Clarifying the timing of initial microbial colonization is essential for determining whether and how priority effects mediate community assembly of vertically transmitted microbes in early life, or whether these microbes arrive into an already established microbial landscape. We used non-sterile frogs of our captive laboratory colony to determine the timing of microbial colonization of poison frog embryos. Embryos within the vitelline membrane exhibit a characteristic C-shaped posture due to the membrane’s restriction on their movement and adopt a straight posture after hatching (Figure 1—figure supplement 1). First, we qualitatively examined bacterial presence in Rv egg jellies and embryos across tadpole developmental stages using a broad-range PCR targeting the 16 S rRNA gene. We manually separated embryos in the vitelline envelope from the surrounding jelly and processed each for DNA extraction (Figure 1A). Bacteria were detected in egg jellies across all developmental stages (formation of the dorsal lip to hatchling) but were not detected in embryos prior to hatching from the vitelline membrane (N=12) (Figure 1—figure supplement 1B). After hatching, bacteria were detected in tadpoles that remained in the jelly and in tadpoles that were transported by their caregiver (Figure 1—figure supplement 1B). Next, we confirmed differences in bacterial load between jelly and embryo by examining variations in 16 S rRNA gene copy numbers using a droplet digital PCR (ddPCR) approach suited for low biomass samples (Abellan-Schneyder et al., 2021). We found that unhatched embryos contained on average 2942 copies of the 16 S rRNA gene per µl (min = 742, max = 11584, std = ± 3086) while jellies contained on average 6128 times as many (min = 457417, max = 66115451, mean = 18029604 ± 18586471; Kruskal-Wallis (KW): chi-squared=18.6667, df = 2, p<0.00001) (Figure 1B, Figure 1—figure supplement 1C, Supplementary file 1a). 16 S rRNA gene copy concentrations were on average 8310 times higher in hatched relative to unhatched embryos (min = 737340, max = 48159335, mean = 24448337 ± 33532414; KW: p=0.0195), but did not differ significantly from concentrations detected in the jelly (KW: p=0.5) (Figure 1—figure supplement 1C, Supplementary file 1a).

We used a cross-foster design to test the hypothesis that frogs carrying tadpoles on their backs serve as a source of microbes for communities on tadpole skin. Siblings of the same clutch (N=6 clutches from 2 Rv pairs) were randomly assigned to three groups: (1) not carried (N=6 tadpoles), (2) carried on the back of their biological parent for 6 hr (N=9 tadpoles), or (3) carried by a heterospecific poison frog (Oophaga sylvatica, Os) for 6 hr (N=9 tadpoles) (Figure 1C). For this design, sequence-based surveys of amplified 16 S rRNA genes were used to assess the composition of skin-associated microbial communities from tadpoles and their adult caregivers (i.e. the frogs carrying the tadpoles, typically referred to as ‘transporting’ frogs). We found that bacterial community composition on caregiver skin was primarily shaped by host species (Rv versus Os; Adonis: Bray-Curtis: F_{1, 12} = 5.6538, R²=0.33949, p=0.001; unweighted Unifrac: F_{1, 12} = 2.0078, R²=0.15435, p=0.013), allowing us to distinguish between these potential sources of tadpole-colonizing microbes (Figure 1—figure supplement 2A). Indeed, we found that after 6 hr of being carried on the back, bacterial community composition on Rv tadpole skin was influenced by caregiving species (Rv versus Os; Adonis: Bray-Curtis: F_{1, 12} = 1.75, R²=0.099, p=0.017; Unifrac: F_{1, 12} = 1.67, R²=0.095, p=0.015). Despite their distinct compositions, we observed substantial overlap between communities of tadpoles carried by Os and Rv (Figure 1—figure supplement 2B), possibly reflecting similarities in tadpoles’ skin communities arising from a shared clutch environment after hatching until transport.

