T. ovatus were purchased from a commercial fish farm in Hainan, China. Before the experiment, the fish were acclimated at 26°C in a flow-through water system for one week. All animal experiments conducted during this study were approved by the Animal Care and Use Committee at Hainan University. Human embryonic kidney (HEK293T) cells were cultured at ambient CO₂ and 37°C in Dulbecco’s modified Eagle’s medium (DMEM, USA), the DMEM contained 1% penicillin and streptomycin (Pen Strept) and 10% foetal bovine serum (FBS). Golden pompano snout (GPS) cells were generously donated by the South China Sea Institute of Oceanology Chinese Academy of Sciences. The GPS cells were grown at 26°C with Leibovitz’s L-15 medium containing 10% FBS (Gibco, USA) (42, 43). V. harveyi was grown at 28°C in Luria-Bertani broth (38).

Based on the transcriptome data, the primers TroIRF3-F/TroIRF3-R were used to clone the open reading frame (ORF) of TroIRF3 by polymerase chain reaction (PCR) amplification ( Table S1 ). Sequencing was performed following gel purification of the PCR products into the pEASY^®-T1Simple Vector (Transgen, Beijing, China). A homology search of the TroIRF3 protein sequence was conducted using BLAST programs on NCBI (http://www.ncbi.nlm.nih.gov/blast). Signal peptide prediction was performed using SignalP 5.1. Thereafter, DNAMAN (Lynnon Biosoft, USA) was used to align the amino acid sequences. SWISS-MODEL was used to predict the TroIRF3’s tertiary structure. A phylogenetic analysis was performed using neighbour-joining methods through MEGA 6.0.

A total of 11 tissues were obtained from 15 healthy T. ovatus (blood, head kidney, liver, intestine, spleen, gill, muscle, skin, brain, stomach and heart). To avoid the affection of the individual difference of fish, the fish tissues were mixed aseptically into a sample with five identical samples. The tissue samples were stored at −80°C until RNA extraction was performed. The T. ovatus were divided into two groups: control and bacterial challenge. The challenge group received 0.1 ml of V. harveyi (2 × 10⁶ CFU/ml), while the control group received the same volume of PBS. At 6, 9, 12 and 24 hours post-injection (hpi), 15 individuals in each group were randomly sacrificed, and the liver, spleen and head-kidney were sampled. Five identical tissues from each group were mixed aseptically to form a sample. The total RNA was extracted from the harvested samples using the Total RNA Extraction Kit (Promega, China). The cDNA was synthesised using the Reverse Transcription System (Promega, China). An SYBR ExScript qRT–PCR kit (LS2062, Promega, China) was used to perform quantitative real-time PCR (qRT-PCR) on a qTOWER3 Real-Time PCR (qPCR) system (Analytik Jena, Germany). The internal reference gene was β-2 microglobulin (B2M), and the relative expression levels of TroIRF3 were calculated using the 2^{^-ΔΔCT} method (44). The primers are summarised in Table S1 .

The eukaryotic expression vector pTroIRF3 was constructed for TroIRF3 overexpression in vivo. TroIRF3 was cloned into the pCN3’s Sma I site and transformed into Escherichia coli, DH5α. pTroIRF3 was extracted using the Endo Free Plasmid Kit (Tiangen, China). The fish (weight 15.2 ± 3.1 g) were randomised into three groups (N=12), with each receiving a 0.1 ml injection of pTroIRF3 (150 μg/ml), pCN3 (empty vector, 150 μg/ml) or PBS (control). To detect the expression level of TroIRF3, the liver, spleen and head-kidney of three fish from each group were collected after five days of intramuscular injection of plasmids. After the isolation of total RNA and cDNA synthesis, TroIRF3 expression was detected by qRT-PCR with primer TroIRF3-RT-F/R ( Table S1 ). B2M was used as the internal reference gene. V. harveyi (10⁵ CFU/fish) was injected into the remaining fish. The liver, spleen and head-kidney of three fish in each group were homogenised at 6, 9 and 12 hpi for bacterial count analysis.

The small interfering RNA (siRNA) of TroIRF3 was synthesised using the T7 RiboMAX™ Express RNAi System (Promega, USA). The primers used for siRNA synthesis are listed in Table S1 . A total of 36 fish were randomly divided into three groups of 12 each and injected with 0.1 ml of PBS (control), siRNA-C (control, 15 μg/fish) or siRNA-TroIRF3 (15 μg/fish). At 12 h after intramuscular injection of siRNA, three fish were sacrificed for liver, spleen and head-kidney collection. Then, qRT–PCR was performed to detect the TroIRF3 expression levels. The rest of the fish were infected with V. harveyi (10⁵ CFU/fish). The liver, spleen and head-kidney of three fish in each group were homogenised at 6, 9, and 12 hpi for bacterial count analysis.

