Mutual Regulation of NOD2 and RIG-I in Zebrafish Provides Insights into the Coordination between Innate Antibacterial and Antiviral Signaling Pathways

Nucleotide-binding oligomerization domain-containing protein 2 (NOD2) and retinoic acid-inducible gene I (RIG-I) are two important cytosolic pattern recognition receptors (PRRs) in the recognition of pathogen-associated molecular patterns (PAMPs), initiating innate antibacterial and antiviral signaling pathways. However, the relationship between these PRRs, especially in teleost fish models, is rarely reported. In this article, we describe the mutual regulation of zebrafish NOD2 (DrNOD2) and RIG-I (DrRIG-I) in innate immune responses. Luciferase assays were conducted to determine the activation of NF-κB and interferon signaling. Morpholino-mediated knockdown and mRNA-mediated rescue were performed to further confirm the regulatory roles between DrNOD2 and DrRIG-I. Results showed that DrNOD2 and DrRIG-I shared conserved structural hallmarks with their mammalian counterparts, and activated DrRIG-I signaling can induce DrNOD2 production. Surprisingly, DrNOD2-initiated signaling can also induce DrRIG-I expression, indicating that a mutual regulatory mechanism may exist between them. Studies conducted using HEK293T cells and zebrafish embryos showed that DrRIG-I could negatively regulate DrNOD2-activated NF-κB signaling, and DrNOD2 could inhibit DrRIG-I-induced IFN signaling. Moreover, knocking down DrRIG-I expression by morpholino could enhance DrNOD2-initiated NF-κB activation, and vice versa, which could be rescued by their corresponding mRNAs. Results revealed a mutual feedback regulatory mechanism underlying NOD2 and RIG-I signaling pathways in teleosts. This mechanism reflects the coordination between cytosolic antibacterial and antiviral PRRs in the complex network of innate immunity.


Activation of DrRIG-I Signaling Inducing DrNOD2 Expression and Vice Versa
After confirming DrRIG-I and DrNOD2 conservation structurally, we further investigated the mutual regulation between DrRIG-I and DrNOD2. In mammals, the activation of RIG-I signaling

Activation of DrRIG-I Signaling Inducing DrNOD2 Expression and Vice Versa
After confirming DrRIG-I and DrNOD2 conservation structurally, we further investigated the mutual regulation between DrRIG-I and DrNOD2. In mammals, the activation of RIG-I signaling pathway could induce the expression of NOD2. However, whether this induction was conserved in zebrafish and whether DrNOD2 signaling activation induced the expression of DrRIG-I remain unknown. In this experiment, a stimulatory mutant of DrRIG-I containing only CARDs (DrRIG-I-CARD) was used, instead of wild-type DrRIG-I, because the latter remains in an auto-inhibited state without ligand stimulation [37]. In addition to DrNOD2, a mutant without the LRR domain (DrNOD2 (∆LRR)) was also used to activate the NOD2 signaling. Our results showed that DrNOD2 was induced at 6 and 12 h post injection (hpi) of DrRIG-I (CARD) and reached its highest level at 6 h. Surprisingly, we also observed the induction of DrRIG-I when DrNOD2 signaling was activated at 24 hpi (Figure 2A,B). Moreover, the induction of DrRIG-I could be observed as early as 12 hpi in the DrNOD2 (∆LRR)-injected group ( Figure 2C). Thus, we inferred a feedback regulatory mechanism between these two cytosolic signaling pathways. pathway could induce the expression of NOD2. However, whether this induction was conserved in zebrafish and whether DrNOD2 signaling activation induced the expression of DrRIG-I remain unknown. In this experiment, a stimulatory mutant of DrRIG-I containing only CARDs (DrRIG-I-CARD) was used, instead of wild-type DrRIG-I, because the latter remains in an auto-inhibited state without ligand stimulation [37]. In addition to DrNOD2, a mutant without the LRR domain (DrNOD2 (ΔLRR)) was also used to activate the NOD2 signaling. Our results showed that DrNOD2 was induced at 6 and 12 h post injection (hpi) of DrRIG-I (CARD) and reached its highest level at 6 h. Surprisingly, we also observed the induction of DrRIG-I when DrNOD2 signaling was activated at 24 hpi (Figure 2A,B). Moreover, the induction of DrRIG-I could be observed as early as 12 hpi in the DrNOD2 (ΔLRR)-injected group ( Figure 2C). Thus, we inferred a feedback regulatory mechanism between these two cytosolic signaling pathways. One-cell stage embryos were injected with DrRIG-I (CARD) (100 pg/embryo). DrNOD2 was induced 6 and 12 h post injection (hpi), with the highest level at 6 hpi; (B and C) NOD2 signaling activation induced the expression of DrRIG-I. One-cell stage embryos were injected with DrNOD2 (B) and DrNOD2 (ΔLRR) (C) (100 pg/embryo). At 24 hpi, the expression of DrRIG-I was upregulated in the DrNOD2-injected group (B); At 12 and 24 hpi, RIG-I was induced in the DrNOD2 (ΔLRR) injection group, with higher expression at 24 hpi (C). All experiments were conducted with three replicates, and 80-100 zebrafish embryos were collected for the analysis. Values are expressed as mean ± SD; *p < 0.05, **p < 0.01.

