Structural and functional studies of PCNA from African swine fever virus

ABSTRACT Proliferating cell nuclear antigen (PCNA) belongs to the DNA sliding clamp family. Via interacting with various partner proteins, PCNA plays critical roles in DNA replication, DNA repair, chromatin assembly, epigenetic inheritance, chromatin remodeling, and many other fundamental biological processes. Although PCNA and PCNA-interacting partner networks are conserved across species, PCNA of a given species is rarely functional in heterologous systems, emphasizing the importance of more representative PCNA studies. Here, we report two crystal structures of PCNA from African swine fever virus (ASFV), which is the only member of the Asfarviridae family. Compared to the eukaryotic and archaeal PCNAs and the sliding clamp structural homologs from other viruses, AsfvPCNA possesses unique sequences and/or conformations at several regions, such as the J-loop, interdomain-connecting loop (IDCL), P-loop, and C-tail, which are involved in partner recognition or modification of sliding clamps. In addition to double-stranded DNA binding, we also demonstrate that AsfvPCNA can modestly enhance the ligation activity of the AsfvLIG protein. The unique structural features of AsfvPCNA can serve as a potential target for the development of ASFV-specific inhibitors and help combat the deadly virus. IMPORTANCE Two high-resolution crystal structures of African swine fever virus proliferating cell nuclear antigen (AsfvPCNA) are presented here. Structural comparison revealed that AsfvPCNA is unique at several regions, such as the J-loop, the interdomain-connecting loop linker, and the P-loop, which may play important roles in ASFV-specific partner selection of AsfvPCNA. Unlike eukaryotic and archaeal PCNAs, AsfvPCNA possesses high double-stranded DNA-binding affinity. Besides DNA binding, AsfvPCNA can also modestly enhance the ligation activity of the AsfvLIG protein, which is essential for the replication and repair of ASFV genome. The unique structural features make AsfvPCNA a potential target for drug development, which will help combat the deadly virus.

the largest pork producer in the world, caused huge economic losses and an immediate pork shortage (38). Although it has been extensively studied for over 100 years, safe and effective vaccines against ASFV are still very limited (39). Instead, ASFV has become a global threat to the agricultural industry in recent years.
ASFV is a large double-stranded DNA (dsDNA) virus and is one of the most complex viruses known to date. The size of the ASFV genome varies between 170 and 190 kb, encoding more than 160 proteins that are involved in entry into host cells, suppression of host immune response (40), DNA replication (41), DNA damage response (42), DNA repair (43), and gene expression and virion assembly of ASFV (44). During replication, small viral DNA fragments are synthesized in the nucleus of the infected cells at early times, while large fragments are synthesized in the cytoplasm at later times. Both small and large DNA fragments are precursors of mature head-to-head cross-linked viral DNAs. Through site-specific nicking, rearrangement, and ligation, the cross-linked DNAs are resolved and form the genomic DNA of ASFV (45,46). ASFV belongs to the Asfivirus genus and is the only member of the Asfarviridae family. Although the reported ASFV protein structures are still very limited, previous studies showed that the structures of ASFV proteins could be very different from their homologous proteins from other species (47)(48)(49). Here, we report the structural and biochemical studies of the AsfvPCNA protein. The crystal structures were determined at atomic resolution and confirmed that AsfvPCNA can assemble into a homotrimeric ring-shaped structure. Different from homologous protein structures, AsfvPCNA has a very unique amino acid sequence and conformation at the IDCL, J-loop, and P-loop regions, which are involved in partner recognition by PCNAs. Although the detailed molecular basis remains to be investigated, AsfvPCNA has a high dsDNA-binding affinity and can enhance the catalytic activity of the AsfvLIG protein, which plays critical roles in both DNA replication and DNA repair pathways of ASFV. Our study not only advances our understanding on the function of AsfvPCNA but also provides a potential target for anti-ASFV drug development.

Crystal structure of AsfvPCNA
AsfvPCNA is a hypothetic sliding clamp family protein; it is coded by the E301R gene that is highly conserved in ASFV strains. The high conservation suggests that AsfvPCNA may play certain functional role in ASFV; however, AsfvPCNA shares no clear sequence similarity with other sliding clamp family members, and the detailed function of AsfvPCNA has not been experimentally verified. To investigate the potential function of AsfvPCNA, we constructed and expressed one His-Sumo-tagged AsfvPCNA protein.
