SNIPE Defense System in the KBase Data Lakehouse Pangenome

Raw report: snipe_defense_system Feeds into: adversarial-research-quality-assurance, callability-limited-comparative-inference, cofitness-network-architecture, condition-dependent-gene-tradeoffs, condition-specific-fitness, cross-species-fitness-transferability, cross-tenant-data-bridging, cultivation-collection-bias-in-ecological-genomics, data-landscape-ownership-and-coverage-bias, environment-embedding-geography, evidence-triangulation-for-functional-annotation, functional-marker-validation, gene-cooccurrence-ecological-guilds, genetic-perturbation-coverage-bias, homology-search-negative-evidence, horizontal-gene-transfer-driven-innovation, ontology-and-category-schema-sensitivity, pangenome-integration, phage-defense-syndromes-and-arms-race, phage-therapy-evidence-translation, provenance-aware-resource-discovery, two-speed-bacterial-genome

Overview

This report evaluates snipe-defense-system across the KBase Data Lakehouse pangenome, connecting its phage-defense mechanism to the metabolic cost of losing the ManYZ transporter, correcting the Pfam assignment of its nuclease domain, and surveying its prevalence, pangenome status, environmental association, and occurrence in phage-therapy targets. The analysis supports the hypothesis that SNIPE homologues are widespread, predominantly mobile, and associated with particular ecological contexts. ₁

Key Findings

Phage resistance without transporter loss

SNIPE constitutively localizes to the inner membrane and cleaves phage DNA as it passes through the ManYZ mannose-transporter pore, potentially providing phage resistance while retaining transporter function. By contrast, manY and manZ loss blocks or sharply reduces phage lambda growth but causes loss of mannose transport and partial glucose-transport defects: the manY (EIIC) knockout JW1807 gave zero phage growth after 24h, with loss of mannose transport and a partial glucose-transport defect, and the manZ (EIID) knockout JW1808 reduced phage growth by 5 orders of magnitude, with the same transport consequences reported (Burmeister et al. 2021). manX loss produces fitness defects specifically on D-mannose and D-glucosamine. Under phage pressure, man mutants sweep to >95% frequency, then decline sharply when phages evolve alternative injection mechanisms. The report frames SNIPE as resolving this trade-off — retaining full transporter function while cleaving phage DNA during injection, and so giving the phage-resistance benefit of man knockouts without the metabolic cost — but offers no direct comparative fitness measurement of SNIPE-bearing against man-knockout strains, so the trade-off resolution remains a mechanistic proposal. ₁

Direct RB-TnSeq (random barcode transposon sequencing) measurements from the Deutschbauer/Price Fitness Browser covered 168 E. coli K-12 experiments for each ManXYZ gene. The worst, average, and strongly deleterious fitness values were: manX, -3.93 worst, -0.249 average, 6 conditions below -1, and 4 below -2; manY, -3.82 worst, -0.171 average, 10 conditions below -1, and 4 below -2; and manZ, -4.14 worst, -0.082 average, 7 conditions below -1, and 4 below -2. All fitness scores are log2 ratios from RB-TnSeq, and carbon-source conditions were identified by the condition_1 field together with expGroup = 'carbon source'. ₁

The strongest defects occurred on D-glucosamine as sole carbon source, where the average manX, manY, and manZ fitness values were -3.79, -2.79, and -3.63, respectively, and on D-mannose as sole carbon source, where the averages were -2.75, -3.00, and -2.74. On D-trehalose, also a sole-carbon-source condition, the values were -1.27, -1.49, and —, and on sodium chlorite they were -1.47, -1.32, and -1.08. Cofitness correlations were 0.851 for manX↔manZ, 0.705 for manX↔manY, and 0.725 for manY↔manZ, supporting their operation as a single operon. ₁

The Fitness Browser contradicts the UniProt “fructose-specific” annotation for ManX: on D-mannose, manX, manY, and manZ fitness values were -2.75, -3.00, and -2.74, while fruA was -0.06; on D-glucosamine they were -3.79, -2.79, and -3.63, while fruA was -1.02; and on D-fructose they were -0.22, +0.15, and -0.07, while fruA was -1.44. These measurements support mannose/glucosamine specificity rather than fructose transport. ₁

