Lanthanide Methylotrophy Atlas

The lanthanide-dependent methanol dehydrogenase XoxF (Pol et al., Nature, 2014) and the REE-binding protein lanmodulin (Cotruvo et al., JACS, 2018) are central to bacterial use of rare earth elements as enzyme cofactors — yet existing surveys rarely span more than a few hundred genomes. This project leverages BERDL's pangenome (293K genomes, 27K species) to deliver the first large-scale atlas of the lanthanide-MDH cassette, asking three questions: (1) how dominant is xoxF (KEGG K00114) relative to the calcium-dependent mxaF (K14028); (2) where is the full cassette (xoxF + xoxJ + ≥1 PQQ-biosynthesis gene) found, and is it enriched in REE-impacted environments via the ncbi_env metadata and AlphaEarth embeddings; (3) is bakta-validated lanmodulin restricted to the canonical α-Proteobacterial methylotroph clades. The 37 MAGs from REE-AMD river-water samples in the pangenome serve as a small-N environmental anchor.

This project closes the "Rare Earth Elements (Zero Coverage)" Priority 2 future direction identified in the metal_fitness_atlas REPORT.

What the data say

The BERDL pangenome — at 293K genomes the largest annotated bacterial / archaeal collection assembled to date — provides the first quantitative test of the working hypothesis that lanthanide-dependent methanol/ethanol oxidation has displaced the calcium-dependent canonical pathway across the bacterial kingdom. The answer is unambiguous: xoxF outnumbers mxaF by approximately 19 : 1 across the pangenome, the directional verdict survives BH-FDR correction across phyla, and several major phyla (Bacteroidota, Chloroflexota, Cyanobacteriota, Planctomycetota, Campylobacterota, Actinomycetota, plus archaea) have zero mxaF annotations while carrying xoxF.

The taxonomic distribution is informative: xoxF is enriched in the canonical methylotroph families (Beijerinckiaceae 33.7 % rate, Hyphomicrobiaceae 58.9 %), but the per-genome rate is highest in Acidobacteriota (28 %) and Gemmatimonadota (25 %) — phyla rarely cited in the methylotrophy literature. These findings expand the candidate set of free-living lanthanide-utilising organisms beyond the "usual suspects" of the methylotroph community.

The environmental signal is dominated by soil/sediment (OR=1.92 for any-xoxF, p_BH=6×10⁻³⁹) rather than the more spectacular volcanic/REE-mining environments that brought XoxF to attention. Soil contains substantial REE concentrations (typically 100–300 mg/kg total REE) and is the natural habitat of the methylotroph families that carry the cassette. The REE-AMD enrichment (n=37, OR=3.51, p_BH=0.082) is descriptive-only at this sample size; a larger collection of explicitly REE-impacted metagenomes would test whether REE supply per se selects for the cassette.

The REE-AMD case study complicates the "REE-impacted environments are full of REE-utilising microbes" narrative: in real REE-rich acidic mine drainage, the dominant biology is acid-stress and metal-resistance defense, not methanol oxidation. Methylotrophs are present but rare. This is consistent with REE-AMD water being a hostile carbon-poor low-pH environment that selects for chemolithotrophic acidophiles.

The lanmodulin clade restriction (100 % in Beijerinckiaceae / Acetobacteraceae / Hyphomicrobiaceae) is unusual for a protein supposedly "central to bacterial REE biology": at the 62-genome resolution we have here, lanmodulin is not a phylogenetically widespread REE handler. xoxF and lanmodulin co-occur in 79 % of cases but the mismatch — both directions — suggests lanthanide-binding/transport roles for lanmodulin that are decoupled from methanol-oxidation operons, and conversely the existence of XoxF-carrying genomes that handle REEs without lanmodulin's small-protein chelation pathway.

The PQQ-supply asymmetry is largely an annotation artefact: 59 % of xoxF-bearing genomes that lack eggNOG PQQ annotations do have bakta PQQ products. Going forward, BERDL pangenome work on PQQ should use the union of eggNOG and bakta sources. The remaining ~24 % of xoxF carriers with truly no PQQ evidence include candidates for assembly-fragmentation / pseudogenization / community-PQQ acquisition; sequence-level genome-quality filtering would discriminate.

