Imminent invasion of the chytrid fungus threatens the last naïve amphibian biodiversity hotspots

Amphibian extinction crisis; Batrachochytrium dendrobatidis; Disease-free 47 refuge; Pathogeography 48 49 This PDF file includes: 50 Main Text 51 Figures 1 to 4 52 53 54 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 3

55 While the amphibian chytrid fungus Batrachochytrium dendrobatidis (Bd) is driving catastrophic 56 biodiversity loss worldwide, some amphibian communities persist seemingly unaffected despite 57 occurring in climates conducive to pathogen establishment. These amphibian communities may 58 remain epidemiologically naïve. As mitigation of Bd is rarely successful after establishment, 59 identifying remaining Bd-free refuges is imperative. Presently, the only known large-scale Bd-free 60 refuge is the island of New Guinea (NG), safeguarding Australasia’s amphibian phylogenetic 61 diversity otherwise devastated by Bd. Following extensive multi-year disease surveillance, we 62 here uncover a second large-scale Bd-free refuge in the Sierra Nevada de Santa Marta (SNSM), 63 a Neotropical biodiversity hotspot in northern Colombia. We detected no evidence of Bd in 64 SNSM-wide screening, while we uncovered the presence of hypervirulent Bd-GPL in adjacent 65 areas of the tropical Andes. Population genomic analyses in an SNSM-endemic anuran found no 66 evidence for demographic bottlenecks indicative of cryptic epizootic decline. Niche modelling 67 highlights the high risk for Bd establishment and Bd-induced declines in both the SNSM and NG, 68 and the important role of lowland environmental barriers in restricting Bd invasion. Infection trials 69 using three SNSM-endemic amphibians reveal varying disease susceptibility. Together, these 70 data identify the SNSM as an epidemiologically naïve refuge likely facing imminent Bd invasion, 71 which could result in the loss of at least 25 endemic amphibian species. We highlight the urgent 72 need for proactive conservation action and strict implementation of biosecurity to safeguard the 73 unique and vast amphibian diversity of the world's last major Bd-free refuges. 74 Significance Statement 75 Amphibian chytridiomycosis caused by Batrachochytrium dendrobatidis (Bd) has driven 76 unprecedented global biodiversity loss. The Neotropics and Australasia comprise epicenters of 77 declines. Identifying remaining Bd-free refuges is crucial to curb further amphibian extinctions, but 78 so far contemporary absence of Bd has only been demonstrated for New Guinea. Here we 79 identify the last known major Bd-free biodiversity hotspot in the Neotropics: the Sierra Nevada de 80 Santa Marta (SNSM) in Colombia. Our results show that amphibian communities in this hotspot 81 are immunologically naïve despite occurrence of hypervirulent Bd lineages nearby and climatic 82 conditions within the SNSM conducive to Bd-induced declines. This creates an imminent risk for 83 Bd-driven declines and highlights the urgent need for preventive actions to avert another wave of 84 biodiversity loss. 85 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 4 Main Text 86

87 The Anthropocene is characterized by the global emergence of infectious diseases, posing major 88 challenges to human and wildlife health (1-3). Amphibian chytridiomycosis, caused by the fungal 89 pathogen Batrachochytrium dendrobatidis (Bd), is considered the most devastating wildlife 90 disease ever recorded (2, 4). Since the 1980s, it has resulted in catastrophic amphibian declines 91 and extinctions worldwide (5, 6), further accelerating the amphibian extinction crisis (7). While the 92 pathogen has been detected in virtually all regions surveyed, Bd-induced declines are distributed 93 unequally across the globe (6, 8). Humid montane environments have emerged as prime 94 landscapes for Bd epizootics (8, 9). Neotropical and Australian highlands stand out as epicenters 95 in terms of the number of species affected and the spatial extent of community-wide declines 96 (10). In both regions, Bd advanced across the landscape in a wavelike pattern from one or 97 multiple introductions (Fig. 1A, B), with spatiotemporally structured near-extinction of once 98 diverse and abundant amphibian communities (11-13). 99 Amidst the epicenters of global amphibian declines, some communities persist seemingly 100 unaffected (14-17). In regions providing environmental conditions that favor Bd-driven amphibian 101 declines, the factors allowing host community survival remain enigmatic. Elevated tolerance or 102 resistance might drive population survival (18) but appear unlikely to explain persistence of intact 103 multi-host communities. Instead, local absence of hypervirulent Bd lineages within Bd-infested 104 regions could explain amphibian community survival (15, 17). Geographic or environmental 105 barriers might sustain localized pathogen absence long term (17). Such Bd-free refuges play a 106 crucial role in averting future disease-driven biodiversity loss and potentially saving extant yet 107 potentially highly susceptible species. Mitigating the impact of emerging infectious disease is 108 most successful in a pre-invasion stage (19-21), thus the timely identification of disease-free 109 refuges is critical to safeguard biodiversity. 110 The only major Bd-free refuge that has so far been uncovered is the island of New Guinea (NG) 111 (15, 16, 22). NG is an amphibian hotspot, containing over 6% of the world’s amphibian species on 112 less than 1% of its landmass (15). Many species are phylogenetically related to the Australian 113 anuran lineages that have suffered extreme Bd-induced declines and extinctions (16). NG thus 114 safeguards a substantial portion of Australasia's evolutionary amphibian diversity (23). In the 115 Neotropical realm, major disease-free refuges have not yet been identified. However, severe 116 declines of amphibian populations throughout the region sharply contrasts with the seemingly 117 unaffected, abundant amphibian communities of the Sierra Nevada de Santa Marta (SNSM) in 118 northern Colombia (14, 24-26). The SNSM is an isolated coastal massif reaching 5,710 meters 119 above sea level and a hotspot of amphibian richness and endemism, harboring at least 25 120 endemic species, including several endemic genera (27-30, SI Appendix, Tables S1, S2). 121 Climatic conditions and community composition closely mirror other Neotropical montane 122 systems that have experienced severe Bd-driven declines, and phylogenetically close Andean 123 relatives of SNSM endemics have collapsed elsewhere (25, 26, 31). 124 Here, we evaluate whether the SNSM represents a Bd-free refuge. We predict (i) pre-epizootic 125 amphibian host densities across species, (ii) presence of hypervirulent Bd lineages in adjacent 126 mountain ranges, (iii) genome-wide SNP data of a focal species to indicate stable, diverse 127 populations, (iv) environmental conditions conducive to establishment of Bd and Bd-driven host 128 declines, and (v) host susceptibility to chytridiomycosis. To test these predictions and assess the 129 vulnerability of amphibians of the SNSM, we combine mountain-range wide Bd surveillance, Bd 130 lineage genotyping, host population genomics, pathogen niche modelling, and experimental 131 infections. Further, we extend our niche modelling of the climatic envelope associated with 132 Bd/i1 driven declines to New Guinea, allowing us to evaluate the risk of community/i1 wide declines 133 should Bd be introduced. Finally, we outline a series of actions to avert yet another wave of 134 disease-driven amphibian declines and extinctions. 135 136 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 5

