ROS-induced HSP70 promotes cytoplasmic translocation of high-mobility group box 1b and stimulates antiviral autophagy in grass carp kidney cells

Autophagy plays many physiological and pathophysiological roles. However, the roles and the regulatory mechanisms of autophagy in response to viral infections are poorly defined in teleost fish, such as grass carp (Ctenopharyngodon idella), which is one of the most important aquaculture species in China. In this study, we found that both grass carp reovirus (GCRV) infection and hydrogen peroxide (H2O2) treatment induced the accumulation of reactive oxygen species (ROS) in C. idella kidney cells and stimulate autophagy. Suppressing ROS accumulation with N-acetyl-l-cysteine significantly inhibited GCRV-induced autophagy activation and enhanced GCRV replication. Although ROS-induced autophagy, in turn, restricted GCRV replication, further investigation revealed that the multifunctional cellular protein high-mobility group box 1b (HMGB1b) serves as a heat shock protein 70 (HSP70)–dependent, pro-autophagic protein in grass carp. Upon H2O2 treatment, cytoplasmic HSP70 translocated to the nucleus, where it interacted with HMGB1b and promoted cytoplasmic translocation of HMGB1b. Overexpression and siRNA-mediated knockdown assays indicated that HSP70 and HMGB1b synergistically enhance ROS-induced autophagic activation in the cytoplasm. Moreover, HSP70 reinforced an association of HMGB1b with the C. idella ortholog of Beclin 1 (a mammalian ortholog of the autophagy-associated yeast protein ATG6) by directly interacting with C. idella Beclin 1. In summary, this study highlights the antiviral function of ROS-induced autophagy in response to GCRV infection and reveals the positive role of HSP70 in HMGB1b-mediated autophagy initiation in teleost fish.

phagy occurs at a basal level that is important for maintaining normal cellular homeostasis. Upon stress, such as nutrient starvation, hypoxia, oxidative stress, and pathogen infection, autophagy is rapidly initiated (1,3,4). Mature autophagosomes fuse with lysosomes to form single-membraned autolysosomes, where enzymes in lysosomes degrade the contents in the autophagosomes (2,5). Autophagy induces the modification of the soluble form microtubule-associated protein 1 light chain 3 (LC3)-I to the lipid-modified form LC3-II (6). LC3-II conjugates on the membrane of autophagosomes and is commonly used as a marker protein for autophagosomes. Beclin 1, the mammalian ortholog of yeast ATG6, is regarded as an essential component for the initiation of conventional autophagy (6). However, the essential role of fish Beclin 1 in autophagy regulation remains unclear, although orthologs of Beclin 1 have been identified in some fish species (7).
High-mobility group box 1 (HMGB1) protein is conserved chromatin-associated nuclear protein (8). Although pathogenic stimulation or stress treatment induce nuclear HMGB1 to shuttle to the cytoplasm, where HMGB1 is further actively secreted or passively released to the extracellular space (9,10), HMGB1 plays multiple roles in physiological and pathological processes in different cellular fractions (11). Recently, HMGB1 has been reported to play important nuclear, cytosolic, and extracellular roles in the regulation of the autophagy process under oxidative stress, starvation conditions, or chemotherapy treatment (12)(13)(14)(15). Reactive oxygen species (ROS) 2 can trigger HMGB1 translocation to the cytosol where HMGB1 interacts with Beclin 1 and subsequently induces autophagy (12). Heat shock proteins (HSPs) are a family of conserved chaperone proteins that act in response to physiological and environmental This work was supported by the National Natural Science Foundation of China (31572648) and the Huazhong Agricultural University Scientific and Technological Self-innovation Foundation (2014RC019). The authors declare that they have no conflicts of interest with the contents of this article. This article contains Figs. S1-S3 and Tables S1 and S2. 1 To whom correspondence should be addressed. stresses (16). HSP70 is involved in the regulation of fundamental cellular processes, such as signal transduction, cell cycle regulation, apoptosis, and innate immunity (17,18). HSP70 has also been reported to inhibit starvation or rapamycin-induced autophagy via the increased v-akt murine thymoma viral oncogene homolog (AKT) pathway (19,20). Grass carp (Ctenopharyngodon idella) is one of the most important aquaculture species in China. Grass carp reovirus (GCRV), a dsRNA virus, causes severe hemorrhagic disease in juvenile grass carp (21). Previous studies demonstrated pivotal roles of grass carp HMGBs (HMGB1a, HMGB1b, HMGB2a, HMGB2b, HMGB3a, and HMGB3b) in the antiviral immune response against GCRV infection (21)(22)(23). Similar to mammalian HMGBs, grass carp HMGBs localize in the nucleus in the resting stage, whereas GCRV infection or pathogenic stimuli induce HMGB nucleocytoplasmic translocation, and subcellular localization of HMGB1b is more sensitive to pathogenic stimuli compared with other HMGBs (24). Intramolecular interaction between HMG boxes and C-tail domains has been identified to mediate nucleocytoplasmic translocation of HMGBs (24). However, the intermolecular interaction that regulates the subcellular localization and functions of fish HMGB1 remains unclear.
In this study, HSP70 was identified to interact with HMGB1b upon GCRV infection. GCRV infection induced ROS-mediated autophagy, which, in turn, inhibited GCRV replication. ROS treatment induced HSP70 translocation to the nucleus, where it interacted with HMGB1b and promoted HMGB1b cytoplasmic shuttling. In the cytoplasm, HSP70 and HMGB1b synergistically enhanced ROS-induced autophagy. HSP70 also promoted HMGB1b-Beclin 1 association via direct interaction with Beclin 1. Our results highlight the antiviral role of ROS-induced autophagy in response to GCRV infection. Differing from that in mammals, this study uncovers that cytoplasmic HMGB1bmediated autophagic activation depends on HSP70 in teleosts.

