Report: SNIPE Defense System in the BERDL Pangenome

Status

Complete — all notebooks executed 2026-03-01.

Summary

Bacteria face a fundamental trade-off in phage defense: the most effective resistance mechanism against phage lambda — losing the ManYZ mannose transporter pore — comes at a steep metabolic cost (loss of mannose uptake, impaired glucose transport). SNIPE resolves this dilemma. By constitutively sitting at the inner membrane and cleaving phage DNA as it passes through ManYZ, SNIPE-bearing strains get phage resistance without sacrificing the transporter. Quantitative fitness data from the Deutschbauer/Price Fitness Browser (168 experiments in E. coli K-12) confirms the trade-off: ManXYZ knockouts show fitness scores of -2.7 to -4.1 on mannose and glucosamine, but are dispensable for fructose — contradicting UniProt’s “fructose transporter” annotation and confirming mannose/glucosamine specificity. Coevolution data show that man knockouts sweep to >95% frequency under phage pressure but collapse when phages evolve alternative entry routes — SNIPE offers a more durable, cost-free alternative. Notably, the Fitness Browser contains one archaeal SNIPE protein (Methanococcus maripaludis JJ, locus MMJJ_RS01635) with both the DUF4041 and Mug113 nuclease domains and 129 experiments of fitness data — the first SNIPE homologue with genome-wide knockout phenotypes.

We surveyed the 293K-genome BERDL pangenome for SNIPE homologues using Pfam domain annotations and found them in 1,696 species across 33 phyla — far more than the >500 reported in the original paper (Saxton et al. 2026, Nature). 86.7% of SNIPE genes are accessory or singleton, consistent with mobile defense island carriage. SNIPE-bearing species occupy statistically distinct environmental niches (22/64 AlphaEarth dimensions, p < 0.05 Bonferroni-corrected). SNIPE was also detected in Klebsiella, a key phage therapy target. These results strongly support H1: SNIPE homologues are widespread, predominantly mobile, and associated with specific ecological contexts.

Corrected Pfam Domain Assignments

The original paper (Saxton et al. 2026) refers to “DUF4041” and “GIY-YIG” by name without citing Pfam accessions. Cross-referencing with InterPro and Pfam databases reveals the correct accessions and an important distinction in the nuclease domain:

Evidence: The E. coli SNIPE protein A0A0A1A5Z2 (558 aa) is annotated by InterPro with PF13250 (positions 232–333) and PF13455 (positions 443–520). InterPro has already renamed IPR025280 from “Domain of unknown function DUF4041” to “SNIPE associated domain” based on this paper’s functional characterization.

Practical implication: The SNIPE nuclease belongs to the GIY-YIG clan (CL0418) but is a distinct Pfam family (PF13455/Mug113) from the canonical GIY-YIG (PF01541). Searching for PF01541 co-occurrence with DUF4041 yields zero hits — not because of annotation artifacts, but because these are genuinely different Pfam families. The correct search strategy uses PF13250 (or the domain name “DUF4041” / “T5orf172”) for the binding domain and PF13455 for the nuclease.

Key Findings

1. SNIPE resolves the phage resistance vs. metabolic cost trade-off

Saxton et al. (2026) showed that SNIPE constitutively localizes to the inner membrane and cleaves phage DNA as it passes through the ManYZ mannose transporter pore. Published knockout and coevolution studies (not from BERDL) on ManYZ reveal why this matters — losing ManYZ is effective phage defense, but metabolically expensive:

Data source: Phage lambda phenotypes from Burmeister et al. 2021 (PMID: 34032565), using the Keio collection (E. coli BW25113). Fitness scores below from the Deutschbauer/Price Fitness Browser (feba.db, queried locally).

Quantitative fitness scores from the Fitness Browser

Direct RB-TnSeq fitness measurements for ManXYZ in E. coli K-12 (Keio collection, 168 experiments; Price et al. 2018) confirm severe, substrate-specific fitness defects:

The strongest defects cluster on two carbon sources:

The three genes are tightly co-fitness correlated (manX↔manZ r=0.851, manX↔manY r=0.705, manY↔manZ r=0.725), confirming they function as a single operon.

