Foot-and-Mouth Disease (FMD) Virus 3C Protease Mutant L127P: Implications for FMD Vaccine Development

ABSTRACT The foot-and-mouth disease virus (FMDV) afflicts livestock in more than 80 countries, limiting food production and global trade. Production of foot-and-mouth disease (FMD) vaccines requires cytosolic expression of the FMDV 3C protease to cleave the P1 polyprotein into mature capsid proteins, but the FMDV 3C protease is toxic to host cells. To identify less-toxic isoforms of the FMDV 3C protease, we screened 3C mutants for increased transgene output in comparison to wild-type 3C using a Gaussia luciferase reporter system. The novel point mutation 3C(L127P) increased yields of recombinant FMDV subunit proteins in mammalian and bacterial cells expressing P1-3C transgenes and retained the ability to process P1 polyproteins from multiple FMDV serotypes. The 3C(L127P) mutant produced crystalline arrays of FMDV-like particles in mammalian and bacterial cells, potentially providing a practical method of rapid, inexpensive FMD vaccine production in bacteria. IMPORTANCE The mutant FMDV 3C protease L127P significantly increased yields of recombinant FMDV subunit antigens and produced virus-like particles in mammalian and bacterial cells. The L127P mutation represents a novel advancement for economical FMD vaccine production.

F oot-and-mouth disease virus (FMDV) is a worldwide threat to food security, production, and trade, with approximately 2.35 billion doses of foot-and-mouth disease (FMD) vaccines administered to livestock annually (1). Efficacious FMD vaccines require production of intact, assembled FMDV capsids to induce a protective immune response. Conventional FMD vaccines are produced by chemical inactivation of purified, virulent FMDV virions, but this production method is inherently risky with respect to accidental release of FMDV. Current U.S. law (21 U.S.C. 113A) prohibits introduction of live FMDV into the U.S. mainland for any purpose, which precludes manufacture of inactivated FMDV vaccines within the United States. Recombinant subunit vaccines are a safer alternative to traditional FMDV vaccines because they express only the FMDV proteins required for assembly of empty FMDV capsids, or virus-like particles (VLPs) (2)(3)(4)(5). Molecular recombinant subunit vaccines containing only the FMDV genes needed to induce immunity have been developed and recently licensed (2)(3)(4).
We sought to identify FMDV 3C mutations that increased yields of recombinant antigens in transfected cells compared to wild-type 3C. To screen for increased yields, FMDV 3C protease variants were linked with Gaussia princeps luciferase (GLuc) expression and separated by the FMDV translational interrupter sequence (Δ1D2A on a bicistronic gene cassette) (22). Mammalian cells expressing the 3C(L127P) mutation had significantly increased recombinant protein production compared to cells expressing wild-type FMDV 3C protease. The 3C(L127P) mutant was further analyzed for the ability to process the P1 polypeptide from multiple FMDV serotypes, for its effect on host proteins, and for the production of VLPs in both mammalian and bacterial cells. Last, we show that cattle immunized with an adenovirus (Ad)-vectored FMDV vaccine using the 3C(L127P) protease are protected from clinical FMD after homologous challenge.

RESULTS
Effect of 3C mutations on transgene expression. The FMDV 3C protease is highly conserved among all seven FMDV major serotypes (23). We evaluated five 3C variants for their effect on recombinant protein (GLuc) yields compared to 3C(wt) using a bicistronic vector (chimera of Gaussia luciferase and the FMDV 3C protease [GLuc-3C]) (Fig. 1A). The variants included three novel 3C mutations (V28K, L127P, and V141T), the C163A mutant that inactivates 3C proteolytic activity (24), and a construct that downregulates 3C activity by combination with an HIV frameshift sequence (HIV-C142T) to reduce 3C translation to approximately 5% of normal expression levels (9).
