Analysis of Porcine RIG-I Like Receptors Revealed the Positive Regulation of RIG-I and MDA5 by LGP2

The RLRs play critical roles in sensing and fighting viral infections especially RNA virus infections. Despite the extensive studies on RLRs in humans and mice, there is a lack of systemic investigation of livestock animal RLRs. In this study, we characterized the porcine RLR members RIG-I, MDA5 and LGP2. Compared with their human counterparts, porcine RIG-I and MDA5 exhibited similar signaling activity to distinct dsRNA and viruses, via similar and cooperative recognitions. Porcine LGP2, without signaling activity, was found to positively regulate porcine RIG-I and MDA5 in transfected porcine alveolar macrophages (PAMs), gene knockout PAMs and PK-15 cells. Mechanistically, LGP2 interacts with RIG-I and MDA5 upon cell activation, and promotes the binding of dsRNA ligand by MDA5 as well as RIG-I. Accordingly, porcine LGP2 exerted broad antiviral functions. Intriguingly, we found that porcine LGP2 mutants with defects in ATPase and/or dsRNA binding present constitutive activity which are likely through RIG-I and MDA5. Our work provided significant insights into porcine innate immunity, species specificity and immune biology.


INTRODUCTION
The innate immune system, as the first line of host defense, relies on the pattern recognition receptors (PRRs) to sense various danger signals from pathogens via detecting the pathogen associated molecular patterns (PAMPs) (1). Additionally, the host damage associated molecular patterns (DAMPs) are also recognized by PRRs to maintain homeostasis (1,2). PRRs encompass Toll-like receptors (TLRs), RIG-I-like receptors (RLRs), NOD-like receptors (NLRs), C-type lectin like receptors (CLRs) and cytosolic DNA receptors (CDRs). Upon activation by PAMPs and DAMPs, the PRRs trigger intracellular signaling, which induces either downstream gene transcription or protease-dependent cytokine secretion. Thus, the produced anti-viral interferons (IFNs), proinflammatory cytokines and chemokines will directly combat pathogens and shape subsequent adaptive immunity.
RLRs are the family of DExD/H-box helicases and the main cytosolic viral RNA sensors including homologous RIG-I, MDA5 and LGP2 (3). The prototypic RIG-I as well as MDA5 were discovered more than 15 years ago in overexpression studies to induce type I interferons (IFNs) (4,5). Both RIG-I and MDA5 have similar domain architecture: N-terminal signal active two tandem caspase recruitment domain (CARDs), middle DExD/Hbox helicase domain and C-terminal repressor domain (CTD) or regulatory domain (RD) (6,7). The effective RNA recognized by RIG-I is the blunt ended 5'-ppp dsRNA (double stranded RNA) with the 5′-terminal nucleotide 2'-O-unmethylated, whereas MDA5 prefers internal binding to the long dsRNA with no end specificity (8,9). Upon recognition of dsRNA, RIG-I and MDA5 activate downstream common adaptor MAVS, which recruit TRAF3/TBK1/IKKe and TRAF6/IKKs to activate transcription factors IRF3 and NF-kB, driving IFN and proinflammatory cytokine expression, respectively (10).
The third member of RLRs is LGP2, with similar helicase and CTD domains as RIG-I and MDA5, but due to lack of N terminal CARDs, has no signaling activity. Instead, it is able to regulate RIG-I and MDA5 signaling likely via its strong capability of binding dsRNA (11). It was initially shown to inhibit RIG-I and MDA5 mediated signaling through multiple possible mechanisms (12)(13)(14). Later on, using the LGP2 gene KO mice, the LGP2 was suggested to positively regulate RIG-I and MDA5 signaling (15,16). Further studies tried to reconcile the controversial regulatory roles of LGP2 by proposing that LGP2 is a concentration dependent biphasic switch: at low concentration when cells are in steady state, LGP2 enhances MDA5 activation to help fight infection, whereas at high concentration in late infection stages, LGP2 exerts a general inhibition to RIG-I and MDA5 signaling to resolve the inflammation (17,18). Nevertheless, the exact role of LGP2 in the regulation of RIG-I/MDA5 and in antiviral signaling still remains an open question (8,19).
RLRs are expressed almost in all mammalian cell types, and play key roles in virus recognition and immune responses (6,20). RIG-I and MDA5 play independent and synergistic roles in sensing different RNA viruses (7,21). Consistent with the short RNA agonists, RIG-I mainly senses most negative-sense singlestranded RNA viruses which generate short 5'-ppp dsRNA during replication. These viruses include Filoviridae (Ebola virus), Paramyxoviridae (Measle virus, Sendai virus and Newcastle disease virus), Pneumoviridae (Respiratory syncytial virus), Orthomyxoviridae (Influenza virus), Bunyaviridae (Hantavirus), Rhabdoviridae (Vesicular stomatitis virus and Rabies virus). RIG-I also senses positive-sense single-stranded RNA viruses such as Flaviviridae (Hepatitis C virus and Japanese encephalitis virus). In contrast, MDA5 recognizes long dsRNA and thus senses the positive-sense single-stranded RNA viruses such as the Picornaviridae (Encephalomyocarditis virus, Poliovirus, Coxasackie virus, and Arteriviruses as well as Hepatitis D virus). On the other hand, both RIG-I and MDA5 cross-detect the same viruses including double-stranded RNA Reoviridae (Rotavirus), positive-sense single-stranded Flaviviridae (Dengue virus and West Nile virus), Coronaviridae (Murine hepatitis virus).
In addition to RNA viruses mentioned above, RLRs are also implicated in other pathogen recognitions. For example, RIG-I is able to sense some DNA viruses such as Adenovirus, Poxvirus and Herpesvirus because these viruses produce short dsRNA through the type III RNA polymerase during replication. Other reports also showed that RIG-I can detect the RNA from bacteria such as Listeria monocytogenes, Helicobacter pylori and Shigella flexneri (22). However, till now, there is a lacking of systemic investigation on RLRs of animal source including Sus scrofa (pig). Porcine (pig) is an important livestock species but also a promising animal model to study several human diseases including infectious diseases (23,24). Accumulating evidences showed the essential roles of RLRs in the antiviral function against a number of porcine viruses such as porcine reproductive and respiratory syndrome virus (PRRSV) (25), porcine epidemic diarrhea virus (PEDV) (26), classical swine fever virus (CSFV) (27), porcine circovirus type 2 (PCV2) (28), porcine deltacoronavirus (PDCoV) (29), seneca valley virus (SVV) (30) etc. Further, the interrelationship between RLR members is still not completely resolved. In this study, we characterized the porcine RIG-I, MDA5 and LGP2, and analyzed their mutual relationship. We found that porcine RIG-I and MDA5 exhibit similar and cooperative functions in recognition of RNA and sensing viruses; porcine LGP2 positively regulates RIG-I and MDA5 by promoting the RNA recognition; and interestingly, the ATPase sites in porcine LGP2 play unique roles in RIG-I and MDA5 signaling.

