Chinese perch were obtained from a fish farm in Hubei Province, China, and maintained in aerated tanks at 28 °C with recirculating fresh water for two weeks prior to the experiments. All animal experiments were conducted in accordance with the Guiding Principles for the Care and Use of Laboratory Animals and were approved by the Institute of Hydrobiology, Chinese Academy of Sciences.

The mandarin fish fry cell line (MFF-1) and human embryonic kidney 293T (HEK293T) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) with 10% fetal bovine serum (FBS; Gibco) and 5% CO₂ at 28°C and 37°C, respectively. The infectious hemorrhagic syndrome virus (IHSV) was propagated in MFF-1 cells, and viral titers were determined according to a previous report (24).

DNase I reagent (Thermo Scientific) was used at 37°C to remove genomic DNA remnants in total RNA samples, which were extracted from fish tissues or cells using TRIzol^® reagent (Ambion) following the manufacturer’s protocol. Next, the RevertAid™ First Strand cDNA Synthesis Kit (Thermo Scientific) was used to synthesize first strand cDNA according to the manufacturer’s instruction. Full-length open reading frames (ORFs) of ripk2, card9, nod1 and nod2 were generated through PCR amplification by using Phanta Max Master Mix high-fidelity DNA polymerase (Vazyme) with gene-specific primers (referring to Supplementary Table 1 ) which were designed based on the predicted sequences (XM_044218208.1 and XM_044195794.1 from NCBI database) and the reported sequences (KY974318 and KY974317) (25). All PCR products were confirmed by sequencing.

The online Translate tool (http://web.expasy.org/translate/) was used to predict amino acid sequences. The conserved domain of proteins and transmembrane region was determined by BLASTP program based on Conserved Domains Database (CDD) and by TMHMM Server v.2.0 program (http://www.cbs.dtu.dk/services/TMHMM-2.0/), respectively. The AnimalTFDB v4.0 software (http://bioinfo.life.hust.edu.cn/AnimalTFDB4/) was used to predict the ISRE site of nod1 and nod2 genes. Multiple alignments of protein sequences were performed by using the Clustal X and GeneDoc program. Phylogenetic trees were generated by using neighbor joining (NJ) method with 1000-time bootstrap in MEGA-X package. For gene collinearity analysis, data were collected from the available assemblies, including Homo sapiens (human, version: GRCh38.p14, NCBI accession: GCF_000001405.40), Mus musculus (house mouse, version: GRCm39, GCF_000001635.27), Gallus gallus (chicken, version: bGalGal1.mat.broiler.GRCg7b, GCF_016699485.2), Danio rerio (zebrafish, version: GRCz11, GCF_000002035.6) and Siniperca chuatsi (version: ASM2008510v1, GCF_020085105.1).

The IFN expression plasmids (pcDNA3.1-IFNh, pcDNA3.1-IFNc and pcDNA3.1-IFN-γ) and promoter luciferase reporter plasmids, including pGL3-NF-κB-pro-Luc, pGL3-IFNh-pro-Luc (HP1) and pGL3-IFNc-pro-Luc (CP1), were described in previous reports (26, 27). The pcDNA3.1/Myc-His(-) and p3XFLAG-CMV-14 vectors were used to construct eukaryotic expression plasmids in this study. The full-length ORF fragments were generated by PCR procedure with the protocol as the follows: initial denaturation for 3 min at 95°C, 35 cycles for 30 s at 95°C, 30 s at 55°C and 2 min at 72°C, followed by a further 10 min at 72°C. The specific primers used in the amplification are shown in Supplementary Table 1 . The ORFs of nod1 and nod2 were inserted into the vector to construct pcDNA3.1-NOD1-Myc and pcDNA3.1-NOD2-Myc through EcoRI/KpnI site, respectively. The ORFs of card9 and ripk2 were cloned into the NotI/XbaI site of p3XFLAG-CMV-14 to obtain p3XFLAG-CARD9-FLAG and p3XFLAG-RIPK2-FLAG, respectively. The card9, ripk2 and mda5 with an HA-tag-encoding sequences (5’-tacccatacgacgtcccagactacgct-3’) were ligated into the EcoRI/XbaI site of p3XFLAG, EcoRI/KpnI site of pcDNA3.1, and NotI/KpnI site of pcDNA3.1 to generate p3XFLAG-CARD9-HA, p3XFLAG-RIPK2-HA and p3XFLAG-MDA5-HA, respectively. The plasmids were extracted and purified by using E.Z.N.A.^® Endo Free Plasmid Mini Kit II (Omega Bio-tek) following the manufacturer’s instruction. Cell transfection was performed according to the previous reports (26, 28), and the used transfection reagents for HEK293T and MFF-1 cells were Lipofectamine 2000 (Invitrogen) and FuGENE^® HD Transfection Reagent (Promega), respectively.

