Optimization of 5′UTR to evade SARS-CoV-2 Nonstructural protein 1-directed inhibition of protein synthesis in cells

Abstract

Abstract

Maximizing the expression level of therapeutic proteins in cells is the general goal for DNA/mRNA therapies. It is particularly challenging to achieve efficient protein expression in the cellular contexts with inhibited translation machineries, such as in the presence of cellular Nonstructural protein 1 (Nsp1) of coronaviruses (CoVs) that has been reported to inhibit overall protein synthesis of host genes and exogenously delivered mRNAs/DNAs. In this study, we thoroughly examined the sequence and structure contexts of viral and non-viral 5′UTRs that determine the protein expression levels of exogenously delivered DNAs and mRNAs in cells expressing SARS-CoV-2 Nsp1. It was found that high 5′-proximal A/U content promotes an escape from Nsp1-directed inhibition of protein synthesis and results in selective protein expression. Furthermore, 5′-proximal Cs were found to significantly enhance the protein expression in an Nsp1-dependent manner, while Gs located at a specific window close to the 5′-end counteract such enhancement. The distinct protein expression levels resulted from different 5′UTRs were found correlated to Nsp1-induced mRNA degradations. These findings ultimately enabled rational designs for optimized 5′UTRs that lead to strong expression of exogenous proteins regardless of the translationally repressive Nsp1. On the other hand, we have also identified several 5′-proximal sequences derived from host genes that are capable of mediating the escapes. These results provided novel perspectives to the optimizations of 5′UTRs for DNA/mRNA therapies and/or vaccinations, as well as shedding light on the potential host escapees from Nsp1-directed translational shutoffs.

Key points

• The 5′-proximal SL1 and 5a/b derived from SARS-CoV-2 genomic RNA promote exogenous protein synthesis in cells expressing Nsp1 comparing with non-specific 5′UTRs.

• Specific 5′-proximal sequence contexts are the key determinants of the escapes from Nsp1-directed translational repression and thereby enhance protein expressions.

• Systematic mutagenesis identified optimized 5′UTRs that strongly enhance protein expression and promote resistance to Nsp1-induced translational repression and RNA degradation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00253-023-12442-2.

Introduction

Expression of engineered proteins that are capable of supplementing defect endogenous proteins in vivo have been applied as gene therapies with growing clinical impact in recent years (Dunbar et al. 2018; Kessler et al. 2015; Saraswat et al. 2009). One of the most commonly used techniques to express exogenous proteins is to deliver promoter-driven protein-coding DNAs into target cells. Another approach to express therapeutic proteins could be delivering synthesized translatable mRNAs into cells as mRNA therapies (Sahin et al. 2014). This approach has been further developed into mRNA vaccinations (Pardi et al. 2018a, b; Ulmer and Geall 2016) and successively applied to combat the current pandemics of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (Chung et al. 2021a, b; Haas et al. 2021; Harris et al. 2021; Shah et al. 2021). Ideal DNA/mRNA-based treatments are aiming to achieve high protein expression level in cells with minimal exogenously delivered DNA/mRNA doses and/or administration frequencies, for that DNA/mRNA preparations may cause adverse immune stimulations, while viral vectors and/or non-viral carries used for delivery have been shown for cellular and/or systemic toxicities (Liang et al. 2018; Mullard 2021; Qin et al. 2022; Weng et al. 2020). Therefore, optimization of protein expression efficiencies is of great importance for DNA/mRNA therapies and vaccinations, which is particularly challenging when non-viral vectors were used (Damase et al. 2021).

To achieve high expression efficiency of exogenous proteins in cells, scientists have tried to optimize the expression cassettes that generally consist of promoters, protein-coding sequences, and non-coding regions. There were several cassette optimization studies focusing on promoters and protein-coding regions (Ho and Yang 2014; Lai and Duan 2019; Papadakis et al. 2004; Wu et al. 2019), yet not much efforts have been made on the optimizations of 5′UTRs regarding to promoting exogenous protein expression in human cells, although it has been broadly reported that 5′UTR is a primary determinant of translation efficiency in which a variety of regulatory elements have been identified (Asrani et al. 2018; Ding et al. 2018; Jackson et al. 2010; Leppek et al. 2018). In the past, many cis-regulatory elements located in human 5′UTRs have been characterized individually, and prediction of protein expression efficiency basing on the primary sequence of 5′UTR, however, is still difficult, except the well-known Kozak sequence that strongly promote protein translation (Kozak 1986; Sample et al. 2019). Therefore, rational design and engineering of artificial 5′UTRs for highly efficient protein expression is still challenging to date. Recently, studies on high-throughput screening of synthetic short 5′UTRs have improved our understanding of how short 5′UTRs may affect protein expression in cells (Cao et al. 2021; Sample et al. 2019). However, the possibility of utilizing short 5′UTR to enhance protein synthesis in differential cellular contexts has not been addressed, particularly in the cellular context that protein synthesis machinery is drastically inhibited.

It has been recently reported that the cellular protein synthesis is strictly inhibited upon infections of SARS-CoV-2 (de Breyne et al. 2020; Dong et al. 2021), making the expression of exogenous proteins in such cellular context highly challenging. Recent studies have found that Nonstructural protein 1 (Nsp1), the most N-terminal cleavage product encoded by ORF1a/b, is responsible for the global inhibition of cellular and exogenous protein synthesis upon infections of several alpha- and beta-CoVs (Huang et al. 2011a; Kamitani et al. 2009, 2006; Lokugamage et al. 2012; Min et al. 2020; Narayanan et al. 2008; Simeoni et al. 2021). Such nsp1-induced shutoff of protein synthesis largely affects the pathogenicity and propagation of CoVs in infected cells, yet the molecular basis of this process was poorly understood until recent structural studies suggested that the interaction between Nsp1 and 40S ribosomal subunit blocks the entry tunnel for translating mRNAs (Banerjee et al. 2020; Schubert et al. 2020; Thoms et al. 2020; Yuan et al. 2020). Previous and recent studies on SARS-CoV and SARS-CoV-2 suggested that the 5′UTR of genomic (g) and sub-genomic (sg) mRNAs of CoVs promotes an escape from Nsp1-directed inhibition of protein synthesis and mediates sufficient production of viral proteins (Mordstein et al. 2020; Nakagawa and Makino 2021; Narayanan et al. 2008; Tanaka et al. 2012; Tidu et al. 2020; Wathelet et al. 2007; Yuan et al. 2021), yet the determinants of such viral 5′UTR-mediated escape was poorly understood (Min et al. 2020; Shi et al. 2020; Simeoni et al. 2021; Yuan et al. 2021).