HSP70 interacts with HMGB1b upon GCRV infection
To define the molecular action of HMGB1b during GCRV infection, we performed affinity purification for HMGB1b from stable CIK cells and analyzed HMGB1b-binding proteins by LC-MS/MS (Fig. 1A). We identified 705 candidate interactive proteins from the proteomic dataset, of which 31 proteins were matched with at least four unique peptides (Ն4) (Table S1). Besides HMGB1b, HSP70, an ortholog of mammalian HSP72 (Fig. S1), is the top candidate and has the most matched peptides (unique peptides ϭ 13). Thus, we investigated the role of HSP70 in HMGB1b-dependent antiviral activity. The interaction between HMGB1b and HSP70 was verified through co-IP upon GCRV infection (Fig. 1, B and C). A standard plaque assay and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay indicated that HMGB1b overexpression significantly promoted cell viability, whereas HSP70 overexpression had no influence on GCRV-induced cell death (Fig. 1, D  and E). The viral titer indicated that HMGB1b, but not HSP70, possessed significant antiviral activity (Fig. 1F). These results imply that HSP70 may play an indirect role in HMGB1b-mediated antiviral immunity.

HSP70 promotes H 2 O 2 -induced cytoplasmic translocation of HMGB1b
Given the previous report that mammalian HSP70 negatively regulates oxidative stress-induced HMGB1 cytoplasmic translocation (10), we established cell lines that stably co-expressed HMGB1a-GFP/RFP-HSP70, HMGB1a-GFP/RFP-HSC70, HMGB1b-GFP/RFP-HSP70, and HMGB1b-GFP/RFP-HSC70 and examined the influence of HSP70s on the H 2 O 2 -induced dynamic localization of HMGB1s. As shown in Fig. 2, A and B, H 2 O 2 treatment did not induce obvious cytoplasmic translocation of HMGB1a upon RFP-HSP70 or RFP-HSC70 overexpression, but for HMGB1b, significant cytoplasmic shuttling of HMGB1b was evoked upon H 2 O 2 treatment in the presence of RFP-HSP70 (Fig. 2C). However, H 2 O 2 treatment induced no or a few cells that displayed an HMGB1b cytoplasmic presence in HMGB1b-GFP/RFP-HSC70 co-expressing and HMGB1b-GFP single-expressing cells (Fig. 2, D and E). Consistently, cell percentage analysis and Western blot (WB) analysis of subcellular fractions displayed a significant cytoplasmic presence of HMGB1b in HMGB1b-GFP/RFP-HSP70 co-expressing cells (Fig. 2, F and G). All of these results indicate that HSP70 is required for H 2 O 2 -induced HMGB1b cytoplasmic translocation, which is contrary in mammals.

H 2 O 2 induces HSP70 nuclear translocation and interaction with HMGB1b
In mammals, HSP72 translocates to the nucleus in response to oxidative stress (10). To elucidate the effect of oxidative stress on grass carp HSP70 subcellular localization, RFP-HSP70 overexpression cells were treated with or without H 2 O 2 , and the cytoplasmic and nuclear proteins were extracted to determine HSP70 localization. WB analysis indicated that H 2 O 2 treatment induced significant endogenous and exogenous HSP70 expression in the nucleus (Fig. 3, A and B). A co-IP assay demonstrated the interaction between HSP70 and HMGB1b in the nuclear fraction after H 2 O 2 treatment (Fig. 3C). This result agrees with a report that oxidative stress induces HSP72 nuclear translocation and subsequent interaction with HMGB1 in the nucleus (10). The interaction between HSP70 and HMGB1b in the nucleus is required for H 2 O 2 -induced HMGB1 cytoplasmic translocation.

