Subsurface Bacillota_B Specialization

The companion clay_confined_subsurface project (PR #227, merged) found that within Bacillota_B (after phylum control), 5/5 deep-clay isolates carry sulfate reduction vs 4/19 soil-baseline (p_BH=0.04) — the only result that survived phylum-stratified testing. Bacillota_B is exactly Bagnoud 2016's recurrent indigenous Mont Terri lineage. The natural follow-up is gene-cluster–level: what else distinguishes these subsurface specialists from their soil congeners? We extend the clay project's framework from a curated 18-marker dictionary to the full BERDL pangenome accessory-genome of Bacillota_B, using eggNOG OGs as the cross-species orthology surrogate. We also use this as the venue for correcting the clay project's H3 IR-side analysis, which used three KO numbers (K07811, K17324, K17323) that turn out to be TMAO reductase, glycerol ABC, and glycerol permease — not iron-reduction markers. The correction (Phase 1) replaces those with PFAM PF14537 multi-heme cytochrome detection + heme-binding motif counting and re-runs the clay H3 test.

Cohort

data/cohort_assignments.tsv:

IR PFAM availability (NB01 gate)

data/ir_pfam_availability.tsv — within-Bacillota_B PFAM count:

PF14537 confirmed silently absent (the documented plant_microbiome_ecotypes pitfall). Multi-heme cytochrome PFAMs are sparse within Bacillota_B (only ~6 clusters total across all four PFAMs); CXXCH motif counting on gene_cluster.faa_sequence carried most of the IR signal in NB06 — which is why the corrected IR rates (40–56%) are far higher than what PFAM-only detection would yield.

H1 OG enrichment (NB03)

data/og_enrichment.tsv: 547 OGs anchor-enriched (q<0.05, fold≥3, n_anchor≥3); 27 OGs anchor-depleted. Top hits at OR=∞ (7–10 anchor positive vs 0 baseline). Functional annotations in data/enriched_ogs_annotated.tsv (NB04).

H2 compactness (NB05)

See Finding 2 table. All four size/OG-count metrics show anchor SIGNIFICANTLY LARGER (p ≤ 0.025), all with large effect sizes (Cohen's d ≥ +1.30). H2 H1 prediction was anchor SMALLER → rejected, opposite direction.

H3 corrected IR (NB06)

See Finding 3 table. With corrected multi-heme cytochrome detection, no significant cohort difference (all pairwise Fisher p ≥ 0.46).

Literature Context

• H1 functional categories align with Beller 2012 Pelosinus HCF1, which carries 2 [NiFe] + 4 [FeFe] hydrogenases plus dissimilatory N-oxide reductase + Cr/Fe reductase + methylmalonyl-CoA pathway in a single subsurface aquifer Firmicute. Our anchor-enriched-OG categories (anaerobic respiration / regulators / sporulation) match Beller's gene-content snapshot at the population scale.
• H2 expansion finding (anchor LARGER than baseline) directly supports Becraft 2021 Ca. Desulforudis audaxviator (PMC8443664), who explicitly note the genome's self-sufficiency: "the assembled genome suggested complete self-sufficiency for this subsurface bacterium ... contains all genes necessary for carbon and nitrogen fixation and encodes all necessary amino acid biosynthesis pathways." Our finding extends this from a single lineage to a within-phylum comparison: Bacillota_B in deep-clay habitats encode ~25% more orthogroups than their soil congeners, consistent with a self-sufficiency adaptation.
• H2 contradicts Tian 2020's "small is mighty" (PMC7137472) at first glance — but Tian's finding was explicit to Patescibacteria/CPR (an episymbiotic superphylum that depends on hosts for many functions) and Tian explicitly notes that non-CPR phyla retain larger genomes. Our result is consistent with Tian's framing: streamlining is a CPR-specific adaptation; cultivable subsurface Firmicutes show the opposite pattern.
• H1 sporulation-revival hits replicate Vandieken 2017's (PMID 28646634) characterization of Baltic Sea subsurface Desulfosporosinus species as spore-forming with broad respiratory versatility. Our anchor cohort includes 3 Desulfosporosinus genomes that contributed to the sporulation-OG signal.
• H3 correction is a real bug fix in PR #231. The clay_confined_subsurface project's IR-side analysis used K07811 (TMAO reductase, threshold 1336.87), K17324 (glycerol ABC ATP-binding, threshold 363.93), K17323 (glycerol ABC permease, threshold 318.27) — none are iron-reduction genes, and KEGG has no canonical KO for Geobacter omcS / Shewanella mtr lineages. The clay project's "Mitzscherling rock-attached signature in shallow clay" framing was an artefact of mismatched gene IDs combined with the standard eggNOG-mapper threshold (which doesn't filter on KO biological identity, just sequence similarity). After correction, the IR comparison loses statistical significance entirely.

Novel Contribution

1. First quantitative gene-cluster–level characterization of subsurface Bacillota_B specialization. Goes beyond the curated marker dictionary used in clay_confined_subsurface to the full BERDL pangenome accessory genome. 547 enriched OGs with functional annotation provide a concrete target list for future biochemistry / fitness-screen / cultivation effort.
2. Direct refutation of "subsurface = streamlining" within cultivable Firmicutes. Existing literature emphasizes Patescibacteria streamlining (Tian 2020) or Ca. Desulforudis self-sufficiency (Becraft 2021) as case studies. Our population-scale within-Bacillota_B comparison shows that gene-content expansion is the genuine signal at this phylum's scale, with statistical support.
3. Methodological correction to a merged BERIL project. The clay_confined_subsurface H3 IR-side bug (mismatched KOs) is documented and fixed; the corrected IR rates are saved at data/clay_h3_ir_corrected.tsv and serve as the basis for a separate correction commit on the clay project's branch.
4. Methodological pattern: triple-signal multi-heme cytochrome detection. Combining curated PFAMs (PF02085, PF22678) with sequence-based CXXCH heme-motif counting on gene_cluster.faa_sequence is a reusable BERDL pattern for detecting multi-heme cytochromes when KEGG KO assignment is unreliable. Applicable across the pangenome whenever iron-reduction or extracellular-electron-transfer questions arise.