In this study, we attempted to identify a minimal sequence derived from SARS-CoV-2 gRNA 5′UTR that is capable of mediating escape from Nsp1-induced inhibition of protein synthesis. Also we studied in depth to acquire the knowledge of structural and/or sequential determinants for the 5′UTR in mediating selective protein expression in cells expressing SARS-CoV-2 Nsp1. These findings are applicable to design potent artificial 5′UTRs that promote sufficient and/or selective protein expression in an Nsp1-dependent manner, providing fresh insights into the Nsp1-induced translational repression upon SARS-CoV-2 infections and perspectives to DNA/mRNA therapies and/or vaccinations in the presence of cellular Nsp1.

Materials and methods

Protein expression and luciferase reporter vectors

Protein expression in cells was achieved by transfecting pcDNAs (Invitrogen™, USA) having EGFP, wildtype, or K164A/H165A mutant SARS-CoV-2 nsp1. Expression level of mutant and wildtype Nsp1 was detected by C-terminal c-myc tags, respectively (Fig. 1A), for that C-terminal tags have been shown not to interfere with Nsp1’s function (Kamitani et al. 2009; Lokugamage et al. 2012; Mendez et al. 2021; Vazquez et al. 2021; Zhou et al. 2021). The dual luciferase reporter vector, pRLFL, was adapted from psiCheck-2™ (Promega, USA) having two CMV promoters driving the expression of Firefly luciferase (Fluc) and Renila luciferase (Rluc) independently (Fig. 1A). To verify if particular 5′-sequences promote an escape from nsp1-directed inhibition of protein synthesis in cells, divergent sequences derived from the 5′-proximal region of SARS-CoV-2 gmRNA or human cellular mRNAs were cloned to the 5′-termimus upstream to Fluc coding sequence, respectively, while the Rluc baring constant 5′UTR (5′-gtcagctagccacc-3′) was taken as control.

Fig. 1.

In vivo reporter assay showing relative protein expression efficiencies in the presence of SARS-COV-2 Nsp1. A Schematic representation of the in vivo reporter assay. The relative protein expression efficiencies were assessed by relative luciferase activity (RLA), in which the Fluc contains engineerable tested 5′UTRs, while the Rluc has a constant 5′UTR (5′-gtcagctagccacc-3′) as control. B RLA derived from the 76-nt SARS-CoV-2 5′-Leader Sequence (5′-LS) relative to the non-specific 5′-sequence (5′-aagcttggcattggtaccgcagccaccggt-3) were shown in cells expressing differential levels of Nsp1 (upper panel). Western blots verifying the cellular level of Nsp1 and EGFP (lower panels). C To eliminate the Nsp1-independent effects on protein expression potentially caused by different nature of each tested 5′UTRs, RLA(Nsp1/EGFP) was calculated for each construct, which is the RLA derived from a certain tested 5′UTR measured in the presence of cellular Nsp1 (with 100 ng pcDNA-SARS2-Nsp1) relative to what is measured in the absence of Nsp1 (with 100 ng pcDNA-EGFP). This approach allowed us to independently assess the relative expression efficiencies mediated by different tested 5′UTRs specific to the presence of cellular Nsp1. Results are shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.)

In vitro synthesis of mRNAs

The cDNA sequences of the reporter transcripts were amplified with PCR using specific primers, in which the forward primer contains additional T7 promoter sequence, while the backward primer has poly-T at the 3′-end (Figs. 2, 3, 4, 5, 6 and 7B). The amplified DNA fragment was separated with agarose electrophoresis, followed by purification with phenol/chloroform and precipitation with ethanol. Purified template DNA was used for synthesizing capped RNA transcripts with RiboMAX™ Large Scale RNA Production System (Promega, USA). Synthesized 5′-capped 3′-A₃₀-tailed RNA transcripts were treated with RNase-free DNase, followed by phenol/chloroform extraction and ethanol precipitation. Purified transcripts were used for in vitro translation and RNA transfection experiments.

Fig. 2.

Luciferase assays revealing specific elements that promote protein expression in the presence of cellular Nsp1. A The sequence of 5′-proximal 295 nts of SARS-CoV-2 gRNA were shown with predicted secondary structures. Start codons of the uORF and polyprotein 1ab were boxed with dotted and solid line, respectively, while the stop codon of the uORF was underlined. Nucleotides predicted to be the transcriptional regulatory sequences (TRS) that define the 5′-leader sequence (LS) were shaded with grey colour. B The tested 3′- and 5′-truncations (5′T and 3′T series) and the deletion mutants (D series) of the 5′-proximal sequence were demonstrated. Dashed line indicates the deleted regions. These sequences were 5′-terminally fused upstream to Fluc coding region in pRLFL. Relative protein expression level derived from each tested 5′UTRs were evaluated by normalized RLA(Nsp1/EGFP) which is the relative luciferase activities (Fluc/Rluc) measured in cells expressing Nsp1 relative to what measured in cells expressing EGFP. The mean ± SD of three independent experiments, in which differentially C truncated, D deleted SARS-CoV-2 5′-proximal sequences, or E individual structural elements were tested. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.)

Fig. 3.

Mutagenesis studies of the 5′-proximal SL1 and 5a/b in mediating protein expression. The RLA(Nsp1/EGFP) for A SL1 and B SL5a/b and their mutants were shown. The sequences corresponding to wildtype SL1 and SL5a/b were indicated, respectively, while the mutated nucleotides were specified. Sequences identical to the wildtype are indicated as “.” while nucleotide deletion were indicated as “-”. The stem-forming base-paring regions of wildtype SL1 and SL5a/b were underlined. The common 5′-agct-3′ and 5′-accggt-3′ flanking sequences derived form cloning junctions were indicated in lower cases, while the start codon of the Fluc reporter were coded in grey colour. Data were shown by the mean ± SD of three independent experiments. P-value: ≥ 0.05(n.s.), < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****)

Fig. 4.

