Crystal Structure of African Swine Fever Virus pS273R Protease and Implications for Inhibitor Design

African swine fever virus, a large and complex icosahedral DNA virus, causes a deadly infection in domestic pigs. In addition to Africa and Europe, countries in Asia, including China, Vietnam, and Mongolia, were negatively affected by the hazards posed by ASFV outbreaks in 2018 and 2019, at which time more than 30 million pigs were culled. Until now, there has been no vaccine for protection against ASFV infection or effective treatments to cure ASF. Here, we solved the high-resolution crystal structure of the ASFV pS273R protease. The pS273R protease has a two-domain structure that distinguishes it from other members of the SUMO protease family, while the unique “arm domain” has been proven to be essential for its hydrolytic activity. Moreover, the peptidomimetic aldehyde compounds designed to target the substrate binding pocket exert prominent inhibitory effects and can thus be used in a potential lead for anti-ASFV drug development.

ABSTRACT

African swine fever (ASF) is a highly contagious hemorrhagic viral disease of domestic and wild pigs that is responsible for serious economic and production losses. It is caused by the African swine fever virus (ASFV), a large and complex icosahedral DNA virus of the Asfarviridae family. Currently, there is no effective treatment or approved vaccine against the ASFV. pS273R, a specific SUMO-1 cysteine protease, catalyzes the maturation of the pp220 and pp62 polyprotein precursors into core-shell proteins. Here, we present the crystal structure of the ASFV pS273R protease at a resolution of 2.3 Å. The overall structure of the pS273R protease is represented by two domains named the “core domain” and the N-terminal “arm domain.” The “arm domain” contains the residues from M1 to N83, and the “core domain” contains the residues from N84 to A273. A structure analysis reveals that the “core domain” shares a high degree of structural similarity with chlamydial deubiquitinating enzyme, sentrin-specific protease, and adenovirus protease, while the “arm domain” is unique to ASFV. Further, experiments indicated that the “arm domain” plays an important role in maintaining the enzyme activity of ASFV pS273R. Moreover, based on the structural information of pS273R, we designed and synthesized several peptidomimetic aldehyde compounds at a submolar 50% inhibitory concentration, which paves the way for the design of inhibitors to target this severe pathogen.

IMPORTANCE African swine fever virus, a large and complex icosahedral DNA virus, causes a deadly infection in domestic pigs. In addition to Africa and Europe, countries in Asia, including China, Vietnam, and Mongolia, were negatively affected by the hazards posed by ASFV outbreaks in 2018 and 2019, at which time more than 30 million pigs were culled. Until now, there has been no vaccine for protection against ASFV infection or effective treatments to cure ASF. Here, we solved the high-resolution crystal structure of the ASFV pS273R protease. The pS273R protease has a two-domain structure that distinguishes it from other members of the SUMO protease family, while the unique “arm domain” has been proven to be essential for its hydrolytic activity. Moreover, the peptidomimetic aldehyde compounds designed to target the substrate binding pocket exert prominent inhibitory effects and can thus be used in a potential lead for anti-ASFV drug development.

INTRODUCTION

Africa swine fever virus (ASFV), a large enveloped double-stranded DNA virus, infects domestic and wild boars and can cause a wide range of clinical symptoms, including anorexia, moribundity, bloody diarrhea, vomiting, high fever, and hemorrhagic signs (1–3). ASFV is the only species of the genus Asfivirus in the Asfarviridae family (4). In the 1920s, an African swine fever (ASF) case was first reported by Wardley at al. and by Montgomery (5, 6), and in recent decades, it has spread from Africa to Europe and Asia, posing a serious risk for further expansion (7, 8). ASF is a highly fatal disease in pigs, and there is no treatment.

ASFV particles have a genome-containing nucleoid, a core shell, an inner lipid membrane, an icosahedral capsid, and an outer lipid envelope (9). This five-layer ASFV structure was further confirmed recently, by cryo-electron microscopy investigation on the whole virus particle (10–12). The extracellular enveloped virions are approximately 250 nm in diameter, and the genome varies between approximately 170 and 193 kbp and encodes between 150 and 167 proteins (13). Among these proteins, some are the structural proteins for virion particle assembly (9, 14–16), some are for genome replication (17–19) and some are critical for viral resistance to host immunity (20–22).

Currently, some studies indicate that ASFV’s main route of entry into porcine macrophages is via endocytosis, although the identity of the receptor is still uncertain (23). When ASFV is uncoated in the endosome and subsequently followed by viral core releasing, the genome is successfully released into the cytoplasm. At the perinuclear area near the microtubule organizing center, the ASFV genome begins to replicate and the genes expressed (24). Similar to positive-strand RNA viruses and retroviruses, ASFV also encodes polyproteins that are proteolytically cleaved by viral proteinases to yield the structural proteins required for virus morphogenesis (25, 26). Two polyprotein precursors, pp220 and pp62, are cleaved by the intrinsic pS273R protease to produce p5, p34, p14, p37, and p150 (derived from pp220) and p15, p35, and p8 (derived from pp62) (14, 27–29). This process plays a key role in ASFV particle maturation and infectivity (30).