H 2 O 2 treatment evokes autophagy in CIK cells
Autophagy is a self-degradative process that is important for maintaining cellular homeostasis (25). A number of physiologically relevant insults, e.g. ROS and rapamycin, can induce autophagy (4,26). In this study, we employed both H 2 O 2 and rapamycin (an autophagy promotor) treatment to induce autophagy, as evidenced by GFP-LC3-labeled autophagosomes and RFP-LAMP2-labeled lysosomes in CIK cells. Compared with control cells, H 2 O 2 and rapamycin treatment significantly enhanced the accumulation of GFP-LC3 puncta, which are called autophagosomes, in the cytoplasm (Fig. 4, A and C). H 2 O 2 and rapamycin treatment induced strict fusion between GFP-LC3-labeled autophagosomes and RFP-LAMP2-labeled HSP70 promotes HMGB1b-mediated antiviral autophagy lysosomes (Fig. 4, B and C). Autophagy activation is accompanied by the conversion of cytosolic LC3-I to membrane-bound LC3-II (27). To assess the formation of endogenous LC3-II, WB analysis was used to examine the expression levels of LC3-I and LC3-II. The specificity of the anti-LC3 antibody (Ab) was verified ( Fig. 4D). WB analysis showed that H 2 O 2 treatment significantly enhanced LC3-II expression and the LC3-II/LC3-I ratio (Fig. 4, E and F). Transmission electron microscopy (TEM) analysis revealed the ultrastructure of double-membraned autophagosomes and monolayer-membraned autolysosomes upon H 2 O 2 and rapamycin stimulation (Fig. 4G). In the late stage of H 2 O 2 and rapamycin treatment (48 h), the number of autolysosomes was more than that of autophagosomes (count of TEM data). Collectively, these findings demonstrate that H 2 O 2 and rapamycin treatments are sufficient to initiate autophagy activation and subsequent enhancement of auto-phagy flux. This is the first evidence of ROS-induced autophagy in fish cells.

GCRV infection promotes autophagy activation by eliciting ROS accumulation
To evaluate the relationship between ROS and GCRV, we examined ROS production in GCRV-infected CIK cells using the DCFH-DA probe. Compared with normal cells, GCRV infection induced excessive ROS production in ϳ84% of cells, as potent as the H 2 O 2 treatment, which was analyzed by flow cytometry (Fig. 5A) and fluorescence microscopy (Fig. 5B). Excessive ROS generation promotes activation of the antioxidant pathway to protect cells from oxidative stress and damage (28). The NF-E2-related factor 2 antioxidant response element (Nrf2-ARE) pathway is capable of stimulating the activity of antioxidant enzymes (superoxide dismutase (SOD), catalase -GFP stably transfected CIK cells were obtained by G418 selection. IP was carried out using GFP trap beads. LC-MS/MS analysis was performed on a Q Exactive mass spectrometer. B and C, verifying the interaction between HSP70 and HMGB1b via co-IP assay. B, CIK cells were cotransfected with HMGB1b-GFP and HSP70-HA for 24 h. Then the cells were infected with GCRV for another 24 h. Co-IP was performed with anti-HA monoclonal Ab and mouse IgG (control) and immunoblot (IB) analysis with the respective Abs as shown. C, both HMGB1b-GFP and GFP overexpression CIK cell lines were infected with GCRV for 24 h. The cell lysate was subjected to IP by GFP trap beads. Then the IP samples and whole-cell lysate (WCL) were used for IB analysis with anti-HSP70, anti-GFP, and anti-␤-tubulin Abs. D-F, antiviral activity analyses of HMGB1b and HSP70 to GCRV infection. CIK cells seeded in 24-well plates overnight were transiently transfected with 600 ng of HMGB1b-GFP, HSP70-HA, or the corresponding empty plasmids, respectively. Twenty-four hours later, the transfected cells were infected with GCRV at an m.o.i. of 0.1. Forty-eight hours later, the supernatants were collected as titer samples. Virus titers were measured by standard plaque assay in 24-well plates and cell viability assay. D, standard plaque assay. CIK cells were plated in 24-well plates for 12 h and treated with titer samples of HMGB1b-GFP, GFP, HSP70-HA, and empty vector (pcDNA3.1) at different dilution rates (0, 1:500, 1:1000, and 1:2000). Thirty-six hours later, the viable cells were fixed with 10% paraformaldehyde and stained with 0.05% (w/v) crystal violet. E, cell viability was assessed by MTT assay in 96-well plates. F, virus titers supernatants were measured by PFU assay. Error bars indicate standard deviation (n ϭ 4). **, p Ͻ 0.01.