The eukaryotic expression vector pTroIRF3-N3 was created by cloning TroIRF3’s ORF into the Bgl II and Sal I sites of pEGFP-N3. Subcellular localisation experiments were performed to examine TroIRF3 localisation in the cell with or without V. harveyi infection. The cells were seeded on 24-well plates and grown overnight, and then Lipofectamine 2000 (Invitrogen, USA) was used to transfect pEGFP-N3 or pTroIRF3-N3 into the GPS cells. At 48 h post-transfection, the cells transfected with pTroIRF3-N3 were treated with 0.1 ml PBS (control) or V. harveyi (1×10⁵ CFU/ml) for 6 h. The GPS cells were washed, stained and visualised with fluorescence microscopy (Leica, Wetzlar, Germany) after fixing in 4% paraformaldehyde.

Different PCR fragments of TroIRF3 were inserted into Bgl II and Sal I sites of pEGFP-N3 vector to constructe TroIRF3-GFP truncation mutants: pTroIRF3-(△NES)-N3, pTroIRF3-1-165-N3, pTroIRF3-1-417-(△NES)-N3, pTroIRF3-64-466-(△NES), or pTroIRF3-127-466-(△NES). Various plasmids were transfected into the GPS cells to detect the NLSs of TroIRF3. The cells were fixed, washed, stained and visualised under fluorescence microscopy at 48 h post-transfection.

Two TroIRF3 variants (KR74/75NG and RK82/84LQ) were designed to further detect the NLSs of IRF3 within the aa 64–127 region. The PCR fragments of KR74/75NG and RK82/84LQ were inserted into Bgl II and Sal I sites of pEGFP-N3 vector to construct two recombinant plasmids: pTroIRF3-KR74/75NG-N3 and pTroIRF3-RK82/84LQ-N3. The primers used are listed in Table S1 . The GPS cells were transfected with plasmids (pTroIRF3-KR74/75NG-N3 or pTroIRF3-RK82/84LQ-N3). After 48 h, the cells were treated with 0.1 ml V. harveyi (1×10⁵ CFU/ml) or PBS (control) for 6 h. Afterwards, the cells were fixed, washed, stained and visualised under fluorescence microscopy.

The TroIRF3 binding site in the promoter of TroIFNa3 was evaluated by luciferase reporter assay. Three truncation mutants of the TroIFNa3 promoter (−1649 to +1, –896 to +1 and −722 to +1) were amplified and inserted into the Kpn I and Xho I restriction sites of the pGL4-basic luciferase reporter plasmid (Promega, USA) to obtain the recombinant plasmid pGL4-TroIFNa3-p1, pGL4-TroIFNa3-p2 and pGL4-TroIFNa3-p3. The primers used are summarised in Table S1 . Lipofectamine 2000 (Invitrogen, USA) was used to cotransfect pRL-TK (0.025 μg), pGL4-TroIFNa3-p1/p2/p3 (0.25 μg) and pTroIRF3 (0.25 μg) into the HEK293T cells. pRL-TK was used as an internal control. After 36 h of transfection, the firefly and renilla luciferase activities were measured using a dual-luciferase reporter assay system (Promega, USA). The activity ratio between the firefly luciferase and the renilla luciferase was used to calculate the luciferase activity. The effect of V. harveyi stimulation on the TroIFNa3 promoter activation mediated by TroIRF3 was detected. After 24 h of transfection, the transfected HEK293T cells were treated with PBS or V. harveyi (1×10² CFU/ml) for 12 h. Luciferase activity was measured in the lysed cells as described above.

Luciferase reporter assay was performed to detect whether the activation of the TroIFNa3 promoter was dependent on the pTroIRF3 concentration. The HEK293T cells were cotransfected with pRL-TK (0.025 μg), pGL4-TroIFNa3-p2 (0.25 μg) and different concentrations of pTroIRF3 (0, 0.01, 0.05, 0.1 or 0.2 μg). After 24 h, the transfected HEK293T cells were treated with V. harveyi (1×10² CFU/ml) for 12 h. Luciferase activity was measured in the lysed cells as described above.

Luciferase reporter assay was performed to detect the role of the basic amino acids (KR74/75 and RK82/84) in the TroIRF3-triggered TroIFNa3-p2 activation. The HEK293T cells were cotransfected with pGL4-TroIFNa3-p2, wild-type TroIRF3 or its variants (pTroIRF3-KR74/75NG or pTroIRF3-RK82/84LQ) and pRL-CMV. After 24 h, the cells were treated with V. harveyi (1×10² CFU/ml) for 12 h. Afterwards, the relative luciferase activity was tested as described above.

All data were analysed by SPSS 16.0 (SPSS, IL, USA) and GraphPad Prism 5 (GraphPad, CA, USA). The statistical significance level of this study was set at 0.05.