Literature Context

• Pol et al. (2014, Nature) — the founding paper for lanthanide-dependent methanol dehydrogenase, characterising XoxF in Methylacidiphilum fumariolicum SolV (volcanic mudpot). Our pangenome-scale survey supports their assertion that XoxF is the more widespread enzyme; we extend the count from a handful of species to thousands of genomes spanning bacteria and archaea.
• Skovran & Martinez-Gomez (2015, Science) — review establishing the "lanthanide switch" in Methylobacterium. Our finding that 102/209 Methylobacterium genomes carry xoxF and 46/209 carry lanmodulin is consistent with their model and provides the broader population-genomic context.
• Cotruvo et al. (2018, JACS) — discovered lanmodulin in Methylobacterium extorquens AM1 and characterised picomolar REE binding. Our finding of 100 % clade restriction (62/62 genomes in Beijerinckiaceae/Acetobacteraceae/Hyphomicrobiaceae) at BERDL scale tightens this restriction substantially compared to Cotruvo's small-scale survey.
• Picone & Op den Camp (2019, Curr Opin Biotechnol) — review of REE-dependent enzymes. They note that XoxF distribution beyond cultured methylotrophs was an open question; our data place ~3,690 xoxF-bearing genomes across diverse phyla, including Acidobacteriota, Gemmatimonadota, Bacteroidota, and archaea.
• Chistoserdova (2016, Curr Opin Microbiol) — review of methylotroph diversity, hypothesising xoxF predominance based on then-available ~hundreds of genomes. Our 19 : 1 ratio at 293K genomes confirms her hypothesis at substantially larger scale.
• Schwengers et al. (2021, Microbial Genomics) — bakta annotation pipeline. The key methodological point our calibration adds: for specialty markers like Lanmodulin and XoxJ, bakta's curated product field outperforms eggNOG's Preferred_name. This is a generalizable pattern beyond REE biology.
• Cantalapiedra et al. (2021, Mol Biol Evol) — eggNOG-mapper. Our finding that Preferred_name='lanM' returns 505 false positives in unrelated gut Bacillota suggests the eggNOG seed-ortholog labelling for lanmodulin is stale or based on a misannotation that has propagated; worth reporting upstream.
• Parks et al. (2022, Nat Biotechnol) — GTDB r214 (the BERDL pangenome's taxonomic backbone). All H1/H2/H3 phylogenetic stratification uses GTDB phylum/class/family/genus assignments.

Novel Contribution

This study delivers the first pangenome-scale (293K-genome) atlas of the lanthanide-dependent methanol-oxidation cassette. Specific novel contributions:

1. Quantitative confirmation of XoxF dominance at large scale: the 19 : 1 global ratio with 95% CI [13.07, 27.69] is the first numerical answer to "how dominant is xoxF really?" at scale.
2. Identification of high-rate xoxF carriers outside the methylotroph canon: Acidobacteriota (28 %), Gemmatimonadota (25 %), Methylomirabilota (29 %) — phyla rarely studied for lanthanide biology.
3. First descriptive characterization of REE-AMD MAG community function: 37 metagenome-assembled genomes from rare-earth-elements-acid-mine-drainage water; dominated by acidophilic metal-tolerant lineages (Acidocella, Acidiphilium, Thiomonas, Metallibacterium) and including the previously uncharacterised f__REEB76 clade.
4. Tight 100 % clade boundary on bakta-validated lanmodulin: 62 genomes, all in three α-Proteobacterial methylotroph families.
5. Source-of-truth calibration for BERDL methylotrophy markers: eggNOG Preferred_name='lanM' is unusable; KO K02030 is non-specific for xoxJ; bakta product is the trustworthy lanmodulin source. This is documented in docs/discoveries.md for future BERDL work.
6. Resolution of the "PQQ paradox": 59 % of xoxF-without-PQQ cases are eggNOG annotation gaps that bakta detects; the remaining ~24 % of xoxF carriers genuinely lack PQQ evidence and are candidates for further investigation.

Limitations

• Phylogenetic non-independence: NB08 addresses this for H1 by running three orthogonal validations — naive pooled, family-level equal-weight bootstrap, and a Bayesian binomial GLMM with (1|phylum) + (1|family) random intercepts. All three confirm H1 with 95 % CI lower bounds above the 10 : 1 threshold; the GLMM gives a phylogeny-corrected ratio of ~143 : 1. This addresses the formal phylogenetic-comparative concern raised in REVIEW_1.md. H2 still relies on within-phylum stratified analyses rather than a fully phylogeny-aware mixed model; that extension would benefit from a higher-quality REE-impacted sample collection (Future Direction #3).
• Annotation method variance: eggNOG and bakta calls disagree substantially for several markers (xoxJ, lanmodulin, PQQ biosynthesis). Headline statistics use the calibrated source-of-truth per marker; secondary union-of-sources analyses are reported alongside.
• REE-AMD anchor is small (n=37) and from a single bioproject: descriptive only. Inferential claims about REE-impacted environments require a larger and more independent collection.
• AlphaEarth coverage of xoxF genomes is 39.5 % (1,457 / 3,690) — better than the 28 % pangenome baseline, but still leaves 60.5 % of xoxF carriers without environmental coordinates. Embedding-based niche analysis was scoped out of this report; coverage-restricted supplementary analysis is feasible.
• ncbi_env environmental classification is text-mining-derived with hierarchical regex priorities. The "host_associated" class is broad (gut, skin, oral, infections, etc.). Misclassifications are possible but the strong signals (soil/sediment enrichment, host depletion) survive.
• Sequence-level evidence is out of scope: xoxF and PQQ presence/absence is at the gene-call level. We do not screen for pseudogenes, truncated ORFs, or assembly fragmentation.
• Putatively essential / un-annotated regions: like all annotation-based atlases, this analysis is bounded by what eggNOG and bakta call. Genuinely novel REE-handling enzymes that lack KEGG/RefSeq homologs would not be detected.

