*2.1. Patient Characteristics*

The study cohort consisted of 41 newly diagnosed, treatment-naive CC patients scheduled for elective surgery and included 23 male (56.1%) and 18 female (43.9%) patients, of whom 24 patients were with RSCC and 17 patients were with LSCC (only CC, no rectal cancer patients), aged from 39 to 90 years. The baseline clinical and pathological characteristics are shown in Table 1; no significant differences were observed between patients with RSCC and LSCC.

**Table 1.** Clinicopathological characteristics of the study participants.



**Table 1.** *Cont.*

## *2.2. Microbiome Profile of the Study Cohort*

2.2.1. The Microbiome Landscape across the Locations

First, we analyzed all sample types (ileal tissue, healthy colon tissue, healthy rectal tissue, tumor tissue, preoperative stool and postoperative stool). The most abundant phyla in all samples were *Firmicutes* and *Bacteroidetes*, followed by *Actinobacteria*, *Proteobacteria, Verrucomicrobia* and *Fusobacteria*, to different degrees (Figure 1a). The microbiota profile at the genus level in all samples is shown in Figure 1b. The profile of the microbiota in the different analyzed samples differed from those found in the quality controls (mock community, water).

To estimate the richness and diversity of the different habitats, the alpha diversity indices were analyzed. We compared the Observed, Chao, ACE, Shannon, Simpson and Fisher indices of the different sample types at the genus level. The overall structure of the microbiota in the microhabitats was significantly different based on all indices: the Observed index (*p* value: 0.000002; (ANOVA) F value: 7.4813) (Figure 2a), the Chao1 index (*p* value: 0.00001; (ANOVA) F value: 6.4832), the ACE index (*p* value: 0.00004; (ANOVA) F value: 5.9738), the Shannon index (*p* value: 0.000000006; (ANOVA) F value: 10.664), the Simpson index (*p* value: 0.00000004; (ANOVA) F value: 9.6125) and the Fisher index (*p* value: 0.000002; (ANOVA) F value: 7.5182). The diversity was lowest in postoperative stool samples, which could be explained by the bowel preparation (mechanical and antibiotics) and surgical stress.

**Figure 1.** Taxonomic analysis of the microbiome in the different habitats of CC patients: represented at the phylum level (**a**) and genus level (**b**).

(**a**)

**Figure 2.** Microbiome diversity comparison between the locations of CC patients: alpha diversity box plot (Observed, *p* value < 0.001) (**a**) and principal coordinate analysis (PCoA) using Jensen–Shannon metric distances of beta diversity (**b**) at the genus level, *p* value < 0.001.

Moreover, a beta diversity analysis was performed. At the genus level, the analysis revealed that the overall structure of the microbiota in the analyzed habitats was significantly different (PCoA Jensen–Shannon (PERMANOVA) F value: 9.5743, R-squared: 0.22074, *p* value < 0.001; Figure 2b).

A linear discriminant analysis (LDA) coupled with effect size measurements (LEfSe) was applied to identify key taxa that were differentially abundant between the analyzed samples. A total of 46 key taxa were identified at the genus level (Figure 3, LDA score > 3, *p* value < 0.05, FDR-adjusted *p* value < 0.1; Supplementary Figure S1).

2.2.2. The Microbiome Communities Are Significantly Different between Tumor and Stool Samples

The early detection of CC is of great prognostic importance, and stool samples are a potential source of microbial biomarkers. We compared tumor tissue and preoperative stool samples and analyzed differences in the microbiota composition. The beta diversity comparisons showed significantly different bacterial community clusters between the tumor and stool samples (PCoA Jensen–Shannon divergence (PERMANOVA) F value: 18.721, R-squared: 0.19558, *p* value < 0.001, Figure 4a). The LEfSe analysis identified

43

35 genera whose abundances significantly differed between the tumor and stool samples (LDA score > 3, *p* value < 0.05, FDR-adjusted *p* value < 0.05; Figure 4b). No significant differences in the alpha diversity were observed between the tumor and stool samples (Figure 5a). The random forest classification machine learning algorithm was used to confirm the data. Using 120 trees, the algorithm achieved the best prediction with a classification error of 0.0253 (Supplementary Figure S1). The top five ranked genera to discriminate between stools and tumors were Flavonifractor, Oscillibacter, Odoribacter, Roseburia and Eggerthella (Supplementary Figure S2).