HSP70 promotes HMGB1b-mediated antiviral autophagy
(CAT), GSH synthetase (GSS), and hemeoxygenase-1 (HO-1)), which contribute to resistance to a variety of diseases (28). To this end, we examined the expression levels of Nrf2, SOD1, SOD2, CAT, GSS, and HO-1 at different time points after GCRV infection and H 2 O 2 and rapamycin treatment. GCRV infection induced up-regulation of Nrf2 and SOD1 from 12 h to 72 h, SOD2 from 12 h to 48 h, GSS from 24 h to 72 h, CAT at 12 h and 48 h, and HO-1 at 48 h (Fig. 5C). After H 2 O 2 treatment, Nrf2, SOD1, and GSS were rapidly inhibited at 12 h and then recovered to normal levels and were up-regulated at the following time points. The mRNA levels of SOD2, CAT, and HO-1 were significantly enhanced at 24 h, 48 h, and 72 h (Fig. 5D). Furthermore, rapamycin treatment also remarkably promoted the expression levels of Nrf2, SOD1, SOD2, CAT, GSS, and HO-1 at different time points (Fig. 5E). These results prove that GCRV infection, oxidative stress, and autophagy initiation evoke antioxidant pathway activation.
To explore whether GCRV can initiate autophagy, we evaluated autophagosome formation at 12 h and 24 h after GCRV infection. As shown in Fig. 6, A and B, GCRV infection markedly enhanced GFP-LC3 punctum formation compared with control cells. WB analysis indicated that GCRV infection significantly increased LC3-II expression and the LC3-II/LC3-I ratio from 6 h to 72 h in CIK cells (Fig. 6, C and D). TEM analysis revealed autolysosomes and virions captured in autolysosomes in GCRV-infected CIK cells (Fig. 6E). These data demonstrate that GCRV infection induces autophagy activation. To investigate the role of ROS in GCRV-induced autophagy initiation, a ROS inhibitor, N-acetyl-L-cysteine (NAC) was used to eliminate ROS. First, the efficiency of NAC was examined by flow cytometry. As shown in Fig. 6F, NAC pretreatment significantly inhibited GCRV-induced ROS in CIK cells. Interestingly, NAC pretreatment remarkably decreased LC-II expression and autophagosome formation (Fig. 6, G and H). A plaque assay,

HSP70 promotes HMGB1b-mediated antiviral autophagy
MTT assay, and virus titer revealed that NAC pretreatment promoted GCRV replication and inhibited cell viability (Fig. 6, I-K). Therefore, suppression of ROS results in inhibition of autophagy activation. Collectively, these results reveal that GCRV infection-induced autophagy activation depends on ROS accumulation.

Autophagy inhibits GCRV replication in CIK cells
The antiviral or proviral roles of autophagy have been witnessed in the past few years (29). To determine the precise role of autophagy in regulating GCRV infection, CIK cells were pretreated with the autophagic inhibitor 3-methladenine (3-MA) or the autophagy promotors rapamycin and H 2 O 2 and then infected with an equal amount of GCRV. The supernatant from the infected cells was harvested to determine the viral yield. Compared with the mock-treated group, 3-MA pretreatment increased the expression of GCRV genes (VP4, VP5, VP6, VP7, and NS38) by more than 200-fold, whereas rapamycin and H 2 O 2 pretreatment significantly reduced viral gene expression (Fig. 7A). Interestingly, H 2 O 2 treatment induced more robust inhibition of GCRV gene expression compared with rapamycin treatment. Consistently, the viral yield in 3-MA-pretreated cells was higher than in the mock-treated group, whereas H 2 O 2 and rapamycin pretreatment significantly decreased the viral titer (Fig. 7B). A standard plaque assay and MTT assay indicated that 3-MA pretreatment strongly weakened cell viability. Nevertheless, H 2 O 2 and rapamycin pretreatment significantly promoted cell viability in CIK cells, and H 2 O 2 had a more protective effect on cell viability (Fig. 7, C and D). However, H 2 O 2induced cell survival and inhibition of GCRV replication were eliminated by the presence of 3-MA, which indicated that autophagy is essential for H 2 O 2 -evoked cell protection and GCRV suppression. These results suggest that autophagy, especially ROS-induced autophagy, significantly restricts GCRV replication.

HSP70 promotes HMGB1b-mediated antiviral autophagy
showed a more significant increase in the LC3-II ratio (Fig. 8, G  and H). To uncover the roles of endogenic HMGB1b and HSP70 in autophagy regulation, HMGB1b-and HSP70-specific siRNA was used to knock down endogenic HMGB1b and HSP70. Among the synthetic siRNA, s1 of HMGB1b and s1 of HSP70 showed the best interference efficiencies in HMGB1b and HSP70 mRNA levels (Fig. 8, I and J). Further results indicated that knockdown of HMGB1b or HSP70 significantly decreased LC3-II/LC3-I intensity and autophagosome formation (Fig. 8, K and L). Both HMGB1b and HSP70 knockdown resulted in stronger inhibition of autophagy activation. Interestingly, HSP70 knockdown has no significant influence on GCRV replication. However, double knockdown of HMGB1b and HSP70 induced a notable decrease in GCRV titer (Fig. 8M), which means that HSP70 inhibits HMGB1b-mediated antiviral function to GCRV. Collectively, these results demonstrate HSP70 promotes HMGB1b-mediated antiviral autophagy that both HSP70 and HMGB1b positively participate in autophagy regulation and that HSP70 and HMGB1b synergistically facilitate autophagy activation, which further inhibits GCRV replication.