Corrected SNIPE domain architecture

The report identifies PF13250 as the correct Pfam assignment for DUF4041, also called the SNIPE-associated domain and associated with InterPro IPR025280. Its domain table lists 1,612 proteins and 2,466 species for that entry, while its Methods section lists 1,612 SNIPE proteins across 1,373 species for IPR025280; the two species counts disagree, the source gives no basis for reconciling them, and both are therefore recorded here unresolved. It identifies PF13455, or Mug113, as the SNIPE nuclease family; PF13455 belongs to the GIY-YIG clan CL0418 but is not the canonical GIY-YIG family. Canonical GIY-YIG is PF01541, InterPro IPR000305, listed with 68,148 proteins covering restriction enzymes and DNA repair, and is explicitly not the SNIPE nuclease. The E. coli SNIPE protein A0A0A1A5Z2 is 558 aa long, with PF13250 at positions 232–333 and PF13455 at positions 443–520. ₁

Among the surveyed KBase Data Lakehouse annotations, zero gene clusters contained both DUF4041/PF13250 and canonical GIY-YIG PF01541; the report attributes this null result not to annotation artifacts but to these being genuinely different Pfam families. The 74,686 PF01541 clusters found in the survey are accordingly identified as non-SNIPE nucleases — restriction enzymes, DNA repair enzymes, and other GIY-YIG family members — and the lack of DUF4041 co-occurrence is read as genuine rather than an annotation artifact. The analysis found 4,572 DUF4041-containing gene clusters, of which 54 carried the description “Meiotically up-regulated gene 113,” consistent with full-length SNIPE proteins containing the Mug113 nuclease annotation. The report therefore recommends searching PF13250, DUF4041, or T5orf172 together with PF13455 rather than using PF01541 as the SNIPE nuclease marker. ₁

Pangenome prevalence and mobility

DUF4041/PF13250 occurred in 4,572 gene clusters across 1,696 species and 33 bacterial and archaeal phyla, substantially exceeding the >500 homologues reported in the original paper. The largest species counts were Pseudomonadota, 556 species; Actinomycetota, 334; Bacillota_A, 275; Bacillota, 206; Bacteroidota, 114; Nitrospirota, 45; Cyanobacteriota, 28; and Planctomycetota, 22. The report reads the patchy distribution across many phyla and families as consistent with horizontal gene transfer of defense islands; consistency with that mechanism is not a demonstration of it. ₁

Of the 4,572 DUF4041-containing clusters, 13.3% were core — 604 clusters — 30.7% were accessory, approximately 1,402 clusters, and 56.1% were singleton, 2,566 clusters; the combined accessory-plus-singleton fraction was 86.7%. The report reads this distribution as strong support for the defense-island mobility hypothesis: with only 13.3% of SNIPE clusters core within their species pangenome, SNIPE is typically gained or lost rather than vertically inherited. Mobility here is inferred from pangenome category composition rather than measured directly. The report’s own assessment summarizes the two results together: prevalence across 1,696 species and 33 phyla, far exceeding the >500 reported in the original paper, and 86.7% accessory/singleton status, which it reads as consistent with mobile defense-island carriage. ₁

The top functional descriptions for DUF4041-containing clusters were T5orf172, the gene-family name for DUF4041, with 2,109 clusters, and “Domain of unknown function (DUF4041)” with 1,566 clusters; histidine kinase followed with 319 clusters and seryl-tRNA aminoacylation with 238 clusters. The report estimates that 80.4% of DUF4041-containing clusters—3,675/4,572—were represented by the descriptions T5orf172 or DUF4041 and therefore formed a high-confidence SNIPE-family set. The remaining approximately 20%—897 clusters—included primary annotations such as histidine kinase, 319 clusters, and seryl-tRNA aminoacylation, 238 clusters, which may reflect DUF4041 being detected as a secondary or minor domain in multi-domain proteins whose primary function is unrelated to phage defense; the report suggests these likely represent annotation noise rather than functional SNIPE proteins, without establishing that classification. Applying the stricter description filter would retain approximately 80% of clusters as a high-confidence SNIPE set, and the report states that the 86.7% accessory rate and the taxonomic distribution patterns are not substantially affected by that subset, because the non-SNIPE annotations are distributed across the same phyla. ₁