DrRIG-I Negative Regulation of DrNOD2-Initiated Signaling
To prove our hypothesis, we determined the function of DrRIG-I in DrNOD2-initiated NF-κB activation in the HEK293T cells and zebrafish embryos. The results showed that administration of DrNOD2 or DrNOD2 (ΔLRR) alone led to a robust NF-κB activation in the HEK293T cells ( Figure  3A) and zebrafish embryos ( Figure 3B). Activation was suppressed when DrRIG-I or DrRIG-I-CARD was co-injected ( Figure 3A,B). The mutant without the CARD domain DrRIG-I (ΔCARD) was also used as negative control, and this mutant could not inhibit the DrNOD2-and DrNOD2 (ΔLRR)-activated NF-κB signaling ( Figure 3B). The DrRIG-I translation in the zebrafish embryos was knocked down using DrRIG-I MO to confirm the negative role of DrRIG-I in NOD2 signaling. The efficiency of MO was examined before use, and significant knockdown efficiency was observed ( Figure 3C). Co-administration DrRIG-I MO resulted in the elevation of the DrNOD2 and DrNOD2 (ΔLRR)-mediated NF-κB activation and TNF-α expression. Good rescue was achieved with the simultaneous injection of DrRIG-I mRNA ( Figure 3D,E). MDP-initiated NOD2 signaling was also suppressed when DrRIG-I or DrRIG-I-CARD was co-injected in the zebrafish embryos ( Figure 3F). These results demonstrated the negative roles of DrRIG-I in NOD2 signaling. No apparent developmental defects were observed (data not shown). (∆LRR) injection group, with higher expression at 24 hpi (C). All experiments were conducted with three replicates, and 80-100 zebrafish embryos were collected for the analysis. Values are expressed as mean ± SD; * p < 0.05, ** p < 0.01.