Beclin 1 functions as a positive regulator of autophagy in CIK cells
Mammalian Beclin 1 is an essential component for the initiation of autophagy (6). Previous studies indicated that HMGB1 and HSP70 are involved in Beclin 1-medialted autophagy (31,32). However, whether piscine Beclin 1 functions in autophagy remains largely unknown. We first identified that Beclin 1 strictly localized in GFP-LC3-labeled autophagosomes (Fig.  S2A). Beclin 1 overexpression increased the percentage of CIK cells containing LC3 puncta. Rapamycin or H 2 O 2 treatment induced more LC3 punctum production in Beclin 1 overexpression cells (Fig. S2B). Consistently, Beclin 1 overexpression enhanced LC3-II accumulation in resting cells (Fig. S2, C and  D). Compared with cells transfected with an empty vector, Beclin 1 overexpression significantly increased H 2 O 2 -and rapamycin-induced LC3-II expression (Fig. S2, E-H). These data suggest that piscine Beclin 1 is an inherent positive regulator of autophagy. To probe the effect of GCRV, H 2 O 2 , and rapamycin on Beclin 1 expression, the protein levels of endogenous Beclin 1 were examined by WB analysis in GCRV-infected, H 2 O 2 -and rapamycin-treated CIK cells. The specificity of the anti-Beclin 1 Ab was identified in GFP-Beclin 1 and Beclin 1-HA overexpression CIK cells (Fig. S2I). H 2 O 2 treatment induced significant up-regulation of Beclin 1 at 24 h and 48 h (Fig. S2, J and K). The protein levels of Beclin 1 were enhanced from 12 h to 48 h after rapamycin treatment (Fig. S2, L and M). After GCRV infection, the protein levels of Beclin 1 were rapidly and remarkably up-regulated from 6 h to 72 h (Fig. S2, N and O). Therefore, Beclin 1 is involved in the GCRV-and H 2 O 2 -induced autophagy pathway.

HSP70 enhances H 2 O 2 -induced HMGB1b-Beclin 1 association
To assess the regulation of HSP70 and HMGB1b in the Beclin 1 autophagy pathway, we investigated the influence of HSP70 and HMGB1b on Beclin 1 expression. As shown in Fig.  9, A and B, HSP70 and HMGB1b overexpression significantly increased the protein levels of Beclin 1. However, knockdown of HSP70 and HMGB1b significantly inhibited the protein level of Beclin 1 (Fig. 9C). Knockdown of both HMGB1b and HSP70 synergistically restrained Beclin 1 expression. Additionally, the HSP70 inhibitor VER-155008 was used to prove HSP70 function. Upon VER-155008 treatment, Beclin 1 expression was sig-nificantly inhibited (Fig. 9D). These results identify the positive roles of HMGB1b and HSP70 in the Beclin 1-mediated autophagy pathway. To explore the regulation mechanism of HSP70 to HMGB1b-Beclin 1 association, the interaction between HMGB1b and Beclin 1 was examined with a co-IP assay. In HMGB1b-GFP stably transfected cells, HMGB1b slightly interacted with Beclin 1. The interaction was significantly enhanced by RFP-HSP70 overexpression and weakened upon VER-155008 treatment (Fig. 9E), indicating that HSP70 plays an essential role in HMGB1b-Beclin 1-mediated autophagy activation. Whether in the resting stage or upon rapamycin or H 2 O 2 treatment, HSP70 and Beclin 1 were clearly colocalized in the cytoplasm (Fig. 9F). The co-IP assay indicated that HSP70 interacted with Beclin 1 under normal conditions and rapamycin or H 2 O 2 treatment, but H 2 O 2 treatment enhanced the interactive intensity between HSP70 and Beclin 1 (Fig. 9G). These results underline that HSP70 -Beclin 1 interaction is essential for ROS-induced HMGB1b-Beclin 1 association.

Discussion
Over the past few years, scientists have attempted to understand the mechanisms of HMGB1 in mediating the cellular and biological responses (33). Some immune receptors, such as receptor for advanced glycation end products (RAGE), Tolllike receptor 2 (TLR2), TLR4, and TLR9, have been indicated to interact with HMGB1 (33,34). This study identified 31 HMGB1b-interacting proteins that were quantified by four or more peptides by LC-MS/MS upon GCRV infection. These proteins were functionally grouped into nucleus proteins, ribosomal proteins, metabolic proteins, growth-associated proteins, and HSP family proteins, which implies multiple functions of HMGB1b in cells. In line with a previous report that H 2 O 2 or lipopolysaccharide (LPS) stimuli induce cytoplasmic HSP72 to translocate to the nucleus and interact with HMGB1 (10, 35), HSP70 is rapidly translocated to the nucleus upon H 2 O 2 treatment in CIK cells, supporting a similar function in mammalian HSP72 and piscine HSP70 in response to oxidative stress. This result may be attributed to the high conservation between fish HSP70 and mammalian HSP72 (Fig. S1). Interestingly, mammalian HSP72 and piscine HSP70 show the opposite effect on the HMGB1 nucleocytoplasmic shuttling. Mammalian HMGB1 functions as a proinflammatory cytokine in response to infection, injury, and stress, which is sufficient to initiate inflammation (10,35,36). However, inflammatory responses are strictly controlled by anti-inflammatory cyto-

