In this study, we produced Wt and Dmt CLuc using Expi293F cells, silkworms, and BY-2 cells, designating the recombinant proteins CLuc_EX, CLuc_SW, and CLuc_BY, respectively. For clarity, we designated Dmt CLuc previously expressed using COS1 cells (Yasuno et al., 2018) as Dmt CLuc_CO in this study.

Expression vectors for recombinant Wt CLuc_EX and Dmt CLuc_EX, pcDNA-CL (Nakajima et al., 2004), and pcDNA-CL (T184A + N404D) (Yasuno et al., 2018) were prepared as previously described. The pcDNA-CL (T184A + N404D) was modified by the insertion of the sequence for His-tag and TEV protease (HHHHHHENLYFQG) following the predicted signal sequence (1–18 a.a.). Expression vectors for deletion mutants were designed with deleted amino-terminal or carboxyl-terminal residues of the mature region by modification of the His-TEV-pcDNA-CL (T184A + N404D) plasmid. These insertion and deletion modifications were outsourced to Genscript Japan. Each expression vector was transfected into Expi293F mammalian cells (Thermo Fisher Scientific, MA, United States) according to the manufacturer’s instructions. On day 5 post-transfection, the culture medium was centrifuged at 350 × g for 10 min, and the resulting supernatant was used for the following analysis, hereafter called the medium fraction. Precipitated cells were lysed with 10 mM Tris-HCl (pH 7.4) by sonication, the cell lysate was centrifuged at 20,000 × g for 5 min, and the resulting supernatant was used as the cell extract fraction. Immunoblot analysis for CLuc proteins was conducted as described previously (Yasuno et al., 2018), except that 10 μL of each medium fraction sample was prepared under reducing conditions with dithiothreitol before electrophoresis and that the anti-CLuc antibody was raised using purified recombinant CLuc in rabbit. The activities of the recombinant CLuc proteins were measured as described previously (Yasuno et al., 2018), and the intensities were obtained as relative light units (RLUs).

Dmt CLuc_EX was S-reduced and alkylated as described previously (Mitani et al., 2017). The protein was digested with a mixture of trypsin and Lys-C endopeptidase and/or chymotrypsin (TL/C/CTL). An aliquot of each digest was heated at 80°C for 2 h in 0.1% TFA to remove sialic acid and to increase the detection sensitivity of glycopeptides. Each digest was analyzed by an LC-MS system using a nanoflow LC (Ultimate 3000, Thermo Fisher Scientific) and Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). MS and MS/MS spectra were obtained by the Orbitrap analyzer with a resolution of 120,000 at 200 m/z. Data dependently selected precursor ions were fragmented by high-energy collision-induced dissociation (HCD) and HCD-fragment-triggered EThcD. Glycan peptide cluster analysis was performed by glycan-heterogeneity-based relational identification of glycopeptide signals on elution profile (Glyco-RIDGE; Togayachi et al., 2018). MS/MS spectra were searched using Mascot (ver. 2.5.1, Matrix Science, MA, United States) with the UniProt protein sequence database for Homo sapiens including sequences of CLuc (gi|41152712) with mutations (T184A + N404D). Search parameters were set as described previously (Mitani et al., 2017) with slight modifications; Hex(1) HexNAc(1)_Nli(ST) or Hex(1) HexNAc(1) NeuAc(2)_Nli(ST); Nli: neutral loss and ignore mass were set, enzyme; semi for each enzyme, and maximum missed cleavage for chymotrypsin; 6.

Recombinant CLuc_SW proteins were produced using the silkworm expression system as described previously (Maeda et al., 1985). CLuc ORF was amplified by PCR and inserted into the pHS02 vector under the control of a polyhedron promoter. The recombinant CLuc_SW was designed to be produced with FLAG-tag at its carboxyl terminus. Silkworm body fluid was collected from 20 individuals and subjected to ion exchange chromatography using the HiTrap Q HP column (GE Healthcare, IL, United States). The activities of elusion fractions using up to 500 mM NaCl were monitored by light emission activity using synthetic cypridinid luciferin as described previously (Yasuno et al., 2018). The active fractions were pooled and concentrated using a centrifuge filter unit (Amicon Ultra-15, Millipore, MA, United States), followed by desalting using a PD-10 column (GE Healthcare) according to the manufacturer’s protocol. The desalted sample was then further purified using anti-FLAG affinity gel (Merck, Germany).

