HAID Group provided healthy L. vannamei with a body weight of about 5 g each. Prior to any experiments, the shrimp were acclimated in aerated artificial seawater (25 salinity) for two days. A commercial shrimp diet was fed to the shrimp on a daily basis (HAID Group).

The open reading frames (ORFs) of L. vannamei Dorsal-L and Dorsal-S were amplified and cloned into pAc5.1-3×HA vector (24) to express the HA-tagged Dorsal-L protein and Dorsal-S proteins, respectively. Extracellular (ec) domain of each Toll (Toll1/2/3/4/5/6/7/8/9-ec), and the cytoplasmic TIR domain of each Toll (TIR1/2/3/4/5/6/7/8/9) were constructed into pAc5.1-3×HA vector to express HA-tagged proteins. The full length of TRAF6 (1–594), and a series of truncated mutants of TRAF6, including TRAF6 (1-100), TRAF6 (101-225), TRAF6 (226-410) and TRAF6 (411-594), were all inserted into pAc5.1-3×HA vector. The full length of TAB2 (1-608) was constructed into pAc5.1-3×FLAG vector (25), while different truncated mutants of TAB2, including TAB2 (1-60), TAB2 (61-320), TAB2 (321-410), TAB2 (411-500), TAB2 (501-530) and TAB2(531-608), were constructed into pAc5.1-GFP vector (26). Besides, two truncated mutants of TAK1, including TAK1 (1-630) and TAK1 (631-758) were also constructed into pAc5.1-GFP vector. All the primers used for vector construction were shown in Supplementary Table 1 .

Protein expression plasmids, including TAK1-GFP, TAK1-V5, TAB2-GFP, MyD88-GFP, Tube-GFP, and Pelle-GFP, and the reporter gene vectors, including DmDpt-pGL3, DmDrs-pGL3, DmAttA-pGL3, Pm411-pGL3, Pm536-pGL3, LvCrustin1-pGL3, 4×NF-κB-pGL3, and pRL-TK, were obtained from our previous studies (27–29).

To explore the potential interaction between LPS and each Toll, the plasmids expressing the extracellular (ec) domain of each Toll (Toll1/2/3/4/5/6/7/8/9-ec-HA), or pAc5.1A-3×HA (as a control), were transfected in Drosophila S2 cell line, respectively. Forty-eight hours post- transfection, cells were lysed in IP lysis buffer (Pierce, cat. no. 87788) with Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific, cat. no. 87786). The supernatants were incubated with 10 μg LPS-EB Biotin (In vivoGen, cat. no. tlrl-lpsbiot) at 4 °C for half an hour, followed by incubation with Monoclonal Anti-biotin agarose antibody produced in mouse (Merck, cat. no. A1559-5ML) at 4 °C for 2 hours.

An in vitro co-immunoprecipitation (Co-IP) assay was performed to detect the interaction between different pair of proteins. In brief, 48 hours after transfection, Drosophila S2 cells were lysed in IP lysis buffer (Pierce, cat. no. 87788) with Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific, cat. no. 87786). The supernatants were incubated with 30 µl of Monoclonal Anti-HA Agarose antibody produced in mouse (Merck, cat. no. A2095) or aAnti-GFP mAb Agarose (MBL International Corporation, cat. no. D153-8) or an Anti-V5 Agarose Affinity Gel antibody produced in micouse (Merck, cat. no. A7345) or ANTI-FLAG M2 Affinity Gel (Merck, cat. no. A2220-25ML) at 4 °C for two hours.

The agarose (from pull-down or co-immunoprecipitation) was washed five times with PBS before being subjected to SDS-PAGE and the western blot assays. In addition, 5% of total cell lysis was used tested as an input control. Membranes were developed with the Omni-ECL enhanced Pico Light Chemiluminescence Kit (EpiZyme, cat. no. SQ101L), and chemiluminescence was detected using the 5200 Chemiluminescence imaging System (Tanon).