After the His-Trap column purification, the tag was removed and the target protein was further purified by size-exclusion S200 16/600 column (Fig. S1A). Two protein peaks were observed: the peak 1 was eluted at the void volume of the column and the peak 2 was eluted at a volume of 78.2 mL, corresponding to the molecular weight of an AsfvPCNA homotrimer. The ratio between OD280 and OD260 is 1.75, suggested that the purified AsfvPCNA protein is free of nucleic acids. The purity of AsfvPCNA was confirmed by the SDS-PAGE gel analysis (Fig. S1B).
Using the purified AsfvPCNA protein, we performed extensive crystallization trials and solved the structures in two different forms, Form I and Form II (Table 1). Form I structure was refined at higher resolution (2.2 Å); therefore, it was used for detailed structural analysis. Form I crystal belongs to P6 3 space group, per asymmetric unit contains one AsfvPCNA molecule (Fig. 1A). Out of the total 301 residues of AsfvPCNA, residues 4-15 at the N-terminus and residues 298-301 at the very C-terminus are disordered. The remaining residues form 8 α-helices and 18 β-strands (Fig. S2), assembled into two domains, Domain I and Domain II ( Fig. 1A and B). Domain I (residues 1-155) is composed of α1-3 and β1-9, and Domain II (residues 181-301) consisted of α6-8 and β10-18. The overall folding of the two domains is similar. All the β-strands are antiparallel (Fig. 1A) and reside in the middle of the structure; the six α-helices of Domain I and Domain II locate on the same side of the β-strands. Similar to all reported PCNA structures, Domain I and Domain II of AsfvPCNA are connected by an IDCL linker ( Fig. 1A and B). AsfvPCNA IDCL is composed of 25 residues, residues 156-180. Although the side chains of some AsfvPCNA IDCL residues are flexible, the main chains of the IDCL linker are well ordered in Form I structure, supported by the clear 2F o −F c electron density maps (Fig. S3A). In addition to the loop structures at its two ends, AsfvPCNA IDCL also contains two short α-helices, α4 and α5, which are formed by residues 164-171 and 174-177, respectively.
Unlike the Form I structure, the Form II AsfvPCNA structure belongs to P1 space group (Table 1); per asymmetric unit contains six AsfvPCNA molecules, which assemble into two homotrimeric ring structures (Fig. S4). Via symmetric operation, Form I AsfvPCNA structure can also form homotrimer (Fig. 1C). The overall conformations of Form I and Form II AsfvPCNA trimers are virtually identical, supported by the very low root mean square deviation (RMSD, 0.2 Å) value over 840 pairs of Cα atoms. The three AsfvPCNA protomers arrange in a head-to-tail fashion, embracing all Domain I and Domain II αhelices at the inner face of the ring. The diameter of the inner ring is approximately 30 Å.
Trimerization of AsfvPCNA is mediated by various types of interactions. As depicted in Fig. 1D, the two strands β9 and β13 from the neighboring AsfvPCNA protomers reside next to each other, forming seven hydrogen bond (H-bond) interactions between their main chains. The side chains of α6 Lys193 and Tyr201 residues form H-bond interactions with the main chains of α3 Ser117 and Leu110 and Met114 residues, respectively (Fig.  1E). Like β9 and β13, the two α-helices are also arranged in an antiparallel orientation. In addition to direct H-bond interactions, AsfvPCNA trimerization is further stabilized by several water-mediated H-bond interactions (Fig. 1F).

Conformational flexibility of AsfvPCNA
In addition to the two structures determined in this study, one additional AsfvPCNA structure was recently deposited in the protein data bank (PDB_ID: 7DRH). Although they possess identical amino acid sequences, the overall folding of the 7DRH structure is very different from that of the Form I or Form II AsfvPCNA structure; the RMSD value between them is close to 1.9 Å over 252 pairs of Cα atoms ( Fig. 2A). The overall folding of Domain I is well conserved in all AsfvPCNA structures, supported by the low RMSD value (0.4 Å).