Differential protein expressions mediated by unstructured A/U-rich. A Unstructured short 5′UTRs having differential A/U/G/C contents were tested for their capabilities of mediating protein expression in the presence of cellular Nsp1 by measuring RLA(Nsp1/EGFP). Sequences of every tested 5′UTRs were indicated. Identical sequences are indicated as “.” while nucleotide deletion were indicated as “-”. Common sequences flanking the tested 5′UTRs derived from the cloning junctions are in lower cases, while the start codon of Fluc reporter is underlined. B The RLA(Nsp1/EGFP) derived from the nt 128–185 and the swapped mutant were shown. Sequences of the tested 5′UTRs are indicated, while the sequence shaded with yellow colour represents the A/U-rich proportion to be swapped. Data were shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.). C Western blots showing the protein levels of Fluc, Rluc, and β-actin in cells expressing either Nsp1 or EGFP, in which differential tested 5′UTRs were fused upstream to Fluc, while the control Rluc 5′UTRs remain constant as illustrated in Fig. 1A. The protein levels of the control Rluc were mostly decreased for all constructs in the presence of cellular Nsp1, validating the effectiveness of Nsp1-directed inhibition of protein synthesis in cells. Certain 5′UTRs resulted in higher Fluc/Rluc ratio in the cells expressing Nsp1 relative to that in cells expressing EGFP, showing their capability of mediating escapes from Nsp1-directed inhibition of protein synthesis

Fig. 5.

Sequence specificities of the 5′UTR-mediated selective protein expressions in the presence of cellular Nsp1. Normalized RLA(Nsp1/EGFP) were shown for the unstructured A/U-rich 5′UTRs with a variety of nucleotide substitutions, respectively. These substitutions include A nt 17 and 24 to be As, Ts, Cs, or Gs; B double Gs gradually shifted from nt 15 and 22 to nt 19 and 26, respectively. The sequences of tested 5′UTRs were shown from the + 1 transcription initiation site (Even et al. 2016). Lower cases represent the common sequences derived from cloning junctions. Sequences identical to the wildtype are indicated as “.”. The ATG translation start codon of Fluc is underlined. Data were shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.). C Western blots showing protein levels of these nucleotide-substituted constructs in the presence of cellular EGFP and Nsp1, respectively

Fig. 6.

Distinct effects resulted from 5′-proximal Gs and Cs to the escapes from Nsp1-directed inhibition of protein synthesis. Normalized RLA(Nsp1/EGFP) were shown for the 5′UTR-mediated escapes from Nsp1-directed inhibition of protein synthesis, in which the 5′-proximus contains A a single G introduced between nt 12 and 26 and B double Cs introduced with and without a pair of Gs, respectively. The sequences of tested 5′UTRs were shown from the + 1 transcription initiation site (Even et al. 2016). Lower cases represent the common sequences derived from cloning junctions. Sequences identical to the wildtype are indicated as “.”. The ATG translation start codon of Fluc is underlined. Data were shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.)

Fig. 7.

Differential translation activities mediated by specific 5′UTRs in the presence of cellular Nsp1. A The RLA showing the SARS-CoV-2 1–295 nts mediate differential protein expression specifically in the presence of wildtype Nsp1 but not K164A/H165 mutant or EGFP. B Schematic representation of transfecting in vitro synthesized mRNAs into cells expressing either wildtype or K164A/H165A Nsp1, in which the Fluc transcripts contain various tested 5′UTRs, while the Rluc has constant 5′UTR as control. C The RLA derived from each 5′UTR in the cells expressing wildtype Nsp1 and in the cells expressing K164A/H165A mutant Nsp1 were shown. Particular 5′UTRs exhibit Nsp1-specific selective translation. D and E Normalized RLA(Nsp1 WT/Nsp1 K164A-H165A) were calculated for the assessment of how the Nsp1-specific selectivity was altered by differential sequence manipulations. The 5′UTR sequences of tested mRNAs are shown. Mutated sequences were specified, while sequences identical to the wildtype were indicated as “.”. The two additional 5′-terminal Gs generated with the in vitro mRNA synthesis by T7 promoters are shown in lower cases, while the start codon of Fluc reporter is underlined. Data were shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.)

Transfection of synthesized RNA

Synthesized 5-capped 3′-A₃₀-tailed RNA transcripts were transfected into cells with Lipofectamine® (Invitrogen, USA) according to manufacturer’s protocol. In brief, cells transfected by either pcDNA expressing wildtype or K164A/H165A mutant nsp1 were cultured for 36–40 h on 48-well plates, followed by transfecting 500 ng Fluc and 250 ng Rluc transcripts per well in triplicate (Fig. 7B). Eight hours after RNA transfection, the cells were subjected to the measurement of luciferase activities.

Measurements of luciferase activities

The relative luciferase activity was measured by Dual-Glo® Luciferase Assay System (Promega, USA) according to manufacturer’s protocol. In brief, transfected cells were lysed with 1 × passive lysis buffer supplied in the kit and subjected to measurement for the relative light units by supplying Fluc-specific substrate, followed by injecting Stop&Glo buffer which terminates Fluc reaction while provides Rluc specific substrate.

Cell culture and co-transfection experiments

HEK-293 T cells were cultured to ~ 60% confluent with Dulbecco’s Modified Eagle Medium (DMEM) in 48-well culture plates, followed by transfection with PolyJet™ In Vitro DNA Transfection Reagent (SignaGen, USA) according to manufacturer’s protocol. In the experiments of co-transfection, 100 ng of pcDNA and 100 ng of reporter pRLFL were premixed and used for transfections for each well (Fig. 1A). Culture media were renewed after 24 h, followed by harvesting the cells 36–40 h post transfection. Harvested cells were subjected to measurements of Fluc and Rluc activities.

Western and Northern blotting

The levels of protein and RNA were verified by Western and Northern blots, respectively. In brief, cultured cells were harvested with trypsinization and lysed by RIPA lysis buffer (25 mM Tris–HCl pH7.6, 150 mM NaCl, 1% NP-40, 1% Sodium deoxycholate, 0.1% SDS) for total protein extraction. Each Western sample contains 30 ug total protein that was denatured in loading buffer (50 mM Tris·HCl, pH 6.8, 2% SDS, 6% Glycerol, 2 mM DTT, 0.01% Bromophenol Blue) by heated at 95 °C for 5 min, followed by separation with denature protein electrophoresis (SDS-PAGE). Separated protein samples were transferred to PVDF membrane, followed by staining with specific primary (anti-c-myc, anti-beta-actin, anti-Fluc, anti-Rluc, anti-EGFP, etc.) and HRP-conjugated secondary (anti-rabbit or anti-mouse) antibodies. RNA samples were prepared following the protocol described by Mordstein et al. (2020) with TRIzol™ Reagent (Invitrogen, USA). Cytosolic fractions of RNAs were separated with denature agarose electrophoresis, followed by transferring to nylon blotting membranes. Transferred membrane were probed with digoxigenin (DIG) labelled Fluc- or Rluc-specific probes and then documented by chemiluminescence imager.

