Viral isolation reveals novel and diverse phages infecting natural stream biofilms

25 Biofilms are ubiquitous microbial assemblages 1, consisting of complex, spatially structured 26 microbial communities that lend emergent properties such as resource retention and resilience 27 against environmental stress 2. These sessile communities have been under a negative 28 spotlight in medical and industry settings but they play important roles in maintaining 29 fundamental functions across a wide range of ecosystems . For example, benthic biofilms in 30 streams and rivers orchestrate carbon and nutrient cycling at the base of lotic food webs 3,4 31 while often acting as sentinels for monitoring climate-induced changes to the ecosystem5,6. 32 33 As the dominant microbial lifestyle for millienia7, biofilms also possess a long eco-evolutionary 34 history with bacteriophages (phages for short), which are viruses that infect bacteria. Virulent 35 phages adopt the lytic cycle exclusively where infected cells are lysed to release assembled 36 virions, while temperate phages alternate between active lytic replication and dormant 37 lysogeny as the phage genome integrates with the bacterial chromosome 8. Through these 38 lifecycles, phages regulate bacterial abundance and foster microbial diversity9, co-evolution, 39 and horizontal gene transfer 10,11. In doing so, phage s are also the main source of microbial 40 mortality in natural systems, affecting broader biogeochemical processes by liberating organic 41 matter via cell lysis 12,13 while also shifting metabolic demands of infected populations by 42 reprogramming their hosts14,15. Phage predation is therefore expected to shape the ecology 43 and evolution of biofilm communities 15. However, the complex interactions between phages 44 and biofilm-dwelling bacteria in nature remain poorly understood. 45 46 Major strides in unravelling phage-biofilm interactions were largely made within the motives of 47 phage therapy research16. Key insights were derived from single-species biofilms of model 48 and clinically-relevant bacterial species such as Escherichia coli, Staphylococcus epidermidis, 49 and Vibrio cholerae. These studies have uncovered the protective effects of the biofilm matrix 50 against phages17, the impact of cell packing density on viral spread across the biofilm 18, and 51 communal signalling amongst biofilm-dwellers under phage invasion19,20. Few have extended 52 these insights to mixed-species biofilm 21,22, yet these systems remain confined to model 53 bacteria, thereby representing a narrow subset of phage-biofilm interactions. In contrast, field-54 based metagenomics of native biofilms have revealed robust coupling and host-specificity 55 between phages and bacteria that extend to global scales23,24, often reporting microbial taxa 56 that are underrepresented in contemporary biofilm research. However, a mechanistic basis 57 underlying these field observations often remain unresolved due to a paucity of matched, 58 culturable phage -bacteria pairs necessary for empirical investigation and validation. For 59 instance, leveraging synthetic assemblages to mimic environmental biofilms and investigate 60 viral-driven mechanisms in spatially structured heterogeneous communities. Many extensive 61 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint and well -coordinated initiatives have been undertaken to develop and maintain phage 62 repositories globally25 such as PhagesDB26, BASEL27, and CPL28. However, these collections 63 remain largely focused on model and clinically relevant bacterial species with environmental 64 strains remain comparatively underrepresented. 65 66 To advance the field beyond these confines , we present the Alpine Lotic Phage (ALP) 67 collection: a resource compris ing of 28 unique phage isolates infecting 1 4 environmental 68 biofilm-forming bacterial species, spanning three taxonomic classes (α- and γ-proteobacteria, 69 and Flavobacteriia). By integrating morphological, phenotypic, and genomic characterization, 70 the ALP collection represents a highly novel and diverse culturable fraction of viruses 71 previously hidden within the viral “dark matter” of natural stream biofilms. We envisage this 72 resource to provide a foundation for empirical and computational studies aimed at 73 understanding phage–biofilm dynamics, adaptation, and ecosystem function within 74 environmentally relevant context s. This collection also offers a valuable reference for 75 advancing phage gene annotation and exploring the biotechnological potential of alpine 76 phages. 77 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint

78 Phage isolation workflow and overall composition of the ALP collection 79 To facilitate the isolation of phages infecting biofilm-forming bacteria in nature, we selected a 80 panel of 3 7 bacterial isolates from a recently established in-house collection derived from 81 stream biofilms. This panel spans 24 genera across eight taxonomic classes , representing 82 bacterial taxa commonly reported in freshwater biofilm communities (Table S1). Notably, 83 genera such as Rhodoferax, Flavobacterium, Massilia, and Sphingomonas are well -84 documented inhabitants of streambed biofilms, with Rhodoferax and Flavobacterium reported 85 as abundant and widespread across glacier-fed streams globally29,30. These taxa contribute to 86 key ecosystem processes, including carbon and nutrient cycling 24, and are increasingly 87 susceptible to climate-driven environmental change6,31. 88 89 All bacterial isolates (and their corresponding phages) grew on standard R2A media and 90 encompassed a spectrum of phenotypes relevant to biofilms including a range of growth rates, 91 surface motility, aggregative behaviour (i.e. floc -forming), and pigment production. Stream 92 water was collected from the confluence of a groundwater-fed stream (La Vièze) and a glacier-93 fed tributary (La Saufla) (Fig.S1), which was subsequently concentrated using tangential flow 94 filtration, followed by 0.45 μm dead-end filtration to remove most prokaryotic and eukaryotic 95 cells while retaining the concentrated viral fraction (Fig.1A). Phages were isolated using soft-96 agar overlays of the concentrated water samples where observed plaques were picked , 97 double plaque -purified, and amplified for downstream characterisation by whole-genome 98 sequencing and transmission electron microscopy (TEM) (Fig.1A). 99 100 In total, 57 phage isolates were recovered across 14 bacterial host genera (Fig.2A). Rahnella 101 inusitata yielded the highest number of isolates (n = 22), followed by Pseudomonas cyclaminis 102 (n = 7) and Pseudomonas haemolytica (n = 6). All isolates were culturable except a single 103 Sphingomonas phage, which could not be amplified sufficiently for DNA extraction and 104 transmission electron microscopy (TEM). High-quality genomic DNA was obtained from 43 of 105 the 56 culturable isolates, whereas the remaining isolates - particularly those infecting P. 106 cyclaminis - were recalcitrant to DNA extraction. Dereplication based on genome identity 107 resulted in a final collection of 28 unique phage s (Fig. 2A; Table S2). All 28 assembled 108 genomes were independently classified as viral by both VIBRANT and geNomad. CheckV 109 assessment indicated that 26 genomes were high quality and 2 (Rahnella phages B311P5 110 and B311P9) were medium quality (Table S3). All genomes were assigned by geNomad to the 111 class Caudoviricetes, with eight further resolved at the family level: six Autographiviridae, one 112 Casjensviridae, and one Schitoviridae (Table S3). Across host genera, between 1 and 9 unique 113 phages were recovered per bacterial genus. Notably, the 22 initial Rahnella-infecting isolates 114 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint dereplicated into nine distinct phages, indicating frequent re-isolation of closely related viruses 115 targeting this host (Fig. 2A). 116 117 118 Fig.1 Phage isolation workflow and bioinformatic assembly pipeline of viral genomes. A) ~120 L 119 of stream water was collected from an alpine stream and was filtered to remove prokaryotes with a 0.45 120 µm filter. The viral fraction retained was concentrated by ~100-fold via tangential flow. The concentrated 121 stream water was screened for phages using soft-agar overlay mixed with bacterial host broth culture. 