137 Landscape-scale screening uncovers the last major pathogen-free Neotropical biodiversity 138 hotspot. Using a combination of quantitative PCR and CRISPR diagnostics on 2,212 amphibian 139 skin swabs from at least 33 species and 27 environmental DNA (eDNA) samples, we were unable 140 to detect Bd in any of the 34 sampled localities across the SNSM and adjacent lowlands (Fig. 1C; 141 see SI Appendix, section 1, for a discussion of previous screening data). Bayesian hierarchical 142 modelling yielded a posterior mean probability of 0.0297 per site that Bd was present but 143 overlooked (95% highest posterior density 2.6e⁻/i1 –0.0872, when assuming a test sensitivity of 144 0.75). Locality-wise posterior probabilities of infection were uniformly low across all sites (SI 145 Appendix, Fig. S1). In the closest Andean mountain range to the SNSM, the Serranía de Perijá 146 (SP) 20 km E of the SNSM, we detected Bd in an eDNA sample (Fig. 1C). Sequence capture 147 confirmed Bd presence in the environmental sample from this site. Phylogenomic reconstruction 148 placed the strain from the SP in the hypervirulent Bd-GPL clade (Fig. 1D). For details and 149 GenBank accession numbers see SI Appendix, Table S3. 150 The SNSM is home to at least 48 amphibian species, including at least three endemic genera and 151 25 endemic species that occur exclusively in montane regions (SI Appendix, Tables S1, S2). 152 Stream surveys from January 2024 in the SNSM revealed amphibian encounter rates of 26.5 ± 153 7.0 (mean ± SD) individuals per person per hour (range 17–39, n = 10 localities). Transect-based 154 surveys of Cryptobatrachus boulengeri conducted in June 2023 showed a mean encounter rate of 155 16.3 ± 13.0 individuals per person per hour (range 1.6–49.7, n = 25 localities) and yearly surveys 156 (2019–2024) of Atelopus laetissimus yielded 0.049 individuals m⁻ ² (range 0.020–0.080) averaged 157 across years (SI Appendix, Table S4). 158 159 Population genomics fail to find evidence for potentially Bd-induced bottlenecks in the 160 recent past. To test whether amphibian populations in the SNSM bear genomic signatures of 161 recent bottlenecks, which could have been caused by past cryptic waves of Bd followed by 162 population recovery, we generated whole-genome sequencing data for 19 individuals of the 163 endemic harlequin toad, Atelopus nahumae, from one site in the northwestern SNSM (SI 164 Appendix, Fig. S2). Observed heterozygosity exceeded heterozygosity expected under Hardy-165 Weinberg equilibrium in every individual genome (Fig. 2A), and all three inbreeding estimators 166 (Fhat1–Fhat3) yielded negative values, indicating heterozygote excess and the absence of 167 inbreeding or allele loss (SI Appendix, Table S5). Runs of homozygosity spanned very short 168 distances (0–100 kb) with only a small number of stretches exceeding 300 kb, inconsistent with a 169 recent bottleneck and increased levels of inbreeding (Fig. 2B) (32). Tajima’s D was strongly 170 skewed towards negative values across genomic windows, consistent with a large, genetically 171 diverse contemporary population (Fig. 2C). 172 Population structure analyses further supported demographic stability. Principal component 173 analysis revealed no clustering of individuals (SI Appendix, Fig. S3), and ADMIXTURE cross-174 validation identified K = 1 as the best-supported model, indicating a single cohesive population 175 with no evidence of fragmentation, admixture, or hidden substructure (SI Appendix, Fig. S4). 176 Based on the effective population size (Ne) estimations across the last 150 generations, there are 177 no signs of a rapid population contraction followed by recovery (SI Appendix, Fig. S5). See SI 178 Appendix, section 2 for details and SI Appendix, Table S5 for sample accession numbers 179 (BioProject PRJNA1390455). 180 181 Niche models reveal environmental conditions conducive to Bd-induced amphibian 182 declines and narrow lowland barriers to Bd invasion. We modelled Bd environmental 183 suitability based on ~6,000 global presence records using six climate variables selected in 184 stepwise optimized experimental runs. Probability of occurrence and binary predictors revealed 185 high environmental suitability for Bd across the entire SNSM below the snowline (Fig. 3A, SI 186 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 6 Appendix, Fig. S6, S7). Environmental suitability was equally high in the adjacent SP and Andean 187 mountain ranges in line with the pathogen’s known distribution in these areas. In contrast, 188 lowland environments surrounding the SNSM, including the ~20 km wide dry valley of the Río 189 César (Fig. 1C) between SNSM and SP, showed low suitability for Bd (Fig. 3A). Binary 190 predictions classified this region as mostly unsuitable. However, some binary predictors 191 recovered larger areas within the Río César valley as suitable (SI Appendix, Fig. S6). For the 192 second Bd-free refuge (15), New Guinea (NG), species distribution modelling revealed high 193 environmental suitability across most of the island with highest suitability in the central highlands 194 and lowest in northern lowlands (Fig. 3C). Binary predictions identified unsuitable regions 195 throughout the lowlands north of the central mountain range, and along the eastern coast (SI 196 Appendix, Fig. S6). 197 To gain further insight on whether Bd introduction may result in amphibian declines, we analyzed 198 the climatic envelope associated with Bd-driven declines from other regions of the world using the 199 same environmental predictors and georeferenced mass mortality events, declines, and 200 extinctions likely induced by Bd (SI Appendix, Fig. S7). The resulting model showed a narrower 201 range of environmental suitability for Bd-induced declines compared to Bd presence, with 202 contraction towards tropical highlands, including high suitability for Bd-driven declines across all 203 montane regions of SNSM and SP (Fig. 3B, SI Appendix, Fig. S7). In NG, the central montane 204 regions showed high suitability for Bd-driven declines with low suitability in both northern and 205 southern lowland areas (Fig. 3D, SI Appendix, Fig. S6, S7). 206 207 Experimental infections highlight community susceptibility and high risk through reservoir 208 dynamics. To assess direct susceptibility to Bd, we experimentally infected three SNSM-endemic 209 amphibian species. Selected species included the stream-associated anurans, Atelopus 210 laetissimus and Cryptobatrachus boulengeri, and the bromeliad-associated salamander, 211 Bolitoglossa savagei. Infected A. laetissimus showed no mortality with two of nine inoculated 212 individuals clearing infection. In the remaining seven individuals, infection loads remained lower 213 than in the other species but increased towards the end of the experiment in five individuals (Fig. 214 4A; SI Appendix, Fig. S8, Table S7). Experimental infection of three additional specimens at a 215 later stage and with higher zoospore dose recovered the same non-lethal infection pattern (SI 216 Appendix, Table S7). In B. savagei, all experimentally inoculated individuals presented an 217 increase in infection load with a single specimen reaching the clinical endpoint for euthanasia at 218 week 10 (Fig. 4B; SI Appendix, Fig. S8), and three individuals presenting skin sloughing towards 219 the end of the experiment. Infection loads did not saturate but increased during the experimental 220 period in the surviving specimens of B. savagei (Fig. 4B; SI Appendix, Fig. S8, Table S7). In 221 contrast, all infected C. boulengeri were euthanized following Bd-induced loss of self-righting 222 ability within 7 weeks post inoculation after a short period of increased skin sloughing and 223 abnormal hind limb posture (Fig. 4C). We did not observe significant differences in body mass 224 change between infected and control individuals during the experiment (linear model on 225 proportional mass change; A. laetissimus: p = 0.607, C. boulengeri: p = 0.535, B. savagei: p = 226 0.385, Fig. 4). Histopathology confirmed epidermal colonization by Bd in all three species. 227 228