Results

The 5′-proximal structural elements that promote escapes from Nsp1-directed inhibition of protein synthesis

To verify how the 5′UTR may affect protein synthesis in the presence of cellular Nsp1 in vivo, we have constructed dual luciferase reporter vector pRLFL, in which Firefly and Renilla luciferases are independently driven by human cytomegalovirus (CMV) promoters, respectively (Fig. 1A). The CMV promoter is one of the most commonly used promoter for strong expression of exogenous proteins in human cells, and its well-studied transcription start site (TSS) makes it suitable for the study of 5′UTRs (Even et al. 2016). In this approach, relative luciferase activity (RLA) represented the expression level of Fluc (with changeable tested 5′UTRs) relative to the control Rluc (with constant control 5′UTR as 5′-gtcagctagccacc-3′) (Fig. 1A). The 5′-leader sequence (LS) which is universally possessed by SARS-CoV-2 sgmRNAs have been suggested to promote an escape from Nsp1-directed inhibition of protein synthesis for sufficient viral protein expression (Lapointe et al. 2021; Yuan et al. 2020). Thus, we first constructed two reporter plasmids, pRLFL-SARS2-5′LS and pRLFL-empty, in which the Fluc 5′UTRs were engineered having the 5′-LS and the non-specific 5′-sequence (5′-aagcttggcattggtaccgcagccaccggt-3′ derived from the cloning vector), respectively, and tested in cells expressing differential cellular level of Nsp1.

To minimize potential transfection error, the total amount of pcDNA to be transfected was kept 100 ng per 1 × 10⁵ cells, in which the pcDNA-SARS2-Nsp1 was serial titrated with pcDNA-EGFP accordingly, followed by transfection of 100 ng reporter pRLFL-SARS2-5′LS or pRLFL-empty. Fig. 1B showed the RLA resulted from 5′-LS relative to the 5′-non-specific sequence in different cellular contexts with increasing levels of cellular Nsp1. In the absence of Nsp1, the 5′-LS resulted in slightly higher RLA than the non-specific 5′-sequence, suggesting that the two Fluc 5′UTRs lead to slightly differential protein expressions relative to the control Rluc 5′UTR. Notably, the differences between the two Fluc 5′UTRs were much higher in the presence of cellular Nsp1, indicating that the 5′-LS is capable of mediating relatively higher protein expression than the non-specific 5′-sequence (Fig. 1B). This result suggested that in the cells expressing Nsp1, exogenous genes with the viral 5′UTRs are preferable over the non-specific 5′UTR for protein synthesis.

To eliminate the potential interferences caused by the differential natures of each tested 5′UTRs, such as length and thermostability, which were irrelevant to the presence of cellular Nsp1, the RLA(Nsp1/EGFP) was calculated for each construct, which is the RLA of a certain tested 5′UTR measured in the presence of cellular Nsp1 (with 100 ng pcDNA-SARS2-Nsp1) relative to what measured in the absence of Nsp1 (with 100 ng pEGFP). This allowed us to independently assess the relative expression efficiencies mediated by different tested 5′UTRs specific to the presence of cellular Nsp1. It was shown that SARS-CoV-2 5′-LS mediates significantly higher protein expression than the non-specific 5′-sequence does in a Nsp1-specific manner (Fig. 1C). Utilizing similar approach, we further tried to identify the minimal sequence of the viral 5′UTR required for promoting the protein expression in cells expressing Nsp1.

The SARS-CoV-2 5′-LS consists of 3 structural elements designated as Stem-loop (SL) 1, 2, and 3 which are followed by SL4 and 5 (Fig. 2A). These structural elements are particularly conserved in alpha- and beta-CoVs and constitute the 5′UTR of gmRNA which translates into polyprotein 1a/b (Chen and Olsthoorn 2010; Chen et al. 2021; Miao et al. 2021; Sun et al. 2021). To verify if these elements alone were capable of mediating sufficient protein expression, the RLA(Nsp1/EGFP) of differentially truncated (5′T and 3′T series) and deleted (D series) SARS-CoV-2 5′-proximal sequences were tested, respectively (Fig. 2B). Figure 2C showed that the truncation of SL1-corresponding 5′-terminal 40 nts (5′T41-295) resulted in significant reduction of protein expression comparing with the 295-nt full-length 5′-proximal sequence (F1-295), while further serial truncations toward to the 3′ gradually caused marginal decline. Notably, a bounce of protein expression appeared between 5′T_150-293 and 5′T_183-293, suggesting that the sequence of nts 183 to 293, which encompasses SL5a/b, promoted protein expression in a certain extent when situated close to the 5′-terminus. Results from 3′T series, however, showed that differential truncations to the 3′ do not largely affect the protein expression as long as the 5′-terminal 76 nts remains (Fig. 2C). Fig. 2D further showed that among the deletion mutants, D_1-44 had the most significantly reduced protein expression. Deletions of other regions, however, did not have pronounced effects. These results suggested that sequences corresponding to SL1 and SL5a/b are critical for mediating sufficient protein expression in the presence of cellular Nsp1.

We then verify the effectiveness of SL1 and SL5a/b alongside SL2/3 and SL4, respectively. It should be noted that a point mutation A107U was applied to construct SL4 to disable the upstream AUG start codon that may interfere the reporter assay. Fig. 2E showed that 5′-proximal SL1 and SL5a/b significantly mediate stronger protein expression comparing with SL2/3 and SL4 in the presence of cellular Nsp1. We also tested other SL-like structural elements derived from Mouse hepatitis virus Packaging Signal (MHV-PS) and human alpha-synuclein iron-responsive-element (SNCA-IRE), respectively (Chen and Olsthoorn 2019; Chen et al. 2007) and found no significant effect (Fig. 2E). These results collectively indicated that the 5′UTRs containing proximal SL1 and SL5a/b derived from SARS-CoV-2 gRNAs, but not other viral or non-viral SLs, are capable of mediating sufficient protein expression in the presence of cellular Nsp1, suggesting that, in addition to the SL-like structural features, there are other critical determinants of 5′UTR-mediated escapes from Nsp1-directed translational inhibition, e.g. specific structural and/or sequence contexts.

