Temporal regulation of MDA5 inactivation by Caspase-3 dependent cleavage of 14-3-3η

The kinetics of type I interferon (IFN) induction versus the virus replication compete, and the result of the competition determines the outcome of the infection. Chaperone proteins that involved in promoting the activation kinetics of PRRs rapidly trigger antiviral innate immunity. We have previously shown that prior to the interaction with MAVS to induce type I IFN, 14-3-3η facilitates the oligomerization and intracellular redistribution of activated MDA5. Here we report that the cleavage of 14-3-3η upon MDA5 activation, and we identified Caspase-3 activated by MDA5-dependent signaling was essential to produce sub-14-3-3η lacking the C-terminal helix (αI) and tail. The cleaved form of 14-3-3η (sub-14-3-3η) could strongly interact with MDA5 but could not support MDA5-dependent type I IFN induction, indicating the opposite functions between the full-length 14-3-3η and sub-14-3-3η. During human coronavirus or enterovirus infections, the accumulation of sub-14-3-3η was observed along with the activation of Caspase-3, suggesting that RNA viruses may antagonize 14-3-3η by promoting the formation of sub-14-3-3η to impair antiviral innate immunity. In conclusion, sub-14-3-3η, which could not promote MDA5 activation, may serve as a negative feedback to return to homeostasis to prevent excessive type I IFN production and unnecessary inflammation.


Introduction
An early onset of type I interferon (IFN) induction and response are the keys for successful viral clearance.During viral infection, several cytosolic sensors could detect viral nucleic acids as the pathogen-associated molecular patterns (PAMPs) [1,2].The RIG-I-like receptors (RLRs), including retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated gene 5 (MDA5), are key cytosolic sensors to recognize PAMP RNA species [1,2].Once bound to non-self RNA, the RLRs undergo major conformational changes to interact with accessory proteins and downstream signaling molecules [1,2].These protein complexes could then engage host protective innate immunity against viral infection via mitochondrial antiviral signaling protein (MAVS), a mitochondria-associated transmembrane protein, through their Caspase activation and recruitment domains (CARDs) [3,4].We have previously showed that upon activation, RIG-I and MDA5 relocalize to mitochondria-associated membrane (MAM) to interact with MAVS, which is the crucial onset of MAVS-mediated IFN production pathway during acute phase of viral infection [3,[5][6][7].
MDA5-mediated antiviral signaling has been shown important in the clearance of Flavivirus, Picornavirus, Paramyxovirus, and Reovirus infections [8].In the laboratories, enterovirus infections, human coronavirus 229E infections and/or high molecular weight poly (I:C) (synthetic dsRNA) transfection are the most common agents to model MDA5 activation.It has been shown that MDA5 has an essential and nonredundant role in detecting RNA virus infection [9].The binding of MDA5 to the long dsRNA will then cooperatively form tandem MDA5 filaments along the dsRNA, and the CARDs of MDA5 will then oligomerize to interact with MAVS for downstream antiviral signaling pathway [10].These reports indicated that similar regulatory event of MDA5 redistribution from the cytosol to the MAM compartment promoted by 14-3-3η to interact with MAVS is critical for MDA5-mediated type I IFN induction.Through interaction with MAVS, MDA5 activation will then lead to the activation of PAMPdriven transcription factors, IFN production, and interferon-stimulated gene (ISG) expressions, resulting in the immediate onset of host antiviral state [11].These studies indicated that not only the PRRs, signaling molecules, and transcription factors are critical, but also the chaperone proteins which facilitate the intracellular redistribution and protein-protein interactions among the signaling molecules and therefore promote the kinetics of type I IFN induction are important to restrict viral propagation and infection.