HSP70 promotes HMGB1b-mediated antiviral autophagy
kines because excessive production of cytokines can lead to inflammatory diseases. Thus, HSP72, serving as an anti-inflammatory cytokine, inhibits ROS-or LPS-induced HMGB1 cytoplasmic translocation and release, which is essential for maintaining cellular homeostasis. Here, ROS rarely induced nuclear export of HMGB1b but caused significant HMGB1b nucleocytoplasmic translocation upon HSP70 overexpression, which implies that HMGB1b is insufficient to sense oxidative stress. This result is consistent with our previous study that pathogenic stimulus-induced nucleocytoplasmic shuttling of grass carp HMGBs is not as intense as that in mammals (24). Therefore, we proposed that mammalian HMGB1 is more sensitive to stress than piscine HMGB1 and that piscine HSP70 is required for the activation of HMGB1b.
The autophagy lysosomal degradation pathway is one of the essential degradation systems in vertebrates and has been validated to play critical roles in cellular homeostasis, development, disease, and infection (2,37). ROS are inducers of autophagy activation (4). Previous studies have uncovered molecular regulatory mechanisms between ROS and auto-

HSP70 promotes HMGB1b-mediated antiviral autophagy
phagy (4,11,12). Our experimental results first revealed that H 2 O 2 , a source of ROS, is sufficient to initiate autophagy and autophagy lysosomal flux in fish cells, which signifies that fish possess a mechanism to alleviate oxidative stress-induced damage via autophagy. Oxidative stress has been found to occur in various viral infections (38). ROS and antioxidant ingredients are a series of crucial signaling molecules in the oxidative stress response (4). GCRV infection induced the accumulation of ROS, enhancement of antioxidant genes, and autophagy activation in CIK cells, whereas NAC-induced ROS inhibition restricted autophagy activation. Given the vital role of H 2 O 2 in autophagy activation, GCRV-induced accumulation of ROS  1:500, 1:1000, and 1:2000). Thirty-six hours later, viable cells were fixed with 10% paraformaldehyde and stained with 0.05% (w/v) crystal violet. J, cell viability was assessed by MTT assay in 96-well plates. K, virus titer supernatants were measured by PFU assay. Error bars indicate standard deviation (n ϭ 4). **, p Ͻ 0.01; *, 0.01 Ͻ p Ͻ 0.05.

HSP70 promotes HMGB1b-mediated antiviral autophagy
may serve as a signaling molecule to initiate autophagy. Mild oxidative stress has been demonstrated to promote cell survival, whereas severe oxidative stress can cause oxidative injury and even death (38). So far, considerable advances have been made in understanding the complex interplay between autophagy and viruses (29, 39 -42). On one hand, autophagy plays a positive role in the host antiviral response through the degradation of cytoplasmic viral components, activation of innate and adaptive immune signaling, promotion of cell survival, and cooperation between autophagy proteins and other immune pathways to restrict viral infections. On the other hand, viruses employ multiple strategies to avoid or antagonize the antiviral function of autophagy by avoiding autophagic capture and suppressing autophagy initiation and maturation (29). Furthermore, viruses exploit the autophagy pathway as a proviral system to keep virally infected cells alive and prolong viral replication (29). Therefore, it is urgent to understand the direct effects of oxidative stress and autophagy on GCRV infection. Our results demonstrate that rapamycin and nontoxic doses of H 2 O 2 pretreatment promote cell survival and restrict GCRV replication, whereas autophagy the inhibitor 3-MA greatly increase viral productive infection. These data suggest that limited oxidative stress and autophagy are beneficial for the host against virus infection in grass carp, which is in contrast to a previous study in epithelioma papulosum cyprini cells (a cell line of cyprinid fish) that spring viremia of carp virus utilizes the autophagy pathway to facilitate its own genomic RNA replication (43). This distinction may be attributed to species and virus specificity. Therefore, this study opens a door for researchers to develop new antiviral drugs targeting ROS and autophagy regulation. More mechanistic investigation of oxidative stressinduced autophagy is essential for exploring therapeutic strategies and pathogenic insights into GCRV infection.
In mammals, abundant evidence emphasizes the subcellular localization-dependent roles of HMGB1 in autophagy modulation (12,14,44,45). Especially upon ROS and starvation stimuli, HMGB1 translocates to cytoplasm, where it interacts with Beclin 1 and evokes the autophagic function of Beclin 1 by dissociating Bcl-2 from Beclin 1 (12,31). Previously, dual roles of HSP70 in autophagy modulation have been reported: on one hand, HSP70 inhibits starvation-induced autophagy by regulating the mTOR/Akt pathway (19,20). On the other hand, acetylated HSP70 binds to the Beclin 1-Vps34 complex and promotes KRAB-ZFP-associated protein 1 (KAP1)-dependent SUMOylation and activity of Vps34, which are essential for autophagic vesicle formation (30). Overexpression of HSP72 augments LPS-induced autophagy through c-Jun N-terminal kinase phosphorylation and Beclin 1 up-regulation (32). In this study, positive roles of HMGB1b and HSP70 in autophagy activation were verified by enhancement of the LC3-II/LC3-I ratio and Beclin 1 level upon HMGB1b or HSP70 overexpression. HSP70 and HMGB1b also synergistically promoted autophagy

