The silkworm fourth instar larvae were purchased from Ehime Sansyu (Ehime, Japan) and reared with an artificial diet (Silkmate S2, Nosan, Japan) under a controlled environment (25°C, 65% ± 5% relative humidity) for 4∼5 days to reach fifth instar larvae for baculovirus infections. The cultured silkworm Bm5 cells were routinely passaged in Sf-900II medium (ThermoFisher Scientific Tokyo, Japan) supplemented with 10% fetal bovine serum (Gibco, Tokyo, Japan) and 1% antibiotic-antimycotic (ThermoFisher Scientific, Tokyo, Japan) at 27°C.

The synthesized VP2 gene of Canine parvovirus (NC_001539.1) by Genewiz (Suzhou, China) was amplified via polymerized chain reaction (PCR) using KOD-Plus-Neo DNA polymerase (Toyobo, Tokyo, Japan). The VP2 gene was subcloned to pFastBac1-SpT/SpC (Spy002 version, SpyTag002/SpyCatcher002) (Keeble et al., 2017) plasmids which were generated in our previous study (Xu et al., 2019). The variants for VP2 with inserted SpT or SnT (Genewiz, Suzhou, China) at Loop 2 (S226_G227) or Loop X (T391_G392) regions were constructed by inverse PCR using phosphorylated primers (T4 Polynucleotide Kinase, NEB, MA, United States) listed in Table 1. The transfer pFastBac plasmids, pFastbac-SpC-EGFP (Xu et al., 2019), pFastBac-SnC-mCherry, pFastBac-SpC-VP2, pFastBac-noSpT-VP2, pFastBac-NSpT-VP2, pFastBac-SpTL2-VP2, pFastBac-SpTLx-VP2, pFastBac-SpT_L2SnT_Lx-VP2, pFastBac-SnT_L2SpT_Lx-VP2, and pFastBac-∆StrepTag-SpT_L2SnT_Lx-VP2 for producing bacmid DNA were transformed into E. coli BmDH10Bac cells (Motohashi et al., 2005). This study used double fusion tags, His6-StrepTagII (HS) or FLAG-StrepTagII (FS), to facilitate the detection and purification among different constructs (Table 2). All the sequences were confirmed directly by DNA sequencing. The corresponding bacmid was subsequently transfected into either cultured Bm5 cells or silkworm fifth instar larvae using 1,2-dimyristyloxypropyl-3-dimethyl-hydroxy ethyl ammonium bromide and cholesterol (DMRIE-C, ThermoFisher Scientific K. K., Tokyo, Japan) transfection reagents (2 μg/1 μL bacmid DNA) or chitosan/bacmid system to obtain recombinant baculoviruses according to the protocols established previously (Kato et al., 2016; Xu et al., 2019). The series infection method obtained the stock of high titer viruses employed for infecting silkworm cells and larvae.

To obtain suitable candidates for protein display on VP2-derived VLPs, the three-dimension structural information was adopted (PDB: 4QYK) and visualized in the CLC Sequence Viewer (ver8.0, Qiagen). As already depicted in several previous reports, four loop regions, Loop 1 (aa V84–D99, at the top of protrusion: V92, N93), Loop 2 (R216–G235; H222–T228), Loop 3 (P295–I306; A300–F303), and Loop 4 (Y409–Y444; N421–N428, T433–N443), have been studied, among which aa 211–240 has been suggested to be not essential for VLP formation (Tsao et al., 1991; Hurtado et al., 1996; Xu et al., 2014). In this study, except for a reported Loop 2 (S226_G227, L2) surface loop region (Rueda et al., 1999), another simulated loop candidate (T391_G392, LoopX, Lx) was marked to the outer surface of VP2 protein (Xie and Chapman, 1996) and VP2-derived VLPs and further used for inserting SpT/SnT.