ManYZ is NOT a fructose transporter

UniProt annotates ManX as a “fructose-specific” PTS component. Fitness Browser data directly contradicts this:

ManXYZ mutants grow normally on fructose (fitness ≈ 0) but are crippled on mannose and glucosamine. FruA (the actual fructose PTS) shows the opposite pattern — essential for fructose, dispensable for mannose. This clean substrate separation confirms ManYZ is a mannose/glucosamine-specific transporter, consistent with its role as the phage lambda entry receptor.

Data source: Fitness Browser SQLite database (feba.db, downloaded from Figshare), queried locally. All fitness scores are log2 ratios from RB-TnSeq (Price et al. 2018). Carbon source conditions identified by condition_1 and expGroup = 'carbon source'.

In phage-bacteria coevolution experiments, man mutants sweep to >95% frequency under phage pressure but decline sharply when phages evolve alternative injection mechanisms (Burmeister et al. 2021). This creates a well-characterized evolutionary dilemma: lose the transporter and starve, or keep it and be vulnerable to phage.

SNIPE resolves this trade-off: rather than losing ManYZ entirely, SNIPE-bearing strains retain full transporter function while cleaving phage DNA during injection. SNIPE provides the phage resistance benefit of man knockouts without the metabolic cost.

Klebsiella co-occurrence: DUF4041 + ManYZ in the same genome

The PhageFoundry Klebsiella genome browser — the only one of four species where we detected DUF4041 (Finding 6) — also has abundant mannose PTS family annotations:

The 3 DUF4041 proteins (JJW41_17025, KFB13_RS08150, KFB31_RS01490) are all 558 aa — the same length as the characterized E. coli SNIPE protein (A0A0A1A5Z2), strongly suggesting full-length two-domain (PF13250 + PF13455) SNIPE proteins. Klebsiella is the only PhageFoundry species with both SNIPE and the PF00358 ManX-family PTS domain by exact Pfam accession. The co-occurrence of SNIPE and its target mannose transporter in the same organism is consistent with the evolutionary rationale — SNIPE is present where ManYZ pores provide a phage entry route worth defending.

Data source: BERDL PhageFoundry browser_eggnog_description and browser_annotation tables, queried for ManYZ-related terms (“mannose”, “PTS”, “phosphotransferase”) and Pfam domains (PF00358/PTS_EIIC, PF02378/PTS_EIID). See NB04 §A5 for the full query code.

Fitness Browser: ManXYZ data confirmed, SNIPE found in Methanococcus

Querying the Deutschbauer/Price Fitness Browser database (48 organisms, 228K genes, 27M fitness scores) revealed:

• ManXYZ: Present only in E. coli K-12 (Keio collection) among the 48 Fitness Browser organisms. Full fitness data for 168 conditions is presented above. No Klebsiella strain in the Fitness Browser carries ManXYZ by gene name.
• SNIPE (DUF4041/PF13250): One organism carries it — Methanococcus maripaludis JJ (locus MMJJ_RS01635), annotated with both PF13250 (DUF4041) and PF13455 (Mug113/GIY-YIG nuclease), plus PF10544 (DUF2525). This is a full two-domain SNIPE protein with 129 experiments of fitness data.

The M. maripaludis SNIPE gene shows mild fitness defects under specific conditions (minimum fitness = -1.16 on formate/acetate), but is dispensable under most conditions. This is the first organism with both a confirmed SNIPE protein and genome-wide fitness data — though as an archaeon, the phage biology context differs significantly from the Enterobacterales phage lambda system.

Among the 48 Fitness Browser organisms, 7 genes total carry PF13455 (the SNIPE nuclease domain), spread across 6 organisms (Azospirillum brasilense, Bacteroides thetaiotaomicron, Cupriavidus, Desulfovibrio vulgaris, Paraburkholderia kururiensis, and M. maripaludis). Only M. maripaludis has the full two-domain SNIPE architecture (PF13250 + PF13455).

Data source: Fitness Browser SQLite database (feba.db, Figshare), GeneDomain and GeneFitness tables. Also available as kescience_fitnessbrowser in BERDL (API was experiencing 503 outage during analysis).