The ASFV protease is encoded by the open reading frame S273R and is classified in the SUMO-1-specific protease family (27). The SUMO protease family contains Ubl (ubiquitin-like protein)-specific proteases (Ulp) in yeast and sentrin-specific proteases (SENP) in mammals (31, 32). These proteases mainly catalyze the deconjugation of SUMOylated proteins; however, certain SUMO proteases are also critical for SUMO precursor maturation and thus indirectly affect SUMO conjugation. These SUMO conjugation and deconjugation processes play important roles in nuclear transport, transcription replication, recombination and chromosome segregation (32, 33). To date, some SUMO proteases have been intensively studied structurally and enzymatically, including Ulp1 and Ulp2 from Saccharomyces cerevisiae (34, 35) and SENP1 (36), SENP2 (37), SENP7 (38), and SENP8 (39) from Homo sapiens. In general, these SUMO proteases contain at least two domains, in which the conserved core domain is responsible for the precise cleavage reaction between the terminal Gly of SUMO and the Lys of the substrate.

Similar to some animal viruses and plant proteases, including the poliovirus proteinase (40), adenovirus proteinase (41), and papain (42), the core domain of the SUMO protease contains conserved catalytic Cys and His residues. In addition, the active-site pocket orients the catalytic triad residues (His-Asp-Cys) for substrate cleavage and positions the SUMO C-terminal diglycine motif in a shallow tunnel that is formed by two Trp residues (32). Sequence analysis shows that there are several “Gly-Gly-Xaa” cleavage sites in two polyproteins, pp220 and pp62. Early studies showed that the mature products of pp220 and pp62 of ASFV are required for the formation of the viral core shell (14). Alejo et al. confirmed that repression of protease expression inhibits polyprotein processing and profoundly impairs infective virus production (30). Evidence from electron microscopic examination also shows that inhibition of proteolytic processing leads to the assembly of defective icosahedral particles containing a noncentered electron-dense nucleoid surrounded by an abnormal core shell of irregular thickness (30). Therefore, the pS273R protease is believed to be an attractive target for the design and development of effective inhibitors for use as ASFV treatments.

Here, we solved the three-dimensional structure of recombinant pS273R protease of ASFV. Subsequently, we synthesized three peptidomimetic aldehydes inhibitors ZW60, ZW30, and ZW22 based on our structural information. The result of biochemical activity test indicate that ZW60 inhibit pS273R enzymatic activity at a submolar 50% inhibitory concentration (IC₅₀) value. Our study could serve as the structural basis for further inhibitor optimization and development of potential drugs for anti-ASFV therapies.

RESULTS

Overall structure of ASFV pS273R protease.

Early research has proved that ASFV pS273R protease is a new SUMO-1-specific protease (27). The ASFV pS273R protease is a 273-amino-acid protein that cleaves the polyprotein precursor pp220 (2,475 amino acids) into the five mature structural proteins p5, p34, p14, p37, and p150 (Fig. 1A) (29, 43) and the polyprotein precursor pp62 (530 amino acids) into the three mature structural proteins p15, p35, and p8 (Fig. 1A) (28, 29).

FIG 1.

Process of ASFV core-shell structure protein maturation and the overall structure of the ASFV pS273R protease. (A) Schematic diagram of pS273R protease cleaving the polyprotein pp220 and pp62 to generate mature protein structures. The black downward-pointing arrow represents the cleavage sites, and the number below the diagram represents the residue sequence numbers of the structural proteins. (B) Overall structure of the pS273R protease. The “arm domain” is lime colored, and the “core domain” is salmon colored. The active site is enlarged in the bottom panel. The catalytic “C232-H168-N187” triad is shown as a stick model with the hydrogen bonds as dashed black lines. The length of the hydrogen bonds is labeled.

In this study, the full-length ASFV pS273R was expressed in Escherichia coli with a hexahistidine tag at the C terminus. During the crystallization process, we obtained two crystal forms: one is the irregular laminar crystal cultured in 0.1 M Tris (pH 8.6) and 19% (wt/vol) polyethylene glycol 3350 (PEG3350), and the other is a cube crystal cultured in 0.2 M ammonium acetate, 0.1 M Tris (pH 8.5), 0.6 mM coenzyme A (CoA), and 20% PEG3350. The former is crystallized in the P2₁2₁2₁ space group with two independent molecules in the asymmetric unit, and the latter is crystallized in the C2 space group, which contains one molecule in the crystallographic asymmetric unit. The crystal structure was determined by using the selenium single-wavelength anomalous dispersion (Se-SAD) method. The former model was refined to 2.3 Å with R_work and R_free values of 19.1 and 22.0%, respectively, and the latter is refined to 2.5 Å with R_work and R_free values of 22.30 and 25.2%, respectively (Table 1). Despite the higher resolution of the former crystal, residues M114 to R119 and D160 to G166 of chain B in the former structure could not be built due to the lack of interpretable electron density data, whereas the loop regions M114 to R119 and D160 to G166 could be traced unambiguously in the latter structure. Besides this difference, the two independent molecules in the former one shared high structural similarity with the latter one such that it could be superimposed within a 0.917-Å root-mean-square deviation (RMSD) over 255 C_α atoms or 0.723-Å RMSD over 248 C_α atoms, respectively. For clarity, we refer to the latter model throughout the remainder of this study.

