Critical periods of development are time points that are especially sensitive to disruptions. Critical periods are important for microbiome assembly and the development of an effective immune system and, therefore, can have large and lasting impacts on host health, even through later life stages. Here, we investigated how a disruption to the microbiome of amphibians (treatment with a cocktail of six antibiotics and one antifungal compound) during early life and subsequent introduction to microbes at different developmental stages influences microbiome assembly and the development of the immune system (lymphoid tissues thymus and spleen). We found that antimicrobial treatments and introduction to microbes altered microbiome assembly (total microbial richness, antifungal microbial richness, composition, and relative abundances) and these changes were dependent on the timing of microbial introduction. Tadpoles treated with antimicrobials and then introduced to microbes at different developmental stages also had higher scaled abundances of bacteria in the phylum Actinobacteriota. However, after 7.5 weeks of tadpole development, we found no effects of treatment on lymphoid organ (thymus and spleen) size or on lymphoid cell counts. Overall, these results suggest that a disruption to the microbiome during early development, and more specifically, the length of the disruption and timing of reintroduction to microbes, can have significant impacts on microbiome assembly, potentially leading to long-term impacts on host health.

Embryo Transport

We received X. tropicalis embryos from the Marine Biological Laboratory, National Xenopus Resource (Woods Hole, MA, USA; Gensollen et al. 2016). The embryos arrived at approximately Nieuwkoop and Faber stage 16 (NF stage; approximately Gosner stage 14; Nieuwkoop and Faber 1994, Zahn et al. 2022). NF stages describe Xenopus development, similar to Gosner staging; however, NF stages are more specific and a better fit for Xenopus developmental studies (Nieuwkoop and Faber 1994; Zahn et al. 2022). Upon arrival, we randomly separated the embryos into two treatment groups and placed them into 50 mL conical tubes (approximately 60 embryos per conical tube). We moved one group inside a laminar flow biosafety 2 cabinet and kept the other outside the biosafety cabinet (i.e., in nonsterile conditions). We rinsed each group three times with sterilized (i.e., autoclaved) deionized (DI) water and then placed the group inside of the biosafety cabinet into an antimicrobial (AMX) cocktail (described below) to soak for 4.5 h (Miller et al. 2023). We treated the nonsterile control group outside the biosafety cabinet in the same manner but replaced the AMX cocktail with an equal volume of autoclaved water (Miller et al. 2023). The temperature during development was approximately 22°C, and we previously confirmed the temperature inside the biosafety cabinet did not differ from the temperature outside the biosafety cabinet (Miller et al. 2023).

Animal Husbandry

We allowed the tadpoles in each group to develop in Petri dishes for approximately 2 weeks and then moved them to three stainless steel tanks per treatment group that were either autoclaved (for the sterile treatment groups) or not (approximate starting density of 45–50 tadpoles per 16.5 × 16 × 26.7 cm tanks; Thunder Group, CA, USA). We fed the tadpoles Xenopus larvae tadpole powder (Carolina Biological; NC, USA) ad libitum. We sterilized the tadpole powder for the animals in the sterile treatment groups via gamma irradiation with a dose of 15 kGy, which we previously validated for sterility (Miller et al. 2023). We also spot cleaned the Petri dishes and tanks as needed with sterile transfer pipettes and added fresh sterile or nonsterile water at least once per week. To prepare the tadpole water, we allowed DI water to dechlorinate for at least 24 h prior to use and then added an aquarium salt mixture (Instant Ocean; Blacksburg, VA, USA) until the conductivity of the water was approximately 1200 microsiemens. We then autoclaved the water for the sterile groups in groups of 1 L glass media bottles (Corning Life Sciences Pyrex; USA) and allowed it to cool to room temperature before adding it to the tanks.

