Improving Lab-Made Self-Amplifying RNA Quality Using Experimental Design

The human embryonic kidney cell line 293 T used in this study was purchased from ATCC and stocked in our laboratory. The liver cancer cell line HuH-7 used in this study was stocked in our laboratory. In total, 293 T cells and HuH-7 cells were cultured in Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin in a humidified incubator at 37 °C with 5% CO₂. All cell culture reagents were purchased from Gibco. SPF-grade BALB/cmice (6–8-week-old, female, 18–22 g) were provided by and housed at the Laboratory Animal Resource Center, National Institute for Food and Drug Control and were used as experimental animals.

The genomic sequence of VEEV-TC83 (GenBank accession no.: MZ399798.1) and the signal peptide sequence and receptor-binding domain (RBD) sequence of the SARS-CoV-2 JN.1 strain S protein (GenBank accession no.: XBC23597.1) were downloaded. Codon optimization was performed for the SARS-CoV-2 JN.1 strain S protein signal peptide sequence and S protein RBD sequence. The pcDNA3.1+ vector was selected to facilitate subsequent synthesis and verification of saRNA. All sequences were assembled and submitted to Sangon Biotech (Shanghai, China) Co., Ltd. for full-length gene synthesis.

Plasmids were digested with the restriction enzyme XbaI, and the linearization of plasmid DNA was assessed using 1% agarose gel electrophoresis. The completely linearized DNA template was purified using the Microcolumn DNA Gel Extraction kit (ZomanBiot, Beijing, China; ZPV202). Purified linearized DNA template was used for the IVT of saRNA. After IVT completion, DNase I was added and digested for 30 min. An equal volume of 8 M LiCl was added, mixed thoroughly, and stored at −80 °C for 30 min. The mixture was then centrifuged at 12,000 rpm for 30 min at 4 °C, washed once with 70% anhydrous ethanol, air-dried, and dissolved in an appropriate volume of nuclease-free water.

A total of 293 T cells were seeded (1 × 10⁶ cells/well) in 6-well plates until 90% confluency; next, 1 μg, 2 μg, and 5 μg of self-amplifying RNA with an encoding receptor-binding domain (RBD) sequence of the SARS-CoV-2 JN.1 strain (Sa-RBD) were transfected into the 6-well plates with Lipofectamine RNAiMAX reagent (Thermo Fisher scientific, MA, USA; 13778075). At 36 h, cells and cell culture supernatants were collected. Cells were centrifuged, had the supernatant discarded, were washed three times with 1× PBS, and were lysed in 100 μL of lysis buffer on ice for 30 min. The lysate was centrifuged at 12,000 rpm for 30 min, and the supernatant was collected. Protein quantification was performed according to the instructions of the BCA Protein Assay kit (Thermo Fisher Scientific, MA, USA; 23225). Based on the BCA quantification result, an appropriate volume of protein loading buffer was added to the supernatant. The mixture was denatured in boiling water for 10 min. Proteins were separated on a 10% SDS-polyacrylamide gel. An appropriate volume of sample was loaded and electrophoresed at 80 V through the stacking gel and 120 V through the separating gel. Protein transfer was performed using the eBlot™ L1 Rapid Wet Transfer System. Membranes were blocked with 5% skimmed milk at room temperature for 2 h. Primary antibody was added and incubated overnight at 4 °C. Membranes were washed three times with 1× PBST, and a secondary antibody was added and incubated at room temperature for 1 h. Membranes were washed three times with 1× PBST, followed by ECL chemiluminescence detection for signal acquisition.

In total, 293 T cells were seeded (1 × 10⁶ cells/well) in 6-well plates until 90% confluency; next, 5 μg of Sa-RBD and linear mRNA were transfected into the 6-well plates with Lipofectamine RNAiMAX reagent (Thermo Fisher scientific, MA, USA; 13778075). At 36 h, cells and cell culture supernatants were collected. Cells were centrifuged, had the supernatant discarded, were washed three times with 1× PBS, and were lysed in 100 μL of lysis buffer on ice for 30 min. The lysate was centrifuged at 12,000 rpm for 30 min, and the supernatant was aspirated. The antibody (Sinobiological; 40592) was diluted to 10 μg/mL in coating buffer, and 100 μL was added per well for overnight coating at 4 °C. Blocking buffer was added and incubated at 37 °C for 2 h. After washing the plate, samples and serially diluted standards were added and incubated at 37 °C for 1 h. After washing, detection antibody was added and incubated at 37 °C for 1.5 h. After washing, horseradish peroxidase (HRP)-conjugated secondary antibody (Zhongshan Golden Bridge Biotechnology; ZB-2305) was added and incubated at 37 °C for 1 h. After washing, freshly prepared chromogenic substrate (Wantai BioPharm; ZDT0175, ZDT0181) was added, developed in the dark for 15 min, and then stop solution (Wantai BioPharm; ZDT0178) was added. The plate was immediately read at dual wavelengths of 450/630 nm.

