Activity-Based Protein Profiling Identifies ATG4B as a Key Host Factor for Enterovirus 71 Proliferation

Enterovirus 71 (EV71), one of the major pathogens of human HFMD, has caused outbreaks worldwide. How EV71 efficiently assesses its life cycle with elaborate interactions with multiple host factors remains to be elucidated. In this work, we deconvoluted that the host ATG4B protein processes the viral polyprotein with its cysteine protease activity and helps EV71 replicate through a chemical biology strategy. Our results not only further the understanding of the EV71 life cycle but also provide a sample for the usage of activity-based proteomics to reveal host-pathogen interactions.

ABSTRACT

Virus-encoded proteases play diverse roles in the efficient replication of enterovirus 71 (EV71), which is the causative agent of human hand, foot, and mouth disease (HFMD). However, it is unclear how host proteases affect viral proliferation. Here, we designed activity-based probes (ABPs) based on an inhibitor of the main EV71 protease (3C^pro), which is responsible for the hydrolysis of the EV71 polyprotein, and successfully identified host candidates that bind to the ABPs. Among the candidates, the host cysteine protease autophagy-related protein 4 homolog B (ATG4B), a key component of the autophagy machinery, was demonstrated to hydrolytically process the substrate of EV71 3C^pro and had activity comparable to that of the viral protease. Genetic disruption of ATG4B confirmed that the enzyme is indispensable for viral proliferation in vivo. Our results not only further the understanding of host-virus interactions in EV71 biology but also provide a sample for the usage of activity-based proteomics to reveal host-pathogen interactions.

IMPORTANCE Enterovirus 71 (EV71), one of the major pathogens of human HFMD, has caused outbreaks worldwide. How EV71 efficiently assesses its life cycle with elaborate interactions with multiple host factors remains to be elucidated. In this work, we deconvoluted that the host ATG4B protein processes the viral polyprotein with its cysteine protease activity and helps EV71 replicate through a chemical biology strategy. Our results not only further the understanding of the EV71 life cycle but also provide a sample for the usage of activity-based proteomics to reveal host-pathogen interactions.

INTRODUCTION

Enterovirus 71 (EV71) is one of the major etiological agents of human hand, foot, and mouth disease (HFMD) worldwide and has aroused severe public health concerns (1). EV71 belongs to the Picornaviridae family and has an icosahedral virus capsid packaging a single-stranded, positive-stranded RNA genome (2–4). After entry into host cells, the genome of EV71 immediately translates a single polyprotein that must be hydrolyzed predominantly by the viral protease 3C^pro to generate individual components for subsequent viral replication processes (5). Therefore, EV71 3C^pro plays a primary role in the viral replication cycle and is known as a major target for antiviral discovery.

Virus-host interactions play key roles in multiple steps of the EV71 life cycle (6, 7). For example, scavenger receptor B2 (SCARB2), P-selectin glycoprotein ligand 1 (PSGL-1), and cyclophilin A (CypA) function as receptors or entry factors to mediate viral entry (8–10). Poly(A) binding protein 1 (PABP1) and poly(rC) binding protein 2 (PCBP2) are required for the genome replication of picornavirus (11, 12). Moreover, viral proteins can regulate host proteins to achieve highly efficient propagation. Remarkably, EV71 3C^pro not only has been found to hydrolyze viral polyproteins but also has been found to process host proteins for viral replication. For example, EV71 3C^pro cleaves cellular interferon regulatory factor 7 (IRF7) (13), the TAK1/TAB1/TAB2/TAB3 complex (14), TIR domain-containing adaptor inducing interferon beta (TRIF) (15), and CstF-64 (16) to achieve antiviral innate immune evasion. It has also been reported that EV71 infection promotes host cell apoptosis through cleavage of the cellular protein PinX1 by EV71 3C^pro (17). However, how host proteases affect viral polypeptides/proteins and facilitate viral replication is elusive, possibly due to a lack of bona fide biological methods.

Because the catalytic activity of proteases is strictly regulated by a myriad of posttranslational modifications, traditional abundance-based proteomic methods cannot distinguish the potential significant divergence between protease abundance and activity (18). Activity-based protein profiling (ABPP), which is a chemoproteomic tool for detecting active enzymes in complex biological systems, was developed to address this issue (19). ABPP has the advantage of directly detecting changes in the catalytic activity of target proteins in complex proteomes and has become an important tool for uncovering virus-host interactions (20). ABPP has been used to detect host enzymes that are differentially regulated during viral propagation, profile the alteration of the functional proteome of a pathogen in response to external cues, and discover related protease inhibitors (21, 22).

One of the major challenges in the design of activity-based probes (ABPs) is to appropriately balance the selectivity of the probe and reactivity toward the target protein (23). Directed ABPs with strong electrophilic reactive groups (such as fluorophosphonate and phosphonate esters) can form a covalent bond with the catalytic residue of the target enzyme active site (7). However, such ABPs have low intrinsic reactivity and label only very limited protease families with high selectivity, impeding detection of the cross talk of diverse host proteases with viruses (24, 25). In contrast, nondirected ABPs with mild electrophilic reactive groups (such as sulfonate esters and α-chloroacetamide) are capable of labeling virtually all nucleophilic residues, which broadens the number of enzyme classes addressable by ABPP (7). However, this type of ABP suffers from poor specificity, resulting in extensive numbers of candidates and thus increasing the difficulty of further validation (26, 27).

Here, we designed ABPs based on rupintrivir and NK-1.8k, which are well-known EV71 3C^pro inhibitors (28, 29), to investigate host factors in response to the hydrolysis of a polypeptide and replication of EV71. Among the identified candidates, we demonstrate that autophagy-related protein 4 homolog B (ATG4B), a major component in autophagosome biogenesis (30), hydrolytically cleaves the substrate of EV71 3C^pro and plays a key role in EV71 replication in host cells. Our work demonstrates the ability of activity-based probes to define host-virus enzymatic cross talk in the virus life cycle.

RESULTS

Design and synthesis of activity-based probes.