To determine whether the composition of the microbiome differs according to clinical factors, additional analyses were performed based on location (RSCC, LSCC) and pathologic parameters (T stage, differentiation, nodal stage, MSS status). The alpha diversity of the whole microbiome of the stool and tumor tissue was significantly different between the RSCC and LSCC groups (Observed index *p* value: 0.014561; (*t* test) statistics: 2.4996; Chao1 index *p* value: 0.017411; (*t* test) statistics: 2.4305). The MSS and MSI tumor groups were slightly but not significantly different (Chao1 index *p* value: 0.0508; (*t* test) statistics: −2.0505) (Figure 5b,c).

The tumor tissue of grade 3 tumors was significantly enriched in *Fusobacterium* and *Parvimonas*, while *Fusicatenibacter, Blautia, Intestimonas* and *Romboutsia* were significantly increased in grade 2 tumors (*p* value < 0.01, FDR-adjusted *p* value < 0.1). There was no significant difference among the T or N stages; we think that this was due to the stage-specific distribution: early T1/2 stages (*n* = 11) compared to T3/4 (*n* = 29) and more N-negative (*n* = 32) than N-positive patients (*n* = 9). In tumor tissue, no significant differences according to MSS status were observed.

In contrast, the preoperative stool of grade 2 patients was associated with *Dialister* and *Intestimonas*, while grade 3 tumors were significantly enriched in E. shigella (*p* value < 0.01, FDR-adjusted *p* value < 0.1). Furthermore, the stool of MSI patients was significantly enriched with *Clostridium\_XIVb* (*p* value < 0.01, FDR-adjusted *p* value < 0.1). Taken together, these findings suggest that the stool microbiome (preoperative) only partly reflects the tumor microbiome.

(**a**)

**Figure 4.** *Cont*.

**Figure 4.** Stool microbiome only partly reflects the microbiome landscape in CC patients: PCoA using Jensen–Shannon divergence of beta diversity between tumor and preoperative stool, *p* value < 0.001 (**a**), LEfSe detected marked differences in the predominance of bacterial communities between tumor and preoperative stool, *p* value < 0.05 (**b**).

**Figure 5.** Microbiome composition according to tumor sidedness and MSS status: (**a**) diversity analysis using the Chao1 alpha diversity index between tumor and preoperative stool; (**b**) overall microbiome of the tumor and preoperative stool according to sidedness RSCC and LSCC; and (**c**) overall microbiome of the tumor and preoperative stool according to MSS status (\* *p* < 0.05, n.s, not significant).

The core microbiome, based on sample prevalence (>50%) and relative abundance (0.01%), is displayed in Figure 6. The core analysis revealed six genera as the core taxa across all samples. Among them, *Parabacteroides* was prevalent in more than half of the samples from the RSCC patients, while *Bifidobacterium* and *Roseburia* were prevalent in more than half of the LSCC patients. Taken together, these findings indicate that RSCC and LSCC harbored a diverse core microbiome, with *Bacteroides* as the predominant genus (Figure 6) in both.

### 2.2.3. The Tumor Microbiome Profile: Significant Differences between RSCC and LSCC

For a deeper understanding of the intratumoral microbiome, we further analyzed the tumor tissue and sidedness (Figure 7). We first assessed the general tumor landscape. The top taxa in RSCC patients (Figure 7b) at the genus level were *Bacteroides* (15%), *Ruminococcus2* (10%), *Blautia* (8%), *Peptostreptococcus* (7%) and *Veillonella* (5%), and the top taxa in LSCC patients (Figure 7c) were *Blautia* (15%), *Bacteroides* (11%), *Streptococcus* (7%), *Parvimonas* (7%) and *Fusobacterium* (6%). The MSI patients (Figure 7d) harbored *Bacteroides* (18%), *Clostridium\_XIVa* (11%), *Corprococcus* (9%) and *Blautia* (8%), while in the MSS patients (Figure 7e), the top taxa were *Bacteroides* (12%), *Blautia* (12%), *Ruminococcus2* (6%) and *Peptostreptococcus* (6%).