Molecular Cloning and Gene Mutations
Total RNA was extracted from PAMs or PK15 using TRIpure Reagent. From the total RNA, the porcine open reading frames (ORFs) of RIG-I, MDA5 and LGP2 were amplified by RT-PCR using the designed primers shown in Supplementary Table 1. The PCR products were digested with SalI/SmaI, SalI/SmaI and SalI/ EcoRV, respectively, and cloned into the SalI/EcoRV sites of the Gateway entry vector pENTR4-2HA or pENTR4-3FLAG which carries 2HA/3FLAG tag to express C-terminal HA or FLAG tagged proteins. The sequence confirmed HA-tagged and FLAG-tagged pRIG-I, pMDA5 and pLGP2 were all transferred from pENTR4 vectors to Destination vectors pDEST47 (Addgene) by LR recombination to obtain the final pcDNA recombinant expression vectors. Expression plasmids pcDNA human RIG-I-FLAG, pcDNA human MDA5-FLAG and pEGFP-C1-pMAVS were previously cloned and stored in our laboratory. If not specifically mentioned, the LGP2, RIG-I and MDA5 refer to those of porcine origin.
For LGP2 subcloning and mutations, the helicase ATP binding domain (L1), Dhelicase ATP binding domain (DL1), helicase CTER domain (L2), CTR domain (L3) and DCTR domain (DL3) fragments were amplified by PCR using PfuUltra II Fusion HS DNA Polymerase from pcDNA pLGP2-HA plasmid template using the designed primer pairs shown in Supplementary Table 1. For LGP2 deletion mutant Dhelicase CTER (DL2), both fragments flanking the deletion site were amplified by PCR from LGP2 plasmid respectively; next, the two flanking fragments together with the Bridge fragment (Fusion primer) as the templates were joined together by fusion PCR using the designed primer pair L1F/L3R shown in Supplementary Table 1. The LGP2 L1, DL1, L2, DL2, L3 and DL3 PCR products digested with NheI and EcoRV were cloned into the same sites of pcDNA vector expressing C-terminal 2HA as we described previously (33). The mutants of LGP2 M1 K30A, MIIa K138E/Y142F, MIII T167A/S169A, and K654E were produced by mutation PCR using pcDNA pLGP2-HA as a template and the Fast Mutagenesis System Kit. The mutation PCR primer pairs were shown in Supplementary Table 1.