122 Visible plaques were picked, double-purified, and amplified in host broth culture to yield pure high-titre 123 phage lysates. These lysates were then leveraged to characterise phages via transmission electron 124 microscopy (TEM) and whole-genome sequencing. B) Summary of DNA sequencing and bioinformatic 125 pipeline to yield unique, high -quality, and complete phage genomes which constitutes the ALP 126 collection. Hybrid assemblies with long -read Oxford Nanopore and short -read MiSeq sequencing 127 platforms were adopted for selected isolates with marginal DNA quantity due to inherent challenges 128 with DNA extraction. Tools adopted in each step of the pipeline are indicated by bolded grey text, where 129 tool versions are referenced in online methods. 130 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint Alpine stream phages exhibit diverse viral and plaque morphologies 131 Unique viral isolates were imaged with TEM to provide a first glance at the morpho types 132 pertaining to phages infecting stream biofilms . Of the 28 phage isolates, we were able to 133 acquire electron micrographs for 18 (Table S2). TEM images revealed a collection of tailed 134 viruses comprising 7 siphoviruses, 6 podoviruses, and 5 myoviruses , highlighting the 135 heterogeneous viral population infecting sessile bacterial communities in streams (Fig.2B). 136 Two myoviruses, Rahnella phage B311P2 and Comamonas phage B146P1, exhibited large 137 virion dimensions with capsid diameters exceeding 100 nm, consistent with structural 138 dimensions expected of jumbophages32. 139 140 In addition to virion structure, plaque morphologies were assessed for the 18 imaged isolates 141 to capture basic phage phenotypes under semi-solid growth conditions. Here, plaque sizes 142 varied markedly across the collection, ranging from small punctate plaques to large clearings 143 up to 280 μm in diameter (Fig.2B). Plaques exceeding 200 µm in diameter were predominantly 144 produced by podoviruses, in line with their smaller virion dimensions which enhances 145 diffusivity in structured matrices relative to larger myo- and siphoviruses. Recent biophysical 146 insights further suggest that larger plaque sizes may also be associated with shorter viral latent 147 periods which lends higher phage virulence33. Plaques also differed in qualitative features, 148 including central turbidity and translucent halos extending beyond the plaque margin. Most 149 isolates produced clear plaques reflecting their effective host -killing capacity. H owever, 150 Massilia phage B343P1 and Pseudomonas phage B427P1 formed turbid plaques, indicative 151 of reduced lytic capacity relative to other ALP isolates. 152 153 In addition, 8 of the 18 imaged phages produced translucent halos surrounding the plaque 154 border, with radii ranging from 30 to 250 µ m (Fig.2B). These halos are consistent with 155 depolymerase activity, whereby phage -encoded enzymes degrade bacterial surface 156 polysaccharides including biofilm exopolysaccharides to facilitate access to host receptors34. 157 Intriguingly, Janthinobacterium phages B503P1 and B503P2 were dereplicated as near-158 identical genomes yet, the isolates exhibited distinct plaque morphologies (Fig.2B). Isolate 159 B503P2 produced larger plaques with more extensive depolymerase halos compared to 160 B503P1, indicating phenotypic divergence despite 99.9% genome identity. Closer inspection 161 revealed a single nucleotide substitution in the tail fibre gene, resulting in a serine -to-162 asparagine substitution at position 590 (S590N; Fig. S2). Although the structural 163 consequences of this substitution is not known, it occurs adjacent to a region partially 164 homologous to a pectate lyase domain , which was associated with depolymerase activity in 165 Acinetobacter phages35. Collectively, these observations reveal the substantial morphological 166 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint and phenotypic diversity of lotic phages where the latter underscores the various viral infection 167 traits leveraged against structured sessile microbial communities. 168 169 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 170 Fig.2 Plaque and virus morphologies of alpine stream phages. A) A total of 28 unique phage 171 isolates were derived from 14 alpine bacterial species in our viral screening effort. Coloured legend 172 represents the bacterial host taxonomic class that were targeted by phages while phage icons represent 173 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint the number of genomes after dereplication (e.g. 2 unique phages were isolated for Flavobacterium sp.) 174 that were isolated per bacterial species. Question mark designates the unknown number of unique 175 phages as dereplication was not possible without successful phage genome extraction (except 176 Sphingomonas sp. phage which only had one isolate but DNA extraction was unsuccessful). B) Phage 177 plaques were assayed using 0.4% soft agar overlay with brightfield images demonstrating a variety of 178 plaque sizes and morphologies. Haloed plaques demonstrating potential depolymerase activity were 179 marked with asterisks. Transmission electron microscopy images of 18 phage isolates revealed a 180 collection of tailed phages consisting of myoviruses, podoviruses, and siphoviruses . Only the 181 Sphingomonas phage isolate was neither successfully sequenced nor imaged due to poor culturability 182 in vitro. 183 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint The ALP collection represents an emerging repository of novel viruses infecting stream 184 biofilm-forming bacteria 185 Given the understudied nature of viruses associated with natural stream biofilms, we assessed 186 the genomic novelty represented by the ALP collection. Briefly, phage genomes were BLAST-187 queried against the NCBI nucleotide database where we retained the top five hits based on 188 sequence coverage and nucleotide identity. The majority of isolates exhibited limited similarity 189 to previously described viruses, with 20% coverage against publicly 191 available viral genomes, indicating that most ALP phages lack close relatives in existing 192

databases and public collections (Table 1). Among these, Pseudomonas phage 193 B508P1 demonstrated the strongest match (94% genome coverage and nucleotide identity ) 194 with a cold-active podovirus VSW-3 infecting Pseudomonas fluorescens in wetlands36. This is 195 followed by Rahnella phage B311P4 with 96% coverage and 87% identity against phage 196 KLB24 infecting plant-associated Klebsiella sp. which was isolated from farmland rain 197 puddles37. Further adding to the relatedness of ALP isolates with other aquatic phages , 198 Pseudomonas phage B072P2 displayed 81–86% genome coverage and ~85% nucleotide 199 identity to podoviruses also infecting P. fluorescens, notably the psychrophilic freshwater 200 phage phiGM22-338. The remaining phages infecting Rahnella, Massilia, Brevundimonas, and 201 Rhodoferax exhibited varying similarity (23–50% coverage, >74% identity) to known phages, 202 except Rahnella phage B311P3 which was partially found amongst Serratia bacterial 203 genomes, implying a possible distant prophage relative within this genus. We also note that 204 Rahnella jumbophage B 311P2 appears distantly related to two other jumbophages 205 Cronobacter phage vB_CsaM_GAP32 and Escherichia phage PBECO 4 , suggesting that 206 these large viruses could share core genomic segments. 207 208 Table 1: List of top two BLAST with >20% query coverage of ALP phages against NCBI database 209 Phage (φ) isolate Closest hits Query coverage % identity Accession # Brevundimonas φ B307P1 Caudoviricetes sp. (MAG)a 38% 80.94% OR222513.1 Caulobacter phage Seuss 14% 77.03% NC_047757.1 Brevundimonas φ B307P2 Caudoviricetes sp. (MAG)a 36% 81.23% OR222513.1 Caulobacter phage Seuss 14% 77.02% NC_047757.1 Massilia φ B343P2 Caudoviricetes sp. (partial MAG)a 68% 85.50% BK020520.1 Caudoviricetes sp. (MAG)a 1% 75.35% OR222459.1 Pseudomonas φ B072P2 Pseudomonas phage phiGM22-3 86% 84.73% MW627366.1 Pseudomonas phage phi2 81% 85.57% NC_013638.1 Pseudomonas φ B508P1 Pseudomonas phage VSW-3 94% 94.39% NC_041885.1 Providencia phage PSTNGR1 2% 74.38% MW145136.1 Pseudomonas φ B529P1 Pseudomonas phage phiPsa300 50% 82.01% NC_073687.1 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint Pseudomonas phage phiPsa347 49% 81.99% NC_073685.1 Rahnella φ B311P2 Cronobacter phage vB_CsaM_GAP32 33% 72.00% JN882285 Escherichia phage PBECO 4 15% 68.00% KC295538 Rahnella φ B311P3 Serratia fonticolab 23% 79.02% CP013913.1 Serratia sp. JSRIV004b 22% 79.07% CP074140.1 Rahnella φ B311P4 Bacteriophage sp. (partial MAG)a 96% 87.22% OP072654.1 Klebsiella phage vB_KM5a1-KLB24 96% 87.13% PP554338.1 Rahnella φ B311P5 Enterobacter phage EC151 27% 74.50% MW464860.1 Klebsiella phage vB_Ko_K4PH164 8% 73.92% OY979482.1 Rahnella φ B311P7 Enterobacter phage EC151 36% 76.76% MW464860.1 Klebsiella phage vB_Ko_K4PH164 22% 75.07% OY979482.1 Rhodoferax φ B534P2 Variovorax phage Gard 38% 80.25% PV920657.1 Variovorax phage VAC_51 32% 79.67% OX359471.1 210 a MAG denotes metagenome-assembled genomes that are either complete or partial. 