DrRIG-I Negative Regulation of DrNOD2-Initiated Signaling
To prove our hypothesis, we determined the function of DrRIG-I in DrNOD2-initiated NF-κB activation in the HEK293T cells and zebrafish embryos. The results showed that administration of DrNOD2 or DrNOD2 (∆LRR) alone led to a robust NF-κB activation in the HEK293T cells ( Figure 3A) and zebrafish embryos ( Figure 3B). Activation was suppressed when DrRIG-I or DrRIG-I-CARD was co-injected ( Figure 3A,B). The mutant without the CARD domain DrRIG-I (∆CARD) was also used as negative control, and this mutant could not inhibit the DrNOD2and DrNOD2 (∆LRR)-activated NF-κB signaling ( Figure 3B). The DrRIG-I translation in the zebrafish embryos was knocked down using DrRIG-I MO to confirm the negative role of DrRIG-I in NOD2 signaling. The efficiency of MO was examined before use, and significant knockdown efficiency was observed ( Figure 3C). Co-administration DrRIG-I MO resulted in the elevation of the DrNOD2 and DrNOD2 (∆LRR)-mediated NF-κB activation and TNF-α expression. Good rescue was achieved with the simultaneous injection of DrRIG-I mRNA ( Figure 3D,E). MDP-initiated NOD2 signaling was also suppressed when DrRIG-I or DrRIG-I-CARD was co-injected in the zebrafish embryos ( Figure 3F). These results demonstrated the negative roles of DrRIG-I in NOD2 signaling. No apparent developmental defects were observed (data not shown). HEK293T cells (1 µg/mL) or one-cell stage embryos (100 pg/embryo) were administered with DrNOD2 or DrNOD2 (∆LRR). The cells/embryos were collected at 24 h post transfection/injection. Luciferase assays showed robust NF-κB activation. This activation was evidently inhibited when DrRIG-I or DrRIG-I-CARD was co-administered with DrNOD2 or DrNOD2 (∆LRR). However, no inhibition was observed in the DrRIG-I (∆CARD) co-administered group; (C) Examination of knockdown efficiency of DrRIG-I MO. The 5 UTR sequence (complement to the MO sequence) of DrRIG was amplified and inserted into the EGFP-N1 vector. One-cell stage embryos were injected with the constructed vector (100 pg/embryo) and the control MO or DrRIG MO (4 ng/embryo). The embryos were collected at 24 hpi, and phase contrast images and GFP fluorescence were observed to examine the knockdown efficiency, scale bar, 400 µm; (D,E) The role of DrRIG-I in DrNOD2-initiated signaling was confirmed via MO-mediated knockdown and mRNA rescue in zebrafish embryos. One-cell stage embryos were injected with DrNOD2 or DrNOD2 (∆LRR) (100 pg/embryo) and control MO or DrRIG MO (4 ng/embryo) or together with DrRIG mRNA (100 pg/embryo). At 24 hpi, NF-κB activation and TNFα production were elevated when DrRIG-I MO was co-administered. Good rescue was achieved with the simultaneous injection of DrRIG-I Mrna; (F) DrRIG-I negatively regulated MDP-initiated NOD2 signaling in zebrafish embryos. One-cell stage embryos were administered with 2 nL (1 µg/µL) of MDP. The embryos were collected at 24 hpi. Luciferase assay results showed robust NF-κB activation, which was inhibited by the co-administration of MDP with DrRIG-I or DrRIG-I-CARD. All luciferase assays and qRT-PCR were conducted with three replicates, and each replicate contained the extracts from the cells of a well from the six-well plate or 80-100 zebrafish embryos. Values are expressed as mean ± SD; * p < 0.05, ** p < 0.01.

DrNOD2 Negative Regulation of DrRIG-I-Initiated Signaling
The roles of DrNOD2 in DrRIG-I-induced IFN signaling were also examined. DrRIG-I (CARD) overexpression induced the ISG15 expression in the HEK293T cells ( Figure 4A) and Mx in the zebrafish embryos ( Figure 4B). ISG15 and Mx are interferon-stimulated genes (ISGs) that indicate the activation of IFN signaling. This activation was suppressed when DrNOD2 or DrNOD2 (∆LRR) was co-injected ( Figure 4A,B). The CARD domain deletion mutant DrNOD2 (∆CARD) could not inhibit the DrRIG-I (CARD)-initiated Mx activation ( Figure 4B). Highly efficient knockdown of DrNOD2 by MO ( Figure 4C) resulted in the robust activation and expression of Mx, and simultaneous injection of DrNOD2 mRNA decreased the level of activation ( Figure 4D,E). Furthermore, low-molecular weight (LMW) poly I:C-initiated RIG-I signaling was suppressed when DrNOD2 or DrNOD2 (∆LRR) was co-injected in zebrafish embryos ( Figure 4F). These results clearly demonstrated the negative mutual regulation between DrNOD2 and DrRIG-I signaling pathways. No apparent developmental defects were observed (data not shown).    Luciferase assay results showed robust Mx activation. However, this activation was inhibited when DrNOD2 or DrNOD2 (∆LRR) was co-administered with LMW poly I:C. All luciferase assays and qRT-PCR were conducted with three replicates, and each replicate contained extracts from the cells of a well from the six-well plate or 80-100 zebrafish embryos. Values are expressed as mean ± SD; * p < 0.05, ** p < 0.01.