Rupintrivir, a reported peptidomimetic EV71 3C^pro inhibitor (28), was selected as the reference compound for the design of ABPs. Rupintrivir has an α,β-unsaturated ester as the warhead at the P1′ position, a γ-lactam ring at the P1 position, and 4-fluorophenyl, isopropyl, and 5-methylisoxazole groups at positions P2 and P3 (Fig. 1A) (28). Rupintrivir binds to the catalytic center of EV71 3C^pro and inactivates the enzyme by forming a covalent bond with the catalytic cysteine residue via the α,β-unsaturated ester group (28).

FIG 1.

Structures of compounds. (A) Structures of rupintrivir and NK-1.8k. (B) Structures of ABPs 1 to 8.

Based on the structure of rupintrivir, we selected the α,β-unsaturated ester group that was used in rupintrivir as the warhead. Previous studies have shown that the replacement of the γ-lactam ring by a δ-lactam ring at the P1 position alters specificity against EV71 3C^pro and its isoenzyme, 3C-like protease (3CL^pro), encoded by Middle East respiratory syndrome coronavirus (MERS-CoV) (31, 32). We therefore decided to use the δ-lactam ring and the γ-lactam ring as recognition moieties to achieve various selectivities of the probes. Moreover, in our previous works, we demonstrated that the structure of P2 interacts with the residues in the catalytic pocket of EV71 3C^pro to significantly increase the binding affinity of the inhibitor (33), while the P3 position has less of an effect on binding and inhibitory activity (31, 33). Because we wanted to detect the cross talk of diverse host factors with the virus, we excluded the groups at P2 and P3 to moderate the binding affinity and specificity of the ABPs against EV71 3C^pro and host factors.

An alkyne handle was introduced by linkers with various lengths to facilitate subsequent detection, enrichment, and identification by copper-catalyzed azide-alkyne cycloaddition (CuAAC) (34) to tetramethyl rhodamine (TAMRA) for fluorescence detection or biotin for enrichment. In addition, three straight-chain probes were synthesized as negative controls to verify the specific recognition of the lactam ring of EV71 3C^pro. Eventually, three ABPs with a δ-lactam ring, two ABPs with a γ-lactam ring, and three straight-chain ABPs as negative controls were synthesized for further investigation (Fig. 1B).

Specificity of ABP labeling of EV71 3C^pro and its isoenzyme.

We first investigated the ability of ABPs to bind recombinant EV71 3C^pro in vitro. As expected, ABPs 1 to 5 showed promising labeling of EV71 3C^pro, while ABPs 7 and 8 displayed a limited binding ability, and ABP 6 had a nondetectable reaction with EV71 3C^pro (Table 1). Consistent with the respective binding affinities, ABP 1 (50% inhibitory concentration [IC₅₀] = 120.2 μM) presented the best enzymatic inhibition activity of EV71 3C^pro, while ABPs 2 to 5 showed less inhibition than ABP 1, and ABPs 6 to 8 did not inhibit the enzymatic activity (Table 2). The modified sites on EV71 3C^pro were analyzed by digesting labeled EV71 3C^pro. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) results showed that all ABPs labeled the active-site Cys147 residue of EV71 3C^pro, except for ABP 3, which labeled Cys171 (Fig. 2A).

TABLE 1.

Labeling of recombinant EV71 3C^pro with ABPs

ABP	% labeling
	EV71 3Cpro	Labeled EV71 3Cpro
1	10	90
2	15	85
3	20	80
4	17	83
5	21	79
6	>95	<5
7	80	20
8	81	19

TABLE 2.

Enzyme-inhibitory activity of ABPs as EV71 3C^pro inhibitors

ABP	IC50 (μM)
1	120.2
2	306.0
3	188.9
4	168.1
5	259.4
6	>800
7	>800
8	>800

FIG 2.

Structures of EV71 3C^pro (A) and MERS-CoV 3CL^pro (B) shown as colored cartoons. The cysteine residues labeled by ABPs are displayed as colored sticks and are labeled in the enlarged insets.

We next used MERS-CoV 3CL^pro (35), an isoenzyme of EV71 3C^pro, to assess the labeling specificity of the ABPs. Both 3C^pro and 3CL^pro are cysteine proteases and share a typical chymotrypsin-like fold with a conserved nucleophilic cysteine residue in the active center (32). MERS-CoV 3CL^pro has two free cysteine residues exposed to the solvent, Cys145 at the entrance of the catalytic pocket and Cys148 in the catalytic center (Fig. 2B) (35). The results showed that ABP 2, ABP 4, and ABP 5 labeled the catalytic Cys148 residue of MERS-CoV 3CL^pro, while ABP 1 and ABP 3 labeled Cys145. A previous study demonstrated that inhibitors with the γ-lactam ring had enhanced inhibitory activities against MERS-CoV 3CL^pro (35). Consistently, the ABPs with the γ-lactam ring all labeled the active site of MERS-CoV 3CL^pro. Collectively, the synthesized ABPs have diverse selectivity against different viral proteases. ABPs with the δ-lactam ring have a greater affinity for EV71 3C^pro, while ABPs with the γ-lactam ring have a greater affinity for MERS-CoV 3CL^pro. As the negative controls, ABPs 6 to 8 did not react with either of the viral proteases.

Activity-based protein profiling with ABPs.

We then carried out an analysis of RD cell lysates infected with EV71 as a background to explore the labeling efficacy of ABPs. RD cells were infected with wild-type EV71 (wt-EV71) at a multiplicity of infection (MOI) of 10 after 15 h until the cytopathic effect (CPE) was obviously observed, and the infected cells were maintained to globally assess ABP activity.