211 b Closest hits with bacterial origins i.e. non-viral hits. 212 213 To further contextualise novelty at a global scale, the ALP genomes were queried against 214 LOGAN, which is a planetary -scale contig database curated from public sequence read 215 archives39. Using a permissive k -mer coverage threshold of 0.25, 16 out of 28 ALP phages 216 returned positive matches (Fig.3A). These matches were geographically widespread, 217 predominantly originating from the European continent but also spanning Asia and North 218 America (Fig.3A, Table S5). The matches were also associated with diverse environment al 219 sources ranging from freshwater ecosystems including rivers, glaciers, groundwater and 220 alpine streams to other environments such as soil, plants, and terrestrial animals, followed by 221 anthropogenic food and wastewater sources (Figs.3A and B) . Overall, 7 phage isolates 222 matched with contigs from a single source while the remaining 9 isolates were associated with 223 contigs from multiple environments of up to 6 sources (Fig.3B). Despite this broad distribution, 224 individual matches generally exhibited low similarity, with a median k-mer coverage score of 225 0.36 (range: 0.25 –0.85) against LOGAN contigs (Fig.3A). Collectively, the sparse 226 representation of phages in both reference and global metagenomic data bases underscores 227 the substantial novelty behind viruses associated with stream biofilms. 228 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 229 Fig.3 Matches of ALP genome sequences against planetary -wide contig database. A) Total 230 number of phage isolates with matches to contigs within the LOGAN database and a histogram 231 depicting the spread of k-mer coverage scores of all positive hits to the database. The map represents 232 the geographical distribution of these hits while the colours represent the source of the contigs identified 233 within the database. Asterisk denote matches to “aquatic (unknown)” sources do not possess any 234 geographical metadata and are thus, not located on the map. B) Cumulative k -mer coverage of 235 sequence matches between a 1kb sequence of the ALP isolates to the LOGAN contigs, where colours 236 also represent the contig source stipulated within the database. 237 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint Auxiliary coding sequences in phages facilitate host takeover for infection and 238 environmental resilience 239 Across the 28 phage isolates, genome sizes spanned over an order of magnitude from 37 to 240 363 kb (Fig.4A). Two isolates, Comamonas phage B146P1 (206 kb) and Rahnella phage 241 B311P2 (363 kb), exceeded 200 kb and were hence, classified as jumbophages consistent 242 with their large virion dimensions (Fig.2B). Differences in phage genome size were matched 243 by expansions in capsid size, with phages below 83 kb exhibiting relatively similar capsid 244 diameters (69 nm on average) . Meanwhile, capsids enlarged progressively as genome size 245 exceed 111 kb, reaching 133 nm in the Rahnella jumbophage B311P2 with a 363 kb genome. 246 Allometric analysis also revealed proportional scaling between genome size and estimated 247 capsid volume (allometric coefficient α = 0. 91), indicating that DNA packing density was 248 relatively conserved across the ALP collection following structural principles of icosahedral 249 dsDNA viral capsids40 (Fig.4B, Table S6). Furthermore, coverage patterns from read mapping 250 identified DNA packaging mechanisms41 for 14 phage isolates: 1 utilized a 3’ cos terminus, 4 251 harboured pac sites, and 9 employed direct terminal repeats (DTR) ( Table S7), where t he 252 Rahnella jumbophage B311P2 possessed a remarkably long DTR of 19,824 bp. Together, 253 these features illustrate the genomic architectural diversity, packaging constraints, and 254 mechanisms among isolates of the ALP collection. 255 256 Genome annotation resolved putative functions for 9 – 54% of predicted coding sequences 257 (CDS), with gene functions predominantly associated with virion structure , nucleotide 258 metabolism, and host lysis (Fig.4A, Table S6). Only one isolate – Rahnella phage B311P3 – 259 was identified as a temperate phage, encoding both integrase and excisionase while also 260 possessing genes consistent with lysogenic capacity such as CII- and CIII-like transcriptional 261 regulators and Ren-like superinfection exclusion factor. In addition, Rahnella phage B311P3 262 was also detected as an integrated prophage within the Rahnella host genome. Meanwhile, 263 larger phages generally encoded substantial tRNAs relative to the entire collection, comprising 264 5 – 11% of total CDS (Fig.4A). This pattern aligns with previous observations that phages with 265 larger genomes leverage self-proprietary tRNAs to sustain viral r eplication as host 266 translational capacity becomes limiting during late-stage infection42. 267 268 Beyond core functions, morons and auxiliary metabolic genes (AMGs) were predicted in 20 of 269 28 phage genomes, ranging from 1 – 8 genes per genome except for the Rahnella 270 jumbophage, which encoded 33 auxiliary genes (Figs.4A and C, Table S8). The majority of 271 these genes however, corresponded to membrane proteins and biocatalytic enzymes of broad 272 unclassified functions. Among fully annotated genes, functions associated with host takeover 273 and anti-defence were prevalent, such as the host mRNA-modulating HicA-like toxin-antitoxin 274 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint system43, NAD synthesis and scavenging factors, followed by anti-restriction and anti-CRISPR 275 systems. Several phages also encoded enzymes implicated in phage DNA modification, such 276 as 2 -oxoglutarate/Fe(II)-dependent oxygenases 44 and phosphoheptose isomerases 45, 277 consistent with strategies to evade intracellular host restriction and CRISPR -Cas immunity 278 (Fig.4C). 279 280 In contrast, relatively few AMGs were predicted to directly enhance host environmental 281 resilience. These included tellurite resistance genes encoded by the Rahnella jumbophage 282 and Pseudomonas phage B529P1, as well as genes involved in antioxidative protection and 283 cellular energetics in other isolates (e.g. gluthathionylspermidine synthase and porphyrin 284 biosynthesis; Fig.4C). These AMGs are consistent with promoting survival and maintaining 285 cellular metabolism in biofilm-associated microenvironments characterised by accumulation 286 of trace metal ions46,47, steep redox gradients 48, and chronic oxidative stress49. Notably, 4 287 Rahnella phage isolates encoded enzymes within queuosine biosynthesis pathway alongside 288 a queuine tRNA -ribosyltransferase, suggesting roles in phage DNA modification for host 289 immune evasion 50 and maintaining translational efficiency to sustain bacterial cells in 290 biofilms51. Despite expectations that temperate phages preferentially retain host -beneficial 291 genes52, only a single AMG was detected in the temperate Rahnella phage B311P3, encoding 292 a sporulation factor of unclear relevance to the non-sporulating Rahnella genus. Collectively, 293 these findings reveal that stream-biofilm phages encode diverse auxiliary gene functions 294 primarily associated with host takeover and antiviral defence evasion, with a smaller subset 295 linked to host cellular and biofilm stress tolerance. The large unannotated portion of ALP 296 genomes also represents a resource for exploring additional phage-encoded adaptations or 297 even, gene products of potential biotechnological relevance. 298 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 299 Fig.4 Genomic features of ALP isolates and profile of auxiliary coding sequences. A) Genome 300 analysis of sequenced phage isolates where the red and green line plots correspond to genome size 301 (kb) and capsid diameter (nm) respectively, which are plotted on the top x-axis, while the coloured bar 302 chart indicates the proportion of genes predicted , corresponding to the bottom x-axis. B) Log -303 transformed plot of capsid volume (approximated as spherical volume: [p ÷ 6] ´ D3, where D is the 304 capsid diameter) against genome length. The regression line is shown in blue with the accompanying 305 allometric exponent (a i.e. gradient of the log-transformed regression), R-squared value, and p-value. 