of 30s at 95°C, 30s at 55°C, and 30s at 72°C, with a final extension for 7 min at 72°C. The 96‐well plate was transferred to the QX200 droplet reader (Bio‐Rad Laboratories), and data were analyzed with Quanta Software version 1.7.4 (Bio‐Rad Laboratories).

2.3. Detection of RAS and BRAF mutations

According to the manufacturer's protocols, RAS and BRAF mutations were detected using ddPCR. We evaluated 12 KRAS mutation sites (G12A, G12C, G12D, G12R, G12S, G12V, G13D, Q16H [183A>C], Q61H [183A>T], Q61K, Q61L, and Q61R) and 13 NRAS mutations sites (G12A, G12C, G12D, G12S, G12V, G13D, G13R, G13V, Q61H [183A>C], Q61H [183A>T], Q61K, Q61L, and Q61R) for RAS mutations using commercial kits, which were the ddPCR KRAS Screening Multiplex Kit, the ddPCR KRAS Q61 Screening Kit, the ddPCR NRAS G12/G13 Screening Kit, and the ddPCR NRAS Q61 Screening Kit. BRAF mutations were determined by BRAF V600E mutations as described previously. ¹⁸

2.4. Statistical analysis

All statistical analyses were conducted using R version 4.2.2 (The R Foundation for Statistical Computing, Vienna, Austria). All reported p‐values are two‐sided, with a significance level of p < 0.05. Quantitative variables were analyzed to compare pks ⁺ E. coli copy numbers using the Mann–Whitney U‐test, the Kruskal–Wallis test, the Wilcoxon matched‐pairs signed‐ranks test, and the Spearman's rank correlation coefficient. For categorical variables, the χ ²‐test or Fisher's exact test was employed. Logistic regression analysis was used to identify factors influencing stage, tumor depth, and lymph node metastasis. Differences in CSS and recurrence‐free survival (RFS) were assessed with Kaplan–Meier analysis, followed by log‐rank analyses. A multivariate analysis of prognostic factors related to CSS and RFS was conducted using the Cox proportional hazards model.

3. RESULTS

3.1. Patients

During the study, 509 patients underwent surgery. Of those, 69 patients were excluded because of preoperative treatment, 6 patients because of hereditary CRC, and 21 patients because of double cancer. In total, 413 patients were included. Normal mucosal tissue was obtained from 244 patients. Seventy‐five patients with Stage 0 or I CRC were registered, but later, one case was identified as having a hereditary tumor, and this case was excluded. Enrolled patients included 146 Stage II CRC, 133 Stage III, and 60 Stage IV cases (Table 1). RAS mutations were detected in 198 patients (47.9%) and BRAF mutations were detected in 27 (6.5%).

TABLE 1.

Characteristics according to the abundance of polyketide synthetase (pks)⁺ Escherichia coli in CRC tissue

Characteristic	All patients (n = 413)	pks + Escherichia coli abundance in CRC tissue			p‐Value
		High (n = 135)	Low (n = 136)	Negative (n = 142)
Sex
Female	174	55	60	59	0.84
Male	239	80	76	83
Age in years (median, [IQR])	72 [65–80]	72 [65–80]	73 [65–81]	72 [63–79]	0.67
Tumor location
Right	158	55	56	47	0.64
Left	144	46	44	54
Rectum	111	34	36	41
Stage (UICC)
0‐I	74	50	12	12	<0.001
II	146	47	47	52
III	133	22	55	56
IV	60	16	22	22
Tumor depth
Tis‐1	30	17	4	9	<0.001
T2	57	35	11	11
T3	240	64	89	87
T4	86	19	32	35
Tumor size, mm (median, [IQR])	45 [30–60]	40 [25–53]	45 [30–60]	50 [30–60]	0.03
Circumference, % (median, [IQR])	70.0 [40.0–100.0]	50.0 [33.0–90.0]	71.0 [42.8–100.0]	75.0 [47.0–100.0]	0.01
Tumor differentiation
Well to moderate	335	110	115	110	0.32
Others	78	25	21	32
Lymphatic invasion
Presence	262	75	85	102	0.02
Absence	151	60	51	40
Venous invasion
Presence	269	80	93	96	0.22
Absence	1	55	43	46
Perineural invasion
Presence	293	83	103	107	0.01
Absence	120	52	33	35
Lymph node metastasis
Presence	179	33	71	75	<0.001
Absence	234	102	65	67
Metastasis
Presence	60	16	22	22	0.55
Absence	353	119	114	120
RAS status
Mutation type	198	65	63	70	0.88
Wild type	215	70	73	72
BRAF status
Mutation type	27	6	14	7	0.08
Wild type	386	129	122	135

Abbreviations: CRC, colorectal cancer; IQR, interquartile range; UICC, the Union for International Cancer Control.

