The intestinal microbiota, an invisible organ supporting a host’s survival, has essential roles in metabolism, immunity, growth, and development. Since intestinal microbiota influences a host’s biology, application of such data to wildlife conservation has gained interest. There are standard protocols for studying the human intestinal microbiota, but no equivalent for wildlife. A major challenge is sampling the intestinal microbiota in an effective, unbiased way. Fecal samples are a popular proxy for intestinal microbiota because collection is non-invasive, convenient, and allows for longitudinal sampling. Yet, it is unclear whether the fecal microbiota is representative of the intestinal microbiota. In amphibians, research on sampling methodology is limited. In this study, we characterize and compare microbiota (small intestine, large intestine, feces) of two Hong Kong stream-dwelling frog species: Lesser Spiny Frog (Quasipaa exilispinosa), and Hong Kong Cascade Frog (Amolops hongkongensis). We found that both species have similar dominant phyla and families, but diverge in terms of the dominant genera. Next, we assess the performance of fecal microbiota in representing the intestinal microbiota in these two species. We found that (1) microbiota of small and large intestine differs significantly, (2) feces are not an appropriate proxy of both intestinal sections, and (3) a set of microbial taxa significantly differs between sample types. Our cautions equating fecal and intestinal microbiota. Sampling feces can avoid sacrifice of an animal, but researchers should avoid over-extrapolation and interpret results carefully.

Frog surveys

Night surveys were conducted within three hours after sunset, from June to August 2021. Each survey included 3–4 surveyors experienced in herpetological field work. Surveyors walked along the stream looking for the two target species using headlamps and flashlights. Individuals were captured with gloved hands, with gloves changed between individuals to prevent cross contamination. After sampling skin microbiota for another study, each individual was stored in separate air-filled plastic bags for transport to laboratory within two hours.

Fecal microbiota

All equipment was sterilized with a 10% bleach solution beforehand to avoid contamination. Upon arrival to the laboratory, each individual was kept separately in sterile plastic boxes for 12–24 hours waiting for defecation. The boxes have a wire mesh floor that allows the feces to fall and avoid contamination by the individual. Each box was humidified by spraying sterile water, and kept in dark to minimize disturbance. Approximately nine hours after capture, boxes were checked hourly for defecation. No food was provided during the process. Fecal samples were collected using sterile forceps and stored in tubes at –80 °C until DNA extraction. Individuals were euthanized for sampling intestinal microbiota (see below) after defecation, or after 24 hours without defecation.

Intestinal microbiota

Each individual was euthanized by overdosing with ethyl 3-aminobenzoate methanesulfonate in the laboratory (MS222; Sigma-Aldrich, St. Louis, MO, USA) (American Veterinary Medical Association, 2020). The entire intestine was dissected from end of the stomach to the cloaca. The small and large intestine were identified morphologically. A 2-cm piece of anterior and posterior end of the intestine was taken as small and large intestine, respectively (Tong et al., 2014). We could not identify intestinal subsections morphologically (e.g., duodenum, jejunum, and ileum of the small intestine), so this information was excluded from our analyses. Each segment was rinsed to remove transient bacteria and intestinal content with 2 mL sterile water using a pipette. Each segment was then cut longitudinally on one side, laid flat, and the gut microbiota was collected by swabbing using an ESwabTM (Copan Diagnostic Inc., California, USA). The swab was drawn across the inner surface of the intestine segment 10 times (1 time = forward and backward). The swabs were stored at –80 °C until DNA extraction. For a negative control, a blank swab was included in subsequent DNA extraction and sequencing. Frog bodies were preserved and accessioned into the herpetology specimen collection of the Lingnan University Natural History Collection.