The primary antibodies used in this study included rabbit anti-GFP N-terminal antibody produced in rabbit (Sigma-Aldrich, cat. no. G1544-100UL), Anti-V5 Epitope Tag Antibody (Merck, cat. no. AB3792), rabbit anti-HA antibody produced in rabbit (Merck, cat. no. H6908-100UL), rabbit anti-FLAG antibody produced in rabbit (Merck, cat. no. F7425), rabbit anti-p-LvDorsal (Genecreate), rabbit anti-LvDorsal (Genecreate), and mouse anti-actin clone C4 antibody (Merc, cat. no. MAB1501) and Monoclonal anti-β-Actin antibody produced in mouse (Merck, cat. no. A2228). The secondary antibodies used were anti-mouse IgG horseradish peroxidase (HRP-conjugate) (Promega, cat. no. W4021), anti-rabbit IgG HRP-conjugate (Promega, cat. no. W4011), anti-rabbit IgG (H+L) F (ab’)2 fragment Alexa Fluor 488 Conjugate (CST, cat. no. 4412S), and anti-mouse IgG (H+L), F (ab’)2 Fragment Alexa Fluor 594 Conjugate (CST, cat. no. 8890S).

The dsRNA-Dorsal-L, dsRNA-Dorsal-S, dsRNA-Dorsal-B (targeting both Dorsal-L and Dorsal-S), and dsRNA-GFP (as a control) were synthesized by in vitro transcription with T7 RiboMAX Express RNAi System kit (Promega, cat. no. P1700) ( Supplementary Table 1 ). The dsRNAs targeting Toll1/2/3/4/5/6/7/8/9, TAK1, and TRAF6 were obtained from our previous study (30, 31). Each shrimp was given an intraperitoneal injection of dsRNAs (2 μg/g shrimp in 50 µl of PBS) or an equivalent volume of PBS for gene silencing. Hemocytes were collected from shrimp at 48 hours after the dsRNA injection to evaluate the efficacy of RNAi ( Supplementary Table 1 ).

V. parahaemolyticus (MCCC: 1A10122), which was isolated from the hepatopancreas of diseased shrimp L. vannamei, was cultured in Luria broth (LB) medium overnight at 37°C. The microbial colony-forming units (CFUs) per milliliter of cultured V. parahaemolyticus were counted on LB agar plates. Shrimp were injected with 50 µl V. parahaemolyticus (approximately 1 × 10⁵ CFU). Six hours later, the hemocytes of three shrimp were sampled for detecting the phosphorylation levels of Dorsal.

Each shrimp was injected with 10 µg of dsRNA. And shrimp was received a second injection with PBS or 50 µl V. parahaemolyticus (approximately 1 × 10⁵ CFU) after 48 hours. The number of died shrimp was recorded every 4 h for five days.

Parallel experiments were also performed to sample shrimp tissues for further detection. Hemocytes were collected from at least three shrimp at 48 hours post dsRNA injection (for qPCR to detect the efficiency of RNAi), 6 hours post V. parahaemolyticus injection (for Western blotting and Immunofluorescence), and 48 hours post V. parahaemolyticus injection (for qPCR to detect the transcription level of LvCrustin1). The control C4 blots were derived from the same samples of the p-Dorsal blot.

QPCR were used to assess the transcription levels of target genes. The cDNA templates for the tissue distribution assay and challenge experiments were obtained according to a previous study (28). The expression levels of Dorsal-S, Dorsal-L, Toll1-9, and Crustin1 were determined and calculated by qPCR and Livak (2^-△△CT) method (32) after normalization to L. vannamei elongation factor 1 alpha (EF1-α). All samples were tested in triplicate. Primer sequences were listed in Supplementary Table 1 .

The Dual-luciferase reporter assays were done as our previous described (24) and measured with Dual-Glo Luciferase Assay System kit (Promega, cat. No. E2940). Western blotting was used to examine 50% of cell lysis as inputs.

The hemocytes from dsGFP- injected shrimp, dsTAK1- injected shrimp or dsTRAF6- injected shrimp at 6 hours post- V. parahaemolyticus infection were collected for immunofluorescence with rabbit anti-LvDorsal and mouse anti-β-Actin antibody.

The data are all displayed as mean ± SD. To compare groups of numerical data, the Student’s t- test was applied. Data was statistically analyzed for survival rates using Kaplan– ± Meier plot (log-rank χ² test). Statistical significance was defined as * p < 0.05, and ** p < 0.01.