Antiviral innate immunity constitutes pro-inflammatory response to initiate adaptive immune response required for virus clearance.Yet, excessive inflammatory responses are highly destructive and could lead to tissue damage.Hence, the antiviral innate immunity requires a rapid, however, tightly regulated response to effectively clear the invading pathogens without harming the host.Several negative regulators of the type I IFN induction pathway have been reported recently, such as apoptotic Caspases and plasminogen activator inhibitor type 2 (PAI-2) (reviewed in [12]).Several studies discovered that the expression of PAI-2 was up-regulated when macrophages were stimulated by LPS [13,14], and this phenotype was dependent on IKKβ-NF-κB signaling pathway [14].Recently, a novel role of PAI-2 in the component of antiviral immunity.During RNA virus infections, sub-14-3-3η are accumulated in both Caspase-3 dependent and independent manners.This discovery sheds light on how viruses manipulate host proteins to evade immune surveillance, potentially offering new targets for therapeutic intervention.

PLOS PATHOGENS
Temporal regulation of MDA5 inactivation by Caspase-3 dependent cleavage of 14-3-3η macrophages has been reported to suppress IL1β processing via inhibiting Caspase-1 activation in macrophages, suggesting that intracellular proteases may be involved in promoting inflammation and innate immunity [15].Mitochondrial apoptotic pathway is known to be induced after RLR stimulations [16].When apoptosis is triggered, the 35-kDa pro-Caspase-3 is cleaved into two active subunits of which molecular weight are 17 kDa and 12 kDa [17].Previous studies had shown that apoptotic Caspases not only mediate proteolysis during apoptosis but also cleave the components of type I IFN induction pathway.It has been well-characterized that activated Caspase 3/7 could cleave several signaling molecules, including cGAS, MAVS, STING and IRF3, to downregulate type I IFN expression [18][19][20].Indeed, the activation of Caspase-3 and other apoptotic Caspases could not only be triggered by proapoptotic signals but also be promoted by the RLR-MAVS-mediated antiviral signaling [16,21], suggesting a role of Caspase-3 as a regulator in the negative feedback loop of type I IFN induction.Pathogenic viruses, such as enteroviruses, utilize this intrinsic negative regulatory mechanism to impair type I IFN induction by promoting the activation of Caspase-3 [22].Not only during enterovirus infections, the ectopic expression of enterovirus proteases 2A and 3C may also achieve the activation of Caspase-3 and thus downregulate the antiviral innate immunity [22,23].
We previously identified that the redistribution of RIG-I from the cytosol to a membrane fraction upon ligand recognition was controlled by 14-3-3ε, and 14-3-3η can facilitate the activation of MDA5 [5,7].There are 7 isoforms in the 14-3-3 family, including 14-3-3β, 14-3-3γ, 14-3-3ε, 14-3-3η, 14-3-3σ, 14-3-3θ, and 14-3-3z.Although the cargo preferences vary among different isoforms, the tertiary structures of the 14-3-3 isoforms are well-conserved [24].According to the structural studies of 14-3-3 proteins, a single 14-3-3 molecule is composed of nine α-helices, indicated as αA to αI [24].The flexible loop structure between αC-αD and αH-αI are the CD loop and HI loop, respectively [24,25].When 14-3-3 protein dimer is formed, the first four α-helices (αA to αD) are pivotal for dimerization, and the αC, αE, αG, and αI contribute to construct the target protein binding groove [26].The ligand binding groove is a hydrophobic and positively charged patch composed by the arginine-arginine-tyrosine triad [26,27].It was proposed that the C-terminal tail of 14-3-3, which varies in peptide lengths and sequences among all isoforms, could be the regulatory domain for the target protein binding affinity [28,29].Because the C-terminal truncation may either enhance or impair the binding affinity of 14-3-3 ligands among different 14-3-3 isoforms, the C-terminal region as well as the αI play a critical role to regulate 14-3-3 interaction to the target proteins.Many viral proteins are also reported to interact with the 14-3-3 family.Studies have showed that both Dengue virus (DENV) and Zika virus (ZIKV) NS3 proteins use a protease-independent phosphomimetic-based mechanism to occupy 14-3-3ε and/or 14-3-3η to retain RIG-I and MDA5 in cytosol and lead to the attenuation of IFNβ induction in the infected cells [5,30], suggesting that 14-3-3 family serves as an important regulator in the type I IFN expression.It is intriguing whether other viruses may also target 14-3-3ε and/or 14-3-3η to downregulate type I IFN induction pathway and antiviral innate immunity.