We used ‘ourcetracker’ (Knights et al., 2011) to determine the relative contribution of the transporting frogs as a source of bacteria detected in tadpoles' skin after 6 hr of transport. Tadpoles of a clutch that were transported by Os always shared a higher proportion of their communities with this heterospecific caregiver than siblings that were not transported (KW: chi²=14.4696, df = 2, p=0.0003; difference in proportion: median = 39.33, min = 0.872%, max = 68.478%) or siblings transported by the biological parent (KW: p=0.015; percentage of higher proportion: median = 37.07%, min = 0.874 %, max = 72.97%) (Figure 1D, Supplementary file 1b). Similarly, all but two transported tadpoles shared a larger proportion of their communities with the transporting species than with the non-transporting species (p=0.001) (Supplementary file 1b). Transported tadpoles (N=18) generally shared higher proportions of their skin communities with the transporting species than did non-transported tadpoles (N=6) (Kruskal-Wallis chi²=19.4029, df = 2, p=0.0001) (Figure 1—figure supplement 2D). Non-transported tadpoles were colonized by between 5 and 57 Amplicon Sequence Variants (ASVs) (median = 12). We conducted a separate analysis defining non-transported tadpoles as potential microbe sources to represent community proportions that were acquired in the clutch before caregiver contact (Figure 1—figure supplement 2C). We found that transported tadpoles shared between 0 and 46.99% (median = 13%) of their communities with non-transported tadpoles. Sourcetracker detected higher proportions of clutch-acquired microbes in tadpoles transported by Os (pairwise Wilcoxon test, p=0.02). This likely reflects the higher degree of similarity between the source clutch and Rv, as the clutch remained in the Rv tank during development, making it harder to distinguish between these sources. Overall, our results suggest that microbes are vertically transmitted during tadpole transport in a poison frog.

To investigate how vertical transmission of microbes during tadpole transport affects microbial skin communities of tadpoles in an ecologically relevant context, we studied the skin microbiome composition and diversity of Rv tadpoles and adults in a natural population. To assess the broader relevance of our findings to frog species that transport their tadpoles, we also sampled two sympatric species with differing reproductive strategies: Allobates femoralis (Af) and Leptodactylus longirostris (Ll) (Figure 2A). In contrast to Rv that transports its cannibalistic tadpoles to individual pools in bromeliads, Af lives in forest leaf litter and shuttles multiple non-cannibalistic tadpoles to the same pool where they grow up gregariously. Ll is a nocturnal leptodactylid frog inhabiting the rock savanna that does not exhibit tadpole transport but deposits eggs directly in ephemeral rock pools where the larvae grow up together (Figure 2—figure supplement 1).

We compared bacterial communities from 137 skin swabs of adult frogs, tadpoles, and the aquatic environment belonging to the tadpoles of these three species. We sampled 44 adults and 21 tadpoles (Gosner stages 29–41, categorized as ‘medium’ and ‘large’) of the species Rv, 10 adults and 14 tadpoles (Gosner stages 34–41, categorized as ‘medium’ and ‘large’) of the species Ll, and 10 adults and eight tadpoles (Gosner stages 25–26, categorized as ‘small’) of the species Af (Supplementary file 1c). Additionally, we monitored the growth of 184 Rv tadpoles in separate bromeliad pools for up to 82 days. Out of 45 tadpoles that were monitored for over 50 days, 13 (three ‘small’ and 10 ‘medium’) tadpoles did not advance in their sizing category, reflecting that the complete larval development in this population spans multiple months. Cannibalistic Rv tadpoles grow up in isolated pools that were sampled as single replicates when collecting the tadpoles. Due to their development in groups, Af and Ll tadpoles were collected from fewer water bodies and represent a narrower developmental window (stages 34–41 for Ll and stages 25 and 26 for Af). If multiple tadpoles were collected from the same aquatic environment (Af and Ll tadpoles), we sampled triplicates of the respective environment.

