Fecal Cloacibacillus porcorum Improves Non-Invasive Diagnosis of Colorectal Adenoma in the Hong Kong Population.

We analyzed fecal metagenomic sequencing data from 492 Hong Kong Chinese subjects, including 170 patients with CRC, 142 with adenoma (63 with non-advanced adenoma (nAA) and 79 with advanced adenoma (AA)), and 180 control subjects [ 22 ]. Relative species abundances were determined using MetaPhlAn (version 4.0, Segata Lab, University of Trento, Trento, Italy) [ 23 ]. Fecal samples were collected from 426 Hong Kong Chinese subjects (127 with CRC, 161 with adenoma, and 138 normal controls) at the Prince of Wales Hospital, the Chinese University of Hong Kong, between 2009 and 2019. Subjects recruited for fecal sample collection included individuals presenting with symptoms such as a change in bowel habits, rectal bleeding, abdominal pain, or anemia, as well as asymptomatic individuals aged 50 years or above undergoing screening colonoscopy, as described in our previous metagenomic study [ 5 ]. Samples were collected either before colonoscopy or one month afterward, when the gut microbiome would be expected to have recovered to baseline [ 17 ]. The exclusion criteria were as follows: (1) use of antibiotics within the previous 3 months; (2) adherence to a vegetarian diet; (3) invasive medical intervention within the previous 3 months; (4) a history of cancer or inflammatory disease of the intestine. Subjects were instructed to collect stool samples in standardized containers at home and immediately store them in their home freezer at −20 °C. Frozen samples were then delivered to the hospitals in insulating polystyrene foam containers and immediately stored at −80 °C until further analysis. Patients were diagnosed by colonoscopic examination and histopathological review of biopsy specimens, when applicable. Informed consent was obtained from all subjects. The study was approved by the Joint Chinese University of Hong Kong, New Territories East Cluster Clinical Research Ethics Committee (the Joint CUHK-NTEC CREC, CREC Ref. Nos.: 2021.126 and 2017.369). DNA extraction, design of primer and probe sequences, and qPCR amplifications on an ABI QuantStudio sequence detection system were performed as previously described [ 13 ]. Primer and probe sequences specifically targeting C. porcorum and other markers are listed in Table S7 . Primer and probe sequences targeting other bacterial gene markers and the 16s rDNA internal control were adopted from our previous study [ 13 ]. Each probe carried a 5′ reporter dye FAM (6-carboxy fluorescein) or VIC (4,7,2′-trichloro-7′-phenyl-6-carboxyfluorescein) and a 3′ quencher dye TAMRA (6-carboxytetramethyl-rhodamine). Primers and hydrolysis probes were synthesized by Invitrogen (Carlsbad, CA, USA). PCR amplification specificity was confirmed by direct Sanger sequencing of PCR products or by sequencing randomly selected TA clones. The relative level of each marker was calculated using the delta Cq method relative to the internal control and expressed as the log value of ‘*10 × 10 6 + 1’. Stool samples ( n = 326) were examined using FIT with the automated quantitative OC-Sensor test (Eiken Chemical, Tokyo, Japan). The quantitative OC-Sensor test was performed as previously described [ 24 ], with a positive cutoff value equivalent to 100 ng hemoglobin per milliliter. The combined score for four bacterial markers (4Bac) using a logistic regression model (4Bac score = I 1 + β 1 × Fn + β 2 × m3 + β 3 × Bc + β 4 × Ch ) was determined in a previous study [ 16 ]. The combined scores for two to five markers with or without FIT using logistic regression models are listed in Table S8 . In the regression models, ‘I’ represents the intercept, ‘β’ represents the regression coefficients, and the markers represent the corresponding Cq values. Cutoff values were determined by receiver operating characteristic (ROC) curve analyses that maximized the Youden index (J = Sensitivity + Specificity − 1) [ 25 ]. Values are expressed as mean ± SD or median (interquartile range (IQR)), as appropriate. Differences in bacterial abundances were determined using the Mann–Whitney U test for comparisons between two groups or the Kruskal–Wallis test with Dunn’s correction for comparisons among three groups. The Jonckheere–Terpstra trend test was used to evaluate changes in marker levels during disease progression from control to adenoma to CRC. Simple and multiple regression analyses were performed to estimate associations between marker levels and factors of interest. Incidence rates among different groups and sensitivities of different markers were analyzed using Fisher’s exact test (two groups) or the Chi-square test (three groups). Combinations of multiple biomarkers were evaluated using logistic regression models to generate values estimating disease incidence relative to controls. ROC curves were used to evaluate the diagnostic performance of bacterial markers and models in distinguishing CRC or adenoma from controls. Pairwise comparisons of ROC curves were performed using a nonparametric approach [ 26 ]. All analyses were conducted using Graphpad Prism 9.5 (Graphpad Software Inc., San Diego, CA, USA) or MedCalc Statistical Software version 18.5 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc.org ; 2018). p < 0.05 was considered statistically significant.