In contrast to Domain I, the relative orientation and the detailed conformation of Domain II are very different in the AsfvPCNA structures ( Fig. 2B and C). When superimposed on Domain I, there are more than 6 Å movement for the residues at the middle of Domain II Full-Length Text β12-β13 and β14-β15 connecting loops (Fig. 3B). Compared to Form I structure, the α6 helix undergoes approximately 20° rotation in the 7DRH structure (Fig. 2B). Large conformational differences were also observed for AsfvPCNA Domain IIs when they were directly superimposed on each other (Fig. 2C); the RMSD value between the Domain IIs is 1.8 Å. Instead of forming a continuous helix (α6) with residues 188 SKQLQQTF 195 , the residues 196 SDLSN 200 are unfolded into loops in the 7DRH structure. The conformations of β12 N-terminus and β13 C-terminus are well conserved, but the N-terminus of β13 is tilted in the 7DRH structure. More dramatically, conforma tional changes occur in the C-terminus of β12, pointing the β12-β13 connecting loop to a completely different direction.
Compared to Domain I and Domain II, IDCL of AsfvPCNA is more flexible, as indicated by the higher B-factors (Fig. 2D). Not surprise, the IDCL also showed obvious conforma tional differences in the 7DRH and Form I AsfvPCNA structures (Fig. 2E). In the 7DRH structure, the regions 162 DM 163 and 177 LKN 179 are completely disordered. However, the conformations of the α4 helix are well conserved in the two structures, likely due to the strong electrostatic interactions between the side chains of IDCL Glu169 and Arg65 of the β2 strand ( Fig. 2E; Fig. S3B).
Different from Form I and Form II structures, the 7DRH structure belongs to I4 space group, which contains one AsfvPCNA molecule per asymmetric unit. Instead of trimer, the 7DRH structure forms a square-like tetramer with a diameter of 50 Å for the inner ring in the crystal lattice (Fig. S5A). Structural comparison showed that the intermolecular interactions mediated by β9 and β13 are conserved in all the AsfvPCNA structures. However, the interactions between α3 and α6 are not observed in the 7DRH structure, likely due to the rotation of α6 (Fig. S5B). In addition to the conformational flexibility of the AsfvPCNA protomer, molecular packing and crystallization condition may also contribute to the formation of trimeric or tetrameric AsfvPCNA structure.
Compared to the 7DRH structure, our AsfvPCNA structure is more similar to the reported PCNA structures in overall folding and assembly. As depicted in Fig. 3; Fig. S6 and S7, our AsfvPCNA structure can superimpose well with the eukaryotic and archaeal PCNA structures; the diameters of their inner rings are comparable. The structural similarity suggests that AsfvPCNA may be able to function as trimer. Previous studies showed that the sequences and conformations of PCNAs could be different at various regions, especially the J-loop, IDCL, and P-loop regions. While J-loop and IDCL were proposed to play important roles in species-specific partner recognition (54), the P-loop plays a regulatory role through post-translational modifications (59,60).
The overall folding is conserved in AsfvPCNA and homologous PCNA proteins, but the sequence of AsfvPCNA is quite unique. Structure-based alignment (Fig. S8) showed that AsfvPCNA IDCL is much longer and has no clear sequence similarity with IDCLs of other PCNAs. IDCLs exhibit loop-like conformation in all reported PCNA structures ( Fig. 3; Fig.  S6 and S7), whereas it forms two short α-helices (α4 and α5) in the AsfvPCNA structure. The length of the P-loop of AsfvPCNA is similar to that of AfPCNA and HvPCNA, but is shorter than other PCNA proteins, especially the eukaryotic PCNAs. AsfvPCNA P-loop contains four residues ( 234 SSNK 237 ), forming one short helix (α7). The P-loops also contain one short helix in the ScPCNA (Fig. 3B), NcPCNA, and AtPCNA structures (Fig.  S6D), but locates at a different position.
AsfvPCNA J-loop ( 143 DFDIDK 148 ) is composed of six residues, which is identical to ScPCNA, NcPCNA, AtPCNA, DmPCNA, and HsPCNA (Fig. S8). The fifth and sixth residues are relatively conserved, but the other four residues are variable. Compared to these eukaryotic PCNA proteins, AsfvPCNA J-loop is more negative in charge; out of the six residues, three are Asp residues (Asp143, Asp145, and Asp147). The length of AsfvPCNA Jloop is similar to that of SsPCNA, but is longer than AfPCNA, HvPCNA, PaPCNA, and TgPCNA. No sequence similarity could be observed between the J-loops of AsfvPCNA and the archaeal PCNA proteins (Fig. S8). Previous studies showed wild-type NcPCNA is not functional in S. cerevisiae, but introducing NcPCNA with ScPCNA J-loop can rescue the growth of S. cerevisiae pol30Δ PCNA deletion strain, indicating the important role of Jloop. In the future, it is worth investigating whether the J-loop of AsfvPCNA also plays a certain functional role in ASFV.