In this study, we reported that MDA5 activation-dependent Caspase-3 activity cleaved 14-3-3η at the C-terminus to truncate out the αI helix of 14-3-3η.The cleaved 14-3-3η, termed as sub-14-3-3η, could strongly interact with MDA5 but could not promote MDA5-MAVSdependent type I IFN induction.Therefore, the cleavage of 14-3-3η may serve as a negative feedback mechanism to prevent excessive type I IFN induction by MDA5-dependent signaling, and pathogenic viruses, such as human coronavirus 229E, have taken the advantage of this intrinsic pathway to downregulate MDA5-dependent antiviral signaling.Besides the antiviral innate immunity, 14-3-3η and MDA5 are both found to be associated with Rheumatic arthritis, and the serum levels of 14-3-3η can serve as the biomarker of this autoimmune disease [31].Understandings in the protein interaction at the molecular levels may benefit us not only in the development of antiviral drugs but also in the development of immune modulatory therapies.
We next assessed whether endogenous sub-14-3-3η was produced when cells were infected with coronavirus and/or enterovirus or were transfected with high molecular weight (HMW) poly (I:C), which were all reported to activate MDA5-dependent antiviral signaling (Fig 1, A, B  and C).Total cell lysates from Huh7 infected with human coronavirus 229E (hCoV-229E) at 0.1 MOI were harvested from 0 to 42 hours post infection (h.p.i) and immunoblotted with anti-N and anti-14-3-3η antibodies to determine whether sub-14-3-3η was accumulated during hCoV-229E infection (Fig 1A).We observed the distinct sub-14-3-3η at 18 h.p.i accompanying with the reduced abundance of full-length 14-3-3η from 24 to 42 h.p.i.along with the accumulation of hCoV-229E N protein and cleaved PARP, one of the main cleavage targets of Caspase-3 in vivo (Fig 1A).Similarly, in enterovirus 71 (EV71)-infected RD cells, we could observe the accumulation of sub-14-3-3η along with the increasing expression of EV71 viral protein and cleaved PARP determined by immunoblotting (Fig 1B).In a time-course experiment of HMW poly (I:C) transfection in Huh7 cells, the inductions of MDA5 and IFIT3 protein levels over time were expected, as both genes were reported as interferon-stimulated genes (ISGs) (Fig 1C).We also observed the degradation of endogenous 14-3-3η upon the activation of Caspase-3 through time, which was indicated by the increase in the cleaved Caspase-3 abundance (Fig 1C).
We next assessed the MAM redistribution of MDA5 upon HMW poly (I:C) transfection, which is a critical step in MDA5-MAVS-mediated type I IFN induction.Huh7 cell lysates were fractionated into cytosol and crude mitochondria-MAM (mito-MAM) fractions, and antitubulin and anti-VDAC1 immunoblotting served as markers of cytosol and crude mitochondria-MAM fractions, respectively.In the vector transfected control cells, MDA5 was almost undetectable in the unstimulated cells, and under the stimulation of poly (I:C) transfection, MDA5 was primarily in the cytosol fraction and partially redistributed to the mito-MAM fraction (Fig 3C).The ratio of MDA5 in mito-MAM fraction to that in cytosol fraction was increased upon the ectopic expression of Myc-14-3-3η (Fig 3D ), which is consistent with our previous reports [5].However, in Huh7 cells expressing Myc-14-3-3ηΔαI, MDA5 redistribution to the mito-MAM fraction during poly (I:C) stimulation was less to that of the vector control cells, even that Myc-14-3-3ηΔαI itself was found in both cytosol and mito-MAM fractions (Fig 3C and 3D).Redistribution of MDA5 in response to poly(I:C) transfection was also assessed in 14-3-3η KD Huh7 cells reconstituted with Myc-tagged 14-3-3η or 14-3-3ηΔαI (S3B Fig) .When transfected with poly (I:C), 14-3-3η KD Huh7 cells were unable to support MDA5 redistribution to the mito-MAM fraction, and only the reconstitution of Myc-14-3-3η but not Myc-14-3-3ηΔαI to 14-3-3η KD Huh7 cells could rescue the mito-MAM redistribution of MDA5 in poly(I:C) transfected cells (S3B Fig) .Unlike full-length Myc-14-3-3η, Myc-14-3-3ηΔαI was unable to promote MDA5 redistribution (Figs 3C-3D and S3B), which may be the reason why ectopic expression of Myc-14-3-3ηΔαI impaired MDA5-dependent IFNβ promoter activity (Fig 2C and 2D).