Western blot analysis that showed 3C(L127P) processed the P1 polyprotein from FMDV serotypes O, A, Asia1, C3, and SAT2 into the individual FMDV capsid proteins VP0, VP1, and VP3 ( Fig. 2C to E). Detection of VP2 in the cell lysates from serotype strains SAT2, O1, A24, Asia1, and C3 demonstrated separation of VP0 into VP2 and VP4, which indicates the occurrence of capsid assembly (25) (Fig. 2C and D). However, differential levels of antibody binding affinity among serotypes resulted in different band intensities among samples. For example, to visualize the presence of SAT2 VP2, additional lysate needed to be loaded (Fig. 2D). While processing of A2006 Turkey P1 into VP0, VP3, and VP1 was seen by Western blotting, VP2 was not detected despite testing increased amounts of lysate (Fig. 2E). Interestingly, polyclonal antibody detection of VP3 resulted in the observation of a doublet of VP3 bands for serotypes O1 Manisa, Asia Shamir, and A2006 Turkey but no such result was observed for A24, C3 Indaial, or SAT2 Egypt 2012 ( Fig. 2C and E). Processing of SAT1 and SAT3 P1 polyproteins was not assessed by Western blotting due to a lack of reactive monoclonal antibodies for those serotypes.
VLP production in mammalian cells. When produced at high concentrations, assembled FMDV capsids and VLPs can be observed as crystalline arrays within cells by transmission electron microscopy (TEM) (22,26). In this study, crystalline arrays were seen in mammalian cells transfected with P1-3C(L127P) constructs encoding P1 polyproteins from FMDV serotypes O, Asia1, and SAT2 ( Fig. 3A to C). Cells expressing FMDV serotype O P1-3C(L127P) were observed producing crystalline arrays in the cytoplasm which disassociated at the plasma membrane (Fig. 3D) to release individual VLPs from the cell (Fig. 3E). Although crystalline arrays were not observed for all FMDV serotypes tested, previous TEM studies indicated that crystalline arrays of VLPs are not always found for subunit vaccine constructs that are efficacious in cattle trials (unpublished data).
Effect of 3C(L127P) on mammalian host proteins. To further characterize the mechanism behind the increased yields of recombinant proteins in mammalian cells associated with 3C(L127P), HEK293-T cells were transfected with plasmids encoding the FMDV O1 Manisa P1 polyprotein and a 3C variant (wt, L127P, or C163A). Cell lysates were analyzed for cleavage products from FMDV and host proteins with equal levels of loading of cell lysates determined by blotting for GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (Fig. 4B). Western blots showed higher yields of FMDV VPs in cells transfected with O1P1-3C(L127P) than in those transfected with O1P1-3C(wt) (Fig. 4A), confirming the luciferase assay results. Cells transfected with O1P1-3C(C163A) had no detectable processed P1 polyprotein, confirming the knockout of proteolytic activity (Fig. 4A).
Degradation of host proteins was evaluated by Western blotting for four known host protein targets: NEMO, histone H3, SAM68, and eIF4AI. Lysates from HEK293-T cells transfected with O1P1-3C(C163A), the negative control, showed no detectable degradation of these four host proteins (Fig. 4C to G). Lysates from cells transfected with O1P1-3C(wt) showed degradation products for all four tested host proteins. The 3C(L127P) mutation completely blocked degradation of NEMO (Fig. 4C) and SAM68 ( Fig. 4F and G) and reduced degradation of eIF4AI (Fig. 4E) and histone H3 (Fig. 4D).
SAM68 is degraded in FMDV-infected cells, resulting in a 40-kDa N-terminal fragment and intracellular redistribution of the protein (15). In this study, lack of a 40-kDa band indicated that 3C(L127P) did not cleave SAM68 (Fig. 4F). Western blots using an antibody to the C terminus of SAM68 showed a unique band at around 20 kDa in the O1P1-3C(wt) Degradation of eIF4AI by 3C(wt) shuts down host translation in mammalian cells (13), and the 3C(L127P) mutant degraded eIF4AI less than 3C(wt) (Fig. 4E). To further characterize the effect of 3C(L127P) on bovine eIF4AI under uniform expression conditions, a pSNAP plasmid encoding bovine eIF4AI was coexpressed in a cell-free system at equimolar concentrations with a second pSNAP plasmid encoding 3C(L127P), 3C(C163A), or 3C(wt). Results showed that degradation of bovine eIF4AI by 3C(L127P) and 3C(C163A) was negligible compared to that by 3C(wt) (Fig. 5).