229 Chytridiomycosis has significantly eroded amphibian diversity at the level of genes, species, and 230 communities. Our results identify a rare and globally significant exception to this pervasive 231 pattern. Five decades after Bd invaded Neotropical highlands and Australia, we provide multiple 232 lines of evidence that the seemingly unaffected amphibian communities in the SNSM, bordered 233 by a Bd-infested Andean landscape, remain epidemiologically naïve rather than having recovered 234 in a post-epizootic state. However, proving pathogen absence is intrinsically difficult (33, 34). By 235 combining extensive surveillance with negative results, high densities of experimentally 236 susceptible host species, and a flagship species lacking a genomic signature of population 237 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 7 decline, we provide independent lines of evidence for the absence of Bd in SNSM amphibian 238 communities. 239 Identifying Bd-free refuges embedded within the world’s epicenters of amphibian decline is critical 240 to safeguarding the remaining phylogenetic diversity of amphibians threatened by disease, 241 especially when these occur in diversity hotspots (15, 17). NG and SNSM currently represent a 242 narrow window of opportunity to prevent Bd introduction. Bd-free refuges represent unique pre-243 epizootic reference ecosystems. Intact montane amphibian assemblages, in the absence of Bd 244 but in a climate conducive to its establishment and impact, offer the opportunity to study pristine 245 community dynamics, and to design and apply conservation strategies before declines occur. 246 Amphibian host abundances in the SNSM were equal to pre-epizootic abundances recorded for 247 closely related Neotropical communities that have since suffered dramatic declines (11, 31). Host 248 genomic diversity as shown for the SNSM-endemic harlequin toad A. nahumae indicates long-249 term demographic stability and substantial adaptive potential, yet such variation is likely to be 250 rapidly eroded following pathogen invasion, as observed elsewhere (35, 36). Today, however, the 251 high standing genetic variation provides a particularly strong basis for informed selection and 252 management of ex-situ founder populations (37). 253 Niche models indicate high suitability for Bd presence and conditions conducive to Bd-induced 254 declines throughout both refuges. While NG amphibian communities might benefit from oceanic 255 isolation and lowland environments poorly suitable for Bd along potential coastal entry points, the 256 SNSM is separated only by a narrow region of low environmental suitability for Bd in the 257 surrounding lowlands and the dry valley to the east towards the Andes. These lowland 258 environments likely act as a climatic barrier that has so far prevented host-mediated Bd dispersal 259 from infected regions, such as the nearby Serranía de Perijá (SP) (17). In parallel, relatively low 260 human impact in the montane zones of both SNSM and SP (SI Appendix, Fig. S9) combined with 261 thermal inactivation of the pathogen’s heat-sensitive zoospores in lowland surroundings may 262 have prevented anthropogenic translocation of Bd (38). However, Bd exhibits substantial 263 ecological and evolutionary diversity (39-41), and other hypervirulent lineages with distinct 264 thermal growth patterns could pass the lowland barrier into the SNSM. Further, binary predictors 265 recover patches of environmentally suitable habitat throughout the lowland barrier. Coupled with 266 the presence of Bd-GPL less than 20 km from the SNSM, Bd introduction is highly likely. In the 267 case of NG, potential Bd emergence might be more complex, with different major Bd clades 268 occurring in surrounding regions, such as the hypervirulent Bd GPL clade in Australia and the 269 poorly studied yet possibly less virulent Asia3 clade on Sulawesi (41). Overall, increasing human 270 activity such as the ongoing expansion of infrastructure, tourism, and trade in SNSM and NG will 271 further exacerbate the risk for unintended human-mediated pathogen introduction through 272 contaminated materials or stow-away amphibians (16, 42, 43). 273 Our experimental infections highlight the likely severity of future Bd invasion for SNSM 274 communities but also raise hope for the persistence of some species. While Cryptobatrachus 275 boulengeri showed rapid, lethal chytridiomycosis, Atelopus laetissimus and Bolitoglossa savagei 276 developed increasing infection loads with the latter showing delayed mortality and no loss of 277 infection, potentially contributing to reservoir dynamics with prolonged transmission and host-278 mediated spread (44, 45). The elevated tolerance of A. laetissimus is exceptional given the 279 unparalleled chytridiomycosis-driven declines in the genus (26, 31) and high individual 280 susceptibility to experimental infection in other Atelopus species (44, 46, 47). Understanding the 281 mechanisms that enable A. laetissimus to survive with Bd infection might provide rare insights for 282 safeguarding the >100 highly threatened emblematic harlequin toad species and underscores the 283 importance of epidemiologically naïve refuges to conserve the global amphibian phylogenetic 284 diversity. This tolerance is likely not the result of past pathogen-host coevolution, as we found no 285 evidence of potentially Bd-induced declines in the closely related and syntopic A. nahumae. 286 Together with A. nahumae, A. laetissimus is part of the phylogenetically most basal clade in the 287 genus (48) and given the apparent lack of epizootic declines in the sister genus Oreophrynella 288 (38), a phylogenetic signal in susceptibility might explain the elevated tolerance. However, our 289 experiments evaluated only adults, while other life stages may exhibit higher disease 290 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 8 susceptibility and disproportionately influence population trajectories during an epizootic (49, 50). 291 The SNSM harbors at least 25 endemic amphibian species, including evolutionarily distinct and 292 endemic genera and several harlequin toad (Atelopus) species that are deeply embedded in 293 regional indigenous culture, where the toad’s seasonal activity patterns inform agricultural cycles 294 (SI Appendix, Table S1). In NG, the number of endemic taxa is even higher, with dozens of 295 genera and hundreds of species (15, 23), yet experimental infection trials are lacking. Relative to 296 ~90 potentially extinct species and a further >400 declined species associated with Bd (6), 297 pathogen introduction to these two large Bd-free refuges could strongly exacerbate global 298 phylogenetic and functional amphibian diversity loss. 299 These insights argue for an explicit and unified emergency framework to safeguard remaining Bd-300 free refuges threatened by likely Bd introduction. Mitigation and prevention strategies for 301 amphibian chytrid fungi have been developed and tested extensively over the past decades (19-302 21, 51-53). Core elements include (i) strict biosecurity protocols for human activities, (ii) continued 303 active surveillance of Bd coupled with host population monitoring in refuges and adjacent areas, 304 (iii) targeted susceptibility and mitigation studies spanning the evolutionary diversity of amphibian 305 species in the refugia, and (iv) rapid development of ex situ capacity for the most vulnerable and 306 evolutionarily distinct species in close collaboration with local and indigenous stakeholders. 