PhageFoundry phage-host interaction matrix (Gaborieau et al. 2024)

The phagefoundry_strain_modelling database — previously unexplored — contains the complete phage-host interaction dataset from Gaborieau et al. 2024, Nature Microbiology:

• 188 E. coli strains (ECOR collection, NILS collection, clinical/environmental isolates) × 96 phages (41 Podoviridae, 32 Myoviridae, 23 Siphoviridae) = 17,672 binary infection outcomes
• ML model: AUC = 0.883, Accuracy = 84.3%, trained on 1,582 gene cluster presence/absence features with SHAP importance scores
• Overall infection rate: 22% (3,929 positive / 13,743 negative)

Phage lambda finding: The Lambdavirus (411_P1, genus Lambdavirus) infects only 1 of 188 strains (NILS06) — a 0.5% infection rate, the lowest of any phage in the dataset. This is consistent with widespread ManYZ variation/loss conferring near-universal lambda resistance across diverse E. coli strains. By comparison, Myoviridae phages achieve 43.4% infection rate, and even Siphoviridae reach 9.7%.

The paper’s key conclusion — that “adsorption factors” (surface receptors) are the primary predictors of phage-host interaction, not antiphage defense systems — directly supports SNIPE’s niche: SNIPE targets the adsorption/injection step by cleaving DNA during ManYZ transit, operating at exactly the interface where phage success is determined.

Data source: BERDL phagefoundry_strain_modelling database — strainmodelling_interaction, strainmodelling_organism, strainmodelling_organism_metadata, and strainmodelling_experiment_metric tables. This database was previously overlooked (documented as “timed out during discovery”); it contains 18 tables with experimental phage-host phenotype data.

No Klebsiella fitness data for SNIPE or ManYZ

PhageFoundry has ManYZ annotations in Klebsiella (28 ManZ proteins, 4,619 PTS_EIIC Pfam hits — see above), but PhageFoundry is GenomeDepot format (genome annotations only, no fitness/TnSeq tables). Generating Klebsiella-specific SNIPE or ManYZ fitness data would require new mutant libraries in natural SNIPE-carrying strains such as K. pneumoniae NCTC9140 or HS11286. A plan for curating external K. pneumoniae Tn-Seq datasets has been developed to address this gap.

2. SNIPE is widespread (1,696 species, 33 phyla)

DUF4041 (PF13250, InterPro IPR025280), the diagnostic SNIPE domain, was detected in 4,572 gene clusters across 1,696 species spanning 33 bacterial and archaeal phyla. This substantially expands the >500 homologues reported in the original paper.

Data source: BERDL kbase_ke_pangenome.eggnog_mapper_annotations table, filtering Pfam columns for “DUF4041”, joined to gene_cluster and genome tables for taxonomy. See notebook 01 and notebook 02.

Top phyla by species count:

The patchy distribution across many phyla and families is consistent with horizontal gene transfer of defense islands.

3. SNIPE genes are predominantly accessory (86.7%)

Pangenome status of DUF4041-containing gene clusters:

• Core: 13.3% (604 clusters)
• Accessory (non-core, non-singleton): 30.7% (~1,402 clusters)
• Singleton: 56.1% (2,566 clusters)

The combined 86.7% accessory+singleton fraction strongly supports the defense island mobility hypothesis. Only 13.3% of SNIPE clusters are core genes within their species pangenome, indicating that SNIPE is typically gained or lost rather than vertically inherited.

Data source: BERDL kbase_ke_pangenome.gene_cluster table, pangenome_class column for the 4,572 DUF4041-containing clusters. See notebook 01.

4. The SNIPE nuclease domain is PF13455 (Mug113), not PF01541 (GIY-YIG)

Zero gene clusters contained both DUF4041 and canonical GIY-YIG (PF01541) in their eggNOG Pfam annotations. InterPro/UniProt analysis of the E. coli SNIPE protein (A0A0A1A5Z2) reveals why: the SNIPE nuclease domain is actually PF13455 (Mug113, “Meiotically up-regulated gene 113”), not PF01541. Both PF13455 and PF01541 belong to the GIY-YIG clan (CL0418) — the paper describes the SNIPE nuclease as “GIY-YIG” at the superfamily level, but at the Pfam family level they are distinct entries.

The E. coli SNIPE protein domain architecture (558 aa):

This means the 74,686 PF01541 (canonical GIY-YIG) clusters we found are not SNIPE nucleases — they represent restriction enzymes, DNA repair enzymes, and other GIY-YIG family members. The lack of DUF4041 + PF01541 co-occurrence is genuine, not an annotation artifact. Among the 4,572 DUF4041 clusters, 54 carry the description “Meiotically up-regulated gene 113”, consistent with full-length SNIPE proteins annotated with both domains.