Discussion
RIG-I and NOD2 are two of the most important cytosolic PRRs participating in the recognition of viral and bacterial invasion in mammals. In the present study, we used zebrafish as an attractive model organism to investigate the NOD2-and RIG-I-mediated immunology. The zebrafish was selected because of its conserved structural and functional characteristics compared with its mammalian counterparts.
We used the zebrafish model to demonstrate that RIG-I signaling activation could induce DrNOD2 expression and that NOD2 signaling activation could induce DrRIG-I expression. These observations suggest a feedback regulatory mechanism between the two signaling pathways. Further evaluation showed that the overexpression of DrRIG-I or DrRIG-I-CARD in the HEK293T cells and zebrafish embryos could significantly inhibit DrNOD2 and DrNOD2 (∆LRR)-activated NF-κB signaling. Similarly, administration of DrNOD2 or DrNOD2 (∆LRR) in cells and embryos could significantly suppress DrRIG-I (CARD)-induced IFN signaling. Further knockdown and rescue experiments also confirmed this proposition. Therefore, negative feedback and mutual regulation exist between DrNOD2 and DrRIG-I signaling pathways. Moreover, the results showed the physiological significance of this mutual feedback regulation. DrRIG-I could negatively regulate MDP-initiated NOD2 signaling, and DrNOD2 could negatively regulate LMW poly I:C initiated RIG-I signaling. NOD2 was recently confirmed to interact with RIG-I in a zebrafish cell line (ZF4) by co-immunoprecipitation analysis [38], and we established that this interaction relies on the CARD domain in HEK293T cells ( Figure S1). Moreover, the interaction of mammalian NOD2 with RIG-I and mutual regulation have been observed in human cell lines [39]. These descriptions supported our present observations. We believed that the mutual regulation between DrNOD2 and DrRIG-I signaling pathways may have resulted from the sequestering of the CARDs of DrRIG-I and DrNOD2 away from their adaptors, MAVS and RIPK2, respectively.
Therefore, DrNOD2 and DrRIG-I could be induced by their corresponding signaling pathway and negative regulate the signaling. We believe that this feedback might be of great biological significance in innate antibacterial or antiviral immunity. For example, when the host NOD2/RIG-I signaling is activated via bacterial/viral infection, a feedback regulation might contribute to maintain the homeostasis of immune responses. This characteristic prevents excessive immune reactions. Meanwhile, the induced NOD2 or RIG-I would alert the host, leading to a much faster and stronger immune response against a secondary bacterial or viral invasion. Thus, dysfunction in this feedback regulation may result in various pathological disorders. Secondary bacterial infection commonly develops after viral infections, and this process is accompanied by bursts of inflammatory responses [19]. In addition, susceptibility to bacterial super infection is usually increased as a consequence of innate antiviral responses [40]. We infer that these observations might be associated, at least partially, with the cross-regulation between NOD2 and RIG-I signaling pathways. However, the exact underlying mechanisms of these processes require further clarification. A comprehensive understanding of the cross-regulation between antibacterial and antiviral signaling pathways has long been a challenging task because of the limited availability of research models. Our findings on the mutual regulation between NOD2 and RIG-I signaling pathways in zebrafish may provide a basis for elucidating the cross-regulatory mechanisms of different innate immune signaling pathways. Finally, our work may also be beneficial for the study on the coordination among innate immune systems from the evolutionary perspective.

Experimental Fish
Wild-type AB zebrafish (Danio rerio) weighing 0.5-1 g and measuring 1-2 cm in length were purchased from the National Zebrafish Resources of China. The fish were kept in tanks with recirculating water at 28 • C and fed daily with commercial pellets at 0.7% of their body weight. The fish were acclimatized and evaluated for overall fish health at least two weeks before the experiments.

Bioinformation Analyses
The DrNOD2 sequence was retrieved from the National Center for Biotechnology Information [41], and the DrRIG-I sequence was obtained from a previous study [37]. Functional domains and motifs in proteins were analyzed using SMART and Pfam databases [42,43]. Tertiary structures were determined through PyMOL [44].