307 These efforts must be accompanied by an active, systematic search for additional refuges that 308 may still persist undetected, particularly in yet poorly surveyed regions that harbor a climate 309 conducive to Bd-driven declines. 310 Unlike during the initial waves of ‘enigmatic’ amphibian declines four decades ago, conservation 311 science now has tailored much effort into researching pathogen dynamics, improving surveillance 312 tools, and applying experimental mitigation approaches to detect, anticipate, and prevent 313 pathogen emergence. Whether these tools can be translated into effective prevention will be 314 determined by outcomes in the world’s last major Bd-free refuges. The SNSM and NG thus stand 315 as parallel, irreplaceable ecosystems for averting further disease-driven biodiversity loss, raising 316 hope for safeguarding global phylogenetic and functional amphibian diversity. 317 318 319 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 9

320 Pathogen screening. Permission to collect samples was granted by Autoridad Nacional de 321 Licencias Ambientales (ANLA), Colombia, under research and collection permits RCM0014-00-322 2024 and RCI003-00-2020, with mobilization permits P02182S9571_N0055, 323 P02182S9691_N0057 and P11487S10331_N0002. Some samples were exported to Germany 324 under ANLA permit Nos. 003790, 003880 and 004120 validated by Secretaría Distrital de 325 Ambiente, Distrito Capital de Bogotá. We conducted several field expeditions during both dry and 326 rainy seasons from 2021 to 2025 targeting premontane and montane amphibian communities in 327 34 localities in the SNSM (Fig. 1C). Additional field expeditions during June to September of 2024 328 targeted adjacent lowland ecosystems as well as the neighboring mountain range Serranía de 329 Perijá (SP; Fig. 1C). Amphibians were sought during nocturnal encounter surveys in an 330 opportunistic manner resulting in the collection of 2,212 swab samples (Fig. 1C; SI Appendix, 331 Table S6). The spatial structure of sampling was limited by the geopolitical situation, which 332 precluded a larger sampling in the eastern, northeastern and southern slopes of the SNSM and 333 adjacent lowlands (Fig. 1C). Swabbing followed 54 with swabs stored in 180 µl DNA/RNA Shield 334 (Zymo Research, Freiburg) for DNA preservation or dry for on-site processing. In addition, we 335 collected 27 water samples (SI Appendix, Table S6) for environmental detection of Bd as 336 described in 55. 337 We deployed two diagnostic approaches (sample information in SI Appendix, Table S6): all eDNA 338 and most swab samples were processed with quantitative PCR (qPCR) in duplicate following 56. 339 For qPCR, samples were extracted using the DNeasy Blood & Tissue Kit (Qiagen, Hilden) with 340 deviations from the manufacturer’s protocol as described in 57 for swabs and in 55 for 341 environmental samples. qPCR reactions were performed in 20µl volumes on a StepOnePlus 342 platform using TaqMan Fast Advanced reagents (Thermo Fisher, Waltham) and triplicate 343 synthetic standards (101-106 ITS copies). We considered samples with sigmoidal amplification 344 curves and Cq ≤ 40 positive, with a standard curve efficiency of 90-110% and R2 ≥ 0.98 required 345 to accept the results of a plate. To guide decisions on sampling design, some swab samples were 346 processed on-site deploying CRISPR diagnostics (CRISPR-Dx) (SI Appendix, Table S6), as 347 follows. We adapted the Cas12-based rapid assay, FINDeM (58), by using the improved fast 348 DNA extraction protocol of 55, followed by isothermal Recombinase Polymerase Amplification 349 and Cas12-activated cleavage of a fluorescence reporter. Performance of both methods has 350 been compared previously (58, 59). FINDeM reaction conditions followed the supplementary 351 protocol in 59 and reactions were run in triplicate alongside controls and visualized in the field 352 with a handheld UV flashlight and UV glasses. 353 We assessed the posterior probability of infection per site for diagnostic zero detections as 354 described in 33 using WinBUGS14 (60) with skin swab data only. We assumed three test 355 sensitivity values for pathogen diagnostics (0.5, 0.75 and 1.0) and perfect specificity given zero 356 detections in our dataset. For posterior probability estimates, we further combined samples by 357 nearest-neighbor distance of ≤ 1 km. 358 As our eDNA screening in areas surrounding the SNSM led to the detection of Bd in 359 environmental samples from the SP to the southeast, we employed a sequence-capture 360 approach on the positive sample as an independent verification of Bd presence and to obtain 361 data for phylogenomic placement of the Bd lineage. Sequence capture followed 61, with library 362 preparation using the Watchmaker Genomics DNA Library Prep Kit. The resulting 26 loci were 363 concatenated into a 6,035 bp alignment with no missing data. Besides the SP sample, the 364 alignment contained a representative panel of whole-genome and sequence-capture Bd samples 365 to determine the phylogenetic position of the SP-derived lineage (SI Appendix, Table S3). We 366 inferred a maximum likelihood phylogeny using IQ-TREE2 version 2.3.6 (62) visualized assuming 367 midpoint rooting as implemented in FigTree version 1.4.4 (63). 368 369 Host population genomics. We selected Atelopus nahumae (Anura: Bufonidae) as a model 370 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 10 species for population genomic profiling due to the high susceptibility of harlequin toads 371 (Atelopus) to Bd (12, 26) and its stream associated life history in montane forests, where Bd-372 driven declines have been common in other harlequin toad species (see above). We evaluated i) 373 genome-wide heterozygosity as an indicator for genetic variation for assessing the potential for 374 genetic adaptation to any future outbreak of Bd and ii) the potential presence of demographic 375 bottlenecks or inbreeding in the recent past that could indicate epizootic decline and subsequent 376 recovery. We collected toeclips of 19 specimens from the northwestern SNSM (SI Appendix, Fig. 377 S2) and extracted DNA using the DNeasy Blood & Tissue Kit. We sequenced whole genomes 378 using DNBSeq technology (T7 platform; BGI, Hong Kong) from PCR-free, 150 bp paired-end 379 libraries, targeting 10× assuming a genome size of 3.5 Gb (64) with coverage reported in SI 380 Appendix, Table S5. Reads were adapter-trimmed and base-quality-filtered using the software 381 FASTP version 0.23.4 (65) (Phred ≥ 20) before mapping against the A. laetissimus reference 382 genome (GenBank accession PRJNA1142550; 64) using the software BWA-MEM2 version 2.2.1 383 (66). After stringent filtering for mapping quality using SAMtools version 1.18 (67) (MAPQ ≥ 30), 384 we used ANGSD version 0.940 (68) to identify SNPs. ANGSD utilizes genotype likelihoods, which 385 handle genotyping and mapping errors better than other SNP calling software at low coverage 386 (68). Subsequently, tfam, tped were generated as input for PLINK version 1.9 (69). Using PLINK 387 we generated VCF files from the tped and tfam as input for iSMC version 1.0.0 (70). We 388 estimated inbreeding coefficients (Fhat1–Fhat3, following 71) (SI Appendix, Table S5) and 389 population structure with PCA in PLINK. In brief, Fhat1 estimates the variance-standardised 390 relatedness of an individual with the average individual of the population according to Hardy-391 Weinberg expectations. Fhat2 corresponds to the basic calculation of excess homozygosity. 