HEK293T cells (about 5 × 10⁶) were seeded into 90 mm cell culture dishes overnight and were co-transfected with 4 µg of each plasmid according to a series of combinations, including pcDNA3.1-NOD1-Myc + p3XFLAG-CMV-14 empty vector, pcDNA3.1-NOD1-Myc + p3XFLAG-CARD9-FLAG, pcDNA3.1-NOD1-Myc + p3XFLAG-RIPK2-HA; pcDNA3.1-NOD2-Myc + p3XFLAG-CMV-14 empty vector, pcDNA3.1-NOD2-Myc + p3XFLAG-CARD9-HA, pcDNA3.1-NOD2-Myc + p3XFLAG-RIPK2-HA; p3XFLAG-RIPK2-FLAG + p3XFLAG-CMV-14 empty vector, and p3XFLAG-RIPK2-FLAG + p3XFLAG-CARD9-HA. 48 hours after transfection, the supernatant medium was removed and the cells were washed with phosphate buffered saline (PBS, pH 7.0). Each sample of cells was added with 400 µL lysis buffer (Thermo Fisher Scientific) containing 4 µL protease inhibitor mixture (Thermo Fisher Scientific), and was lysed on the rotational mixer at 4 °C for 24 hours. Next, cell fragments were removed by centrifugation at 2000 × g for 10 min, and the supernatants were incubated by overnight rotating with anti-FLAG or anti-HA Immunomagnetic Beads (Dia-An Biotech) at 4°C. After wash for five times with PBS, protein samples were collected from beads through boiling.

Western blotting assay was performed as described in a previous study (3, 27). The primary antibodies included mouse monoclonal antibody (mAb; anti-FLAG M2, #F1804, Sigma-Aldrich) with the dilution ratio of 1:5000, mouse anti-Myc-Tag mAb (#2276, Cell Signaling Technology, 1:5000), rabbit anti-HA-Tag mAb (#3724, Cell Signaling Technology, 1:5000), and the secondary antibodies included HRP-conjugated mouse anti-rabbit IgG (#D110065-0100, Sangon Biotech, 1:5000) and HRP-conjugated goat anti-mouse IgG (#D110087-0100, Sangon Biotech, 1:5000).

MFF-1 cells were seeded onto coverslips in 24-well plates and were co-transfected with 0.25 µg of each plasmid in same combinations as described above. 24 hours after transfection, the supernatants were removed, and the cells were washed with PBS and fixed by using 4% paraformaldehyde at 28°C for 40 min. Next, the cells were washed four times with PBS and incubated with 0.2% Triton X-100 to permeate the cell membrane at 28°C for 12 min. After being washed with PBS for four times, the cells were blocked by 5% BSA at 37°C for 1 hour and incubated with the tag-antibodies described above and the fluorescein-labelled secondary antibodies, including Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 594 (Thermo Fisher Scientific) or Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488 (Thermo Fisher Scientific). Subsequently, the coverslips with labelled cells were sticked on the microscope slides by using ProLong™ Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific). The cells were observed under a Leica confocal microscope (40 × magnification, oil immersion lens).

MFF-1 cells were seeded in 48-well plates and were transfected with a series of plasmid combinations (empty/expression plasmids + luciferase reporter plasmids), including 0.1 μg pcDNA3.1/p3XFLAG-CMV-14, 0.1 μg pcDNA3.1-NOD1-Myc/pcDNA3.1-NOD2-Myc/p3XFLAG-RIPK2-FLAG/p3XFLAG-CARD9-HA/p3XFLAG-MDA5-HA, 0.1 μg pGL3-NF-κB-pro-Luc/pGL3-IFNh-pro-Luc (HP1)/pGL3-IFNc-pro-Luc (CP1)/pGL3-NOD1-pro-Luc (wt)/pGL3-NOD1-pro-Luc (mut)/pGL3-NOD2-pro-Luc (wt)/pGL3-NOD2-pro-Luc (mut) and 0.02 μg pRL-TK. 24 hours after transfection, the cells of each well were washed once with PBS and were lysed by adding 100 μL Passive Lysis Buffer (Promega) for half hour. The Luciferase^® Reporter Assay System (Promega) was used to determine the luciferase activity by running GloMax^®-Multi JR detection system (Promega).