ABPs tagged with TAMRA-azide were used to optimize the concentration of ABPs and the reaction time by in-gel fluorescence detection (Fig. 3A and B). In order to avoid the nonspecific labeling caused by high concentrations of ABPs and the instability of the proteome, most ABPs with strong signals at a concentration of 200 μM at 60 min were chosen as the final reaction conditions (Fig. 3C and D). We subsequently performed labeling reactions with ABPs 1 to 8 under these optimal conditions. The results showed that ABPs 1 to 4 had similar proteome-labeling profiles and that ABP 5 produced weaker labeled bands, while the labeling achieved by ABPs 6 to 8 was not confirmed (in particular, the band near 25 kDa was missing) (Fig. 3E). Next, densitometry of 3C^pro with different ABPs was performed by using ImageJ (Fig. 3F). ABP 5 was the most efficient in labeling the EV71 3C^pro probe, but its ability to label the proteome was significantly weaker than those of the others. Combined with the binding affinity and enzymatic inhibition results, these results verify that ABP 1 is the most suitable ABP for further ABPP analysis based on its promising reactivity and selectivity.

FIG 3.

Reactivity and selectivity of ABPs in proteomes. (A) Concentration-dependent labeling of RD cell lysates infected by wt-EV71. (B) Time-dependent labeling of RD cell lysates infected by wt-EV71. (C) Quantification of the intensity of concentration-dependent labeling. The intensity of labeling of EV71 3C^pro was quantified by using ImageJ, normalized to the maximum concentration of ABPs and graphed as a percentage. (D) Quantification of the intensity of time-dependent labeling. The intensity of labeling of EV71 3C^pro was quantified by using ImageJ, normalized to the maximum reaction time of ABPs and graphed as a percentage. (E) Comparison of the reactivity of ABPs 1 to 8 in RD cell lysate infected by wt-EV71 at an MOI of 10 after 15 h and incorporation of TAMRA-azide using CuAAC. (Top) In-gel fluorescence; (bottom) Coomassie brilliant blue (CBB) staining. (F) Quantification of the intensity of different ABPs labeling EV71 3C^pro. The intensity of labeling of EV71 3C^pro was quantified by using ImageJ, normalized to the value for β-actin and graphed as a percentage. (G) Comparative ABPP of ABP 1 labeling. (H) Competitive ABPP of ABP 1 labeling.

To assess the variety in the amounts of viral or host cysteine proteases upon EV71 infection, we carried out a comparative analysis of native RD cells and EV71-infected RD cells to observe the disparities. The cells were harvested, lysed, incubated with TAMRA-azide-coupled ABP 1, separated by SDS-PAGE, and analyzed by in-gel fluorescence scanning. ABP 1 labeling specificity was validated by a clear difference in the signals observed before and after infection. Upon EV71 infection, the intensity of the band at approximately 25 kDa, corresponding to EV71 3C^pro, was clearly stronger than that in native RD cells (Fig. 3G).

We also performed competitive ABPP by using NK-1.8k (Fig. 1A), a previously reported EV71 3C^pro inhibitor, by occupying the catalytic pocket (29). As expected, competitive inhibition with NK-1.8k resulted in a decrease in ABP 1 labeling (Fig. 3H). The competitive ABPP assay indicated that ABP 1 and NK-1.8k had overlapping targets, which correspond to the decreasing band intensities with increasing inhibitor concentrations. When the inhibitor concentration was increased to 100 nM, the band corresponding to EV71 3C^pro almost completely disappeared, suggesting that the binding affinity of NK-1.8k for EV71 3C^pro is higher than that of ABP 1 and that the presence of NK-1.8k at a high concentration completely blocks the binding of ABP 1 to EV71 3C^pro.

In summary, these results show that ABP 1 is the most efficient ABP, and we thus employed ABP 1 for the profiling of cellular targets involved in EV71 infection.

Target identification of competitive ABPP.

Gel-based approaches offer a method to rapidly screen global ABP alterations (19). We enriched the ABP 1-labeled proteins by using streptavidin beads after conjugation of the probe to biotin-azide via CuAAC. The candidates were separated by SDS-PAGE and analyzed by LC-MS/MS with label-free quantification after in-gel trypsin digestion (Fig. 4A) (36).

FIG 4.

Target identification by competitive ABPP and comparative ABPP. (A) Scheme of ABPP methods used to identify proteins. (B) Representative ratio plot for proteins identified by competitive ABPP from the active proteome treated with ABP 1 versus NK-1.8k. (C) Cellular component analysis of proteins identified by competitive ABPP. (D) Biological process analysis of proteins identified by competitive ABPP. (E) Representative ratio plot for proteins identified by comparative ABPP from the active proteome infected versus mock. (F) Cellular component analysis of proteins identified by comparative ABPP. (G) Biological process analysis of proteins identified by comparative ABPP. ER, endoplasmic reticulum.

We first performed competitive ABPP to study the specific targets of ABP 1 in competition with NK-1.8k. The proteomes were treated with NK-1.8k for 60 min, followed by incubation with ABP 1 for another 60 min. Three replicate assays were performed to overcome biological and experimental variations, and proteins that were present in at least two experiments were selected as candidates. Because NK-1.8k is an inhibitor of EV71 3C^pro, it is not surprising that NK-1.8k selectively targeted EV71 3C^pro with a quantification ratio of 73, which is remarkably higher than those of the other candidates. There were 144 proteins that were found to be labeled by ABP 1 (Fig. 4B and Fig. 5A). Gene ontology (GO) analysis showed that the identified proteins are mainly distributed in the cytosol, nucleus, and mitochondria, while some of the proteins are distributed in the endoplasmic reticulum, endosome, and lysosome (Fig. 4C). Interestingly, although the identified proteins are involved in diverse biological processes, three are involved in biological processes related to the viral life cycle, including viral gene expression, viral transcription, the viral life cycle, and translation (Fig. 4D).

FIG 5.

(A) Overlap of the proteins identified by competitive ABPP and comparative ABPP. (B and C) LC-MS/MS analysis of the IMPDH protein incubated with (C) or without (B) ABP 1. “ca” represents the carbamidomethyl Cys, and “ab” represents the Cys bound by ABP 1.

Target identification using comparative ABPP.

We next performed comparative ABPP by comparing native and EV71-infected RD cells to ABP 1. There were 146 proteins found to be labeled by ABP 1 (Fig. 4E and Fig. 5A). GO analysis showed that the proteins identified in the comparative ABPP are distributed across major intracellular compartments (Fig. 4F) and represent several processes related to viral infection (Fig. 4G), as observed for competitive ABPP.