392 Fhat3 is based on Wright's F (72) and calculates the correlation between uniting gametes. We 393 calculated Tajima’s D with VCFtools (73) for scaffolds ≥ 10 mbp. Genome-wide heterozygosity 394 and Tajima’s D provide information on overall genetic diversity and allele frequency distributions 395 and can indicate populations inconsistent with recent severe bottlenecks. However, these metrics 396 alone cannot distinguish long-term demographic stability from population recovery following a 397 recent decline, as rapid post-bottleneck expansion may restore heterozygosity and generate 398 negative Tajima’s D values. We therefore additionally quantified Runs of Homozygosity using 399 BCFtools (74) for scaffolds ≥ 10 mbp, which provide insights into recent severe bottlenecks and 400 inbreeding over few generations (32, 75), when potential cryptic Bd-induced declines would have 401 been most likely. Population structure was further examined using ADMIXTURE (version 1.3.0; 402 76). In addition, we explored recent changes in effective population size (Ne). We used iSMC to 403 calculate the per-individual recombination rate (SI Appendix, Table S5) to obtain the cM/Mbp 404 values necessary for implementation of GONE2 (77). Because iSMC is based on the sequentially 405 Markovian coalescent (SMC) method, it is designed for large timeframes of hundreds of 406 thousands of years. Accurate estimations of the past 100 years are not possible with SMC 407 methods. The software package GONE2 uses linkage disequilibrium to estimate Ne in recent past 408 generations. We therefore implemented GONE2 to accurately estimate the recent demographic 409 history of A. nahumae. A mutation rate of 3e-9 per site per generation was assumed based on an 410 estimation for Atelopus manauensis (78). Details on population genomic methods are provided in 411 SI Appendix, section 2. 412 413 Niche Modelling for Bd presence and Bd-induced declines. To assess environmental 414 suitability for Bd throughout SNSM and NG, we compiled global occurrence records for Bd 415 (spanning the period 1981–2024) from the Amphibian Disease Portal (79; downloaded 21 May 416 2025). We refrained from using regional subsets of Bd occurrences, as applied in previous work 417 (80), because lineage-specific training data are not available at sufficient resolution and the 418 complete global niche is most relevant for risk assessment given that the source of any future 419

into the Bd-free refuges cannot be predicted. We removed duplicate records by 420 coordinates, resulting in ~6,000 unique geo-referenced occurrences, 75% of which were used as 421 training points in machine-learning correlative presence-only distribution modelling following the 422 principle of maximum entropy. Climatic predictors were clipped to global land surfaces and 423 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 11 included 66 numeric CHELSA version 2.1 variables (SI Appendix, Table S8, 30 arc seconds 424 resolution, 1981–2010; 81-84), which provide increased spatial resolution compared to previous 425 environmental suitability models for Bd (9, 79). We modeled Bd environmental suitability in 426 MaxEnt v3.4.4 following 9. In brief, we used linear (L), product (P), and quadratic (Q) features to 427 reduce overfitting (85). We ran ten replicate models and assessed model performance using area 428 under the Receiver Operating Characteristic curve and omission rate (86). Predictor variables 429 were reduced from 66 to six based on contribution metrics, jackknife tests, and response curve 430 interpretability in two stepwise exploratory runs (SI Appendix, Fig. S10, S11). Selected variables 431 include Annual Precipitation (bio12), Growing Season Length (gsl), Number of Growing Degree 432 Days above 0 °C (ngd0), Net Primary Production on Land as Carbon Mass Flux (npp), Monthly 433 Potential Evapotranspiration (pet_penman_min) and Surface Downwelling Shortwave Flux in Air 434 (rsds_range) (SI Appendix, Fig. S12). Continuous clog-log outputs were further converted to 435 binary predictions using three common thresholds (SI Appendix, Table S9): 10th percentile 436 training presence, maximum training sensitivity plus specificity, and maximum test sensitivity plus 437 specificity (86-88). To further assess the putatively more confined climatic envelope in which Bd 438 induces amphibian declines (89, 90), in a second run we compiled and georeferenced 969 global 439 locality records at which amphibian mass mortalities, declines or extinctions most likely induced 440 by Bd have been reported (SI Appendix, Fig. S12). We used the six climatic variables selected 441 with presence data for MaxEnt modelling as described above. 442 443 Infection trials. We used three representative species from the premontane and high montane 444 SNSM amphibian communities to assess susceptibility to Bd infection. Selected species were 445 phylogenetically unrelated and ecologically distinct (30), including the bromeliad-associated 446 salamander, Bolitoglossa savagei (Plethodontidae) and the stream-associated anurans, 447 Cryptobatrachus boulengeri (Hemiphractidae) and Atelopus laetissimus (Bufonidae). All animal 448 experiments were approved by the institutional animal care and use committee (Comité 449 Institucional para el Cuidado y Uso de Animales, CICUA) of Universidad de los Andes, Bogotá 450 under code C.FUA_24-020. Collection of live individuals was authorized by Autoridad Nacional de 451 Licencias Ambientales (ANLA) under Resolución 002166 with mobilization permit 452 P11487S10853_N0005. Animals were collected during nocturnal surveys and transported to a 453 controlled climate chamber (~20°C, ~80% relative humidity, 12 h light cycle) at the Universidad 454 de los Andes, Bogotá. All frogs tested negative upon arrival using CRISPR-Dx, as described 455 above. Live animals were weighed before inoculation and after euthanasia on a digital scale to 456 the nearest 0.01 g and measured with calipers to the nearest 0.1 mm prior to inoculation. Animals 457 were housed individually in terraria with wet paper towels as substrate, plastic hides and ad-458 libitum provision of feeder insects. Individuals were randomly assigned to treatment groups. After 459 a one-week acclimatization period, we experimentally infected six individuals per species 460 alongside six negative controls. The experiment was terminated 10 weeks after inoculation. 461 Experimental setup followed 91: In brief, we used the hypervirulent Bd-GPL isolate JEL423 for 462 infection trials, cultured on 0.75× strength TGhL plate media at 20°C. For the inoculum, we rinsed 463 plates with 3 ml sterile distilled water and counted spores using a Neubauer improved 464 hemocytometer. Spores were diluted to a final concentration of 5 × 106 spores ml-1 and 465 amphibians were inoculated in random order with 1 ml for 24 h in individual plastic tubes. 