A total of 293 T cells were seeded (1 × 10⁶ cells/well) in 6-well plates until 90% confluency; next, 5 μg of Sa-RBD and linear mRNA were transfected into the 6-well plates. The culture medium was replaced at 12 h post-transfection, and cells were collected at 12 h, 24 h, 36 h, 48 h, 60 h, 72 h, 84 h, 96 h, 108 h, and 132 h post-transfection, respectively, washed three times with PBS, and lysed in 100 μL of lysis buffer. qPCR was performed using the One Step TB Green^® PrimeScript™ RT-PCR kit II (Takara, Beijing, China; RR086A), according to the manufacturer's instructions. The target RNA content at 12 h was set as 100%, and results were relatively quantified using the 2^(−ΔΔCt) method.

The Design of Experiments (DoE) for optimizing self-amplifying mRNA IVT process parameters was designed using JMP software (JMP Pro 17). A Central Composite Design (CCD) within Response Surface Methodology (RSM) was employed to generate the experimental design. Experimental data were obtained according to the design and analyzed using JMP software (JMP Pro 17).

Self-amplifying mRNA obtained from the IVT experiments according to the DoE design was assessed for integrity and yield. The concentration of self-amplifying mRNA obtained from IVT was measured using NanoDrop. The integrity of self-amplifying mRNA was detected using the Qsep100 Fully Automated Capillary Electrophoresis System (BIOptic) with an N3 cartridge, with the 9000 nt RNA Ladder analyzed simultaneously with the samples.

Sa-RBD with 37%, 64%, and 80% integrity, obtained from process optimization, was encapsulated using microfluidic mixing technology (MIANANO, INano L, Shanghai, China) with a lipid mixture (SM102). The encapsulated product was buffer-exchanged by ultrafiltration, and an equal volume of 20% sucrose buffer was added. Self-amplifying mRNA vaccines with 37%, 64%, and 80% integrity were named 0.37 V-Sa-RBD, 0.64 V-Sa-RBD, and 0.80 V-Sa-RBD, and stored at −20 °C. The encapsulation efficiency of the vaccines was detected using the Quant-iT™ RiboGreen™ RNA reagent and kit (Invitrogen, MA, USA; R11490). The average particle size and polydispersity index (PDI) of the vaccines were detected using the Malvern Zetasizer Nano ZS (Malvern, UK).

Blood was collected from the retro-orbital venous plexus of mice at 2 weeks after the first immunization and 5 weeks after the second immunization, with 200 μL collected per mouse. The collected whole blood was left at room temperature for 2 h and centrifuged at 6000 rpm for 10 min; the serum was collected and stored at −20 °C. Spleens were ground in RPMI-1640 complete medium, and the spleen homogenate was collected. The homogenate was centrifuged at 500× g for 10 min, the supernatant was discarded, and the pellet was gently resuspended. Three milliliters of red blood cell lysis buffer were added, mixed by inversion, and lysed for 3 min. An equal volume of RPMI-1640 complete medium was added to stop the lysis, the solution was centrifuged at 500× g for 10 min, and the supernatant was discarded. Splenocytes were resuspended in 3 mL of cell freezing medium, frozen using a gradient freezing protocol, and ultimately stored at −80 °C.