For each species, the aquatic environment of the tadpoles displayed a higher average phylum-level diversity than the inhabiting tadpoles or adults (Supplementary file 1c). As previously reported for Neotropical frogs (Hughey et al., 2017), ASVs belonging to Phylum Proteobacteria dominated the microbial community across all species and life stages (mean relative abundance: 70.9%), followed by Bacteroidota (10.3%), Actinobacteriota (4.1%), Firmicutes (4%), Verrucomicrobiota (2.3%), and Planctomycetota (1.1%) (Figure 2—figure supplement 2). The abundance of the frog pathogen Batrachochytrium dendrobatidis (Bd) in the rock savanna was low, with detection in two samples of Rv adults and one Rv tadpole, but no positive Ll sample. We detected Bd on one adult Af individual collected in primary forest but not on tadpoles of this species collected in the same area. In all species, tadpole communities contained fewer taxa with known Batrachochytrium-growth inhibiting function than did environmental samples or adults (Supplementary file 1c).

We found no significant differences in the observed ASV richness, diversity (Shannon), or evenness between tadpoles and adults of the poison frogs Rv and Af. In contrast, Ll tadpoles were significantly less diverse than Ll adults (Figure 2B; Supplementary file 1d). We followed up on this finding by comparing the diversity of tadpole-associated communities across species and found higher microbial diversity and evenness in Rv and Af tadpoles compared to Ll tadpoles (Figure 2C). Differences in community diversity of tadpoles were not a reflection of habitat diversity: microbial communities associated with Rv aquatic environments were less diverse and even compared to those in Af and Ll habitats (Figure 2D). These variations may, therefore, reflect differences in life history traits among the three species.

To assess how microbes transmitted during parental care shape communities associated with tadpoles, we tested the hypothesis that tadpoles of the transporting species Rv and Af would have communities more similar to conspecific adults compared to tadpoles of the non-transporting species Ll. We first evaluated variation in bacterial communities associated with adults, tadpoles and their microenvironment by using a principal coordinate analysis (PCoA) to construct ordinations based on Bray-Curtis dissimilarities to explore factors that influence skin community composition (Figure 3A). Tadpoles of all species clustered separately from adults, each other, and their aquatic environment (PERMANOVA on genus level, df = 8, R²=0.52909, F=18.117, p adj = 0.001) (Figure 3A and B; Supplementary file 1e). Consistent with findings in pond-spawning anurans (Kueneman et al., 2014; Prest et al., 2018), skin communities varied significantly across life stages in all species (PERMANOVA on genus level: species: df = 5, R²=0.29674, F=11.139, p=0.001, life stage: df = 2, R²=0.23148, F=20.331, p=0.001). Parental care also explained some of the observed variance (PERMANOVA on genus level: df = 2, R²=0.07513, F=5.4831, p=0.001). To further examine patterns of similarities between life stages, we compared the core communities of adults, tadpoles, and their aquatic environments across a range of prevalence and abundance cutoffs prevalence 100% or 75%; relative abundance no cutoff, ≥0.1% (low) and ≥1% (high) (Supplementary file 1f). Independent of parenting strategy and species, tadpoles shared core communities of their microbiome with their environments, but not with adult individuals. Adult poison frogs never shared core taxa with the aquatic environment of their tadpoles (Figure 3C; Supplementary file 1f), even though adults were commonly found in the nurseries. In contrast, the aquatic environment contributed substantially to the skin community of adult Ll, which shared high- and low-abundance (≥0.1%) core genera with the pond water, but not with the pond-inhabiting tadpoles (Figure 3C; Supplementary file 1g).

To assess if microbes transmitted during tadpole transport constitute a stable part of skin communities associated with tadpoles, we compared community distances between skin communities of tadpoles and adults of the three frog species. We predicted that tadpoles and adults might have more similar skin microbiomes in the transporting species Rv and Af than in the non-transporting species Ll. We found no difference in community distances between tadpoles and adults in Af and Ll (iterated beta regression averaged over 999 random subsamples of similar sample size (IBR): Unifrac: p=0.3). Communities associated with Rv tadpoles were more dissimilar from adult communities than microbiota of Ll tadpoles and adults (IBR: Unifrac: p=0.008). Taken together, our results indicate that microbes acquired during tadpole transport do not create lasting similarities in composition between tadpoles and adults of wild populations.