HEK293T cells were seeded into 6 well plates and were transfected with 1 μg pcDNA3.1 empty vector (control), pcDNA3.1-IFNh, pcDNA3.1-IFNc or pcDNA3.1-IFN-γ to generate recombinant protein of IFNh, IFNc or IFN-γ, respectively. The control group and recombinant proteins of IFNh/IFNc/IFN-γ in the cell precipitation and supernatant medium were detected by Western blotting with IFN-specific antibodies, respectively (27). Then, the control and supernatant medium with recombinant IFNs were diluted 10 times for incubation. MFF-1 cells were seeded into 48 well plates and were transfected with 0.01 μg pRL-TK and 0.1 μg pGL3-NOD1-pro-Luc (wt)/pGL3-NOD1-pro-Luc (mut)/pGL3-NOD2-pro-Luc (wt)/pGL3-NOD2-pro-Luc (mut), respectively. After transfection for 24 hours, the MFF-1 cells were incubated separately with control medium and recombinant protein of IFNh (final concentration of 107.5 ng/mL)/IFNc (266.67 ng/mL)/IFN-γ (57.92 ng/mL) for 24 hours to detect the luciferase activity.

To investigate the tissue distribution of nod1, nod2, card9 and ripk2, three healthy Chinese perch with a body length of about 10 cm were sacrificed to collect different organs/tissues, including fin, head-kidney, truck-kidney, spleen, skin, liver, gill, eye, and intestine, for qRT-PCR detection. To investigate whether the four genes (nod1, nod2, card9 and ripk2) could be induced by poly(I:C), IFN-γ, IFNh and IFNc, MFF-1 cells (5.0 × 10⁵) were seeded in 12-well plates and were separately transfected with 50 μg (final concentration of 41.67 μg/mL) poly(I:C) by using 1 μL Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) or directly incubated with 100 μL prepared culture medium supernatant containing the recombinant protein of IFNγ/IFNh/IFNc, as described above. The samples were collected at 3, 6, 12 and 24 hours post-stimulation, respectively. For IHSV infection, MFF-1 cells (5.0 × 10⁵) were seeded in 12-well plates and were infected with IHSV (MOI = 0.01) or without virus (control group), and the samples were collected at 12, 24, 36 and 48 hours post-infection, respectively.

The cDNA synthesis from total RNA was performed with All-in-One RT SuperMix Perfect for qPCR kit (Vazyme). A total of 20 μL reaction system consisted of 10 μL PowerUP™ SYBR™ Green Master Mix (ABI), 2 μL primer mixture, 7 μL sterile water, and 1 μL cDNA template. The qRT-PCR was performed on the QuantStudiO™ 3 Real-Time PCR Instrument (96-Well Block, ABI) with the following protocol: initial denaturation for 2 min at 50°C, 2 min at 95°C, 50 cycles for 15 s at 95°C, 30 s at 58°C, followed by a further 15 s at 95°C, 1 min at 60°C, 1 s at 95°C. The β-actin was used as the internal reference gene to calculate the folding change of expression level of target genes using the 2^(-ΔΔCt) method. The specific primers are listed in Supplementary Table 1 .

All experiments were repeated separately at least three times. Results were analyzed statistically with Student’s t-test in SPSS 16.0 software. Significant differences were indicated with * P < 0.05, and ** P < 0.01, respectively. The data were presented as mean ± standard error (SE).

The nucleotide sequences of two adaptors, RIPK2 and CARD9, were cloned from mixed cDNA samples of Chinese perch tissues and were deposited in the GenBank database with the accession numbers, OR713117 and OR713116, respectively. The full-length ORF of ripk2 and card9 contained 1806 and 1623 base pairs, respectively, which were predicted to encode 601 and 540 amino acids. The RIPK2 and CARD9 protein sequences shared medium identity with human (46.6%/35.6%), mouse (46.4%/37.6%), chicken (44.6%/35.8%) and zebrafish (69.8%/43.2%), respectively. The multiple alignments and domain prediction revealed that vertebrate RIPK2 has a conserved serine-threonine kinase domain (KD) at N-terminal region and a conserved C-terminal CARD domain, and CARD9 has a N-terminal CARD domain and a coiled-coil (CC) domain, while no transmembrane motif (TM) was found in the two molecules ( Supplementary Figures 1 , 2 ). Phylogenetic analysis showed that RIPK2 and CARD9 were formed into a separate clade with their fish homologues, respectively ( Figure 1 ). Moreover, in synteny analysis, vertebrate card9 was found to be linked with conserved small nuclear RNA activating complex polypeptide 4 (snapc4) from fish to human, and tetrapod ripk2 is located on a conserved genetic region which contains matrix metallopeptidase 16 (mmp16) and oxidative stress induced growth inhibitor family member 2 (osgin2) ( Figure 2A ). However, fish ripk2 was sandwiched with agouti signaling protein, nonagouti homolog (mouse) 2b (asip2b) and coiled-coil domain containing 39 (ccdc39) ( Figure 2B ).