Mammalian TLR receptors function as PRRs that directly sense microbial components (33). To date, nine tolls have been identified in the shrimp, L. vannamei (31). To explore whether any of the nine Tolls could interact with LPS, pull-down assays were performed in Drosophila S2 cells ( Figure 1A ). The extracellular parts of Toll1, Toll2, Toll3, Toll5, and Toll9 were found to bind to LPS ( Figure 1B ), implying that shrimp Tolls which recognize LPS may have a pattern similar to mammalian TLRs that detect pathogenic PAMPs.

MyD88 acts as an adaptor protein linking various Tolls/TLRs and downstream signal transduction proteins in the Toll/TLR pathways within mammals and Drosophila (31). Whether or not MyD88 can act as a downstream signaling mediator of the Toll pathway in shrimp has not yet been revealed. All nine Tolls and MyD88 contain one TIR domain. Thus, we explored the potential interaction between Toll and MyD88 using co-immunoprecipitation assays. We observed that GFP-tagged MyD88 was immunoprecipitated with each TIR domain of the nine Tolls (HA tag) with anti-HA antibody agarose affinity gels ( Figure 2A ). In addition, according to our previous studies, shrimp Tube (IRAK4 ortholog) acts as an adaptor protein to link MyD88 and Pelle (IRAK1 ortholog) to form a functional complex, MyD88-Tube-Pelle, that regulates Toll signaling (27). However, the downstream events of Pelle in shrimp and other invertebrates are still unclear. We inferred that shrimp TRAF6 might play a role in mediating signal transduction by interacting with Tube and Pelle. To address this, shrimp TRAF6 was co-expressed with Pelle or Tube, and we observed that HA-tagged TRAF6 interacted with GFP-tagged Pelle but not with the GFP-tagged Tube ( Figure 2B ). These findings suggest that, similar to TLR signaling in mammals, shrimp Tolls that sense LPS may initiate a signaling cascade involving the MyD88-Tube (IRAK4)-Pelle (IRAK1)-TRAF6 complex.

Since mammalian TRAF6 can form a complex with TAB2-TAK1 to activate NF-κB, we reasoned that shrimp TRAF6 could be involved in the formation of the TRAF6-TAB2-TAK1 complex. In vitro co-immunoprecipitation assays were performed to confirm the presence of the shrimp TRAF6-TAB2-TAK1 complex, and the interaction domains were determined using a series of truncated mutants ( Figure 3A ). We demonstrated that TRAF6 could only interact with TAK1 in the presence of TAB2, while TAB2 could bind to both TRAF6 and TAK1 directly ( Figure 3B ), suggesting that TAB2 functions as an adaptor linking TRAF6 and TAK1. The results also showed that the full-length TAB2 (1-608) was immunoprecipitated with TAK1 (631-758), which contained the coiled-coil region ( Figure 3C ). The C-terminal MATH domain of TRAF6 interacted with TAB2 ( Figure 3D ). Interestingly, although the coiled-coil region of TAB2 (321-410) was vital for its interaction with TAK1 ( Figure 3E ), the full-length TAB2 was required for its interaction with TRAF6 ( Supplementary Figure 1 ). In particular, TAB2 interacted with TAK1 through a combination of their coiled-coil regions, while the entire length of TAB2 was required to form a certain conformation to interact with the C-terminus of TRAF6 (MATH domain) ( Figure 3F ). Overall, our results indicate that shrimp TAB2 links TRAF6 to TAK1.