Unstructured 5′UTRs with high 5′-proximal A/U content result in selective enhancement of protein expression in the presence of cellular Nsp1

By analysing the sequence bias of all reference alpha- and beta-CoV gRNAs listed in NCBI database, we found that the 5′-proximal sequences of these CoV gRNAs consist of majorly As and/or Us (Table S1), particularly within the first 40 nts of the 5′end corresponding to SL1. We wondered how the alterations of the A/U-content and/or the structural integrities of SL1 may affect protein expression in the presence of cellular Nsp1. Mutagenesis studies showed that disruptions of the SL-structure by introducing multiple mismatches to the stem region did not significantly reduce the relative protein expression (Fig. 3A), suggesting that the integrity of the SL-like structure may not be the sole determinant for SL1-mediated escape from Nsp1-directed inhibition of protein expression. Intriguingly, introducing multiple C to G, to A, or to T mutations that seriously alter the sequence context and disrupt the structural integrity resulted in differential effects. Fig. 3A showed that substitutions of C for G, but not C for T, greatly lowered the protein expression. This result suggested that high content of Gs in the 5′UTR is unfavourable for protein expression in the presence of cellular Nsp1, regardless of the integrity of the specific structure of SL1. We found similar trends in SL5a/b mutagenesis studies that substituting the loop sequences of SL5a and b with hexa-G, but not hexa-A, hexa-C, or hexa-T, decreases the protein expression, although the SL-like structure is mainly preserved. Deletion of the bulges in SL5a/b, on the other hand, does not have pronounced effect (Fig. 3B). These results collectively suggested that the sequence contexts of 5′UTRs are critical determinants of protein expression in the presence of cellular Nsp1. To be more specific, Gs are unfavourable for 5′UTR to mediate efficient protein expression in the presence of cellular Nsp1.

We further study how the specific sequence contexts of 5′UTRs may affect protein expression in the presence of cellular Nsp1 independent to particular RNA structures, by adapting an unstructured A/U-rich sequence derived from nts 128–156 of SARS-CoV-2 gRNA. Fig. 4A showed that substitutions of A for C reduced the relative protein expression, while substitutions of A for G substantially abolished its ability of mediating efficient protein synthesis in the presence of cellular Nsp1. Similar trends have been found in substitutions of T for C and G, respectively (Fig. 4A). These data suggested that 5′UTRs with high A/U content are capable of promoting protein expression in the presence of cellular Nsp1, while the presence of Gs is exceptionally unfavourable. It should be noted that internal deletion of the region corresponding to nts 128–156 does not significantly alter the protein expression as long as the SL1-corresponding region remains (Fig. 2D). These results implied that the expression efficiency of exogenous proteins in the presence of cellular Nsp1 is predominantly determined by the sequence located proximally to the 5′end, while sequences distal to the 5′end have merely minor effect, in agreement with what we have found for SL5a/b. We further verified this finding by swapping the high-A/U portion of nts 128–156 with the downstream non-A/U-rich sequences. Fig. 4B showed that the swapped 5′UTR is incapable of mediating relatively high protein expression in the presence of cellular Nsp1. Western blots were in agreement with luciferase assays, showing relatively higher level of Fluc proteins for those 5′UTRs with high proximal A/U content, while sequence swapping or G substitutions greatly reduce the level of Fluc proteins in the presence of cellular Nsp1 (Fig. 4C). Here we also observed that the structured 76-nt 5′-LS and the 295-nt full-length 5′UTR of SARS-CoV-2 gmRNA do promote escapes from Nsp1-directed inhibition of protein synthesis, showing relative less-repressed expression of Fluc than Rluc (Fig. 4C). However, a more noticeable Nsp1-dependent enhancement was detected for the untrusted 5′UTRs with 5′-proximal A/U-rich sequences. For instance, in the background of strongly inhibited expressions of Rluc in all constructs that validate the effectiveness of Nsp1-directed translational repression in cells, the A/U-rich nt 128–185 and nt 128–156 resulted in relatively higher levels of Fluc protein in the presence of cellular Nsp1 than in the presence of EGFP. On one hand, these AU-rich 5′UTRs seemed to be less translational active in the absence of Nsp1, but on the other hand, they exhibit relatively high activity in the cells expressing Nsp1. This unique feature made the A/U-rich sequences capable of mediating Nsp1-dependent selective protein expressions, showing high fold changes of protein level in the presence of cellular Nsp1 relative to that in the absence of Nsp1.

The 5′-proximal Cs and Gs distinctly promote and impede the protein expression in cells expressing Nsp1, respectively

As the A/U-rich 5′-proximal sequences were found to promote protein expression in cells expressing Nsp1, we further verified the sequence specificities and the potential positional effect of these 5′UTRs in mediating protein expression in the presence of cellular Nsp1. Fig. 5A showed that a couple of G substitutions significantly reduced relative protein expression, while C substitutions resulted in an enhancement. This result revealed the distinct effects caused by Cs and Gs in the background of the A/U-rich short 5′UTR, respectively. We further test the positional effect by gradually shifting the couple of Gs away from the 5′-terminus. Figure 5B showed that the 5′-proximal Gs have stronger effects on abolishing the Nsp1-specific enhancement of protein expression than the 5′-distal Gs. Previous reports have shown that Nsp1 induces degradations of host mRNAs (Burke et al. 2021; Huang et al. 2011b; Nakagawa and Makino 2021; Nakagawa et al. 2018). Here we also found that the mRNA level is significantly lower in the presence of cellular Nsp1 (Fig. 5C), particularly the construct SP9 which had Gs close to the 5′-proximus and showed strongly inhibited protein expression in the presence of Nsp1 (Fig. 5B). The extremely A/U-rich constructs SP28 and 29, however, showed lower mRNA stability independent to Nsp1 while retain relatively less degraded mRNA level in the presence of Nsp1 than SP9. This finding was similar to what has been found for nt 128–156 of SARS-CoV-2 5′UTR that resulted in relatively low basal protein expression but exhibited Nsp1-dependent enhancement (Fig. 4C). We further verified the necessity of the 5′-proximal locations for the Gs to show their effects. Fig. 6A showed that the Gs located specifically in a window of 12th to 17th positions in the 5′UTRs resulted in the most significant reduction of protein expression. In contrast to the Gs, 5′-proximal Cs were found to enhance the protein expression in the presence of cellular Nsp1, by comparing FG1 with FG3, 5, and 7 (Fig. 6B). Notably, positional effect was also observed for such enhancement mediated by Cs that 5′-proximal Cs result in higher relative protein expression than the distal Cs. However, a couple of G substitutions were found to seriously impede the protein expression (Fig. 6B). These results indicated that the 5′-proximal Cs promote the protein expression in the presence of cellular Nsp1, while Gs are strong negative effectors in contrast.