To test how offspring transport in Rv affected the community composition in tadpoles, we conducted a field experiment where we reared transported (N=10) and non-transported (N=10) tadpoles in artificial cups. Transported tadpoles were collected from the backs of their caregiver (N=8), while we bypassed parental care in the non-transported group by transferring eggs to cups before the tadpoles hatched (Figure 4A). Two out of ten non-transported tadpoles died before sampling. Richness and evenness of tadpole communities did not differ significantly between experimental conditions (ANOVA: Observed ASVs: df = 1, F=2.7234, p.adj=0.3552; Shannon: df = 1, F=1.2067, p.adj=0.8646; evenness: df = 1, F=0.0009, p.adj=2.9313), though PCoA on Unifrac distances revealed differences in microbiome composition attributable to the presence or absence of transport (Figure 4B) (PERMANOVA: df = 1, R²=0.11759, F=2.1321, p=0.002). Compared to transported tadpoles, non-transported tadpoles showed less overlap of genera in their microbial community with their aquatic environment (Figure 4C). Specifically, the genera Pelomonas, Rhodomicrobium (Proteobacteria), Mycobacterium (Actinobacteriota), and Candidatus Koribacter (Acidobacteriota) primarily colonized transported tadpoles, while the genus Limnohabitans (Proteobacteria) was found on non-transported tadpoles. Cetobacterium and Burkholderia-Caballeronia-Paraburkholderia were found to colonize tadpoles from both experimental conditions, but at different abundances (ANCOMBC 2, p<0.05; Supplementary file 2).

We further tested the hypothesis that related tadpole-adult pairs with a history of direct contact would have more similar communities than unrelated tadpole-adult pairs without such a history. One month after transport, we found that skin communities of tadpoles were not more similar to the transporting parent than to unrelated adults of the population (iterated beta regression averaged over 999 random subsamples of similar sample size (N=10) (IBR): Unifrac: p=0.27), confirming our results from the population sampling. However, a Sourcetracker analysis on unrarefied data revealed notable relative contributions of the parental skin community to the community on four transported tadpoles (tadpoles 1, 2, 6, 8) (min = 0% to max = 17%, median = 0%, IQR = 10.5; Supplementary file 1h). We followed up on this finding by directly identifying taxa shared between caregivers and the respective transported offspring without imposing an abundance cutoff. We found that 8 out of 10 tadpoles shared one or two ASVs with the frog that had transported them in relative abundances ranging from 0.02 to 19.3% (Figure 4D; Supplementary file 1h). Shared taxa belong to the genera Akkermansia, Elsteraceae, Aquitalea, Methylocella, Methylobacterium-Methylorubrum, Staphylococcus, Pseudomonas, and Acinetobacter. Seven out of ten shared microbes constituted members of the ten most abundant ASVs found on the respective transporting frog (Supplementary file 1h). Absolute abundances of shared ASVs likely originating from the parental source pool (as identified by Sourcetracker) after one month of growth ranged from 7804–172,326 sequence copies per tadpole (Supplementary file 1i). Finally, we determined the success of shared ASVs in colonizing transported and non-transported experimental tadpoles. Even though at least six (min: 6, avg: 6.38, max: 7) of the ten ASVs were consistently present in the aquatic environment of non-transported experimental tadpoles, only one (family Enterobacteriaceae) successfully colonized their skin (Figure 4D, not transported tadpoles; Supplementary file 1j).

Captive-bred Rv of our laboratory colony are kept non-sterile and reproduce year-long. Males attend the terrestrial clutch of 3–5 eggs and transport larva to water when they reach Gosner stage 24–25 where they hatch from the vitelline membrane. Rv were housed in pairs and provided with horizontally mounted film canisters as egg deposition sites, and film canisters filled with water (treated with reverse osmosis R/O Rx, Josh’s Frogs, Owosso, MI) for tadpole deposition (Goolsby et al., 2023).