3.2. Pks ⁺ E. coli in tumor tissue and normal mucosal tissue

Pks ⁺ E. coli was detected in tumor tissue from 271 of 413 patients (65.6%) and 152 of 244 normal mucosal tissue (62.3%). The median copy number of pks ⁺ E. coli was 0.70 copies/ng in tumor tissue and 0.50 copies/ng in normal tissue. There was no significant difference in pks ⁺ E. coli copy numbers between tumor and normal mucosal tissue (p = 0.44; Figure 1). When comparing matched pairs of the 244 tumor and normal mucosal tissue, the pks ⁺ E. coli copy number was also not significantly different (p = 0.66; Figure 2A). However, at each stage, there were significant differences in pks ⁺ E. coli copy numbers between matched pairs of tumor tissue and their respective normal mucosal tissue. In Stages 0–I (46 patients), the difference was significant (p < 0.001; Figure 2B), as it was in Stage III (78 patients, p = 0.002; Figure 2D). However, no significant difference was observed in Stage II (83 patients, p = 0.85; Figure 2C) or Stage IV (37 patients, p = 0.72; Figure 2E).

FIGURE 1.

Comparison of polyketide synthetase (pks)⁺ Escherichia coli copy numbers between normal mucosa and tumor tissue. The box plots show the median with interquartile range (25th percentile and 75th percentile). There was no significant difference in pks ⁺ E. coli copy numbers between tumor and normal mucosal tissue (Mann–Whitney U‐test, p = 0.44).

FIGURE 2.

Comparison of polyketide synthetase (pks)⁺ Escherichia coli copy numbers in matched pairs of tumor and normal mucosal tissue. The box plots show the median with the interquartile range (25th percentile and 75 percentile): (A) including all patients (244 patients) (B); including patients of Stages 0–I (46 patients); (C) including patients of Stage II (83 patients); (D) including patients of Stage III (78 patients); and (E) including patients of Stage IV (37 patients). In Stages 0–I and III patients, the copy number of pks ⁺ E. coli in tumor tissue was significantly greater than that in normal mucosal tissue (Wilcoxon matched‐pairs signed‐ranks test, p < 0.001, p = 0.0018).

Pks ⁺ E. coli copy number did not differ based on tumor location (e.g., right‐side colon, left‐side colon, or rectum, p = 0.73, Kruskal–Wallis test) or location of normal mucosa (p = 0.18). There was a significant difference among the five groups (mucosal tissue and Tis‐4 tumor tissue, p < 0.001, Kruskal–Wallis test). Pks ⁺ E. coli copy number was greater in Tis‐T1 tumor tissue than in normal mucosal tissue and T3 and T4 tumor tissue (Figure 3A). T2 tumor tissue also showed a greater pks ⁺ E. coli copy number compared to normal mucosal tissue and T3 and T4 tumor tissue(Figure 3B). Furthermore, the pks ⁺ E. coli copy number in tumor tissue from patients without lymphatic metastasis was significantly lower than that in tumor tissue of patients with lymphatic metastasis (p < 0.001, Kruskal–Wallis test). A significant difference was also observed among the five groups (normal mucosal tissue and Stage I–IV tumor tissue, p < 0.001, Kruskal–Wallis test). Remarkably, Stage I tumor tissue had the highest pks ⁺ E. coli copy number (Figure 4).

FIGURE 3.