Recent structural studies suggested that the C-terminal domain (approximately spanning from a.a. 145 to 180) of Nsp1 folds into two helices upon binding to the mRNA entry tunnel of the 40S ribosomal subunit (Banerjee et al. 2020; Schubert et al. 2020; Thoms et al. 2020; Yuan et al. 2020). Consequently, loading and accommodation of cellular mRNAs to the entry tunnel was physically prevented, and the translation in cells is generally shutoff. These studies have shown that Nsp1 with K164A and H165A mutations loses its translational repressive activity (Nakagawa and Makino 2021). Thus, we tested if the 5′UTR-mediated selective protein expression is sensitive to the K164A/H165A mutations. Fig. 7A showed that the viral 5′-proximal 1–295 nts, in comparison with the non-specific 5′-sequence, resulted in significant differential relative protein expressions in cells expressing wildtype Nsp1 but not the K164A/H165A mutant or EGFP. This finding suggested that the Nsp1-directed translational repression is involved in the differential expression mediated by viral 5′UTR in the presence of cellular Nsp1. We then tested if exogenously delivered mRNAs baring particular 5′UTRs also showed Nsp1-specifc differential expression. To do so, in vitro synthesized translatable Fluc mRNAs with various tested 5′UTRs were premixed with control Rluc mRNAs and transfected into cells expressing wildtype Nsp1 and the defect K164A/H165A mutant, respectively (Fig. 7B). In this approach, RLAs represented the relative translation activities mediated by certain tested Fluc 5′UTRs relative to the control Rluc 5′UTRs in the two cellular contexts, respectively. For transcripts with the non-specific 5′-sequence as their 5′UTRs, the relative translation activity in cells expressing wildtype Nsp1 did not significantly differ from that in cells expressing the K164A/H165A mutant (Fig. 7C), suggesting that the non-specific 5′-sequence was not capable of mediating significant escapes from Nsp1-directed translational repression comparing with the control 5′UTRs. However, the viral structural elements, SL1 and SL5a/b, and the unstructured AU-rich sequence nt 128–156 were shown to mediate escapes, leading to relatively higher RLA in the presence of wildtype Nsp1 than that in the presence of Nsp1 mutant, respectively. The nt 128–156 sequence with A to G mutations, however, failed to mediate apparent selective translation (Fig. 7C), in accordance to the negative roles we have found for the G residues. For better assessment of the translational activities in an Nsp1-specific manner, the RLA(Nsp1 Wt/Nsp1 K164A-H165A) that is the RLA of a certain tested 5′UTR measured in cells expressing wildtype Nsp1 relative to what measured in cells expressing the mutant was calculated. Fig. 7D showed that the G residues were exceptionally unfavoured to mediate efficient translation in the presence of Nsp1 in comparison with other bases, while Fig. 7E revealed that the 5′-proximal Gs more significantly reduced the translation efficiency than the distal Gs did. These results collectively indicated that Nsp1-dependent selective translation can be achieved by delivering synthesized mRNAs with particular 5′UTRs into cells. And the sequence preferences and the positional effects also applied to these delivered mRNAs.

The optimized 5′UTR for strong protein expression resistant to the translationally inhibitory Nsp1

Taking advantage of the knowledge we have found for the 5′UTR-mediated escape from Nsp1-directed translation inhibition, it is possible to design an artificial 5′UTR to achieve high exogenous protein level regardless of the presence of translational repressive SARS-CoV-2 Nsp1. One of the most simplified principles for 5′UTR optimizations could be to introduce CAT-motifs to the 5′-proximal regions of the mRNA transcripts that increase the local A/U content with the pro-expression Cs while leaving out Gs. Fig. 8A showed that the 5′UTR with CAT-motifs, CAT2, results in strong protein expression in the presence of EGFP and Nsp1, respectively. In comparison with the 5′-sequence derived from the 5′UTR of Eukaryotic elongation factor 2 (EEF2), in which the terminal oligopyrimidine (TOP) track has been suggested to generally promote translation efficiency (Weber et al. 2021) and preferentially escape from the global suppression of translation induced by SARS-CoV-2 Nsp1 (Rao et al. 2021), the CAT motif-containing 5′UTR leads to higher protein level in the presence of cellular Nsp1 (Fig. 8A), probably due to the absence of Gs in the particular 5′-proximal window which is a key determinant for the 5′UTR-mediated escapes (Fig. 6A). We further verified the relative translational activities by studying CAT2 5′UTR in cells expressing Nsp1. Fig. 8B showed that in the presence of cellular Nsp1, CAT2 mediated significant higher relative translation than its mutant baring two C to G mutations. However, such difference was much less pronounced in the presence of K164A/H165A mutant, showing that such selective enhancement of translation is specific to the presence of functional Nsp1 and impeded by the G residues located in the 5′-proximal window. Fig. 8C showed the results of a detail study on the sequence specificity of the 5′UTR-mediated escapes from Nsp1-directed translational repression by delivering synthesized transcripts into cells expressing wildtype and mutant Nsp1, respectively. It confirmed that the Gs located in a particular window of the 5′-proximus resulted in the most significant Nsp1-directed translational repression (Fig. 8C). It also verified that additional distal Cs may not promote translation further when the proximal Cs were present. We have also found that wildtype Nsp1 induced significant RNA degradations to the exogenous transcribed mRNAs comparing with the K164A/H165A mutant (Fig. 8D), as previously reported for host mRNAs (Burke et al. 2021; Huang et al. 2011b; Nakagawa and Makino 2021; Nakagawa et al. 2018). It was shown that the constructs resulting in less inhibited protein synthesis accordingly have less significant mRNA decays (Fig. 8D). For instance, the construct CAT2 which showed relatively high protein expression in the presence of cellular Nsp1 (Fig. 8A) exhibited significantly less Nsp1-induced mRNA degradation (Fig. 8D). However, construct CAT2-mut which has a couple of 5′-proximal Gs exhibited strong Nsp1-dependent mRNA degradation (Fig. 8C). These results again revealed the correlations between the escapes of translational repression and mRNA degradations.