Effect of expression of the 3C mutants in Escherichia coli. We evaluated the ability of the 3C(L127P) mutant to enable bacterial expression as a practical platform for which 3C(wt) toxicity had previously been problematic. When 3C protease mutants were expressed in E. coli under conditions of Lac I regulation, approximately 1,000 more colonies expressing the 3C(C163A) or 3C(L127P) mutants survived after overnight induction (IPTG [isopropyl-␤-D-thiogalactopyranoside]) compared to E. coli transfected with the other three mutants or the wild-type 3C gene (Fig. 7), demonstrating that the 3C(L127P) mutation also mitigated the detrimental effects of 3C expression in the bacterial system.

DISCUSSION
The results of this study show that the FMDV 3C(L127P) mutant significantly increased yields of recombinant FMDV capsid proteins compared to wild-type 3C protease and produced FMDV VLPs in both mammalian and bacterial cells. Furthermore, cattle immunized with a single dose of Ad5-FMD P1-3C(L127P) vaccine were protected from clinical disease after challenge with homologous FMDV.
The increased yields of recombinant vaccine antigens associated with use of 3C(L127P) appear to result from reduced degradation of host proteins allowing better expression of transgenes in transformed cells. The L127 residue resides on the B 2 ␤-strand in published crystal structures (30,31). It has been previously hypothesized that the nearby A 2 -B 2 loop in the FMDV 3C protease may contribute to substrate specificity (30). The results showing reduced cleavage of host proteins NEMO, SAM68, histone H3, and eIF4AI by 3C(L127P) collectively would support the conclusion that this region of the 3C protease can play a critical role in substrate specificity despite being distal to the active site. In particular, reduced degradation of eIF4AI by 3C(L127P) indicates less inhibition of host cell translation than was seen with 3C(wt). Shutdown of this pathway by 3C(wt) has been a hindrance for multiple molecular vaccine platforms, particularly those which utilize cap-dependent translation for antigen expression. NEMO inhibits nuclear factor kappa-B kinase subunit gamma and supports type I interferon production (16). The lack of NEMO degradation by 3C(L127P) compared to 3C(wt) might mitigate this viral immune evasion mechanism and might possibly enhance the efficacy of recombinant FMD vaccines containing a 3C(L127P) protease gene.
A reduction in toxic effects on host cells by 3C(L127P) should also improve production yields of recombinant subunit FMD vaccines and vaccine potency. For example, current methods of adenovirus-vectored FMDV vaccine production seek to reduce or eliminate expression of 3C(wt) in order to limit toxicity to host cells during the manufacturing process. The reduced toxicity associated with the 3C(L127P) protease may improve manufacturing yields by limiting toxicity in the system or may lower costs by eliminating the inhibitors currently required during the production cycle. Moreover, vaccination with Ad-FMD vectors using the 3C(L127P) protease may have benefits with respect to intrinsic antigen expression. In this study, cattle immunized with a single  The most striking result of using 3C(L127P) was the clear demonstration of abundant intact FMDV VLPs produced in E. coli indicated by intracellular crystalline arrays as seen by electron microscopy. To our knowledge, this is the first demonstration of FMDV VLP crystalline array structures forming in a bacterial cell. Crystalline array structures were not seen in E. coli transfected with FMD vaccine constructs expressing 3C(wt) (unpublished results). Incorporation of the 3C(L127P) protease into FMD vaccine constructs may enable bacterial expression of FMD subunit vaccines as a safe and economical production method that could significantly reduce costs compared to FMD vaccine production in eukaryotic cells. As such, the 3C(L127P) mutant advances the potential to use bacterial expression as a production platform for recombinant FMD subunit vaccines, a long-sought goal of the FMD vaccine community.