The copy numbers of polyketide synthetase (pks)⁺ Escherichia coli between tumors at various tumor depths and normal mucosal tissues. The box plots depict the median along with the interquartile range (25th and 75th percentiles). The copy number of pks ⁺ Escherichia coli was greater in Tis‐T1 tumor tissue than in normal mucosal tissue (p = 0.02), T3 (p = 0.013), and T4 (p = 0.0055) tumor tissue (A, Mann–Whitney U‐test). The copy number of pks ⁺ E. coli was greater in T2 tumor tissue than in normal mucosal tissue (p < 0.001) and T3 (p < 0.001) and T4 (p < 0.001) tumor tissue (B, Mann–Whitney U‐test).

FIGURE 4.

Comparison of polyketide synthetase (pks)⁺ Escherichia coli copy numbers between normal mucosal and tumor tissue of various stage. The box plots show the median with the interquartile range (25th percentile and 75 percentile). The copy number of pks ⁺ E. coli in Stage I tumor tissue was the greatest and significantly greater than that in other four groups (p < 0.001, Mann–Whitney U‐test).

We investigated the correlations between pks ⁺ E. coli and Fusobacterium nucleatum, as well as between pks ⁺ E. coli and Parvimonas micra in patients with available residual specimens (a total of 180 patients). There were no significant correlations between pks ⁺ E. coli and either Fusobacterium nucleatum or Parvimonas micra (Figure S1).

3.3. The relationship between pks ⁺ E. coli in colorectal cancer tissues and clinicopathologic features

We categorized pks ⁺ E. coli‐positive patients into two groups based on the median (1.53 copies/ng for pks ⁺ E. coli copy number): the pks ⁺ E. coli‐high group and the pks ⁺ E. coli‐low group. Patients in whom pks ⁺ E. coli was not detected in tumor tissue were classified as pks ⁺ E. coli‐negative. A comparison of clinicopathological features among these three groups is presented in Table 1. Patients with high abundance of pks ⁺ E. coli in tumor tissue exhibited shallower tumor depth (p < 0.001), smaller tumor sizes (p = 0.03), and reduced tumor circumferences (p = 0.01) and more frequently had no lymphatic invasion (p = 0.02), no perineural invasion (p = 0.01), and no lymph node metastasis (p < 0.001).

To assess associations between pks ⁺ E. coli expression in tumor tissue and factors like tumor depth and lymph node metastasis, we conducted a logistic regression analysis (Table 2). High level of pks ⁺ E. coli in cancer tissue was significantly associated with shallower tumor depth in both univariable (p < 0.001; Table 2) and multivariable logistic regression analyses (hazard ratio [HR] = 5.0, 95% confidence interval [CI] = 2.3–11.3, p < 0.001; Table 2). Similarly, the absence of lymph node metastasis was linked to high abundance of pks ⁺ E. coli in univariable (p < 0.001; Table 3) and multivariable logistic analyses (HR = 3.0, 95% CI = 1.8–5.1, p < 0.001; Table 3).

TABLE 2.

Univariate and multivariate analyses for predicting T factor in colorectal cancer patients (pTis −2 vs. pT3–4)

Variables	Univariate			Multivariate
	HR	95% CI	p‐value	HR	95% CI	p‐value
pks + Escherichia coli (high vs. low and negative)	4.4	2.7–7.2	<0.001	5.0	2.3–11.3	<0.001
Lymph node metastasis (absence vs. presence)	4.9	2.7–9.1	<0.001	1.3	0.5–3.2	0.55
Tumor size (<45 mm vs. ≥45 mm)	9.3	5.0–19	<0.001	2.2	0.8–6.5	0.13
Circumference (<70% vs. ≥70%)	18.1	8.7–44.4	<0.001	8.5	2.6–31	<0.001
Pathology (well to moderate vs. others)	2.3	1.2–5.2	0.03	3.1	1.2–9.2	0.03
Venous invasion (absence vs. presence)	4.0	2.5–6.7	<0.001	2.5	1.1–5.4	0.02
Lymphatic invasion (absence vs. presence)	3.0	1.9–4.9	<0.001	0.9	0.4–2.0	0.84
Perineural invasion (absence vs. presence)	25.1	14–48.1	<0.001	18.0	8.4–41.4	<0.001
RAS status (WT vs. MT)	1.2	0.8–2.0	0.37	–	–	–
BRAF status (WT vs. MT)	2.2	0.8–9.5	0.20	–	–	–

Abbreviations: CI, confidence intervals; HR, hazard ratio; pks, polyketide synthetase.