DNA extraction

DNA extraction was conducted at the laboratory of Lingnan University. DNA was extracted from swabs using the QIAamp®BiOstic® Bacteremia DNA Kit (Qiagen GmbH, Hilden, Germany) following amendments verified by the manufacturer. The amendments are summarized as follows—in the first step, 450 µL of MBL buffer was added to each PowerBead tube: (1) swabs were defrosted then placed in PowerBead tubes, (2) feces were subsampled (250 mg) and placed in PowerBead tubes, and (3) the subsequent steps followed the manufacturer’s protocol starting from “vortex for 10 s and incubation at 70 °C for 15 min”. A blank sample was included for every batch of DNA extraction as a negative control. DNA extractions were quantified by Qubit™ 3 Fluorometer with the dsDNA BR Assay Kit (Thermo Fisher Scientific Inc., MA, USA).

Library preparation and sequencing

Sequencing libraries were prepared according to the Illumina protocol (“16S Metagenomic Sequencing Library Preparation Part #15044223 Rev. B protocol"). The V4 region of the 16S rRNA gene was amplified by PCR using The Earth Microbiome Project V4 primer pair coupled with Illumina adapter overhang sequences (515F [Parada]: 5’- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGYCAGCMGCCGCGGTAA-3’, 806R [Apprill]: 5’- GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGACTACNVGGGTWTCTAAT-3’) (Caporaso et al., 2011; Apprill et al., 2015; Parada et al., 2016). For each reaction, 2 ng of DNA was amplified with 5X reaction buffer, 1 mM of dNTP mix, 500 nM of each PCR primer, and Herculase II fusion DNA polymerase (Agilent Technologies, Santa Clara, CA). Thermal cycling conditions were as follows: 3 min at 95 °C; 25 cycles of 30 s at 95 °C, 30 s at 55 °C, and 30 s at 72 °C; followed by 5 min at 72 °C. The 1st PCR product was purified with AMPure beads (Agencourt Bioscience, Beverly, MA). For each sample, 2 µL of the 1st PCR product was amplified for library construction using a unique Nextera XT Index Primer (Illumina, Inc., San Diego, CA). Thermal cycling conditions were the same as the 1st PCR, but with only 10 cycles. The 2nd PCR product was then purified using AMPure beads (Agencourt Bioscience, Beverly, MA), and quantified using two approaches: (1) qPCR according to the qPCR Quantification Protocol Guide (KAPA Library Quantification kit for Illumina Sequencing platforms) and (2) TapeStation system and D1000 ScreenTape (Agilent Technologies, Waldbronn, Germany). The pooled products were paired-end sequenced with Herculase II Fusion DNA Polymerase Nextera XT Index V2 Kit on an Illumina MiSeq PE300 (Illumina, Inc., San Diego, CA). A negative control was included in each lane to test for contamination. All library preparation and sequencing work was conducted at Macrogen (Seoul, South Korea).

Bioinformatics work was performed using Quantitative Insights into Microbial Ecology 2 (QIIME 2, version 2022.11) (Bolyen et al., 2019) and R (R Core Team 2021). Raw sequence data were demultiplexed by Macrogen (Seoul, South Korea). We evaluated the quality of demultiplexed reads and trimmed to remove poor-quality bases. Sequence quality control and denoising were performed with DADA2 (Callahan et al., 2016) to generate amplicon sequence variants (ASVs) using q2-dada2. Representative sequences were aligned with MAFFT (Katoh and Standley, 2013) using q2-alignment. To maximize taxonomic classification accuracy, we assigned taxonomy to ASVs using a weighted Bayes classifier that incorporates information on environment-specific taxonomic abundance (Kaehler et al. 2019), pre-trained on the 16S 515F/806R region in SILVA 138 (Quast et al., 2013) (MD5: b9476399080d189b4c9917d1246e7c69) (Bokulich et al., 2018; Robeson II et al., 2021) using q2-feature-classifier (Bokulich et al., 2018). Sequences were removed if unassigned at the domain level or assigned to be mitochondria/chloroplast/archaea. The format of the metadata file used in QIIME2 was validated by the cloud-based Google Sheets add-on Keemei (Rideout et al., 2016). The output QIIME2 files were exported to R for subsequent analysis using R package “qiime2R” (Bisanz 2018).