Like IDCL, J-loop, and P-loop, the C-terminal tail (C-tail) is also important for the function of many PCNAs (7,61,62). The sequences of the C-tail are relatively conserved in eukaryotic and archaeal PCNAs (Fig. S8). As observed in many PCNA structures, the first four residues of the C-tail are ordered, but the following residues are disordered, indicating their high flexibility in conformation. The C-tails of eukaryotic and archaeal PCNAs all contain a number of negatively charged residues (either Asp or Glu) at or near structures. Domain I, Domain II, and IDCL of Form I structure are colored as in Fig. 1A, whereas 7DRH structure is colored in blue in all panels.
Full-Length Text the very C-termini. However, the C-tail ( 297 LNNTI 301 ) of AsfvPCNA has no sequence similarity with either the eukaryotic or the archaeal PCNAs (Fig. S8). Although it is ordered in the 7DRH structure, the four residues ( 298 NNTI 301 ) are all disordered in the Form I and Form II AsfvPCNA structures, suggesting that AsfvPCNA C-tail is very flexible and can undergo large conformational changes in solution.

Comparison of AsfvPCNA with homologous proteins from other viruses
In addition to ASFV, PCNA homologous proteins are also present in many other viruses. Although the homologous protein structures for many viruses are still unavailable, the structures of HSV-1 virus UL42, HCMV virus UL44, KSHV virus PF-8, and Epstein-Barr virus BMRF1 have been reported (33)(34)(35)63). As depicted in Fig. 4A, UL42, UL44, PF-8, and BMRF1 are composed of two domains: Domain I and Domain II, which are located at the N-and C-termini, respectively; the overall folding and the relative orientations between the two domains are similar to that of AsfvPCNA. However, the functional states of these Full-Length Text viral proteins are different from AsfvPCNA (Fig. 4B). Instead of trimer, UL42 exists as monomer. UL44, PF-8, and BMRF1 all function as dimers with a C-shaped conformation; the two protomers are arranged in a head-to-head manner. In the AsfvPCNA structure, the three protomers are arranged in a head-to-tail manner: Domain I of one protomer interacts with Domain II of the neighboring protomer. Like AsfvPCNA, the Domain I and Domain II of UL42, UL44, PF-8, and BMRF1 are also connected by an IDCL linker (Fig. 4A). As confirmed by their crystal structures, the IDCL linkers play a critical role in partner recognition by these viral sliding clamp proteins. The IDCL linker of AsfvPCNA is longer and shares no sequence similarity with those of UL42, UL44, PF-8, and BMRF1 (Fig. 4C), which may lead to the different IDCL conformations observed in the structures. The predicted DNA-interacting surfaces of UL42 and UL44 contain several Lys and Arg residues ( Fig. S9C and D), but their detailed distributions and conformations are very different from those of AsfvPCNA, ScPCNA, and HsPCNA ( Fig. 5B; Fig. S9C and D).

AsfvPCNA possesses high dsDNA-binding affinity
Like other eukaryotic and archaeal PCNA structures, the ring-shaped AsfvPCNA structure can be divided into four parts: the outer surface, the inner surface, the top side, and the bottom side. The top side is also termed the C side because the C-termini of the PCNA protomers all protrude from this side. As depicted in Fig. 5A, the electrostatic potential Full-Length Text of AsfvPCNA is unevenly distributed. The C side is highly negative at the external region, due to the presence of Glu160, Asp162, Glu164, and Glu167 of IDCL and Glu276 and Glu277 of Domain II. The internal region of the C side is mainly composed of hydropho bic and positively charged residues, such as Lys222 and Lys255.
Compared to other parts, the inner surface of AsfvPCNA is more positive in charge (Fig. 5A). The side chains of Lys48, Lys55, Lys113, and Arg120 of Domain I and Lys189, Lys192, and Lys260 of Domain II all point toward the inner channel. Instead of the same protomer, these positively charged residues from the neighboring protomers are close to each other in space (Fig. 5B). HsPCNA and ScPCNA possess similar number of positively charged residues (Lys or Arg) at the inner surface of rings. The locations of these residues are conserved in HsPCNA and ScPCNA (Fig. S9A and B). However, out of the seven positively charged residues of AsfvPCNA, only Lys55 and Lys113 are relatively conserved in position.