**Figure 7.** Taxonomic analysis of the tumor microbiome composition: Pie chart showing the abundance profile of the tumor samples (**a**), RSCC subgroup (**b**), LSCC subgroup (**c**), MSI subgroup (**d**) and MSS subgroup (**e**) at the genus level.

A comparison of alpha diversity revealed significant differences between RSCC and LSCC at the genus level. Based on the Chao1 (*p* value: 0.018981; (*t* test) statistics: 2.4735;

Figure 8) and Observed (*p* value < 0.05) indices, the alpha diversity was significantly higher in LSCC than in RSCC (Figure 8a). There were no significant differences in alpha diversity based on sex, age, T stage, N stage or differentiation, while for the MSS status, these indices were significantly different (Chao1 index *p* value: 0.014618; (*t* test) statistics: −2.8349, Figure 8b).

**Figure 8.** Tumor microbiome diversity comparison: the alpha diversity analysis revealed significant differences between RSCC and LSCC (**a**) and between MSS and MSI patients (**b**) at the genus level (\* *p* < 0.05).

The differential abundance analysis, which shows the highest power to compare groups, especially for less than 20 samples per group, revealed a significant increase in the abundance of *Haemophilus* and *Veilonella* in the tumor tissue of RSCC patients, while increased *Bifidobacterium, Akkermansia, Roseburia* and *Ruminococcus* were associated with LSCC (genus level, *p* value < 0.001, FDR-adjusted *p* value < 0.05). The FDR-adjusted LEfSe analysis revealed two significantly different genera, *Bifidobacterium* and *Romboutsia*, in LSCC patients (genus level, *p* value < 0.05, LDA > 3.0, FDR-adjusted *p* value < 0.05). The original LEfSe analysis revealed 10 significantly different genera: *Bifidobacterium, Romboutsia, Clostridium\_III, Ruminococcus, Anaerostipes, Akkermansia, Clostridium\_sensu\_stricto* and *Asaccharobacter* in LSC patients and *Haemophilus* and *Veillonella* in RSCC patients (genus level, *p* value < 0.05, LDA > 3.0). In regard to MSS status, the original LEfSe analysis revealed seven significantly different genera: *Asaccharobacter, Actinomyces, Eubacterium, Pseudoflavonifractor, Fusicatenibacter* and *Anaerostipes* in tumor specimens from the MSS patients and Clostridium\_III in tumor tissue from the MSI patients (genus level, *p* value < 0.05, LDA > 3.0). The FDR-adjusted LEfSe revealed no significant differences. The abundances of *Fusobacterium, Peptostreptococcus* and *Desulfotomaculum* were significantly different in grade 3 tumor specimens (original LEfSe, genus level, *p* value < 0.05, LDA > 3.0), but no significant differences were identified based on the FDR-adjusted *p* value (<0.05).

## 2.2.4. The Microbiome of the Terminal Ileum: Tumor-Associated Alterations

We next assessed the general ileum landscape (Figure 9). The most abundant phylum was Firmicutes, followed by Bacteroidetes and Proteobacteria. The top 5 taxa at the family level were *Lachnospiraceae* (32%), *Streptococcaceae* (18%), *Bacteroidaceae* (8%), *Enterobacteriaceae* (8%) and *Verrucomicrobiaceae* (6%) (Figure 9a). The terminal ileum core microbiota, defined as genera with a threshold over 50%, are displayed in Figure 10. The typical ileal microbiota is dominated by the facultative anaerobic genus *Streptococcus* and the strict anaerobic genera *Bacteroides*, *Lachnospiraceae\_incertae\_sedis* and *Clostridium cluster XIV*.
