AI-aided engineering and production optimization of chimeric hetero-viral capping enzymes for mRNA vaccines

Abstract

The 5’ cap structure exerts profound impacts on mRNA vaccine stability, translational efficiency, and evasion of innate immunity, encompassing key enzymatic components: triphosphatase (TPase), guanylyl transferase (GTase), and bifunctional methyltransfer-ases (N⁷G-MTase/2’-O-MTase). However, conventional multi-enzyme capping systems for mRNA vaccines are plagued by inherent inefficiencies. Herein, we report an AI-driven novel hetero-viral chimeric capping enzyme for mRNA vaccines, developed through bioinformatics mining of the DeepSeek and UniProt databases. This single-molecule biocatalyst integrates all capping functions by fusing a rabbit fibroma virus (RFV)-derived TPase/GTase domain with bifunctional methyltransferase domains from flavivirus (FV), respiratory syncytial virus (RSV), and rotavirus (RV). Following systematic tag-host optimization, a soluble and functional RFV-FV chimera was achieved via N-terminal maltose-binding protein (MBP) fusion in E. coli BL21(DE3) pLysS, whereas RFV-RSV and RFV-RV chimeras formed insoluble aggregates or were trapped in chaperone complexes. Complementary assays confirmed that the RFV-FV capping enzyme exhibited catalytic activities comparable to the commercial vaccinia virus-derived capping enzyme: TPase (96.76%±1.63%), GTase (100.69%±2.95%), and MTase (77.87%±3.85%). Our work presents a versatile cross-viral RFV-FV capping enzyme that enables large-scale prokaryotic production and ensures mRNA vaccine quality via its multifunctional catalytic activity, thereby streamlining mRNA vaccine manufacturing and establishing an AI-driven paradigm for designer enzyme engineering.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-47383-z.

Introduction

Messenger RNA (mRNA) vaccines have emerged as a prominent focus in biopharmaceuticals, owing to their promising applications in infectious disease prevention, tumor immunotherapy, and genetic disorder intervention^1–3. Their standardized manufacturing workflow encompasses antigen screening, in vitro mRNA transcription, 5’ capping, 3’ poly(A) tailing, and isolation-purification^4,5. A core advantage lies in harnessing host cellular machinery for antigen expression, requiring vaccine mRNA to exhibit structural characteristics consistent with endogenous eukaryotic mRNAs. Notably, the 5’ cap structure impacts mRNA stability, modulates translational efficiency, and enables evasion of innate immunity^6,7. Eukaryotic 5’ cap biogenesis proceeds via a four-step enzymatic cascade: (1) RNA triphosphatase (TPase) hydrolyzes precursor mRNA’s 5’-triphosphate to 5’-ppN (2) guanylyl transferase (GTase) transfers GMP from GTP to form 5’-GpppN (3) N⁷-methyltransferase (N⁷G-MTase) mediates guanine N⁷-methylation using SAM (4) 2’-O-methyltransferase (2’-O-MTase) methylates the initiating nucleotide’s 2’-O position⁸. Multienzyme systems suffer from mismatched kinetics and intermediate accumulation, increasing production complexity and costs⁹. Thus, developing a single-molecule enzyme integrating all capping activities is critical to addressing mRNA vaccine manufacturing bottlenecks.

In the prolonged coevolutionary arms race with hosts, viruses have evolved highly distinctive multifunctional 5’ capping enzymes¹⁰. Via natural domain integration, these enzymes condense the catalytic functions of multiple eukaryotic enzymes into a single molecule, serving as valuable templates for artificial tool enzyme design. For example, Vaccinia virus-encoded capping enzyme comprises two subunits (D1/D12): D1 integrates TPase, GTase, and N⁷G-MTase activities, while D12 enhances D1’s methyltransferase activity via allosteric modulation¹¹. In the Flavivirus (FV) family, West Nile virus non-structural protein 5 integrates N⁷G-MTase and 2’-O-MTase in a single module, with a temporally ordered catalytic cascade regulated by domain conformational switches^12–14. Notably, respiratory syncytial virus (RSV) L protein and rotavirus VP3 exhibit divergent cross-family N⁷G-MTase and 2’-O-MTase functional integration modes, highlighting viral enzymes’ evolutionary plasticity^15,16.

Despite these unique features, current research has notable limitations: it predominantly focuses on antiviral therapeutics targeting structural features^13,16, while enzymatic properties critical for mRNA vaccine manufacturing remain incompletely elucidated¹⁷. The dominant commercial vaccinia virus capping enzyme (D1/D12 heterodimer) lacks intrinsic 2’-O-MTase activity, requiring exogenous VP39 for Cap1 modification¹⁸. Additionally, long-held application patents, high licensing fees, and supply chain risks drive up production costs^19,20. Developing a novel capping enzyme with simplified architecture, integrated functions, and patent freedom is thus urgent to overcome mRNA vaccine manufacturing bottlenecks. Recent studies have scarcely investigated cross-family integration or structure-guided engineering, thereby impeding the advancement of “all-in-one” vaccine tool enzymes²¹. AI-assisted protein structure prediction tools empower rational protein design, functional modification, and cross-domain fusion^22–24. Accurate prediction of inter-domain arrangement, linker conformation, and active site exposure reduces randomness in chimeric enzyme engineering and improves functional integration efficiency²⁵.