COG (clusters of orthologous groups — broad functional categories) annotations of the DUF4041-containing clusters were distributed as D, cell cycle, 1,533; T, signal transduction, 1,118; J, translation, 1,066; and M, cell wall/membrane, 834. The report notes that the membrane-associated COG M category aligns with SNIPE’s inner-membrane localization — an alignment of annotation categories, not a localization measurement. ₁

Environmental association

Among species with AlphaEarth embeddings, SNIPE-bearing species numbered n=1,069 and non-SNIPE species n=13,977. Welch’s t-tests with Bonferroni correction found significant differences in 22 of 64 dimensions at p < 0.05; the largest effect was dimension A19, with Cohen’s d = 0.26 and p = 5.0e-14. Effect sizes ranged from |d| = 0.11–0.26, indicating a modest but consistent environmental shift rather than strict habitat segregation. Principal component analysis (PCA) of the embeddings showed SNIPE-bearing species broadly distributed but with a detectable environmental bias. The dimension comparison used Welch’s t-test with Bonferroni correction across 64 tests, and effect sizes were Cohen’s d with pooled standard deviation. ₁

AlphaEarth dimensions were derived from satellite remote-sensing imagery at sample collection latitude/longitude coordinates and can capture signals such as vegetation indices, temperature, precipitation, and land cover. The report does not assign a named biological interpretation to A19 or the other dimensions. AlphaEarth coverage was 28.4% of genomes and was biased toward environmental isolates with latitude/longitude metadata; the resource supplies 64-dimensional environmental vectors for 83K genomes, that same 28.4% coverage. ₁

Fitness Browser and phage-host evidence

The Fitness Browser contained 48 organisms, 228K genes, 27M fitness scores, and 7,552 experiments. ManXYZ was present only in E. coli K-12, the Keio collection, among those organisms — a comparison made by gene name across the 48 Fitness Browser organisms, so this is a gene-name-query result rather than an unrestricted biological absence — and no Klebsiella strain in the Fitness Browser carries ManXYZ by gene name. One full two-domain SNIPE protein occurred in Methanococcus maripaludis JJ at locus MMJJ_RS01635. This protein carried PF13250, PF13455, and PF10544 and had 129 experiments of fitness data; its minimum fitness was -1.16 on formate/acetate and it was dispensable under most conditions. It is the first organism with both a confirmed SNIPE protein and genome-wide fitness data, but as an archaeon its phage-biology context differs significantly from the Enterobacterales phage lambda system. ₁

PF13455 occurred in 7 genes across 6 Fitness Browser organisms: Azospirillum brasilense, Bacteroides thetaiotaomicron, Cupriavidus, Desulfovibrio vulgaris, Paraburkholderia kururiensis, and M. maripaludis. Only M. maripaludis had the complete PF13250-plus-PF13455 architecture. ₁

The phagefoundry_strain_modelling database holds the complete phage-host interaction dataset from Gaborieau et al. 2024, Nature Microbiology. It had been previously overlooked — documented as having “timed out during discovery” — and contains 18 tables with experimental phage-host phenotype data. ₁

It covered 188 E. coli strains, from the ECOR and NILS collections plus clinical and environmental isolates, and 96 phages, comprising 41 Podoviridae, 32 Myoviridae, and 23 Siphoviridae, with 17,672 binary infection outcomes, of which 3,929 were positive and 13,743 negative, giving a reported overall infection rate of 22%. The report writes this as 188 strains × 96 phages = 17,672 outcomes, a statement that is arithmetically inconsistent; the figures are reproduced here exactly as reported and no corrected product is supplied. ₁

Its machine-learning model used 1,582 gene-cluster presence/absence features with SHAP importance scores and achieved AUC = 0.883 and accuracy = 84.3%. ₁