Data source: Co-occurrence analysis from BERDL eggnog_mapper_annotations Pfam columns (notebook 01). Domain architecture of E. coli SNIPE protein verified via InterPro and UniProt REST APIs.

5. SNIPE-bearing species occupy distinct environmental niches

Comparison of AlphaEarth 64-dimensional environmental embeddings between SNIPE-bearing (n=1,069) and non-SNIPE species (n=13,977):

• 22 of 64 dimensions are significantly different (Bonferroni-corrected p < 0.05)
• Largest effect: dimension A19 (Cohen’s d = 0.26, p = 5.0e-14)
• Effect sizes are small-to-medium (|d| = 0.11–0.26), suggesting a consistent but modest environmental shift rather than strict habitat segregation
• PCA shows SNIPE species are broadly distributed but with a detectable environmental bias

AlphaEarth dimensions are derived from satellite remote-sensing imagery at sample collection lat/lon coordinates (Goormaghtigh et al. 2025). The dimensions capture environmental signals such as vegetation indices, temperature, precipitation, and land cover. The biological interpretation of individual dimensions (e.g., A19) requires reference to the AlphaEarth documentation; this analysis identifies the statistical signal but does not attempt to map specific dimensions to named environmental variables.

Caveat: AlphaEarth coverage is 28.4% of genomes, biased toward environmental isolates with lat/lon data.

Data source: BERDL kbase_ke_pangenome.alphaearth_embeddings table (64 dimensions, 83K genomes), joined to SNIPE species labels from Finding 2. See notebook 03.

6. SNIPE detected in phage therapy target (Klebsiella)

PhageFoundry genome browser search across 4 species:

The detection of DUF4041 in Klebsiella is clinically relevant — SNIPE could affect phage therapy efficacy in this pathogen.

Data source: BERDL PhageFoundry genome browser databases (browser_eggnog_description table), searching for defense-related terms across 4 species. See notebook 04.

7. Functional annotations are consistent with SNIPE

Top functional descriptions for DUF4041-containing clusters:

COG category distribution: D (cell cycle, 1,533), T (signal transduction, 1,118), J (translation, 1,066), M (cell wall/membrane, 834). The membrane-associated COG M category aligns with SNIPE’s inner membrane localization.

False-positive estimate: The top two descriptions — T5orf172 (2,109) and DUF4041 (1,566) — account for 3,675/4,572 clusters (80.4%) and represent bona fide SNIPE-family proteins. The remaining ~20% (897 clusters) have primary annotations like Histidine kinase (319) and seryl-tRNA aminoacylation (238), suggesting DUF4041 was detected as a secondary/minor domain in multi-domain proteins whose primary function is unrelated to phage defense. These likely represent annotation noise rather than functional SNIPE proteins. A stricter filter (Description containing “T5orf172” or “DUF4041”) would retain ~80% of clusters as a high-confidence SNIPE set; the 86.7% accessory rate and taxonomic distribution patterns are not substantially affected by this subset (the non-SNIPE annotations are distributed across the same phyla).

Data source: BERDL kbase_ke_pangenome.eggnog_mapper_annotations table, description and COG_category columns for the 4,572 DUF4041-containing clusters. See notebook 01.

Hypothesis Assessment

• H0 (null): SNIPE domain genes are rare or absent in the BERDL pangenome and show no taxonomic or environmental enrichment — i.e., their distribution is random with respect to phylum, habitat, and pangenome status.
• H1 (alternative): SNIPE homologues are detectable across multiple bacterial phyla, are predominantly accessory (not core) genes within species pangenomes (consistent with defense island mobility), and are enriched in environmental/non-clinical isolates where phage pressure is high.

H1 is supported on all four axes:

1. Evolutionary rationale: SNIPE resolves the well-characterized ManYZ phage resistance vs. metabolic cost trade-off — providing phage defense without sacrificing transporter function
2. Prevalence: 1,696 species across 33 phyla — far exceeding the >500 reported in the original paper
3. Accessory status: 86.7% accessory/singleton — consistent with mobile defense island carriage
4. Environmental niche: 22/64 AlphaEarth dimensions significantly differ between SNIPE and non-SNIPE species

H0 (random distribution) is rejected for taxonomic distribution and environmental niche association.