We used individually packed sterile transfer pipettes (Samco, Thermo Scientific, 7.7 mL, cat # 202-1SPK) with a cut tip to move eggs and embryos. All tools were treated with 8.25% Sodium Hypochlorite (CloroxPro) and wiped down with 70% Ethanol before use and between tissues and tadpoles. The experimenter always wore gloves that were changed between individuals and disinfected with 70% Ethanol before handling frogs.

Triplicates of five different developmental stages (Gosner, 1960) before (stages 10, 18, 23, 24) and after hatching from the vitelline membrane (stages 24, 25) were collected from three different breeding pairs. We manually separated the jelly from the egg sac with autoclaved forceps, a procedure commonly done in preparation for microinjections. The embryo in the vitelline membrane was transferred to a small petri dish with sterile, filtered deionized water and opened using forceps. Tadpoles were transferred to a new petri dish with clean water twice to remove transient bacteria from the skin and were then euthanized and homogenized. Organic matter was removed from the jelly, and both jelly and tadpole tissues were flash frozen and stored at –80°C until DNA isolation. Isolation of DNA was performed as detailed below, and 1 µL of isolated DNA was used for PCR amplification of the full length 16 S rRNA region (~1500 bp) with 1 µL of 10 µM forward 27 F (5’-AGAGTTTGATCMTGGCTCAG-3’) and reverse primer 1492 R (5’-TACGGYTACCTTGTTAYGACTT-3’) (Fredriksson et al., 2013), 12.5 µL of OneTaq Hot Start Quick Load polymerase (New England Biolabs) and 9 µL water. We followed an amplification protocol with denaturation for 30 s at 94°C followed by 30 cycles of denaturation for 15 s at 94°C, annealing for 30 s at 58°C and extension for 2 min at 68°C, and a final extension period of 5 min at 68°C. To determine the presence or absence of amplification, we visualized 5 µL of each PCR product on a 1% Agarose gel (90 mV for 90 min) with 2 µL of 1 kb GeneRuler for sizing.

Cross-fostering tadpoles onto non-parental frogs has been used previously to study navigation in poison frogs (Pašukonis et al., 2017). According to our experience, successful adherence to both parent and heterospecific frogs depends on the developmental readiness of tadpoles, which must have retracted their gills and be capable of hatching from the vitelline envelope through vigorous movement. Another factor influencing cross-fostering success is the docility of the frog during initial attachment, as erratic movements easily dislodge tadpoles before adherence is established. Rv are small, jumpy frogs that are easily stressed by handling, making experimental fostering of tadpoles- even their own- impractical. Therefore, we favored an experimental design where tadpoles initiate natural transport, and parental frogs pick them up with a 100% success rate. We chose the poison frog Os as foster frog because adults are docile, parental care in this species involves transporting tadpoles, and skin microbial communities differ from Rv- a critical prerequisite for our Sourcetracker analysis. The use of the docile Os as the foster species enabled a 100% cross-fostering success rate, with no notable differences in adherence strength of tadpoles after 6 hr.

Siblings of one clutch were randomly assigned to one of three treatments: (1) no transport, (2) transport by the biological caregiver, or (3) transport by the heterospecific surrogate frog Os. Tanks of Rv breeding pairs were monitored for clutches of at least three eggs daily. Embryos assigned to group one or three were placed into a petri dish in a separate tank without access to adults when tadpoles reached a developmental stage close to hatching (Gosner stage 22) to bypass transport by the biological father. The remaining eggs in the parental tank were monitored with security cameras (Wyze v3, Goolsby et al., 2023) that notify the user of any movement in the canister. After tadpole pickup by the biological father, all water canisters for tadpole deposition were removed from the tank to standardize transport time to 6 hr for both groups. Siblings of group three were transferred to a new petri dish, hatched using a sterile brush, and directly transferred to the back of a surrogate frog (as described in Pašukonis et al., 2017) that was caught in a fresh plastic bag and rinsed with sterile water. The surrogate frog was placed into a plastic containment (Sterilite 16428012) with moist, autoclaved paper towels for the duration of the experiment. After 6 hr, the transporting biological father and the surrogate frog were caught in a fresh plastic bag. Each tadpole was removed from the back and washed twice to remove transient bacteria by transferring them to new petri dishes with 10 ml sterile water. Non-transported tadpoles were removed from the jelly and washed in a similar way. Each frog was rinsed with sterile water, and a skin swab of their back was collected using a sterile Puritan applicator (Puritan Medical Products, 25–206 1PD BT) (Caty et al., 2024b). All tadpoles were euthanized, their skin was collected, and skin and applicators were stored at –20°C until DNA extraction. We chose to sample whole skin instead of swabs to detect small amounts of transferred bacteria on and in the skin.