PCNAs can interact with various types of partner proteins, including polymerase, ligase, and Fen1 endonuclease that all involve in DNA replication and/or repair pathway. To enhance the catalytic activity of the partner protein, PCNA should be able to form a stable complex with the partner protein and slide freely along the substrate DNA. The structures of PaPolD-PCNA-DNA (Fig. S10A) and ScPolδ-PCNA-DNA (Fig. S10B) have been reported (26,27), showing that DNA was held by the polymerase and threaded through the central channel of the PCNA ring. The orientation of the DNA is roughly Full-Length Text perpendicular to the plane of the ring. Similar orientations were also observed for DNA and β clamp in the EcPolIII-β clamp-DNA complex structure (Fig. S10C) (64).
We could not obtain any AsfvPCNA crystal in the absence of dsDNA, whereas Form I and Form II AsfvPCNA crystals readily grew when 10 bp dsDNA (5′-CCCATCGTAT-3′; 5′-ATACGATGGG-3′) is present in the sample. In the structures, extra electron density maps could be observed in the central channel (Fig. S11A), suggesting the existence of dsDNA. The existence of dsDNA could be further supported by staining the crystals using GelRed, a nucleic acid-specific stain ( Fig. S11B and C). Likely, due to the dynamic binding or sliding of the DNA, the electron density is diffused, and no DNA was included in the final structures. Weak electron density was also observed for DNAs in some previously reported PCNA/DNA complex structures (50). To further confirm the DNA-binding ability of AsfvPCNA, we performed in vitro EMSA assays (Fig. 5C). Compared to single-stranded (ssDNA), the dsDNA-binding affinity of AsfvPCNA is higher (Fig. 5D). The calculated dissociation values (K d ) are 0.16 µM and 0.52 µM for the dsDNA and ssDNA, respectively. Taken together, these observations suggested that AsfvPCNA possesses high dsDNAbinding affinity, and the dsDNA is likely bound in the central channel with an orientation similar to those in the reported polymerase ternary complex structures (Fig. S10).

AsfvPCNA modestly enhances the ligation activity of AsfvLIG
ASFV is one of the most complex dsDNA viruses known to date. In addition to PCNA, the genome of ASFV also encodes many other proteins involved in DNA replication and repair, such as replicative DNA polymerase, topoisomerase II, and ligase (65). Maybe due to the difficulties in expression and purification, the structures of ASFV replicative polymerase and topoisomerase II remain elusive. The structure of ASFV ligase (AsfvLIG) was determined by our group previously (48). Like HsLIG (Fig. S12A) and all other homologous proteins, AsfvLIG is composed of three domains: the DNA-binding domain (DBD) at the N-terminus, the adenylation domain (AD) in the center, and the OB-fold domain (OB) at the C-terminus (Fig. S12B).
Interestingly, although they share similar domain architectures, the sequence similarities between AsfvLIG and the homologous proteins are very low, especially at the DBD domain region. The DBD of canonical ligase is approximately 280 amino acids in size and is of α-fold in nature. The DBD domain of AsfvLIG (amino acids 1-120) is much shorter; in addition to α-helices, it also contains several β-strands. The SsPCNA-SsLIG-DNA ternary complex structure (PDB_ID: 7RPX) was recently reported (28), showing that DNA was bound in the central channel of SsPCNA (Fig. 6A). The IDCL loop of SsPCNA plays the most critical roles in the ternary complex assembly; it directly recognizes the PCNA-interacting peptide (PIP) motif of SsLIG, which locates in the DBD domain.