306 C) Frequency table of morons and AMGs predicted in 20 phage isolates using nucleotide-based 307 (pharokka) and protein structure-based (phold) algorithms. Gene functions are broadly categorised into 308 6 groups. Each cell within the table is separated to four quadrants which are colour -coded based on 309 phold annotation confidence i.e. (blue = high, green = medium, and low = orange), while the bottom -310 right red quadrant indicates genes predicted by the nucleotide-based algorithm, pharokka. The number 311 within each quadrant shows the frequency of gene functions predicted in accordance with the type of 312 algorithm employed and confidence level for phold annotations. 313 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint Phage infections in microfluidics reveal host -specific architectural responses of 314 environmental isolate biofilms. 315 Microfluidic platforms have emerged as powerful systems to resolve phage-biofilm interactions 316 with high spatial and temporal precision , lending crucial insights into mechanistic responses 317 of biofilms under phage infection 17,20,21,53 and viral propagation strategies in structured 318 communities18,54. Yet, their application to environmentally derived phage -host systems 319 remains limited. To demonstrate the utility of the ALP collection, we performed time -course 320 infections of environmental isolate biofilms in microfluidic devices. We selected three 321 representative bacterial isolates exhibiting high, moderate, and low biofilm-forming capacities 322 (Massilia sp., R. inusitata, and P. fluorescens, respectively), based on crystal violet staining of 323 static overnight biofilms (Fig.5A). These isolates were seeded into microfluidic channels and 324 allowed to establish for 24 h ours under continuous flow prior to infection with their 325 corresponding phages: Massilia phage B343P1, Rahnella jumbophage B311P2, and 326 Pseudomonas phage B508P1 (Fig.5B). 327 328 Over the course of 96 hours following infection, biofilms were monitored using brightfield 329 microscopy. These environmental isolates are non -fluorescent and display heterogeneous 330 growth patterns under fluid flow (Fig.5B), highlighting the exploratory potential of extending 331 microfluidic phage-biofilm assays beyond fluorescently tagged model systems. Here, time -332 course imaging revealed distinct architectural characteristics across the three biofilm species. 333 Massilia sp., a floc -forming isolate, developed heterogeneous biofilms with microcolonies 334 reaching up to ~200 µm wide. Meanwhile, environmental P. fluorescens, despite limited biofilm 335 formation in static assays, formed dense communities with patterned fronts that aligned with 336 the direction of flow. In contrast, R. inusitata formed comparatively homogeneous and low-337 density lawn-like biofilms spanning the channel surface (Fig.5B). Notably, Massilia sp. and P. 338 fluorescens biofilms exhibited quasi -cyclical development even in the absence of phage, 339 characterized by initial disruption of heterogeneous biofilms under flow between 48 – 72 hours 340 followed by structured re-growth. 341 342 Phage predation reduced biofilm surface area across all systems by 48 h post -infection and 343 altered the structure of Massilia sp. and P. fluorescens communities. Despite continuous viral 344 pressure, Massilia sp. and P. fluorescens biofilms recovered but adopted architectures distinct 345 from both their early developmental states and uninfected controls. Massilia sp. biofilms 346 formed sparser microcolonies, whereas the structured fronts of P. fluorescens appeared 347 distorted despite unchanged flow conditions (Fig.5B, 96 hpi). These changes imply potential 348 trade-offs between biofilm formation and phage resistance in these two isolates. In contrast, 349 R. inusitata biofilms failed to recover under sustained jumbophage pressure, and surviving 350 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint cells frequently exhibited filamentation similar to infections caused by E. coli jumbophage 351 Sharanji55. Together, these findings demonstrate that phages can actively reshape biofilm 352 architecture and induce host-specific responses that dictate biofilm recovery under sustained 353 viral pressure. By extending microfluidic infection assays beyond model systems, these 354 experiments illustrate the tractability of environmentally derived phage -host pairs within 355 contemporary biofilm platforms and provides a framework for investigating virus -biofilm 356 dynamics in structured natural communities. 357 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 358 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint Fig.5 Biofilm forming capacity of environmental isolates and time -course phage infection in 359 microfluidics. A) Crystal violet absorbance A590 values for static biofilms grown in 96-well plates for 24 360 hours with three replicates (n = 3) per bacterial species. Species highlighted in bold were adopted for 361 microfluidic experiments as representative for their varying capacities to form biofilm (i.e. high, 362 moderate, and low). Error bars represent the SEM of triplicate A590 values. B) Brightfield images of 24 363 h-old biofilms established in microfluidic devices which were subsequently infected with 106 phages/mL 364 under continuous flow at 0.1 µL/min and monitored up to 96 hours post-infection. Massilia sp. biofilms 365 were infected with phage B343P1, R. inusitata with jumbophage B311P2. and P. fluorescens with phage 366 B508P1. Arrowheads in Massilia sp. biofilm highlights examples of microcolonies observed over the 367 experimental duration while arrowheads in R. inusitata community indicate cell filamentation under 368 sustained jumbophage pressure. 369 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint

370 Biofilms represent the dominant microbial lifestyle on Earth, forming highly structured and 371 ubiquitous communities that underpin essential ecosystem processes1. As obligate bacterial 372 parasites, phages have coexisted with biofilms over evolutionary timescales , and 373 metagenomic surveys consistently reveal strong coupling between viral and bacterial 374 community composition across diverse environments8,23,24,56. Despite the central role of 375 phages in driving microbial diversity, carbon cycling, and horizontal gene transfer57, our 376 understanding of phage-biofilm interactions remains disproportionately informed by model and 377 applied phage-host systems. Consequently, the diversity, biology, and ecological roles of 378 phages in natura l multispecies biofilms remain poorly understood. The ALP collection 379 addresses these gaps by providing a culturable, full -genome resource of novel , 380 environmentally derived phages that extend our current understanding of viral diversity and 381 traits beyond model and clinical biofilms. 382 383 Mountain stream ecosystems are structured and sustained by benthic biofilms that drive 384 primary production, respiration, and nutrient cycling3. Through these processes, they support 385 the functioning o f river networks that supply freshwater to meet the demands of millions 386 worldwide58. Moreover, stream biofilms are also sensitive indicators of environmental change 387 and are increasingly affected by climate warming6,29,31. Therefore, there is growing urgency to 388 understand how biofilms and their associated viral communities respond with accelerating 389 climatic pressures. By targeting native stream bacteria from benthic biofilms , the ALP 390 collection captures a fraction of viral diversity that is largely absent from existing phage and 391 genome repositories. Although derived from alpine stream systems, its ecological relevance 392 extends broadly as benthic biofilms are pervasive across aquatic ecosystems. 393 394 The limited representation of freshwater biofilm -associated phages in reference databases 395 and sequence archives underscores the novelty observed across the ALP collection. Only 5 396 of 28 isolates exhibited >50% genome coverage with previously described freshwater phages 397 (Table 1), while queries against the curated LOGAN database revealed sparse and low -398 coverage matches across geographically diverse environments (Fig.3). The latter also 399 indicates that potential relatives – or perhaps the ALP isolates themselves – are rarely 400 recovered as near -complete genomes in metagenomic datasets , reflecting persistent 401 technical challenges in viral metagenomic assemblies 59,60. In contrast , DNA from purified 402 isolates are more amenable to yield complete genomes thereby, providing greater confidence 403 in identifying features such as terminal repeats that may inform phage DNA packaging 404 strategies41. Moreover, the ALP isolates were sequenced and characterised with minimal 405 laboratory passage, providing a complementary perspective of genomic content that reflect 406 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint their recent life history in nature and minimising genetic drift caused by extended propagation 407 under laboratory conditions 61. In these contexts, the ALP collection not only expands 408 taxonomic representation but also offers high-resolution genomes for facilitating bioinformatic 409 analyses and ecological interpretation of biofilm-associated viruses in freshwaters. 