Review papers with raw data transparency

Key quantitative findings (from paper)

• Study: paired tumor and distant-normal colon tissues from 116 Malaysian individuals (pooled MAL1–MAL3 cohorts) Queen et al. cohort & methods
• Biofilms visualized on 68.8% of tumors (77/111) and more frequent on right-sided tumors (91.7% right vs 58.7% left; p=0.0003) Biofilm prevalence
• F. nucleatum detected by 16S in 86.6% of tumors (97/112) and significantly enriched in tumors vs paired normal (p<0.0001) F.n. prevalence & enrichment
• Fna C2 (subspecies animalis clade 2) is the dominant clade enriched in biofilm-positive tumors, with clade-level typing from sequence variants and PCR corroboration Fna C2 enrichment evidence

Critical assessment — strengths

• Multi-modal design (spatial FISH imaging + 16S sequencing + PCR + culture) increases internal validity for localization and taxonomic signals versus sequencing-only studies Multi-modal methods
• Paired tumor–normal sampling reduces between-subject confounding and strengthens tumor-specific enrichment claims.
• Clade-level focus on Fna C2 follows recent independent discovery of a CRC-dominant F. nucleatum clade (Zepeda-Rivera et al. 2024), adding reproducibility across cohorts Zepeda-Rivera 2024

Critical assessment — limitations & blindspots

1. Generalisability: cohort is from a single Malaysian center; ethnicity mix (Malay/Chinese/Indian) is a strength for regional diversity but findings may not generalize globally; global meta-analyses show geographic heterogeneity in Fusobacterium prevalence Global prevalence meta-analysis
2. Functional inference: PICRUSt2 predictions are hypothesis-generating but infer function from 16S — not equivalent to shotgun metagenomes, metatranscriptomes, metabolomics, or direct enzyme assays. The paper correctly frames PICRUSt2 as predictive but experimental validation is required PICRUSt2 caveat
3. Strain-level inference & culture yield: only 13 Fusobacterium isolates recovered from 111 tumors; authors cite long-term storage (5–10 years) as a likely cause for low culture recovery — this prevents whole-genome-resolved linkage of clade-specific genes (e.g., fadA/fap2 variants) to in situ phenotype Culture yield limitation
4. Spatial sampling mismatch: sequencing and FISH were performed on different biopsy regions; paper notes spatial heterogeneity in tumors and that absolute fusobacterial abundance can be localized (dense blooms) and underrepresented by bulk sequencing — a recognized measurement bias.
5. Association vs causation: as with most human microbiome cross-sectional work, enrichment does not prove causality; multiple mechanistic routes are plausible (bloom-after-tumor, facilitation-of-growth, immune-selection), and disentangling them needs longitudinal and interventional data (animal colonization with clinically-derived strains, isogenic mutants, or longitudinal human sampling) Garrett/Brennan review context

Potential biases authors acknowledged & we must consider

• Selection bias (single hospital surgical cohort), storage effects on culture, and limitations of 16S taxonomic resolution addressed by authors (Resphera ambiguous calls used when needed) Author bias statements
• Confounders such as antibiotic exposure (standard IV prophylaxis used perioperatively, but no oral antibiotics) and mechanical bowel prep were common across patients and may affect luminal communities but less likely deeply tissue-resident biofilms.

What the data robustly support

• F. nucleatum is commonly detected in these colorectal tumors and is enriched in tumors relative to paired distant normal tissue in this cohort (statistically robust by paired Wilcoxon tests) Paired comparisons
• Biofilm-positive tumors often harbor polymicrobial consortia that include Fusobacterium, and Fna C2 is the clade most strongly associated with these biofilms in this dataset.

What remains uncertain or needs next-step experiments

1. Are Fna C2 strains causally promoting progression, or are they ecological 'bloomers' that exploit tumor niches? Answer requires strain-resolved genomes from tumor isolates + orthogonal in vivo tests.
2. Do predicted PICRUSt2 pathway enrichments reflect real metabolic shifts? Need shotgun metagenomics, metatranscriptomics, and targeted metabolomics (e.g., 1,2-propanediol metabolites, CMP-pseudaminic acid, glycosylation-related molecules) from spatially co-registered biopsies.
3. What host features (e.g., Gal-GalNAc expression, immune infiltration, MSI/CIMP status) co-vary with Fna C2/biofilms in this cohort? Integration of tumor molecular pathology would sharpen mechanistic links (authors note some associations in literature) Mechanistic literature context

Concrete next experiments (high-value, falsifiable)

1. Obtain fresh tumor biopsies and perform: spatially-resolved shotgun metagenomics + metatranscriptomics + targeted metabolomics co-registered with FISH to validate PICRUSt2 predictions and localize metabolic activity to Fusobacterium-dense regions.
2. Isolate fresh Fusobacterium strains from biofilm-positive tumors, sequence genomes, and compare virulence/adhesin repertoires (fadA, fap2 variants) between tumor-derived Fna C2 strains and oral isolates; then test isogenic mutants in murine CRC models for tumor promotion.
3. Longitudinal sampling (adenoma→carcinoma) to test whether biofilms and Fna C2 precede tumor progression or bloom after tumor microenvironment changes; falsification: showing stable absence of Fna C2 prior to tumor formation would contradict a driver role.

Author reviews (one-click)