Oophaga sylvatica (Os) (Funkhouser, 1956; CITES Appendix II, IUCN Conservation status: Near Threatened) is a large, diurnal poison frog (Family Dendrobatidae) inhabiting lowland and submontane rainforests in Colombia and Ecuador. While male Os care for the clutch of up to seven eggs, females of this species transport 1–2 tadpoles at a time to water-filled leaf axils where tadpoles complete their development (Pašukonis et al., 2022; Silverstone, 1973; Summers, 1992). Notably, females return regularly to these deposition sites to provision their offspring with unfertilized eggs.

Ranitomeya variabilis (Rv) (Zimmermann & Zimmermann 1988; CITES Appendix II, IUCN Conservation status: Least Concern) is a small diurnal poison frog (Family Dendrobatidae) with male uniparental care that lives largely arboreal and uses bromeliads as a resource for reproduction. Adult frogs defend territories, are polygamous and typically lay clutches of 3–4 eggs into small arboreal bromeliads (Catopsis berteroniana) that are abundant on Clusia trees in the rock savanna. Male frogs transport hatched tadpoles to individual water bodies that form in the leaf axils of the large terrestrial tank bromeliads (Aechmea aquilega). The cannibalistic tadpoles grow up in individual pools of ~80 ml (Poelman et al., 2013) where they feed on algae and leaf debris until they complete their development after about 3–5 months (Poelman and Dicke, 2007).

Allobates femoralis (Af) (Boulenger, 1884; CITES Appendix II, IUCN Conservation status: Least Concern) are polygamous diurnal poison frogs (Family Aromobatidae) that inhabit the understory of primary tropical forests. Females are attracted by advertisement calls of the territorial males and deposit clutches of 11–25 eggs in the leaf litter (Fischer et al., 2020; Stückler et al., 2019; Weygoldt, 1980). Fathers care for their offspring by attending clutches and shuttle hatched-tadpoles to water bodies where the social tadpoles grow up sharing a pool (Lescure, 1976; Ringler et al., 2013; Ringler et al., 2018).

Leptodactylus longirostris (Ll) (Boulenger, 1882; IUCN Conservation status: Least Concern) are medium sized nocturnal leptodactylid frogs (Family Leptodactylidae) that inhabit the rock outcrops. They shelter in rock crevices and ground covering vegetation during the day and dwell in seasonal rock pools on the Inselberg plateau during nights. They breed in the rainy season and lay their eggs in ephemeral rock pools. After clutch deposition, they do not care further for their tadpoles, which collectively grow up in the deposition pond and feed on algae and plant debris.

The natural study site of Rv and Ll is situated on top of the mountain 'Inselberg' in vicinity to the Centre Nationale de la Recherche Scientifique managed research station (4°5’ N, 52°41’ W) within the Nature Reserve ‘Les Nouragues’ in French Guiana (Bongers et al., 2011). Patches of Clusia trees are separated by bare granite rocks and exposed to extreme environmental conditions (e.g. temperature oscillations between 18°C–75°C; Sarthou, 2001). Rv and Ll (IUCN Conservation status: Least Concern) inhabit the granite outcrop of the study site, Af inhabit forest floors of primary terra-firme forests surrounding the campsites ‘Inselberg’ and ‘Saut Pararé’ (4°02’N/52°41’W) in the Nouragues reserve.