Dmt CLuc_SW was purified by anion exchange chromatography using a MonoQ column (Cytiva, MA, United States). The purification was performed under the conditions of 20 mM HEPES-NaOH pH 7.5, 100–600 mM NaCl in a linear gradient. Finally, gel filtration chromatography using a Superdex 200 Increase column (Cytiva) was carried out under the conditions of 150 mM NaCl and 20 mM HEPES-NaOH pH 7.5 to make the final purified sample. The sample at this stage showed a single band in SDS-PAGE.

The Dmt CLuc_BY coding sequence was amplified by PCR from chemically synthesized cDNA with a Sma1_cLUC_ F (5′-ggg​ccc​ggg​ATG​AAG​ACT​TTA​ATA​CTC​GCT-3′) and Sal1_cLUC_R (5′-CCC​GTC​GAC​TCA​CTT​GCA​CTC​ATC​TGG​CA-3′) primer pair and cloned into a p35SHSPG vector (Oshima et al., 2011). Then, CaMV35Spro::Dmt CLUC::HSPter was transferred to a pBCKK T-DNA vector (Mitsuda et al., 2006) and transformed into Rhizobium radiobacter GV3101.

In a previous work, we reported the expression of recombinant CLuc deficient in N-glycosylation binding sites (Dmt CLuc_CO) in a mammalian adherent cell, COS1. In the present study, we first focused on Expi293F cells, which have been adapted to grow in high density and to give higher recombinant protein yields, and tried a suspension culture of the Expi293 expression system, with the expectation of high expression of CLuc. Each expression vector for Wt or Dmt CLuc_EX was transfected into Expi293F cells. After 5 days of incubation, the accumulation of CLuc_EX in the medium was confirmed by immunoblot analysis using an anti-CLuc antibody (Figure 1A). Mobility shift differences between Wt and Dmt CLuc_EX were presumably due to deficiencies in N-glycans. Over 90% of the luciferase activities were observed in the medium in both Wt and Dmt CLuc_EX, indicating that the majority of CLuc produced in Expi293F cells was secreted into the medium efficiently. The activity of CLuc_EX was determined by measuring the luminescence intensity using synthetic cypridinid luciferin. The specific activities of Wt and Dmt CLuc_EX in the medium were calculated by dividing the luminescence intensities by the estimated intensities of their immunoblot CLuc_EX bands (Figure 1B). The visibility of the immunoblot band differed among multiple experiments, and the estimation range was broad. Thus, we performed 3 independent experiments and calculated the average of the specific activity ratio for Dmt CLuc_EX relative to that for Wt CLuc_EX. A typical result is shown in Figure 1B, where the ratio was approximately 120%, but the average ratio was 90.7 ± 21.8%. These data showed that the specific activities of Wt and Dmt CLuc_EX were almost the same, unlike the previous result of approximately 20% for Dmt CLuc_CO expressed in COS1 cells.

Confirmation of N-glycosylation site mutations and search for O-glycosylation site(s) of Dmt CLuc_EX expressed in Expi293F cells by LC-MS.

Dmt CLuc_EX protein samples digested with chymotrypsin and/or a mixture of trypsin and Lys-C endopeptidase were analyzed by LC-MS/MS, and the obtained MS/MS spectra (HCD) were searched using Mascot to identify Dmt CLuc_EX-derived peptides (Supplementary Tables S1–S6). For the trypsin + Lys-C (TL) digest, 81% of peptides were identified, and a peptide containing mutation T184A was included. In addition, for the chymotrypsin digest, 87% of peptides were identified, including a peptide with an N404D mutation (Supplementary Tables S1, S3). These results indicate that both N-glycosylation sites were mutated and lost.