CRISPR gRNA Mediated Stable KO Cells and Homozygous KO PAM Cell Clones
The CRISPR gRNAs targeting porcine RIG-I, MDA5 and LGP2 were designed using the web tool from Benchling (www. benchling.com). For porcine RIG-I, MDA5 and LGP2, four, three and three gRNAs were chosen, respectively, based on the predicted high scores. The encoding DNA sequences are shown in Supplementary Table 2. The annealed gRNA encoding DNA pairs were ligated with BsmBI digested lentiCRISPRv2 vector (Addgene), and the efficiency and specificity of these gRNA expressing lentiviral vectors targeting porcine RIG-I, MDA5 and LGP2 were demonstrated as shown in Supplementary Figure 2. The effective gRNAs are gRNAs 2 and 4 for pRIG-I, the gRNAs 1, 2 and 3 for pMDA5, and gRNAs 1, 2 and 3 for pLGP2. The gRNA expressing lentiviruses were generated by co-transfecting each effective lentiCRISPRv2-gRNAs with package plasmids psPAX2 and pMD2.G into 293T cells using Lipofectamine 2000. For RIG-I, MDA5 or LGP2, the supernatants containing different gRNA expressing lentiviruses were mixed equally and used to infect the PAMs or PK-15 cells, then the infected cells were selected with 1 mg/ml puromycin. After puromycin selection, the CRISPR vector control, RIG-I KO, MDA5 KO and pLGP2 KO stable PAMs and PK-15 cells were obtained and ready for subsequent experiments.
The RIG-I KO, MDA5 KO and LGP2 KO stable PAMs were subcloned by limited dilution and monoclonal cells were chosen. The individual grown cell clones were screened by RT-PCR using the designed primers shown in Supplementary Table 2: First, the PCR products were run by non-denatured polyacrylamide gel electrophoresis and stained with silver dye. Second, the cell clones with desired silver staining would be examined further by cloning the genomic PCR products into T vector using pClone007 Versatile Simple Vector Kit (TsingKe Biological Technology, Beijing, China) and inserted fragments were multiply sequenced and analyzed for base insertion and deletion (indel) mutations. After screening of 49 RIG-I, 10 MDA5 and 8 LGP2 PAM cell clones, we obtained RIG-I -/-, MDA5 -/and LGP2 -/-PAM homozygous KO cell clones.

Co-Immunoprecipitation and Western Blotting
For Co-immunoprecipitation, cells in 6-well plate (8×10 5 cells/ well) were transfected with different combinations of plasmids for 24 h, and then subjected to different treatments to activated receptor signaling. Cells were harvested and lysed in 500 ml RIPA buffer (50 mM Tris pH 7.2, 150 mM NaCl, 1% sodium deoxycholate, 1% Triton X-100) containing protease inhibitors on ice for 30 min. The cell lysates were centrifuged at high speed of 10,000g to pellet cellular debris and the cleared supernatants were transferred to new tubes. The 25-50 ml (5-10%) cleared lysates was kept as input controls and the remained were incubated with 1 mg specific antibody at 4°C overnight with shaking, further incubated with Protein A/G PLUS-Agarose (sc-2003, Santa Cruz Biotechnology) for 2-3h. Later, the agarose was thoroughly washed with RIPA and eluted with 40 ml 2×SDS sample buffer. The elution and input were both subjected to following Western-blot analysis. Protein samples containing 1×SDS sample buffer were boiled at 100°C for 5-10 min and resolved on 6-10% SDS-polyacrylamide gels. The protein bands on gels were transferred onto PVDF membranes and the membranes blocked with 5% nonfat dry milk Tris-buffered saline, pH 7.4, with 0.1% Tween-20 (TBST). Next PVDF membranes were sequentially incubated with various primary antibodies and HRP-conjugated goat anti-mouse or anti-rabbit IgG. The protein signals were detected with enhanced chemiluminescence (ECL) substrate (Tanon, China) and visualized by Western blot imaging system (Tanon, China).

Porcine RLR Protein Purification and dsRNA Binding Assay
The HA-tagged pLGP2, pRIG-I and FLAG-tagged pLGP2, pMDA5 were transferred from pENTR4 vectors to Destination vector pDEST527 (Addgene) by LR recombination to obtain the prokaryotic expression plasmids. Then the protein expression in bacteria DE3/BL21 was induced by IPTG. The proteins in the supernatant of bacterial lysates were purified according to the instructions of Protino ® Ni-TED 2000 Packed Columns Kit, while the proteins in precipitation of bacterial lysates were purified according to the instructions of Micro Protein PAGE Recovery Kit. Specifically, the proteins in supernatant were loaded onto the Protino Ni-TED Packed Column for binding, then washed by LEW buffer and eluted by elution buffer according to the instructions of Protino Ni-TED 2000 Packed Columns Kit. The proteins in precipitation were separated by SDS-PAGE gel, and the target protein were cut off and put into a clean tube, then grinded into small pieces. Next, appropriate amount of solution A was added and the tube was shaked overnight at room temperature. Next day, the centrifugal supernatant was taken and the precooled solution B was added to be mixed at 4°C for 30 minutes. The supernatant was discarded after high-speed centrifugation, and the precipitated protein was dissolved with proper amount of PBS according to the instructions of Micro Protein PAGE Recovery Kit.
For dsRNA binding assay, 10 mg purified protein was mixed with 1mg dsRNA analogue poly (I:C) HMW Biotin in 500 mL PBS, incubated at 25°C for 1h with slow rotating. Then 30 mL resuspended streptavidin agarose was added and incubated at 4°C on a rotating device overnight. Next day, the protein-dsRNAagarose was washed 3 times by centrifugation, and bound proteins eluted with 100 ml 2×SDS sample buffer by heating at 100°C for 10 min. Finally, the eluted proteins were analyzed by Western blotting.