Within the scope of this work, we selected the Rabbit fibroma virus (RFV)-derived TPase/GTase domain as the N-terminal scaffold, achieving single-molecule integration of all capping activities (TPase, GTase, and MTase) to obviate the need for auxiliary enzymes^10,25. AI-aided Uni-fold linker optimization mitigated steric clashes between cross-viral domains and ensured optimal exposure of the active sites^26–28. Following systematic tag and host system screening^29–32, we obtained soluble Rabbit fibroma virus-Flavivirus (RFV-FV) chimeric enzyme in the form of free multimers via the synergistic effects of MBP-tag and E.coli BL21(DE3) pLysS procaryotic cells. This patent-unencumbered enzyme exhibits catalytic activities comparable to those of commercial vaccinia virus capping enzymes, with values of TPase: 96.76%±1.63%, GTase: 100.69%±2.95% (no statistically significant differences were observed P > 0.05) and MTase: 77.87%±3.85%. we report a novel versatile hetero-viral RFV-FV facilitates large-scale prokaryotic expression and upholds mRNA vaccine modification through its integrated multifunctional catalytic properties, thereby streamlining mRNA vaccine manufacturing and establishing an AI-driven paradigm for designer enzyme engineering.

Materials and methods

Screening of viral capping enzymes

Screening of Viral Capping Enzymes Leveraging advances in AI-driven enzyme engineering and viral protein mining³³, we preliminarily curated virus-encoded multifunctional capping enzymes using two LLM platforms (https://chat.deepseek.com; https://www.doubao.com). To ensure reliable and reproducible screening outcomes, all candidates underwent with pre-defined inclusion criteria: (1) Experimentally validated capping-associated functional domains by PubMed and UniProt (https://www.uniprot.org) functional analysis (2) Distinct viral family representation for phylogenetic diversity (3) Resolved crystal/cryo-EM structures in PDB (https://www.rcsb.org) (4) Prokaryotic heterologous expression feasibility. Amino acid and nucleotide sequences of selected enzymes were retrieved from NCBI (https://www.ncbi.nlm.nih.gov).

Prediction and visualization of engineered chimeric enzymes structures

Hetero-viral engineered chimeric enzymes 3D structural modeling was conducted using Uni-Fold (https://www.bohrium.com/apps/uni-fold) that is a high-precision protein 3D structure prediction platform enhanced and refined based on Google DeepMind’s AlphaFold. Chimeric amino acid sequences were submitted to the Bohrium Uni-Fold workflow with default parameter settings. Following 3D modeling of the engineered chimeric enzymes, the resultant structural models were visualized using ChimeraX-1.6.1 software to delineate critical structural characteristics.

Assessing the reliability of optimized models for hetero-viral chimeric enzymes

The quality of optimized predictive models was evaluated using the UCLA-DOE LAB SAVES V6.1 platform (https://saves.mbi.ucla.edu). Ramachandran plots (PROCHECK) mapped all protein residues’ φ/ψ dihedral angles (x-/y-axes); models were validated if > 90% of residues fell within the most favored/allowed regions (key φ/ψ signatures: α-helices − 57°/−47°, parallel β-sheets − 119°/113°, antiparallel β-sheets − 139°/117°)³⁴. ERRAT quantified noncovalent interaction errors (residues x-axis, error values y-axis). values exceeding the 95% confidence interval indicated reduced structural rationality. The Overall Quality Factor (OQF) is the percentage of residues with error values below the 95% threshold: ≥95% (comparable to highly refined native structures), about 91% (no obvious conformational defects), ≥ 80% (suitable for preliminary structures)³⁵.

Expression of chimeric capping enzymes

Genes encoding novel chimeric capping enzymes were synthesized by GenScript Biotechnology and the gene fragments were cloned into fusion-tagged expression vectors Via homologous recombination. Prokaryotic expression constructs were heat-shocked into E. coli competent cells (Rosetta (DE3), Transetta2(DE3), BL21(DE3) pLysS, pGro7/BL21). Colonies were inoculated into 50 mL LB medium (with antibiotics) and cultured overnight (37 °C, 200 rpm), then diluted 1:50 (v/v) into 500 mL TB medium (with antibiotics) and grown to OD₆₀₀ 0.6–0.8 (37 °C, 200 rpm), followed by induction with 0.2 mM IPTG (16 °C, 200 rpm, 20 h). For pGro7/BL21, 2 mg/mL L-arabinose was added at OD₆₀₀ 0.2–0.4 to induce GroEL chaperone, then 0.2 mM IPTG induction (16 °C, 200 rpm, 20 h) at OD₆₀₀ 0.6–0.8. Eukaryotic pPICZA was chemically transformed into Pichia pastoris X33 competent cells. Colonies were inoculated into 2 mL YPD (28 °C, 200 rpm) to OD₆₀₀ 2–8, diluted 1% (v/v) into 10 mL BMGY (28 °C) to OD₆₀₀ 2–6, centrifuged, resuspended in 500 mL BMMY (100 µg/mL Zeocin™, 2 L flasks), and induced (28 °C, 160 rpm, 5 d) with 0.5% methanol supplementation every 24 h.

Purification of chimeric capping enzymes

Chimeric capping enzyme-expressing cells were resuspended in lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.5% Triton X-100, 1 mM PMSF). Cells were lysed by high-pressure disruption (1400 bar); lysates were centrifuged (4 °C, 15000 rpm, 40 min) and filtered through a 0.22 μm membrane. His-tagged proteins on a HisTrap HP 5 mL column eluted with buffer A1: 50 mM Tris-HCl, 500 mM NaCl, 5% glycerol, pH 8.0 containing 300 mM imidazole. MBP-tagged proteins on an MBP Trap HP 5 mL column eluted with buffer A2: 50 mM Tris-HCl, 300 mM NaCl, pH 8.0 containing 10 mM maltose. GST-tagged proteins on a GST Trap HP 5 mL column eluted with buffer A3: 50 mM Tris-HCl, 300 mM NaCl, pH 8.0 containing 10 mM reduced glutathione. Size-exclusion chromatography (SEC) was performed on a HiLoad 16/60 Superdex 200 pg column (Cytiva, 20 mM Tris-HCl pH 7.5/150 mM NaCl) at 4 °C (1 mL/min), and eluted fractions were analyzed by SDS-PAGE for purity and molecular weight confirmation.