Discussion
For a long time, MDA5 has been known as an important PRR which raises the alarm and triggers the type I IFN induction pathway against the invading RNA viruses [35,36].Comparing to the auto-suppressive mechanism of RIG-I, which is another member of the RLR family, how MDA5 activities are controlled still requires further investigation.Recently, it was demonstrated that 14-3-3η formed complex with MDA5 and played a key regulatory role in MDA5 activation which enhanced the MDA5 oligomerization and redistribution [5].While investigating the role of 14-3-3η in MDA5 activation, we noticed that in addition to the full-length 14-3-3η, a 14-3-3η sub-isoform was able to interact with MDA5 even with high affinity (Figs 2B and S2B).Accordingly, we proposed that to prevent the overactivation of MDA5, a negative feedback was achieved by 14-3-3η sub-isoform to temporally regulate the MDA5-mediated type I IFN induction pathway.These results suggested that once type I IFN signaling was activated to a certain degree, the apoptosis related Caspases would be triggered to clean away the infected and/or damaged cells as well as to prevent the constitutive activation of MDA5 and over-inflammation.The apoptotic Caspases, especially Caspase-3, would mediate the cleavage of 14-3-3η to produce the sub-14-3-3η.Although the truncated form of 14-3-3η lacking the αI helix did not show dominant-negative effect toward MDA5 activation (Figs 2 and 4), due to the strong interaction between sub-14-3-3η and MDA5, we proposed that sub-14-3-3η would occupy the MDA5 and prevent MDA5 interaction to the full-length 14-3-3η and therefore reduce the mitochondrial redistribution of MDA5 to interact with MAVS.This hypothesis was partly supported by our results (S3A Fig) that the addition of sub-14-3-3η may interrupt the interaction between MDA5 and full-length 14-3-3η.The other possibility is that sub-14-3-3η may strongly bind to N-MDA5 and thus prevent MAVS from interacting with the CARDs of MDA5.Via producing sub-14-3-3η to inhibit the MDA5 activities, cells can return to homeostasis when the infection is resolved.This regulation is particularly important in MDA5-dependent signaling, as MDA5 can be activated when the intracellular abundance of MDA5 is high [5].Further studies have indicated that cellular RNAs play a significant role as PAMPs in activating the RLR-MAVS signaling pathway.Although the specific host ligands for MDA5 remain less understood, MDA5 can become active when mitochondrial RNA degradation is impaired, allowing mitochondrial double-stranded RNA to escape into the cytoplasm [37].Additionally, the expression of endogenous retroviral elements can trigger the MDA5-MAVS signaling axis to express type I IFNs [38,39].Our findings of Caspase-3 dependent sub-14-3-3η accumulations observed during acute RNA viral infections may also contribute to the control of MDA5 deactivation under sterile inflammation.