410 411 Consistent with the global predominance of tailed dsDNA viruses 24,62, the ALP collection 412 comprises myo-, sipho-, and podovirus morphotypes supported by TEM imaging and genome 413 annotation (Figs.2B and 4A). Capsid-genome scaling of the ALP isolates also followed 414 allometric relationships consistent with icosahedral dsDNA phage s, where scaling models 415 reported proportional expansion in capsid volume with increasing viral genome sizes 40,63 416 (Fig.4B). This ensures that DNA packing density is maintained to preserve the internal capsid 417 pressure required for genome ejection 40,63, thereby reflecting the conserved infection 418 mechanobiology amongst ALP isolates. Consequently, the ALP collection also expands the 419 empirical basis of these capsid -genome scaling models 63 by contributing physical 420 measurements of virions derived from these novel isolates. Importantly, the conserved capsid-421 genome allometry also underscores the constraints of phages operating within environmental 422 biofilms. For instance, larger capsids are expected to reduce phage diffusivity in biofilm 423 matrices64, yet simultaneously accommodate larger genomes that may encode beneficial 424 auxiliary genes, such as those enhancing host metabolism to support viral replication. While 425 not directly tested here, this hints that biofilm-associated phages likely occupy a constrained 426 multidimensional trait space, maintaining internal capsid pressure while balancing gene 427 content, and diffusion through biofilms. Whether matrix-degrading depolymerases can partially 428 alleviate diffusional constraints also remains unclear. This framework can be extended to 429 stream environments where hydraulic forces (e.g. turbulent flow and shear) and oligotrophy 430 may further shape viral structural stability and auxiliary gene content. 431 432 The compact and densely packed nature of phage genomes impose s strong constraints on 433 gene content, which limits retention of genes beyond core replicatory functions65. Accordingly, 434 morons and AMGs were sparse among ALP isolates and were largely associated with host 435 takeover, metabolic reprogramming, and suppression of antiviral defences (Figs.4A and 4C). 436 This pattern aligns with the predominance of virulent phages in the collection, prioritising 437 genes that optimise the lytic cycle which typically occur at shorter timescales than lysogeny 438 (Fig.4C). In contrast, only a minor subset of AMGs was putatively linked to longer-term host 439 environmental resilience, notably metalloid resistance and oxidative stress mitigation. These 440 functions are congruent with overcoming stream biofilm-associated stressors, where EPS 441 matrices are known to concentrate heavy metals and resident cells experienc ing chronic 442 oxidative stress46,48. Within the context of virulent phages, these AMGs may enhance phage 443 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint production indirectly by temporarily sustaining host physiology under environmental stress 444 rather than promoting long-term host persistence. Nonetheless, the true functional diversity of 445 the ALP collection is likely underestimated given the overwhelming proportion of unknown 446 gene functions (Fig.4A), despite recent major strides with protein structure-based annotations 447 in revealing the functions of >50% of genes within an average phage genome66. 448 449 Beyond genomic features, the presence of multiple infection phenotypes especially amongst 450 phages infecting the same species (Fig.2B; Rhodoferax and Massilia phages), coupled with 451 host-specific structural responses to phage infection (Fig.5B) suggests that viral invasion of 452 biofilms is unlikely dictated by a dominant infection strategy. Instead, competing viral traits and 453 unique host-specific adaptations in biofilms may stabilise microbial and viral diversity 56 by 454 promoting localised negative frequency -dependent “kill-the-winner” dynamics and limiting 455 selective sweeps across the structured community. These effects are also likely amplified 456 within natural multispecies biofilms as several taxa are competing within a spatially 457 heterogenous niche, generating a mosaic of localised virus -microbe interactions where viral 458 suppression is tempered and bacterial adaptation is dependent on the microenvironment 67 459 and local phage pressure. In this context, environmental phages function not only as agents 460 of mortality but as active architects of biofilm structure (Fig.5B) and community composition, 461 promoting turnover and preserving functional resilience of environmental biofilm by 462 maintaining microbial community diversity. 463 464 In summary, the ALP collection captures substantial genotypic and phenotypic diversity among 465 biofilm-associated phages in streams. This curated set of 28 unique phage isolates expands 466 upon the viral “dark matter” that remains poorly represented in current culture collections and 467 databases, particularly among underexplored natural environments where biofilms are 468 pervasive. Crucially, the collection provides a foundation for necessary empirical 469 investigations into virus-microbe interactions across scales. For example, integrating the ALP 470 collection with synthetic communit ies68,69 and microfluidic platform s18,21 offers a tractable 471 framework to finely dissect viral -driven mechanisms in heterogeneous biofilms. These 472 experiments are also extendable to mesoscales such as flume models 70 or interconnected 473 fluidic systems of biofilm networks 71 to explore and validate emergent ecological properties 474 and evolutionary dynamics as phages propagate across broader biofilm landscapes. Beyond 475 experimental applications, the viral genomes offer valuable references for advancing virome 476 research and computational tools, with the prospect of uncovering environmental gene 477 functions that may provide the biotechnological innovation required to navigate our rapidly 478 changing ecosystems in a warming planet. 479 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint

480 Bacterial host and phage culture. All bacterial host isolates used in this study were 481 cultivated at ambient room temperature (~22°C) in 1´ Reasoner’s 2a (R2A) medium (Neogen, 482 USA) which was prepared following manufacturer’s protocol. Agar (Merck, Germany) was 483 added to a final concentration of 0.4% or 1.5% for soft-agar overlay assay and standard agar 484 media plates, respectively. Phages were propagated via broth or soft-agar medium depending 485 on the viral isolate. Broth amplification was started with a 1:50 dilution of an overnight host 486 culture in 10 mL of fresh media broth followed by spiking with 20 µL of phage lysate (minimum 487 ~106 PFU/mL). The mixture is then incubated at room temperature with agitation overnight or 488 until complete lysis i.e. broth clearance. The amplification reaction is centrifuged at 5000 ´g 489 for 10 minutes to pellet bacterial debris and the viral supernatant is sterilised with 0.45 µm 490 syringe filters (Sartorius, Germany). Amplification via soft-agar medium begins with mixing 500 491 µL of overnight host culture with 50 µL of serially-diluted phage lysate and 2 mL of 0.4% R2A 492 agar. The 0.4% mixture is casted on 1.5% agar media to form a soft-agar overlay and the plate 493 is incubated at room temperature overnight. Plates of phage serial dilutions exhibiting near 494 complete lysis are then scraped with 5 – 10 mL of sterile 1´ PBS (Invitrogen, Lithuania) and 495 transferred to a 50 mL Falcon tube. The scrapings were vortexed vigorously followed by 496 centrifugation at 5000 ´g for 10 minutes to pellet agar debris and the viral supernatant was 497 sterilised with 0.45 µm syringe filters. The phage lysates were stored at 4 °C for short -term 498 purposes and at -80°C in 20% glycerol final concentration for long-term storage. 499 500 Phage isolation from stream water. The water column in lotic systems functions as a 501 dynamic conduit that integrates viral production from upstream biofilms through downstream 502 dispersal, where waterborne phages represent a subset of mobile and infectious viruses 503 dispersing across biofilm-associated hosts. Hence, concentrating viral particles from the water 504 column enables scalable and reproducible recovery of viable phages. 