Antimicrobial Treatments

The antimicrobial (AMX) cocktail consisted of 500 μL of penicillin G: streptomycin (10,000 units mL−1: 10 mg mL−1; antibiotic), 50 μL of amphotericin B solution (250 μg mL−1; antifungal), 200 μL kanamycin sulfate (25 ug mL−1; antibiotic), 0.53 μL of sulfamethoxazole: trimethoprim (13.3 mg L−1: 2.67 mg L−1; antibiotic), and 1.2 mg enrofloxacin (final concentration 30 mg L−1; antibiotic) added to an autoclaved 250 mL glass beaker with autoclaved DI water to create a total volume of 40 mL (Miller et al. 2023). We then homogenized and filter-sterilized the AMX cocktail inside a biosafety cabinet using a 0.22 μm sterile syringe filter. Following the treatments with either AMX or the sham solution, we rinsed the embryos three additional times with sterile water. We then moved the embryos to their respective sterile Petri dishes either inside or outside the biosafety cabinet with 50 mL of either sterile or nonsterile water with aquarium salt, depending on treatment group. After the AMX and sham treatments, we randomly and evenly separated the embryos (approximately N = 220–240 embryos per treatment group) inside the biosafety cabinet into three additional groups. We used a labeling system, denoted by “AMX_No.”, for each group that reflected if the group received AMX or sham treatments and the length of time in sterile conditions (No. of days). We held the first group inside the biosafety cabinet and in sterile conditions (i.e., fed sterile food, held in sterile water) from embryo arrival to the lab throughout the duration of the entire experiment (i.e., through week 10 of the experiment; AMX). We held a second group inside the biosafety cabinet and in sterile conditions from embryo arrival to the lab until the thymus was developed, or, approximately experimental day 5 (Du Pasquier et al. 2000; Robert and Ohta 2009; Lee et al. 2013), when we then removed the tadpole tanks from inside the biosafety cabinet, fed the group nonsterile food, and maintained them in nonsterile water (AMX_5). This change in husbandry exposed them to the various microbes present in the nonsterile environment. We maintained the next group inside the biosafety cabinet in sterility from embryo arrival to the lab until experimental day 25, when we then removed the tadpole tanks from inside the biosafety cabinet and exposed them to nonsterile conditions (AMX_25). This time point of removal from sterile conditions corresponded with the development of both the thymus and spleen (Du Pasquier et al. 2000; Robert and Ohta 2009; Lee et al. 2013). We also included a nonsterile, conventionally raised treatment group that was not treated with antimicrobials and held in nonsterile conditions for the entirety of the experiment (i.e., fed nonsterile food and maintained in nonsterile water; No AMX). A visual schematic of the experimental design is included in the Supporting Information (Figure S1).

Tadpole Microbiome DNA Extractions

Five weeks after initial embryo arrival to the lab, we haphazardly selected N = 10 tadpoles per treatment group (approximately NF stage 55–56; Nieuwkoop and Faber 1994) for whole-body 16S rRNA targeted amplicon microbiome sequencing analysis (Miller et al. 2023). We first filter-sterilized MS-222 and humanely euthanized the tadpoles (National Research Council Committee on Care, and Use of Laboratory Animals 1986). We measured tadpole mass to the nearest 0.1 g and snout-vent length (hereafter, SVL) to the nearest 0.1 mm. We homogenized the tadpoles using the Disruptor Genie homogenizer (Scientific Industries, Bohemia, NY, USA) and used the QIAmp PowerFecal Pro DNA Kit (Qiagen, Valencia, CA, USA) to extract DNA. We homogenized the tadpoles and extracted the DNA following the Qiagen protocol associated with the QIAmp PowerFecal Pro DNA Kit.

Sequencing and Sequence Data Pre-Processing

We sent extracted DNA to the Idaho State University Molecular Research Core Facility (RRID:SCR_012598) for sequencing and included N = 4 water blanks (i.e., molecular grade water extracted using the same QIAmp DNA Kit). For sequencing, a bacterial 16S rRNA gene fragment from the V4 region was polymerase chain reaction (PCR) amplified from each DNA extract and was sequenced on the Illumina MiSeq (Illumina Inc., San Diego, California, USA) using methods previously described (Miller et al. 2023). For all microbiome data apart from the antifungal richness analysis (described below), demultiplexed FASTQ files, with primers and adapters removed, were processed in R (v. 4.4.1; R Core Team, n.d.) using the dada2 pipeline (v. 1.32.0; R Core Team, n.d.). Reads were trimmed, merged, and chimeras removed. We performed taxonomic assignment using the SILVA bacterial 16S rRNA database (v. 138.1; Quast et al. 2012).