Expression and purification of all proteins in cultured Bm5 cells or silkworm larvae were performed according to our previous reports (Xu et al., 2019; Xu et al., 2022). Briefly, the fat body from silkworm larvae at 5 days post-infection (dpi) was resuspended and sonicated in a lysis buffer (100 mM Tris-HCl pH 8.4, 0.15 M NaCl, 1 mM EDTA, 0.1% NP-40, 1 × proteinase inhibitor (Roche, Tokyo, Japan)). The clear supernatants were then analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by either Coomassie brilliant blue (CBB) staining or western blotting. Protein samples were separated on 12% SDS-PAGE gel and then transferred onto a polyvinylidene fluoride (PVDF) membrane by a semidry transfer cell (Bio-Rad, Hercules, CA, United States). Subsequently, the membrane was blocked for 1 h in TBST buffer (0.1%v/v Tween 20 in Tris-buffered saline) containing 5% w/v skimmed milk (Wako, Tokyo, Japan) at room temperature. The washed membrane was then incubated with anti-StrepTagII primary antibody (1:5,000, anti-mouse, QIAGEN, Hilden, Germany), followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibody (1:10,000, GE Healthcare, Piscataway, NJ, United States). The final development was carried out using Immobilon western chemiluminescence HRP substrate (Merck Millipore, Darmstadt, Germany), which was visualized by VersaDoc MP imaging systems (Bio-Rad) for verifying expressed proteins.

After the expression was validated, protein purification was performed using fat body lysates from baculovirus-infected silkworm larvae (10∼20). Briefly, the lysates were centrifuged at 8,000 g for 30 min at 4°C to remove the insoluble material, followed by filtration with a 0.45-μm filter (Millipore). The × 10 diluted supernatants in a binding buffer (100 mM Tris–HCl pH 8.4, 0.15 M NaCl, 1 mM EDTA) were applied to a StrepTrap column (QIAGEN, Hilden, Germany). After × 10 washing, the target protein was eluted by a binding buffer containing 2.5 mM desthiobiotin (IBA Lifesciences, Germany). The purified proteins were concentrated by Amicon Ultra-15 (MWCO 30 kDa, Millipore) and buffer-exchanged in PBS buffer as demanded. The final protein concentration for each purified protein in this study was estimated by a BCA protein quantification kit (Bio-rad).

The recombinant baculoviruses stock was directly employed to infect/co-infect cultured cells in a 6-well plate or injected/co-injected into the silkworm larvae on the third day of the fifth instar. At 4 dpi, cultured cells or fat body tissues from 10 silkworm larvae were collected and lysed in a lysis buffer (100 mM Tris–HCl pH 8.4, 0.15 M NaCl, 1 mM EDTA in 0.1% NP-40) with complete EDTA-free protease inhibitor tablet (1 tablet/100 mL; Roche). The clear supernatant was subjected to SDS-PAGE and western blot to detect specific adducts after in vivo ligation in silkworms. The in vitro ligation assay between purified SpT/SpC- or SnT/SnC-tagged proteins was investigated in the mixture with the corresponding proteins partners under different ratios where a final 20 μL volume was adjusted with PBS buffer. The ligation reaction was performed at room temperature (∼25°C) or 4°C for an indicated time from 1∼24 h in PBS, where the VP2 proteins tested (NSp-VP2, SpTL2-VP2, SnTL2-SpTLx, and ∆StrepTag-SnTL2-SpTLx) are self-assembled into VLPs. The reaction samples mixed with 1 × SDS sampling buffer were heated in 99°C for 10 min before the SDS-PAGE assay to denature proteins and disturb VLPs. The ligation efficiency was estimated by investigating the formed adduct between binding partners from three independent CBB-stained SDS-PAGE.

The purified and concentrated proteins from VP variants were subjected to TEM to confirm the VLP formation and morphology following our previous publications (Suhaimi et al., 2019; Utomo et al., 2022). Briefly, a 20 μL protein solution drop was loaded onto the surface of a carbon film-supported copper grid (200 mesh, Nisshin Em Co. Ltd., Tokyo, Japan) and incubated at room temperature for 30 s. The grid was three-time washed with PBS and negatively stained with phosphotungstic acid (2% v/v). The immune-TEM was performed to investigate the possible VLP surface display in loop regions. The VLP sample on the grid was firstly blocked in 2% bovine serum albumin (BSA) for approximately 5 min, followed by wash and incubation with primary antibody (anti-His6, 1:30 in PBS, MBL, Nagoya, Japan) for 1 h in room temperature. After washing with PBS, the grid was washed six times and incubated with a secondary antibody of goat anti-rat IgG-conjugated 12 nm gold beads (1:50 in PBS, FUJI- FILM Wako Pure Chemical) for 1 h. Subsequently, the PBS-washed grid was negatively stained with phosphotungstic acid (2% v/v), which was visualized in the TEM apparatus (JEM-2100F, JEOL, Ltd., Tokyo, Japan). The DLS was used to analyze the size distribution of formed VLPs which was measured by the Zetasizer Nano series (Malvern Inst. Ltd., Malvern, United Kingdom).