Dorsal is the NF-κB transcription factor that acts downstream of the Toll pathway in both shrimp and Drosophila (31, 34). Dorsal phosphorylation may reflect the activation of the Toll pathway (31). To determine whether LvDorsal responded to V. parahaemolyticus infection, the hemocytes from infected shrimp were detected by western blotting using the p-LvDorsal antibody based on phosphorylation Ser342 in the ‘VQLLRPSDKST’ peptide. Interestingly, the band appeared to be over 70 kDa ( Figure 4A ), whereas the LvDorsal band was only ~44 kDa. As a result, we believe in addition to the previous LvDorsal, another Dorsal of L. vannamei may exist. A new LvDorsal (ROT84343.1) was discovered after blasting in National Center for Biotechnology Information (NCBI), and the sequence was identified using PCR. Because ROT84343.1 was substantially larger than the previous LvDorsal (ACZ98167.1), we termed ROT84343.1 LvDorsal-L and ACZ98167.1 LvDorsal-S. The LvDorsal-L sequence is shown in Supplementary Figure 2A . Similar to Dorsal in other species, architecture prediction showed that LvDorsal-L possessed a Rel-homology domain (RHD) DNA-binding domain, followed by an RHD dimer domain ( Supplemental Figure 2B ). Multiple sequence alignment revealed that the full-length amino acid sequence of LvDorsal-S was almost fully consistent with the N-terminal of LvDorsal-L, with only three aa missing in the N-terminal and 10 aa differences in the C-terminal ( Supplementary Figure 3 ).

qPCR was used to determine the tissue distribution of LvDorsal-L and LvDorsal-S. Both LvDorsal-S and LvDorsal-L mRNA were detected in all examined tissues ( Figure 4B ). Compared with LvDorsal-S, LvDorsal-L was the major dorsal isoform in all tested tissues. LvDorsal-L was highly expressed in hemocytes, gills, and intestines. LvDorsal-S was relatively abundant in hemocytes and showed low expression levels in other tissues. Therefore, hemocytes were chosen to measure transcriptional changes in LvDorsal-S and LvDorsal-L after PAMP and pathogen infection. The results showed that from 4 to 72 hours post V. parahaemolyticus infection, LvDorsal-L plays a continuous role in the immune response induced by V. parahaemolyticus ( Supplementary Figure 4 ).

AMPs play a vital role in protecting the host from invading microorganisms and are primarily induced by NF-κB. Dual-luciferase reporter assays were performed to determine the effect of LvDorsal-S and LvDorsal-L on the regulation of AMP induction. As shown in Figure 4C , LvDorsal-L significantly activated the promoters of AMP genes, including the Drosophila antimicrobial peptide DmDpt (~7.21-fold), DmDrs (~124.74-fold), DmAttA (~152.70-fold), Penaeus antimicrobial peptide Pm411 (~2.27-fold), Pm536 (~1.64-fold), LvCrustin1 (~245.49-fold), and artificial 4×NF-κB (~80.88-fold). Unexpectedly, overexpression of LvDorsal-S only slightly increased the promoter activities of DmAttA (~2.04-fold) and LvCrustin1 (~2.16-fold) and even decreased DmDpt (~0.61-fold) and Pm536 (~0.59-fold) promoter activity. The activation of the DmDrs, Pm411, or 4×NF-κB promoters did not change significantly during LvDorsal-S overexpression. LvCrustin1, also known as LvCrustina-2, belongs to type II crustin, which has antibacterial activity against both gram-positive and gram-negative bacteria and was the first crustin identified in L. vannamei (35). These results suggest that LvDorsal is involved in the regulation of shrimp and Drosophila antimicrobial peptides, while LvDorsal-L is the isoform with stronger inductive activity.

RNAi of LvDorsal-S, LvDorsal-L, and LvDorsal-B was used to investigate whether LvDorsal-S and LvDorsal-L were involved in host defense against V. parahaemolyticus. The dsRNA-LvDorsal-S and dsRNA-LvDorsal-L primers were situated at the C-terminus of LvDorsal-S and LvDorsal-L, respectively. In contrast, the dsRNA-Dorsal-B primers were situated in the RHD DNA-binding domain to knockdown both LvDorsal-S and LvDorsal-L, respectively ( Figure 5A ). As shown in Figure 5B , dsRNA-LvDorsal-S and dsRNA-LvDorsal-B remarkably suppressed the transcription of LvDorsal-S, while LvDorsal-L was inhibited by dsRNA-LvDorsal-L and dsRNA-LvDorsal-B ( Figure 5C ). The shrimp were then injected with V. parahaemolyticus 48 h post-dsRNA injection, and survival rates were monitored for 120 h after the infection. As shown in Figure 5D , the survival rates in the dsRNA-LvDorsal-L and dsRNA-LvDorsal-B groups were lower than those in the dsRNA-GFP control group, indicating that shrimp were more vulnerable to V. parahaemolyticus infection. Notably, the survival rate of the dsRNA-LvDorsal-S group was not significantly different (p = 0.28) but demonstrated a lower trend than the control group. To gain more information on the effect of LvDorsal on LvCrustin1 induction during V. parahaemolyticus infection, a parallel experiment was performed to explore the transcription levels of LvCrustin1 in LvDorsal-silenced shrimp following infection. In accordance with the survival rates, the transcription level of LvCrustin1 was much lower in the dsRNA-LvDorsal-L and dsRNA-LvDorsal-B groups than in the control group ( Figure 5E ). Overall, our data suggests that LvDorsal-L, rather than LvDorsal-S, is a key factor in protecting the host from V. parahaemolyticus infection and inducing LvCrustin1 production.