As aforementioned, both the IDCL linker of AsfvPCNA ( Fig. 3 and 4) and the DBD domain of AsfvLIG (Fig. S11B) are unique, which raises the question whether AsfvPCNA can enhance the ligation activity of AsfvLIG. To this end, we performed in vitro ligation assays. As depicted in Fig. 6B (left panel), AsfvLIG is active in the absence of AsfvPCNA; however, the activity is relatively weak. Addition of 0.5 and 1.0 molar ratios of AsfvPCNA (trimer) can slightly enhance the ligation activity. Addition of 2 molar ratio of AsfvPCNA in the reaction system leads to an approximately twofold enhancement in DNA ligation catalyzed by AsfvLIG. No further enhancement is observed when more AsfvPCNA is included in the reaction system (Fig. S13). To test whether the enhancement by AsfvPCNA is specific, we purified the ScPCNA protein (Fig. S14) and performed ligation assays under the same condition. In contrast to AsfvPCNA, the addition of ScPCNA has no impact on DNA ligation catalyzed by AsfvLIG (Fig. 6B, right panel). Like ScPCNA, the DNA-binding affinities of many other archaeal and eukaryotic PCNAs are also weak, whereas these PCNA proteins can significantly enhance the catalytic activities of their partner proteins. Instead of AsfvPCNA-DNA binding, we believe that the potential AsfvPCNA-AsfvLIG interaction may play a more significant role in the enhancement of the catalytic activity of AsfvLIG. the IDCL linker, the P-loop, and the C-tail. It is of note that one trimeric AsfvPCNA structure was recently reported by Wu and coworkers, showing that the N-terminal extension of AsfvPCNA is also unique and plays an important role in the stabilization of AsfvPCNA homotrimer in solution (66).
Compared to eukaryotic and archaeal PCNAs, the IDCL linker of AsfvPCNA is much longer and contains two short α-helices (α4 and α5) near the C-terminus (Fig. 1A). Via interacting with the PIP-motif (PCNA interaction peptide) of the partner proteins, IDCL plays the most critical role in partner recognition by PCNA. The canonical PIP-motif is composed of Q-xx-ψ-x-x-θ-θ, in which x represents any residue, ψ and θ are hydrophobic (e.g., M, L, and I) and aromatic (e.g. F and Y) residues, respectively (67,68). By exploiting a random peptide display library, a novel PIP-motif with a sequence of K-A-(A/L/I)-(A/L/ Q)-x-x-(L/V) was discovered. This novel motif was also termed the KA-motif and has been found in several PCNA partner proteins (69). Although both AsfvPCNA and AsfvLIG show obvious conformational differences from their homologous proteins ( Fig. 3; Fig.  S11), our in vitro ligation assay results (Fig. 6B) confirmed that AsfvPCNA can modestly enhance the ligation activity of AsfvLIG. Sequence analysis did not identify any canonical PIP-like or KA-like motif in AsfvLIG, but structural superposition suggested that the β4-β5 linker of AsfvLIG DBD is likely involved in interaction with AsfvPCNA IDCL (Fig. 6C). The conformation of AsfvLIG β4-β5 linker is very different from the canonical PIP-or KA-motif. We speculated that AsfvLIG β4-β5 linker and/or AsfvPCNA IDCL will undergo certain conformational changes to form stable interactions.
To enhance the catalytic activities of polymerase, ligase, and nuclease, PCNA needs to encircle and freely slide along the substrate DNA. However, eukaryotic and archaeal PCNAs, such as HsPCNA, ScPCNA, and SsPCNA, exist as stable trimer with a closed ring-shaped overall structure. In all known cellular life, PCNA loading onto DNA requires the help of the RFC factor, which is composed of five protein molecules. A very recent study showed that PCNA loading is a coordinated and stepwise process, including ATP binding by RFC, PCNA-RFC complex formation, large-scale expansion of RFC, PCNA opening, DNA binding, PCNA closure, ATP hydrolysis, and RFC ejection (70). Unlike the eukaryotic and archaeal PCNAs, our in vitro assays showed that AsfvPCNA has strong dsDNA-binding ability ( Fig. 5C and D). The K d value of AsfvPCNA is comparable to that of the viral sliding clamp proteins, such as UL42 and UL44, which do not require RFC for DNA binding. In addition to positively charged Lys and Arg residues located at the inner surface, the recent study suggested that the N-terminal extension also plays a certain role in DNA binding by AsfvPCNA (66). As demonstrated by our structures (Fig. 1C) and the 7DRH structure (Fig. S4), AsfvPCNA can exist as either trimer or tetramer with a central channel. At present, it is unclear whether both trimeric and tetrameric AsfvPCNAs are functional and whether RFC or a similar factor is required for opening and loading AsfvPCNAs onto target dsDNA in vivo.