TABLE 1.

Data collection and refinement statistics^a

Parameter	SeMet-pS273R	Native-pS273R
PDB accession no.	6LJ9	6LJB
X-ray source	SPring-8 BL41XU	SSRF BL19U1
Wavelength (Å)	0.97911	1.0083
Space group	P212121	C2
Unit cell parameters (Å; °)	a = 60.2, b = 83.4, c = 108.8; α = β = γ = 90.0	a = 220.8, b = 38.0, c = 33.0; α = γ = 90.0, β = 93.4
Resolution range (Å)	50.00–2.30 (2.34–2.30)	50.00–2.50 (2.54–2.50)
No. of unique reflections	23,913 (974)	9,380 (418)
Completeness (%)	96.8 (80.0)	95.2 (79.9)
Redundancy	11.6 (7.7)	5.5 (3.3)
I/σ(I)	21.1 (3.5)	11.7 (2.4)
Rmerge (%)	12.1 (64.6)	13.1 (34.8)
CC1/2	0.99 (0.91)	0.99 (0.91)
Refinement statistics
Resolution range (Å)	48.52–2.30 (2.39–2.30)	32.97–2.49 (2.58–2.50)
No. of reflections used in refinement	21,865 (1,507)	8719 (440)
No. of reflections used for Rfree	1,093 (72)	467 (16)
Rwork (%)	19.1 (24.2)	22.3 (28.2)
Rfree (%)	22.0 (28.8)	25.2 (28.4)
No. of nonhydrogen atoms	4,429	2,255
Protein	4,321	2,177
Solvent	108	78
Avg B-factors	47.3	39.4
Protein	47.5	39.5
Solvent	41.8	36.6
RMSD
Bond length (Å)	0.003	0.005
Bond angle (°)	0.88	1.04
MolProbity clash score	5.2	19.1
Ramachandran (%)
Favored	96.9	95.4
Allowed	3.1	4.6
Outliers	0	0

^aNumbers in parentheses indicate results for the highest-resolution shell. R_merge = Σ_hΣ_l|I_ih−〈I_h〉|/Σ_hΣ_I〈I_h〉, where〈I_h〉is the mean of the observations I_ih of reflection h. R_work = Σ (ǁF_p(obs)|−|F_p(calc)ǁ)/Σ|F_p(obs)|; R_free is an R factor for a preselected subset (5%) of reflections that was not included in refinement.

As revealed in this structure analysis, the ASFV pS273R protease comprises twelve α-helices, seven β-strands, and one 3₁₀-helix, which fold into two closely but distinct domains (Fig. 1B and Fig. 2A): a “core domain” resembling a SUMO-protease fold (34) and an “arm domain” formed by approximately the first 83 amino acids. The “core domain” consists of a sandwich-like fold structure, in which the β1 and β4 to β7 strands are sandwiched between α7, α8, and α9 from one side and α6, α10, and α11 on the other side (Fig. 1B and Fig. 2A). The β2 and β3 strands are situated close to the α6 helix, which are variable regions compared to the previously determined SENP2-SUMO2 complex (37) and Rickettsia CE protease (44). The “arm domain” consists of five α-helices, containing α1 to α5, and one 3₁₀ helix (Fig. 1B). The catalytic triad, composed of H168, N187 and C232, is situated at one end of the central β-strands in the “core domain.” C232 is situated at the end of α10, and H168 and N187 are situated at the end of the β5 and β6 strands, respectively. In addition, H168 bridges C232 and N187 by hydrogen bonds (Fig. 1B and Fig. 2A).

FIG 2.

Sequence alignment and structure superimposition of the ASFV pS273R protease to other cysteine proteases. (A) SENP1 (Homo sapiens sentrin-specific protease 1), SENP2 (Homo sapiens sentrin-specific protease 2), and ULP1 (Saccharomyces cerevisiae ubiquitin-like-specific protease 1). Strictly and relatively conserved residues are highlighted in red and yellow boxes, respectively. The secondary structural elements of ASFV pS273R are shown above the alignment. The residues of the catalytic triad are marked with upright green triangles. (B) Superimposition of ASFV pS273R with ChlaDUB1-Ulp1 (ChlaDUB1, chlamydial deubiquitinase 1), SENP2-SUMO-1, and adenain (adenovirus protease). The color of ASFV pS273R is the same as in Fig. 1B. All structures are shown as cartoons. ChlaDUB1, slate; Ulp1, red; SENP2, dark gray; SUMO-1, magenta; adenain, cyan. The enlarged active sites are displayed at the bottom left corner of every small panel.

