The pQE-30 plasmid and Escherichia coli BL21(DE3) were purchased from Novagen Inc. (Madison, WI, USA) for T7 RNAP heterologous expression. The site-directed mutagenesis kit and dsRNA quantitative kit were supplied by Vazyme Co., Ltd. (Nanjing, China). ATP, GTP, CTP, UTP, pyrophosphatase, and murine RNase inhibitor were purchased from Vazyme Co., Ltd. (Nanjing, China). MgCl₂, glycerol, dithiothreitol (DTT), and other chemicals were purchased from Aladdin Industrial Inc. (Shanghai, China) and other commercial sources.

The nucleotide sequence of T7 RNAP was obtained from GenBank NC_001604.1 and optimized using GeneOptimizer^® expert software for improved expression in E. coli. Optimized foreign nucleotide sequences have been targeted into plasmid pQE-30 digested with BamHI and HindIII (inserted His-tag sequences at 5ʹ terminal of foreign nucleotide sequences), which was then used for the transformation of E. coli BL21 (DE3) cells or as the original material for site-directed mutagenesis. The oligonucleotide primers required for site-specific mutations are detailed in Supplementary Table S1.

E. coli BL21 (DE3) harboring T7 RNAP-carrying recombinant pQE-30 was cultured in Luria-Bertani (LB) broth by supplementing 100 mM ampicillin at 37°C on a shaker (220 rpm). When the OD at 600 nm reached ∼0.6, 0.5 mM IPTG was added to the growing culture to induce T7 RNAP expression for 4–6 h at 37°C. Subsequently, cells were harvested by centrifugation (4°C, 10,000×g, 20 min). Thereafter, the cells were washed twice, resuspended in Tris-HCl buffer (pH 8.0, 50 mM pH 8.0), and disrupted by sonication (260 W, 3-s pulse, 5-s pause). Crude enzymes were obtained by centrifugation (4°C, 10,000×g, 20 min). The protein content was analyzed using SDS-PAGE.

Nucleic acid impurities were removed from the crude enzyme by pretreatment with the addition of 0.1 vol of 3% polyethyleneimine (PEI). The sample was diluted with the same volume of SP-binding buffer (20 mM MES, 100 mM NaCl, 5% v/v glycerol, 1 mM dithiothreitol, pH 6.5) and loaded onto a 1 mL SP column equilibrated with 10 mL SP-binding buffer. The sample was eluted with 5 mL SP elution buffer (40 mM MES, 500 mM NaCl, 5% v/v glycerol, 1 mM DTT, pH 7.5). Subsequently, T7 RNAP was diluted with the same volume of Ni-binding buffer (300 mM NaCl, 0.2 mM DTT, 20 mM imidazole, 5% v/v glycerol, 20 mM Tris-HCl pH 7.5) loaded onto a His trap HP (GE Healthcare). The purified T7 RNAP was eluted with 5 mL of Ni-elution buffer (200 mM NaCl, 0.2 mM DTT, 450 mM imidazole, 5% v/v glycerol, 20 mM Tris-HCl pH 7.5). Purified T7 RNAP was desalted using a HiTrap Desalting Column (GE Healthcare Corp., USA). The sample content was measured using a BCA Protein Assay Kit (Vazyme Co., Ltd., Nanjing, China) with bovine serum albumin (BSA) as a standard.

The activity of the samples (T7 RNAP wild type [WT] and mutants) was estimated as X U/μL. T7 RNAP standard (Vazyme Co., Ltd, Nanjing, China) and samples were diluted by dilution buffer (100 mM NaCl, 0.1 mM EDTA-2Na, 0.1% g/g Triton X-100, 10% v/v glycerol, 50 mM Tris-HCl pH 8.0) to 0 U/μL, 0.05 U/μL, 0.1 U/μL, 0.2 U/μL, 0.3 U/μL, and 0.6 U/μL. The reaction mixture (final concentrations of 0.5 mM ATP/GTP/CTP/UTP, 2 U/μL murine RNase inhibitor, 0.4% μg/μL DNA template, 20% v/v T7 RNAP, 1× transcription buffer pH 8.0) was added to a 96-well plate and incubated for 30 min at 37°C or other temperatures (the T7 RNAP standard was only incubated at 37°C). Subsequently, the nucleic acid dye BR reagent (Vazyme Co., Ltd., Nanjing, China) was added. Fluorescence detection was performed at an excitation wavelength of 630 nm, emission wavelength of 680 nm, and a gain value of 126. A linear regression equation was constructed using a microplate reader to determine slope and k. Specific activity was defined as follows: Y   Standard = k 1 X + b , (1) Y   Sample = k 2 X + b , (2) Activity of sample = k 1 / k 2 × Estimated value of activity  U / μ L , (3) Specific activity of sample = Activity   of   protein   U /   μ L   Concentration   of   protein     μ g / μ L . (4)