Plasmid Constructions
The open-reading frame (ORF) of DrNOD2 was inserted into pCMV-HA (Beyotime, Shanghai, China) between the EcoRI and XhoI sites to construct the eukaryotic expression vector (pCMV-DrNOD2). The LRR motif-deleted mutant (∆LRR) construct (pCMV-DrNOD2 (∆LRR)) and the CARD motif-deleted mutant (∆CARD) construct (pCMV-DrNOD2 (∆CARD)) were cloned and inserted into the pCMV-HA between the sites of EcoRI and XhoI. The CARD motif-deleted mutant (∆CARD) construct of DrRIG-I was cloned and inserted into the pcDNA6 between the sites of BamHI and EcoRI. The NF-κB luciferase vector was purchased from Clontech (Palo Alto, CA, USA), while the pRL-TK vector was obtained from Promega (Madison, WI, USA). Human ISG15 and zebrafish Mx promoters (hISG15-pro-luc and DrMx-pro-luc), as well as the RIG-I expression plasmids pcDNA6-DrRIG-I and pcDNA6-DrRIG-I (CARD), were constructed in our previous study [37,45]. All primers used in plasmid construction are shown in Table S1.

Morpholino Oligonucleotide (MO) and Capped mRNA
The translation-blocking MOs of DrRIG-I and splice junction MO of DrNOD2 were designed, synthesized by Gene Tools (Philomath, OR, USA), and dissolved in water (2 mM). The MO sequences used were as follows: DrRIG-I MO, 5 -GATTCTCCTTCTCCAGCTCGTACAT-3 ; and DrNOD2 MO, 5 -ACCTGCCAAAAATCCAACATGGTTA-3 . The 5 UTR sequence (complement to the MO sequence) of DrRIG was amplified with DrRIG-I F1 and R1 primers (Table S1), and then cloned into the EGFP-N1 vector to evaluate the translation blocking efficiency of DrRIG-I MO. Along with DrRIG-I MO or standard control MO (4 ng/embryo), the constructed vector was injected into one-cell stage embryos (100 pg/embryo). The embryos were collected at 24 hpi, and the GFP fluorescence was visualized through an Olympus MVX10 MacroView. The splice inhibition efficiency of DrNOD2 MO was examined through RT-PCR. The one-cell stage embryos of zebrafish were injected with DrNOD2 MO (4 or 6 ng/embryo) or standard control MO (6 ng/embryo). The embryos were collected at 24 hpi for RNA isolation and cDNA reverse transcription. A forward primer in exon 1 and reverse primer in exon 2 were used to detect the deletion of exon 2 (Table S1). Capped DrRIG-I and DrNOD2 mRNA was synthesized in vitro using a Message Machine kit (Ambion, Thermo Fisher Scientific, Waltham, MA, USA) according to the manual and then solubilized in DEPC water.

Luciferase Assay
HEK293T cells or zebrafish embryos were transfected (1 µg/mL) or injected (100 pg/embryo) with relative stimulant plasmids and NF-κB/ISG15/Mx luciferase reporter vectors. The pRL-TK renilla luciferase reporter plasmid was used as internal control. An empty control plasmid was then added to ensure same amounts of the total DNA. Subsequently, the cells and embryos were lysed at 24 hpi, and dual-luciferase reporter assay was performed as described previously [46]. Luciferase activity was normalized to pRL-TK activity and expressed as fold stimulation relative to the control.

Induction Assay
pcDNA6-DrRIG-I (CARD) or pCMV-DrNOD2 or pCMV-DrNOD2 (∆LRR) (100 pg/embryo) was injected into one-cell stage embryos. The mRNA levels of NOD2 or RIG-I relative to β-actin were examined at 6, 12, and 24 hpi using RT-PCR to determine whether RIG-I-initiated signaling induced NOD2 production and vice versa. The empty plasmid injection group was set as control. RT-PCR was conducted with the following parameters: (1) 40 cycles of amplification at 95 • C for 30 s and 60 • C for 20 s; (2) melting curve analysis at 95 • C for 5 s, 65 • C for 15 s, and 95 • C for 15 s; and (3) cooling at 40 • C for 30 s. Relative gene expression was calculated using the 2 −∆∆Ct method with NOD2/RIG-I, which were initially normalized against β-actin. The primers used are shown in Table S1. Each PCR trial was performed in triplicate and repeated at least thrice.