Spike-specific IgG antibodies were detected using ELISA. The spike proteins of JN.1 were individually coated at 1 µg/mL on 96-well plates overnight at 4 °C. The subsequent methods were performed as described in. Section 2.5

To avoid edge effects, 250 μL of cell maintenance medium was added to the peripheral wells of a 96-well culture plate. Row 2 received 142.5 μL of cell maintenance medium. Cell control wells received 150 μL of cell maintenance medium, virus control wells received 100 μL of cell maintenance medium, and the remaining wells received 100 μL of cell maintenance medium. Test serum was pre-diluted in the cell maintenance medium at specific dilutions, and 7.5 μL per well (in duplicate) was added to row 2 of the 96-well culture plate, followed by three-fold serial dilution. Fifty microliters of 1 × 10⁴ TCID₅₀/mL SARS-CoV-2 pseudovirus were added to each well, mixed thoroughly, and incubated at 37 °C, 5% CO₂ for 1 h for neutralization. HuH-7 cell suspension was diluted to 2 × 10⁵ cells/mL, and 100 μL was added per well. The plate was incubated at 37 °C, 5% CO₂ for 20–24 h. Subsequently, 150 μL of cell maintenance medium was aspirated, 100 μL of luciferase detection reagent was added, and luminescence was measured using a high-throughput multi-mode microplate detection system.

ELISpot plates were coated with purified anti-mouse IFN-γ antibody (BD Biosciences; 51-2525KC) and incubated overnight at 4 °C. Plates were washed with DMEM complete medium and blocked with DMEM complete medium at room temperature for 2 h. After discarding the blocking solution, 50 μL of 10 μg/mL peptide pool and 50 μL of cells (4 × 10⁶/mL) were added, and incubated at 37 °C, 5% CO₂ for 20 h. Plates were washed twice with pre-chilled deionized water and three times with 1× PBST. Biotinylated Detection antibody (BD Biosciences; 51-1818KZ) was diluted 1:250 in antibody dilution buffer, and 100 μL was added per well to the ELISpot plate, incubated at room temperature for 2 h. Plates were washed three times with 1× PBST. Streptavidin–HRP was diluted 1:100 in antibody dilution buffer, and 100 μL was added per well to the ELISpot plate, incubated at room temperature for 2 h. Plates were washed four times with 1× PBST and twice with 1× PBS, blotted dry, and 100 μL of chromogenic substrate was added per well for 10 min. The reaction was stopped by washing with deionized water, air-dried at room temperature, and spots were counted using an automated ELISpot reader (CTL-ImmunoSpot^® S6, Cleveland, OH, USA).

Statistical analysis of all experimental results was performed using GraphPad Prism 9 (GraphPad Software Inc., USA). T-test was used to compare statistical differences between two groups. Analysis of variance (ANOVA) was used to compare statistical differences among multiple groups. p < 0.05 was considered statistically significant and denoted by _*, p < 0.01 was considered significantly different and denoted by **, p < 0.001 was considered highly significantly different and denoted by ***, and p < 0.0001 was considered extremely highly significantly different and denoted by ****. JMP software was used for IVT experimental design and data model fitting.

The saRNA sequence (Figure 1A) was ligated into the pcDNA3.1(+) vector to construct the plasmid (P-Sa-RBD), and the linearized template (L-P-Sa-RBD) was obtained. The proportion of supercoiled plasmid and the purity of the linearized template were detected (Figure S1A–C). Using this system, we synthesized saRNA expressing EGFP with different base modifications [23,24,25,26]. Unmodified saRNA demonstrated stronger and longer-lasting expression in 293 T cells (Figure S2). Therefore, we used the unmodified IVT system to obtain saRNA encoding RBD (Sa-RBD) (Figure 1B). The expression of RBD protein and NSP2 protein was detected by Western blot 36 h post-transfection with Sa-RBD (Figure 1C,D). At the same transfection dose of 5 μg, the RBD RNA content in saRNA-transfected cells showed a self-amplifying trend—increasing initially, then decreasing and stabilizing—whereas cells transfected with non-replicating mRNA exhibited no such change in RBD RNA content (Figure 1E). At 36 h post-transfection with equal doses, the RBD protein level in saRNA-transfected cells was significantly higher than in non-replicating mRNA-transfected cells (Figure 1F). These results indicate that the saRNA backbone confers both expression and self-amplification capabilities.

Response Surface Methodology (RSM) was selected for experimental design to optimize the IVT process for saRNA vaccines. The integrity and yield of the IVT product were set as the response variables for each experiment. Based on prior knowledge, five parameters potentially affecting IVT integrity and yield were initially identified. The levels for each factor were set as follows: Mg²⁺ concentration (6~75 mM), T7 RNA polymerase concentration (5~10 U/μL), IVT reaction time (2~4 h), cap analog concentration (4~10 mM), and DNA template input amount (30~60 ng/μL). Other IVT parameters were fixed at NTPs (10 mM), 1 M Tris-HCl (40 mM), 1 M DTT (10 mM), spermidine (2 mM), pyrophosphatase (0.002 U/μL), and RNase inhibitor (1 U/μL). Central Composite Design (CCD) can more accurately describe the relationship between response variables and factors. Therefore, a multi-factor CCD was implemented using JMP software across defined levels for each factor, yielding a total of 28 IVT experiments (Table S1).

