The medaka (Oryzias latipes) lines were obtained from the National Bioresource Project (NBRP) Medaka and maintained at Utsunomiya University: inbred strain Hd-rR (strain ID: IB178), standard line OK-cab (strain ID: MT830), and cab-Tg (rag1:egfp), a transgenic medaka expressing egfp under the control of an immature lymphocyte-specific rag1 promoter (strain ID: TG848) (22). Fish were maintained in temperature-controlled tanks (26°C) with a water circulation system under a 14 h light/10 h dark cycle. This study was conducted in accordance with the ethical guidelines of the Utsunomiya University Animal Experimentation Committee, and the experimental protocols were approved by the committee (approval no. A20c-0012). The developmental stages of medaka were designated as previously described (28).

Alt-R CRISPR-Cas9 crRNA targeting exon 3 of il2rg (ENSORLG00000001795), Alt-R CRISPR-Cas9 tracrRNA, and the Cas9 protein were purchased from Integrated DNA Technologies (IDT; Singapore). The sgRNA sequences are listed in Table S1 . RNA–protein complexes were prepared according to the manufacturer’s instructions. Glass capillaries with 1.0 mm outer diameter were pulled (temperature 62°C, force: two light and two heavy weights provided by the manufacturer) using a micropipette puller PC-10 (Narishige Instruments, Tokyo, Japan). The capillary was filled with crRNA (0.75 mM), tracrRNA (1.5 mM), and 0.25 mg/mL Cas9 protein. Then 1–2 nL of the mixture was injected into 1-cell Cab embryos using a FemtoJet 4i (Eppendorf, Hamburg, Germany). G0 fish were crossed with wild-type (WT) Cab fish to establish the il2rg mutant strains. il2rg^del4 mutation was chosen for phenotypic analysis. The primers used for genotyping are shown in Table S1 .

The larvae and adult medaka were anesthetized with 0.1% (v/v) 2-phenoxyethanol and immobilized in 3% methylcellulose. Images were captured using a DP73 digital camera (Olympus, Tokyo, Japan) under an MZ16F fluorescence stereo microscope (Leica, Houston, TX, USA).

RNA was extracted using the NucleoSpin RNA kit (Macherey-Nagel, Düren, Germany) with on-column DNase treatment. The RNA was reverse transcribed using a QuantiTect Reverse Transcription Kit (Qiagen). To detect immunoglobulin Mu (igm) and t-cell receptor beta chain (tcrb) cDNA after VDJ recombination, nested PCR was performed using primer sets targeting the variable (V) and constant (C) regions. Two V regions highly expressed in wild-type kidneys were chosen for RT-PCR analysis, respectively (data not shown). Amplification was performed in a thermal cycler using the following program: 1 min at 95°C, cycles of 10 sec at 95°C, 30 sec at 59°C, 30 sec at 72°C, and 5 min at 72°C. The cycle numbers were as follows: actb (actin beta), 25 cycles; igm (igV_H-Cm) and tcrb (tcrVb-Cb2), 30 cycles plus nested 30 cycles. The primers used for RT-PCR are listed in Table S1 .

Quantitative PCR (qPCR) was performed using THUNDERBIRD SYBR qPCR Mix (Toyobo, Osaka, Japan) and LightCycler 480 (Roche, Basel, Switzerland). The primers used for qPCR are listed in Table S1 .

The WT and il2rg mutants derived from incrosses of heterozygous il2rg mutants were raised together. After genotyping, they were maintained in a water tank separated by a mesh to share the same water environment. Intestinal bacterial DNA was extracted using a QuickGene DNA Tissue Kit (Kurabo, Osaka, Japan) following the method described by Wong et al. (29) with some modifications. Intestines were harvested from WT and il2rg^-/- medaka using sterile tweezers, transferred to microfuge tubes with 0.8 g zirconia beads and 200 μL Tissue lysis Buffer MDT, and immediately frozen in liquid nitrogen. Intestines were then quickly thawed at 65°C, 50 μL lysozyme (50 mg/mL) was added, and incubated at 30°C for 45 min. Subsequently, the bacterial lysates were processed following the protocol of Bacterial Genomic DNA Extraction from Stool (DF-1) of the QuickGene DNA tissue kit (Kurabo).