T7 RNAP WT and variants were diluted by dilution buffer to 100 U/μL and pre-incubated at 45°C for different time intervals (0 min, 20 min, 40 min, and 60 min). Subsequently, the residue activity was measured using Eq. 4.

The melting temperature was measured as follows: T7 RNAP WT and 20 μL of variants (no less than 1 μg/μL in 200 mM NaCl, 0.2 mM DTT, 5%v/v glycerol, 0.1 mM EDTA-2Na, 50 mM Tris-HCl, pH 8.0) were added to 0.15 mL PCR tubes, and then measured by Prometheus Panta (NanoTemper, Germany) varying temperatures from 25°C to 95°C, with a heating speed of 1 °C/min.

The reaction mixture of circRNA synthesis (final concentrations of 7.5 mM ATP/GTP/CTP/UTP, 2 U/μL Murine RNase inhibitor, 0.005 U/μL pyrophosphatase, 2.5% μg/μL DNA template, 15 U/μL T7 RNAP, 1× transcription buffer (Vazyme Co., Ltd, Nanjing, China)) was added to a 96-well plate and incubated for 90 min at 37°C, 40°C, 45°C, 48°C, and 50°C. Subsequently, the RNA products were purified using clean RNA magnetic beads (Vazyme Co., Ltd., Nanjing, China). The RNA product yield was measured using a Nano-300 instrument. After denaturing the RNA products (incubated for 5 min at 70°C and then iced for 5 min at 0°C), the ratio of cyclization was measured by capillary electrophoresis (CE).

After circRNA synthesis with 20 mM Mg²⁺, 25 mM Mg²⁺, 30 mM Mg²⁺, 35 mM Mg²⁺, and 40 mM Mg²⁺ as the cofactor, the RNA product yield was measured using a Nano-300 instrument. dsRNA content was measured using a dsRNA quantitative kit supplied by Vazyme Co., Ltd (Nanjing, China). This kit uses a double-antibody sandwich enzyme-linked immunosorbent assay (ELISA) to detect dsRNA content in the in vitro transcription system.

The PIE method is an effective, easy, and widely used approach for RNA cyclization (Wesselhoeft et al., 2018; Chen and Lu, 2021). A one-step reaction for circRNA synthesis was attempted using T7 RNAP WT as the catalyst. Before catalysis, purified T7 RNAP WT was prepared. The optimized T7 RNAP WT gene, which was obtained from Bacteriophage T7, was amplified and overexpressed in E. coli BL21 (DE3). Additionally, the 6×His tag was expressed as an N-terminal fusion to the T7 RNAP. The encoded His-tagged T7 RNAP was purified as described in Section 2 and was obtained from the cell-free extract of the recombinant E. coli strain carrying the plasmid pQE-30 (Figure 2A). Subsequently, the purified His-tagged T7 RNAP produced a single band on SDS-PAGE with a molecular mass of approximately 100 kDa (Figure 2B).

The activity of T7 RNAP WT at different temperatures was then measured. The activity decreased rapidly as the temperature increased from 37°C to 50°C. The specificity activity of T7 RNAP WT at 45°C was only 22.09 ± 3.7 U/μg, which was more than 10 times less than the original specificity activity at 37°C. Thus, 37°C was an appropriate temperature for using T7 RNAP WT for RNA synthesis.

Cofactors Mg²⁺ and GTP needed for cyclization are included in linear RNA precursor synthesis systems (Wesselhoeft et al., 2018). Thus, the one-step method for circRNA synthesis by T7 RNAP WT at 37°C was evaluated without adding any extra reagent. circRNAs were synthesized (Figure 3). However, nearly 50% of the linear RNA precursors were retained. The large amount of retained precursors is probably due to the low reaction temperature (Wesselhoeft et al., 2018). T7 RNAP WT is unstable at higher temperatures. Consequently, circRNA synthesis can be developed via a two-step approach for circRNA precursors by sequential conversion via transcription and cyclization. T7 RNAP thermostability modification should be designed for effective circRNA synthesis.