The integrity and yield of the obtained saRNA were detected. Prediction models for integrity and yield were fitted using the Least Squares Method. For the saRNA integrity model: Root Mean Square Error (RMSE) = 0.0148, R-squared (R²) = 0.99838, p = <0.0001 (Figure 2A). For the self-amplifying mRNA yield model: RMSE = 89.034, R² = 0.98, p = <0.0001 (Figure 2B).

JMP software was used to analyze the effects and interactions of different factors on integrity and yield, obtaining LogWorth values for each factor to reflect their contributions to the predictive model (Table 1). In the predictive model for saRNA integrity, Mg²⁺ concentration (p < 0.00001, LogWorth = 10.017) and IVT reaction time (p < 0.00001, LogWorth = 3.685) had the most significant effects on saRNA integrity, and these two exerted interaction effects (p < 0.00001, LogWorth = 3.740). The effect of Mg²⁺ concentration on saRNA integrity was non-linear. Under different IVT reaction time conditions, as Mg²⁺ concentration gradually increased within the 6~75 mM range, the integrity of IVT products all showed an initial increase followed by a decrease. When Mg²⁺ concentration was 6 mM, IVT reaction time had no significant effect on integrity, and integrity was almost 0; but when Mg²⁺ concentration rose to 75 mM, integrity significantly decreased with increasing IVT time. According to the above data, it can be concluded that under lower IVT reaction time conditions, selecting moderate Mg²⁺ concentration and lower T7 polymerase can obtain IVT products with higher integrity, conserving more materials and time (Figure 3A).

In the predictive model for self-amplifying mRNA yield, Mg²⁺ reaction concentration (p = 0.0001, LogWorth = 3.987), IVT reaction time (p = 0.03713 LogWorth = 1.430) and DNA template input amount (p = 0.4082, LogWorth = 1.389) had significant effects on saRNA yield. The effects of the DNA template input amount and Mg²⁺ reaction concentration on self-amplifying mRNA yield had interaction effects (p = 0.05427, LogWorth = 1.265). Similarly to the integrity model, the effect of Mg²⁺ reaction concentration on self-amplifying mRNA yield was non-linear. Under different DNA template input amount conditions, as Mg²⁺ reaction concentration gradually increased within the 6 mM~75 mM range, the integrity of IVT products all showed an initial increase followed by a decrease. When Mg²⁺ concentration was at 6 mM, the yield was not affected by the DNA template input amount; but when Mg²⁺ reaction concentration rose to 75 mM, yield increased with increasing DNA template input amount. The above data indicate that selecting moderate Mg²⁺ reaction concentration and higher DNA template input amount helps improve the yield of saRNA (Figure 3B).

Using JMP software, the two response variables (yield and integrity) were fitted for grouped data analysis. Through desirability function settings, the importance of integrity was set to 3 and yield to 1. The optimal process parameters were obtained: a DNA template input amount of 60 ng/μL, Mg²⁺ concentration of 52 mM, IVT reaction time of 2 h, cap analog concentration of 10 mM, and T7 RNA polymerase concentration of 5 U/μL. Under these parameters, IVT was expected to obtain a yield of 1121.5 μg and integrity of 86.5% (Table 2).

The design space is a multivariate combination and interaction range of input variables and process parameters that has been proven to ensure product quality. Operating within this space ensures the quality meets preset criteria. Based on product quality requirements and process capability, the criterion for self-amplifying mRNA integrity was set at 80%, and the criterion for yield was set at 600 μg/100 μL. Monte Carlo Simulation was performed to obtain the process parameter design space meeting this quality criteria. Results showed the design space: DNA template input amount of 50~60 ng/μL, Mg²⁺ concentration of 50~55 mM, IVT reaction time of 2~2.2 h, cap analog concentration of 7~10 mM, T7 RNA polymerase concentration of 5~7.5 U/μL (Table 3). The defect rate was 0% after 10,000 simulations of this model.