The V3 and V4 regions of the 16S rRNA gene were amplified using primer pairs with Illumina overhang sequences ( Table S1 ). Index tag sequences were added to the amplicons using the Nextera XT Index Kit v2 (Illumina) and sequenced using the MiSeq Reagent Kit v3 (Illumina). The 301-bp paired-end reads were trimmed to remove low-quality ends (quality score < 15) and adapters and merged using the DADA2 plugin of QIIME2 (https://qiime2.org/). The curated sequences were annotated using the QIIME2 q2-feature-classifier plugin and Silva 132 99% OTUs (full-length, seven-level taxonomy) classifier pre-trained to the full-length Silva database to assign taxonomy to all ribosomal sequence variants.

Alpha- and beta-diversity analyses were performed using the q2-diversity plugin wrapped in QIIME2 at a sampling depth of 15,000 sequences. Alpha diversity metrics (Shannon and Chao1 indices) were calculated, and statistical significance was determined using the Kruskal–Wallis test. For Beta diversity analysis, principal coordinate analysis (PCoA) was performed using Bray–Curtis and weighted UniFrac metrics and visualized using a QIIME2 emperor visualizer (ver. 2022.2.0). The statistical significances of weighted UniFrac metrics distances were calculated using Permutational Multivariate Analysis Of Variance (PERMANOVA).

Data were obtained from 2-3 independent experiments ( Table S2 ).

The intestinal tracts were harvested from adult Hd-rR strain using sterile tweezers. After washing the surface of the intestine with sterile water, it was crushed in sterile phosphate-buffered saline (PBS) using a Power Masher II (Nippi Incorporated, Tokyo, Japan). Diluted intestinal bacteria was aerobically cultured on solid LB media at 28°C overnight, and the colonies were picked up for liquid culture. Bacterial DNA was extracted as described by Wong et al. (29) with some modifications. The bacterial pellet was suspended in 200 μL enzymatic lysis buffer (20 mM Tris-Cl (pH 8.0), 2 mM EDTA (pH 8.0), and 1% Triton X-100) and immediately frozen in liquid nitrogen. Subsequently, the samples were quickly thawed at 65°C, 20 μL 200 mg/mL lysozyme was added, and incubated at 30°C for 45 min. Next, 10 μL proteinase K (10 mg/mL) and 230 μL 40% guanidine hydrochloride were added, and the sample was incubated at 56°C for 30 min. Bacterial lysates were processed using a Mag Extractor-Genome (Toyobo) for DNA extraction.

The bacterial 16S rRNA gene was amplified using primers (16SrRNA 27F and 16SrRNA 1491R). Amplification was performed in a thermal cycler using the following program: 1 min at 95°C, 40 cycles of 10 sec at 95°C, 30 sec at 59°C, 30 sec at 72°C, and 5 min at 72°C. The amplicons were directly sequenced using 16SrRNA 27F, 16SrRNA 1491R and 16SrRNA 907R primers with BigDye Terminator v3.1Cyle Sequencing Kit (Thermo Fisher Scientific, Waltham, MA, USA) and analyzed using 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The primer sequences are listed in Table S1 . Bacterial families were determined using the National Center for Biotechnology Information (NCBI) BLAST program (www.ncbi.nlm.nih.gov/cgi-bin/blast). A phylogenetic tree was generated with MegAlign (Lasergene, DNASTAR, Madison, WI) using MUSCLE alignment and a maximum likelihood model with 1000 bootstraps.

One of the type A clones of Aeromonas isolated from the intestine of an adult Hd-rR strain was spread on a solid Brain Heart Infusion (BHI) medium (BD, Franklin Lakes, NJ). A single colony was picked and cultured in liquid BHI medium at 25°C overnight, then 1:100 dilution was further cultured at 25°C for 4 h until the growth phase was reached. Subsequently, the culture was centrifuged at 2000 × g for 10 min, and the supernatant was replaced with PBS. The OD600 of the culture was measured using an Od meter (Thomas Scientific, Swedesboro, NJ, USA). Two months post-fertilization (mpf), WT medaka were exposed to the bacteria at 6.5×10⁷ colony forming units (CFU)/mL in filtered water using the bath method. CFU of the bacteria were calculated as the colony numbers of diluted bacteria after overnight culture on the solid BHI medium. Twenty percent of the water-containing bacteria was replaced with filtered water every day.