We further cross-checked the targets identified from the competitive ABPP and comparative ABPP to obtain a total of 133 common targets (Fig. 5A). As the purpose of this study was to understand how host proteases function in the virus life cycle, we focused our attention on proteases that could hydrolyze peptides, including UCHL1, USP4, USP15, and ATG4B (Table 3). UCHL1, USP4, and USP15 are deubiquitinating enzymes (DUBs), which share homologous structures and conserved cysteines to catalyze the hydrolysis of the C-terminal peptide extensions of ubiquitin (Ub) with a conserved catalytic mechanism (37). ATG4B is a key component of the autophagy machinery and is a cysteine protease that cleaves pro-LC3 isoforms to form LC3-I and cleaves phosphatidylethanolamine from LC3-II to regenerate LC3-I for further membrane biogenesis (38). ATG4B also plays roles in autophagy-related antiviral innate immunity or, conversely, in helping the virus to achieve efficient proliferation (39).

TABLE 3.

Cysteine proteases identified by comparative ABPP and competitive ABPP

Accession no.	Protein name	Gene name	MW (kDa)a	Ratio by comparative ABPPb	Ratio by competitive ABPPb
B8YLV9	3C protease	EV71 3C	20		73.11
P09936	Ubiquitin carboxyl-terminal hydrolase isozyme L1	UCHL1	24.8	2.64	1.04
Q13107	Ubiquitin carboxyl-terminal hydrolase 4	USP4	108.6	1.00	1.63
Q9Y4E8	Ubiquitin carboxyl-terminal hydrolase 15	USP15	112.4	1.02	1.74
Q9Y4P1	Autophagy-related protein 4 homolog B	ATG4B	44.3	1.78	1.47

^aMW, molecular weight.

^bThe reported ratios are the means from three biological replicates.

We also identified that PRDX1, PRDX3, and PRDX6, a series of peroxide reductases that contain a single conserved cysteine as the active site, were upregulated, suggesting that EV71 infection may induce oxidative stress in host cells. In addition, for IMP dehydrogenase 2 (IMPDH2), it was found that the selective druggable site Cys140 was covalently modified by the α,β-unsaturated ester in ABP 1 via the Michael addition reaction (Fig. 5B and C) (40).

ATG4B cleaves the substrate of EV71 3C^pro.

We next verified whether the identified host proteases can process the substrate of EV71 3C^pro. The peptide substrate is derived from the natural substrate of EV71 3C^pro, and the cleavage activity is measured using fluorescence resonance energy transfer (FRET). Cleavage of the glutamine-glycine linkage resulted in an intensified fluorescence signal when excited at 340 nm with a high dynamic range (Fig. 6A) (41). Because UCHL1, USP4, and USP15 have similar structures and functions (37), UCHL1 was selected as a representative for subsequent investigations. The results showed that EV71 3C^pro hydrolyzed the peptide substrate and released a strong fluorescence signal (Fig. 6B). ATG4B also reacted with the substrate, and the intensity of the released fluorescence signal was over 50% of that generated by EV71 3C^pro (Fig. 6B). UCHL1 showed only slight activity (∼10%) against the EV71 3C^pro substrate (Fig. 6B). We further determined the Michaelis-Menten constants of EV71 3C^pro and ATG4B. The K_m and V_max of ATG4B against the EV71 3C^pro substrate were both equivalent to 50% of those of EV71 3C^pro (Fig. 6C).

FIG 6.

ATG4B cleaves the substrate of EV71 3C^pro. (A) Structure of fluorogenic peptide substrate for EV71 3C^pro. (B) Progress curve of the time-dependent interaction of different proteases with the substrate of EV71 3C^pro. EV71 3C^pro, ATG4B, and UCHL1 were diluted into assay buffer containing the peptide substrate. A full substrate reaction without protease served as a negative control. Assays were conducted in black 96-well plates. (C) Determination of the K_m and V_max of the peptide substrate for EV71 3C^pro and ATG4B. The enzyme velocities obtained from the recorded time courses are plotted versus the substrate concentration, and the kinetic constants K_m and V_max were calculated by using the Michaelis-Menten equation. Each data point represents the average from three replicates. The error bars represent the standard errors of the means (SEM). (D) Ability of EV71 3C^pro, ATG4B, and ATG4B C74A to process the recombinant protein substrate EV71 3CD C147A. The reaction products were analyzed by 12% SDS-PAGE and Coomassie brilliant blue staining. The names and positions of the detected proteins are shown at the right.

The EV71 3CD protein substrate of EV71 3C^pro was subsequently expressed and purified (5), and its active-site residue Cys147 was mutated to alanine to avoid EV71 3CD self-hydrolysis. Over 90% of the protein substrate was hydrolyzed by EV71 3C^pro, while approximately 50% of the protein substrate was hydrolyzed by ATG4B (the two bands of ATG4B resulted from the degradation of the His tag) (Fig. 6D). Strikingly, when rupintrivir and NSC 185058 (an inhibitor of ATG4B) (42) were added separately, the hydrolysis activity of EV71 3C^pro and ATG4B was abolished (Fig. 6D). Moreover, the ATG4B C74A mutant with defective protease activity lost its ability to process the protein substrate (Fig. 6D). Collectively, these experiments confirm that ATG4B could process not only the peptide substrate but also the protein substrate of EV71 3C^pro.

ABP 1 targets ATG4B by covalently binding at Cys74.

To verify the binding of ABP 1 to ATG4B, we incubated ATG4B with TAMRA-azide-tagged ABP 1. ATG4B labeling was confirmed by in-gel fluorescence and Coomassie brilliant blue staining (Fig. 7A). ATG4B was then incubated with ABP 1 in the absence or in the presence of β-mercaptoethanol (BME) or dl-dithiothreitol (DTT) for competitive binding to ATG4B via the thiol (Fig. 7B). The results showed that BME and DTT completely abolished ATG4B binding to ABP 1, indicating that ABP 1 covalently binds to the thiol of cysteine.

FIG 7.