The full length of TroIRF3 cDNA contains a 1398 bp ORF encoding 465 amino acids (GenBank Accession No. AWY04222.1). In theory, it has a molecular mass of 52 kDa and an isoelectric point of 4.89. According to the amino acid alignments, TroIRF3 contains three conserved domains, namely, DBD, IAD and SRD ( Figure 1A ). Among the selected IRF3 sequences, TroIRF3 was closest to that of Seriola lalandi dorsalis (85.95%), followed by those of Siniperca chuatsi and Lateolabrax japonicus (81.94% and 82.37%, respectively) ( Figure 1A ). According to the SWISS-MODEL online software, TroIRF3 has a similar three-dimensional structure to human IRF3 ( Figure 1B ). In addition, the IRF3 from fish and mammals were located in two separate branches, while TroIRF3 was inserted in the fish cluster. TroIRF3 shows a close evolutionary relationship with S. lalandi dorsalis IRF3 ( Figure 1C ).

The expression level of TroIRF3 was detected in 11 normal tissues (spleen, head-kidney, skin, muscle, gill, heart, liver, brain, stomach, blood and intestine) by qRT-PCR. TroIRF3 was constitutively distributed in all the tissues examined. TroIRF3 expression was high in the gill and liver but low in the spleen and blood ( Figure 2A ).

The level of TroIRF3 expression following V. harveyi infection was detected using qRT-PCR ( Figure 2B ). The level of TroIRF3 expression in the liver was increased from 6 hpi to 24 hpi (4.1-, 30.9-, 4.1-, and 2.7-folds at 6, 9, 12, and 24 hpi, respectively), the expression level of TroIRF3 in the spleen was increased from 6 hpi to 24 hpi (3.5-, 40.2-, 6.5-, and 7.8-folds at 6, 9, 12, and 24 hpi, respectively), and the expressions of TroIRF3 in the head-kidney were increased from 9 hpi to 24 hpi but manifested no difference at 6 hpi (35.5-, 8.8-, and 4.0-folds at 9, 12, and 24 hpi, respectively). Based on these results, TroIRF3 may contribute to antibacterial immunity.

To clarify TroIRF3’s role in antibacterial immunity, we assessed the effects of TroIRF3 overexpression on the fish’s antibacterial ability. An increase in TroIRF3 expression was observed in fish treated with pTroIRF3, indicating the overexpression of TroIRF3 ( Figure S1A ). Compared with the control group, the number of bacteria in the tissues decreased in the TroIRF3-overexpressing fish ( Figure 3A ). The bacterial load in the liver was significantly decreased in the TroIRF3-overexpressing fish compared with the control fish by 7.1- and 8.3-fold at 6 and 9 hpi, respectively. The pattern of bacterial load in the spleen of the pTroIRF3-overexpressing fish was similar to that in the liver, which decreased significantly with 5.0- and 10.9-fold at 6 and 9 hpi, respectively. Compared with the control fish, TroIRF3 overexpression significantly reduced the number of bacterial colonies in the head-kidney by 6.8- and 9.6-fold at 9 and 12 hpi, respectively. No significant difference in bacterial loads was observed between the empty vector-injected fish and the control fish, which showed that the vector had no effect on the fish’s antibacterial immune responses.

The effects of TroIRF3 on antibacterial immune responses were further investigated, and TroIRF3 was knocked down using siRNA administration. The TroIRF3 gene was decreased in vivo using siRNA ( Figure S1B ). Compared with the control fish, the number of bacteria in the tissues obviously increased in the siTroIRF3-treated fish ( Figure 3B ). Compared with the control fish, the TroIRF3 knockdown fish exhibited an increased number of bacteria in the liver by 4.2- and 1.5-fold at 6 and 9 hpi, respectively. At 6, 9 and 12 hpi, the siTroIRF3-treated fish had higher bacterial colony counts in the spleen (15.4-, 3.3- and 4.9-fold). Compared with the control group, the siTroIRF3-treated fish had 9.5- and 2.8-fold higher bacterial loads in the head-kidneys at 6 and 9 hpi, respectively. The PBS-treated group and siTroIRF3-C-treated group (siRNA-C) showed similar bacterial loads, which indicated that the siRNA-C had no effect on the fish’s antibacterial immune responses. Together, TroIRF3 was involved in the antibacterial response against V. harveyi in vivo.

