IDDF2024-ABS-0170 Metagenomic analyses highlight strain-level, age-independent and immunotherapy-associated microbial signatures in colorectal cancer

<h3>Background</h3> To characterize the microbiome in early-onset colorectal cancer (aged &lt;50, EOCRC) and late-onset CRC (aged ≥50, LOCRC) patients. <h3>Methods</h3> 460 CRC patients (167 EOCRC) were recruited in Guangzhou, and deep shotgun metagenomic sequencing was applied to stool and saliva samples. <i>De novo</i> genome assembly, binning, inStrain, StrainPhlAn and MetaPhlAn analyses were performed. Stool metagenomes from a Fudan study (Yang <i>et al.</i>, 2021) consisting of 50 EOCRC, 50 LOCRC and 100 age-matched controls were also included. <h3>Results</h3> Despite age being a determining factor for the gut microbiome, our analyses found limited associations between the gut microbiome and age in CRC. Well-known CRC-associated taxa (IDDF2024-ABS-0170 Figure 1. CRC-associated taxa and genes enriched in both EOCRC and LOCRC (a)) and genes (IDDF2024-ABS-0170 Figure 1. CRC-associated taxa and genes enriched in both EOCRC and LOCRC (b)) were consistently enriched in both EOCRC and LOCRC. In the strain-level analysis on <i>Fusobacterium nucleatum</i> (<i>Fn</i>), we identified two <i>Fn</i> subspecies <i>animalis</i> Clades (<i>Fna</i> C1 and C2), with <i>Fna C2</i> as the dominant strain (IDDF2024-ABS-0170 Figure 2. Distribution of <i>Fna</i> C1 and <i>Fna</i> C2 in CRC patients (a)). This observation is in line with a <i>Nature</i> study (Zepeda-Rivera <i>et al.</i>, 2024). We confirmed that <i>fadA</i> abundance did not differ in patients carrying either <i>Fna</i> C1 or C2 (IDDF2024-ABS-0170 Figure 2. Distribution of <i>Fna</i> C1 and <i>Fna</i> C2 in CRC patients (b)). More importantly, we found that both the abundances of <i>Fna</i> C1 and C2 were associated with mismatch repair and HER2 expression status (IDDF2024-ABS-0170 Figure 2. Distribution of <i>Fna</i> C1 and <i>Fna</i> C2 in CRC patients (c,d)), suggesting their potential roles in CRC clinical outcomes. We identified two distinct <i>Bacteroides fragilis</i> (<i>Bf</i>) strain clusters with high genome-wide similarity to the reference genomes of strain NCTC 9343 and Q1F2 respectively (IDDF2024-ABS-0170 Figure 3. Two distinct <i>Bf</i> strain clusters in CRC patients (a)). Intriguingly, the abundance and prevalence of <i>Bf</i> toxin gene <i>bft</i> were significantly higher in patients with strain NCTC9343 than Q1F2 (IDDF2024-ABS-0170 Figure 3. Two distinct <i>Bf</i> strain clusters in CRC patients (b)). Only one strain cluster was identified for <i>Escherichia coli</i> (<i>Ec</i>) (IDDF2024-ABS-0170 Figure 4. Correlations between <i>Ec</i> and <i>pks</i> genomic island (a)). A positive correlation between <i>pks</i> genomic island and <i>Ec</i> was observed in the Guangzhou cohort, but no correlation in the Fudan cohort, indicating population heterogeneity. Lastly, we found that the baseline compositions of gut and oral microbiome had remarkable differences between patients with different responses in anti-PD1 treatment (IDDF2024-ABS-0170 Figure 5. Gut and oral microbes associated with anti-PD1 treatment response). <h3>Conclusions</h3> Our results highlight striking similarities in microbial signatures between EOCRC and LOCRC, emphasizing the important roles of microbiome in CRC biology and treatment.