ABP 1 targets ATG4B by covalently binding at Cys74. (A) ABP 1 binds to ATG4B. ABP 1 was incubated with recombinant ATG4B protein at increasing concentrations for 1 h, and the incorporation of TAMRA-azide was determined using CuAAC. The labeled protein was detected by fluorescence (Fluo.). (B) ABP 1 covalently binds to ATG4B at the thiol of cysteine. The recombinant ATG4B protein was preincubated with BME or DTT for 1 h and further incubated with ABP 1 for 1 h. The labeled protein was detected by fluorescence. (C and D) LC-MS/MS analysis of ATG4B incubated with (D) or without (C) ABP 1. “ca” represents the carbamidomethyl Cys, and “ab” represents the Cys bound by ABP 1.

The active catalytic-triad site of ATG4B consists of Cys74, His280, and Asp278 (30). To determine whether the catalytic cysteine residue was attacked, ATG4B was incubated with ABP 1, followed by digestion and LC-MS/MS analysis. The molecular weight of the carbamidomethyl peptide NFPAIGGTGPTSDTGWGCMLR was 2,195 Da (Fig. 7C), but the molecular weight of this peptide was reduced to 2,486 Da after incubation of ATG4B with ABP 1 (Fig. 7D). The molecular weight difference of 291 Da is exactly the difference between the molecular weight of ABP 1 and the molecular weight of the carbamidomethyl group. We infer from the MS/MS spectrum that the shift of this molecular weight occurs at Cys74, which further indicates that ABP 1 covalently binds to ATG4B at Cys74 (Fig. 7C and D).

Taken together, these results show that ATG4B can process the EV71 3C^pro substrate and suggest that ATG4B is likely to be a host factor that facilitates EV71 replication in host cells.

Silencing of ATG4B expression attenuates EV71 proliferation.

We next verified the biological impact of ATG4B on EV71 proliferation in cell-based assays. We introduced short hairpin RNAs (shRNAs) targeting ATG4B by lentiviral vectors into RD cells to downregulate the expression of endogenous ATG4B. Compared with cells that received a nontargeting control shRNA (sh-NC), sh-ATG4B-treated cells were significantly less susceptible to EV71 infection (Fig. 8A). As RD-sh-ATG4B-3 cells displayed the greatest silencing and inhibition efficiency, we used this shRNA in the following experiment and refer to it as sh-ATG4B here. When cells were infected with wt-EV71 at both an MOI of 0.01 after 24 h and an MOI of 20 after 12 h, the EV71 RNA levels in RD-sh-ATG4B cells were diminished to approximately 50% of the levels observed in RD-sh-NC cells (Fig. 8B). Next, we generated genomic wt-ATG4B (pATG4B^wt) and ATG4B C74A (pATG4B^C74A) constructs in the pcDNA plasmid vector and introduced them into RD cells. Both pATG4B^wt and pATG4B^C74A rescued the expression level of ATG4B (Fig. 8C, left). The overexpression of pATG4B^wt increased the EV71 RNA level, but the overexpression of pATG4B^C74A did not enhance EV71 replication (Fig. 8C, right). Moreover, the expression level of EV71 VP1, one of the EV71 capsids, was also significantly reduced in RD-sh-ATG4B cells, which was consistent with the EV71 RNA level (Fig. 8D and E). Next, RD-sh-NC and RD-sh-ATG4B cells were infected with EV71-Gluc virus (a stable Gaussia luciferase EV71 reporter virus) (43) at a low MOI, and EV71 proliferation was measured over the next 3 days by monitoring the Gaussia luciferase activity. This analysis revealed that EV71 was persistently suppressed in RD-sh-NC cells compared to RD-sh-ATG4B cells (Fig. 8G).

FIG 8.

Silencing of ATG4B expression attenuates EV71 proliferation. (A) Quantification of the ATG4B expression level and virus infection of RD-sh-NC and RD-sh-ATG4B cells. Infections were quantified by measuring the luciferase activity in relative luminescence units. Virus infection is expressed as a percentage relative to that in RD-sh-NC cells. The RNA expression levels were quantified by qRT-PCR, normalized to the value for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and graphed as a percentage of the levels determined in RD-sh-NC cells. Each data point represents the average from three replicates. The error bars represent the SEM. (B) Quantification of infection of 4RD-sh-NC and RD-sh-ATG4B cells that were infected with wt-EV71 at an MOI of 0.01 after 24 h and an MOI of 20 after 12 h by qRT-PCR. (C) RD cells containing pATG4B^wt or pATG4B^C74A were infected with EV71-Gluc at the same time. The RNA expression level of ATG4B (left) and the RNA level of EV71 (right) were quantified by qRT-PCR. (D) Abundance of ATG4B in knockdown RD cells. RD cells were infected with shRNA recombinant lentiviruses that were specific for ATG4B (RD-sh-ATG4B) and control shRNA (RD-sh-NC). ATG4B and GAPDH (an internal control) were detected by Western blot analysis. (E) Detection of expression of EV71 VP1 in RD-sh-NC and RD-sh-ATG4B cells that were infected with wt-EV71 at an MOI of 1 after 24 h by Western blot analysis. (F) Detection of expression of EV71 VP1 and ATG4B in HEK293T-sh-NC, HEK293T-sh-ATG4B, HeLa-sh-NC, HeLa-sh-ATG4B, Vero-sh-NC, and Vero-sh-ATG4B cells that were infected with wt-EV71 at an MOI of 20 after 24 h by Western blot analysis. (G) RD-sh-NC or RD-sh-ATG4B cells were infected with EV71-Gluc at an MOI of 0.01. Infection by EV71 was detected by the Gaussia luciferase activity in relative luminescence units (RLU) at different times postinoculation, as indicated. (H) Quantification of ATG4B expression levels in HEK293T-sh-NC, HEK293T-sh-ATG4B, HeLa-sh-NC, HeLa-sh-ATG4B, Vero-sh-NC, and Vero-sh-ATG4B cells by qRT-PCR. (I) Quantification of infection of HEK293T-sh-NC, HEK293T-sh-ATG4B, HeLa-sh-NC, HeLa-sh-ATG4B, Vero-sh-NC, and Vero-sh-ATG4B cells that were infected with wt-EV71 at an MOI of 20 after 24 h by qRT-PCR. Statistical significance was calculated by Student’s t test. N.S., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.005; ****, P < 0.001.