Frogs were caught in fresh plastic bags upon encounter, rinsed with sterile water to remove transient bacteria and soil particles and then swabbed with 20 strokes each on the dorsal and lateral sides as well as on each leg and between the toes using Sterile Polyester Applicators (Puritan Medical Products, 25–206 1PD BT). Due to limitations in the availability of dry ice at the remote study site, applicators were directly transferred into 800 µL of buffer CD1 (a lysis buffer containing chaotropic salts) of the Qiagen PowerSoil Pro extraction kit and frozen until DNA extraction. Previous studies have found no difference in OTU richness and evenness between native frozen and lysis buffer-stored swabs (Hallmaier-Wacker et al., 2018).

To avoid disturbing microbial communities of tadpoles and their habitats before sampling, free-swimming tadpoles were visually categorized as ‘small’ (with a body length of less than 6 mm, typically Gosner stages 25 and 26), ‘medium’ (with a body length exceeding 6 mm but no visible dorsal pattern, stages 26–33), ‘large’ (visible yellow dorsal pattern, stages 34–41) or ‘metamorph’ (front limbs visible, stages 42–46). Tadpoles of stages 29–41 (Gosner) were collected from bromeliad pools for swabbing when they at least doubled the body size of freshly hatched tadpoles, at the earliest four weeks after deposition. The water and sediment of the nursery was sampled prior to tadpole extraction by submerging a sterile applicator tip into the leaf axil pool and moving it over all plant material contained in the water for 10 s. Rv tadpoles were extracted from bromeliad leaf axils with a custom-designed vacuum extractor consisting of an inlet hose wide enough to allow the passage of a late-stage tadpole and a suction hose with air filter, both tightly connected to the lid of a one-liter bottle. The inlet hose was submerged into the bromeliad pool while applying suction to the second hose to create a vacuum that transferred the pool content into the collector. After each use, all parts of the collector were cleaned with soapy water and rinsed, followed by a sodium hypochlorite disinfection for 1 hr and a rinse in 70% Ethanol as described for the dissection tools. After collection, tadpoles were transferred to a petri dish with sterile water twice, using individual sterile transfer pipettes with a cut tip. For swabbing, tadpoles were collected in the bulb of a new sterile pipette. The water was discarded, and the tadpole was swabbed on all body parts with a sterile applicator tip by moving the tip over its body for 10 s. We further processed swabs in the same way as adult swabs. Af tadpoles of stage 25–26 (Gosner) were collected from an artificial pool in the vicinity of camp ‘Saut Parare.’ Ll tadpoles of Goser stage 34–41 were collected from three natural pools in the study plot ‘Inselberg.’ We sampled triplicates of each pool containing multiple tadpoles. All tadpoles were processed for swabbing in a similar way as Rv tadpoles.

To directly compare parental frogs with transported and non-transported tadpoles while reducing predation and variation in the quality of their nurseries, we reared 10 tadpoles that were collected from the back of eight caregivers and ten tadpoles that hatched from eggs without experiencing transport in polypropylene cups (USP #77876, 60 ml). Each cup was punctured at the 50 ml water level, covered with a sterilized mosquito net secured with a rubber band to minimize predation, and attached to the stem of a Clusia tree. All tadpoles received water collected from bromeliads as their initial aquatic environment with Clusia leaves as shelter and food. Cups were exposed to rain and during dry periods, rainwater was collected and mixed with leaf litter to refill cups. Triplicates of water samples were obtained using the procedure described for bromeliad water. We caught transporting frogs upon encounter, transferred tadpoles from their backs to cups, and swabbed the parenting frog as described above. To avoid transport, eggs from five clutches were transferred to cups before tadpoles hatched. Tadpoles were sampled after growing for at least 26 days (min: 26, max: 77), when all tadpoles of a group exceeded ~5 mm in body size, as smaller individuals exhibited increased mortality after swabbing in previous lab experiments. After swabbing, tadpoles were measured and photo documented.