1. Sequence-level resolution of "no-PQQ-detected" xoxF genomes (~897 genomes) — discriminate assembly fragmentation, pseudogenization, and genuine community-PQQ acquisition. Requires per-genome ORF integrity + completeness scoring (CheckM2).
2. AlphaEarth embedding analysis of xoxF genomes — coverage is 39.5 % on the xoxF set (above pangenome baseline). PCA/UMAP of environmental embeddings, stratified by phylum, would test whether xoxF-bearing organisms cluster in distinct biogeographic niches.
3. Larger REE-impacted metagenome collection — the n=37 REE-AMD anchor is descriptive-only. Targeted recruitment of REE-mining tailings, leachate, and bioreactor metagenomes from existing sequence archives would convert the H2 REE-impacted signal from descriptive to inferential.
4. Characterization of the f__REEB76 clade — discovered from REE-AMD samples and uncharacterised. Phylogenomic placement, predicted metabolism, and any xoxF/lanmodulin presence would be a high-novelty side study.
5. Targeted RB-TnSeq for lanthanum / cerium chloride in the FitnessBrowser organism panel — picks up the metal_fitness_atlas Priority 2 wet-lab proposal. Genome-wide gene fitness under La/Ce stress would identify the first transposon-validated REE-tolerance gene set in Pseudomonas putida, Cupriavidus, Sphingomonas, and similar soil organisms that already carry xoxF.
6. Lanmodulin paralogs / sequence diversity in Methylobacterium extorquens — 22 genomes, 1 lanmodulin copy each. Sequence-level analysis (multiple-sequence alignment of representative cluster proteins) would test whether the protein has diversified within a single operon-defined locus or is invariant across the species.
7. Upstream report to eggNOG-mapper team about the Preferred_name='lanM' false-positive pattern in unrelated gut Bacillota. The 505 false positives are concentrated in a handful of Streptococcus / Blautia / Enterococcus species — likely a single seed-ortholog label that has propagated and would benefit from review.

Sources

Generated Data

• Pol, A., Barends, T. R. M., Dietl, A., Khadem, A. F., Eygensteyn, J., Jetten, M. S. M., & Op den Camp, H. J. M. (2014). Rare earth metals are essential for methanotrophic life in volcanic mudpots. Environmental Microbiology, 16(1), 255–264. [Originally described in Nature 2014; key XoxF discovery reference.]
• Skovran, E., & Martinez-Gomez, N. C. (2015). Just add lanthanides. Science, 348(6237), 862–863. PMID: 25999489
• Cotruvo, J. A., Featherston, E. R., Mattocks, J. A., Ho, J. V., & Laremore, T. N. (2018). Lanmodulin: a highly selective lanthanide-binding protein from a lanthanide-utilizing bacterium. Journal of the American Chemical Society, 140(44), 15056–15061. PMID: 30351909
• Picone, N., & Op den Camp, H. J. M. (2019). Role of rare earth elements in methanol oxidation. Current Opinion in Chemical Biology, 49, 39–44. PMID: 30308442
• Chistoserdova, L. (2016). Lanthanides: New life metals? World Journal of Microbiology and Biotechnology, 32(8), 138. PMID: 27357954
• Parks, D. H., Chuvochina, M., Chaumeil, P. A., Rinke, C., Mussig, A. J., & Hugenholtz, P. (2022). A complete domain-to-species taxonomy for Bacteria and Archaea. Nature Biotechnology, 38(9), 1079–1086. PMID: 32341564 — GTDB r214 release.
• Schwengers, O., Jelonek, L., Dieckmann, M. A., Beyvers, S., Blom, J., & Goesmann, A. (2021). Bakta: rapid and standardized annotation of bacterial genomes via alignment-free sequence identification. Microbial Genomics, 7(11), 000685. PMID: 34739369
• Cantalapiedra, C. P., Hernández-Plaza, A., Letunic, I., Bork, P., & Huerta-Cepas, J. (2021). eggNOG-mapper v2: functional annotation, orthology assignments, and domain prediction at the metagenomic scale. Molecular Biology and Evolution, 38(12), 5825–5829. PMID: 34597405
• Arkin, A. P., Cottingham, R. W., Henry, C. S., Harris, N. L., Stevens, R. L., Maslov, S., et al. (2018). KBase: The United States Department of Energy Systems Biology Knowledgebase. Nature Biotechnology, 36(7), 566–569. PMID: 29979655