Western blotting analysis

Target protein expression was analyzed by Western blotting. Equal 2 µL volumes of uninduced and IPTG‑induced cell lysates were denatured in Laemmli buffer at 95 °C for 5 min and separated by 10% SDS‑PAGE at 120 V. Proteins were transferred to a PVDF membrane at 100 V for 2 h at 4 °C in Tris‑glycine buffer. The membrane was blocked with 5% non‑fat milk in TBST for 1 h at room temperature, then washed three times with TBST (5 min each). It was incubated with HRP‑conjugated anti‑6×His monoclonal antibody (ABclonal, Cat# AE104) diluted 1:5000 in TBST containing 3% non‑fat milk for 1 h at room temperature, then washed five times with TBST (5 min each). Protein bands were visualized using ECL substrate after 2–3 min incubation at room temperature, and the membrane was exposed to X‑ray film for 3–5 min in the dark.

Quantification of TPase activity

TPase activity was quantified by detecting inorganic phosphate (Pi) release from RNA substrates using the Malachite Green Phosphate Assay Kit (Sigma-Aldrich, MAK370). Experiments were performed in triplicate, with results presented as mean ± SD. A Pi standard curve (0–40 µM) was generated by serial dilution of 1 mM Pi in 1× reaction buffer; 80 µL of each standard and 20 µL kit working reagent were added to a 96-well plate, incubated 10 min at 25 ± 1 °C in the dark, and absorbance at 620 nm (OD₆₂₀) measured. Sample Pi concentrations were calculated via standard curve linear regression. The 40 µL TPase reaction system contained 20 µM pppG-12mer (Sangon Biotech), 19 µL 1× reaction buffer, and 1 µL enzyme (1 µg/µL). Controls included: (1) positive Vaccinia Capping Enzyme (Cat No.2460 A); 2’-O-Methyltransferase VP39 (Cat No.2470 A; Takara Bio) (2) negative reaction buffer (3) blank ultrapure water. Reactions were incubated 10 min at 25 °C, terminated with 10 µL chromogenic working reagent, incubated 30 min, and A₆₂₀ measured. TPase activity (pmol·min⁻¹·µg⁻¹ enzyme) was calculated after subtracting background absorbance from negative/blank controls.

Determination of kinetic parameters

Kinetic parameters, including the Michaelis constant (Kₘ), maximum reaction velocity (Vmax), turnover number (kcat), and catalytic efficiency (kcat/Kₘ), were determined by measuring initial reaction rates across a range of 12nt-mRNA substrate concentrations from 0.5 to 20 µM. The absorbance at 620 nm (A₆₂₀) was monitored, and the molar concentration of the reaction product Pi was calculated using a Pi standard curve. The concentration of 12nt-mRNA substrate was set as the abscissa, while the concentration of the product Pi was used as the ordinate. All kinetic data were subjected to nonlinear regression fitting according to the Michaelis–Menten equation using GraphPad Prism 9.0. The turnover number (kcat) was calculated using the equation: kcat = Vmax / [E]. The [E] represents the concentration of active enzyme used in the assay (1 µg/µL, equivalent to 7.84 µM). Catalytic efficiency was determined as the ratio of kcat to Kₘ (kcat/Kₘ).

Quantification of Guanylyl-transferase (GTase) activity

GTase activity was quantified by detecting pyrophosphate (PPi) released during RNA capping, using the EnzChek^® Pyrophosphate Assay Kit (Thermo Fisher, E-6645). Experiments were run in triplicate (results: Mean ± SD), with statistical analyses performed in GraphPad Prism 9.0 (one-way ANOVA; P < 0.05 = significant). A PPi standard curve (0, 1, 2, 5, 10, 20, 40 µM) was generated by serially diluting 1 mM PPi in 1× GTase reaction buffer. To a 96-well plate (Corning^® 3915), 80 µL assay mix and 20 µL standard were added, incubated at 25 ± 1 °C (10 min, dark), and A_360nm was measured (R²≥0.99 for quantification). The 20 µL GTase system contained 40 µM 5’-triphosphate RNA, 1 µL 0.5 mM GTP (Sigma-Aldrich, G8877), 18 µL 1×buffer, and 1 µL purified enzyme with the concentration 1 µg/µL. Positive controls (VV cap enzyme GTase standard), Negative controls (no enzyme), substrate (1 µL ultrapure water). Reactions (37 °C, 10 min) were terminated with 3µL 0.5% trifluoroacetic acid (5 min, RT), assayed for A_360nm, and activity expressed as PPi production (nmol·min⁻¹·ug⁻¹).

Assay of methyltransferase (MTase) activity

MTase activity was quantified by detecting S-adenosylhomocysteine (SAH) produced from S-adenosylmethionine (SAM) during RNA capping, using MTase Glo™ Kit (Promega, V7601) on a GloMax Navigator luminometer (Promega, USA). Experiments were run in triplicate (mean ± SD). Statistics: GraphPad Prism 9.0 (one-way ANOVA, Dunnett’s test); P < 0.05 significant. An SAH standard curve (0, 0.05, 0.25, 0.75, 1.00 µM) was generated by diluting 1 µM SAH in 1×buffer. To a 96-well white plate (Corning^® 3912), 20 µL standard and 5 µL 5×MTase-Glo™ Reagent were added, incubated at 25 ± 1 °C, 30 min. After adding 25 µL Detection Solution RLU was measured (R² ≥ 0.99). The 20 µL MTase system contained 40 µM RNA, 1 µL 2 mM SAM, 1 µL 10 mM GTP, 17 µL 1×buffer, and 1 µL purified enzyme with the concentration 1 µg/µL. Controls: positive (vaccinia MTase standard), negative (1×buffer), substrate (1 µL ultrapure water). Reactions (37 °C, 10 min) were terminated with 2 µL 0.5% TFA (5 min, RT). SAH was quantified as above, activity = SAH production (nmol·min⁻¹·µg⁻¹).