To verify whether the impact of ATG4B on EV71 infection is a general rather than a cell-specific effect, we generated stable ATG4B silencing in HEK293T, HeLa, and Vero cells (Fig. 8H). We infected these stable knockdown cells with EV71 at an MOI of 20 and found that the decreased EV71 RNA level and VP1 expression level were comparable to those in RD ATG4B knockdown cells (Fig. 8F and I), indicating that the inhibition of EV71 proliferation by the absence of ATG4B is not cell type specific. All of these results verified the significant impact of ATG4B on the EV71 life cycle.

ATG4B functions during EV71 replication.

It has been found that ATG4B is required for the translation of incoming hepatitis C virus (HCV) RNA and, thereby, for the initiation of HCV replication (39). To identify the steps in the EV71 life cycle affected by ATG4B, we first performed single-cycle infections in RD-sh-NC and RD-sh-ATG4B cells that had been infected with EV71(FY)-Luc pseudotype virus (44). We found that the luciferase activity in both RD-sh-NC and RD-sh-ATG4B cells was almost consistent until 4 h postinfection (hpi) (Fig. 9A). The augmentation of luciferase activity indicated that the replication of EV71 began at 4 hpi in both RD-sh-NC and RD-sh-ATG4B cells. Although replication began in both cell types almost simultaneously, replication in RD-sh-ATG4B cells was significantly slower than that in RD-sh-NC cells (Fig. 9A). This result demonstrates that ATG4B is required for EV71 infection at a postentry step.

FIG 9.

ATG4B functions in EV71 replication. (A) Growth curves of EV71(FY)-Luc pseudotype virus in RD-sh-NC or RD-sh-ATG4B cells at an MOI of 20. Infection by EV71 was detected by firefly luciferase activity, and infection is expressed as a percentage relative to that in RD-sh-NC cells. RD-sh-NC and RD-sh-ATG4B cells were transfected with EV71 subgenomic replicon RNA. (B) RD-sh-NC and RD-sh-ATG4B cells were transfected with EV71 subgenomic replicon RNA. Infection by EV71 was detected by luciferase activity in relative luminescence units, and infection is expressed as a percentage relative to that in RD-sh-NC cells. (C) Detection of expression of ATG4B in RD-sh-NC and RD-sh-ATG4B cells containing ATG4B overexpression plasmids by Western blot analysis. (D) Inhibition of EV71(FY)-Luc pseudotype virus infection of RD cells by rupintrivir, pirodavir, and NSC 185058 at various times of addition and DMSO as a negative control. Infection by EV71 was detected by firefly luciferase activity, and infection is expressed as a percentage relative to the value for mock. (E and F) ATG4B can help EV71 replicate, and inactive EV71-Gluc-3C_mut is rescued by ATG4B overexpression. RD-sh-NC and RD-sh-ATG4B cells were transfected with EV71-Gluc or EV71-Gluc-3C_mut. In addition, RD-sh-ATG4B cells were transfected with EV71-Gluc or EV71-Gluc-3C_mut plus the pcDNA empty plasmid, pATG4B^wt, pATG4B^shRes, or pATG4B^C74A-shRes individually. The expression level of ATG4B (E) and the RNA level of EV71 (F) were quantified by qRT-PCR. Statistical significance was calculated by Student’s t test. N.S., not significant; *, P < 0.05; ***, P < 0.005; ****, P < 0.001.

To verify whether ATG4B impacts EV71 replication inside host cells, we transfected EV71 subgenomic replicon RNA lacking the P1 region (44) into RD cells containing sh-NC or sh-ATG4B, allowing the observation of alterations in viral replication and excluding any possible impact on the virus entry step. As expected, the replication of EV71 RNA was sharply downregulated in RD-sh-ATG4B cells (Fig. 9B).

We next analyzed whether NSC 185058 affects viral replication but not the entry process. EV71(FY)-Luc pseudotype virus was used to exclude the possibility of an impact of virus reinfection (44). RD cells were treated with rupintrivir, pirodavir (a reported capsid binding compound) (45), and NSC 185058 at −4, −2, 0, 2, 4, 6, 8, and 10 hpi, where 0 hpi is the time when inhibitors were supplied immediately after virus infection. The results showed that the inhibition of EV71 by NSC 185058 was similar to that by rupintrivir, although the inhibition efficiency against EV71 was lower than that of rupintrivir (Fig. 9D). Treatments with rupintrivir and NSC 185058 at −4 to 6 hpi displayed an antiviral effect, while the antiviral effect of pirodavir showed a dramatic decrease from 4 hpi (Fig. 9D). All of these results support the hypothesis that ATG4B functions in the EV71 replication step but not in the entry step.

Next, we engineered a recombinant EV71 3C^pro mutant replicon (EV71-Gluc-3C_mut) (43) through the introduction of a single cysteine-to-alanine substitution at the Cys147 position of EV71 3C^pro. The complete genome length was sequenced to ensure that no mutation occurred during reverse transcription-quantitative PCR (qRT-PCR) and cloning procedures. Equal amounts of EV71-Gluc and EV71-Gluc-3C_mut RNAs were transfected into RD-sh-NC and RD-sh-ATG4B cells to measure their RNA expression levels. At 24 h posttransfection, the RNA level in RD-sh-ATG4B cells was approximately 45% of that in RD-sh-NC cells transfected with EV71-Gluc. The RNA levels in both RD-sh-NC and RD-sh-ATG4B cells transfected with EV71-Gluc-3C_mut were less than 20% of the level in RD-sh-NC cells transfected with EV71-Gluc (Fig. 9F). We subsequently generated genomic RNA interference (RNAi)-resistant ectopic wt-ATG4B (pATG4B^wt-shRes) and ATG4B C74A mutant (pATG4B^C74A-shRes) constructs in the pcDNA plasmid vector. The pATG4B^wt plasmid rescued the expression level of ATG4B, while the pATG4B^wt-shRes and pATG4B^C74A-shRes plasmids performed better than the pATG4B^wt plasmid, achieving an almost 90% rescue level (Fig. 9C and E). The overexpression of ATG4B enhanced EV71-Gluc infection and increased the EV71 RNA level of EV71-Gluc-3C_mut by approximately 30% (Fig. 9F). However, the pATG4B^C74A-shRes plasmid could rescue the expression level of ATG4B but not the EV71 RNA level (Fig. 9E and F). These results together demonstrate that ATG4B is a key host factor participating in the replication stage of EV71 by hydrolytically processing the polyprotein of the virus.