To search for other glycosylated site(s), EThcD MS/MS spectra were obtained by triggering with HCD fragments of m/z = 138, 204, or 366, that is, diagnostic fragment ions of glycan corresponding to [HexNAc-CH₆O₃]⁺, [HexNAc]⁺, or [HexHexNAc]⁺, respectively. Several O-glycosylated peptides, which were modified with glycan at a serine or threonine residue, were found by the Mascot search, peptide sequences were identified from HCD spectra, and glycosylated sites were assigned by EThcD spectra (Supplementary Tables S2, S6, Supplementary Figures S1–S5). In a TL digest, the HCD MS/MS spectrum of peptide sequence PPNTVPTSCEAK (residues 27–38) having a Hex(1) HexNAc(1) NeuAc(2) modification was found, presenting a series of diagnostic ions of glycan including a NeuAc-H₂O fragment (m/z = 274) and a ladder-like signal of glycopeptides suggesting a NeuAc-Hex-HexNAc-peptide (Supplementary Figure S1). A partial sequence of the core peptide was assigned. In an acid-treated digest with chymotrypsin + trypsin + Lys-C (CTL), glycopeptides having the same core sequence, PPNTVPTSCEAK, carrying 1 and 2 Hex(1) HexNAc(1) modification(s), were identified. A partial peptide sequence could be assigned from their HCD spectra, but the modification sites were not identified because there are 3 potential Ser/Thr in the peptide (Supplementary Figure S2). Glyco-RIDGE analysis of the acid-treated TL digest revealed a cluster of the core peptide having different glycan compositions, that is, members carrying glycans of Hex:HexNAc:dHex = 0:0:0 (no glycosylation), 0:1:0, 1:1:0, 1:2:0, 2:2:0, and 3:3:0 (data not shown). Extracted ion chromatograms are shown in Supplementary Figure S3 for the signals of 0:0:0, 0:1:0, 1:1:0, and 2:2:0. Signals of 1:2:0 and 3:3:0 were too weak to be seen in the figure. The signal of non-glycosylated peptide was highest even when the scale was reduced to 1/100, suggesting that the glycan occupancy of the peptide was quite low, about 2% of the area. The glycopeptide carrying Hex(1) HexNAc(1) showed twin broad peaks. The EThcD spectra of glycopeptide ions at 3 different retention times were compared (Supplementary Figure S4), revealing multiple O-glycosylated sites in the peptide sequence: Thr-30 and Thr-33 or Ser-34. MS/MS acquired at time (2) in Supplementary Figure S3 suggested that both Thr-30 and Thr-33 were glycosylated (Figure 2). The EThcD spectrum of the peptide having 2 Hex(1) HexNAc(1) was found, and Thr-30 had the Hex(2) HexNAc(2) modification (Supplementary Figure S5). In addition, another sequence EGECIDSSCGTCTR (residues 39–52) was found to have Hex(1) HexNAc(1); however, the site could not be assigned. The rate of O-glycosylation of the peptide was found to be quite low (<1/10⁴) from the area of the extracted ion chromatogram (data not shown).

Deletion mutants were constructed to remove the region around O-glycosylation residues at Thr-30 and Thr-33 (Figure 3A). Protein expression using the Expi293 expression system was confirmed by immunoblot analysis (Figure 3B), and specific activities were determined based on the immunoblot band intensity (Figure 3C). Specific activities were halved in mutants with deletions at residues 19–24 and 19–30 when compared to the full length of Dmt CLuc_EX. However, no activities were exhibited by the mutants with further deletions at residues 19–36, 19–42, or 19–48, even when the protein amounts were almost equal to those of the full-length protein. These results suggested that the amino acid residues at 31–36 (VPTSCE) were important for CLuc function. This region was highly conserved with the luciferase of the related cypridinid species, including cysteine residues (Supplementary Figure S6).