466 Negative controls were sham-inoculated. In the case of A. laetissimus, due to slow infection 467 progression during the first weeks of the experiment, incongruent with the high susceptibility of 468 other Atelopus species (47), we inoculated three additional specimens that were not inoculated in 469 the first trial. Inoculation of the additional specimens was conducted as described above but using 470 a higher dose (8 × 106 spores ml-1) four weeks post initial inoculation, to confirm tolerance of A. 471 laetissimus to experimental infection with an increased sample size and an independent Bd 472 inoculum (SI Appendix, Table S7). Individuals were inspected daily for clinical signs and skin 473 swabs were collected every seven days and at the end of the experiment. Swabs were preserved 474 in DNA/RNA Shield and analyzed with qPCR as described above. We set the loss of self-righting 475 ability as a humane endpoint for euthanasia with an overdose of lidocaine. Upon euthanasia, we 476 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 12 preserved skin sections in 5 % buffered formalin for histopathologic confirmation of epidermal Bd 477 colonization (91). Sublethal effects were assessed as proportional change in body mass using the 478 ln ratio of terminal to initial body mass, ln(terminal weight/initial weight). For each species, we fit 479 linear models with infection status (inoculated vs. control) as the focal effect, using the natural 480 logarithm of initial mass as covariate. Inference used HC3 heteroscedasticity-robust standard 481 errors to reduce small sampling bias (92). 482 483 Hygiene and biosafety protocols. During all field surveys, each individual was handled with a 484 fresh pair of nitrile gloves (93). Prior to each field visit, boots and other equipment were 485 thoroughly cleaned and subsequently disinfected with 70% ethanol or Virkon S and no sampling 486 consumables previously brought to other field sites were employed. Infection experiments were 487 carried out in a climate chamber at Universidad de los Andes with all materials in the climate 488 chamber being disinfected with 70% ethanol and waste being autoclaved at the end of the 489 experiment. 490 491 492 Acknowledgments 493 We are grateful to Amanda B. Quezada Riera, Andrés Rocha Usuga, María José Navarrete 494 Méndez and Viktoria Ferner for their help in the field. We thank Angie Sánchez Galán of the 495 Museo de Historia Natural C. J. Marinkelle at the Universidad de los Andes for her help with 496 obtaining permits. We are further grateful to Deborah Bower and Philipp Böning for valuable 497 discussions and to Elvia Mora, Juana Diaz, Karin Fischer and Sabine Naber for their help in the 498 lab. We want to thank the Arhuaco and Kogi communities of the Sierra Nevada de Santa Marta 499 as well as the families from San Pedro de la Sierra and Palmor for their hospitality and support 500 throughout our field expeditions. We are grateful to Fundación Atelopus for supporting logistics on 501 field expeditions. We further thank the Colombian environmental authorities for issuing permits. 502 This work was partially funded by the Forschungsfonds of Trier University, the Wilhelm-Peters-503 Fonds of the German Society of Herpetology DGHT, and the National Geographic Society (Grant 504 number NGS-66509C-20). Computational resources on the HPC Elwetritsch at RPTU were 505 provided by Rechenzentrumsallianz Rheinland-Pfalz (RARP). Amadeus Plewnia is funded by the 506 Research Foundation Flanders (FWO) under PhD fellowship fundamental research 1104226N. 507 508

509 510 1. P. Daszak, A. A. Cunningham, A. D. Hyatt, Emerging infectious diseases of wildlife - threats to 511 biodiversity and human health. Science. 287, 443–449 (2000). 512 2. M. C. Fisher et al., Emerging fungal threats to animal, plant and ecosystem health. Nature. 513 484, 186–194 (2012). 514 3. A. Plewnia, P. Böning, S. Lötters, Mitigate diseases to protect biodiversity. Science. 379, 1098 515 (2023). 516 4. C. Bellard, P. Genovesi, J. M. Jeschke, Global patterns in threats to vertebrates by biological 517 invasions. Proc. R. Soc. B. 283, 20152454 (2016). 518 5. L. Berger et al., Chytridiomycosis causes amphibian mortality associated with population 519 declines in the rain forests of Australia and Central America. Proc. Natl. Acad. Sci. U.S.A. 520 95, 9031–9036 (1998). 521 6. B. C. Scheele et al., Amphibian fungal panzootic causes catastrophic and ongoing loss of 522 biodiversity. Science, 363, 1459–1463 (2019). 523 7. J. A. Luedtke et al., Ongoing declines for the world's amphibians in the face of emerging 524 threats. Nature. 622, 308–314 (2023). 525 8. M. C. Fisher, T. W. J. Garner, S. F. Walker, Global emergence of Batrachochytrium 526 dendrobatidis and amphibian chytridiomycosis in space, time, and host. Annu. Rev. 527 Microbiol. 63, 291–310 (2009). 528 9. S. Lötters et al., The link between rapid enigmatic amphibian decline and the globally emerging 529 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 13 chytrid fungus. EcoHealth. 6, 358–372 (2010). 530 10. A. J. Crawford, K. R. Lips, E. Birmingham, Epidemic disease decimates amphibian 531 abundance, species diversity, and evolutionary history in the highlands of central Panama. 532 Proc. Natl. Acad. Sci. U.S.A. 107, 13777–13782 (2010). 533 11. K. R. Lips et al., Emerging infectious disease and the loss of biodiversity in a Neotropical 534 amphibian community. Proc. Natl. Acad. Sci. U.S.A. 103, 3165–3170 (2006). 535 12. K. R. Lips, J. Diffendorfer, J. R. Mendelson III, M. W. Sears, Riding the wave: reconciling the 536 roles of disease and climate change in amphibian declines. PLoS Biol. 6, e72 (2008). 537 13. B. C. Scheele et al., After the epidemic: Ongoing declines, stabilizations and recoveries in 538 amphibians afflicted by chytridiomycosis. Biol. Conserv. 206, 37–46 (2017). 539 14. H. D. Granda Rodríguez, A. del Portillo Mozo, J. M. Renjifo, F. Bolaños, ¿Están declinando 540 todos los Atelopus de alta montaña? El caso de las ranas arlequín de la Sierra Nevada de 541 Santa Marta, Colombia. Herpetotropicos. 7, 21–30 (2012). 542 15. D. S. Bower et al., Amphibians on the brink. Science. 357, 454–455 (2017). 543 16. D. S. Bower et al., Island of opportunity: can New Guinea protect amphibians from a globally 544 emerging pathogen? Front. Ecol. Environ. 17, 348–354 (2019). 545 17. C. J. Hoskin, H. B. Hines, R. J. Webb, L. F. Skerratt, L. Berger, Naïve rainforest frogs on 546 Cape York, Australia, are at risk of the introduction of amphibian chytridiomycosis disease. 547 Aust. J. Zool. 66, 174–178 (2018). 548 18. L. F. Grogan, M. J. Mangan, H. I. McCallum, Amphibian infection tolerance to 549 chytridiomycosis. Phil. Trans. R. Soc. B. 378, 20220133 (2023). 550 19. D. C. Woodhams et al., Mitigating amphibian disease: strategies to maintain wild populations 551 and control chytridiomycosis. Front. Zool. 8, 8 (2011). 552 20. T. W. J. Garner et al., Mitigating amphibian chytridiomycosis in nature. Phil. Trans. R. Soc. B. 553 371, 20160207 (2016). 554 21. V. Thomas et al., Mitigating Batrachochytrium salamandrivorans in Europe. Amphib. Reptil. 555 40, 265–290 (2019). 556 22. D. S. Bower et al., Disease surveillance of the amphibian chytrid fungus Batrachochytrium 557 dendrobatidis in Papua New Guinea. Conserv. Sci. Pract. 2, e256 (2020). 558 23. P. M. Oliver et al., Melanesia holds the world’s most diverse and intact insular amphibian 559 fauna. Commun. Biol. 5, 1182 (2022). 560 24. L. A. Rueda-Solano et al., Epidemiological surveillance and amphibian assemblage status at 561 the Estación Experimental de San Lorenzo, Sierra Nevada de Santa Marta, Colombia. 562 Amphib. Reptile Conserv. 10, 7–19 (2016). 563 25. S. V. Flechas et al., Current and predicted distribution of the pathogenic fungus 564 Batrachochytrium dendrobatidis in Colombia, a hotspot of amphibian diversity. Biotropica. 565 49, 685–694 (2017). 566 26. S. Lötters et al., Ongoing harlequin toad declines suggest the amphibian extinction crisis is 567 still an emergency. Comm. Earth Environ. 4, 412 (2023). 568 27. A. G. Ruthven, The amphibians and reptiles of the Sierra Nevada de Santa Marta, Colombia. 569 Misc. Publ. Mus. Zool. Univ. Mich. 8, 1–69 (1922). 570 28. J. L. Pérez-González, L. R. Mejía-Quintero, L. C. Jiménez-López, L. A. Rueda- Solano, 571 Anfibios de Santa Marta y sus alrededores, Colombia. Field Guides. 