Results

AI-assisted enzyme engineering elucidates well-defined multifunctional domain architecture in hetero-viral chimeric capping enzymes

The 5’ capping of mRNA vaccines mimics the natural 5’ cap of eukaryotic mRNA, facilitating site-specific enzymatic modification to enhance mRNA stability and translational efficiency (Fig. 1A). Eukaryotic mRNA capping involves three steps: (1) triphosphatase hydrolyzes the 5’-terminal triphosphate (2) guanylyl transferase transfers GMP from GTP to the 5’-diphosphate (3) guanine-N7 and 2’-O-methyltransferases with S-adenosylmethionine as donor respectively to forming the mature cap (Fig. 1B).

Fig. 1.

AI-Assisted enzyme engineering elucidates well-defined multifunctional domain architecture in hetero-viral chimeric capping enzymes. (A) mRNA vaccine core architecture: 5’ cap (red box), antigen ORF (blue box), 3’ poly(A) tail (green box). The enlarged high‑resolution chemical structure of the 5’ cap is shown within the red box. (B) Eukaryotic mRNA 5’-capping cascade: sequential multi-enzymatic reactions catalyzed by TPase, GTase and MTase. (C) Surface representation of rabbit fibroma virus (RFV) enzyme domains: TPase (1–233aa, purple), GTase (234–537aa, green); remaining regions in gray. (D) Surface representation of Flaviviridae (FV) Nsp5 MTase domains: N⁷G-MTase (1–222aa, beige color), 2’-O-MTase (223–300aa, magenta). (E) Surface representation of rotavirus (RV) VP3 MTase (181–560aa) domains: N7G-MTase (brown), 2’-O-MTase (pink). (F) Predicted surface representation of the RFV-RV chimeric enzyme: TPase (purple), GTase (green), MTase (golden yellow, light pink). (G) Predicted surface representation of the RFV-FV chimeric enzyme: MTase (beige, magenta). (H) Predicted surface representation of the RFV-RV chimeric enzyme: MTase (brown, pink color).

To engineer a multifunctional chimeric capping enzyme for mRNA vaccine production, we screened DeepSeek and identified 11 viral candidates (S Fig. 1 and S Table 1). We found Rabbit Fibroma Virus have a conserved N-terminal 1–537 aa segment with triphosphatase (TPase) and guanylyl transferase (GTase) domains (Fig. 1C). Multifunctional methyl-transferase domains were found in other viruses: West Nile Virus (Flaviviridae) NsP5 (1–300 aa) has N⁷G- and 2’-O-methyltransferase activities (Fig. 1D). Rotavirus VP3 (181–560 aa) also has dual MTase functions (Fig. 1E). RSV L protein (1765–2165 aa) has two MTase domains. We constructed “all-in-one” enzymes by fusing RFV’s TPase/GTase scaffold with the above MTase domains, using flexible (GGGGSGGGGS) or rigid linkers. This integrated all activities into one molecular enzyme, generating RFV-RSV, RFV-FV, RFV chimeric capping enzyme (Figs. 1F–H).

Ramachandran plots were generated to assess the stereochemical plausibility of backbone conformations in three engineered and optimized chimeric enzyme models, with 99.7%, 99.1%, and 99% of non-Gly/Pro residues localized to the Most Favored and Additional Allowed Regions (S Figs. 2A-C), confirming adherence to the stereochemical rules of naturally folded proteins and validating structural plausibility. Furthermore, the three models yielded core ERRAT Overall Quality Factors (OQF) of 91.774, 88.113, and 87.356, respectively (S Figs. 3A-C). Over 80% of residues had error values below the 95% confidence interval threshold. This confirms that the atomic noncovalent interactions of these optimized models align with those of naturally folded high-quality proteins and validates the structural integrity of the engineered constructs.

Fig. 2.

RFV-FV capping enzyme with His-tags displayed expression in diverse cells but assembled into inclusion bodies or chaperone complexes. (A-B) Schematic diagram of the structure of the histidine-tagged RFV-FV chimeric capping enzyme and characteristics of the recombinant expression vector (C) Electrophoresis results of the RFV-FV chimeric capping enzyme purified by affinity chromatography (D) Purification of RFV-FV by SEC and electrophoresis results (E) Characteristics of the eukaryotic expression vector for Pichia pastoris (F-G) SDS-PAGE and Western Blot analysis of the eukaryotically expressed RFV-FV chimeric capping enzyme after affinity purification.

Fig. 3.