TABLE 3.

Univariate and multivariate analyses for predicting N factor in colorectal cancer patients (absence vs. presence)

Variables	Univariate			Multivariate
	HR	95% CI	p‐value	HR	95% CI	p‐value
pks + Escherichia coli (high vs. low and negative)	3.4	2.2–5.5	<0.001	3.0	1.8–5.1	<0.001
Tumor depth (Tis‐2 vs. T3‐4)	4.9	2.7–9.1	<0.001	1.6	0.7–3.5	0.26
Tumor size (<45 mm vs. ≥45 mm)	1.8	1.2–2.6	0.005	1.4	0.8–2.5	0.28
Circumference (<70% vs. ≥70%)	1.9	1.3–2.8	0.001	1.0	0.6–1.8	0.99
Pathology (well to moderate vs. others)	1.8	1.1–3.0	0.02	1.5	0.8–2.6	0.20
Venous invasion (absence vs. presence)	2.0	1.3–3.1	0.001	1.1	0.7–1.9	0.69
Lymphatic invasion (absence vs. presence)	5.3	3.4–8.6	<0.001	4.3	2.6–7.3	<0.001
Perineural invasion (absence vs. presence)	4.0	2.5–6.6	<0.001	2.1	1.1–4.0	0.02
RAS status (WT vs. MT)	1.1	0.7–1.6	0.66	–	–	–
BRAF status (WT vs. MT)	1.7	0.8–3.8	0.19	–	–	–

Abbreviations: CI, confidence intervals; HR, hazard ratio; pks, polyketide synthetase.

Next, we categorized patients into three groups based on pks ⁺ E. coli in mucosal tissue as described above: the pks ⁺ E. coli ‐high group, the pks ⁺ E. coli ‐low group, and the pks ⁺ E. coli‐negative group. High abundance of pks ⁺ E. coli in mucosal tissue was significantly associated with sex (p = 0.02). There was no relationship between pks ⁺ E. coli in mucosal tissue and CRC progression (Table S1).

3.4. Effect of pks ⁺ E. coli in tumor tissue on prognosis

Patients with lower abundance or absence of pks ⁺ E. coli in tumor tissue had shorter CSS time when analyzed both among three groups (log‐rank test, p = 0.02; Figure 5A) and two groups (log‐rank test, p = 0.01; Figure 5B). Both the univariable analysis (HR = 4.3, 95% CI = 1.3–14.2, p = 0.02) and multivariable Cox regression analysis (HR = 6.4, 95% CI = 1.7–25.6, p = 0.005) indicated that the pks ⁺ E. coli‐low and ‐negative groups were significantly associated with a shorter CSS compared to the pks ⁺ E. coli‐high group (Table 4). In contrast, there was no significant difference for CSS dividing into various CRC stages (Figure S2).

FIGURE 5.

The influence of polyketide synthetase (pks)⁺ Escherichia coli copy number on cancer‐specific survival of Stages I–IV patients: (A) divided into three groups (high, low, and negative) and (B) divided into two groups (high and low or negative). Each statistical analysis included a log‐rank test.

TABLE 4.

Univariate and multivariate analyses for predicting cancer‐specific survival in CRC patients: Cox regression model result.

Variables	Univariate			Multivariate
	HR	95% CI	p‐value	HR	95% CI	p‐value
pks + Escherichia coli (low, negative vs. high)	4.3	1.3–14.2	0.02	6.4	1.7–25.6	0.005
Tumor location (right vs. left, rectum)	3.1	1.4–6.7	0.004	2.2	1.0–5.2	0.06
Tumor depth (T4 vs. Tis‐3)	7.5	3.5–16	<0.001	3.7	1.5–9.0	0.003
Lymph node metastasis (presence vs. absence)	11.1	3.3–37	<0.001	5.0	1.4–18	0.01
Pathology (others vs. well to moderate)	4.3	2.0–9.0	<0.001	4.5	2.0–10	<0.001
Venous invasion (presence vs. absence)	13.9	1.9–102.5	0.01	20.9	2.3–190.2	0.01
Lymphatic invasion (presence vs. absence)	4.5	1.4–15	0.01	1.7	0.5–6.0	0.42
Perineural invasion (presence vs. absence)	3.4	1.0–11	0.04	0.5	0.1–2.0	0.33
RAS status (MT vs. WT)	1.1	0.5–2.2	0.89	–	–	–
BRAF status (MT vs. WT)	2.9	1.1–7.7	0.03	2.5	0.8–7.5	0.12

Abbreviations: CI, confidence intervals; CRC, colorectal cancer; HR, hazard ratio; pks, polyketide synthetase.