678, 1–3 (2015). 572 29. S. Arroyo et al., A new genus of terraranas (Anura: Brachycephaloidea) from northern South 573 America, with a systematic review of Tachiramantis. Syst. Biodivers. 20, 1-25 (2022). 574 30. J. Erens et al. A field-deployable eDNA metabarcoding workflow including de novo reference 575 assembly for characterizing understudied biodiversity hotspots. bioRxiv [Preprint] (2025). 576 https://doi.org/10.1101/2025.09.14.676136 (accessed 16 January 2026). 577 31. E. La Marca et al., Catastrophic population declines and extinctions in Neotropical harlequin 578 frogs (Bufonidae: Atelopus). Biotropica. 37, 190–201 (2005). 579 32. A. Talavera et al., Genomic insights into the Montseny brook newt (Calotriton arnoldi), a 580 Critically Endangered glacial relict. Iscience, 27, 108665 (2024). 581 33. A. A. Cunningham et al., Apparent absence of Batrachochytrium salamandrivorans in wild 582 urodeles in the United Kingdom. Sci. Rep. 9, 2831 (2019). 583 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 14 34. H. Sentenac et al., Accounting for bias in prevalence estimation: The case of a globally 584 emerging pathogen. J. Appl. Ecol. 60, 2007–2017 (2023). 585 35. A. Q. Byrne et al., Whole exome sequencing identifies the potential for genetic rescue in 586 iconic and critically endangered Panamanian harlequin frogs. Glob. Change Biol. 27, 50–587 70 (2020). 588 36. K. E. Jaynes et al., Harlequin frog rediscoveries provide insights into species persistence in 589 the face of drastic amphibian declines. Biol. Conserv. 276, 109784 (2022). 590 37. M. J. Davidson et al., Exposure to low doses of Batrachochytrium dendrobatidis reveals 591 variation in resistance in the Critically Endangered southern corroboree frog. Glob. Ecol. 592 Conserv. 60, e03587 (2025). 593 38. P. J. R. Kok et al., Historical biogeography of the palaeoendemic toad genus Oreophrynella 594 (Amphibia: Bufonidae) sheds a new light on the origin of the Pantepui terrestrial biota. J. 595 Biogeogr. 45, 26–36 (2018). 596 39. S. J. O’Hanlon et al., Recent Asian origin of chytrid fungi causing global amphibian declines. 597 Science. 360, 621–627 (2018). 598 40. J. Voyles et al., Diversity in growth patterns among strains of the lethal fungal pathogen 599 Batrachochytrium dendrobatidis across extended thermal optima. Oecologia. 184, 363–373 600 (2017). 601 41. A. Q. Byrne et al., Cryptic diversity of a widespread global pathogen reveals expanded threats 602 to amphibian conservation. Proc. Natl. Acad. Sci. U.S.A. 116, 20382–20387 (2019). 603 42. P. J. R. Kok et al., Tourism may threaten wildlife disease refugia. Conserv. Lett. 2022, e12902 604 (2022). 605 43. J. D. A. Serrano, L. F. Toledo, L. P. Sales, Human impact modulates chytrid fungus 606 occurrence in amphibians in the Brazilian Atlantic Forest. Perspect. Ecol. Conserv. 20, 607 256–262 (2022). 608 44. G. V. DiRenzo et al., Fungal infection intensity and zoospore output of Atelopus zeteki, a 609 potential acute chytrid supershedder. PLoS ONE. 9, e93356 (2014). 610 45. B. C. Scheele et al., Reservoir-host amplification of disease impact in an endangered 611 amphibian. Conserv. Biol. 31, 592–600 (2016). 612 46. K. R. Lips, Mass mortality and population declines of anurans at an upland site in western 613 Panama. Conserv. Biol. 13, 117–125 (1999). 614 47. A. R. Ellison et al., Fighting a losing battle: vigorous immune response countered by pathogen 615 suppression of host defenses in the chytridiomycosis-susceptible frog Atelopus zeteki. 616 Genes Genomes Genet. 4, 1275–1289 (2014). 617 48. S. Lötters et al., A roadmap for harlequin frog systematics, with a partial revision of 618 Amazonian species related to Atelopus spumarius. Zootaxa. 5571, 1–76 (2025). 619 49. A. Abu Bakar et al., Susceptibility to disease varies with ontogeny and immunocompetence in 620 a threatened amphibian. Oecologia. 181, 997–1009 (2016). 621 50. A. W. Waddle et al., Population-level resistance to chytridiomycosis is life-stage dependent in 622 an imperiled anuran. EcoHealth. 16, 701–711 (2019). 623 51. L. Berger et al., Advances in managing chytridiomycosis for Australian frogs: Gradarius firmus 624 victoria. Annu. Rev. Anim. Biosci. 20, 113–133 (2024). 625 52. D. H. Olson et al., Preparing for a Bsal invasion into North America has improved multi-sector 626 readiness. Front. Amph. Rept. Sci. 2, 1–21 (2024). 627 53. L. E. B. Goodyear et al., Global assessment of interventions for mitigation of amphibian fungal 628 disease is dominated by geoeconomic trends and antifungal success. bioRxiv [Preprint] 629 (2025) https://doi.org/10.1101/2025.07.03.662529 (accessed 16 January 2026). 630 54. A. D. Hyatt et al., Diagnostic assays and sampling protocols for the detection of 631 Batrachochytrium dendrobatidis. Dis. Aquat. Org. 73, 175–192 (2007). 632 55. A. Plewnia, H. Krehenwinkel, C. Heine, An isothermal workflow for low-cost and PCR-free 633 field-based community metabarcoding. Method. Ecol. Evol. 16, 1413–1424 (2025). 634 56. I. Standish et al., Optimizing, validating, and field-testing a multiplex qPCR for the detection of 635 amphibian pathogens. Dis. Aquat. Org. 129, 1–13 (2018). 636 57. P. Böning et al., Alpine salamanders at risk? The current status of an emerging fungal 637 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 15 pathogen. PLoS ONE. 19, e0298591 (2024). 638 58. B. D. Hoenig et al., FINDeM: A CRISPR-based, molecular method for rapid, inexpensive and 639 field-deployable organism detection. Method. Ecol. Evol. 14, 3055–3067 (2023). 640 59. B. D. Hoenig, P. Böning, A. Plewnia, C. L. Richards-Zawacki, A simplified, CRISPR-based 641

for the detection of Batrachochytrium salamandrivorans. EcoHealth. 22, 161–171 642 (2024). 643 60. D. J. Lunn, A. Thomas, N. Best, D. Spiegelhalter, WinBUGS - A bayesian modelling 644 framework: concepts, structure, and extensibility. Stat. Comput. 10, 325–337 (2000). 645 61. K. P. Mulder et al., Sequence capture identifies fastidious chytrid fungi directly from host 646 tissue. Fung. Genet. Biol. 170, 103858 (2024). 647 62. B. Q. Minh et al., IQ-TREE 2: New models and efficient methods for phylogenetic inference in 648 the genomic era. Mol. Biol. Evol. 37, 1530–1534 (2020). 649 63. A. Rambaut, FigTree v1.4.4. Institute of Evolutionary Biology, University of Edinburgh (2018). 650 64. V. L. N. Araújo et al., Genome assembly of the endangered Santa Marta harlequin toad, 651 Atelopus laetissimus, endemic to the Sierra Nevada de Santa Marta of Colombia. J. Hered. 652 116, 673–679 (2025). 653 65. S. Chen, Y. Zhou, Y. Chen, J. Gu, fastp: an ultra-fast all-in-one FASTQ preprocessor. 654 Bioinform. 34, 884–890 (2018). 655 66. M. Vasimuddin, S. Misra , H. Li, S. Aluru, Efficient architecture-aware acceleration of BWA-656 MEM for Multicore Systems. IEEE International Parallel and Distributed Processing 657 Symposium (IPDPS). (2019). 658 67. H. Li et al., The sequence alignment/map format and SAMtools. Bioinform. 25, 2078–2079 659 (2009). 660 68. T.S. Korneliussen, A. Albrechtsen, R. Nielsen, ANGSD: Analysis of next generation 661 sequencing data. BMC Bioinform. 15, 356 (2014). 662 69. C. C. Chang et al., Second-generation PLINK: rising to the challenge of larger and richer 663 datasets. GigaScience. 4, s13742-015-0047-8 (2015). 664 70. V. Barroso, G., Puzović , N., & Dutheil, J. Y. Inference of recombination maps from a single 665 pair of genomes and its application to ancient samples. PLoS Genet. 15, e1008449 (2019). 666 71. J. Yang, S. H. Lee, M. E. Goddard, P. M. Visscher, GCTA: a tool for genome-wide complex 667 trait analysis. Am. J. Hum. Genet. 88, 76–82 (2011). 668 72. S. Wright, Coefficients of inbreeding and relationship. Am. Nat. 56, 289–384 (1922). 