Limitations

• Annotation sensitivity: eggNOG Pfam annotations may miss divergent SNIPE homologues. The SNIPE nuclease domain is PF13455 (Mug113, GIY-YIG clan), not PF01541 (canonical GIY-YIG); only 54/4,572 DUF4041 clusters show Mug113 co-annotation, suggesting many SNIPE nuclease domains go undetected
• DUF4041 specificity: while DUF4041 (T5orf172) is strongly associated with SNIPE, some hits may represent non-SNIPE proteins
• AlphaEarth bias: 28.4% genome coverage, skewed toward environmental isolates
• Defense enrichment: the planned analysis of COG V defense gene density in SNIPE vs non-SNIPE species could not be completed via REST API (requires Spark Connect for the 93M × 132M row join)
• NMDC metagenomes: ecosystem breakdown was limited — only 2 samples matched the API query format; interactive MCP tool queries would improve coverage

Methods

Data Sources

• BERDL Pangenome (kbase_ke_pangenome): 293K genomes, 93M eggNOG annotations with Pfam domains, 132M gene clusters with core/accessory/singleton status
• Fitness Browser (kescience_fitnessbrowser / feba.db): 48 organisms, 228K genes, 27M fitness scores, 7,552 experiments (Price et al. 2018). Queried locally via SQLite (Figshare)
• AlphaEarth Embeddings: 64-dimensional environmental vectors for 83K genomes (28.4% coverage)
• PhageFoundry Genome Browsers: 4 species-specific databases (Acinetobacter, Klebsiella, P. aeruginosa, P. viridiflava)
• PhageFoundry Strain Modelling (phagefoundry_strain_modelling): 188 E. coli strains × 96 phages, binary infection outcomes + ML model (Gaborieau et al. 2024, Nature Microbiology)
• NMDC: National Microbiome Data Collaborative metagenome annotations
• InterPro/UniProt (IPR025280): 1,612 SNIPE proteins across 1,373 species, saved to data/interpro_snipe_proteins.csv

Domain Detection

SNIPE homologues identified by DUF4041 (PF13250 / IPR025280) Pfam domain in eggNOG annotations. GIY-YIG (PF01541) was queried independently but found to be the wrong Pfam family — the SNIPE nuclease is PF13455 (Mug113, GIY-YIG clan CL0418). Gene cluster pangenome status (core/accessory/singleton) from BERDL gene_cluster table.

Statistical Tests

• AlphaEarth dimension comparison: Welch’s t-test with Bonferroni correction (64 tests)
• Effect sizes: Cohen’s d with pooled standard deviation

Outputs

Data files (data/):

• duf4041_clusters.csv — 4,572 gene clusters with DUF4041
• giy_yig_species.csv — 728 species with GIY-YIG (from 1,000-cluster sample)
• snipe_both_domains.csv — 4,572 DUF4041 clusters (used as primary SNIPE marker; named for the original “both domains” query strategy but contains all DUF4041 hits after the co-occurrence analysis showed zero DUF4041+PF01541 overlaps — see Finding 4)
• snipe_taxonomy.csv — taxonomy for 1,696 SNIPE species
• all_species_taxonomy.csv — taxonomy for all 27,690 species
• snipe_alphaearth_stats.csv — 64-dimension statistical comparison
• snipe_species_embed_labels.csv — SNIPE labels for 15,046 species with embeddings
• snipe_phylum_summary.csv — per-phylum species/family/genus counts
• phagefoundry_defense_hits.csv — defense gene hits across 4 PhageFoundry species
• nmdc_duf4041_ecosystems.csv — NMDC ecosystem category counts
• fitness/feba.db — Fitness Browser SQLite database (7.4 GB, from Figshare)

Figures (figures/):

• snipe_phylum_distribution.png — phylum-level bar chart
• snipe_pangenome_status.png — core vs accessory pie chart
• snipe_alphaearth_pca.png — PCA of environmental embeddings

Reference

Based on PubMed article PMID 41741653. Saxton DS et al. Nature. 2026. DOI: 10.1038/s41586-026-10207-1