Colorectal cancer (CRC) is one of the most common malignancies worldwide [ 1 ]. Most CRCs begin as small polyps. Some polyps, particularly adenomas, gradually develop into cancer. Early cancer detection can facilitate successful treatment, whereas early adenoma detection can prevent CRC and reduce its incidence. However, although currently available non-invasive CRC screening tests perform well in detecting CRC, their sensitivity for adenoma detection remains limited. CRC incidence is higher in more developed regions than in less developed regions, and this increased incidence is believed to be associated with dietary differences [ 2 , 3 ]. Recent evidence has shown that an altered gut microbiome is associated with colorectal tumorigenesis. Abnormalities in gut microbiota composition have been implicated as potentially important etiological factors in the initiation and progression of CRC [ 4 ]. With the widespread application of metagenomic analyses to investigate the intestinal microbiota, an increasing number of bacterial species have been identified as being positively associated with CRC [ 5 , 6 , 7 , 8 ]. Recent basic research has established a critical role for the intestinal microbiota [ 9 ] and specific bacterial species, such as Fusobacterium nucleatum ( Fn ) [ 10 , 11 , 12 ], in promoting colorectal tumorigenesis. Bacteria such as Fn [ 13 ], Clostridium symbiosum [ 14 ], and species within the genera Parvimonas , Porphyromonas , and Parabacteroides [ 15 ] have been identified as potential markers for the CRC diagnosis. However, current knowledge regarding biomarkers for colorectal adenoma detection remains limited. We previously identified and validated bacterial markers for the non-invasive diagnosis of CRC and adenoma using metagenome sequencing and targeted qPCR [ 5 , 13 ]. Specifically, we developed a qPCR test targeting four bacterial markers for CRC and adenoma diagnosis: Fn, the bacterial gene marker m3, Clostridium hathewayi (Ch), and Bacteroides clarus (Bc) [ 16 ]. Three of these markers (Fn, m3, and Ch) are enriched in the stools of patients with CRC or adenoma, whereas Bc is enriched in healthy individuals. Although this four-marker panel demonstrates superior performance compared with currently available non-invasive diagnosis tests for CRC and adenoma, further improvement in adenoma detection sensitivity is still needed. Therefore, this study aimed to identify additional bacterial species based on metagenomic analyses that could enhance adenoma diagnosis and to develop new bacterial marker panels for CRC and adenoma using targeted quantification via qPCR.