Soluble production of an RFV-FV chimeric capping enzyme enabled by an MBP tag and E. coli BL21(DE3) pLysS strains. (A) Schematic diagrams illustrating the fusion of the MBP tag and the GST tag to the N-terminus of the RFV-FV enzyme. (B–C) Schematic diagrams of the MBP- and GST-tagged recombinant vectors. (D) SDS-PAGE analysis showing the affinity purification of MBP-tagged RFV-FV enzyme in E.coli Rosetta (DE3) cell. (E) SDS-PAGE analysis demonstrating the affinity purification of MBP-tagged RFV-FV enzyme in E.coli pGro7/BL21 (DE3) cell, which co-expresses GroEL/GroES chaperones to enhance protein folding. (F) SDS-PAGE analysis showing the SEC purification of MBP-tagged RFV-FV enzyme in E.coli pGro7/BL21 (DE3) cell, which co-expresses GroEL/GroES chaperones to enhance protein folding. (G) SDS-PAGE analysis of the affinity purification process of MBP-tagged RFV-FV enzyme expressed in E.coli BL21(DE3) pLysS cell, which reduces basal expression of toxic proteins via T7 lysozyme. (H) SDS-PAGE analysis of the SEC-purified fractions of MBP-tagged RFV-FV enzyme from E.coli BL21(DE3) pLysS, confirming the homogeneity of the purified recombinant enzyme.

RFV-FV capping enzyme with His-tags displayed expression in diverse cells but assembled into inclusion bodies or chaperone complex

To achieve soluble expression of recombinant RFV-FV chimeric capping enzyme, we first constructed a prokaryotic vector harboring the N-terminal His-tagged target gene (Fig. 2A and B). The vector was transformed into E. coli Transetta2 (DE3) and Rosetta (DE3) competent cells for expression, followed by purification via His-Trap affinity chromatography. For both strains, no 280 nm UV absorption peak (specific to the target) was detected in eluates (S Figs. 4A and B). SDS-PAGE showed a prominent ~ 100 kDa band concentrated in inclusion bodies, matching the theoretical molecular weight 99.7 kDa (Fig. 2C, S Figs. 5 and S Fig. 4C). These data confirmed His-tagged RFV-FV was exclusively expressed as insoluble inclusion bodies in these E. coli strains. Then we used E. coli pGro7/BL21(DE3) that co-expresses GroES–GroEL chaperones to enhance solubility. The recombinant protein was purified via a two-step strategy: affinity chromatography (~ 1000 mAU 280 nm) followed by SEC and SDS-PAGE showed ~ 100 kDa (RFV-FV) and ~ 60 kDa (chaperone) bands (Fig. 2D, S Fig. 4D and S Fig. 6), indicating soluble RFV-FV multimers that formed stable chaperone complexes.

To leverage eukaryotic post-translational modifications, we expressed RFV-FV in Pichia pastoris X33 using the pPICZA vector (Fig. 2E). Whole-cell SDS-PAGE and whole-cell western blot confirmed a ~ 100 kDa target expression band throughout 12–96 h induction, with significant upregulation at 24 h (Fig. 2F, G and S Figs. 7 and 8). Affinity chromatography of Pichia pastoris X33 lysates detected a 280 nm peak in 10 mM imidazole eluates, but no clear ~ 99 kDa band was observed (S Figs. 9 A and 9B). Thus, RFV-FV was successfully expressed in Pichia pastoris X33 but exhibited instability and proneness to degradation.

Soluble production of an RFV-FV Chimeric capping enzyme enabled by an MBP tag and E. coli BL21(DE3) pLysS strains

Owing to the inherent limitations of His-tag fusion in promoting protein solubility, this study focused on optimizing the soluble expression of the RFV-FV chimeric capping enzyme. The N-terminal maltose-binding protein (MBP) and Glutathione S-Transferase (GST)-tagged RFV-FV recombinant vectors were transformed into three E. coli strains Rosetta (DE3), pGro7/BL21(DE3), and BL21(DE3) pLysS cells respectively (Figs. 3A-C). The solubility and expression profiles of RFV-FV were systematically analyzed using affinity purification and SDS-PAGE. In E. coli Rosetta (DE3) cells, neither affinity chromatography nor SDS-PAGE detected soluble RFV-FV in eluates (Fig. 3D, S Figs. 10 A and 10B). Instead, SDS-PAGE of inclusion body fractions revealed a dominant 139 kDa band that matching RFV-FV’s theoretical molecular weight (Fig. 3D and S Fig. 10B). These results indicating that the protein was mainly sequestered in inclusion bodies.

In pGro7/BL21(DE3) cells, affinity chromatography showed a distinct elution peak at 280 nm about 2000 mAU, and SDS-PAGE confirmed a 139.2 kDa band corresponding to RFV-FV (Fig. 3E, S Figs. 11 A and 11B). SEC analysis revealed a ~ 1000 kDa peak at 50.12 mL and SDS-PAGE of this fraction detected both RFV-FV and 60 kDa chaperone, confirming soluble RFV-FV/GroES–GroEL chaperone complex (Fig. 3F and S Fig. 12). In E. coli BL21(DE3) pLysS cells, affinity chromatography exhibited a clear elution peak about ~ 600 mAU, with SDS-PAGE verifying the 139.2 kDa RFV-FV band (Fig. 3G, S Figs. 13 A and 13B). SEC analysis revealed a single symmetric elution peak at 51.24 mL, while SDS-PAGE analysis confirmed a single dominant band at 139.2 kDa, collectively confirming that RFV-FV exists as free soluble multimers (Fig. 3H and S Fig. 14). To systematically quantify RFV-FV purity across all purification stages, SDS-PAGE electrophoresis coupled with ImageJ software densitometry was employed. Densitometric quantification revealed a final purity of RFV-FV of 78.46 ± 0.66%, which satisfies the requirements for subsequent enzymatic activity characterization (Fig. 3H and S Table 2). To expand the strategy repertoire, a GST-tagged RFV-FV construct was also transformed into E. coli Rosetta (DE3) and pGro7/BL21(DE3). However, neither affinity purification nor SDS-PAGE detected soluble GST-RFV-FV ~ 123 kDa in eluates (S Figs. 15 A-C).