MATERIALS AND METHODS
Preparation of pTarget GLuc-⌬1D2A-3C constructs. The pTarget GLuc-Δ1D2A plasmid vector was constructed as previously described (22). Amplification of the 3C protease gene from an FMDV Asia1 Lebanon 1989 (GenBank accession no. AY593798) noninfectious template was performed using OneTaq 2ϫ master mix with Standard Buffer (New England BioLabs) and primers XmaI-3C-F (CTACCCGGGCCG AGTGGTGCCCCAC) and 3C-NotI-R (TAGCGGCCGCTACTCGTGGTGTGGTTC). The PCR product was purified using a PCR purification kit (Qiagen). Both the PCR product and pTarget GLuc-Δ1D2A vector were digested with XmaI and NotI-HF restriction enzymes (New England BioLabs). Ligations were performed using T4 DNA ligase (Roche) and transformed into NEB 5-alpha competent E. coli (New England BioLabs). Plasmids were isolated using a QIAprep Spin Mini-prep kit (Qiagen) and were amplified with primers T7 (TAATACGACTCACTATAGGG) and Seq-R (TTACGCCAAGTTATTTAGGTGACA) for sequencing, and results were analyzed with Sequencher 4.8 software (Genecodes).
Luciferase assay. Luciferase activity in harvested cell culture medium was measured using a 96-well BioSystems Veritas luminometer (Turner Biosystems) with 20 l of sample in each well. Luminescence output was measured immediately upon injection of 25 l of 50 g/l coelenterazine solution (Nano-Light Technologies, Pinetop, AZ) using an integration time of 0.5 s both before and after injection of substrate. Preinjection readings were used to establish a baseline at the time of injection and were subtracted from postinjection values for data analysis. Replicate data were averaged together to give the overall luciferase activity in relative luciferase units per half-second (RLU/0.5 s). Six or seven replicates were used for each sample (see figure legends). Data were analyzed initially by single-factor analysis of variance (ANOVA), which indicated significance. The two-tailed t test (with unequal levels of variance) was then used to compare luciferase RLU/0.5 s data between one construct and another. A post hoc one-tailed t test analysis was used to determine whether the luciferase RLU/0.5 s data from the 3C(L127P) constructs were different from the data determined for the other 3C mutants and wild-type constructs.
Construction of pSNAP FLAG-3C and pET O1P1-2A plasmids. The FLAG-3C(wt) gene was synthesized by GenScript and inserted into pSNAP-tag (T7)-2 vector (New England BioLabs) using restriction enzymes NdeI and XhoI as described above. To produce V28K, V141T, and C163A mutants, site-directed mutagenesis was performed using the primers and protocol described above. To produce L127P and C142T mutants, site-directed mutagenesis was performed as described above using the following primer sets: 3C L127P-MF (CCGACGTTGGGAGACCGATTTTCTCCGGTGA) and 3C L127P-MR (TCACCGGAGAAAAT CGGTCTCCCAACGTCGG) as well as 3C C142T-MF (AAGGACATTGTAGTGACCATGGATGGAGACAC) and 3C C142T-MR (GTGTCTCCATCCATGGTCACTACAATGTCCTT), respectively. For construction of the pET O1P1-2A plasmid, a NotI restriction site was inserted into a Champion pET SUMO expression system (Thermo Fisher). The FMDV O1 Manisa P1-2A polypeptide gene was synthesized (GenScript) with flanking NheI and NotI restriction sites. The NheI and NotI restriction sites were used to insert the P1-2A gene into the modified pET SUMO plasmid, replacing the SUMO sequence.
Transformation of pSNAP FLAG-3C constructs. To evaluate the effect of 3C expression in bacteria, pSNAP FLAG-3C plasmids were transformed into T7 Express competent E. coli (New England BioLabs) and incubated in 10 ml of 100 g/ml carbenicillin Terrific broth on a shaker for 3 h at 37°C. After incubation, viable E. coli cells were quantified using the BacTiter-Glo microbial cell viability assay (Promega). For all samples, equivalent numbers of viable E. coli cells were mixed in 200 l containing 100 g/ml carbenicillin in Terrific broth, and 75 l was plated on LB plates containing either 100 g/ml carbenicillin or 100 g/ml carbenicillin, 0.1 mM IPTG, and 60 g/ml X-Gal (5-bromo-4-chloro-3indolyl-␤-D-galactopyranoside). Plates were incubated at 37°C overnight and examined for growth.