To identify bacteria enriched during normal–adenoma–CRC progression as potential biomarkers for CRC screening, we analyzed metagenomic data from 180 controls, 142 patients with adenoma, and 170 patients with CRC in a Hong Kong Chinese population. ANCOM-BC2 identified 16 species with known classifications that were significantly elevated in both adenoma and CRC compared with normal controls ( Figure 1 A). Spearman’s rank correlation analysis identified 17 candidate species exhibiting increasing trends across normal–adenoma–carcinoma progression ( p < 0.0001; Figure 1 B). Both ANCOM-BC2 and Spearman’s rank correlation analyses consistently identified three species enriched across disease stages: Cloacibacillus porcorum (Cp), Cloacibacillus evryensis , and Dialister pneumosintes ( Figure 1 C). Among these, only Cp was significantly increased in both adenoma and CRC compared with normal controls, whereas C. evryensis and D. pneumosintes were not significantly increased in adenoma compared with controls ( Figure 1 D). Among the identified candidates, 21 species exhibited increasing trends along the normal–adenoma–carcinoma axis, as determined by the Jonckheere–Terpstra trend test ( Table S1 ), and their relative abundances were further compared between groups ( Figure S1 ). Desulfovibrio fairfieldensis , identified by ANCOM-BC2, was significantly increased in both adenoma and CRC compared with normal controls. Intestinimonas butyriciproducens , identified by Spearman’s rank correlation analysis, exhibited an increasing trend, although the increase in adenomas compared with controls was not statistically significant ( Figure 1 D). Among these five candidates, the incidence rates of Cp, D. fairfieldensis , and I. butyriciproducens were significantly higher in the adenoma and CRC groups than in the control group ( Figure 1 E). Because the incidence rate of D. fairfieldensis was less than 15% in both the adenoma and CRC groups, we focused on evaluating Cp and I. butyriciproducens as potential candidates for the non-invasive diagnosis of CRC and adenoma. Cp and I. butyriciproducens have not previously been used as biomarkers for the diagnosis of colorectal neoplasms. Therefore, we established qPCR assays for their targeted quantification. Primers and probes were designed to target the selected species-specific marker genes based on MetaPhlan. The qPCR results exhibited significant positive correlations between the quantification of the two species and their relative abundances obtained from metagenome sequencing ( n = 280; both p < 0.0001; Figure 2 A). The qPCR results further demonstrated that the relative levels of both Cp and I. butyriciproducens significantly differed among the three groups. However, only Cp was significantly increased in both adenoma and CRC samples compared with controls (Kruskal–Wallis test, p < 0.05; Figure 2 B). Importantly, the relative level of Cp assessed by qPCR also showed a significant increasing trend from normal to adenoma and CRC ( p < 0.0001 by the Jonckheere–Terpstra trend test). ROC analysis indicated that Cp could significantly distinguish CRC and adenoma from controls, whereas I. butyriciproducens could only distinguish CRC from controls ( Figure 2 C). These results highlight the potential of Cp as a non-invasive diagnostic marker for CRC and adenoma. According to the qPCR results, the prevalence of Cp was significantly higher in patients with adenoma and CRC than in normal controls ( Figure 3 (A1)), consistent with the metagenome sequencing results ( Figure 3 (A2)). The relative level of Cp, which was significantly elevated in patients with adenoma or CRC compared with controls, showed no significant difference between non-advanced and advanced adenomas or among different TNM stages, as assessed via qPCR ( Figure 3 (B1)). These findings are consistent with the metagenome sequencing results ( Figure 3 (B2)). We further assessed the diagnostic performance of Cp using qPCR and compared it with the three CRC-enriched bacterial markers (Fn, m3, and Ch) from our previous 4Bac panel [ 16 ]. Fecal Cp alone yielded AUCs of 0.657 (95%CI: 0.583 to 0.725; p < 0.0001) for CRC and 0.618 (95%CI: 0.550 to 0.682; p < 0.0001) for adenoma ( Figure 4 ). At cutoff values maximizing the Youden index, Cp showed sensitivities of 37.1% at a specificity of 93.6% for CRC and 30.8% at a specificity of 92.3% for adenoma. We then compared Cp with the markers Fn, m3, and Ch. For distinguishing CRC from controls, Cp was less effective than Fn and m3 and showed no significant difference compared with Ch. For distinguishing adenoma from controls, m3 performed best, whereas Cp showed no significant difference compared with Fn and was significantly better than Ch ( p = 0.0335) ( Figure 4 ). The relative levels of all four markers (Fn, m3, Ch, and Cp) were significantly associated with