Given that LvDorsal-L was the predominant isoform that responded to V. parahaemolyticus infection in shrimp, we presumed that activation of LvDorsal-L was critical for shrimp action against V. parahaemolyticus. Considering that LvToll1/2/3/5/9 could bind to LPS, we wondered whether Toll1, Toll2, Toll3, Toll5, and Toll9 were involved in LvDorsal-L activation during V. parahaemolyticus infection. To select the proper organ for the following experiments, the tissue distribution of LvTolls was assessed using qPCR. The results indicated that all nine Tolls were highly expressed in hemocytes ( Figure 6A ), therefore, hemocytes were chosen as the target tissue in the evaluation of the silencing efficiencies for each Toll by qPCR 48 h post dsRNA injection ( Figure 6B ). Next, we challenged RNAi-treated shrimp with V. parahaemolyticus and subsequently analyzed the phosphorylation level of LvDorsal by western blotting. As shown in Figure 6C , the phosphorylation levels of LvDorsal were inhibited in the dsRNA-LvToll1, dsRNA-LvToll2, dsRNA-LvToll3, dsRNA-LvToll5, and dsRNA-LvToll9 groups, suggesting that LvToll1/2/3/5/9 may be involved in regulating LvDorsal-L activation during V. parahaemolyticus infection.

Although we indicated that LvToll1/2/3/5/9 activated LvDorsal-L, it was still unclear whether the TRAF6-TAB2-TAK1 complex, a critical component of the TLR pathway, stimulated LvDorsal-L activity. In this study, RNAi was used to knock down LvTRAF6 or LvTAK1 in shrimp, followed by infection with V. parahaemolyticus. As shown in Figure 7A , the transcription of LvTRAF6 was successfully inhibited by dsRNA-LvTRAF6, 48 h post dsRNA injection. Efficient silencing of LvTAK1 was observed in the dsRNA-LvTAK1 group after 48 h post-dsRNA injection using qPCR ( Figure 7B ). Six hours post-V. parahaemolyticus infection, the phosphorylation of LvDorsal-L in hemocytes was much lower in the dsRNA-LvTRAF6 and dsRNA-LvTAK1 groups than in the dsRNA-GFP group ( Figures 7C, D ). In accordance with the Dorsal-L phosphorylation results, immunofluorescence experiments in the hemocytes demonstrated that endogenous LvDorsal-L (green fluorescence) co-localized less with the nucleus after 6 h post-infection in both dsRNA-LvTRAF6 and dsRNA-LvTAK1 hemocytes, compared to dsRNA-GFP hemocytes ( Figures 7E ). These findings strongly suggested that the TRAF6-TAB2-TAK1 complex in shrimp is essential for LvDorsal-L activation during V. parahaemolyticus infection.

In summary, during V. parahaemolyticus infection, LvToll1/2/3/5/9 recognized LPS and recruited the TRAF6-TAB2-TAK1 complex via the MyD88-Tube-Pelle cascade, leading to the activation of LvDorsal-L, which plays a major antibacterial role in L. vannamei via the induction of LvCrustin1, implying that shrimp may fight bacterial infection via the Toll-TRAF6/TAB2/TAK1-Dorsal pathway ( Figure 8 ).