Structural similarity of ASFV protease with other cysteine protease members.

Previous research stated the ASFV pS273R protease is a new viral member of the SUMO-1-specific protease family mainly on the basis of primary sequence prediction and a follow-up enzymatic assay (27). Here, we further confirm their previous study by providing more structural information. The traditional SUMO proteases contain at least two domains, a conserved C-terminal protease fold and an N-terminal domain (31, 34, 36, 37). Our sequence alignment similarly indicates that only the “core domain” (located at the C terminus) shares a high degree of sequence similarity with the conserved C termini of SENP1, SENP2, and Ulp1 (Fig. 2A).

In a PDB search using the DALI server (45), ASFV pS273R shows the highest degree of structural similarity with several cysteine proteases, especially the core domain. We can classify these proteins into three groups: mammalian SUMO-proteases (31), Chlamydia and Rickettsia bacteria effector proteases (44, 46), and viral cysteine proteases (41). The results of the superimposition of the pS273R with three representative cysteine proteases (Chlamydia effector protease, SENP2-SUMO-1 complex, and adenoviral protease) show that only the “core domain” shares a high degree of structural similarity (Fig. 2B). This superimposed result also revealed that the “arm domain” of pS273R has a novel structure folding pattern.

Pruneda et al. identified and explained the remarkable dual Lys63-deubiquitinase (DUB) and Lys-acetyltransferase activities in the Chlamydia effector ChlaDUB1 (46). Moreover, these two enzymes seemingly use the same catalytic cysteine residue. The structural alignment shows that the catalytic triad (C232-H168-N187 in pS273R and C345-H275-D292 in ChlaDUB1) are basically in the same spatial location (Fig. 2B, upper panel). This catalytic triad in ChlaDUB1 maintains Chlamydia deubiquitination activity to interfere with host defense, which appears to have a substantial effect in mammalian infections (46, 47). Similarly, another protein, SENP2, which shares a high degree of structural similarity to the crystal structure of ASFV pS273R (DALI Z score of 12.6 and RMSD of 3.5 Å), contains the same catalytic triad (C548-H478-D495) (Fig. 2B, middle panel). This active site plays a key role in SUMO maturation and deconjugation, and the function of SENP2 (and other SENP proteases) contributes to the regulation of cellular pathways involved in apoptosis, differentiation, development, stress response, and the cell cycle (37). It is worth noting that adenovirus also encodes a protease essential for the development of infectivity. ASFV pS273R shares a high degree of structural similarity to the adenoviral protease (DALI Z score of 9.9 and an RMSD of 3.0 Å, Fig. 2B, bottom panel) (41). The catalytic triad (C122-H54-E71) shares similarity with the ASFV pS273R protease (Fig. 2B, bottom panel), and the conserved active-site residues are required for processing several capsid and core precursor proteins. More importantly, some inhibitors have recently been designed for adenovirus protease (48, 49), and they can be used as references for the development of an anti-ASFV drug design based on the pS273R protease structure.

The “arm domain” of ASFV protease is a novel fold structure.

As discussed above, the “arm domain” is a novel fold structure with no structural homology found through a DALI search. We carefully analyzed the structure of the pS273R and carried out experiments using truncated mutants, ThermoFluor assays, and enzyme activity tests to further clarify its function.

First, many hydrophobic or electrostatic interactions are established between the “arm domain” and the “core domain” (Fig. 3A): residues L4 and I7 of the “arm domain” form a hydrophobic region with the aromatic residues F238 and F265 from the “core” domain. In addition, in the vicinity of this area, three groups of hydrogen-bond interactions are also found. The first hydrogen bond group is formed between residue S9 in the “arm domain” and H268 in the “core domain.” The second hydrogen-bond interaction is formed by the main chain comprising residues H268 and S10 and the side chain consisting of residue S12 with the basic residue K77; the main chain consisting of residues P11, T267, and F270 interacts with residue R82 to form the last group hydrogen-bond interaction. In general, these interactions facilitate the flexible N-arm next to the α12 helix in the “core domain.”

FIG 3.

Characterization of the protease properties. (A) Interactions between the “arm domain” and the “core domain.” The residues that participate in the interactions are shown as sticks and are colored according to the domains. The hydrogen bonds are shown as dashed black lines. (B) Enzymatic activity assay. RFU, relative fluorescence units. (C) Results of the thermal stability test of ASFV pS273R. The T_m was identified by plotting the first derivative of the fluorescence emission value as a function of temperature (−dF/dT). The T_m is represented by the lowest part of the curve.