In Stages I–III patients, the pks ⁺ E. coli high group showed better RFS compared to the low or negative group (log‐rank test, p = 0.001; Figure 6A). The pks ⁺ E. coli high group showed better RFS compared to the low and negative group (log‐rank test, p < 0.001; Figure 6B). In Stages II–III patients, the pks ⁺ E. coli high group showed better RFS compared to the low or negative group (Log rank test, p = 0.02; Figure 6C) and the high group showed better RFS compared to the low and negative group (log‐rank test, p = 0.009; Figure 6D). Furthermore, both univariable analysis (HR = 3.0, 95% CI = 1.3–7.0, p = 0.01; Table 5) and multivariable Cox regression analysis (HR = 3.1, 95% CI = 1.3–7.3, p = 0.01; Table 5) confirmed that lower or absent pks ⁺ E. coli in tumor tissue was significantly associated with shorter RFS.

FIGURE 6.

The influence of polyketide synthetase (pks)⁺ Escherichia coli copy number on relapse‐free survival (RFS) of Stages I–III patients. Kaplan–Meier survival curves for RFS rates of Stages I–III patients according to the abundance of pks ⁺ E. coli divided by high, low, negative (A), and high, both low and negative (B), and of Stages II–III patients according to the abundance of pks ⁺ E. coli divided by high, low, negative (C), and high, both low and negative (D). Each statical analysis included a log‐rank test.

TABLE 5.

Univariate and multivariate analyses for predicting relapse‐free survival in Stages II–III colorectal cancer patients: Cox regression model results

Variables	Univariate			Multivariate
	HR	95% CI	p‐value	HR	95% CI	p‐value
pks + Escherichia coli (low, negative vs. high)	3.0	1.3–7.0	0.01	3.1	1.3–7.3	0.01
Tumor location (right vs. left, rectum)	1.0	0.6–1.7	0.89	–	–	–
Tumor depth (T4 vs. Tis‐3)	4.3	2.5–7.4	<0.001	3.2	1.8–5.6	<0.001
Lymph node metastasis (presence vs. absence)	3.2	1.7–5.9	<0.001	2.2	1.1–4.1	0.02
Pathology (others vs. well to moderate)	1.1	0.5–2.2	0.82	–	–	–
Venous invasion (presence vs. absence)	3.1	1.3–6.9	0.005	2.8	1.3–6.4	0.01
Lymphatic invasion (presence vs. absence)	3.5	1.5–8.2	0.004	1.9	0.8–4.6	0.18
Perineural invasion (presence vs. absence)	2.2	0.9–5.2	0.07	–	–	–
RAS status (MT vs. WT)	0.9	0.5–1.6	0.76	–	–	–
BRAF status (MT vs. WT)	0.6	0.2–2.0	0.42	–	–	–

Abbreviations: CI, confidence intervals; HR, hazard ratio; pks, polyketide synthetase.

The prognostic influence pks ⁺ E. coli on normal mucosa was evaluated in a cohort of 244 patients who had available samples of normal mucosal tissue. Patients with pks ⁺ E. coli positive in their normal mucosal tissue had significantly better CSS (log‐rank test, p = 0.019; Figure 7A). In Stages II and III patients, patients with pks ⁺ E. coli positive in their normal mucosal tissue had equivalent RFS compared to patients with pks ⁺ E. coli negative in their normal mucosal tissues (log‐rank test, p = 0.53; Figure 7B).

FIGURE 7.

The influence of polyketide synthetase (pks)⁺ Escherichia coli copy number of normal mucosal tissue on cancer‐specific survival (CSS) of Stages I–IV patients and on relapse‐free survival of Stages II–III patients. (A) Patients with pks ⁺ E. coli positive in their normal mucosal tissue had significantly better CSS (log‐rank test, p = 0.019). (B) In Stages II and III patients, patients with pks ⁺ E. coli positive in their normal mucosal tissue had equivalent RFS compared to patients with pks ⁺ E. coli negative in their normal mucosal tissue (log‐rank test, p = 0.53).