Whole kidney marrow and spleen cells were prepared as previously described (30) with some modifications. Cells were obtained by pipetting the organs into 1 mL ice-cold 1% fetal bovine serum (FBS) in 0.9× PBS. After centrifugation, the pellet was gently mixed with 1 mL distilled water at 20°C by pipetting 4 times to lyse the erythrocytes by osmotic shock to prepare splenocytes, and immediately 1 mL of 1.8× PBS was added. Cells were washed with 1% FBS in 0.9× PBS by centrifugation and then filtered using a 40 μm-stainless mesh. Flow cytometry (FCM) was performed using a FACSLyric flow cytometer (BD Biosciences). Data analysis was performed using Flowjo10.6.2 (BD Biosciences).

A colony of Aeromonas isolated from the intestine of an adult Hd-rR strain was picked and transferred to liquid LB medium. OD600 values were measured using a UV-Vis Spectrophotometer V-630 (Jasco, Tokyo, Japan). By counting colony numbers after overnight culture on LB agar plate, CFU per OD600 were calculated so that 1× 10⁶ CFU/mL of Aeromonas culture was added to larvae.

Germ-free medaka were produced following the method used to produce germ-free zebrafish described by Melancon et al. (14) and Pham et al. (13) with some modifications. Eggs of Hd-rR strain were cultured in non-sterile embryonic culture medium (ECM) comprising 0.1% (w/v) NaCl, 0.003% (w/v) KCl, 0.004% (w/v) CaCl₂-2H₂O, 0.016% (w/v) MgSO₄-7H₂O at 28°C for 6 days.

At 6 days post-fertilization (dpf) (stage 36), embryos were transferred into sterile antibiotics-embryonic culture medium (AB-ECM) (200 μg/mL ampicillin, 5 μg/mL kanamycin, 0.5 μg/mL amphotericin B in ECM, filtered using 0.22 μm filters) and cultured at 28°C for 4 h. The embryos were then transferred to a sterile beaker containing AB-ECM. After washing three times with sterile ECM, embryos were soaked in 0.1% PVP-I solution (0.1% PVP-I in ECM, filtered using a 0.22 μm filter) for 2 min. After washing three times with sterile ECM, the embryos were transferred to another sterile beaker with 0.003% bleach solution (0.003% sodium hypochlorite in ECM, filtered using 0.22 μm filter) and incubated at room temperature for 30 min. For germ-free condition, embryos were washed three times with sterile ECM and transferred into 25 cm² flasks with sterile ECM.

At 12 dpf, when all embryos hatched, 200 μL of the culture medium was spread onto the solid culture media to test for bacterial contamination. Larvae were transferred to new 25 cm² flasks with sterile ECM (for germ-free), 5 μm-filtered fish system water (for conventional control), or 1 × 10⁶ CFU/mL Aeromonas in sterile ECM (for gnotobiotic) with mashed gamma-irradiated powdered food (Otohime S1, Marubeni Nisshin Feed Co., LTD, Tokyo, Japan).

Two days after culture in respective mediums (14 dpf), 200 μL of the culture medium of all groups was spread onto solid culture media to test for bacterial contamination. The larvae were then transferred to new 25 cm² flasks containing sterile ECM. Subsequently, two days after culturing (16 dpf), the larvae were used for phenotypic analyses.

A total of 200 μL germ-free, conventional, or gnotobiotic culture medium containing larvae was spread onto solid LB or trypticase soy agar (TSA; Thermo Fisher Scientific) media for aerobic culture at 28°C for 2 days. In addition, anaerobic culture was performed in a solid TSA medium at 28°C for 4 days. AnaeroPack-Anaero (Mitsubishi Gas Chemical, Tokyo, Japan) was used for anaerobic culture, and the anaerobic status was checked using an Anaero-indicator (Mitsubishi Gas Chemical, Tokyo, Japan).