Fig. 8.

Selective enhancement of protein expression by engineered 5′UTR with CAU motifs. A Western blots showing the protein expression levels of the reporters by delivering pRLFLs into cells expressing EGFP or Nsp1 as illustrated in Fig. 1A, in which the Fluc were engineered baring CAT2 and EEF2 sequences as their 5′UTR, respectively. The former is an artificial sequence containing two CAU motifs, while the later encompasses the 5′-terminal sequence of human EEF2 mRNA. Significantly inhibited Rluc validated the general effectiveness of Nsp1-directed inhibition, while the expression level of Fluc varied depending on the 5′UTRs. B In vitro synthesized mRNA transcripts baring CAT2 5′UTR were transfected into cells either expressing wildtype or K164A/H165A mutant Nsp1 as illustrated in Fig. 6B, showing significant translational enhancement in cells expressing wildtype Nsp1. The two C to G mutations were shown to largely abolish the Nsp1-specific selective translation. Data were shown by the mean ± SD of three independent experiments. P-value: < 0.05(*), < 0.01(**), < 0.001(***), < 0.0001(****), ≥ 0.05(n.s.)

The host gene-derived 5′-proximal sequences that mediate the escapes

With the knowledge of the sequence contexts critical for escaping from the Nsp1-directed translational repression, we were able to investigate if any 5′-proximal sequences derived from host genes promote sufficient protein expression in the presence of cellular Nsp1, although it had been previously suggested that Nsp1 may induced global shutoff of host protein synthesis (Beyer and Forero 2021; Fung et al. 2016; Nakagawa and Makino 2021; Tanaka et al. 2012; Zhang et al. 2021b). Fig. 9 showed that the 5′-proximal sequences derived from Adhesion G Protein-Coupled Receptor E1 (ADGRE1), Butyrophilin Subfamily 3 Member A2 (BTN3A2), CD68 Molecule (CD68), and TRAF-Interacting Protein With Forkhead-Associated Domain Family Member B (TIFAB) mediate relatively stronger escapes in the presence of cellular Nsp1, while 5′-sequences derived from CCR4-NOT Transcription Complex Subunit 10 (CNOT10), Afamin (AFM), Caspase 4 (CASP4), and Killer Cell Lectin Like Receptor C4 (KLRC4) did not result in pronounced protein expression. This result again concluded that the presence of G at the specific 5′-proximal window significantly abolish the capabilities of A/U-rich 5′UTRs in mediating sufficient escapes (Fig. 9). Among these 5′-sequences, some contain upstream AUG start codons that potentially interfere the translations of reporter. However, we did not observe much difference between constructs ADGRE1 and ADGRE1-ΔG, in which the upstream start codon was removed in the later transcript. Collectively, these finding strongly suggested that particular host genes, in which 5′-proximus exhibits the sequence contexts similar to the optimized 5′UTRs we have identified, could be sufficiently expressed regardless of the high cellular level of nsp1 upon SARS-CoV-2 infections. These host genes capable of escaping from Nsp1-directed inhibition could provide evolutionary advantages for individuals to survive from CoV infections, including genes related to fundamental cellular process and/or antiviral responses.

Fig. 9.

Escapes from nsp1-directed inhibition of protein synthesis mediated by 5′-proximal sequences derived from host genes. The RLA measured in cells expressing Nsp1 relative to that measured in cells expressing EGFP were shown, in which the Fluc mRNAs contain differential 5′-proximal sequences derived from human ADGRE1, BTN3A2, CD68, TIFAB, CNOT10, AFM, CASP4F, and KLRC4, respectively. Data were shown by the mean ± SD of three independent experiments. The proximal Cs were highlighted with red colour, while Gs with blue

In conclusion, besides the viral structural element SL1 and SL5a/b, we have shown that unstructured 5′UTRs bearing high A/U content in the 5′-proximal regions are capable of promoting an escape from Nsp1-directed inhibition of protein synthesis, leading to sufficient expressions of exogenous proteins in the presence of cellular Nsp1. In addition, the 5′-proximal Cs and Gs were found to promote and impede protein expression in an Nsp1-specific manner, respectively. With this new knowledge, specific 5′UTRs can be rationally optimized to achieve regulated and/or enhanced protein expressions even in the presence of the translational inhibitor Nsp1. Here we have reported a comprehensive study on non-viral 5′UTRs that are capable of mediating escapes from Nsp1-directed translational repression and mRNA degradation. Such selective enhancement mediated by engineered 5′UTRs can be achieved either by delivering plasmid-based DNA vectors or synthesized translatable mRNAs, improving the DNA/mRNA therapy’s tool kits for sufficient protein expression in particular cellular contexts with SARS-CoV-2 Nsp1 or other similar translational repressors. Finally, identification of host genes which are potential escapees from Nsp1-directed inhibition provided fresh perspectives for future studies on host-virus interactions.

Discussion

The optimization of 5′UTRs for exogenous protein expression

To date, rational engineering of 5′UTRs for enhanced protein expression is still challenging, although quite a few regulatory elements that control gene expression in cells have been identified (Asrani et al. 2018; Cao et al. 2021; Ding et al. 2018; Sample et al. 2019). It is intriguing, yet even more challenging, to predict or design pro-expression 5′UTRs in specific cellular context, such as in the presence of translational repressor in particular. Here we have revealed the key determinants for the 5′UTRs to mediate escapes from SARS-CoV-2 Nsp1-directed inhibition of protein synthesis. This new knowledge made the rational design of pro-expression 5′UTRs possible to achieve strong and/or selective expression of exogenous proteins in the presence of cellular Nsp1. These pro-escaping engineered 5′UTRs, such as the CAT motif-containing sequence, are particularly applicable for DNA/mRNA therapies and/or vaccinations to those patients experiencing or having the risk of SARS-CoV-2 infections. Presumably, delivered DNA or mRNA having the engineered 5′UTRs should result in sufficient protein expression, regardless of the Nsp1-directed translational inhibition upon infections. However, to what extent the escapes from translational inhibition may take place upon infection remains to be studied in the future. Besides translational repression, SARS-CoV-2 has been recently proposed to use a multipronged strategy to impede host protein synthesis, including inhibitions of nuclear export of cellular mRNAs (Finkel et al. 2021; Nakagawa and Makino 2021; Puray-Chavez et al. 2022; Zhang et al. 2021a). This nature highlights the advantage of mRNA therapy for that cytosolic delivery of translatable synthesized mRNAs can bypass the potential barrier of nuclear exportation. Nonetheless, an alternative approach to promote the mRNA export when DNA vectors were used could be the introduction of AU-rich elements (AREs) or constitutive transport element (CTE) into the 3′UTRs (Katahira 2015). These optimized transcripts/cassettes that promote protein expressions could be applied to next-generation mRNA/DNA vaccines for versatile and enhanced vaccination (Callaway 2022; Kumar et al. 2022).