The emerging roles of chaperone proteins in antiviral innate immunity have been revealed in the past decade.In previous reports, members of 14-3-3 protein family were shown to be important for cells to maintain normal functions by manipulating the activities of various signaling pathways, and in the RLR-related type I IFN induction pathway, MDA5 was not the only RLR that is regulated by 14-3-3 chaperone protein [7,40].14-3-3ε was showed to promote the RIG-I/TRIM25/14-3-3ε translocon formation and therefore transduced the active Z-DEVD-FMK 50 μM for 1 hour and subsequently infected with hCoV-229E at 0.1 MOI for 0 to 36 hours.Immunoblotting was performed to detect the viral infection and other protein levels.(F) Wildtype and CASP3 KD Huh7 cells were infected with EV71 at 5 MOI for 0 to 24 hours, and immunoblotting was performed to determine the viral infection and other protein levels. https://doi.org/10.1371/journal.ppat.1012287.g004

PLOS PATHOGENS
Temporal regulation of MDA5 inactivation by Caspase-3 dependent cleavage of 14-3-3η signaling to MAVS [7].RIG-I and MDA5 are both critical to respond to RNA virus infections, and their activities require to be well controlled to prevent overactive inflammation.In this study, we proposed the negative feedback in the regulated relation between MDA5 and 14-3-3η.As RIG-I and MDA5 both contain the Caspase activation and recruitment domains (CARDs) at their N-terminus, it is intriguing to determine if there is a similar mechanism between RIG-I and 14-3-3ε, especially when 14-3-3ε is already showed to be substrate of Caspase-3 [28].In previous reports, 14-3-3ε had decreased binding affinity to its target protein, Bad, after Caspase-3 cleaved 14-3-3ε at Asp238 and removed partial C-tail [28].While preparing this report, a study revealed that EV71 3C protease cleaved 14-3-3ε at a site close to a known Caspase-3 cleavage site at the C-tail, and the cleavage impaired the ability of 14-3-3ε to interact with RIG-I [41].It is fascinating to know whether it is a common regulatory method for cells to mediate cleavage on other 14-3-3 proteins by Caspase-3 after infection for limiting the overactivation of RLR-signaling.
Nevertheless, previous reports have shown that to prevent the induction of type I IFN, DENV NS3 and ZIKV NS3 can compete 14-3-3ε with RIG-I and/or 14-3-3η with MDA5 [30,42].It is intriguing whether certain viral infections may promote the formation sub-14-3-3η to antagonize MDA5 signaling.For instance, it has been reported that enterovirus infection induced apoptotic Caspases, including Caspase-3, to cleave MDA5 at the C-terminus and subsequently dampened the induction of type I IFN [43].It was not fully understood how the cleaved MDA5 during enterovirus infection, which remained the intact N-terminus CARDs, would lose its activity to promote IFNβ induction.Our report provides another aspect to explain the phenotype observed in previous study, which is that Caspase-3 activation during EV71 infection targets not only MDA5 but also 14-3-3η to antagonize type I IFN induction.Our results suggested that Caspases other than Caspase-3 may also be involved in the formation of sub-14-3-3η, as in CASP3 KD Huh7 cells, sub-14-3-3η was still observed during EV71 infection (Fig 4F).In fact, expressions of many viral proteins without the full-course of infection were also able to trigger apoptotic Caspases activation [22].Caspases involved in apoptosis, notably Caspase-3 and Caspase-7, regulate the production of excessive interferons by cleaving several essential proteins, such as cGAS in cytosolic DNA sensing and MAVS in cytosolic RNA sensing [20].Moreover, IRF3, a shared downstream element in both pathways, can be targeted for cleavage by both Caspase-3 and Caspase-8 [44,45].Additionally, Caspase-8 modulates the activation of IFN pathways by directly interacting with components of the RIG-I signaling complex, thereby preventing their excessive activation [46].It may require further in vitro cleavage analysis of 14-3-3η with a panel of purified Caspases in the presence or absence of MDA5 and its RNA ligands to elucidate the role of Caspase 3 directly.We also believe that there are other viral proteins that may potentially induce sub-14-3-3η accumulation to control antiviral innate immunity.In fact, the peptide sequences of 14-3-3η across different species are highly conserved (S4E Fig) , and therefore targeting 14-3-3η as an antagonistic mechanism to downregulate antiviral innate immunity might also be evolutionarily conserved.In summary, we proposed a model to describe sub-14-3-3η cleaved by Caspase-3 as a negative feedback of MDA5-dependent signaling (Fig 5).These findings might provide inspirations to discover more in the field of antiviral innate immunity.