Generally, if the verification results of the extreme worst-case combination parameters from the design space meet the preset criteria, the design space is considered compliant with production requirements. Verification was conducted according to the extreme worst-case combination parameters, the optimal process parameters, the upper and lower boundary combinations of the design space. All results satisfied the preset criteria (Figure 4A,B). Additionally, we applied the optimal process parameters to prepare longer saRNA sequences (9.2k nt, 11.5k nt) to confirm its applicability for longer saRNA. The results showed that IVT products could still meet the preset criteria (Figure 4C).

Cells and supernatants were collected 36 h after transfecting 293 T cells with 1 μg, 2 μg, or 5 μg of Sa-RBD at 37%, 64%, or 80% integrity. Target protein expression was detected by Western blot. Under different transfection doses, as integrity increased, the target protein bands in both cells and culture supernatants gradually intensified. At all doses, Sa-RBD with 80% integrity exhibited the strongest protein expression (Figure 5A). In addition, dose-escalation cannot sufficiently compensate for reduced expression levels caused by low-integrity Sa-RBD. The amount of target protein in cells and supernatants detected by ELISA showed consistent trends with the Western blot results (Figure 5B,C).

Sa-RBDs with different integrity were encapsulated by LNP to achieves saRNA vaccines. Mice were immunized with three different doses (0.3 μg, 1 μg, 3 μg) of the 80% integrity saRNA vaccine (0.80 V-Sa-RBD). Results showed that, 2 weeks post-immunization, all three dose groups achieved seroconversion for RBD-specific binding antibodies and neutralizing antibodies. Geometric Mean Titers (GMTs) showed a dose–response relationship. The 3 μg single-dose immunization induced the highest serum binding antibodies and neutralizing antibodies at 2 weeks, with binding antibody GMT at 1:55127 and pseudovirus neutralizing antibody GMT at 1:467 (). Thus, the 3 μg dose was used for evaluation of the immunogenicity of SaRBD vaccines with different integrities. Figure S3

Mice were immunized with 3 μg of saRNA vaccines at 37%, 64%, and 80% integrities (Table S2). Two weeks post-immunization, the seroconversion rate for spike protein-specific binding antibodies was 100% in all groups. The 37% integrity vaccine failed to effectively induce pseudovirus neutralizing antibodies. Pseudovirus neutralizing antibody GMTs for the 64% and 80% integrity vaccines were 1:48 and 1:411, respectively. The pseudovirus neutralizing antibody titer induced by the 80% integrity vaccine was 137-fold higher than that of the 37% integrity vaccine (p < 0.0001) and 16.1-fold higher than that of the 64% integrity vaccine (p < 0.0001) (Figure 6A,B). The second and third vaccine doses were administered at 2 weeks and 7 weeks after the first immunization, respectively. Five weeks after the second immunization, GMTs for spike protein-specific binding antibodies in the 37%, 64%, and 80% integrity groups were 1:6906, 1:335865, and 1:973253, respectively. Pseudovirus neutralizing antibody GMTs were 1:3, 1:230, and 1:1172, respectively. The pseudovirus neutralizing antibody titer induced by the 80% integrity vaccine was 390.7-fold higher than that of the 37% integrity vaccine (p < 0.0001) and the 64% integrity vaccine was 76.7-fold higher than that of the 37% integrity vaccine (p < 0.0001). Two weeks after the third immunization, GMTs for spike protein-specific binding antibodies in the 37%, 64%, and 80% groups were 1:696492, 1:1204447, and 1:2982650, respectively. Pseudovirus neutralizing antibody GMTs were 1:73, 1:811, and 1:8083, respectively. The pseudovirus neutralizing antibody titer induced by the 80% integrity vaccine was 110.7-fold higher than that of the 37% integrity vaccine and 9.96-fold higher than that of the 64% integrity vaccine (p < 0.01) (Figure 6A,B).

Spleens were harvested 2 weeks after the third immunization to prepare splenocyte suspensions. Splenocytes were stimulated with RBD peptide pools, and the frequency of IFN-γ-secreting T cells was detected. Results showed that the 37% integrity group failed to induce a specific T cell response. The mean spot numbers per 1 × 10⁶ cells for the 37%, 64% and 80% integrity groups were 0, 138, and 370, respectively. The number of IFN-γ spots produced by the 80% integrity group was significantly higher than that of the 64%, 37%, and PBS groups (p < 0.0001). These results indicate that integrity severely impacts both the antibody response and the Th1-type T cell response levels induced by saRNA vaccines (Figure 6C).