The Qiagen PowerSoil Pro Kit was used to extract DNA from all swabs and tissues. The protocol was adapted for use with swabs as described previously (Caty et al., 2024a). DNA concentrations were quantified using a Qubit. Samples were pooled by volume and the V4 region of the 16 S rRNA gene was amplified using 515 F (GTGYCAGCMGCCGCGGTAA) and 806 R (GGACTACNVGGGTWTCTAAT) primers (Bletz et al., 2017) and barcoded using standard Illumina unique dual indices (UDIs). We performed two separate sequencing runs: Laboratory collected samples were sequenced in a 2x300 nt paired-end configuration on an Illumina MiSeq v3 run, field collected samples in a 2x250 nt paired-end configuration on a NovaSeq 6000 SP Flowcell (Roy J Carver Biotechnology Center, University of Illinois).

We annotated in-line barcodes based on the first 7 bases of each sequencing read (umi-tools, Smith et al., 2017), split out reads that matched each known barcode combination (grep, Project GNU, 1998), trimmed the remaining primer sequences from the sequencing reads (cutadapt, Martin, 2011) counted the number of sequencing reads in each file and removed files (including negative controls) with low read numbers (<100 reads) from the dataset. We processed the remaining reads with the R Divisive Amplicon Denoising Algorithm package ‘dada2’ (version 1.28.0) (Callahan et al., 2016). Taxonomy was assigned using the Silva 138 database (Ref NR99) (Quast et al., 2013). The count table, taxonomy table, and sample-associated data were integrated into a phyloseq object using the R package ‘phyloseq’ (version 1.44.0) (McMurdie and Holmes, 2013) for further data analysis. Any ASVs detected in control sequencing reactions without DNA as well as ASVs with a phylum designation that was ‘NA,’ Eukaryotic, belonging to the family of Mitochondria or the class of Chloroplasts were removed from the datasets. Samples with very low reads (clearly below the first quartile) were excluded from each group.

Amplification targeted the second internal transcribed spacer (ITS2) region of the fungal ribosomal cistron using primers ITS3 and ITS4 (White et al., 1990). Full primer sequences included Illumina Nextera transposase adapters (underlined) and were as follows: ITS3-5’-TCG TCG GCA GCG TCA GAT GTG TAT AA GAG ACA GGC ATC GAT GAA GAA CGC AGC-3’ and ITS4-5’-GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA GTC CTC CGC TTA TTG ATA TGC-3’. Reaction mixes contained 1 x AccuStart II PCR SuperMix (Quantabio), 0.4 µM forward and reverse primers, and 2 µl of template DNA. The thermal cycling regime consisted of an initial denaturation and enzyme activation step at 94°C for 3 min, followed by 30 cycles of 94°C for 45 s, 50°C for 60 s, and 72°C for 90 s, with a final extension step at 72° for 10 min. Ten additional PCR cycles were used to tag diluted amplicons with unique dual indices (Illumina; cycling regime as above but with annealing at 54°C). Amplicons were pooled, cleaned using AMPure XP beads (Beckman Coulter), and sequenced on an Illumina MiSeq instrument (2×300 nt) at the DNA Services Lab, Roy J. Carver Biotechnology Center, University of Illinois at Urbana-Champaign. Reads processing was consistent with the DADA2 ITS workflow (https://benjjneb.github.io/dada2/ITS_workflow.html). Raw reads were trimmed of non-biological sequence using cutadapt (Martin, 2011). DADA2 was used to filter and truncate the reads, infer amplicon sequence variants (ASVs), and remove chimeras (Callahan et al., 2016). Taxonomy was assigned using DADA2’s implementation of the RDP naïve Bayesian classifier (Wang et al., 2007) against a UNITE reference database (version 9.0; Abarenkov et al., 2023). The ASV count table, taxonomy table, and sample-associated data were integrated into a single object in R using the phyloseq package (McMurdie and Holmes, 2013).

All statistical analyses were performed in R (v 4.3.2) (R Core Team, 2023), and ‘ggplot2’ (v 3.4.4) (Wickham, 2016) was used for visualizations if not stated otherwise. Illustrations of all species were created in Adobe Illustrator (v 28.3) (Adobe Inc, 2024) and all figures were compiled in InkScape (v 1.1) (Inkscape Project, 2021).