Moreover, the removal effects of carboxyl-terminal regions of Dmt CLuc_EX were investigated. A deletion mutant (19–314) designed to remove the second VWD-like domain exhibited no luciferase activity even though protein production was confirmed at the predicted size of 36 kDa (Figure 3B), suggesting that the residues at 315–553 were indispensable regions for CLuc function. Deletion of the residues at 453–553, which was designed to remove the carboxyl region behind the second VWD-like domain, resulted in a significant reduction of protein production, observed as a slight band around 50 kDa (Figure 3B). Deletion of the region at 453–553 might lead to the instability of the protein or to a reduction of secretion efficiency in this expression system.

The recombinant proteins of Dmt and Wt CLuc_SW were produced using the silkworm expression system. First, each recombinant protein was simply purified using anti-FLAG affinity gel, and a single band was observed for each case in the SDS-PAGE analysis (Figure 4). Depending on the protein, treatment with a reducing agent differently affected the mobility in the gel, as shown previously (Mitani et al., 2017). The Wt CLuc_SW band looked larger than the Dmt CLuc_SW band, and this difference was more obvious under the non-reducing conditions (Figure 4). This size difference would be due to the glycan modification, suggesting that an N-glycan chain would be added to CLuc when produced in silkworm, similar to the case with COS1 and BY-2 cell expressions, as previously reported (Mitani et al., 2017; Yasuno et al., 2018). The luciferase activities of Dmt and Wt CLuc_SW were 4.24 × 10⁷ and 1.67 × 10⁸ RLU ng⁻¹, respectively. These data indicated that almost 25% activity was retained in the Dmt CLuc_SW even without glycan modification. These data were similar to those obtained using a mammalian expression system, which resulted in approximately 20% activity for Wt CLuc_CO (Yasuno et al., 2018). We also produced another mutant, N182D + N404D, and expected an improvement in its specific activity. This mutant recombinant protein was successfully produced (Figure 4), but its activity was almost the same as that of Dmt CLuc_SW (4.18 × 10⁸ RLU ng⁻¹). Thus, we decided to use Dmt CLuc_SW for further analysis.

Next, to improve the recombinant protein purity, each sample was successively purified by FLAG-tag affinity chromatography and anion exchange chromatography. The results showed that Dmt CLuc_SW was separated into two major peaks (Supplementary Figure S7A). Considering the peak area ratio of the chromatograph, the existence ratio of these fractions was estimated to be about 1:3. The two separated peaks of Dmt CLuc_SW were collected, and each sample was subjected to gel filtration chromatography. The higher molecular weight fraction was eluted in the void volume, and the lower one was eluted in the volume corresponding to the molecular weight of the CLuc_SW monomer. Next, electrophoresis under native conditions was performed on each of these fractions. As shown in Supplementary Figure S7B, the Wt CLuc_SW and the Dmt CLuc_SW, which were derived from the lower molecular weight fraction in the gel filtration chromatography, showed a single band near the calculated molecular weight in each case, while the Dmt CLuc_SW, which was obtained as the void fraction in the gel filtration, showed a much higher molecular weight than the monomer. This aggregation is probably due to misfolding, since glycosylation is known to be involved in the correct folding of secreted proteins during translation (Wang et al., 2018). These aggregated Dmt CLuc_SW did not show any luciferase activity. These data seemed reasonable because only 25% of Dmt CLuc_SW was produced as a monomer, probably with normal activity, while the remaining 75% was aggregated during protein production, resulting in no luciferase activity. Thus, our first estimates of the specific activity of the Dmt CLuc_SW when purified only with anti-FLAG affinity were artificially low due to the presence of a high percentage of aggregated, non-functional luciferase. In the case of Dmt CLuc_CO expressed using COS1 cells, the specific activity was 20% of that of the Wt CLuc_CO (Yasuno et al., 2018). Although we did not perform further purification for this protein due to the limited amount of protein produced, 80% of the recombinant protein may have aggregated during protein production, as in the case of the silkworm expression system.

Wt CLuc_BY was efficiently expressed using cultured tobacco BY-2 cells (Mitani et al., 2017), and thus, we tried to produce Dmt CLuc_BY. However, none of the clones produced recombinant Dmt CLuc_BY even though we tested 4 independent BY-2 callus clones. This result suggested that the N-glycosylation of CLuc_BY was essential for its recombinant protein production in BY-2 cells.