Tadpole Dissections

We haphazardly selected N = 10 tadpoles from all treatment groups approximately 7.5 weeks after initial embryo arrival in the lab. We chose this time point because the tadpoles were large enough to dissect while not yet actively going through metamorphosis, when their lymphocyte numbers are known to decrease (Rollins-Smith 1998). We euthanized the tadpoles using MS-222 and measured mass and SVL (Miller et al. 2023). We then measured the size of both the left and right thymus using a dissecting microscope and an eyepiece with a reticle at 30× total magnification (AmScope 10× Microscope Eyepiece with Reticle 23 mm; AmScope, CA, USA). We dissected both the thymuses and added them to 200 μL of sterile amphibian phosphate-buffered saline (APBS; 25 mL H2O, 100 mL PBS; Quast et al. 2012). We homogenized the thymuses in the liquid by gently pulling them apart using fine-tipped forceps and counted the thymus cells present in the solution using a hemocytometer (Edholm 2018; Rollins-Smith et al. 1996). We dissected the spleens using fine-tipped forceps and micro-dissection scissors, floated them in 200 μL of APBS, and measured the diameter using a dissecting microscope and an eyepiece with a reticle at 30× total magnification (Edholm 2018; Rollins-Smith et al. 1996). To calculate the diameter of both the thymuses and spleens, we converted the eyepiece reticle measurements to micrometers using a microscope stage calibrator (AmScope Microscope Stage Calibration Slide; AmScope, CA, USA).

Statistical Analyses

We conducted statistical analyses using R (v. 4.0.2; R Core Team, n.d.), choosing to run nonparametric tests when the assumption of normality was violated. We ran a nonparametric Kruskal–Wallis test with a post hoc Dunn test to look for differences in two alpha diversity metrics (Chao1 and Shannon diversity). We used the Chao1 metric as a nonparametric measure of species richness and Shannon diversity as a measure of both richness and evenness (Kim et al. 2017). We assessed beta diversity by calculating the Bray–Curtis dissimilarity to quantify compositional differences between samples. We performed non-metric multidimensional scaling (NMDS) ordination using the phyloseq (v. 1.48.0) package. We generated ordinations from Bray–Curtis dissimilarity and Jaccard distance calculated from relative abundance (proportional) data, and experimental groups were visualized using 95% confidence ellipses. We tested community composition differences among groups using permutational multivariate analysis of variance (PERMANOVA; 999 permutations) and assessed differences in multivariate dispersion using permutational analysis of multivariate dispersion (PERMDISP), with post hoc Tukey's HSD tests. We used the vegan package (v. 2.7) to conduct all multivariate analyses. We evaluated differential taxonomic abundance at the amplicon sequence variant (ASV) level using Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC2), in the ANCOMBC package (v. 2.12.0), with Holm-adjusted p-values used to control for multiple testing across taxa. The phylum Actinobacteriota was differentially abundant in the AMX_25 group, and given its relevance to our research questions, we further examined this phylum using targeted group comparisons. We analyzed relative abundance data using a Kruskal–Wallis test, followed by pairwise Wilcoxon rank sum test with Holm-adjusted p-values. We used the AmphiBac v. 2023.2 database to determine the antifungal richness and the percent of antifungal bacteria present in the microbiome (Bletz and Woodhams 2023; expanded from 36). We chose to look at this metric as a proxy for immune function because the richness of bacteria in the microbiome with antifungal capabilities is a known indicator of susceptibility to disease in amphibians (Woodhams et al. 2023; Nava-González et al. 2021; Jiménez et al. 2022). We first generated representative sequences from samples using the QIIME 2 platform paired- end sequence pipeline, trimming sequences at 250 bp based on quality scores as well as previous studies (Estaki et al. 2020; Schuck et al. 2024). We then matched our bacterial sequences using a 99% similarity to the AmphiBac database of skin bacteria previously tested in culture for the ability to inhibit the growth of a fungal pathogen (i.e., Batrachochytrium dendrobatidis; Woodhams et al. 2023). We rarified the data to the minimum sequencing depth and calculated antifungal bacterial richness (i.e., the richness of bacteria that matched to bacteria in the AmphiBac database) and the percent of antifungal bacteria present in the microbiome (i.e., the number of antifungal bacterial sequences divided by the total number of sequences for each sample, or, the proportion of antifungal bacteria). We then compared the percent of predicted antifungal bacteria and antifungal bacterial richness among each treatment group with a non- parametric Kruskal–Wallis test and post hoc Dunn tests using Holm adjusted p -values. For all body condition data, we first calculated body condition as tadpole mass divided by tadpole SVL and then ran ANOVAs with post hoc Tukey's HSD tests to look for pairwise differences among treatment groups. For the thymus diameter data, we took the average diameter of the two thymuses for each tadpole. Then, for the thymus diameter, spleen diameter, and thymus cell count data, we ran non-parametric Kruskal–Wallis tests with post hoc Dunn tests to look for pairwise differences using Holm-adjusted p-values. Lastly, we plotted the diameter of the thymus (y-axis) against tadpole mass (x-axis) to see if there was a relationship between tadpole size and thymus size. We ran a Pearson's correlation, a linear model including a mass by treatment group interaction, and then a post hoc ANOVA to look for pairwise differences among treatment groups.