Viruses infection, high molecular weight poly(I:C) stimulation and inhibitor treatment
Huh7 cells were infected with human coronavirus-229E (hCoV-229E) at a 0.01, 0.1 or 1 MOI in serum-free medium and incubated at 37˚C for one hour.After that, the cells were washed by PBS and then incubated in DMED containing 2% FBS.To infect Huh7 cells with Sendai virus (SeV), cells were exposed to 100 HA unit (HAU) of SeV in DMEM supplemented with 10% FBS at 37˚C.For enterovirus 71 (EV71) infection, RD cells or Huh7 cells were respectively infected with EV71 at 1 or 5 MOI in serum-free medium and incubated at 37˚C for one hour.Cells were then washed with PBS and incubated in DMEM with 2% FBS.In addition, high molecular weight poly(I:C) (Invitrogen) were transfected into Huh7 cells by using TransIT- When MDA5-mediated type I IFN induction is activated, it will lead to the activation of Caspase-3.Then, 14-3-3η is cleaved by active Caspase-3, which results in the formation of sub-14-3-3η.Sub-14-3-3η competes with full-length 14-3-3η to interact with MDA5.Subsequently, MDA5 bound with sub-14-3-3 losses IFNβ-inducing function.Therefore, by forming sub-14-3-3η to block MDA5 signaling as a negative feedback, over-activation of type I IFN induction can be prevented in cells.

Transfection and dual luciferase IFNβ reporter assay
Fugene 6 (Promega) was used to transfect plasmids into Huh7 cells.The detailed protocols were based on the instructions of manufacturer.For dual luciferase IFNβ reporter assay, pIFNβ-Luc, pCMV-Renilla-Luc, and other plasmids were transfected into Huh7 cells for 48 hours.After harvesting and lysing the cells, the luciferase activities of IFNβ promoter and Renilla were monitored via Dual-Luciferase Assay Kit (Promega), according to the protocols in manufacturer's instructions.

Semi-denaturing detergent agarose gel electrophoresis (SDD-AGE)
The experiment procedures were previously described [5].Cell pellets were washed with PBS and then lysed in a cold Triton X-100 lysis buffer (50mM Tris pH = 7.5, 137 mM NaCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1% Triton X-100) supplemented with a protease inhibitor (Roche) for 20 minutes.The cell lysates were centrifuged to remove the cell debris to obtain the protein samples.Each protein sample was incubated in 4X SDD-AGE sample buffer (2X TBE, 4% glycerol, 8% SDS, bromophenol blue) and then separated by using 1.5% agarose gel contained 0.1% SDS in 1X SDD-AGE running buffer (1X TBE and 0.1%SDS) at 4˚C, 80V for 80 minutes.The proteins were then transferred to the NC membrane by capillary action.The NC membrane was blocked in TBST buffer with 5% bovine serum albumin (BSA) overnight.Myc-14-3-3ηΔαI for 24 hours, followed by mock-transfection or 1 μg/mL HMW poly(I:C) transfection for 18 hours.Cell lysates were separated into cytosol and mito-MAM fractions.Immunoblotting was utilized for detecting the redistribution of endogenous MDA5.

Fig 5 .
Fig 5. Illustration of the proposed model during viral infection.Viral infection triggers the activation of MDA5 and RIG-I anti-viral signaling.When MDA5-mediated type I IFN induction is activated, it will lead to the activation of Caspase-3.Then, 14-3-3η is cleaved by active Caspase-3, which results in the formation of sub-14-3-3η.Sub-14-3-3η competes with full-length 14-3-3η to interact with MDA5.Subsequently, MDA5 bound with sub-14-3-3 losses IFNβ-inducing function.Therefore, by forming sub-14-3-3η to block MDA5 signaling as a negative feedback, over-activation of type I IFN induction can be prevented in cells.
protease induced apoptosis promoted the degradation of 14-3-3η.(A) Continued to Fig 4A, immunoblotting was used to confirm the ectopic expression.(B) Continued to Fig 4B, immunoblotting was used to confirm the ectopic expression.(C) Continued to Fig 4C and 4D, immunoblotting was utilized to confirm the ectopic expression in Huh7 cells, 14-3-3η KD Huh7 cells and MDA5 KD Huh7 cells.(D) Constant amount of Myc-14-3-3η and empty vector, EV71 2A-eGFP or EV71 2A H21L-eGFP were co-transfected into Huh7 cells for 72 hours.The protein levels were analyzed by immunoblotting.(E) Whole amino acid residues of 14-3-3η from different species were aligned.The sequences of 14-3-3η across different species were highly conserved.(TIF) S1 Raw dataxlsx.Raw data of experimental values behind the means and standard deviations used to build graphs.(XLSX)