Second, we used a fluorescence resonance energy transfer (FRET)-based system to determine pS273R enzyme activity. Compared to the wild-type enzyme, a mutant enzyme generated by any single mutation in the catalytic triad (C232-H168-N187) to Ala results the total loss of pS273R protease enzyme activity (Fig. 3B), a result that basically verifies the critical role of the catalytic triad (27). Moreover, previous research has shown that a 45 to 65 amino acid Ulp1 domain in the N-terminal immediately adjacent to the conserved protease domain (domain 403-621) modulates the activity of Ulp1 (34). Thus, we generated three truncated mutants (ΔN1-20, in which the first 20 residues of the “arm domain” are deleted; ΔN1-83, in which the whole “arm domain” is deleted; and ΔC256-273, in which the last 18 residues in the “core domain,” which interact with the ΔN1-20, are deleted) to explore the function of the unique “arm domain” of the ASFV pS273R protease. All these results show that, although three truncated mutants still maintain structural stability, they totally lose enzyme activity (Fig. 3B and C).

Inhibitor design based on crystal structure information.

The outbreak of ASF in some Asian countries has caused tremendous losses to the pig breeding industry and poses a tremendous threat to food safety. Currently, there is no effective vaccine or drug for the prevention or treatment of ASF. Due to its critical role for virus reproduction, protease is an attractive target for antiviral drug development.

To develop antivirals for the treatment of ASFV, we designed peptidomimetic inhibitors based on the structural information of active site and the characteristics of substrate cleavage sequences. It is worth noting that the active pocket of ASFV pS273R protease is a shallow tunnel with the C232-H168-N187 triad inside and the hydrophobic residues F111, M113, and the hydrogen bond donor residue Y121 outside. We designed ZW60 by maintaining Gly at P1 and P2 position, residues derived from the substrate at the P3 position, followed by cinnamic acid, a common component for drug synthesis with a large π-system. In addition, an aldehyde moiety, was incorporated at P1′ position to form a covalent bond with the active site residue C232 to significantly increase the binding affinity. In comparison, we also test two inhibitors ZW22 and ZW30, originally designed for SARS-CoV main protease, as the control. These two compounds contain the same aldehyde moiety as the “warhead,” while the moieties of P1, P2, and P3 were replaced with a lactam ring to mimic Gln, a Leu, and a Phe or a 2-aminobutyric acid (Abu), respectively. The molecular docking results show that ZW60 interacts with pS273R mainly through three potent interactions: (i) a reversible covalent bond between C232 and the aldehyde group of the P1′ subunit; (ii) hydrogen bond interactions with L87, N89, N158, H168, W169, and Y121; and (iii) a hydrophobic interaction stabilized between the P4 subunit and the F111 and M113 residues (Fig. 4A). In contrast, with the relative bulky side chain incorporated at P1 and P2 positions in ZW30 and ZW22, the hydrogen bond interactions between carboxyl oxygen with the residues W169 and L87 will be significantly decreased. The relatively better IC₅₀ values for ZW60 (IC₅₀ = 0.900 ± 0.037 μM) comparing to ZW22 (IC₅₀ = 3.440 ± 0.144 μM) and ZW30 (IC₅₀ = 2.605 ± 0.145 μM) by enzymatic assay further confirmed the importance of a small moiety at the P1 and P2 positions (Fig. 4B).

FIG 4.

ASFV pS273R inhibition assay. (A) Molecular docking between pS273R and the small molecule inhibitors ZW60, ZW30, and ZW22. The residues and compounds are shown as stick models. The hydrogen bonds are shown as dashed black lines. The pseudoatom ligand center is shown as a green sphere. (B) Inhibitory effect of different small molecule inhibitors. IC₅₀, the half-maximal inhibitory concentration. The structural formula of the small molecule inhibitor is shown at the bottom of the panel. The P1, P2, P3, and P4 sites are marked and colored differently. The P1′ sites are indicated by dashed dark blue circles.

DISCUSSION

Many viruses use a replication strategy involving the translation of a large polyprotein or precursor polypeptide that is cleaved by viral and/or cellular proteases. This method of genome organization has many benefits to the virus, such as condensation of genetic material and temporal and spatial regulation of protein activity that depends on the polyprotein cleavage state (50). RNA viruses, including poliovirus (51), flavivirus (52), and coronavirus, adopt this strategy (53). ASFV, a DNA virus, expresses two polyprotein precursors, pp220 and pp62, which are processed by the pS273R protease. A previous study showed that the ASFV pS273R is a new viral member of the SUMO-1-specific protease family (27). All six proteolytic cleavage sites in pp220 and pp62 are downstream of the second Gly (the P1 site) of the consensus sequence Gly-Gly-Xaa, which is also recognized as a cleavage site critical to the maturation of adenovirus structural proteins (54). In addition, for the vaccinia virus, a similar cleavage site, Ala-Gly-Xaa, is crucial for the maturation of structural proteins (55).