The Lambdavirus phage 411_P1 infected only 1 of 188 strains, NILS06 — a 0.5% infection rate and the lowest of any phage in the dataset. Comparison infection rates were 43.4%, 2,531/5,828, for Myoviridae; 13.0%, 978/7,520, for Podoviridae; and 9.7%, 420/4,324, for Siphoviridae; the Lambdavirus rate is given as 0.5%, 1/188. ₁

The report proposes widespread ManYZ variation or loss, conferring near-universal lambda resistance across diverse E. coli strains, as the explanation for the low Lambdavirus infection rate; the interaction matrix is consistent with this reading but does not establish it. The report attributes to Gaborieau et al. the comparative finding that adsorption factors — surface receptors — rather than antiphage defense systems are the primary predictors of phage-host interaction, and argues that this places SNIPE at exactly the adsorption/injection interface where phage success is determined; that application to SNIPE is an extrapolation, not a direct test of SNIPE within the interaction matrix. ₁

Klebsiella detection

PhageFoundry detected DUF4041 in Klebsiella but not in Acinetobacter, P. aeruginosa, or P. viridiflava in the four-species browser comparison. The corresponding counts were: Klebsiella, DUF4041 1, GIY-YIG 1, nuclease 98, restriction 42, and abortive infection 2; Acinetobacter, 0, 3, 91, 33, and 2; P. aeruginosa, 0, 2, 125, 54, and 2; and P. viridiflava, 0, 0, 90, 41, and 1. ₁

The Klebsiella DUF4041 hit was 1 description covering 3 proteins—JJW41_17025, KFB13_RS08150, and KFB31_RS01490—each 558 aa, the same length as the characterized E. coli SNIPE protein A0A0A1A5Z2; the report infers full-length two-domain PF13250-plus-PF13455 proteins from this length match rather than confirming the architecture directly. The same genome browser contained 28 mannose PTS IID/ManZ proteins, 1 mannose PTS IIA/ManX protein, and 4,619 PF02378 PTS_EIIC annotations against comparator counts of 891 for Acinetobacter, 1,067 for P. aeruginosa, and 260 for P. viridiflava, plus 571 PF00358 PTS_EIIA_1 annotations against 0, 0, and 0 in those same three comparators, making Klebsiella the only PhageFoundry species with both SNIPE and the PF00358 ManX-family PTS domain by exact Pfam accession. The co-occurrence is consistent with SNIPE defending a mannose-transporter phage-entry route, but no Klebsiella SNIPE or ManYZ fitness data were available. The report calls the Klebsiella DUF4041 detection clinically relevant and proposes that SNIPE could affect phage-therapy efficacy in this pathogen; that is a proposal, not a demonstrated effect. ₁

The report’s own Pfam labelling is internally inconsistent: its annotation table labels PF02378 as PTS_EIIC and PF00358 as PTS_EIIA_1, while its query footnote labels PF00358 as PTS_EIIC and PF02378 as PTS_EIID. The conflict is recorded here unresolved, since the source gives no basis for preferring either labelling. ₁

Data Sources and Outputs

The pangenome source was kbase_ke_pangenome on the KBase Data Lakehouse: 293K genomes, 93M eggNOG annotations with Pfam domains, and 132M gene clusters with core/accessory/singleton status. ₁

The remaining sources were the Fitness Browser (kescience_fitnessbrowser / feba.db), queried locally via SQLite; the AlphaEarth embeddings; the four species-specific PhageFoundry genome browsers, for Acinetobacter, Klebsiella, P. aeruginosa, and P. viridiflava; PhageFoundry Strain Modelling (phagefoundry_strain_modelling), holding 188 E. coli strains × 96 phages of binary infection outcomes plus the machine-learning model; NMDC, the National Microbiome Data Collaborative metagenome annotations; and InterPro/UniProt entry IPR025280. SNIPE homologues were identified by the DUF4041 domain, PF13250 / IPR025280, in eggNOG annotations, with PF01541 queried independently and then rejected as the wrong Pfam family. ₁