The VLP formation of CPV can be obtained via the self-assembly of capsid VP2 protein in various protein expression systems, including BEVS with cultured insect Sf9 cells, silkworm Bm5 cells, silkworm larvae, and pupae (Choi et al., 2000; Feng et al., 2014; Jin et al., 2016; Chang et al., 2020). Previous literature has also found that the N-terminal genetically fused GFP can be used as a protein display strategy (Gilbert et al., 2006). Thus, as a continuous effort for our previous SpyBEVS (Xu et al., 2019), we designed several SpT-SpC chemistry constructs at terminal regions to establish protein display on CPV-LPs in silkworm-BEVS (SpyVLP-BEVS), which are illustrated in Figures 1A, B. As shown in Figure 1C, the VP2 with N-terminal SpT or SpC can be expressed in silkworm-BEVS, while SpC-VP2 (81.3 kD) has a lower expression level. We then purified SpT-fused VP2 (SpT-VP2, 70.8 kD) proteins from fat body lysates of recombinant baculovirus-infected silkworm larvae (∼3.5 mg/10 silkworm larvae, Figure 1C). The considerable amount of purified VP2 from silkworm-BEVS offered an excellent platform as a good candidate for further protein engineering.

Subsequently, we performed an in vitro binding assay between purified SpC-EGFP (43.8 kD) (Xu et al., 2019) and SpT-VP2 at 4°C overnight. We understand that the efficiency of the protein conjugation between SpC and SpT partners depends mainly on the conformation and location of SpC/SpT since the amidation itself is tolerant even to various harsh conditions (Zakeri et al., 2012; Veggiani et al., 2016; Xu et al., 2019; Wong et al., 2020). Previous results have reported that purified SpT-VP2 monomers expressed from silkworm-BEVS were efficiently self-assembled into VLP forms (Feng et al., 2011; Feng et al., 2014). Although the result from SDS-PAGE (Figure 1D) showed an invisible portion of the adduct after coupling, the result from western blot (Figure 1E) showed that only a tiny portion of SpC-EGFP was ligated to SpT-VP2 (∼115 kD). This means that the N-terminal SpT-tagged VP2-derived VLPs exposed SpT in a limited abundance, possibly due to the improper location and/or structural orientation of the presented SpT on the surface of the assembled VLPs. One can also notice the multiple bands for SpT-VP2 and adducts in Figure 1E, which might be caused by the VP2 protein degradations during the reaction in PBS buffer at room temperature, indicating that proteinase inhibitors may be required to prevent such degradations. The low display level is consistent with a previous study of N-terminal EGFP-fused VP2-derived VLP since only 20% (12 out of 60, N-terminal fused EGFP protrudes through the 5-fold axis cannon structures) display rate was simulated. A 16.7% (10/60) could be experimentally achieved on the VLP surface (Gilbert et al., 2006). Before further modifications of the location of SpT, we confirmed the VLP formation of the self-assembled SpT-VP2 in TEM (Figure 1F) and investigated the possible protein display by immuno-TEM using anti-His6 for SpC-EGFP (Figure 1G). The intact VLP structure and distribution (20–30 nm) were validated (Figure 1H) in this study, consistent with previous studies (Choi et al., 2000; Gilbert et al., 2006). This result provides direct evidence that VP2-derived SpT-SpC partner can be successfully expressed in silkworm-BEVS and be realized as a VLP display system. However, the protein display on SpT-VP2-derived VLP was not sufficiently achieved, as confirmed in Figure 1G, where no gold particles were found on VLPs, indicating that the location of SpT needs further optimization for better surface display.