669 73. P. Danecek et al., The variant call format and VCFtools. Bioinform. 27, 2156–2158 (2011). 670 74. P. Danecek et al., Twelve years of SAMtools and BCFtools. GigaScience. 10, giab008 (2021). 671 75. F. C. Ceballos et al., Runs of homozygosity: windows into population history and trait 672 architecture. Nat. Rev. Genet. 19, 220–234 (2018). 673 76. D. H. Alexander, K. Lange, Enhancements to the ADMIXTURE algorithm for individual 674 ancestry estimation. BMC Bioinform. 12, 246 (2011). 675 77. E. Santiago, C. Köpke, A. Caballero, Accounting for population structure and data quality in 676 demographic inference with linkage disequilibrium methods. Nat. Commun. 16, 6054 677 (2025). 678 78. R. F. Jorge, A. P. Lima, A. J. Stow, Selection and localised genetic structure in the threatened 679 Manauense Harlequin Frog (Bufonidae: Atelopus manauensis). Conserv. Genet. 23, 559–680 574 (2022). 681 79. D. H. Olson, K. L. Ronnenberg, C. K. Glidden, K. R. Christiansen, A. R. Blaustein, Global 682 patterns of the fungal pathogen Batrachochytrium dendrobatidis support conservation 683 urgency. Front. Vet. Sci. 8, 685877 (2021). 684 80. S. V. Flechas et al., Current and predicted distribution of the pathogenic fungus 685 Batrachochytrium dendrobatidis in Colombia, a hotspot of amphibian biodiversity. 686 Biotropica. 49, 685–694 (2017). 687 81. D. N. Karger et al., Climatologies at high resolution for the earth’s land surface areas. Sci. 688 Data. 4, 170122 (2017). 689 82. D. N. Karger et al., Climatologies at high resolution for the earth’s land surface areas. 690 EnviDat. (2021) https://www.doi.org/10.16904/envidat.228 (accessed 20 November 2025). 691 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 16 83. D. N. Karger et al., CHELSA-TraCE21k high resolution (1 km) downscaled transient 692 temperature and precipitation data since the Last Glacial Maximum. Clim. Past. 19, 439–693 456 (2023). 694 84. P. Brun et al., Global climate-related predictors at kilometer resolution for the past and future. 695 Earth Syst. Sci. Data. 14, 5573–5603 (2022). 696 85. S. J. Phillips, M. Dudik, Modeling of species distributions with Maxent: new extensions and a 697 comprehensive evaluation. Ecography. 31, 161–175 (2008). 698 86. S. J. Phillips, R. P. Anderson, R. E. Schapire, Maximum entropy modeling of species 699 geographic distributions. Ecol. Model. 190, 231–259 (2006). 700 87. C. Liu, G. Newell, M. White, On the selection of thresholds for predicting species occurrence 701 with presence‐ only data. Ecol. Evol. 6, 337–348 (2016). 702 88. S. J. Phillips et al., Opening the black box: an open-source release of Maxent. Ecography. 40, 703 887–893 (2017). 704 89. B. C. Scheele et al., An invasive pathogen drives directional niche contractions in amphibians. 705 Nat. Ecol. Evol. 7, 1682–1692 (2023). 706 90. B. Thumsová et al., Ecological niche models based on occurrence records of the amphibian 707 chytrid fungus overestimate lethal chytridiomycosis risk. J. Appl. Ecol. 62, 995–1006 708 (2025). 709 91. M. S. Greener et al., Presence of low virulence chytrid fungi could protect European 710 amphibians from more deadly strains. Nat. Commun. 11, 5393 (2020). 711 92. J. G. MacKinnon, H. White, Some heteroskedasticity-consistent covariance matrix estimators 712 with improved finite sample properties. J. Econom. 29, 305–325 (1985). 713 93. V. Thomas et al., Instant killing of pathogenic chytrid fungi by disposable nitrile gloves 714 prevents disease transmission between amphibians. PLoS ONE. 15, e0241048 (2020). 715 716 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 17 Figures and Tables 717 718 Figure 1. Invasion of Batrachochytrium dendrobatidis (Bd) in the global epicenters of 719 amphibian declines. A) Hypothetical spatiotemporal spread of Bd in the northern Andes and 720 adjacent Central America, approximated by Bd presence records and years when likely Bd-721 induced host declines were first observed. Information derived from (12, 31), updated with ‘year 722 last seen’ of highly susceptible harlequin toads from (26) and additional records from (25) and this 723 study. B) Hypothetical spatiotemporal spread in Australasia and adjacent SE Asia, derived from 724 Bd presence data and temporal onset of host declines reported in (5, 13, 79). Note that Bd might 725 be native in adjacent SE Asia, but surveys were only carried out post 2000. C) Bd sampling in the 726 Sierra Nevada de Santa Marta and adjacent Colombian versant of the Serranía de Perijá (SP) 727 along the Venezuelan border; gray circles are Bd-negative, the yellow dot is Bd-positive (black 728 dots are major cities). Black inset in A) shows location of C) and Fig. 3A and 3B. Black inset in B) 729 shows the location of Fig. 3C and 3D. D) Maximum likelihood phylogeny of Bd based on a 730 concatenated alignment of 26 genome-wide loci totaling 6,035 bp generated by sequence capture 731 from the positive sample collected from the SP (see Methods). Values indicated on the nodes are 732 derived from ultrafast bootstrap and SH-aLRT test, respectively. Nodes <90% bootstrap support 733 were collapsed. The SP sample (highlighted in bold) is placed phylogenetically in the 734 hypervirulent GPL clade, and this site is proximal to the Bd-free SNSM, and may be the most 735 likely source of colonization of the SNSM by Bd. 736 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 18 737 Figure 2. Host population genomics find no genomic signatures indicative of past Bd-738 induced declines in the Sierra Nevada de Santa Marta. A) Observed heterozygosity (turquoise 739 dots) exceeds heterozygosity expected under Hardy-Weinberg equilibrium (green triangles) 740 averaged across SNPs per individual in the representative species, Atelopus nahumae. B) 741 Frequency of consecutive homozygous sections in the genome over all individuals. C) Tajima's D 742 for the A. nahumae population calculated over 500,000 bp windows for scaffolds ≥ 10 mbp of 743 each individual. Photo: Jaime Culebras, Photo Wildlife Tours. 744 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 19 745 Figure 3. Niche models recover environmental barriers to pathogen introduction and 746 suggest high risk for community declines following Bd invasion in the last major Bd-free 747 refuges. A) Environmental suitability for Bd presence in the Sierra Nevada de Santa Marta 748 (SNSM). Lowland climatic barriers of low environmental suitability towards the adjacent Serranía 749 de Perijá and Andes are the likely cause of contemporary pathogen absence in the SNSM. B) 750 Climatic envelope of Bd-driven declines suggests high risk for SNSM endemic amphibian 751 communities. C) Environmental suitability for Bd presence in New Guinea (NG). D) Climatic 752 envelope of Bd-driven declines suggests high risk for montane amphibian communities in NG. 753 Warmer colors suggest higher suitability. 754 755 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint 20 756 Figure 4. Experimental infection of three amphibian species from the Sierra Nevada de 757 Santa Marta, Colombia. A) Atelopus laetissimus, tolerant to Bd infection, B) Bolitoglossa 758 savagei, low susceptibility to Bd infection with delayed mortality underscoring potential reservoir 759 function, C) Cryptobatrachus boulengeri, high susceptibility to Bd infection. Triangles depict 760 mortality (euthanasia following chytridiomycosis-induced loss of self-righting ability). D) 761 Percentage of body mass retained over the infection trial suggesting no Bd-induced effects. 762 Photos: Amadeus Plewnia and Jaime Culebras, Photo Wildlife Tours. 763 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 4, 2026. ; https://doi.org/10.64898/2026.02.02.703360doi: bioRxiv preprint