Expression and Western blot analysis of pSNAP FLAG-3C and pET O1P1-2A transformed E. coli. For expression in bacteria, the pSNAP FLAG-3C and pET O1P1-2A plasmids were transformed into T7 Express competent E. coli (New England BioLabs) and grown in LB broth containing 100 g/ml carbenicillin and 50 g/ml kanamycin (Teknova) overnight at 37°C. Cultures were split, and an equal amount of fresh media was added. To induce expression, IPTG was added to reach a final concentration of 1 mM, and cultures were incubated on a shaker overnight at 30°C.
Induced bacterial cultures were pelleted by centrifugation at 6,800 ϫ g for 3 min. Supernatant was removed and the pellet dissolved in 500 l of B-PER bacterial protein extraction reagent (Thermo). After the resuspended bacteria were subjected to one freeze-thaw cycle, 5 l of DNase I (New England BioLabs) was added, and the treated bacteria were incubated at 37°C for 1 h. Protein expression and processing were evaluated by Western blotting using precast polyacrylamide (4% to 12%) NuPAGE Bis-Tris gels transferred to a 0.2-m-pore-size polyvinylidene difluoride (PVDF) membrane and probed with F1412SA (anti-VP0 and anti-VP2) mouse monoclonal antibody at a 1:50 dilution (32) and 12FE9.2.1 (anti-VP1) mouse monoclonal antibody at a 1:50 dilution (33); both dilutions were performed using PBS with Tween 20 (PBS-T). Blots were blocked in 5% milk for 1 h, followed by three 5-min washes performed with 1ϫ PBS-T buffer. Primary antibodies were incubated for 1 h at room temperature, and membranes were washed three times with 1ϫ PBS-T buffer for 5 min each time. Secondary antibody of goat anti-mouse horseradish peroxidase (HRP) (KPL) was applied to membranes for 1 h at room temperature followed by three 5-min washes with 1ϫ PBS-T. Blots were developed with SigmaFAST 3,3=diaminobenzidine (Sigma) for 1 h at room temperature or until bands were clearly visible followed by two washes with double-distilled water (ddH 2 O).
Transfection and harvesting of mammalian cells transfected with P1-2A-3C-SGLuc constructs. We transfected HEK293-T cells with mpTarget P1-2A-3C(L127P)-SGLuc and quantified luciferase activity as previously described for pTarget GLuc-Δ1D2A-3C constructs (22). Cells were harvested by resuspension in fresh growth media and centrifugation at 6,800 ϫ g for 5 min to pellet cells. Cells were resuspended in 200 l of M-PER mammalian protein extraction reagent (Invitrogen). Samples were mixed with 4ϫ NuPAGE loading buffer (Invitrogen) and heated at 95°C for 10 min before centrifugation and loading onto NuPAGE Novex 4% to 12% Bis-Tris protein gels (Invitrogen) and run in 1ϫ MES Buffer at 200 V for 35 min. Transfer to membranes was performed using 0.2-m-pore-size PVDF precut blotting membranes (Invitrogen) with 1ϫ Transfer Buffer (Invitrogen).
Western blot analysis of mpTarget P1-2A-3C(L127P)-SGLuc constructs. Loading of cell lysate samples was normalized to luciferase readings from media harvested from mpTarget P1-2A-3C(L127P)-SGLuc transformations. Western blot analysis was performed to investigate the presence of processed VPs using F1412SA (anti-VP0 and VP2) mouse monoclonal antibody at a 1:50 dilution (32), an anti-VP3 rabbit polyclonal antibody at a 1:250 dilution (made by us), 6HC4.1.3 (anti-VP1) mouse monoclonal antibody at a 1:50 dilution (34), and 12FE9.2.1 (anti-VP1) mouse monoclonal antibody at a 1:50 dilution (33); all dilutions were in PBS-T. Blots were blocked in 5% milk for 1 h, followed by three 5-min washes with 1ϫ PBS-T buffer. Blots were incubated with primary antibodies for 1 h at room temperature, and then membranes were washed three times for 5 min each time with 1ϫ PBS-T buffer. Secondary antibodies, either goat anti-mouse HRP antibody (KPL) or goat anti-rabbit HRP antibody (KPL) at a 1:500 dilutions, were applied to membranes for 1 h at room temperature followed by three 5-min washes with 1ϫ PBS-T. Blots were developed with SigmaFAST 3,3=-diaminobenzidine (Sigma) for 1 h at room temperature followed by two washes with ddH 2 O. In order to detect VP2 in SAT2 samples, we used a nonnormalized blot with an increased amount of the L127P cell lysate.