Whole larvae were fixed in Bouin’s solution overnight. Paraffin sectioning was performed as described in a previous study (31). Coronal sections (5 μm thick) of larval intestines were stained with hematoxylin and eosin. For Periodic acid-Schiff (PAS) staining, the deparaffinized sections were placed in a 0.5% periodic acid solution (Fujifilm Wako, Osaka, Japan) for 5 min and rinsed twice with distilled water. Sections were stained with Schiff’s reagent (Fujifilm Wako) for 5 min, washed with a sulfurous acid solution (Fujifilm Wako) three times, rinsed under running tap water, and counterstained with hematoxylin. For acidic mucin staining, deparaffinized sections were placed in 3% acetic acid for 3 min and stained with Alcian blue solution (pH 2.5; Fujifilm Wako) for 30 min. The sections were rinsed under running tap water and counterstained with hematoxylin. Images were captured using a DP72 camera (Olympus) under a BX60 microscope (Olympus).

RNA- sequencing (RNA-seq) analysis was performed as described in a previous study (32). Whole bodies of germ-free (pools of two fish, n = 3) and conventional (pools of two fish, n = 3) larvae at 14 dpf were kept in RNAprotect reagent (Qiagen, Hilden, Germany) at 4°C. RNA was extracted using QIAzol (Qiagen), followed by DNase treatment and purification using an RNeasy Micro Kit (Qiagen). RNA-seq libraries were prepared using the Kapa Stranded mRNA-Seq Kit (Nippon Genetics, Tokyo, Japan). Six libraries (n = 3/group) were mixed equally and sequenced by Macrogen (Japan) using HiSeqX_Ten (151 bp, paired-end). The obtained raw reads were trimmed using the Trimmomatic software ver. 0.39 with the following parameters: CROP:150, SLIDINGWINDOW:4:15, MINLEN:50 (33). The filtered high-quality reads were mapped to the reference genome using HISAT2 (ver. 2.1.0) (34). The genomic DNA sequence of Oryzias latipes from the NCBI database (ASM223467v1) was used as the reference genome. Genes were quantified using StringTie software (ver. 1.3.4). Gene expression variation analysis was performed using Ballgown software (ver. 2.18.0).

GO enrichment analysis was performed for the 53 genes downregulated in germ-free larvae ( Table S3 ) using Panther software (version 17.0; http://pantherdb.org/). The analysis was performed using annotation datasets of the complete GO biological processes and the reference list of human genes. Eighteen genes that were uncharacterized or lacked human orthologs were excluded from the analysis.

Ten microliters of LPS (4.5 mg/mL) obtained from Escherichia coli O111:B4 (Product Number: L2630, Lot Number: 099M4002V, Sigma-Aldrich, St. Louis, MO) in distilled water was injected into the yolk of 9 dpf (stage 39) embryos. RNA was extracted 2 h post-injection.

To assess the effects of a defective adaptive immune system on the intestinal microbiota of medaka, immunodeficient medaka (il2rg^-/- ) were generated by genome editing using the CRISPR/Cas9 system, targeting exon 3 of il2rg ( Figure S1A ). A mutant line with a 4 bp deletion in il2rg was established, which was predicted to cause a frameshift and encode only 75 appropriate amino acids, followed by 12 unrelated amino acids and a stop codon ( Figure S1B ). The resulting protein lacked most of the fibronectin type III domain (FNIII) and the entire transmembrane domain (TM) ( Figure S1C ) and was considered a null allele. The homozygous form of this frameshift mutant did not exhibit any gross developmental defects. However, crossing the mutant with rag1:egfp transgenic medaka, which expresses EGFP under the control of the immature lymphocyte-specific rag1 promoter (22), revealed a severe reduction in EGFP-expressing cells in the larval thymus, suggesting defects in T cell development ( Figure 1A ). This severe reduction in T-cells was continuously observed in the adult thymus ( Figure 1B ).