Confocal Fluorescence Microscopy
PAMs grown on 15 mm glass bottom cell culture dish (NEST, China) were transfected with pLGP2-HA and pRIG-I-FLAG, pLGP2-HA and pMDA5-FLAG, pRIG-I-HA and pMDA5-FLAG plasmids, respectively, using TransIT-LT1 Transfection Reagent. Twenty-four hours later, the transfected cells were infected or not with EMCV (0.01 MOI). Then the cells were fixed with 4% paraformaldehyde at RT for 30 min, and permeabilized with 0.5% Triton X-100 for 20 min. After washing with PBS, the cells were sequentially incubated with primary anti-HA rabbit pAb

Promoter Driven Luciferase Reporter Gene Assays
293T cells grown in 96-well plates (3×10 4 cells/well) were cotransfected by Lipofectamine 2000 with ISRE-firefly luciferase reporter or ELAM (NF-kB)-firefly luciferase (Fluc) reporter (10 ng/well) and b-actin Renilla luciferase (Rluc) reporter (0.2 ng/well), together with the indicated porcine RLR expressing plasmids or vector control (10-20 ng/well). The total DNA per well was normalized to 50 ng by adding empty pcDNA vector. PAMs grown in 96-well plates were similarly transfected using the TransIT-LT1 Transfection Reagent. About 24 hours posttransfection, the cells were subjected for stimulation or not, and then cells were harvested, lysed, and both Fluc and Rluc activities were sequentially measured using the TransDetect Double-Luciferase Reporter Assay Kit. The results were expressed as fold inductions of ISRE or ELAM (NF-kB)-Fluc compared with that of vector control after Fluc normalization by corresponding Rluc.

Virus TCID50 and Plaque Assays
Vero and Marc-145 cells in 96-well plates (5×10 4 cells) were grown at 90% confluency, then infected with 0.1 mL cell supernatants from PEDV infected IPEC-J2 and PRRSV infected Marc-145 cells, respectively. The cell supernatants were tenfold serially diluted in 2% FBS DMEM medium before infections of Vero and Marc-145 cells, and each dilution had four to six replicates. The cytopathic effects (CPEs) were observed daily for 4-6 days, and TCID50 was calculated using the Reed-Muench method.
Vero and Marc-145 cells in 24-well plates (3×10 5 cells) or 12well plates (8×10 5 cells) were grown into monolayer, then infected for 2 h with the tenfold serially diluted cell supernatants (0.2 mL for 24-well plate and 0.4 mL for 12-well plate) from virus infected cells. The infected cells were washed and overlaid with an appropriate volume of immobilizing medium of 1:1 mixture of warmed 2× DMEM with 4% FBS and a stock solution of heated 1.6% low melting agarose. Plague formation could take 1-6 days depending on the viruses being analyzed. The immobilizing medium were discarded by tipping and cells were fixed and stained with crystal violet cell colony staining solution (0.05% w/v crystal violet, 1% formaldehyde, 1×PBS and 1% methanol) for 1h at room temperature. After staining, cells were washed with tap water until the clear plaques appeared. The plaques were counted and photos were taken.

Statistical Analysis
All the experiments are representative of three similar experiments and the representative experimental data in graphs were shown as the mean ± SD of triplicate wells. The statistical analysis was performed with t-test or one-way ANOVA where appropriate which are built within the software GraphPad Prism 6.0 (San Diego, USA).