CRC and adenoma diagnosis, and none were associated with sex, CRC stage, lesion location, or body mass index in univariate analysis ( Table S2 ). While Fn, m3, and Ch significantly increased with age, Cp showed no association with age. Multivariate analysis confirmed that Fn, m3, Ch, and Cp were all independently associated with CRC and adenoma diagnosis, while Fn was also associated with age ( Table S3 ). Further univariate and multivariate analyses of diagnostic factors confirmed that the diagnostic value of Cp is also independent of other markers ( Table S4 ). Logistic regression further demonstrated that Cp is an independent marker for the diagnosis of both adenoma and CRC ( Table S5 ). We further assessed combinations of Cp with markers from our previous 4Bac panel (Fn, Ch, m3, and Bc) for CRC and adenoma diagnosis ( Table S6 ). Logistic regression models were constructed to distinguish patients with CRC/adenoma cases from controls, incorporating all five markers or subsets thereof, with individual markers sequentially removed based on their importance in the model. The results showed that the five-marker model (Fn, Ch, Bc, m3, and Cp) performed best for CRC diagnosis, achieving an AUROC of 0.923 (all p < 0.05 compared with combinations containing fewer markers, as demonstrated by ROC curve comparisons; Figure 5 A,B). For adenoma, the five-marker model showed no significant difference compared with the four- and three-marker panels when Ch and/or Bc were removed ( Figure 5 A,B). Comparisons between the two- and three-marker models, involving Fn and m3 with and without Cp, showed no significant difference for CRC diagnosis. However, inclusion of Cp significantly improved diagnostic performance for adenoma ( Figure 5 A,C). We further compared the diagnostic performance of the 5Bac model incorporating Cp with the 4Bac model and m3, the key marker for adenoma diagnosis. Although the 5Bac model (4Bac + Cp) showed no significant difference compared with 4Bac in CRC diagnosis ( p > 0.05 by ROC curve comparison), it exhibited slightly higher sensitivity (88.6% vs. 85.7%) at 85% specificity ( Figure 6 ). For adenoma diagnosis, the addition of Cp significantly improved the diagnostic performance of 4Bac ( p = 0.002 by ROC curve comparison). Additionally, the 5Bac model performed significantly better than m3 alone for adenoma ( p = 0.048). At 85% specificity, the 5Bac model showed higher sensitivity for adenoma (58.7%) than 4Bac (44.8%) and m3 (41.6%) ( Figure 6 ). We further evaluated the role of Cp combined with FIT in improving the diagnostic performance of the 4Bac model for CRC and adenoma. Logistic regression models were constructed by combining bacterial markers with FIT to distinguish CRC or adenoma cases from normal controls. Performance was assessed separately for CRC, AA, and nAA, as FIT can detect a small proportion of AA but not nAA [ 16 ]. When combined with FIT, the 5Bac model (4Bac + Cp) showed no significant difference compared with 4Bac in CRC diagnosis (5Bac + FIT vs. 4Bac + FIT). However, both models performed significantly better than their corresponding models without FIT (both p < 0.001 by ROC curve comparisons) ( Figure 7 A). For advanced adenoma, both FIT and Cp significantly improved the diagnostic performance of 4Bac (both p < 0.05), and the 5Bac + FIT model performed significantly better than all other models (all p 80%), models including Cp showed higher sensitivity ( Figure 7 A). In our tested cohort, FIT alone detected 72.4% of CRC, 19.3% of AA, and none of the non-advanced adenoma cases at a specificity of 98.7% ( Figure 7 B). Comparison between 5Bac + FIT and 4Bac showed that the inclusion of Cp and FIT significantly increased detection rates for CRC and AA (both p < 0.01). When comparing models with and without FIT, it was evident that FIT improved CRC detection ( p = 0.041 for 4Bac + FIT vs. 4Bac; p = 0.066 for 5Bac + FIT vs. 5Bac). When comparing models with and without Cp, it was evident that Cp improved detection rates for AA and nAA, although the increase was significant only for AA ( p = 0.019 for 5Bac + FIT vs. 4Bac + FIT) ( Figure 7 B). The diagnostic sensitivities of FIT, bacterial markers, and their combinations for CRC were further compared across TNM stage subsets ( Figure 7 C). The bacterial marker models demonstrated higher sensitivities than FIT for stage I–III cancers but not for late-stage IV disease. The combination of 4Bac or 5Bac with FIT significantly increased sensitivity for stage I–III cancers and also improved detection of stage IV cancers, although not significantly. Cp improved the sensitivity of 4Bac for detecting stage II–IV cancers, although not significantly. These results demonstrate that bacterial marker panels outperform FIT for detecting stage I–III CRC, and their combination further enhances non-invasive CRC diagnosis.