120 L of fresh water 505 was collected from a stream confluence located at Champéry, Switzerland that was fed by 506 both groundwater-fed and glacier-fed streams (Fig.S1: 46.164 N, 6.8610 E; 1051 m above sea 507 level). Water samples were immediately transported back to our facility and concentrated via 508 tangential flow using a hollow fibre cartridge with 1.15 m 2, 100 000 kDa membrane ( GE 509 Healthcare, USA ) to a final volume 2 L which was filtered via 0.45 µm Sterivex (Merck, 510 Germany) to exclude eukaryotic and most prokaryotic organisms but not viruses. The filtered 511 water was further concentrated to 500 mL using a tangential flow cassette with 100 000 kDa 512 (Sartorius, Germany). The concentrated water is then screened for phages against 40 513 previously isolated bacterial host strains via soft -agar overlay by mixing 500 µL of overnight 514 host culture with 1 mL of concentrated water and 2 mL of 0.4% agar before casting on 1.5% 515 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint agar media plates. Phage plaques identified were cored with sterile wide-bore pipette tips and 516 resuspended vigorously in 100 µL of 1 ´ PBS. Phages in PBS were then serially diluted and 517 plated using soft agar overlay to perform a subsequent round of coring resulting in double -518 purified phage isolates. Purified phages were amplified in broth or soft-agar media to generate 519 high viral titres (108 – 1010 PFU/mL) for downstream analyses. 520 521 Transmission electron microscopy and plaque morphology imaging. TEM was 522 performed with negative stain. Each sample was adsorbed on a glow -discharged carbon-523 coated copper grid 400 mesh (EMS, Hatfield, PA, USA) washed with deionized water and 524 stained with Uranyl Acetate 1% for 30 seconds. Observations was made using a Talos L120C 525 electron microscope (Thermo Fisher, Hillsboro, USA) operated at 120 kV. Digital images were 526 collected using a CMOS camera Ceta -S (Thermo Fisher, Hillsboro, USA) 4098 ´ 4098 527 pixels.), using a defocus range between -1.5 µm and -2.5 µm. Images of plaque morphologies 528 were obtained on serially -diluted pure phage cultures on 0.4% soft -agar with a Scan 30 0 529 automatic colony counter (Interscience, France). 530 531 Phage DNA extraction and long -read sequencing. 1 mL of phage lysate at 10 8 PFU/mL 532 minimum, were initially treated with 10 U of DNase ( Thermo Scientific, USA) with 1´ DNase 533 buffer (diluted from 10´ stock comprising of 0.1 M Tris pH 7.5, 1 mM CaCl2 and 25 mM MgCl2) 534 for 2 h at 37 °C to eliminate bacterial DNA within the lysate. DNase was then inactivated by 535 incubating at 75°C for 30 minutes. Removal of bacterial DNA was verified via 16S rRNA PCR 536 (V3/V4 primers 341F: 5’ -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCT 537 ACGGGNGGCWGCAG-3’ and 785R: 5’- GTC TCGTGGGCTCGGAGATGTGTATAAGAGACA 538 GGACTACHVGGGTATCTAATCC-3’) with the following cycling conditions: initial denaturation 539 at 95°C for 3 minutes, 30 cycles at 95 °C for 30 seconds, 55 °C for 30 seconds, 72°C for 30 540 seconds, 5-minute final extension using 1 µL of template; and subsequently, gel 541 electrophoresis with a ~550 bp product if PCR-positive for 16S rRNA. PCR negative lysates 542 were extracted for phage DNA using two methods. For most phage isolates, Norgen Biotek 543 column-based phage extraction kit (Norgen Biotek, Canada) was used following 544 manufacturer’s protocol including 20 µL Proteinase K (PanReac AppliChem, Switzerland) 545 treatment and an extended incubation at 55°C for 1 h during viral lysis step. However, isolates 546 that are recalcitrant to column -based extraction were extracted following a modified DNA 547 isopropanol precipitation protocol. Here, 10% SDS (Fisher Scientific, UK) was added to the 548 phage lysate at a final concentration of 0.5%, mixed well, and allowed to incubate at ambient 549 room temperature for 30 minutes. Sodium acetate ( Sigma-Aldrich, USA) solution was then 550 added to the lysate to a final concentration of 0.3 M followed by 0.6 – 0.7 volume of ice-cold 551 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint isopropanol (Merck, Germany). The mixture was mixed by inversion and immediately 552 centrifuged at 15 000 ´g at 4°C for 30 minutes. The supernatant was discarded, and the DNA 553 pellet was washed with 500 µL of freshly prepared 70% ethanol (Fisher Scientific, UK) and 554 pelleted at 15 000 ´g at 4°C for 30 minutes, twice. Once residual ethanol from the final wash 555 was dried, the pellets were resuspended in 50 µL TE buffer overnight at 4°C for complete DNA 556 resuspension. Extracted DNA concentration, quality, and integrity was assessed using Qubit® 557 Fluorometer with 1 ´ dsDNA High Sensitivity kit ( Invitrogen, USA ), Nanodrop® ( Scientific, 558 USA), and Genomic TapeStation® (Agilent, USA) following manufacturers’ protocols, 559 respectively. DNA library preparation and MinION long-read sequencing were performed using 560 Native Barcoding Kit 24 V14 (Oxford Nanopore Technologies, UK) following manufacturer 561 protocol, using short fragment buffer to maximise read preservation at final wash step prior to 562 eluting the genomic library for sequencing. When long-read data were insufficient to assemble 563 complete phage genomes, DNA sequencing was performed on the Illumina MiSeq i100 564 platform using the TruSeq™ DNA PCR-Free Library Prep Kit (Illumina, USA) with a random 565 fragmentation (Covaris E220 focused ultrasonicator). We caution that certain phage isolates 566 especially those infecting P. cyclaminis, remain notoriously recalcitrant to DNA extraction while 567 the Sphingomonas phage isolate was limited to low amplification titre of <104 PFU/mL despite 568 our best efforts, which not only hampered DNA extraction but also transmission electron 569 microscopy. 570 571 Phage genome assembly, annotation, and bioinformatic analyses . Long-reads were 572 base-called and demultiplexed using Dorado (v0.9.1) with high-accuracy model and remaining 573 adapters and barcodes were removed using Porechop (v0.2.4). Any reads mapping to the 574 bacterial host genome were also removed. Sequenced reads were assembled into contigs 575 with Flye (v2.9.5) when using long -reads only and with Unicycler (v0.5.1) when performing 576 hybrid assembly. Contigs were quality -checked to verify their viral origins using VIBRANT 577 (v1.2.1) and geNomad (v1.9.0). Where possible, we also derived taxonomic information of our 578 completed genomes with geNomad to assess the novelty of our viral collection against publicly 579 available databases. We then complemented CheckV (v1.0.3) with an in-house read-mapping 580 visualiser theBIGbam (v0.1; unpublished) to evaluate phage genome completeness based on 581 nucleotide sequence and read mapping features, respectively. We dereplicated our collection 582 of complete phage genomes via nucleotide identity using scripts provided by CheckV with 583 85% coverage and 95% nucleotide identity thresholds. Phage contigs assembled as head-to-584 tail concatemers were identified as overcomplete genomes, and genome lengths were 585 automatically corrected by retaining a single -repeat unit corresponding to the expected 586 complete phage genome. Genes were predicted using PHANOTATE (v1.5.0) and annotated 587 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint with a combination of pharokka (v1.8.2) and phold (v1.1.0) using sequence and structural 588 homology, respectively. DNA packaging mechanisms were also determined with theBIGbam 589 using the logic from PhageTerm33. Nucleotide BLAST to NCBI database was performed with 590 default settings using the dereplicated genome sequences as the query sequence. 591 Meanwhile, queries to planetary -scale LOGAN contig database were initiated by randomly 592 sampling 1 kb segments from the genomes of each dereplicated phage isolate. The 1 kb query 593 sequence was then interrogated against all available contigs in LOGAN with the threshold 594 setting set to the lowest value at 0.25 k-mer coverage (query date 01-Dec-2025). Cumulative 595 k-mer coverage scores were calculated by summing the individual coverage values per 596 source. All data was analysed and plotted with R (v.4.2.0). 