Protein expression was analyzed using SDS-PAGE (Supplementary Figure S2, Supplementary Material). There was no difficulty in the expression of T7 RNAP WT and mutants. The T7 RNAP specificity of the WT and its variants was further determined. As the data showed, the method described above was reliable for hotspot selection; both variants, G788A and C530S, improved protein activity by more than 50% at 42°C (Figure 7A). Based on further research, binding with M0 confirmed that the specificity increased with increasing temperature. Up to 45°C, the activity of T7 RNAP M0+G788A reached the highest (508.73 ± 5.40 U/μg), which was much higher than T7 RNAP WT at 37°C (Figure 7A; Figure 7B). In most cases, a loss of specificity activity was observed at 48°C. In particular, single-site mutants of T7 RNAP were inactivated. Only M0+G788A retained an activity of 168.08 ± 20.46 U/μg, whereas M0 and M0+C530S fell below 100 U/μg. M0+G788A resulted in optimized multiple hotspot-aggregated T7 RNAP variants, which retained 63.08 ± 8.92 U/μg when the temperature rose to 50°C. Obtaining better variants through multiple hotspot aggregations is an effective method for improving enzyme thermostability (Boulain et al., 2013; Elias et al., 2023).

It was observed that T7 TNAP M0+G788A performed much better than M0+C530S under higher-temperature conditions (≥48°C). To further understand T7 RNAP M0+G788A, its thermostability was measured under different time intervals of high-temperature incubation at 45°C. Under incubation conditions, T7 RNAP WT and G788A were inactivated after 20 min, whereas T7 RNAP M0+G788A retained more than 95% of its initial activity. After 60 min of incubation, T7 RNAP M0+G788A retained >300 U/μg activity. However, the basement-M0 dropped to 78.9 ± 5.88 U/μg, which was less than 20% of its initial activity. T7 RNAP M0+G788A still retained more than 50% of its initial activity after being incubated for 80 min at 45°C (Table 1). The results indicated that T7 RNAP M0+G788A, designed using the above strategy, was effective.

Measurement of the protein melting temperature (Tm) is another commonly accepted method for protein thermostability analysis (Senisterra and Finerty, 2009; Watanabe et al., 2016; Elias et al., 2023). The Tm of T7 RNAP was determined by dynamic light scattering (DLS), as previously reported (Prometheus Panta). As shown in Supplementary Figure S3, the Tm values of T7 RNAP WT and M0+G788A were estimated to be 39.91°C ± 0.03°C and 45.19°C ± 0.01°C, respectively. A thermostable variant was successfully constructed by combining M0 and G788A, achieving a high melting temperature that increased by 5°C.

Substitution with M0+G788A (S430P, N433T, S633P, F849I, F880Y, and G788A) significantly improved T7 RNAP stability at high temperatures. The optimized M0+G788A mutant showed higher enzyme activity than the WT. Furthermore, after incubation at 45°C for 80 min, 274.22 ± 14.77 U/μg of residue activity was retained. This variant was proficient in one-step circRNA synthesis.

Before applying the T7 RNAP variant to one-step circRNA synthesis, different yields of RNA products were measured using T7 RNAP WT and its variants as catalysts. From the data shown in Figure 8A, by adding 300U of T7 RNAP WT to the reaction system, 197.3 ± 17.38 μg RNA products were acquired at the reaction temperature of 37°C after incubation for 90 min. As the reaction temperature rose to 50°C, 192.6 ± 16.2 μg circRNA products were achieved through 300U of M0+G788A as the catalyst. No significant decrease in the circRNA expression was observed. As we previously observed, WT was inactive at 50°C. When replaced by M0 as the catalyst, only 138.7 ± 3.13 μg circRNA products were measured (Figure 8A). Compared with M0, M0+C530S improved the yield of circRNA to 167.3 ± 7.67 μg at 50°C, which was still lower than using T7 RNAP WT as the catalyst at 37°C. Based on these findings, the problem of high-temperature intolerance of the catalyst used in circRNA synthesis was solved by constructing the T7 RNAP variant M0+G788A.