Next, we investigated the transcripts of rearranged lymphocyte antigen receptors in the kidney, equivalent to the mammalian bone marrow as a site of hematopoiesis. Reverse transcription PCR (RT-PCR) analysis detected VDJ-recombined igm transcripts in B cells but not tcrb transcripts in the il2rg mutant ( Figure 1C ). qPCR analysis of adult kidney cells revealed a severe reduction in il2rg expression in il2rg mutants ( Figure 1D ). Under these conditions, the expression of pan B cell marker cd79, early B cell marker ebf1, and relatively mature B cell marker cd22l in the WT and il2rg mutant did not differ significantly ( Figure 1D ). In contrast, the expression of T cell markers cd8a and cd4-2 was significantly decreased in il2rg mutants ( Figure 1D ). Among other blood populations, the expression of the neutrophil marker mpx was comparable, but that of the natural cell maker nkl.1 was significantly reduced in il2rg mutants than those in WT medaka ( Figure 1D ). These results suggest that il2rg mutation affects T and NK cells, as observed in the zebrafish il2rg mutants (11, 27); nevertheless, they grew to adult stage in filtered water with a weekly water change (mortality rate at 3 mpf: 0/15), despite being sensitive to lower water quality (mortality rate: 11/14 in a recirculating system sharing water with other medaka tanks).

The analysis of the intestinal bacterial composition of the WT and il2rg mutants at 2, 2.5, and 3 mpf revealed that most OTUs were represented by Proteobacteria, Fusobacteriota, and Firmicutes at the phylum level in both the WT and il2rg mutants at all time points ( Figures 2A , 3A ). However, the abundance of Proteobacteria was significantly higher in the il2rg mutants than that in WT at 2.5 and 3 mpf. Especially at 3 mpf, more than 90% of the intestinal microbiota of il2rg mutants comprised Proteobacteria. Furthermore, the abundance of Fusobacteriota was inversely correlated with that of Proteobacteria.

At the family level, Aeromonadaceae was abundant in WT adult medaka intestines ( Figures 2B , 3B ), which was consistent with the previous report (25). The abundance of Aeromonadaceae gradually increased in the il2rg mutants at 2.5 and 3 mpf and was significantly higher than that in the WT ( Figures 2B , 3B ). This enrichment of Aeromonadaceae was observed repeatedly in independent experiments ( Table S2 ). Other significant changes at the family level included a temporary increase in Shewanellaceae at 2.5 mpf and an increase in Enterobacteriaceae at 3 mpf in il2rg mutants.

To compare the overall composition of intestinal bacteria between the WT and il2rg mutants at each time point, alpha diversity indices were analyzed. Shannon indices, which indicate the evenness of OTUs, showed that the WT was unchanged at all three time points ( Figure 4A ). Although the evenness of the OTUs in the il2rg mutant was comparable to that of the WT at 2 and 2.5 mpf, it was significantly decreased at 3 mpf ( Figure 4A ), which mainly reflected the increased abundance of Aeromonadaceae. In contrast, Chao1 indices, which indicate the richness of OTUs, were comparable between WT and il2rg mutants at all time points, although the index of WT medaka at 3 mpf was higher than that at earlier time points ( Figure 4B ).

Next, we quantified the similarity of bacterial composition among each fish and each group. Analysis of beta-diversity indices using weighted UniFrac and Bray–Curtis plots revealed a stable pattern in WT at all time points. In contrast, the intestinal bacterial composition of il2rg mutants showed dynamic shifts at 2.5 and 3 mpf ( Figures 4C, D ). Considering the phylogenetic distances to the datasets, weighted UniFrac indices were chosen for the quantitative measurement of OTUs. The calculation of the distances indicated that the bacterial composition of il2rg mutants was significantly distant from that of the WT at 2.5 and 3 mpf ( Figure 4E ).

Next, we identified readily culturable bacteria derived from the medaka intestine. A culture of aseptically harvested intestines from adult WT medaka under aerobic conditions revealed uniform shapes and colors of the colonies ( Figure S2A ). The 16S rRNA gene sequences from the 16 colonies were grouped into three types: type A (13 colonies), type B (two colonies), and type C (one colony). The phylogenetic tree of the 16S rRNA gene sequences of the Aeromonas family showed that all three types were close to Aeromonas and away from Oceanimonas ( Figure S2B ). Type A and C sequences were the closest to those of A. intestinalis, recently isolated from human feces (35). The type B sequence was identical to the FPC-0866 strain, derived from eels and defined as A. hydrophila (36). However, A. hydrophila strain ATCC 7966 was relatively phylogenetically distant from the FPC-0866 strain, indicating that the resolution of 16S rRNA gene sequencing was not high enough to identify the species of Aeromonas precisely. Nevertheless, all three types are close to Aeromonas; therefore, we considered Aeromonas a representative bacterium of adult medaka.