597 598 Crystal violet static biofilm assay. Bacterial overnight cultures were adjusted to OD600 ~0.02 599 with fresh media and 190 µL was aliquoted into each well of a 96-well flat-bottom polystyrene 600 plate (Greiner Bio-One, Switzerland). Biofilms were then grown statically in triplicate for each 601 bacterial species for 24 h. Following static incubation, planktonic cells in the supernatant were 602 discarded by carefully inverting the plate into a waste basin and the biofilms were wash thrice 603 with MilliQ water (Merck Millipore, Germany) by submerging the plate in a separate filled basin 604 and discarding the water in waste. The 96 -well plate was then inverted on a paper towel to 605 remove excess liquid and is allowed to air-dry. Crystal violet staining was performed by adding 606 200 µL of 0.01% Crystal violet solution (Sigma-Aldrich, USA) in each well and stained at 607 ambient temperature with low-speed orbital shaking for 30 minutes. Excess Crystal violet was 608 discarded, and the stained biofilms were washed thrice with MilliQ water and allowed to air -609 dry. 200 µL of 70% ethanol (Fisher Scientific, UK) was then added to each well and incubated 610 at ambient room temperature for 30 minutes with low orbital shaking to solubilise the Crystal 611 violet. Absorbance values at 590 nm were obtained (A 590) using Biotek Synergy H1 plate 612 reader (Agilent, USA) and the datapoints were blanked with A590 values from negative control 613 wells containing sterile media. 614 615 Microfluidic biofilm time -course phage infection. Polydimethylsiloxane (PDMS) 616 microfluidic devices with channel dimension of 40 µm height, 250 µm width, and 1 mm length 617 were purchased (Wunderlichips, Switzerland) and sterilised with 70% ethanol followed by UV 618 radiation for 30 minutes. Microfluidic biofilms were established by first seeding the channels 619 with 10 µL of overnight bacterial cultures adjusted to OD 600 ~0.1 in fresh media. Cells were 620 allowed to attach under static conditions for 2 h at ambient room temperature. Devices were 621 then connected a media-filled syringe via autoclave-sterile PTFE tubing (ID: 0.56 mm and OD: 622 1.07 mm, Fisher Scientific, Switzerland) with the outlet emptying into a waste Eppendorf tube. 623 Media was infused continuously with a syringe pump (New Era Pump Systems, USA) at 0.1 624 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint µL/min flow rate and biofilms were grown for 2 h at ambient room temperature. Following 625 overnight growth, sterile media syringes were swapped for 10 6 PFU/mL of phages diluted in 626 sterile media and phages were infused continuously at 0.1 µL/min flow rate for a further 96 h. 627 Control biofilms not infected with phages were maintained with sterile media infusions. Biofilm 628 development was then monitored over the 96-h duration under default brightfield settings via 629 Zeiss Axio Zoom.V16 microscope with the Plan-NEOFLUAR Z 1.0 ´/0.25 FWD 56 mm 630

and 112 ´ zoom magnification (Carl Zeiss Microscopy, Germany) . Images were 631 acquired and processed with Zen (blue edition) software (Carl Zeiss Microscopy, Germany). 632 633 Data, code, and phage culture availability. All code used for long-read preprocessing, 634 assembly and genome annotation were integrated in a Snakemake (v7.32.4) pipeline 635 available at: https://github.com/bhagavadgitadu22/PhageID. Sequencing reads and 636 assembled genomes will be deposited in NCBI under BioProject: PRJNA1423127 following 637 manuscript review. Purified phage cultures will also be submitted to CRBIP Biological 638 Resource Center of Insitut Pasteur following manuscript acceptance. All plots were generated 639 with R and raw data are available via supplementary data tables. 640 641 Acknowledgements. We thank Martina Gonzalez and David Touchette for sharing their 642 bacterial culture collection from streambed biofilms as bait for our phage isolation efforts. We 643 also thank Florence Jagorel and Marc Monot from the group PhagoMics, Phages.fr, for their 644 advice and support, and the Biomics Platform, C2RT, Institut Pasteur, Paris, France, supported 645 by France Génomique (ANR -10-INBS-09) and IBISA. We finally thank Tom Battin for his 646 guidance and provision of laboratory facility in support of this work. This work was supported 647 by the Swiss National Science Foundation under grant no.212726 awarded to Hannes Peter. 648 649 Author contributions . W.H.C and H.P. conceived the study, and W.H.C designed 650 experiments and interpreted data. W.H.C, M.B., and A.H. carried out field expeditions, sample 651 processing, and phage isolations. W.H.C and A.H. purified phages and maintained the phage 652 collection throughout the study. W.H.C, A.H., and F.B. performed phage genome extractions 653 and sequencing. M.B. established bioinformatic pipelines and analysed genomes . W.H.C 654 performed experiments, viral database screens, and data analyses. D.D. performed electron 655 microscopy imaging. W.H.C prepared figures. W.H.C. and H.P. wrote the manuscript. H.P. 656 supervised the study. All authors reviewed, edited, and commented on the manuscript. 657 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint

658 1. Flemming, H. C. & Wuertz, S. Bacteria and archaea on Earth and their abundance in 659 biofilms. Nature Reviews Microbiology 2019 17:4 17, 247–260 (2019). 660 2. Flemming, H. C. et al. Biofilms: an emergent form of bacterial life. Nature Reviews 661 Microbiology 2016 14:9 14, 563–575 (2016). 662 3. Battin, T. J., Besemer, K., Bengtsson, M. M., Romani, A. M. & Packmann, A. I. The 663 ecology and biogeochemistry of stream biofilms. Nature Reviews Microbiology 2016 664 14:4 14, 251–263 (2016). 665 4. Thorpe, A. C. et al. National-scale biogeography and function of river and stream 666 bacterial biofilm communities. Nature Communications 2025 16:1 16, 10571- (2025). 667 5. Michoud, G. et al. Unexpected functional diversity of stream biofilms within and across 668 proglacial floodplains despite close spatial proximity. Limnol. Oceanogr. 68, 2183–669 2194 (2023). 670 6. Bourquin, M. et al. Predicting climate-change impacts on the global glacier-fed stream 671 microbiome. Nature Communications 2025 16:1 16, 1264- (2025). 672 7. Hall-Stoodley, L., Costerton, J. W. & Stoodley, P. Bacterial biofilms: from the Natural 673 environment to infectious diseases. Nature Reviews Microbiology 2004 2:2 2, 95–108 674 (2004). 675 8. Chevallereau, A., Pons, B. J., van Houte, S. & Westra, E. R. Interactions between 676 bacterial and phage communities in natural environments. Nature Reviews 677 Microbiology 2021 20:1 20, 49–62 (2021). 678 9. Hu, C. et al. To kill or to piggyback: Switching of viral lysis-lysogeny strategies 679 depending on host dynamics. Science of The Total Environment 959, 178233 (2025). 680 10. Piel, D. et al. Phage–host coevolution in natural populations. Nature Microbiology 681 2022 7:7 7, 1075–1086 (2022). 682 11. Borin, J. M. et al. Rapid bacteria-phage coevolution drives the emergence of 683 multiscale networks. Science (1979). 382, 674–678 (2023). 684 12. Shiah, F. K. et al. Viral shunt in tropical oligotrophic ocean. Sci. Adv. 8, 2829 (2022). 685 13. Gilbert, N. E. et al. Seasonal enhancement of the viral shunt catalyzes a subsurface 686 oxygen maximum in the Sargasso Sea. Nature Communications 2025 17:1 17, 352- 687 (2025). 688 14. Rosenwasser, S., Ziv, C., Creveld, S. G. van & Vardi, A. Virocell Metabolism: 689 Metabolic Innovations During Host–Virus Interactions in the Ocean. Trends Microbiol. 690 24, 821–832 (2016). 691 15. Howard-Varona, C. et al. Phage-specific metabolic reprogramming of virocells. The 692 ISME Journal 2020 14:4 14, 881–895 (2020). 693 16. Pires, D. P., Melo, L. D. R. & Azeredo, J. Understanding the Complex Phage-Host 694 Interactions in Biofilm Communities. Annu. Rev. Virol. 8, 73–94 (2021). 695 17. Vidakovic, L., Singh, P. K., Hartmann, R., Nadell, C. D. & Drescher, K. Dynamic 696 biofilm architecture confers individual and collective mechanisms of viral protection. 697 Nature Microbiology 2017 3:1 3, 26–31 (2017). 698 18. Winans, J. B., Zeng, L. & Nadell, C. D. Spatial propagation of temperate phages 699 within and among biofilms. Proc. Natl. Acad. Sci. U. S. A. 122, e2417058122 (2025). 700 19. Prentice, J. A., van de Weerd, R. & Bridges, A. A. Cell-lysis sensing drives biofilm 701 formation in Vibrio cholerae. Nature Communications 2024 15:1 15, 2018- (2024). 702 20. Vaidya, S. et al. Bacteria use exogenous peptidoglycan as a danger signal to trigger 703 biofilm formation. Nature Microbiology 2025 10:1 10, 144–157 (2025). 704 21. Winans, J. B., Wucher, B. R. & Nadell, C. D. Multispecies biofilm architecture 705 determines bacterial exposure to phages. PLoS Biol. 20, e3001913 (2022). 706 22. Gliźniewicz, M. et al. Advances in bacteriophage-mediated strategies for combating 707 polymicrobial biofilms. Front. Microbiol. 14, 1320345 (2023). 708 23. Bekliz, M., Pramateftaki, P., Battin, T. J. & Peter, H. Viral diversity is linked to bacterial 709 community composition in alpine stream biofilms. ISME Communications 2, (2022). 710 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 24. Michoud, G. et al. Mapping the metagenomic diversity of the multi-kingdom glacier-fed 711 stream microbiome. Nature Microbiology 2024 10:1 10, 217–230 (2025). 