HSP70 promotes HMGB1b-mediated antiviral autophagy
HSP70 promotes HMGB1b-mediated antiviral autophagy activation. In the HMGB1-Beclin 1-mediated autophagy pathway, cysteines Cys-23 and Cys-45 in the box A domain in HMGB1 are required to interact with Beclin 1, and Cys-106 in the box B domain promotes cytosolic localization and sustained autophagy (12). Coincidentally, the three cysteines are conserved in grass carp HMGB1b (Fig. S3), which prompted us to explore whether cytoplasmic HMGB1b can promote autophagy like mammalian HMGB1. Interestingly, in HMGB1b-GFP stably transfected cells, H 2 O 2 treatment induced a weak interaction between HMGB1b and Beclin 1, whereas HMGB1b-Beclin 1 association was enhanced by HSP70 overexpression but prevented by treatment with the HSP70 inhibitor VER-155008. Considering the foundation role of HSP70 in ROS-induced HMGB1b cytoplasmic translocation, this result further confirms the indispensable role of HSP70 in cytoplasmic HMGB1b-regulated autophagy. The direct interaction between HSP70 and Beclin 1 might be responsible for HMGB1b-Beclin 1 association. All of these findings uniformly add to the accumulating evidence of HSP70 in autophagic regulation.
In conclusion, this study demonstrates that GCRV infection and H 2 O 2 treatment induced the accumulation of ROS in CIK cells, which further enhanced autophagy activation. Nevertheless, ROS-induced autophagy, in turn, dramatically restricted GCRV replication. We clarified a novel mechanism for autophagy regulation: ROS induces HSP70 to translocate to the nucleus, where it interacts with HMGB1b, which results in HMGB1b cytoplasmic translocation and subsequent HMGB1b-Beclin 1 autophagic pathway activation. In the cytoplasm, HSP70 directly interacts with Beclin 1, which facilitates the association between HMGB1b and Beclin 1. These data provide credible evidence for the pivotal role of autophagy in antiviral defense and illuminate a new autophagic regulation mechanism in fish, which lays a foundation for further developing a more effective therapeutic strategy in the context of GCRV infection.
GCRV-GZ1208, a type I GCRV strain, was kindly provided by Dr. Qing Wang. For viral infection, CIK cells were plated for 24 h in advance and then infected with GCRV-GZ1208 at a multiplicity of infection (m.o.i.) of 0.1 as described previously (46).

WB and co-IP analyses
Cytosol and nuclear proteins were extracted using cytosol/ nuclear protein isolation kits (Beyotime). WB and co-IP experiments were conducted as in our previous study (46). IP of GFP fusion proteins was performed using GFP trap beads (Chromotek) that consist of a single-domain anti-GFP Ab conjugated to an agarose bead matrix. All results were obtained from three independent experiments.

HSP70 promotes HMGB1b-mediated antiviral autophagy TEM and confocal microscopy
To analyze the ultrastructures of autophagosomes and autolysosomes, CIK cells were treated with or without H 2 O 2 or rapamycin or infected with GCRV in 6-well plates for 24 h, 36 h, and 48 h. Then the cells were fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) for 1 h at room temperature. Ultrathin sections were prepared as described previously (6). Images were viewed on an HT-7700 transmission electron microscope (Hitachi, Japan).
For confocal microscopy, CIK cells were plated onto coverslips in 12-well plates and transfected with the indicated plasmids for 16 h, and the stably transfected cell lines were directly seeded in 12-well plates. Then the cells were treated with rapamycin or H 2 O 2 or infected with GCRV for 24 h and 36 h. Subsequently, the cells were washed, fixed, and stained as reported previously (24). Finally, the samples were observed with a confocal microscope (SP8, Leica). GFP-LC3-labeled puncta were autophagosomes. In this study, cells with at least five GFP-LC3-positive puncta were regarded as autophagy activation (41).

qRT-PCR assay
Total cellular RNA isolation and cDNA synthesis were performed according to a previous report (22). qRT-PCR was established in the Roche LightCycler 480 system, and EF1␣ HMGB1b-GFP and RFP-HSP70 stably cotransfected cells were pretreated with or without VER-155008 for 12 h. HMGB1b-GFP stably transfected cells were used as a control. All cells were treated with 0.15 mM H 2 O 2 for 24 h. IP was conducted with GFP trap beads and IB with Beclin 1 and GFP Abs. WCL was used to analyze the indicated proteins. Right panel, the relative protein levels of Beclin 1, which interacts with HMGB1b, related to the total level of Beclin 1 in WCL, were quantified by ImageJ. Error bars indicate S.D. (n ϭ 3); **, p Ͻ 0.01. F, HSP70 co-locates with Beclin 1. CIK cells were cotransfected with GFP-Beclin 1 and RFP-HSP70. Then the transfected cells were treated with rapamycin (Rap) or H 2 O 2 for 24 h and subjected to confocal microscope observation. G, HSP70 interacts with Beclin 1. CIK cells were co-transfected with GFP-Beclin 1 and HSP70-HA for 16 h and then treated with or without rapamycin or H 2 O 2 for 24 h. Co-IP was performed with anti-HA monoclonal Ab. Mouse IgG was used as a control. WCL was subjected to IB with anti-GFP, anti-HA, and anti-␤-tubulin Abs, respectively.