DISCUSSION

In this work, we show that ABPs based on inhibitors of EV71 3C^pro are capable of identifying individual protein alterations during EV71 infection and demonstrate that ATG4B is a host factor that plays a key role in EV71 replication by hydrolyzing a viral polyprotein-like protease.

The identification of virus-host interactions is important to understand the biology underlying viral infections but is hindered by conventional biological methods. The hydrolysis of host proteins by viral proteases such as EV71 3C^pro has been demonstrated to be essential for viruses to escape host innate immune responses (13–16). However, knowledge of how host proteases function on viral polypeptides/proteins and facilitate viral replication is rare. ABPP provides a new way to investigate host factors that are involved in the viral life cycle. Here, we introduce a strategy for designing ABPs based on an inhibitor of a viral protease with balanced reactivity and selectivity. The P2 group for inhibitor binding, as well as the P3 group, can be eliminated to moderate the binding affinity and specificity of ABPs against EV71 3C^pro and increase tolerance to detect diverse host proteases. The strategy used in this work may be expanded to discover new interactions between host proteases and a wide range of viral proteases.

Autophagy is an intracellular degradation process in which cytoplasmic components, including organelles, are directed to the lysosome by a membrane-mediated process (46). As a key component of the autophagy machinery, ATG4B plays key roles in autophagosome biogenesis by processing and recycling LC3-I/II via its cysteine protease activity (47, 48). Previous studies have shown that autophagy is activated upon viral or bacterial infection and serves as an innate host defense mechanism (49). 3C^pro plays a crucial role in the viral life cycle and virus-host interactions, and its protease activity is indispensable. It was found that Seneca Valley virus (SVV) 3C^pro induced apoptosis through its protease activity (50), which was similar to that of coxsackievirus B3 (CVB3) 3C^pro (51). In contrast, ATG4B has been shown to be related to the translation of incoming HCV RNA, but the mechanism remains unclear (39). We demonstrated that ATG4B can hydrolytically process an EV71 3C^pro substrate and thus plays a key role in EV71 replication.

Taken together, our results demonstrate the usefulness of ABPP in virus-host interactions and further the understanding of how EV71 hijacks a host innate immunity-related protein for its life cycle.

MATERIALS AND METHODS

Cell lines and viruses.

Cell lines used in the experiment were purchased from the ATCC. The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco) containing 10% fetal bovine serum (FBS) (Gibco), 100 U/ml penicillin, and 100 mg/ml streptomycin (Gibco) at 37°C with 5% CO₂ in a humidified incubator.

The plasmids containing human EV71 strain AnHui1 (GenBank accession no. GQ994988.1) and EV71-Gluc strain HeN09 (GenBank accession no. AGH30721.1) were kindly provided by Bo Zhang from the Wuhan Institute of Virology. The EV71(FY)-Luc pseudotype virus system containing plasmid pcDNA6-FY-capsid and pEV71-Luc-replicon lacking the P1 region was provided by Wenhui Li (National Institute of Biological Sciences) (44). The EV71s were amplified in RD cells, quantified by determining the 50% tissue culture infective dose (TCID₅₀) per 1 ml in RD cells as previously described (52), and used for all experiments.

Proteome extraction and labeling.

RD cells were grown to 90% confluence and incubated with EV71 at an MOI of 20 for 15 h to transfect them with the virus. Cells were harvested by centrifugation at 1,000 rpm, washed twice with cooled phosphate-buffered saline (PBS), resuspended in 0.8 ml 0.1% NP-40 in PBS, and homogenized by sonication. The samples were subjected to centrifugation (40 min at 4°C at 20,000 × g) to isolate the soluble fraction, and the supernatant was collected. The protein concentration was assayed by the bicinchoninic acid (BCA) assay, and prepared soluble proteome samples were diluted to 2 mg/ml in PBS.

ABPs were dissolved in dimethyl sulfoxide (DMSO) to yield stock concentrations of 25 mM. Proteome samples (50 μl of 2 mg/ml proteome for fluorescence detection and 1 ml of 2 mg/ml proteome for enrichment experiment) were treated with the desired concentrations of different ABPs, and the solutions were incubated for 1 h at room temperature. A click reaction was performed with 200 μM TAMRA-azide or biotin-azide (Click Chemistry Tools), 1 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), 100 μM Tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA), and 1 mM CuSO₄ for 30 min at room temperature. ABPs coupled with TAMRA-azide were directly fractionated by SDS-PAGE and captured in fluorescence detection mode. ABPs tagged with biotin-azide were centrifuged to pellet the protein at 12,000 × g after the click reaction. The protein pellet was washed with cold methanol twice and resuspended by sonication in 1.2% SDS in PBS. The samples were heated (3 min at 90°C) and then added to 5 ml PBS containing 100 μl Pierce streptavidin agarose resins (Thermo Fisher Scientific). After incubation for 3 h, the beads were washed with 0.2% SDS in PBS (5 ml for 10 min), PBS (3 times with 5 ml), and water (3 times with 5 ml). Finally, the beads were boiled (10 min at 95°C), and the samples were fractionated by SDS-PAGE and silver staining.

Experimental conditions for in-gel trypsin digestion.