The subcellular localisation of TroIRF3 with or without V. harveyi stimulation was studied using a fluorescence microscope. EGFP is detected in both cytoplasm and nucleus, TroIRF3 was located exclusively in the cytoplasm in the uninfected cells and translocated to the nucleus after infection with V. harveyi ( Figure 4A ). Given the presence of both NLS and NES in IRF3, its predominant cytoplasmic localisation suggested constitutive activities of NES before virus infection, nuclear exports dominated nuclear imports (45). The regulation function of IRF3 was dependent on its nuclear location. To define the TroIRF3’s NLS domain, the NES was inactivated to prevent interference. The protein of TroIRF3 (△NES) was located in the nucleus and cytoplasm of the GPS cells, suggesting that TroIRF3 is capable of entering and staying in the nucleus without NES ( Figure 4C ). To produce different mutants, TroIRF3’s N-terminal and C-terminal regions were truncated ( Figure 4B ). The C-terminal truncations (aa 1–417 and 1–165) and N-terminal truncations (aa 64–466) fused with GFP retained translocation into the nuclei, but the aa 127-466 did not ( Figure 4C ). According to these experiments, the TroIRF3’s NLS domain is located at the N-terminus of aa 64–127.

According to a previous study, mice IRF3 nuclear import requires both basic residue clusters (KR77/78 and RK86/87) (45). The TroIRF3’s NLS domain lies between amino acids 64 and 127. Site-directed mutagenesis was then constructed to determine whether they contribute to the nuclear accumulation of TroIRF3. In TroIRF3 mutants (KR74/75NG and RAK82/84LAQ), basic amino acids (KR74/75 and RK82/84) were replaced with uncharged amino acids ( Figure 5A ). The results of the subcellular localisation assays showed that V. harveyi infection caused TroIRF3, but not KR74/75NG and RK82/84LQ, to be localized in the nucleus ( Figures 4A , 5B ). Therefore, the NLS of TroIRF3 is composed of amino acid residues KR74/75 and RK82/84, which are critical in TroIRF3 nuclear accumulation.

To determine whether NLS affects TroIRF3-mediated antibacterial defences, the number of V. harveyi in the fish treated with pTroIRF3 or the indicated TroIRF3 mutants (pTroIRF3-KR74/75NG and pTroIRF3-RK82/84LQ) was tested. TroIRF3 overexpression significantly suppressed V. harveyi replication in the fish, which was in line with our previous observations ( Figure 6 ). However, the number of invaded bacteria in pTroIRF3-KR74/75NG- or pTroIRF3-RK82/84LQ-treated fish was enhanced compared with that in the wild-type pTroIRF3-treated fish ( Figure 6 ). These results showed that basic amino acids KR74/75 and RK82/84 are required for the efficient antibacterial function of TroIRF3.

To investigate the effects of NLS on the TroIRF3-mediated immune gene expression, the expression levels of immune genes (IFNa3, MXI, TRAF6, Viperin1 and MAVS) in the spleen of the fish treated with pTroIRF3 or the indicated TroIRF3 mutants (pTroIRF3-KR74/75NG and pTroIRF3-RK82/84LQ) were tested. In the pTroIRF3-, pTroIRF3-KR74/75NG- and pTroIRF3-RK82/84LQ-treated fish, the expression level of TroIRF3 and modified TroIRF3 mutants were significantly upregulated compared with the control fish ( Figure S2 ). TroIRF3 remarkably enhanced the production of IFNa3, MXI, TRAF6, Viperin1 and MAVS. However, the expression level of the immune genes (IFNa3, MXI, TRAF6, Viperin1 and MAVS) in the pTroIRF3-KR74/75NG- or pTroIRF3-RK82/84LQ-treated fish was significantly reduced compared with that in the pTroIRF3-treated fish ( Figure 7 ). Furthermore, the fish treated with pCN3, pTroIRF3- KR74/75NG or pTroIRF3-RK82/84LQ showed the same expression level of immune genes ( Figure 7 ). These results showed that basic amino acids KR74/75 and RK82/84 are required for immune gene induction by TroIRF3.

Based on the predicted binding sites, consecutive truncated mutants of the TroIFNa3 promoter were constructed to investigate the TroIRF3-binding region in the TroIFNa3 promoter ( Figure 8A ). TroIFNa3-p2 had a higher response to TroIRF3 than the other mutants, suggesting that the region consists of –896 bp to +1 bp from the transcriptional start site, which contains the binding site for IRF3 ( Figure 8A ). The activation of the TroIFNa3 promoter by TroIRF3 was further enhanced by V. harveyi stimulation ( Figure 8B ). Furthermore, the activation of TroIFNa3-p2 is dependent on the TroIRF3 concentration ( Figure 8C ).

Subsequently, the potential effects of NLS on TroIRF3-mediated TroIFNa3-p2 activation were investigated using the luciferase reporter assay. TroIRF3 exhibited remarkably enhanced TroIFNa3-p2 activation following V. harveyi infection, whereas the other variants (pTroIRF3-KR74/75NG and pTroIRF3-RK82/84LQ) did not exhibit TroIFNa3-p2 activation under the same conditions ( Figure 8D ).