Here, we present the high-resolution crystal structure of ASFV pS273R. The pS273R protease contains two distinct domains named the “arm domain” and “core domain.” The “core domain” is a sandwich-like structure, in which the three central β-strands are sandwiched by seven α-helixes in two groups. This “core domain” shares similar structure fold with other members of the SUMO family. At the active site of the ASFV, the pS273R protease has a catalytic triad of C232-H168-N187 that resembles that of other cysteine protease catalytic triads (Cys-His-Asp/Glu) (34, 41, 44, 56). This active site usually catalyzes the ubiquitin or a ubiquitin-like protein (such as SUMO, smt3, NEDD8, and ISG15 [IFN-stimulated gene 15]) during maturation or deconjugation activities (31, 34, 36, 37, 39, 57); is involved in the proteolysis of a polyprotein that leads to mature structural proteins in some viruses, including ASFV, poliovirus, flavivirus, and coronavirus (29, 41, 51–53, 58); or interrupts eukaryotic host response processes by bacterial effector proteases (44). It is worth noting that ISG15 molecule is a ubiquitin homolog that is rapidly upregulated after viral infection. Recently, some studies found that the virus protease has the additional function of removing ubiquitin and ISG15 from host cell proteins during their evasion of downregulate host innate immune responses (59). Therefore, it is possible that the ASFV pS273R protease may antagonize the host IFN-1 pathway by deubiquitinating specific proteins.

The traditional SUMO protease family includes Ubl (ubiquitin-like protein)-specific proteases (Ulp) in yeast and sentrin-specific proteases (SENP) in mammals, which have a large peptide length, ranging from 573 amino acids in the case of SENP3 (also known as SSP3 or SMT3IP2) to 1,112 amino acids in the case of SENP6 (also known as SUSP1 or SSP1) (31). However, the conserved catalytic domain is typically located close to the C termini of Ulp/SENPs, whereas N-terminal domains frequently direct subcellular localization. In fact, for the ASFV pS273R protease, as well as the substrate structural proteins pp220 and pp62, all are located at the core shell in the virion (30). The virus assembly process strictly depends on the correct spatial and temporal maturation of the pp220 and pp62 polyproteins. We assume that the “arm domain” plays an important role in the recruitment of its substrate polyprotein or even some unknown host protein. A DALI search indicated that the “arm domain” shares no similarity to any structure fold; however, the precise functional role of this “arm domain” remains elusive. Here, we confirmed that the “arm domain” is not required for the stability of the whole protein but is necessary for the enzyme catalytic activity. Previous studies showed that a 45 to 65 amino acid Ulp1 domain immediately adjacent to the conserved protease domain (i.e., domain 403-621) in the N-terminal modulates the activities of Ulp1 (34). In addition, to be activated, the adenovirus protease also needs its 11-amino-acid peptide as a cofactor (54, 60). Therefore, further work is expected to be carried out to elucidate the actual function of the “arm domain.”

Finally, our structure is expected to facilitate structure-based antiviral design against ASFV. Since 2018, the ASFV has been spread to China, Vietnam, Mongolia, and Korea, with more than 30 million pigs culled. Until now, there has been no efficient treatment available, and the development of a potent antiviral drug is an urgent task. The ASFV protease pS273R, a cysteine protease, plays a crucial role in the viral life cycle and thus represents an attractive therapeutic target. Previous studies showed that aldehyde derivatives of the protease substrates are potent inhibitors of cysteinyl proteases (61, 62). Altmann’s research group developed a series of tetrapeptide nitrile compounds against adenain (48, 49). These compounds mimic the consensus substrate cleavage sites and show excellent biochemical activity, indicating that the small moiety at the P1 and P2 sites may be essential for potent inhibitor design. In our study, we synthesized three peptidomimetic aldehydes and evaluated their biochemical activity against the ASFV pS273R protease. The biochemical activity assay shows that the small moiety at the P1 and P2 sites in ZW60 significantly facilitate its binding with active site. We anticipate that this work will pave the way for the development of inhibitors to target this severe pathogen.

MATERIALS AND METHODS

Protein production.

The codon-optimized wild-type cDNA of ASFV pS273R protease (UniProtKB P0C9B9) was synthesized by GENEWIZ. The full-length pS273R protease was cloned into pET-22b (Novagen) vector with NdeI and XhoI restriction sites using the cloning primers. The sequences of the primers were: 5′-TACATATGAGTATTCTGGAAAAGATCAC-3′ (forward) and 5′-GTGCTCGAGCGCGATACGGAACAGATG-3′ (reverse). The accuracy of the inserts was verified by sequencing.