After silver staining, each lane was cut into 4 slices in a LoBind tube. A solution containing 200 μl K₃Fe(CN)₆ (30 mM) and Na₂S₂O₃ (100 mM) was added to the tube until the color faded. The solution was removed, and the gel pieces were dehydrated by the addition of 40 μl acetonitrile. The pieces were reduced with DTT (10 mM) and alkylated with iodoacetamide (55 mM). The pieces were added to 40 μl of a working trypsin solution and incubated for 18 h at 37°C. The peptides were extracted with a solution containing 50% acetonitrile and 5% formic acid. For LC-MS/MS analysis, the peptides were separated by a 120-min gradient elution at a flow rate of 0.30 μl/min with a Thermo-Dionex Ultimate 3000 high-performance liquid chromatography (HPLC) system, which was directly interfaced with a Thermo Scientific Q Exactive mass spectrometer. The analytical column was a homemade fused silica capillary column (75-μm internal diameter [ID], 150-mm length; Upchurch, Oak Harbor, WA) packed with C₁₈ resin (300 Å, 5 μm; Varian, Lexington, MA). The mobile phase consisted of 0.1% formic acid, and mobile phase B consisted of 80% acetonitrile and 0.1% formic acid. The Q Exactive mass spectrometer was operated in the data-dependent acquisition mode using Xcalibur 2.1.2 software, and there was a single full-scan mass spectrum in the orbitrap (m/z 300 to 1,800 m/z; 70,000 resolution), followed by 20 data-dependent MS/MS scans at a 27% normalized collision energy (high-energy collisional dissociation [HCD]).

Protein identification and label-free quantification.

MaxQuant_1.5.5.1 was used for protein identification and label-free quantification using 20 ppm for the first-search peptide tolerance and 4.5 ppm for the main-search tolerance. Acetyl (protein N terminus), oxidation (M), carbamidomethyl (C), and ABPs labeled were all set as variable modifications. Trypsin was specified as the proteolytic enzyme with N-terminal cleavage to proline and two missed cleavages allowed. The precursor ion mass tolerances were set at 20 ppm for all MS scans acquired in an orbitrap mass analyzer, and the fragment ion mass tolerance was set at 20 millimass units (mmu) for all MS2 spectra acquired. When the q value was <1%, the peptide spectrum match (PSM) was considered to be correct. The false discovery rate (FDR) was determined based on PSMs when searched against the reverse, decoy database, and the protein FDR was set as 0.01. The “I = L,” “iBAQ,” and “match between runs” (0.7-min match and 20-min alignment time windows) options were enabled. Searches were performed against the UniProt database (4 September 2017).

Enzyme activity test.

An in vitro enzyme activity test was performed by using purified recombinant proteases, and the fluorescence change associated with the cleavage of a FRET-labeled substrate, NMA-I-E-A-L-F-Q-G-P-P-K(DNP)-F-R (where NMA is N-methylacetamide and DNP is 2,4-dinitrophenyl), was measured over time. The assays were performed with 100-μl samples containing 1 μM protease and 20 mM the substrate in assay buffer (50 mM HEPES, 100 mM NaCl [pH 7.5]). The reaction mixture was incubated at 30°C in Varioskan Flash (Thermo Fisher Scientific).

For the determination of K_m and V_max, the protease was assayed at a given concentration in the presence of various concentrations of the substrate ranging from 2 μM to 90 μM. The initial velocities of the enzymatic reactions from three independent assays were determined and fitted to the Michaelis-Menten equation with nonlinear regression analysis using GraphPad Prism.

The protease assay with protein substrates was carried out using 13 μM EV71 3CD C147A, 2 μM EV71 3C^pro, 20 μM rupintrivir, 4 μM ATG4B, and 40 μM NSC 185058 in a final reaction mixture volume of 20 μl in assay buffer. After incubation at 30°C for 8 h, the reactions were terminated, and the products were resolved using 12% SDS-PAGE.

Generation of stable knockdown cell lines.

shRNA recombinant lentiviruses (LV10-NC and LV-ATG4B) were produced by Suzhou GenePharma, and the virus titers were determined to be 7 × 10⁸ transducing units (TU)/ml. RD, HEK293T, HeLa, and Vero cells were infected at an MOI of 100. All knockdown cell lines were confirmed at 72 h postinfection by Western blot analysis. For the stable knockdown cell lines, the cells were incubated in selection medium containing 5 μg/ml puromycin (Solarbio) beginning 48 h after transduction, and ATG4B knockdown cells were stable after approximately 2 weeks. The following shRNAs were used in this study: NC (5′-TTCTCCGAACGTGTCACGT-3′), ATG4B-1 (5′-GGTGTGGACAGATGATCTTTG-3′), ATG4B-2 (5′-GCCCACTACTTCATCGGCTAC-3′), and ATG4B-3 (5′-GAAGCTTGCTGTCTTCGATAC-3′).

Time-of-addition assay.

RD cells were treated with rupintrivir at a concentration of 0.4 μM, pirodavir at a concentration of 5 μM, and NSC 185058 at a concentration of 5 μM (with DMSO as a negative control), either concurrently with 100 TCID₅₀s of EV71(FY)-Luc (0 h) or at intervals of −4, −2, 0, 2, 4, 6, 8, and 10 hpi. The antiviral activity was compared with the luciferase activity of the mock, which indicated that there are no inhibitors.

qRT-PCR analysis.

The cells were harvested after EV71 infection. Total cellular RNA was extracted and purified by using the EasyPure RNA kit (Transgen Biotech). cDNA is synthesized from isolated RNA by using the iScript cDNA synthesis kit (Bio-Rad). qRT-PCR was performed with iTaq universal SYBR green supermix (Bio-Rad), and the EV71, ATG4B, and EV71 transcript levels were determined by ΔΔC_T methods. The following primers were used for qRT-PCR: EV71 F (5′-TGAATGCGGCTAATCCCAACT-3′), EV71 R (5′-AAGAAACACGGACACCCAAAG-3′), ATG4B F (5′-GGTGTGGACAGATGATCTTTGC-3′), ATG4B R (5′-CCAACTCCCATTTGCGCTATC-3′), GAPDH F (5′-CCCACTCCTCCACCTTTGACG-3′), and GAPDH R (5′-CACCACCCTGTTGCTGTAGCCA-3′).