HSP70 promotes HMGB1b-mediated antiviral autophagy
was employed as an internal control gene for cDNA normalization. qRT-PCR amplification was carried out in a total volume of 15 l containing 7.5 l of BioEasy Master Mix (SYBR Green) (Hangzhou Bioer Technology Co., Ltd.), 5.1 l of nuclease-free water, 2 l of diluted cDNA (200 ng), and 0.2 l of each genespecific primer (10 M) ( Table S2). The data were analyzed as described previously (22,46).

LC-MS/MS
GFP and HMGB1b-GFP stably transfected CIK cells were screened by G418 as reported previously (24). Cell lines stably expressing GFP and HMGB1b-GFP were seeded in 10-cm dishes and infected with GCRV for 24 h. Then the cells were collected for IP with GFP trap beads. The protein samples were eluted by SDT lysis. A portion of the IP samples was used to examine IP efficiency by WB with GFP Ab. LC-MS/MS analysis was performed on a Q Exactive mass spectrometer (Thermo Scientific). MS/MS spectra were searched using the MASCOT engine (Matrix Science) against the Actinopterygii UniProt sequence database and a grass carp transcriptome database in the NCBI SRA browser (Bioproject accession number SRP049081) (47).

Antiviral activity assay
To evaluate the antiviral activities of HMGB1b and HSP70, CIK cells were transfected with HMGB1b-GFP and HSP70-HA for 16 h and infected with GCRV at an m.o.i. of 0.1 for 36 h. The culture medium was collected as titer samples. To assess the influence of autophagy on GCRV replication, CIK cells were pretreated with or without 100 nM rapamycin, 5 mM 3-MA, and 0.15 mM H 2 O 2 for 4 h and infected with GCRV. 36 h after GCRV infection, the supernatants were collected and filtered for titer samples. Virus titers in the supernatants were determined by standard plaque assay and the PFU method (48). Virus titerinduced cell viability was analyzed by MTT assay. In brief, approximately 1 ϫ 10 4 CIK cells/well were seeded into a 96-well plate and cultured for 24 h. Then the medium was replaced with fresh medium supplemented with different titer samples, and the cells were incubated for 24 h. Then, 10 l of MTT solution was added to the medium and incubated for another 4 h. Subsequently, 100 l of formazan solution was added to each well. The optical density was observed at 570 nm using a microplate reader.

siRNA-mediated knockdown
Transient knockdown of endogenous HMGB1b and HSP70 in CIK cells was achieved by transfection of siRNA targeting HMGB1b and HSP70 mRNA, respectively. Three siRNA sequences (HMGB1b s1, GGAAGACAAUGUCUGCCAATT; HMGB1b s2, GGAGAGAAGAAGAGGCGAUTT; HMGB1b s3, GCAGCCCUAUGAGAAGAAATT; HSP70 s1, GCUGAU-UGGCAGGAAGUUUTT; HSP70 s2, GGACAAGUCUCAG-AUCCAUTT, HSP70 s3, CCAAUCAUCACUAAGCUUUTT) targeting different regions of each gene were synthesized by GenePharma (Jiangsu, China). The silencing efficiencies of the siRNA candidates were evaluated by qRT-PCR and compared with those in the negative control siRNA provided by the supplier. Our preliminary experiment indicated that s1 of HMGB1b and s1 of HSP70 possess the best silencing efficiency at a final concentration of 100 nM in mRNA levels. For WB, the LGP2-FLAG overexpression CIK cell line was plated in 6-well plates and transfected with s3 using FuGENE 6. The subsequent experiments were performed with s1 of HMGB1b and HSP70, respectively. For WB analysis, CIK cells were seeded in 6-well plates for 24 h. The cells were transfected with the indicated siRNA for 16 h and treated with H 2 O 2 for another 24 h. Then the cells were lysed for WB analysis.
Author contributions-Y. R. and J. S. conceptualization; Y. R. and Z. L. data curation; Y. R. validation; Y. R. and Q. W. investigation; Y. R. visualization; Y. R., Q. W., and C. Y. methodology; Y. R. writing-original draft; Y. R., Z. L., and J. S. writing-review and editing; C. Y. formal analysis; L. L. resources; J. S. supervision; J. S. project administration; H. S. and X. X. data analysis; J. J. performance of the experiments.