Characterization of Porcine RIG-I, MDA5 and LGP2 Signaling Activity
We isolated and cloned porcine RLRs RIG-I, MDA5 and LGP2, which were deposited into GenBank with the accession numbers MF358966, MF358967 and MF358968, respectively. We confirmed that these porcine RLRs are all IFN stimulated genes (ISG), upregulated by IFN and virus stimulation (Supplementary Figure 1A). The desired HA or FLAG tagged protein expressions of these porcine RLR members were verified by Western blotting (Figures 1A, C). In terms of signaling activity, we found that both porcine RIG-I and MDA5 exhibited constitutive signaling activity in ISRE and NF-kB promoter assay, with MDA5 harboring higher constitutive activity (Figures 1B, D). Upon stimulation by the transfection of double stranded RNA (dsRNA) analogue polyinosinic: polycytidylic acid (poly I:C-LMW), the activities of RIG-I and MDA5 were further heightened ( Figures 1B, D). In contrast, LGP2 did not show any signaling activity with or without stimulation in ISRE and NF-kB promoter assay (Figures 1B, D).
Next, we sought to examine the ligand recognition of dsRNA and viruses by porcine RIG-I and MDA5. In this respect, we chose two types of synthetic analog of viral dsRNA: LMW poly I:C and HMW poly I:C with the length of 0.1-0.6 kb and 1.5-10+ kb, respectively (Supplementary Figure 1B), as well as two viruses vesicular stomatitis virus (VSV) and encephalomyocarditis virus (EMCV). In general, short dsRNA (< 300 bp) was preferentially recognized by human/mice RIG-I whereas long dsRNA (> 1kb) was mainly recognized by human/ mice MDA5 (34)(35)(36). Although LMW poly I:C is not as a commonly used agonist for RIG-I as the 5'-ppp dsRNA, it was confirmed as an effective RIG-I stimulator in several studies including our previous study (37), whereas HMW poly I:C is well known as a agonist for MDA5. In terms of virus recognitions, VSV and EMCV were recognized by RIG-I and MDA5, respectively (34,(37)(38)(39). We first examined the recognition pattern of these poly I:C and viruses by human RIG-I and MDA5 in transfected cells, the results showed as previously reported (Supplementary Figure 1C). In contrast, as shown in Figure 2, the poly I:C (LMW or HMW) activated the porcine RIG-I and MDA5 mediated downstream gene transcriptions in a similar trend (Figure 2A and Supplementary Figure 1D), and as did the two viral stimulations ( Figure 2B and Supplementary Figure 1D). These results might implicate that porcine RIG-I and MDA5 do not bear the feature of preferential recognition of dsRNA and viruses, instead, both have similar and redundant ligand recognition patterns.

The Porcine LGP2 Positively Regulates RIG-I and MDA5 Signaling Activity
The relationship of LGP2 with RIG-I and MDA5 has been elusive (8,9). Therefore, we embarked to explore the regulation of porcine RIG-I and MDA5 by LGP2. First, PAMs were transfected with different concentrations of porcine LGP2, and stimulated with LMW poly I:C and VSV, the downstream gene transcriptions were detected (Figures 3A, B). In both cases, the positive regulation of porcine RIG-I and MDA5 mediated downstream gene transcriptions by LGP2 was observed in a dose dependent manner.
Second, to further examine the regulation of porcine LGP2 in RIG-I/MDA5 mediated signaling, the CRISPR/Cas9 method was used to knockout endogenous porcine LGP2, RIG-I and MDA5, respectively. The gRNAs designed for LGP2, RIG-I and MDA5 were tested for efficacy and specificity ( Supplementary Figures  2A-C) and the effective lentiviral gRNAs were used for lentivirus package and preparation of the stable KO PAMs and PK15 cells. As shown in Figure 4, the LMW poly I:C and VSV induced downstream IFNb, ISG60 and IL8 gene transcription in LGP2 KO PAMs were all decreased relative to CRISPR control PAMs ( Figures 4A, B) and similar results were obtained for IFNb and ISG56 gene transcription in stable LGP2 KO PK15 cells ( Figures  4C, D). Also, EMCV induced ISG56 and IL8 gene transcriptions were decreased in LGP2 KO PK15 cells ( Figure 4E). There results clearly suggested that porcine LGP2 positively regulates the collective signaling activity of RIG-I and MDA5. Additionally, in RIG-I KO and MDA5 KO cells, the LMW poly I:C, VSV and EMCV induced downstream gene transcriptions were all decreased between both types of KO cells. Together with the results showing that ectopic RIG-I and MDA5 respond similarly to distinct RNAs or viruses, our results also suggested that porcine RIG-I and MDA5 play a similar and cooperative role in recognition of different RNAs and viruses.
To ensure the roles of endogenous porcine LGP2, RIG-I and MDA5 in the signaling activity, we further generated the porcine LGP2, RIG-I and MDA5 homozygous KO PAM cell clones ( Supplementary Figures 2D-F). Upon stimulations of either HMW or LMW poly I:C, the IFNb, ISG56, IL1b and IL8 gene transcriptions were all impaired in LGP2 -/-, RIG -/and MDA5 -/-PAMs relative to CRISPR control PAMs ( Figure 5A). The EMCV activated TBK1 and IRF3 phosphorylation and downstream ISG56 production were also decreased in LGP2 -/-, RIG -/and MDA5 -/-PAMs relative to CRISPR control PAMs ( Figure 5B). Next we stimulated the PAM cell clones with a panel of viruses including VSV, Sendai virus (SeV), EMCV, swIAV H9N2 and Herpes simplex virus (HSV-1). All these virus infection induced gene transcriptions of IFNb, ISG56, TNFa, IL1b and IL8 were reduced in in LGP2 -/-, RIG -/and MDA5 -/-PAMs compared with control PAMs ( Figure 5C and Supplementary Figure 3A). Simultaneously, the virus replications of VSV, SeV, EMCV, H9N2 and HSV-1 were significantly upregulated in LGP2 -/-, RIG -/and MDA5 -/-PAMs ( Figure 5D and Supplementary Figure 3B). Generally, the reductions of kinase phosphorylations, downstream gene transcriptions and protein production, and the upregulations of virus replications were relatively comparable in three types of KO PAM cell clones, which again implicates a cooperative relationship between three porcine RLR members. To dissect the regulatory role of LGP2 in RIG-I or MDA5 signaling alone, we used the MDA5 -/-PAM (with RIG-I) and RIG -/-PAM (with MDA5) transfected with different amounts of LGP2, and found that in both cases, LGP2 enhanced the VSV activated  infection in WB and RT-qPCR, respectively ( Figures 6A, B).
LGP2 in transfected IPEC-J2 cells inhibited PEDV M gene transcription in RT-qPCR and lowered the PEDV titers in TICD50 and plaque assays ( Figure 6C).
LGP2 in Marc-145 cells inhibited both PRRSV N protein expression ( Figure 6D) and gene transcription (left, Figure 6E) at 36 h post infection. In addition, LGP2 reduced the PRRSV titers in TCID50 and plaque assays (middle and right, Figure 6E).