We explored the relationship between total copy number of E. coli (including pks ⁺ E. coli and pks ⁻ E. coli) and the prognosis of patients with available residual specimens (a total of 179 patients), finding that E. coli does not have a significant effect on prognosis (p = 0.88; Figure S3).

4. DISCUSSION

This is the first study to examine the variance in the abundance of pks ⁺ E. coli between tumor and normal mucosal tissue, as well as its effect on CRC prognosis, using digital PCR. The detection rate of pks ⁺ E. coli via digital PCR was comparable to, or better than, that in prior studies on tumor and normal mucosal tissue. ¹⁴ , ¹⁵ , ¹⁹ Our results suggest that tumor tissue in Stages 0–I CRC contained higher levels of pks ⁺ E. coli compared to both normal mucosal tissue and more advanced stages of CRC. The abundance of pks ⁺ E. coli in tumor tissue was significantly associated with shallower tumor depth and reduced lymph node metastasis. Additionally, patients with elevated levels of pks ⁺ E. coli in tumor tissue had better cancer‐specific and RFS.

The present results suggest that pks ⁺ E. coli might contribute to carcinogenesis but might not influence cancer progression. To date, no studies have compared the presence of pks ⁺ E. coli in CRC tissue with its presence in normal mucosal tissue. We clearly demonstrated that more abundant pks ⁺ E. coli was found in tumor tissue compared to normal mucosal tissue in Stages 0–I cases; however, this trend was not observed in Stage II or higher cancers. The increased presence of pks ⁺ E. coli in tumor tissue during Stages 0–I might indicate that the local accumulation of this bacteria by selective sweep, as previously described, ²⁰ contributes to carcinogenesis. Numerous researchers have concluded that pks ⁺ E. coli, which produces the genotoxic substance, colibactin, is involved in carcinogenesis of CRC. ⁸ , ²¹ , ²² , ²³ , ²⁴ Colibactin causes DNA double‐strand breaks at positions corresponding to somatic mutation hot spots in the CRC genome. ⁸ When human intestinal organoids were infected with pks ⁺ E. coli, the mutation signature aligned with many oncogenic driver gene mutations found in human CRC. ²¹ Our study, using clinical samples, corroborates these findings from basic studies.

Interestingly, pks ⁺ E. coli abundance was lower in advanced cancers. As of now, no studies have explored the relationship between abundance of pks ⁺ E. coli and tumor progression. This observation suggests that pks ⁺ E. coli might not contribute significantly to progression of CRC. As the cancer advances, its cellular diversity increases. ²⁰ This diversification could result in a decreased proportion of tumor cells infected with pks ⁺ E. coli. Our study did not yield findings that corroborate this hypothesis, emphasizing the need for further investigations using animal models or organoids.

In tumor tissue, greater pks ⁺ E. coli copy number was associated with favorable prognosis for CRC. In this study, the group with high numbers of pks ⁺ E. coli exhibited both longer CSS and longer recurrence‐free survival compared to the groups with a low number or no pks ⁺ E.coli. Moreover, highly abundant pks ⁺ E. coli was an independent predictor of favorable prognosis. Impressively, Patients with pks ⁺ E. coli positive in their normal mucosal tissue had significantly better CSS. Only one other study has investigated the association between the presence of pks ⁺ E. coli in tumor tissue and CRC prognosis. ¹⁸ That study also reported favorable CSS associated with abundant pks ⁺ E. coli. No studies have examined the prognostic effect of pks ⁺ E. coli on normal mucosal tissue. Our analysis of tumor tissue and normal mucosal tissue suggests that tumors implicated in carcinogenesis by pks ⁺ E. coli might have a favorable prognosis. Furthermore, total copy number of E. coli does not have a significant effect on prognosis. Nevertheless, research on pks ⁺ E. coli, particularly regarding pks ⁺ E. coli of normal mucosal tissue is limited, and it is premature to draw definitive conclusions. While no association was found between pks ⁺ E. coli and CD3‐positive or CD8‐positive cells, the relationship between the tumor microenvironment and pks ⁺ E. coli remains unclear (Figure S4).