The recombinant plasmid of ASFV pS273R protease was transformed into Escherichia coli strain BL21(DE3) (TransGen Biotech, Beijing, China) and overexpressed. The cells were cultured at 37°C in 800 ml of Luria-Bertani medium containing 100 μg/ml ampicillin. Once the optical density at 600 nm (OD₆₀₀) reached 0.5 to 0.6, the culture was transferred to 16°C. Protein expression was then induced by adding 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for an additional 16 to 18 h. The cells were harvested and resuspended in lysis buffer (20 mM Tris [pH 8.0], 300 mM NaCl) and homogenized with a low-temperature ultrahigh pressure cell disrupter (JNBIO). The lysate was centrifuged at 25,000 × g for 30 min at 4°C to remove cell debris. The supernatant was then loaded twice onto a Ni-NTA column preequilibrated with lysis buffer. Resin was washed seven to eight times with 100 ml of wash buffer (20 mM Tris, 300 mM NaCl, 20 mM imidazole [pH 8.0]), and the target protein was eluted with elution buffer containing 20 mM Tris, 300 mM NaCl, and 500 mM imidazole (pH 8.0). The protein was further purified on a Superdex 200 (GE Healthcare) column equilibrated with 20 mM Tris-HCl and 150 mM NaCl (pH 8.0). SDS-PAGE analysis revealed >95% purity of the final purified recombinant protein. Fractions from the single major peak were pooled and concentrated to 10 mg/ml for crystallization.

The selenomethionine (SeMet) derivative were prepared as previously described (63). In brief, SeMet derivative of ASFV pS273R protease was produced in the methionine-auxotrophic E. coli strain B834(DE3) that was grown in minimal medium supplemented with 3% glucose, 30 mg/liter l-selenomethionine, and 100 μg/ml ampicillin. When OD₆₀₀ reached 0.6, the culture was transferred to 16°C, and another 30 mg/liter l-selenomethionine was added. Protein was induced by incubating with 0.5 mmol/liter IPTG for an additional 18 h. SeMet substituted ASFV pS273R protease was purified using the same conditions as described for the wild-type protein.

Crystallization.

Initial crystallization conditions were screened by the sitting-drop vapor-diffusion method using commercial crystal screening kits at 16°C, including the index, crystal screen, PEG/ion, and SaltR_X from Hampton Research and Wizard I to IV from Emerald BioSystems. The protein and reservoir solutions were mixed in a ratio of 1:1, and all conditions were equilibrated against 100 μl of reservoir solution in a 48-well format.

The laminar crystals of pS273R protease appeared after 1 week in 0.2 M lithium sulfate monohydrate, 0.1 M Tris (pH 8.5), and 25% (wt/vol) PEG3350. Larger crystals were obtained by sitting-drop vapor diffusion in a 48-well format using a 1:1 ratio of well solution to protein at 4 mg/ml in 0.2 M lithium sulfate monohydrate, 0.1 M Tris (pH 8.6), and 19% (wt/vol) PEG3350. Further optimization with additive and detergent screens (Hampton Research) and using the hanging-drop vapor-diffusion method was performed, and the final optimized crystals were grown in 0.2 M lithium sulfate monohydrate, 0.1 M Tris (pH 8.2), and 23% (wt/vol) PEG1500 by the sitting-drop vapor-diffusion method. The SeMet derivative crystals were also cultured under this condition.

The cube crystals of pS273R protease appeared after 5 days in 0.2 M ammonium acetate, 0.1 M Tris (pH 8.5), and 25% PEG3350. After the same optimization procedure, the final optimized crystals were grown in 0.6 mM CoA, 0.2 M ammonium acetate, 0.1 M Tris (pH 8.5), and 20% PEG3350. Crystals were transferred to 4 M sodium formate solution, and the regular piece crystal was desquamated using a needle. Crystals were cryoprotected in 4 M sodium formate solution and cooled in a dry nitrogen stream at 100 K for X-ray data collection.

X-ray data collection, processing, and structure determination.

The selenomethionine SAD data were collected at beamline BL41XU (Spring8, Japan) and the native data were collected at beamlines BL17U and BL19U (SSRF, China) under cryogenic conditions at 100 K. All data sets were processed using the HKL3000 package (64). Excluding the first selenium of each polypeptide, five of seven selenium atoms in the asymmetric unit were located and refined, and the SAD data phases were calculated and substantially improved by solvent flattening using the PHENIX program (65). A model was manually built into the modified experimental electron density using COOT (66) further refined in PHENIX. Model geometry was verified using the program MolProbity (67). Structural figures were drawn using the program PyMOL (http://www.pymol.org).

Mutant protein production.