In this study, we identified novel bacterial species markers for the diagnosis of CRC and adenoma using metagenomic analysis and further validated them through targeted quantification using qPCR. Among the identified and validated candidate species, C. porcorum emerged as the most promising marker, showing significant increases in both adenoma and CRC samples compared with controls. We compared the diagnostic performance of Cp with that of our previously identified bacterial markers, including m3, Fn, Ch, and Bc. Additionally, we developed new bacterial marker panels, with or without FIT, for the diagnosis of CRC and adenoma. Our findings indicated that the addition of Cp significantly enhanced the diagnostic performance of the previously identified bacterial markers for CRC and adenoma, including both non-advanced and advanced adenomas, whereas combining these markers with FIT further improved the diagnosis of CRC and advanced adenoma. Targeted detection of bacterial markers identified through shotgun metagenomics represents a promising strategy for clinical application due to its cost-effectiveness and ease of implementation. In this study, qPCR-based quantification of the bacterial markers demonstrated good performance for the diagnosis of CRC and adenoma. In particular, the panel including Fn, m3, and Cp yielded AUCs of 0.897 (95%CI: 0.844 to 0.937) for CRC and 0.770 (95%CI: 0.709 to 0.824) for adenoma. The further addition of Bc and/or Ch increased the AUCs for CRC to above 0.92. Combining these markers with FIT further increased the diagnostic sensitivities for CRC and advanced adenoma; however, detection of non-advanced adenomas was not improved due to the known limitations of FIT in this group. The optimal panel, which included all five bacterial markers and FIT, achieved sensitivities of 64.3% for adenoma and 96.2% for CRC without compromising specificity. This improvement was primarily driven by Cp, which is independent of fecal hemoglobin and other bacterial markers. By detecting non-bleeding or FIT-negative lesions, Cp may help address a critical diagnostic gap in early adenoma screening. The maintenance of high specificity (84.6%) alongside improved detection suggests that this approach may offer a favorable net clinical benefit without imposing an excessive burden on downstream colonoscopy services. However, its meaningful clinical utility needs to be evaluated in future prospective studies to confirm its real-world effectiveness and cost-efficiency by assessing downstream colonoscopy demand and net clinical benefit. Because the true performance of the markers cannot be accurately determined from case–control samples alone, rigorous external validation in large, independent, multi-ethnic cohorts is essential to ensure robustness and generalizability across diverse screening populations. Given the single-time-point sampling design of this study, longitudinal studies are needed to confirm the temporal stability of Cp. Such studies should also collect detailed dietary and medication histories to control for potential transient confounding factors. Although some normal control individuals showed relatively high levels of Cp, these individuals also tended to have elevated levels of other CRC-associated bacteria included in the 4Bac panel; therefore, adding Cp improved sensitivity without compromising the specificity of the 4Bac panel. In our cohort, multivariate and logistic analyses confirmed that Cp was independently associated with CRC and adenoma and showed no significant association with potential confounding factors, including age, sex, body mass index, CRC stage, or lesion location. Samples in this study were collected either before colonoscopy or one month afterward, when the gut microbiome would be expected to have recovered to baseline [ 17 ]. Potential confounding factors, including bowel preparation, dietary patterns, underlying disease conditions, and population heterogeneity, may influence microbiome profiles. Future prospective studies should rigorously control for these variables to confirm the specificity of Cp. The current study focused on individuals aged ≥50 years, as this age group represents the primary target population for routine CRC screening. However, the performance of Cp as a diagnostic marker for CRC and adenomas in individuals younger than 50 years remains unclear, particularly among high-risk groups with early-onset CRC, such as those with hereditary cancer syndromes. Future studies involving younger populations are therefore warranted to further define the diagnostic utility of Cp for early-onset CRC. Cp, a mucin-degrading bacterium, was first isolated from the intestinal tract of swine [ 18 ]. Its role in humans remains largely unknown, and it has primarily been reported in animals. Following fecal microbiota transplantation that improved chronic diarrhea in a cynomolgus monkey, the relative abundance of Cp in the gut microbiota was significantly reduced [ 19 ]. As a mucin-degrading bacterium, enrichment of Cp may compromise the integrity of the intestinal mucus barrier, thereby potentially facilitating the translocation of pro-inflammatory factors or pathogens that promote epithelial cell transformation. Notably, Cp has been associated with epizootic rabbit enteropathy and bacteriemia in humans [ 20 , 21 ], supporting its potential role as an opportunistic pathogen. These findings suggest a potential pathogenic role for Cp. However, further in vitro and in vivo functional investigations are warranted to determine whether Cp contributes to colorectal tumorigenesis.

This study identified C. porcorum as a valuable marker for the non-invasive diagnosis of CRC and adenoma using metagenomic analysis. Additionally, we developed a new bacterial marker panel based on targeted quantification via qPCR, which has potential for implementation in CRC screening.