The qRT-PCR was used to detect the expression level of nod1, nod2, ripk2 and card9 genes in Chinese perch organs/tissues, including head-kidney, trunk-kidney, spleen, liver, intestine, fin, gill, eye, and skin. As shown in Figure 3A , the four genes were constitutively expressed in all detected organs/tissues, and card9 mRNA was mainly enriched in head-kidney, spleen, and gill, while ripk2 gene was expressed with highest transcription level in spleen and gill. Interestingly, nod1 and nod2 genes showed high expression levels in head-kidney.

To investigate whether nod1, nod2, ripk2 and card9 can be induced by PAMPs and virus, the expression of the four genes was detected in poly(I:C)-treated and IHSV-infected MFF-1 cells, respectively. Compared to the control, poly(I:C) stimulation and IHSV infection all resulted in significant increase in the expression of ripk2 and card9 genes at different time points, and the induced expression of nod1 and nod2 genes was also observed ( Figures 3B, C , Supplementary Figure 3 ). For further investigating whether nod1, nod2, ripk2 and card9 genes could be upregulated by IFNs, the expression of the four genes was detected in MFF-1 cells incubated with recombinant type I (IFNh and IFNc) and type II (IFN-γ) IFN proteins, which have been confirmed to induce ISGs, such as mx, pkr and viperin ( Supplementary Figures 4A-E ). It was observed that all the IFNs significantly induced the expression of nod1, nod2, ripk2 and card9 genes at 3 and 6 hours post-stimulation ( Figure 4 ).

For understanding inducible regulatory elements of nod1 and nod2 genes, 5’-flanking sequences of the proximal promoter were identified. A potential ISRE site was found in the proximal promoters of both nod1 and nod2 genes, which were cloned to construct luciferase reporter plasmids with wildtype (wt) or mutant (mut) ISRE site ( Figures 5A, B ). The recombinant IFN proteins (IFNh, IFNc and IFN-γ) significantly activated ISRE-wt promoter plasmids of both nod1 and nod2 genes, but not the ISRE-mut plasmids ( Figures 5C–H ).

It has been reported that RIPK2 and CARD9 as adaptors play a role in NLR-mediated immune signaling (11, 29, 30). In order to investigate whether RIPK2 and CARD9 are involved in NOD1/NOD2 signal complex, eukaryotic expression plasmids of the four genes were constructed. Co-IP assay was performed to determine the potential interaction of NOD1, NOD2, RIPK2 and CARD9 through pair-wise co-transfection of expression plasmids with tagged fusion protein, respectively. It was revealed that NOD1-Myc was associated with CARD9-Flag and RIPK2-HA ( Figures 6A, B ), and anti-HA affinity magnetic-immunoprecipitated CARD9 and RIPK2 were associated with NOD2-Myc ( Figures 6C, D ). Interestingly, CARD9-HA and RIPK2-Flag showed activity of interaction ( Figure 6E ). In addition, subcellular colocalization analysis also revealed the interaction of tagged NOD1, NOD2, RIPK2 and CARD9 through immunofluorescence. Colocalizations were observed for NOD1-CARD9 ( Figure 7A ), NOD1-RIPK2 ( Figure 7B ), NOD2-CARD9 ( Figure 7C ), NOD2-RIPK2 ( Figure 7D ), and CARD9-RIPK2 ( Figure 7E ).

For understanding the immune signal mediated by NOD1, NOD2, RIPK2 and CARD9, promoter reporter plasmids of NF-κB, IFNh, IFNc, NOD1 and NOD2 were co-transfected with expression plasmids of NOD1, NOD2, RIPK2, CARD9 and MDA5 in MFF-1 cells, respectively. As shown in Figures 8A–C , the overexpression of NOD1, NOD2, RIPK2 and CARD9 provoked significantly the activity of luciferase reporter plasmids of NF-κB, IFNh and IFNc in MFF-1 cells, respectively. Furthermore, it was observed that the co-expression of these signal molecules, including NOD1 + CARD9, NOD1 + RIPK2, NOD2 + CARD9, NOD2 + RIPK2 and CARD9 + RIPK2, induced a significantly higher level of NF-κB, IFNh and IFNc promoter activity compared with the overexpression of these signal molecules alone, respectively ( Figures 9A–C ). Interestingly, the overexpression of NOD1, NOD2, RIPK2, CARD9 and MDA5 provoked significantly the activity of nod1 and nod2 promoters with wildtype ISRE, respectively, but not the promoters with mutant ISRE ( Figures 10A, B ).