Two saved outputs carry naming caveats. giy_yig_species.csv lists 728 species with GIY-YIG drawn from a 1,000-cluster sample, so it is a sample-derived count rather than an exhaustive species count. snipe_both_domains.csv, named for the original “both domains” query strategy, in fact contains all 4,572 DUF4041 clusters, because the co-occurrence analysis found zero DUF4041+PF01541 overlaps. ₁

Figures

• snipe_phylum_distribution.png — phylum-level bar chart. ₁
• snipe_pangenome_status.png — core vs accessory pie chart. ₁
• snipe_alphaearth_pca.png — PCA of environmental embeddings. ₁

The project is based on PubMed article PMID 41741653: Saxton DS et al., Nature, 2026, DOI 10.1038/s41586-026-10207-1. ₁

Caveats and Unresolved Gaps

The report states that eggNOG Pfam annotations may miss divergent SNIPE homologues, and that only 54/4,572 DUF4041 clusters showed Mug113 co-annotation, suggesting that many SNIPE nuclease domains may go undetected. DUF4041 is strongly associated with SNIPE but may also occur in non-SNIPE proteins. ₁

The AlphaEarth analysis covered 28.4% of genomes and was skewed toward environmental isolates with geographic metadata, limiting generalization to the full pangenome. The planned comparison of COG V defense-gene density could not be completed through the REST API because it required Spark Connect for a 93M × 132M row join. ₁

NMDC metagenome ecosystem analysis was limited because only 2 samples matched the API query format; the report identifies interactive MCP-tool queries as a way to improve coverage. PhageFoundry provides genome annotations in GenomeDepot format but no fitness/TnSeq tables for Klebsiella, despite its ManYZ annotations there, so Klebsiella-specific SNIPE or ManYZ fitness remains unmeasured in that resource; strain-specific mutant libraries in natural SNIPE-carrying strains such as K. pneumoniae NCTC9140 or HS11286, or curated external K. pneumoniae Tn-Seq datasets, would be needed to test SNIPE and ManYZ fitness directly. ₁

The report’s H1 assessment is supported across prevalence, accessory status, environmental association, and the proposed ManYZ trade-off, while H0 is rejected for taxonomic distribution and environmental niche association. However, the environmental result establishes statistical association rather than a specific causal habitat mechanism, and the archaeal Fitness Browser example does not establish the Enterobacterales phage-lambda mechanism in archaea. ₁

Slots Into

• phage-defense-syndromes-and-arms-race — SNIPE’s proposed DNA-cleavage mechanism, the ManYZ resistance-versus-metabolic-cost trade-off, phage-host interaction evidence, and the corrected PF13250/PF13455 architecture. ₁
• pangenome-integration — the detection of 4,572 DUF4041 clusters across 1,696 species and 33 phyla, with 86.7% accessory or singleton status. ₁
• environment-embedding-geography — the AlphaEarth comparison showing significant differences in 22 of 64 dimensions, with explicit coverage and sampling-bias caveats. ₁
• condition-specific-fitness — the 168-condition ManXYZ fitness measurements and the substrate-specific separation of mannose/glucosamine from fructose phenotypes. ₁
• functional-marker-validation — the correction of the SNIPE nuclease marker from PF01541 to PF13455/Mug113, the zero DUF4041+PF01541 co-occurrence, and the T5orf172/DUF4041 description filter as a confidence screen. ₁
• phage-therapy-evidence-translation — the single Klebsiella DUF4041 detection among four PhageFoundry species and the proposal, not demonstration, that SNIPE could affect phage-therapy efficacy. ₁
• condition-dependent-gene-tradeoffs — the manY/manZ knockout phage-resistance benefit set against loss of mannose transport and partial glucose-transport defects. ₁
• homology-search-negative-evidence — only 54/4,572 DUF4041 clusters carrying Mug113 co-annotation, so missing nuclease annotation is weak evidence of a missing domain. ₁
• cultivation-collection-bias-in-ecological-genomics — AlphaEarth coverage of 28.4% of genomes, biased toward environmental isolates with coordinate metadata. ₁
• cross-tenant-data-bridging — the NMDC query matching only 2 samples and the previously overlooked phagefoundry_strain_modelling database. ₁