Co-Immunoprecipitation Assay
pcDNA6-DrRIG-I (CARD) (Myc tag) and pCMV-DrNOD2 (HA tag) were transfected into HEK293T cells. At 48h post transfection, cells were lysed with cold lysis buffer (1% Triton X-100, 150 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl (pH 7.4)) containing protease inhibitor mixture (Roche, Basel, Switzerland) for 30 min at 4 • C. lysates were centrifuged for 15 min at 14,000 rpm and the supernatants were incubated with mouse anti-HA Ab (Abcam, Cambridge, MA, USA) at 4 • C overnight and then incubated with protein A-agarose beads (Roche) for 4 h. The obtained samples were subjected to Western blot assays using rabbit anti-c-Myc tag and HRP-conjugated goat anti-rabbit IgG Ab (Abcam) antibodies, and visualized with ECL reagents as described before.

Mutual Regulation between DrNOD2 and DrRIG-I
The mutual regulation between DrNOD2 and DrRIG-I was examined in the HEK293T cells and zebrafish embryos. The role of DrRIG-I in the DrNOD2-activated NF-κB signaling was analyzed by administering pCMV-DrNOD2 or pCMV-DrNOD2 (∆LRR), alone or together, with pcDNA6-DrRIG-I, pcDNA6-DrRIG-I (CARD), or pcDNA6-DrRIG-I (∆CARD) and the NF-κB luciferase reporter plasmid and internal control plasmid pRL-TK to the HEK293T cells or one-cell stage zebrafish embryos. The CARD deletion mutant was used as the negative control. The role of DrNOD2 in DrRIG-I-induced IFN signaling was analyzed by administering pcDNA6-DrRIG-I (CARD), alone or together, with pCMV-DrNOD2, pCMV-DrNOD2 (∆LRR), or pCMV-DrNOD2 (∆CARD). The IFN luciferase reporter plasmid (hISG15-pro-luc in HEK293T and DrMx-pro-luc in embryo) and control plasmid pRL-TK were also administered. The CARD deletion mutant was used as a negative control. The cells/embryos were harvested and lysed at 24 h post transfection/injection for dual-luciferase reporter assay.

Morpholino-Mediated Knockdown and Capped mRNA-Mediated Rescue
Knockdown and rescue experiments were conducted to further confirm the roles of DrRIG-I in DrNOD2-initiated pathways and DrNOD2 in DrRIG-I-initiated pathways. For the former, one-cell stage embryos were injected with pCMV-DrNOD2/pCMV-DrNOD2 (∆LRR) (100 pg/embryo) and DrRIG-I MO (4 ng/embryo), or together with capped DrRIG-I mRNA (100 pg/embryo). For the latter, one-cell stage embryos were injected with pcDNA6-DrRIG-I (CARD) (100 pg/embryo) and DrNOD2 MO (4 ng/embryo) or together with capped DrNOD2 (100 pg/embryo) mRNA. The embryos in each group were harvested at 24 hpi and lysed for dual-luciferase reporter assay. Furthermore, qRT-PCR was conducted to examine the relative expression levels of TNFα (for NF-κB signaling) and Mx (for IFN signaling). We further confirmed the physiological significance of the mutual negative regulatory roles between DrRIG-I and DrNOD2 by evaluating the role of DrRIG-I in MDP-initiated NOD2 signaling and DrNOD2 in LMW poly I:C-initiated RIG-I signaling. For DrRIG-I, one-cell stage embryos were administered with 2 nL (1 µg/µL) of MDP, alone or together, with pcDNA6-DrRIG-I or pcDNA6-DrRIG-I (CARD) (100 pg/embryo), as well as NF-κB luciferase reporter plasmids. The embryos were collected at 24 hpi, and luciferase assays were conducted to examine the NF-κB activation level. For DrNOD2, one-cell stage embryos were administered with 4 nL (1 µg/µL) of LMW poly I:C alone or together with pCMV-DrNOD2/pCMV-DrNOD2 (∆LRR) (100 pg/embryo), as well as Mx luciferase reporter plasmids. The embryos were collected at 24 hpi, and luciferase assays were conducted to examine Mx activation level.

Statistical Analysis
Data from three independent experiments were expressed as mean ± SD. Groups were compared statistically using Student's t-test for paired samples. Statistical significance were considered at * p < 0.05 and ** p < 0.01.