Porcine LGP2 Interacts With RIG-I and MDA5 Upon Activation and Promotes Their RNA Binding
We speculated that porcine LGP2 may positively regulate RIG-I and MDA5 via several ways: interacting directly with RIG-I and MDA5 to promote their activation; promoting the ligand RNA binding by RIG-I and MDA5; helping the interactions between RIG/MDA5 and the downstream common adaptor MAVS. The interactions between porcine LGP2, RIG-I and MDA5 in transfected cells were first detected by Coimmunoprecipitation. Without cell activation, no any interaction between three porcine RLRs could be observed (upper panels, Figures 7A-C and Supplementary Figure 4). With EMCV infection, the interactions of LGP2 and RIG-I, as well as LGP2 and MDA5 appeared, but no interaction between RIG-I and MDA5 occurred which are consistent with previously observed (13) (lower panels, Figures 7A-C). Accordingly, the cytosolic co-localizations of LGP2 and RIG-I, as well as LGP2 and MDA5 were obvious in transfected PAMs and 293T cells after cell stimulation by EMCV and HMW poly I:C (Figures 7D-F and Supplementary Figure 5). The interactions between the three porcine RLRs and the adaptor MAVS in the presence or absence of EMCV infection were also examined, but significant interaction could not be detected (Supplementary Figure 4). We further examined the RNA binding of porcine LGP2, RIG-I and MDA5 using the HMW poly I:C pull-down assay. The RNA pull-down assay was performed with purified porcine LGP2, RIG-I and MDA5 proteins ( Figure 7G). HMW poly I:C could successfully pull down LGP2 and MDA5 (Lanes 1, 3, Figure 7H), which is consistent with previous knowledge that both LGP2 and MDA5 bind similarly the blunt-ended dsRNA such as poly I:C (11,40). However, HMW poly I:C failed to pull down porcine RIG-I (Lanes 2, Figure 7H) even though previous studies showed that RIG-I could bind to RNA with higher affinity than MDA5 (17,41). We reasoned that HMW poly I:C, due to its long length, is not the optimal ligand for RIG-I. In the presence of porcine LGP2, the RNA binding of MDA5 was definitely enhanced (Line 5 versus line 3, Figure 7H). Similarly, in the presence of porcine LGP2, the RNA bound RIG-I appeared (Line 4 versus line 2, Figure 7H), suggesting an enhancement of the RNA binding of RIG-I by LGP2. In addition, in the presence of porcine RIG-I, the RNA bound MDA5 was subjected to a slight decrease (Line 6 versus line 3, Figure 7H).