Collectively, soluble production of the RFV-FV chimeric capping enzyme was enabled by the MBP tag and E. coli BL21(DE3) pLysS strain. In contrast, GST tagging and expression in Rosetta (DE3) or pGro7/BL21(DE3) failed to yield soluble RFV-FV chimeric capping enzyme. This finding lays a robust foundation for large-scale preparation and subsequent functional characterization of the chimeric enzyme.

RFV-RSV and RFV-RV capping enzymes with various tags are expressed in different cell but form inclusion bodies or chaperone complex

To broaden the research scope, this study systematically investigated RFV-RSV and RFV-RV chimeric capping enzymes’ expression in prokaryotic and eukaryotic host cells. For RFV-RSV enzyme’s soluble expression and purification, three N-terminally tagged (His, MBP, GST) recombinant vectors were constructed and transformed into E. coli Rosetta (DE3) and pGro7/BL21 (DE3) (Fig. 4A). Following transformation of His-tagged RFV-RSV into pGro7/BL21(DE3), Ni-NTA affinity purification yielded about 110 kDa band that matched its theoretical molecular weight of 112.5 kDa, confirming specific expression (Fig. 4B, S Figs. 16 A and 16B). SEC purification showed an elution peak at 50.2 mL (~ 600 kDa), and SDS-PAGE analysis confirmed that the Elution1 contained the ~ 110 kDa enzyme and ~ 60 kDa GroEL (Fig. 4C, D and S Fig. 17). The enzyme was soluble but formed a chaperone complex. MBP-tagged RFV-RSV vector was transformed into both E. coli Rosetta (DE3) and pGro7/BL21 (DE3) strains. Post-induction, no target protein was detected in eluates, but a ~ 150 kDa band that matching MBP-fusion’s 152 kDa existed in inclusion body pellets (Fig. 4E and S Figs. 18 A and 18B). GST-tagged vector (Fig. 4F) expressed in Rosetta (DE3) showed no ~ 134.8 kDa band (S Figs. 19 A and 19B), indicating GST-tag abrogated expression, possibly due to conformational conflicts.

Fig. 4.

RFV-RSV and RFV-RV capping enzymes with various tags are expressed in different cell but form inclusion bodies. (A) Schematic diagram of the His/MBP/GST-tagged RFV-RSV domain (B) Electrophoretic analysis results of His-tagged RFV-RSV following affinity purification (C-D) Electrophoretic analysis results of His-tagged RFV-RSV after purification by SEC (E) Electrophoretic analysis of MBP-tagged RFV-RSV following affinity purification (F) Schematic diagram of the domain architecture of N-terminal His-tagged RFV-RV (G-H) Western Blot and SDS-PAGE analysis results of His-tagged RFV-RV after expression in the eukaryotic yeast Pichia pastoris X33 cells (I) Electrophoretic analysis results of His-tagged RFV-RV following expression in Pichia pastoris X33 cells and subsequent affinity purification.

His-tagged RFV-RV was transformed into Rosetta (DE3) and pGro7/BL21(DE3) strains. For these two strains, 280 nm UV absorption peaks were detected in eluates at ~ 680 mAU and ~ 2000 mAU, respectively (S Fig. 20A and B). SDS-PAGE analysis revealed a prominent ~ 110 kDa band concentrated in pellets—matching its theoretical molecular weight of 110.9 kDa—whereas no target protein was found in eluates, confirming the formation of insoluble inclusion bodies (S Figs. 20 C and 20D). His-tagged RFV-RV enzyme expressed in Pichia pastoris X33 cells induced with 1% methanol (12–96 h) exhibited 110.9 kDa bands (Figs. 4G-H, S Fig. 21 and S Fig. 22). Lysates from cells induced for 24 h showed an elution peak with 10 mM imidazole (~ 700 mAU at 280 nm) but no target band in the eluted fractions (S Figs. 23 A and 23B). A band about 110 kDa was detected, confirming that RFV-RV formed insoluble inclusion bodies (Fig. 4I and S Figs. 23B). Collectively, RFV-RSV and RFV-RV bearing different tags were expressed in different host cells but formed inclusion bodies or chaperone complexes.

The TPase activity of the RFV-FV chimeric capping enzyme is comparable to that of the commercial vaccinia virus capping enzyme

Based on previous findings, we characterized the in vitro mRNA capping activity of the RFV-FV chimeric enzyme via product quantification of TPase, GTase, and MTase. TPase hydrolyzes the 5’-triphosphate moiety of mRNA to release inorganic phosphate (Pi), which forms a green ternary complex with malachite green and ammonium molybdate that exhibits strong absorbance at 620 nm (Fig. 5A). A standard curve plotting A₆₂₀ against Pi concentration demonstrated excellent linearity (Y = 0.02027X − 0.01054, R²=0.9985, 95% confidence interval), confirming the robustness of the assay (Fig. 5B and S Table 3). For activity quantification, we optimized RNA substrate concentration and reaction time using a commercial vaccinia virus (VV) cap enzyme as the positive control, which contains the D1/D12 and VP39. Concentrations of 8.0–20 µM elicited a dose-dependent increase in activity, with a significant elevation at 20 µM (P < 0.05), which was thus identified as the optimal substrate concentration (Fig. 5C and S Table 4).

Fig. 5.