The 5′UTR-mediated escape may similarly apply to endogenous host genes

A big question in Nsp1-directed translational inhibition is that if certain host genes, e.g. antiviral response-related genes, are particularly susceptible or resistant to such inhibition, since differential regulation on these cellular proteins may largely determine the severity of virus infection. Our experimental results have revealed the 5′UTR-mediated escape for exogenously delivered DNA/mRNA, and it is very intriguing to study in the future whether such 5′UTR-mediated escape also applies to endogenous host genes, with similar or dissimilar sequence contexts. Several previously reported proteomic analysis in virus-infected cells have found that the cellular protein expression profile is largely changed upon virus infection, yet limited consensus knowledge was found regarding the translational regulations in these studies, despite some abundant stress/infection-related proteins were shown upregulated (Bojkova et al. 2020; Meyers et al. 2021; Stukalov et al. 2021), probably due to that various infection protocols, cell culture models, and proteomic analysis strategies were used. Also, the drastically altered transcription profile post infection makes it challenging to accurately justify the translational efficiency in genome-wide scale for non-abundant signalling proteins/transcripts. Nonetheless, some other studies have suggested that cellular TOP-containing 5′UTRs are preferably translated in the presence of Nsp1 (Rao et al. 2021; Weber et al. 2021). The TOP, however, did not promote stronger escape (Figs. 4A and 8A) comparing with artificial 5′UTRs, and thus it could be a general, rather than Nsp1-specific, enhancer of translation in stressed cells (Weber et al. 2021; Yang et al. 2022). In the future, it is of urgent importance to study Nsp1-regulated expression profile and its clinical relevance, particularly those are involved in antiviral signalling/response. On the other hand, SARS-CoV-2 has been recently reported to alter the transcription profile of host genes upon infections and potentially result in various mRNA isoforms with variant 5′-proximal sequences for the same genes (Banerjee et al. 2020; Iida et al. 2021), adding another layer of complexity to the genome-wide studies of Nsp1-direced translational regulation in infected cells. Nonetheless, it was reported that the overall translation activities of cellular mRNAs were significantly repressed 5 h post infection due to the expression of repressive Nsp1 (Finkel et al. 2021). Puzzlingly, ribosome foot-printing suggested that the host mRNAs substantially regain translation activities in 8 h post infection relatively to viral mRNAs, while the cellular Nsp1 level is still high. One of the possible explanations is that in the later stage, more variant host cell mRNAs with alternative 5′-proximal sequences which are capable of escaping from Nsp1-directed translational inhibition were accumulated. Future studies shall provide more evidences.

The ultimate 5′-terminus of CoV gRNAs: to structure or not to structure?

The molecular basis of how CoV gmRNA and sgmRNA 5′UTRs may promote efficient production of viral proteins in the presence of high cellular level of Nsp1 was still poorly understood. Recent reports and this study have shown that SL1 alone is capable of promoting an escape from nsp1-directed inhibition of protein synthesis (Bujanic et al. 2022; Shi et al. 2020; Vora et al. 2022). It was suggested that SL1 specifically interact to Nsp1 via its unique structure and discharge the blockage of mRNA entry tunnel (Simeoni et al. 2021). However, we showed that the integrity of secondary structure may not be absolutely needed for SL1 to mediate an escape from translational inhibition, in agreement with some recent studies that report poor structural relevance of SL1 to protein expression (Bujanic et al. 2022; Slobodin et al. 2022). Also, in these studies, they have found that the presence of specific C residues and the absence of G residues in SL1 are critical for sufficient protein expression, respectively, in spite of that the SL-like structure of the SL1 was mainly preserved in their mutagenesis experiments. Here in this study, we have showed that the A/U content is generally high in the 5′-proximal regions of CoV gRNAs and verified the capability of A/U-rich 5′-sequences to mediate strong protein expression in the presence of cellular Nsp1 without specific structures. Yet, the 5′-proximal regions of CoV gRNAs are generally found to be (loosely) structured. One explanation could be that a structured terminus is pivotal for genome replication (Li et al. 2008; Liu et al. 2007; Madhugiri et al. 2018), although mRNAs with strong 5′-proximal structure were generally known to reduce protein synthesis (Doma and Parker 2006). As a result, a dynamically instable SL1 predominantly made of A-U and non-canonical (such as A·C) base-pairs is characteristically present at the 5′-terminuses of CoV gRNAs (Li et al. 2008; Liu et al. 2007; Madhugiri et al. 2018). Such characteristic feature is an example showing that a single RNA cis-acting element may finetune its structural and sequential contexts to fulfil its multiple roles in different scenarios.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

SCC: Conceptualization, methodology, investigation, validation, writing—original draft, supervision, funding acquisition. CTX: Validation, investigation. CFC: Validation, investigation. TYC: Validation, investigation. CCL: Investigation. PWFu: Investigation. CHY: Supervision, writing—review and editing, project administration, funding acquisition.

Funding

This work is supported by the programmes of Research Center for Epidemic Prevention Science (MOST 111-2321-B-006–009) and Columbus Young Scholar Fellowship (MOST 109-2636-B-006–005; MOST 110-2636-B-006-010; MOST 111-2636-B-006 -013; NSTC 112-2636-B-006-009) from Ministry of Science and Technology and National Science and Technology Council, Taiwan. The National Cheng Kung University, Taiwan, and Higher Education Sprout Project, Ministry of Education to the Headquarters of University, Taiwan, also provide funding.

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.