Porcine LGP2 ATPase Sites Are Not Required for Its Positive Regulation of RIG-I and MDA5
LGP2 comprises three domains: N-terminal helicase ATP binding domain (ATP bind), helicase C-terminal domain (CTER) and C-terminal regulatory domain (CTR) with the conserved helicase motifs I, II and III in the helicase ATP binding domain. These motifs I, II, III are implicated in ATPase hydrolysis activity of LGP2, whereas the K654 site in CTR is involved in dsRNA binding (42,43). To dissect the roles of individual domains of porcine LGP2 in the regulation of RIG-I and MDA5 signaling, the each LGP2 domains and domain deletion mutants were produced ( Figure 8A). To analyze the roles of ATPase hydrolysis and RNA binding of porcine LGP2 in the regulation, the helicase motif mutants MI (K30A), MIIa (K138E/Y142F), MIII (T167A/S169A) and RNA binding mutant K654E were prepared ( Figure 8A). The correct expressions of these LGP2 mutant proteins were confirmed by Westernblotting ( Figure 8B). Surprisingly, when these porcine LGP mutants were transfected into PAMs, the downstream IFNb, ISG56, ISG60, IL1b and TNFa transcriptions were significantly activated ( Figure 8C). All the LGP2 mutants also exhibited ISRE activity in promoter assay but no NF-kB activity appeared ( Figure 8D). Furthermore, after these transfected PAMs were infected with VSV, corresponding to the inductions of downstream IFNb, ISGs and proinflammatory cytokines, most of the LGP2 mutants showed anti-VSV activity, which was even greater than wild type LGP2 ( Figure 8E).
Since LGP2 itself is signaling inactive to IFN, we wondered if the constitutive activity of LGP2 mutants was coming from its regulating receptors RIG-I and MDA5. Therefore, the LGP2 mutants were transfected into RIG-I -/and MDA5 -/-PAMs, and downstream gene transcriptions were analyzed. As shown in Figure 9A, relative to control PAMs, the constitutive signaling activity to induce downstream IFNb gene transcription by LGP2 mutants in RIG -/and MDA5 -/-PAMs were both dampened. Nevertheless, the constitutive activity of LGP2 mutants MI, MIII and K654E in RIG -/-PAMs and LGP2 L2 domain in MDA5 -/-PAMs still remained, suggesting certain level of redundancy between porcine RIG-I and MDA5. To examine the antiviral activity of porcine LGP2 and its mutants, LGP2 and the constitutively active MIII mutant (T167A/S169A) were expressed in LGP2 -/-PAMs, and the anti-VSV, anti-SeV and anti-HSV-1 activity measured ( Figure 9B). As expected, VSV, SeV and HSV-1 replications in LGP2 -/-PAM were upregulated, whereas the reconstitution of LGP2 expression suppressed all the viral replications, and reconstitution with active LGP2 MIII mutant further reduced the viral replications ( Figure 9B). Accordingly, the upregulations of downstream gene transcriptions of IFNb and ISG56 by active LGP2 MIII mutant were observed in the absence as well as in the presence of virus infections ( Figure 9C).