Transmission electron microscopy (TEM). For TEM of cell cultures, transfected HEK293-T cells were grown in T-75 flasks and transfected with mpTarget P1-2A-3C(L127P)-SGLuc constructs. Transfected HEK293-T cells were incubated overnight at 37°C in 5% CO 2 prior to processing. Transformed bacterial cells were induced, cultured, and pelleted as described above prior to processing. For TEM processing, cells were fixed in 2% glutaraldehyde-NaHCa (Heuser's) buffer, postfixed with 1% tannic acid followed by 1% osmium tetroxide, stained en bloc with 4% uranyl acetate, embedded in 2% agarose, dehydrated through a graded series of acetone, and embedded in Spurr's resin (Electron Microscopy Sciences). Ultrathin (80-nm) sections were cut on a Leica UC6 microtome, stained with uranyl acetate and lead citrate, and imaged on a Hitachi 7600 analyzer with a 2-k-by-2-k AMT camera at 80 kV.
Cell-free protein synthesis and Western blotting. The bovine eIF4AI gene (GenBank accession no. 77735406) was synthesized by GenScript with the addition of an N-terminal FLAG tag and a C-terminal myc tag and cloned into pSNAP-tag (T7)-2 vector (New England BioLabs) using cut sites NdeI and NotI. Nucleotide sequences for 3C(wt), 3C(L127P), and 3C(C163A) with N-terminal FLAG tags were also cloned into pSNAP-tag (T7)-2 vector (New England BioLabs) using cut sites NdeI and NotI. Cell-free protein synthesis used a PURExpress in vitro protein synthesis kit (New England BioLabs) with the modification that two DNA plasmids were added in equimolar amounts. Western blots of cell-free synthesis products used anti-eIF4AI (ab31217, Abcam) antibody to detect eIF4AI and anti-DYKDDDDK (635691; TaKaRa) antibody to detect FLAG-tagged 3C.
Vaccination and challenge. Prior to conducting the cattle study, prior approval was obtained from the Plum Island Animal Disease Center Institutional Review Board and the Institutional Animal Care and Use Committee. Ten Holstein heifers, 6 to 7 months old, were randomly assigned to three groups as follows: group 1, buffer control cattle (n ϭ 2); group 2, 1 ϫ 10 9 PFU vaccine dose cattle (n ϭ 4); group 3, 1 ϫ 10 8 PFU vaccine dose cattle (n ϭ 4). Both the experimental vaccines and the control vehicle (final formulation buffer [Lonza]) were formulated with ENABL C1 adjuvant (VaxLiant). Cattle were challenged intradermolingually with 10 4 50% bovine infectious doses of FMDV O PanAsia-2 at 14 days postvaccination (dpv). Clinical FMD (pedal and oronasal epithelial lesions) was assessed by visual inspection at 3, 6, 10, and 14 days postchallenge (dpc). Serum samples collected at 0, 7, and 14 dpv and at 6 and 14 dpc were evaluated for virus neutralizing test (VNT) antibody titers against FMDV O PanAsia-2 and human adenovirus serotype 5 (Ad5) by serial dilution of heat-inactivated serum samples. VNT antibody titers were determined by a constant virus, decreasing serum microneutralization test. VNT titers were measured in BHK-21 cell cultures by the use of 100 to 150 50% tissue culture infective doses (TCID 50 ) of FMDV serotype O PanAsia/IBD-10/ARS-PAK/2010 and in HEK293 cells using 250 TCID 50 of Ad5. VNT titers were calculated by using the Spearman-Kärber method based on cytopathic effect (CPE). VNT titers of Ն0.9 log 10 were scored as positive (3). Plasma samples were collected at 0 to 5 dpc and were evaluated for viremia by real-time reverse transcription-PCR (rRT-PCR) (A positive C T [threshold cycle] value was Յ40) and FMDV isolation on LFBK-␣v␤6 cells (3). Protection from clinical FMD or viremia was analyzed statistically with Fisher's exact test.