712 25. Resch, G. et al. Between Centralization and Fragmentation: The Past, Present, and 713 Future of Phage Collections. PHAGE: Therapy, Applications, and Research 5, 22 714 (2024). 715 26. Russell, D. A. & Hatfull, G. F. PhagesDB: the actinobacteriophage database. 716 Bioinformatics 33, 784–786 (2017). 717 27. Maffei, E. et al. Systematic exploration of Escherichia coli phage–host interactions 718 with the BASEL phage collection. PLoS Biol. 19, e3001424 (2021). 719 28. Fletcher, J. et al. The Citizen Phage Library: Rapid Isolation of Phages for the 720 Treatment of Antibiotic Resistant Infections in the UK. Microorganisms 12, 253 (2024). 721 29. Touchette, D. et al. Experimental evidence on the impact of climate-induced 722 hydrological and thermal variations on glacier-fed stream biofilms. FEMS Microbiol. 723 Ecol. 101, 163 (2025). 724 30. Ezzat, L. et al. Diversity and biogeography of the bacterial microbiome in glacier-fed 725 streams. Nature 2025 637:8046 637, 622–630 (2025). 726 31. Kohler, T. J. et al. Global emergent responses of stream microbial metabolism to 727 glacier shrinkage. Nature Geoscience 2024 17:4 17, 309–315 (2024). 728 32. Harding, K. R., Kyte, N. & Fineran, P. C. Jumbo phages. Current Biology 33, R750–729 R751 (2023). 730 33. Valdez, A., Sun, H., Weiss, H. H. & Aranson, I. Biomechanical modeling of 731 spatiotemporal bacteria-phage competition. Communications Physics 2025 8:1 8, 732 139- (2025). 733 34. Knecht, L. E., Veljkovic, M. & Fieseler, L. Diversity and Function of Phage Encoded 734 Depolymerases. Front. Microbiol. 10, 502215 (2020). 735 35. Oliveira, H. et al. Ability of phages to infect Acinetobacter calcoaceticus-Acinetobacter 736 baumannii complex species through acquisition of different pectate lyase 737 depolymerase domains. Environ. Microbiol. 19, 5060–5077 (2017). 738 36. Qin, K. et al. Isolation and characterization of wetland VSW-3, a novel lytic cold-active 739 bacteriophage of Pseudomonas fluorescens. Can. J. Microbiol. 63, 110–118 (2017). 740 37. Gittrich, M. R. et al. Isolation and characterization of 24 phages infecting the plant 741 growth-promoting rhizobacterium Klebsiella sp. M5al. PLoS One 20, e0313947 742 (2025). 743 38. Qin, X. et al. Psychrophilic phage phiGM22-3 efficiently controls Pseudomonas 744 fluorescens contamination in cold-stored milk. Int. J. Food Microbiol. 411, 110525 745 (2024). 746 39. Chikhi, R. et al. Logan: Planetary-Scale Genome Assembly Surveys Life’s Diversity. 747 bioRxiv 2024.07.30.605881 (2025) doi:10.1101/2024.07.30.605881. 748 40. Chaudhari, H. V., Inamdar, M. M. & Kondabagil, K. Scaling relation between genome 749 length and particle size of viruses provides insights into viral life history. iScience 24, 750 102452 (2021). 751 41. Garneau, J. R., Depardieu, F., Fortier, L. C., Bikard, D. & Monot, M. PhageTerm: a tool 752 for fast and accurate determination of phage termini and packaging mechanism using 753 next-generation sequencing data. Scientific Reports 2017 7:1 7, 8292- (2017). 754 42. Yang, J. Y. et al. Degradation of host translational machinery drives tRNA acquisition 755 in viruses. Cell Syst. 12, 771-779.e5 (2021). 756 43. Encina-Robles, J., Pérez-Villalobos, V. & Bustamante, P. The HicAB System: 757 Characteristics and Biological Roles of an Underappreciated Toxin-Antitoxin System. 758 International Journal of Molecular Sciences 2024, Vol. 25, Page 12165 25, 12165 759 (2024). 760 44. Martinez, S. & Hausinger, R. P. Catalytic Mechanisms of Fe(II)- and 2-Oxoglutarate-761 dependent Oxygenases. Journal of Biological Chemistry 290, 20702–20711 (2015). 762 45. Mahler, M. et al. Phage arabinosyl-hydroxy-cytosine DNA modifications result in 763 distinct evasion and sensitivity responses to phage defense systems. Cell Host 764 Microbe 33, 1173-1190.e9 (2025). 765 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 46. Richard, H., Carpenter, E. J., Komada, T., Palmer, P. T. & Rochman, C. M. Biofilm 766 facilitates metal accumulation onto microplastics in estuarine waters. Science of The 767 Total Environment 683, 600–608 (2019). 768 47. Goff, J. L. et al. Tellurite Adsorption onto Bacterial Surfaces. Environ. Sci. Technol. 55, 769 10378–10386 (2021). 770 48. Boles, B. R. & Singh, P. K. Endogenous oxidative stress produces diversity and 771 adaptability in biofilm communities. Proc. Natl. Acad. Sci. U. S. A. 105, 12503–12508 772 (2008). 773 49. Chiang, B. Y. et al. Protein S-thiolation by glutathionylspermidine (Gsp): The role of 774 Escherichia coli Gsp synthetase/amidase in redox regulation. Journal of Biological 775 Chemistry 285, 25345–25353 (2010). 776 50. Hutinet, G. et al. 7-Deazaguanine modifications protect phage DNA from host 777 restriction systems. Nature Communications 2019 10:1 10, 5442- (2019). 778 51. Díaz-Rullo, J. & González-Pastor, J. E. tRNA queuosine modification is involved in 779 biofilm formation and virulence in bacteria. Nucleic Acids Res. 51, 9821–9837 (2023). 780 52. Bailey, Z. M., Igler, C. & Wendling, C. C. Prophage maintenance is determined by 781 environment-dependent selective sweeps rather than mutational availability. Current 782 Biology 34, 1739-1749.e7 (2024). 783 53. Simmons, E. L. et al. Biofilm Structure Promotes Coexistence of Phage-Resistant and 784 Phage-Susceptible Bacteria. mSystems 5, (2020). 785 54. Bond, M. C., Vidakovic, L., Singh, P. K., Drescher, K. & Nadell, C. D. Matrix-trapped 786 viruses can prevent invasion of bacterial biofilms by colonizing cells. Elife 10, e65355 787 (2021). 788 55. Magar, S. et al. Characterization and genomic analysis of Sharanji: a jumbo 789 bacteriophage of Escherichia coli. Virol. J. 22, 67 (2025). 790 56. Peter, H., Michoud, G., Busi, S. B. & Battin, T. J. The role of phages for microdiverse 791 bacterial communities in proglacial stream biofilms. Frontiers in Microbiomes 2, 792 1279550 (2023). 793 57. Breitbart, M., Bonnain, C., Malki, K. & Sawaya, N. A. Phage puppet masters of the 794 marine microbial realm. Nature Microbiology 2018 3:7 3, 754–766 (2018). 795 58. Immerzeel, W. W. et al. Importance and vulnerability of the world’s water towers. 796 Nature 2019 577:7790 577, 364–369 (2019). 797 59. Rose, R., Constantinides, B., Tapinos, A., Robertson, D. L. & Prosperi, M. Challenges 798 in the analysis of viral metagenomes. Virus Evol. 2, (2016). 799 60. Bekliz, M., Brandani, J., Bourquin, M., Battin, T. J. & Peter, H. Benchmarking protocols 800 for the metagenomic analysis of stream biofilm viromes. PeerJ 2019, e8187 (2019). 801 61. Subedi, D. & Barr, J. J. Temporal Stability and Genetic Diversity of 48-Year-Old T-802 Series Phages. mSystems 6, (2021). 803 62. Zhong, Z. P. et al. Glacier-preserved Tibetan Plateau viral community probably linked 804 to warm–cold climate variations. Nature Geoscience 2024 17:9 17, 912–919 (2024). 805 63. Luque, A., Benler, S., Lee, D. Y., Brown, C. & White, S. The Missing Tailed Phages: 806 Prediction of Small Capsid Candidates. Microorganisms 2020, Vol. 8, Page 1944 8, 807 1944 (2020). 808 64. Dunsing, V., Irmscher, T., Barbirz, S. & Chiantia, S. Purely Polysaccharide-Based 809 Biofilm Matrix Provides Size-Selective Diffusion Barriers for Nanoparticles and 810 Bacteriophages. Biomacromolecules 20, 3842–3854 (2019). 811 65. Comeau, A. M., Bertrand, C., Letarov, A., Tétart, F. & Krisch, H. M. Modular 812 architecture of the T4 phage superfamily: A conserved core genome and a plastic 813 periphery. Virology 362, 384–396 (2007). 814 66. Bouras, G. et al. Protein structure-informed bacteriophage genome annotation with 815 Phold. Nucleic Acids Res. 54, (2026). 816 67. Jo, J., Price-Whelan, A. & Dietrich, L. E. P. Gradients and consequences of 817 heterogeneity in biofilms. Nature Reviews Microbiology 2022 20:10 20, 593–607 818 (2022). 819 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint 68. Wilde, J. et al. Assessing phage-host population dynamics by reintroducing virulent 820 viruses to synthetic microbiomes. Cell Host Microbe 32, 768-778.e9 (2024). 821 69. Castledine, M., Padfield, D., Schoeman, M., Berry, A. & Buckling, A. Bacteria–phage 822 (co)evolution is constrained in a synthetic community across multiple bacteria–phage 823 pairs. Microbiology (United Kingdom) 171, 001577 (2025). 824 70. Depetris, A. et al. Biophysical properties at patch scale shape the metabolism of 825 biofilm landscapes. npj Biofilms and Microbiomes 2022 8:1 8, 5- (2022). 826 71. Carrara, F., Altermatt, F., Rodriguez-Iturbe, I. & Rinaldo, A. Dendritic connectivity 827 controls biodiversity patterns in experimental metacommunities. Proc. Natl. Acad. Sci. 828 U. S. A. 109, 5761–5766 (2012). 829 830 .CC-BY 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted March 26, 2026. ; https://doi.org/10.64898/2026.03.26.713887doi: bioRxiv preprint