Active site His168, Gln187, and Cys232 were mutated to Ala to measure the pS273R mutant protein protease enzyme activity. Similarly, in order to explore the function of the “arm domain,” three truncation mutants, ΔN1-20, ΔN1-83, and ΔC256-273, were produced. Complementary mutant primers were designed to introduce alanine at the targeted site. The sequences of the primers were as follows: H168A, 5′-AGCACCGGCACGGGCAAAGCCTGGGTTGCGATC-3′ (forward) and 5′-GGCTTTGCCCGTGCCGGTGCTGAAATCGGTGTT-3′ (reverse); N187A, 5′-TGGAGCATCGAGTACTTCGCCAGCGCCGGTAAT-3′ (forward) and 5′-GGCGAAGTACTCGATGCTCCAGCAATCGCCGCG-3′ (reverse); C232A, 5′-CAGCGCAGCCAGACGGAGCCGGTCCGTACAGT-3′ (forward) and 5′-GGCCTCCGTCTGGCTGCGCTGATGACGGATGTT-3′ (reverse); ΔN1-20, 5′-TACATATGGACAGCTGCCTCAGCAAAAAGAT-3′ (forward) and 5′-GTGCTCGAGCGCGATACGGAACAGATG-3′ (reverse); ΔN1-83, 5′-TACATATGAACACCGGTCTGCTGACGAACT-3′ (forward) and 5′-GTGCTCGAGCGCGATACGGAACAGATG-3′ (reverse); and ΔC256-273, 5′-TACATATGAGTATTCTGGAAAAGATCAC-3′ (forward) and 5′-GTGCTCGGTGCTGATGAAGTGCGTGT-3′ (reverse) (Tsingke, China). PCRs were carried out using pET22b-ASFV pS273R wild-type plasmid as the template. The site-directed mutations of ASFV pS273R protease were generated by using a Fast mutagenesis system (TransGen Biotech, China). The sequences of all the mutation constructs were verified by DNA sequencing (Tsingke). Recombinant pS273R mutations protein were expressed and purified as wild-type proteins.

FRET-based in vitro ASFV pS273R assay.

The protease activity of ASFV pS273R was measured by a FRET-based assay as previously described (68), using the commercially synthesized fluorogenic peptide (MCA)GYFNGG↓GDK(DNP)NP that corresponding to substrate peptide of ASFV pS273R protease with methoxy-coumarin-acetic-acid at the N terminus and 2,4-dinitrophenyl conjuncted with Lys residue at the C terminus (GL Biochem, Shanghai, China). This substrate peptide consists of eleven amino acids derived from the cleavage site present between structural proteins p5 and p34 derived from pp220 precursor. The numbering was assigned to the substrate residues in which the P1 and P1′ positions include the residues at the scissile bond (69).

The enzymatic reactions of pS273R were performed in 96-well black microplate with a 100-μl reaction volume in 20 mM Tris-HCl–200 mM NaCl (pH 8.0) as the assay buffer at 25°C. Reactions were initiated by the addition of 1 μM enzyme from a concentrated stock. The fluorescence was monitored at 393 nm with an excitation wavelength of 328 nm using the multimode plate reader Varioskan Flash (Thermo Scientific). All reactions were conducted in triplicate.

ASFV pS273R inhibition assay.

For the assessment of pS273R inhibition, three inhibitors were designed based on our crystal structure information. All three inhibitors are peptidomimetic compounds, with the aldehyde group moiety as the warhead. Based on the substrate sequence, the inhibitor ZW60 was designed to contain Gly at the P1 and P2 positions, Val at P3 followed by cinnamic acid, and the aldehyde warhead at P1′ to form a covalent bond with the active-site Cys residue. The inhibitor ZW30 contains a lactam ring at the P1 position, Leu at the P2 position, Phe at the P3 position, and the other parts remaining the same as with ZW60. The inhibitor ZW22 is the same with ZW30 in construction, except for the 2-aminobutyric acid (Abu) at the P3 position. The final required concentrations of inhibitors were made by dilution in the assay reaction buffer. The concentrations of dimethyl sulfoxide in the final reactions did not exceed 0.5% (vol/vol).

ASFV pS273R protease inhibition assay was performed in vitro using purified pS273R preincubated with different inhibitors concentrations (from 0.0244 to 100 μM) at 37°C for 30 min. We then added fluorogenic peptide to the reaction well, followed by incubation at 37°C for 60 min. The progress of enzyme inhibition was monitored by measuring the decrease in the fluorescence. All enzyme reactions with inhibitors were performed in triplicate in 96-well black plates. Nonlinear regression analysis was performed using GraphPad Prism 7.0 to calculate the IC₅₀ value.

Molecular docking.

Molecular docking was performed using the AutoDock Vina program (70) to investigate the detailed interactions between our compounds and ASFV pS273R. For the predocking treatment, all small molecules were optimized to their local energy minima. Due to the similarity active site between the adenovirus protease and ASFV protease, we consulted the adenovirus protease with the inhibitor complex as a model.

Data accessibility.

Atomic coordinates and structure factors for ASFV pS273R protease has been deposited in the Protein Data Bank under accession number 6LJ9 (SeMet derivative, in space group P212121) and 6LJB (native protein, in space group C2), respectively.