To examine the effect of intestine-derived Aeromonas on adult medaka, WT medaka were exposed to 6.5 × 10⁷ CFU/mL of one of the type A clones. Although this exposure did not cause lethality (14/14), exposed fish commonly showed vasodilation of the liver and an enlarged spleen ( Figure S2C ), suggesting an inflammatory response. FCM analysis of kidney and spleen cells showed that the exposure did not change the populations of kidney cells; however, the lymphoid population overwhelmed the myeloid population in the spleen of the exposed fish ( Figures S2D, S2E ), implying lymphocyte involvement in the antibacterial response.

Next, we sought to establish a method for maintaining medaka larvae under germ-free conditions. To this end, we followed one of the protocols for zebrafish germ-free culture (14) with some modifications for medaka. PVP-I and bleach treatment of eggs at 6 dpf before hatching and culture of eggs and larvae thereafter with sterile AB kept the larvae sterile, which was confirmed by the absence of colonies after the larval culture medium was spread on LB agar medium for aerobic culture and TSA agar medium for aerobic and anaerobic cultures ( Figure S3 ). Gnotobiotic conditions were also achieved by adding 1 × 10⁶ CFU/mL Aeromonas derived from adult medaka intestine to a germ-free culture.

Next, to analyze the structural changes in the intestines of wild-type germ-free larvae, we compared the anterior, middle, and posterior intestinal structures (37) under germ-free, conventional, and gnotobiotic conditions. The anterior intestine contained small absorptive vacuoles in its epithelial layer. However, no apparent structural differences were observed between the conventional, germ-free, and gnotobiotic conditions. No structural differences were observed in the posterior intestines ( Figure S4 ).

However, apparent structural differences were observed in the middle intestine. In the germ-free intestine, although the thickness of the intestinal tube itself was comparable to that in the conventional condition ( Figures 5A, B ), the thickness of the epithelial layers was significantly increased compared to that in the conventional condition ( Figures 5A, C ). Close observation of the germ-free epithelium revealed enlarged absorptive vacuoles. Although such vacuoles were also present in the conventional intestine, their number was significantly higher in the germ-free intestine ( Figures 5A, D ). These vacuole structures were not goblet cells because they did not stain with PAS (for neutral mucin) or with Alcian blue (for acidic mucin) ( Figure 5E ). The increased epithelium thickness and absorptive vacuole enlargement were negated by adding Aeromonas bacteria to the gnotobiotic conditions ( Figures 5A, C, D ).

To investigate the effects of immune-related gene expression in germ-free medaka, RNA-seq analysis of whole bodies of medaka larvae grown under conventional or germ-free conditions was performed. Fifty-three genes were significantly downregulated, and six were significantly upregulated in germ-free larvae ( Tables S3, S4 ). Gene ontology analysis of the downregulated genes showed significant enrichment of defense response (GO:0006952) genes (p = 0.0018; Figure 6A ), which included the regulation of intestinal absorption (GO:1904478) genes (hamp and apoa4a) (p = 0.00024).

qPCR analysis of whole bodies of germ-free larvae showed a tendency to decrease the expression of seven genes (grn, epx, hamp, hp, c1q, mrc1, and apoa4), except fasn under germ-free conditions compared to conventional conditions ( Figure 6B ). Interestingly, the expression of grn, epx, hamp, hp, c1q, and mrc1, excluding apoa4, was partially recovered under gnotobiotic conditions ( Figure 6B ). qPCR analysis using intestinal RNA samples showed identical expression patterns— the expression of five genes (epx, hamp, hp, c1q, and mrc1), except grn decreased in germ-free conditions and recovered in gnotobiotic conditions ( Figure 6C ). qPCR analysis of whole bodies of larvae injected with LPS, a toll-like receptor 4 (TLR4) signaling inducer, showed increased hamp, hp, and c1q expression ( Figure 6D ), suggesting that the expression of these genes is affected by the detection of bacterial components by TLRs.