Owing to their important biological functions, PCNAs have been considered a potential antibacterial and anticancer drug target (31). Various types of PCNA inhibi tors have been identified, including natural products, small molecule inhibitors, and peptides. Peptide inhibitors generally mimic the PIP-motif in sequence, and they share similar PCNA-binding mode with many known PCNA partners, such as Fen-1 nuclease. Peptide inhibitors can also be derived from other PCNA-interacting motif, such as the AlkB homolog 2 PCNA-interacting motif (APIM). Compared to the canonical PIP-motif, the APIM motif is short; it is only composed of five residues, (K/R)(Y/Y/W)(L/I/V/A)(L/I/V/ A)(K/R) (71). Although they are very different in sequence, these peptide inhibitors normally bind to the PIP-binding pocket and form H-bond and hydrophobic interactions with the IDCL linker of PCNA, blocking the binding of PCNA partners. Like griselimycins (32), nonsteroidal anti-inflammatory drugs, 3,3′,5-triiodothyronine (T3), and some other small molecule inhibitors also target the PIP-binding pocket, inhibiting partner protein binding and interacting with the IDCL linkers of the sliding clamp proteins (72,73). The sequence and conformation of AsfvPCNA IDCL is very unique (Fig. 3 and 4), which may allow the development of AsfvPCNA-specific peptide inhibitors.
In addition to the PIP-binding pocket, peptide and small molecule inhibitors can also bind to other regions of PCNA. For example, the small molecule T2AA (an ana log of triiodothyronine) can bind at the interface between the trimer subunits, block PCNA ubiquitination, and inhibit partner interaction (73). It was reported that T2AA can regulate how cells respond to DNA damage; cells treated with T2AA are more sensitive to the cancer therapeutic cisplatin and have a lower survival rate. Binding of small molecule and canonical PIP-like or APIM-like peptide inhibitors usually does not alter the overall structures of PCNAs. In contrast, binding of some small proteins, such as Thermococcales inhibitor of PCNA (TIP), can lead to obvious conformational changes of PCNA (74). TIP alone is partially disordered, whereas it becomes structured upon PCNA binding. TIP can interact with both Domain I and Domain II, altering the relative orientation between them and preventing PCNA from functional trimer assembling. As revealed by the structural superposition, Domain I and Domain II of AsfvPCNA can undergo large conformational changes ( Fig. 2A and B). Small molecule and peptide that block oligomerization or fix the structure in a nonfunctional state will certainly disrupt the function of AsfvPCNA.
ASFV is one of the most complex DNA viruses known to date, its genome encodes more than 160 proteins. We previously reported the crystal structures of AsfvAP, AsfvPolX, and AsfvLIG proteins, which revealed many ASFV-specific structural features, such as the narrow abasic-binding site in AsfvAP (47), the novel 5′-P-binding pocket in the finger domain of AsfvPolX (49), and the unique DBD domain in the N-terminus of AsfvLIG (48). Here we show that AsfvPCNA is also very different from the homologous PCNA proteins in the J-loop, IDCL, the P-loop, and the C-tail regions ( Fig. 3; Fig. S6 to S8). ASFV is the only member of the Asfarviridae family; the majority ASFV proteins share very low sequence similarity with the homologous proteins, which may explain the unique structural features possessed by AsfvAP, AsfvPolX, AsfvLIG, and AsfvPCNA. In the future, it is worth studying the structure and function of other ASFV proteins, such as the replicative DNA polymerase. Most likely, these proteins also contain some unique structural features that can be utilized as target for the development of ASFV-specific inhibitors and help combat the deadly virus.

Protein expression and purification
The codon optimized AsfvPCNA gene (Table S1) and ScPCNA (Table S2) gene were synthesized by GENEWIZ Co., Ltd, Suzhou, China. The target genes were amplified by polymerase chain reaction and subcloned into the pET-28a-Sumo vector between BamHI and XhoI restriction sites. The recombinant plasmids were transformed into Escherichia coli strain BL21 (DE3) and overexpressed. The cells were cultured at 37°C in 1 L Luria-Ber tani medium containing 50 µg/mL of kanamycin. When the OD600 value reached 0.6-0.8, protein expression was induced by the addition of isopropyl β-D-1-thiogalacto-pyra noside at a final concentration of 0.05 mM. The induced cultures were then grown at 18°C for an additional 18-22 h.