Based on the structure of VP2 (PDB ID: 4QYK) and previous reports (Brown et al., 1994; Hurtado et al., 1996; Feng et al., 2011; Xu et al., 2014), we propose two loop sites, Loop 2 (S226_G227) and Loop X (T391_G392), as candidates for inserting SpT (Xie and Chapman, 1996; Rueda et al., 1999). Theoretically, these two loop regions are supposed to display at a 1: 1 ratio on the protein and assembled VLPs (60 per VLP), as shown in Figure 2A (1 ×, monomer) and 2B (5 ×, pentamer). Considering the peptide length of SpC (SpyCatcher002, 113 aa) and SpT (SpyTag002, 14 aa) (Keeble et al., 2017; Brune and Howarth, 2018), we propose that the shorter SpT peptide might be suitable to be inserted into the loop region of the VP2 protein without affecting its VLP formation. To test our strategy, we designed a series of chimeric VP2 variants without (noSpT, 68.1 kD) or with SpT at different locations, N-terminus (NSpT, 70.8 kD), Loop 2 (SpTL2, 70.5 kD), and LoopX (SpTLx, 70.5 kD) (Figure 2C). The expression level and correct molecular weight for each construct were verified in Bm5 cells. The adducts from several combinations for co-infection (virus ratio = 1:1) between each VP2 variant and SpC-EGFP were also validated in Figure 2D. As expected, the specific binding activity of each VP2 to SpC-EGFP was nicely achieved, indicating that the ligation occurred efficiently in cultured Bm5 cells. This conjugation reaction was in a relative expression level-dependent manner, where the remaining free SpT-VP2 or SpC-EGFP were significantly decreased in co-infection samples (Lanes 9–11) as compared to that in single infection samples (Lanes 2, 4–6).

We purified the VP2 variants from silkworm larvae for verifying in vitro binding and display assay. As demonstrated in Figure 3A, all purified proteins were obtained in decent amounts (mg scale per 10 silkworms) from only 10 silkworm larvae. However, there was a slight decrease in final yields for two loop variants (∼2.3 mg for SpTL2, ∼1.1 mg for SpTLx), possibly due to the relatively higher loss during Strep-column purification. To confirm whether the purified VP2 proteins could form VLPs correctly and efficiently, we conducted gentle purification of formed VLP from purified VP2 monomers where a standard 20% sucrose cushion was employed (Huhti et al., 2010; Kato et al., 2022). After ultracentrifuge at 122,000 × g, the fractions from the supernatant (monomers and nonVLPs) and precipitation (VLPs) were loaded to SDS-PAGE. It was shown in Figure 3B that the NSpT-VP2 and SpTL2-VP2 formed VLP mainly. However, no visible band of VLPs was detected in SpTLx-VP2 samples, indicating some problems for SpTLx-VP2 to form structurally correct VLPs after SpT was inserted into the LoopX region since no intact VLP formation was identified in TEM (Hurtado et al., 1996).

Subsequently, we continue to test if the formed SpT-decorated VLPs could be used as a carrier to further surface display SpC-fused proteins (SpC-EGFP) by in vitro mixing the two partners directly in a PBS-buffered solution. As demonstrated in Figure 3C, the specific ligation between SpT proteins (NSpT, SpTL2, and SpTLx) and saturated SpC-EGFP occurred as expected, where no shifted bands were observed under noSpT-VP2 conditions. Since we have already confirmed in Figure 3B that VLP formation is not sufficiently ensured in SpTLx-VP2, the adducts from the mixture of SpTLx-VP2 and SpC-EGFP should come mainly from the SpTLx-VP2 monomer other than formed VLPs. Afterward, we then focused on the SpTL2-VP2-derived VLPs. The purification using 20% sucrose cushion for SpC-EGFP mixed with NSpT or SpTL2 showed a significant increase in the VLP displaying SpC-EGFP in SpTL2 as judged by the ratio of the relative band intensity between bare VLP and EGFP-displaying VLP (EGFP-VLP) (Figure 3D). The display abundance, either in percentage (%) or number per VLP, was estimated from three independent binding assays (Figures 3E, 3F). As depicted in Figure 3G, protein display abundance was improved from 5.43% in NSpT-VP2 to 80.3% in SpTL2-VP2, which is 3–48 per VLP when calculated for the number of proteins. It indicates that the loop-inserted SpT-exposing to the surface of formed VLPs directly contributes to the significant enhancement of protein display on VLPs. The obtained results of the VLP formation for SpTL2-VP2 and the control from NSpT-VP2 validated by TEM are shown in Figure 3H.