DISCUSSION
Human/mice RLRs RIG-I and MDA5 utilize distinct and similar activation mechanism. Under steady state, RIG-I CARDs is bound and auto-inhibited by the Hel-2i region of the helicase. Upon binding of RIG-I C-terminal RD to RNA ends, the conformation of RIG-I changes so that the RNA further binds to Hel-2i, and thus the auto-inhibition of CARDs is lifted (20). Further, through K63ubiquitination/polyubiquitin binding of the CARDs and/or the filament formation by RIG-I RD-helicase along the dsRNA chain, the CARDs are tetramerized into signaling effective stable lockwasher structure (7,44). In the absence of RNA, MDA5 adopts an open and flexible structure, and thus ectopic MDA5 exhibits higher constitutive activity, which was also observed for porcine MDA5 (Figures 1B, D). Nevertheless, MDA5 has very similar activation mechanism where the RD-helicase binds with dsRNA and conformational change exposes the N terminal CARDs, which form tetramer structure. Because of the long dsRNA recognized by MDA5, the CARDs tetramer is stabilized mainly through the filament formation along the dsRNA by RD-helicase, and thus less dependent on K63-polyubiquitin (7,44). In both cases, the effective CARDs tetramer will serve as a platform and nucleate downstream adaptor MASV on the mitochondrial and peroxisomal membranes, which aggregates into prion-like signaling complex (10,44). Despite that previous studies demonstrated a differential recognition of dsRNAs and viruses by human/mice RIG-I and MDA5 (7,21,(34)(35)(36), we observed the similar recognitions of different dsRNAs and viruses and activations of downstream gene transcriptions by porcine RIG-I and MDA5. First, in transfected cells, porcine RIG-I and MDA5 responded similarly to two different poly I:C or two different RNA viruses (Figure 2 and Supplementary Figure 1D). In parallel, we confirmed LMW poly I:C and VSV are preferred by human RIG-I, whereas HMW poly I:C and EMCV are favored by human MDA5 in transfected cells (Supplementary Figure 1C). Second, in porcine RIG KO and MDA5 KO stable cells and homozygous PAM cell clones, the downstream signaling, gene transcriptions and protein expression are all reduced in a similar way in response to various dsRNAs and viruses (Figures 4, 5). Third, in the antiviral assay, porcine RIG-I and MDA5 exhibited comparable strengths against Influenza virus and PRRSV, two porcine RNA viruses ( Figure 6). Not only presenting similarities, porcine RIG-I and MDA5 also showed cooperation in the recognition of RNAs and viruses since in both RIG KO and MDA5 KO cells, there were significant reductions of activity upon stimulations (Figures 4, 5). Therefore, it appears that certain differences between porcine RIG-I and MDA5 and their human counterparts can be discerned. Importantly, we found porcine LGP2 positively regulates porcine MDA5 as well as RIG-I activity. Positive regulation of MDA5 activation by physiological low concentration of LGP2 was widely accepted (15,45,46) and the biochemical and molecular mechanism recently became clear (17,18,43,47): MDA5 binds dsRNA weakly and the filament formation on long dsRNA chain is slow and unstable. On the contrary, LGP2 binds dsRNA with the highest affinity among three RLRs, regardless of 5'-PPP and RNA length, but not forming filaments (18,40,48). By utilizing its ATP hydrolysis activity, LGP2 is able to not only increase initial binding of dsRNA by MDA5, but also stabilize the shorter MDA5 filaments on dsRNA which are signaling active, thus promoting MDA5 activity. Consistent with these findings and conclusions, we showed that porcine LGP2 positively regulates MDA5 (Figures 3-5), interacts with MDA5 and enhances its dsRNA binding (Figure 7). On the other hand, the positive regulation of RIG-I by LGP2 has not been well recognized so far with contradictory conclusions (8,9,15,46,49,50). In our conditions using transfected cells and gene KO cells (Figures 3-5), we concluded that porcine LGP2 positively regulates RIG-I as well. Further, porcine LGP2 interaction with RIG-I and the enhanced porcine RIG-I binding to dsRNA in the presence of LGP2 provide a direct evidence for positive regulation (Figure 7). Additionally, we showed that porcine LGP2 serves as an broad antiviral protein against various porcine viruses including swIAV H9N2 and PR8 H1N1 which should be sensed mainly by porcine RIG-I (51) (Figure 6 and not shown). Actually, there have been a number of studies which demonstrated the broad spectrum of anti-RNA virus functions by LGP2 from different species (52)(53)(54)(55)(56)(57). Several RNA helicases including DDX3, DHX29, DHX36 and DDX60 are required for both RIG-I and MDA5 mediated IFN production after RNA virus infection (58). Likely and similarly, porcine LGP2 is involved in both RIG-I and MDA5 activation and downstream IFN and other gene transcriptions. In addition to its anti-RNA virus activity, LGP2 is also implicated in anti-DNA virus and anti-intracellular bacterial replications (59). All RLR members are ATPase proteins and ATP hydrolysis plays regulatory roles in RLR activation. For RIG-I, ATP hydrolysis was reported to drive RIG-I translocation along 5'-PPP dsRNA chain, forming signaling active filaments (60,61). For MDA5, ATP hydrolysis was shown to actually dissociate MDA5 from long RNA chain (62,63), probably contributing to shorter and signaling effective filaments. For LGP2, ATP hydrolysis enables it to efficiently engage diverse dsRNA species, which promotes MDA5 recognition of dsRNA, shorter filament formation and activation (47). In this case, both ATP hydrolysis and RNA binding of LGP2 are required for positive regulation of MDA5 (18,43). However, the roles of ATP hydrolysis in RLR activation and antiviral signaling are very complex. RIG-I, MDA5 and LGP2 all contain conserved helicase motifs I-VI in helicase domain, with all essential for ATP hydrolysis activity and some for RNA binding (42). Among the mutants, RIG-I MIII and MDA5 MI, MIII exhibited constitutive activity, suggesting that the ATP hydrolysis of RIG-I and MDA5 was not required for their activation and signaling (42). In our study, we initially sought to explore the role of ATP hydrolysis activity in the LGP2 function as a regulator, but we observed idiosyncratic phenotypes of porcine LGP2: The ATPase inactive mutants MI, MIIa and MIII exhibited obvious constitutive activity in transfected PAMs (Figures 8, 9). More surprisingly, each individual domains (L1, L2, L3) and DL3 mutant also presented constitutive activity, whereas all the mutants including RNA binding defective mutant K654E were able to induce proinflammatory IL1b transcription ( Figure 8C). These constitutive activities of porcine LGP2 mutants came from RIG-I and MDA5 since the activities of most mutants disappeared in RIG-I -/or MDA5 -/-PAMs ( Figure 9A), implying that complexity of porcine LGP2 regulation of RIG-I and MDA5.
Although we do not know the underlying mechanism regarding unique regulation of RIG-I and MDA5 by porcine LGP2 which shares 82% AA identity to human LGP2, it is warranted for further investigation. Collectively, in this study, we characterized signaling functions of porcine innate immune RLRs RIG-I, MDA5 and LGP2, and investigated the regulation of RIG-I and MDA5 by porcine LGP2 and the underlying mechanisms ( Figure 10). We found that porcine RIG-I and MDA5 have similar and cooperative signaling relationship, and porcine LGP2 positively regulates RIG-I and MDA5 signaling via the enhancement of RNA binding by the two receptors ( Figure 10). The ATPase activity and RNA binding of porcine LGP2 are not required for its positive regulation of RIG-I and MDA5 activity; instead porcine LGP2 exhibits complex regulation opening the doors for further study.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors. LGP2, binding with dsRNA in a similar way to MDA5, is able to promote the dsRNA binding by both MDA5 and RIG-I, and thus downstream signaling. Intriguingly, during the positive regulation by LGP2, the ATPase hydrolysis and dsRNA binding activity seem dispensable. The solid lines refer to confirmed interactions in this study, whereas the dashed lines are based on previous knowledge.