Cell was harvested and resuspended in lysis buffer (20 mM Tris pH 7.0, 500 mM NaCl, and 25 mM imidazole) and homogenized with a low-temperature ultra-high pressure cell disrupter. The lysate was centrifuged at 25,000 × g for 30 min at 4°C to remove cell debris. The supernatant was loaded onto a HisTrap HP column equilibrated with lysis buffer. The fusion protein was eluted using elution buffer (20 mM Tris pH 7.0, 500 mM NaCl, and 500 mM imidazole) with a gradient. The fractions containing the desired fusion proteins were pooled and dialyzed against dialysis buffer (20 mM Tris pH 7.0 and 500 mM NaCl) at 4°C for 3 h, and Ulp1 protease was added to the sample during the dialysis process. The sample was again loaded onto the HisTrap HP column. The target proteins were collected, concentrated, and loaded onto a HiLoad 16/600 Superdex S200 column equilibrated with gel filtration buffer (20 mM Tris pH 7.0, 200 mM NaCl, and 2 mM DTT).
Selenomethionine-substituted AsfvPCNA protein was expressed in M9 medium supplemented with 60 mg/L Se-Met and purified using a procedure similar to that of the native protein. AsfvLIG protein was expressed and purified as previously described. Purity of all proteins was analyzed using a 15% SDS-PAGE gel, and the samples were stored at −80°C until use.

Crystallization and data collection
Substrates DNA1 (5′-CCCATCGTAT-3′) and DNA2 (5′-ATACGATGGG-3′) were dissolved in gel filtration buffer. The DNA was annealed by heating to 95°C and slowly cool ing to room temperature. Prior to crystallization, AsfvPCNA and the annealed dsDNA were mixed, the final concentration of AsfvPCNA is 15 mg/mL (corresponding to 0.14 mM AsfvPCNA trimer); the concentration of dsDNA is 0.21 mM. Initial crystallization conditions were screened by the sitting-drop vapor-diffusion method using commercial crystal screening kits at 18°C. The drop contained an equal volume (0.2 µL) of protein sample and reservoir solution and was equilibrated against 50 µL of reservoir solution in a 96-well format. Form I AsfvPCNA crystals were grown in the buffer composed of 0.1 M Tris pH 8.5 and 20% (w/v) PEG1000, whereas Form II crystals were grown in 0.1 M sodium chloride, 0.1 M BICINE pH 9.0, and 20% v/v PEG550 buffer.
All crystals were cryoprotected using their mother liquor supplemented with 25% glycerol and snap-frozen in liquid nitrogen. The diffraction data were collected at beamline BL18U1 at the Shanghai Synchrotron Radiation Facility. Data processing was carried out using the imosflm or HKL3000 program. The data collection and processing statistics are summarized in Table 1.

Structure determination and refinement
Form I AsfvPCNA structure was solved by the single-wavelength anomalous diffraction method (75) with the Autosol program embedded in the Phenix suit (76). The initial model was built using the Autobuilt program and then refined against the diffraction data using the Refmac5 program of the CCP4 suite (77). The 2F o −F c and F o −F c electron density maps were used as guides for the building of the missing amino acids using COOT (78). Form II AsfvPCNA structure was solved by molecular replacement using the Form I structure as the search model with the phaser program of the CCP4 suite (79). The final refinement of both structures was performed using the phenix.refine program. The structural refinement statistics are also summarized in Table 1.

Electrophoretic mobility shift assay
ssDNA (5′-FAM-CCCATCGTAT-3′) or dsDNA (5′-FAM-CCCATCGTAT-3′ and 5′-ATACGA TGGG-3′) was mixed with different molar ration of AsfvPCNA in binding buffer (20 mM Tris pH 7.0 and 500 mM NaCl). The final concentrations are 10 nM for both ssDNA and dsDNA. Twenty microliters of reaction mixtures was incubated on ice for 1 h and then analyzed on 6% native PAGE gels with 0.5 × TBE (Tris-borate-EDTA) buffer. The gel was imaged using Typhoon FLA 9000. The intensities of the substrate bands were quantified by ImageQuant TL. The percentage of binding, for each protein concentration, was calculated. Data were then fitted to the equation Y = B max *X h /(K d h + X h ) using nonlinear regression (curve fit) in GraphPad Prism. The dissociation constants (K d ) were determined from the regression curve.