Although various studies have clarified the relationship between gut bacteria and CRC progression, none have clarified the relationship between pks ⁺ E. coli and tumor proliferation. ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ Fusobacterium nucleatum (F. nucleatum) promotes cancer proliferation through activation of β‐catenin signaling via FadA and Fap2 and by suppressing anti‐tumor immune responses. ³¹ , ³² Additionally, F. nucleatum inhibits E‐cadherin expression, while promoting transcription factors associated with epithelial–mesenchymal transition. ³³ , ³⁴ Parvimonas micra activates the Ras/ERK/c‐Fos signaling pathway by upregulating expression of miR‐218‐5p, which subsequently inhibits expression of protein tyrosine phosphatase receptor type R. ³⁵ Consequently, CRCs with high P. micra expression are associated with poorer prognosis. ²⁷ , ²⁸

In this study, by using ddPCR and frozen samples, we detected pks ⁺ E. coli in a more tumor tissue than prior research. This facilitated more precise analysis of pks ⁺ E. coli, given that it is identifiable even in samples with only trace abundance. The positivity rate of pks ⁺ E. coli in tumor tissue in this study was 65.6%, markedly higher than the 9.4% in a previous study that used quantitative PCR with formalin‐fixed, paraffin‐embedded samples. ¹⁹ The positivity rate of pks ⁺ E. coli in normal mucosal tissue in our study (62.3%) was consistent with earlier studies that had a smaller patient count (38 and 21 patients, respectively). ¹⁴ , ¹⁵ This is likely because, as our study indicates, the level of pks ⁺ E. coli in normal mucosal tissue exceeds that in tumor tissue.

This study has several limitations. First, it was a single‐center study. Although clinically diagnosed hereditary tumors were excluded, the absence of genetic testing means that some might still have been included. In our survival analysis, we did not consider the effects of postoperative chemotherapy, primarily because the sample size was too small to segregate into two groups based on whether chemotherapy was administered. This study did not account for postoperative complications that could influence prognosis. Additionally, this study did not include patients who underwent endoscopic treatment. However, obtaining frozen samples from those treated with endoscopy might have adversely affected clinical diagnosis.

In conclusion, tumor tissue in Stages 0–I CRC contained higher levels of pks ⁺ E. coli compared to both normal mucosal tissue and more advanced stages of CRC. Additionally, high levels of pks ⁺ E. coli were associated with a favorable prognosis. These results suggest that pks ⁺ E. coli might contribute to CRC carcinogenesis, but it might not exacerbate tumor progression.

AUTHOR CONTRIBUTIONS

Toshimitsu Miyasaka: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; visualization; writing – original draft. Takeshi Yamada: Conceptualization; data curation; project administration; validation; writing – review and editing. Kay Uehara: Investigation. Hiromichi Sonoda: Investigation; writing – original draft. Akihisa Matsuda: Conceptualization; investigation; writing – original draft. Seiichi Shinji: Investigation. Ryo Ohta: Investigation. Sho Kuriyama: Conceptualization; data curation; investigation; methodology. Yasuyuki Yokoyama: Investigation. Goro Takahashi: Investigation. Takuma Iwai: Investigation. Kohki Takeda: Investigation. Koji Ueda: Investigation. Shintaro Kanaka: Investigation; writing – original draft. Ryuji Ohashi: Formal analysis; validation. Hiroshi Yoshida: Supervision; writing – review and editing.

FUNDING INFORMATION

This study was supported by the Department of Gastrointestinal and Hepato‐Biliary‐Pancreatic Surgery, Nippon Medical School, Tokyo, Japan.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENT

Approval of the research protocol by an institutional review board: The study was approved by the Institutional Review Board of the Nippon Medical School Hospital (approval no. B‐2022‐506).

Informed Consent: N/A.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.

Supporting information

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Table S1.

Data S1.

Miyasaka T, Yamada T, Uehara K, et al. Pks‐positive Escherichia coli in tumor tissue and surrounding normal mucosal tissue of colorectal cancer patients. Cancer Sci. 2024;115:1184‐1195. doi: 10.1111/cas.16088