The TPase activity of the RFV-FV chimeric capping enzyme is comparable to that of the commercial vaccinia virus capping enzyme. (A) Schematic diagram of the principle for enzyme activity detection via the malachite green-ammonium molybdate colorimetric method. (B) Standard curve of inorganic phosphate concentration versus absorbance. (C) Optimization of RNA substrate concentration for the enzymatic reaction using commercial VV capping enzyme as the positive control, showing absorbance values at different RNA substrate concentrations. (D) Optimization of reaction time for TPase activity using commercial VV capping enzyme as the positive control. TPase activity A₆₂₀ was measured at 25 °C at the indicated times with 20 µM pppG‑12mer. Data are Mean ± SD (n = 3 independent experiments); P > 0.05, not significant. (E) Quantitative analysis of the absorbance of RNA triphosphatase reaction product complexes for the RFV-FV cap enzyme. (F) Catalytic reaction rates of different enzymes calculated from the standard curve. (G) Evaluation of the reaction rate of the RFV-FV cap enzyme using the reaction rate of commercial vaccinia virus as the 100% standard.

For reaction time, we measured the TPase activity of the VV enzyme at 25 °C across different reaction times, and significant differences were observed between 5 min and the 10–30 min groups (P < 0.0001), whereas no significant differences were detected among 10, 20, and 30 min (10 vs. 20 min: P = 0.215; 10 vs. 30 min: P = 0.2511). Therefore, 10 min was selected as the optimal reaction time (Fig. 5D and S Table 5).

TPase Activity quantification results show that the Negative control exhibited minimal absorbance (0.033 ± 0.011), but the Positive VV cap (0.223 ± 0.012) and RFV-FV (0.220 ± 0.009) enzyme showed significantly higher absorbance (P < 0.0001) (Fig. 5E and S Table 6). Specific TPase activities of VV enzyme were 46.085 ± 2.327 pmol·min⁻¹·µg⁻¹ and 45.492 ± 1.722 pmol·min⁻¹·µg⁻¹ for RFV-FV (P > 0.05) (Fig. 5F and S Table 7), with RFV-FV displaying 96.76%±1.63% of the VV capping enzyme (P > 0.05), with no statical differences (Fig. 5G and S Table 7). These results demonstrated that the TPase activity of the RFV-FV is comparable to that of the commercial VV capping enzyme.

To quantitatively evaluate the catalytic properties of RFV-FV enzyme relative to the commercial VV capping enzyme, we performed Michaelis–Menten equation kinetic analyses. The results revealed that the VV capping enzyme exhibited a Km of 78.84 µM, a Vmax of 114.1 µM/min, a kcat of 14.55 min⁻¹, and a catalytic efficiency (kcat/Km) of 0.185 µM⁻¹·min⁻¹ (S Figs. 24 A-B and S Table 8). Meanwhile, the RFV-FV chimeric enzyme showed a Km of 11.08 µM, a Vmax of 17.41 µM/min, a kcat of 2.22 min⁻¹, and a catalytic efficiency (kcat/Km) of 0.20 µM⁻¹·min⁻¹ (S Figs. 24 C-D and S Table 9, and 10). Notably, the RFV-FV kcat/Kₘ value, representing overall catalytic efficiency, was comparable to that of the commercial VV capping enzyme.

Both GTase and MTase activities of the RFV-FV are analogous to the corresponding activities of the commercial vaccinia virus capping enzyme

GTase activity of the RFV-FV chimeric capping enzyme was quantified via a pyrophosphate (PPi) assay. GTase transfers the guanosine monophosphate (GMP) moiety from GTP to RNA’s 5’-end, releasing PPi. Liberated PPi reacts with 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG) (plus purine nucleoside phosphorylase, PNP) to form a product with a 360 nm absorbance (Fig. 6A). A PPi standard curve that A₃₆₀ to PPi concentration, yielded Y = 0.01007X + 0.01419 (R² = 0.9958, 95% CI) with good replicate fitting (Fig. 6B and S Table 11). Negative control absorbance was low (0.0413 ± 0.0053), while positive control and RFV-FV showed higher values 0.2136 ± 0.0049 and 0.2123 ± 0.0047 (P < 0.0001 vs. negative control) and no difference existed between RFV-FV and positive control (Fig. 6C and S Table 12). GTase specific activities that calculated via standard curve of positive control VV enzyme and RFV-FV were 39.362 ± 1.005 and 38.735 ± 0.942 pmol·min⁻¹·µg⁻¹ (P > 0.05) (Fig. 6D and S Table 13).

Fig. 6.

The GTase and MTase activity of the RFV-FV chimeric capping enzyme is as the same as the commercial vaccinia virus capping enzyme. (A) Schematic of GTase and MTase activity detection principles using a pyrophosphate assay kit and the MTase-Glo Chemiluminescent Assay Kit. (B) Standard curve of pyrophosphate concentration versus absorbance for GTase-catalyzed reaction products. (C) Quantification of absorbance at 360 nm for GTase reaction product complexes of RFV-FV. (D) GTase catalytic reaction rate calculated from the standard curve. (E) Standard curve of S-adenosylhomocysteine (SAH) concentration versus absorbance for MTase-catalyzed reaction product complexes. (F) Quantification of absorbance at 560 nm for MTase reaction product complexes of RFV-FV. (G) MTase catalytic reaction rate calculated from the standard curve. (H) Evaluation of GTase and MTase reaction rates of the RFV-FV capping enzyme, using the reaction rate of commercial VV capping enzyme as the 100% standard.

MTase activity was assessed via MTase-Glo™ Kit. MTase uses S-adenosylmethionine (SAM) as methyl donor, producing S-adenosylhomocysteine (SAH). SAH triggers a luciferase reaction with a 560 nm signal. SAH standard curve (560 nm signal vs. concentration) gave Y=742870X + 39,254 (R² =0.994, 95% CI) with good replicate fitting (Fig. 6E and S Table 14). Negative control signal was minimal 342517.33 ± 23306.24, while positive control and RFV-FV showed higher signals 1850579.00 ± 25674.76 and 1538168.67 ± 40396.82 (both P < 0.0001 vs. Negative control) respectively (Fig. 6F and S Table 15). MTase specific activities of VV enzyme and RFV-FV were 3.954 ± 0.0586 and 3.113 ± 0.1433 pmol·min⁻¹·µg⁻¹ (P > 0.05) respectively (Fig. 6G and S Table 16).