To further realize chemically orthogonal protein display on SpT-VP2-derived VLPs, we introduced SnoopTag-SnoopCatcher (SnT-SnC) chemistry to SpTL2-VP2 as explained in Figure 4A for dual plug-and-display (Brune et al., 2017). Based on this concept, we generated a series of DNA constructs expressing His6-Strep (HS)-tagged VP2 proteins with SnTL2-SpTLx (#3, 72.7 kD), SpTL2-SnTLx (#4, 72.7 kD), HS-tag-free (∆StrepTag) VP2 with ∆StrepTag-STSnTL2-SpTLx (#5, 69.1 kD), as well as SpC-EGFP (#1, 43.8 kD) and SnC-mCherry (#2, 43.5 kD). The expression of each construct, except #5 in cultured Bm5 cells, was verified either in sole infection or co-infection (Figure 4B). The triple coinfections of #1,2,3 (SpC-EGFP, SnC-mCherry, and SnTL2-SpTLx; 160 kD) or #1,2,4 (SpC-EGFP, SnC-mCherry, and SpTL2-SnTLx; 160 kD) resulted in adducts (indicated by arrows) with higher molecular weight than #1,4 (116.5 kD), #2,3 (116.2 kD), #1,3 (116.5 kD), and #2,4 (116.2 kD) double coinfections (Table 2). To facilitate the purification of the antigen-displaying VLP, we further removed fusion tags from SnTL2-SpTLx-VP2, focusing more on the fusion tags from SpC/SnC-tagged antigens. With the HS-tag-free VP2 (#5, 69.1 kD), western blot analysis for samples of co-infection combinations in #1,5 (112.9 kD), #2,5 (112.6 kD), and #1,2,5 (156.4 kD) also showed specific VP2 adducts covalently ligated with either SpC-EGFP (#1), SnC-mCherry (#2), or both (#1 and #2), as indicated by arrows (Figure 4C). These results are similar to the results in Figure 4B. The above results from SpT/SnT double chemistry systems suggest that the ligation among SpT, SpC, SnT, and SnC occurs specifically and orthogonally in silkworm cell-based BEVS.

Aiming at better production of above each protein partner, we then designed to purify SpC-EGFP:SnTL2-SpTLx (#1,5) from the fat body lysates of 10 silkworm larvae. We infected silkworm larvae with a mixture of recombinant baculoviruses at a co-infection rate of about 1:2 (SpC-EGFP:SnTL2-SpTLx). Under this tuned condition, the purification of Strep-tagged SpC-EGFP proteins should ensure a majority of SpC-EGFP:SnTL2-SpTLx-VP2 present in the lysates and the final elution. As expected, the purification results from Figure 4D (left panel) proved that we successfully obtained the ligated adduct of SpC-EGFP:SnTL2-SpTLx-VP2 (VP2-EGFP) with a small fraction of SpC-EGFP presented in the final elution. The VLP assembly for VP2-EGFP-SnTLx (VLP-EGFP with free SnT displayed) was verified in the precipitation after ultracentrifugation through a 20% sucrose cushion (Figure 4D, right panel), which is beyond our expectation since the SpT-L2 inserted chimeric VP2 failed to self-assemble VLPs. To further test the in vitro binding capacity of SnT exposed to the VLP surface, we incubated purified VLP-EGFP-SnT and SnC-mCherry proteins for 48 h at 4°C, the results of which are shown in Figure 4E. Specific shifted bands with the highest molecular weight that appeared as confirmed either in CBB staining or western blot are considered as the SnC-mCherry and SpC-EGFP double-displayed VLP products. Moreover, the intact and efficient VLP formation for purified SnTL2-SpTLx-VP2 (bare VLP carrier) and SpC-EGFP:SnTL2-SpTLx-VP2 (SpC-EGFP displaying VLP) were validated by TEM (Figure 4F). Together with the purified VLP-EGFP from 20% sucrose cushion (Figures 4D, F), our study indicates that we successfully developed a VP2-VLP platform for possible multiple protein display based on orthogonal SnT-SnC and SpT-SpC chemistry.