As measured in Section 3.1, nearly 50% of linear RNA precursors were retained in the one-step reaction for circRNA synthesis by T7 RNAP WT at 37°C (Figure 8B, yellow). The efficiency of T7 RNAP variant M0+G788A was measured when one-step circRNA synthesis was conducted at 50°C. As expected, linear RNA precursors were reduced to 4.5 ± 0.28% (Figure 8B Red, Supplementary Table S2). More than 95% of the linear precursors were cyclized when transcribed without any additional reagents.

To understand the relationship between temperature and cyclization ratio, one-step circRNA cyclization reactions were conducted at different temperatures. As shown in Figure 8B, the ratio of cyclization gradually increased as the temperature increased from 37°C to 50°C. The precursor residue decreased sharply as temperature increased. As the temperature rose to ≥45°C, the precursor residue retained was ≤7.5 ± 0.14%, which was less than 10% of total RNA products. When the temperature rose to 50°C, the precursor residue only retained 4.5 ± 0.28% of total RNA products (Figure 9, Supplementary Table S2). Precursors decreased by more than 90% compared to the precursors retained in the low-temperature (37°C) reaction.

Wesselhoeft et al. performed high-temperature incubation to form rings and required step-by-step addition and incubation to obtain high-cyclization-rate IVT products (Wesselhoeft et al., 2018). The T7 RNAP M0+G788A variant constructed in this study successfully catalyzed a high cyclization ratio of circRNA products at high temperatures without any additional reagents. CircRNA formation in only one step was proven to be successful.

In addition to the simplified synthesis process, circRNA synthesis also faces the common problems of in vitro transcription. Various byproducts are produced during product-templated transcription (Donghi and Schnabl, 2011; Wu et al., 2020; Dousis et al., 2023). Among them, the dsRNA byproduct, which may result in a strong immune response against RNA drugs, has been detected by researchers (Kato et al., 2006; Pichlmair et al., 2006; Goubau et al., 2014). dsRNA byproducts were produced at a proportion of 1.98 ± 0.04 ×10^–4 of the total yield of RNA products catalyzed by T7 RNAP WT at 37°C (Supplementary Table S2). Although their content was relatively low, the strong immune response caused by dsRNAs cannot be ignored (Wu et al., 2020; Dousis et al., 2023).

There are several features that drive the production of dsRNA byproducts, such as the 3′-extension of RNA transcribed in vitro, the production of antisense RNAs and so forth (Cavac et al., 2021; Dousis et al., 2023). In a previous study, modified ribonucleotides were shown to reduce the synthesis of antisense RNA, which might induce dsRNA production (Nance and Meier, 2021). Karikó et al. used modified ribonucleotides to reduce the production of dsRNA and improve the efficiency of mRNA translation (Karikó and Weissman, 2007). Because of this innovation, she won the 2023 Nobel Prize for Physiology and Medicine. When modified ribonucleotides are used for circRNA synthesis, the splicing efficiency is dramatically reduced, which induces the relatively low rate of cyclization (Wesselhoeft et al., 2019). Thus, this method has proven to be unsuccessful for dsRNA-reduced circRNA synthesis.

Wu et al. found that the high-temperature transcription reaction could reduce the production of 3′-extension products of mRNA (Wu et al., 2020). This method may also apply to the dsRNA-mediated synthesis of circular RNA. We measured the amount of dsRNA produced by high-temperature transcription catalyzed by the T7 RNAP variant M0 + G788A. The rate of dsRNA production via T7 RNAP WT as the catalyst was plotted at 100. As shown in Figure 10, the content of dsRNA byproducts decreased rapidly as the reaction temperature rose. As the temperature rose to 45°C, relative dsRNA content decreased to 31.54% ± 3.21% of its initial level. When the temperature was further raised to 50°C, only 18.6% ± 4.13% of dsRNA byproducts were produced in in vitro transcription compared to the production of dsRNA in the reaction at 37°C. Thus, the application of T7 RNAP M0+G788A in high-temperature circRNA synthesis not only simplified the synthesis process but also reduced dsRNA production. However, there are still some limitations via T7 RNAP M0+G788A as the catalyst for dsRNA decrease. Transcription in the high-temperature vitro reaction cannot decrease the production of antisense RNAs, which is an important feature that drives the production of dsRNA byproducts (Wu et al., 2020; Nance and Meier, 2021).