Limitations

• Cohort size: 10 anchor vs 62 baseline. With Fisher's exact this is fine for detecting large effects (the 547 enriched OGs are robust); marginal effects with anchor counts of 3–5 should be treated descriptively rather than inferentially. Genus-level phylogenetic confounding is partly mitigated by the cohort spanning four orders (Desulfotomaculales, Desulfitobacteriales, Ammonifexales, Thermacetogeniales) but not eliminated.
• Anchor cohort is borehole/porewater-dominated by construction (per the clay-project lesson on cultivation bias toward porewater isolates). Rock-attached Bacillota_B may differ in gene content; this comparison cannot test that.
• OG-level orthology is hierarchical: we used Firmicutes-level OGs (eggNOG tax 1239) where present, falling back to Bacteria/root. Some enriched OGs are at root level, which is coarse. Ideally a Bacillota_B-specific OG level would be used, but eggNOG OG IDs do not have a Bacillota_B-specific tier yet (Bacillota_B is GTDB-defined, eggNOG uses NCBI taxonomy where Bacillota_B genomes are spread across Firmicutes / Negativicutes / Tissierellia at the legacy class level).
• Keyword-based functional categorization undercounts (the "other_or_unannotated" 462-OG bucket contains substantial real anaerobic-respiration / electron-transfer signal that the keyword scanner missed). Manual reclassification or an LLM-based functional-category extractor would refine the count.
• Phase 1 IR correction is robust for the multi-heme cytochrome signal but doesn't address whether the original clay project's H3 narrative ("BERDL clay cohort matches Bagnoud porewater pattern") is fully recovered or partially undercut. The SR-side is robust; the IR-side narrative loses force. A revised clay-project REPORT.md amendment will note this.

1. Apply the correction commit to the clay project's branch. Outputs at data/clay_h3_ir_corrected.tsv + figures/clay_h3_ir_corrected.png are ready. Cherry-pick or correction PR will add a notebooks/07_h3_iron_reduction_correction.ipynb + REPORT.md amendment to clay_confined_subsurface clarifying that the IR-side H3 finding is artefactual (SR-side stands).
2. Refine the OG functional-category labels with an LLM-based scanner. The 462-OG "other / unannotated" bucket contains substantial real anaerobic-niche signal the keyword scanner missed (COG1977 molybdopterin, DsrEFH, 2-oxoglutarate ferredoxin oxidoreductase). An LLM-based pass over the eggNOG Description + bakta product text would surface these for accurate counting per category.
3. Genus-level decomposition of the H1 signal. The 10-genome anchor is genus-clumped (3 BRH-c8a + 2 BRH-c4a + 2 Desulfosporosinus + Desulforudis + Ch130 + 1 other). Per-genus analysis would distinguish which enriched OGs are genus-specific (lineage-marker noise) from which are recurrent across multiple subsurface-specialist genera (the genuine deep-clay signature). With n=10 anchor this is at the limit of statistical resolution; expansion via BacDive linkage to additional clay-isolated Bacillota_B (or future BERDL pangenome ingests) would improve power.
4. H2 expansion mechanism. Anchor genomes are ~1 Mbp larger than baseline. Where does that extra DNA go? Mobile-element-borne accessory operons (cf. Goff 2024 ORFRC mobilome paper)? Sporulation regulons? Anaerobic-respiration accessories? A breakdown of the "extra" gene fraction by COG / KEGG-pathway membership would localize the expansion functionally.
5. Apply the gene-cluster–level enrichment framework to other within-phylum subsurface comparisons. The same NB01–04 pipeline can be retargeted to within-Bacteroidota / within-Pseudomonadota / within-Acidobacteriota deep-clay vs soil comparisons. Each will test whether the H1 functional-category signal generalizes or is Bacillota_B-specific.

Sources

Generated Data

Full bibliography in references.md. Primary citations:

• Bagnoud A et al. (2016). Reconstructing a hydrogen-driven microbial metabolic network in Opalinus Clay rock. Nat Commun 7:12770. PMID: 27739431.
• Beaver RC, Neufeld JD (2024). Microbial ecology of the deep terrestrial subsurface. ISME J 18:wrae091. PMID: 38780093.
• Becraft ED et al. (2021). Evolutionary stasis of a deep subsurface microbial lineage [Ca. Desulforudis audaxviator]. ISME J 15:2830-2842. PMID: 33824425.
• Beller HR et al. (2012). Genomic and physiological characterization of Pelosinus sp. HCF1. AEM 78:8791-8800. PMID: 23064329.
• Hilpmann S et al. (2023). Presence of uranium(V) during uranium(VI) reduction by Desulfosporosinus hippei. Sci Total Environ 875:162593. PMID: 36889400.
• Methé BA et al. (2003). Genome of Geobacter sulfurreducens. Science 302:1967-1969. PMID: 14671304.
• Mitzscherling J et al. (2023). Clay-associated microbial communities in Opalinus Clay rock formation. MicrobiologyOpen 12:e1370. PMID: 37642485.
• Tian R et al. (2020). Small and mighty: adaptation of superphylum Patescibacteria. Microbiome 8:51. PMID: 32252814.
• Vandieken V et al. (2017). New Desulfosporosinus species from Baltic Sea subsurface sediments. IJSEM 67:1887-1893. PMID: 28646634.