To assess enzymatic activity, we calculated the ratio of RFV-FV activity relative to the commercial VV enzyme. The results showed that the GTase activity of RFV-FV reached 100.69% ± 2.95% of the VV enzyme, whereas its MTase activity was 77.87% ± 3.85% (Fig. 6H and S Table 17). In conclusion, the RFV-FV chimeric capping enzyme exhibits TPase, GTase, and MTase activities that are all comparable to those of the commercially available VV capping enzyme—underscoring its significant potential as an alternative tool for mRNA capping in biotechnological applications.

Discussion

Addressing the bottlenecks of complex workflows and high production costs associated with traditional multi-enzyme capping systems in mRNA vaccine manufacturing, this study pioneered a Hetero-viral family domain fusion strategy (S Fig. 1). Three chimeric capping enzymes integrating the sequential catalytic activities of TPase, GTase, and MTase were successfully designed and constructed. Leveraging AI-assisted structure prediction (Uni-fold) and systematic heterologous expression optimization, we identified that the combination of an MBP tag and E. coli BL21 (DE3) pLysS strain enables high-efficiency soluble expression of the RFV-FV chimeric enzyme. Characterization demonstrated that its TPase (96.763%±1.631%), GTase (100.69%±2.95%), and MTase (77.87%±3.85%) activities are statistically in distinguish from those of the commercial vaccinia virus capping enzyme, confirming that Hetero-viral family domain fusion does not compromise the integrity or synergistic functionality of the catalytic domains.

The core innovations of this study are threefold: First, it breaks the long-standing limitation of conventional intra-viral family enzyme chimerization²¹. Through multi-database (DeepSeek, PDB and Uniports) domain mining, we achieved Hetero-family integration enzyme providing a novel paradigm for the design of multifunctional artificial enzymes²⁷. Second, via systematic tag-host combination screening, we established a high-efficiency soluble expression system for RFV-FV chimeric enzyme. Although RFV‑FV was purified to only approximately 78% homogeneity, with trace host-derived Escherichia coli contaminating proteins still present in the SEC-isolated product (Fig. 3H and S Table 17). As 5’-cap modifying enzymes are strictly eukaryote-specific, the residual contaminating proteins are highly unlikely to possess functional 5’-cap modifying activity. Therefore, the catalytic activity detected in this study was predominantly ascribed to RFV‑FV, and the potential interference of impurities with the enzyme activity assay was essentially eliminated. Third, this work validates the feasibility of replacing complex multi-enzyme cocktails with a single-molecule chimeric enzyme, furnishing a core tool for simplifying mRNA capping workflows.

In the context of mRNA vaccine production, the RFV-FV chimeric enzyme offers substantial practical value: (1) Notably, RFV-FV acts as a single-component enzyme capable of complete cap formation, which simplifies the reaction system and reduces production costs, in contrast to the multi-subunit vaccinia system that requires the combined action of D1/D12 and VP39; (2) The high efficiency of the prokaryotic expression system facilitates large-scale manufacturing, meeting the demands of vaccine industrialization. Under optimized induction conditions, chimeric RFV-FV enzyme expressed in Terrific Broth medium was purified via two steps. UV spectrophotometric quantification showed a purified protein concentration of 1.2 ± 0.3 mg/mL (total ~ 10 mL), corresponding to a final yield of approximately 12 mg/L. The result confirms that the MBP tag and E. coli BL21(DE3) pLysS cells system achieves soluble, functional RFV-FV enzyme expression at the laboratory scale and exhibits excellent industrial scalability; (3) Catalytic activity comparable to commercial enzymes ensures capping efficiency and mRNA product quality, laying a solid foundation for enhanced vaccine stability and translational efficacy. Our controlled experiments revealed that the relative activities of RFV-FV versus the commercial VV cap enzyme were 96.76% ± 1.63% for TPase, 100.69% ± 2.95% for GTase, and 77.87% ± 3.85% for MTase, respectively (Figs. 5 and 6). These results clearly demonstrate that the catalytic performance of RFV-FV is comparable to that of the commercial vaccinia capping system.

Future studies will focus on two key aspects to fully validate the practical application potential of RFV-FV: (1) Determining the capping efficiency, modification accuracy and uniformity of the enzyme on vaccine IVT mRNA of various lengths using LC-MS and RNA 5’-end sequencing. (2) Constructing luciferase and GFP reporter gene systems to systematically evaluate the effects of RFV-FV-mediated capping on mRNA translational efficiency, stability and in vivo immunogenicity via cell-free in vitro translation, eukaryotic cell transfection and in vivo animal experiments, with multi-dimensional comparisons to the commercial vaccinia virus capping system. These advancements are expected to establish the RFV-FV as a core tool in the production of mRNA vaccines and RNA therapeutics, driving technological upgrading in RNA drug industrialization.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Dilidaer Shahatibieke: writing of the first draft; Xiaohui Tang: visualization of the data; Dilidaer Shahatibieke and Xuanfang Zheng: preparation of experimental reagent materials and experimental exploration; Aibibuhan Abulaizi, Abudoureyimu Abula: supervision of the experimental process and guidance in the first draft writing; Abudoureyimu Abula*: review and revision of the first draft.

Funding

This work was supported by the Youth Program of National Natural Science Foundation of China (Grant No. 32201032).

Data availability

All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).

Declarations

Competing interests

The authors declare no competing interests.