Engineering a PAM-flexible SpdCas9 variant as a universal gene repressor

The RNA-guided CRISPR-associated Cas9 proteins have been widely applied in programmable genome recombination, base editing or gene regulation in both prokaryotes and eukaryotes. SpCas9 from Streptococcus pyogenes is the most extensively engineered Cas9 with robust and manifold functionalities. However, one inherent limitation of SpCas9 is its stringent 5′-NGG-3′ PAM requirement that significantly restricts its DNA target range. Here, to repurpose SpCas9 as a universal gene repressor, we generate and screen variants of the deactivated SpCas9 (SpdCas9) with relaxed 5′-CAT-3′ PAM compatibility that can bind to the start codon ATG of almost any gene. Stepwise structure-guided mutations of the PAM-interacting residues and auxiliary PAM-proximal residues of the SpdNG (5′-NG-3′ PAM) create a PAM-flexible variant SpdNG-LWQT that preferentially accommodates 5′-NRN-3′ PAMs. SpdNG-LWQT is demonstrated to be effective in gene repression with the advantage of customizable sgRNA design in both Escherichia coli and Saccharomyces cerevisiae. This work validates the feasibility of purposeful PAM expansion of Cas9 towards signature PAMs and establishes a universal SpdCas9-based gene repressor.

T he bacterial type II clustered, regularly interspaced, short palindromic repeats (CRISPR/Cas9) systems have been extensively harnessed for programmable genome editing in both prokaryotic and eukaryotic cells via RNA-guided cleaving or nicking the gene target of interest [1][2][3][4][5] . Engineering and repurposing the nuclease-deficient Cas9 (dCas9) has derived various applications, including base editing, DNA transposition, gene interference (CRISPRi), and activation (CRISPRa) [6][7][8][9][10][11][12] . Recruitment of (d)Cas9 to target DNA needs two specificity checkpoints: either a dual CRISPR RNA (crRNA)-trans-activating CRISPR RNA (tracrRNA) guide or a chimeric single-molecule guide RNA (sgRNA) that specifically pairs with the target DNA strand, and a protospacer adjacent motif (PAM) immediately downstream of the protospacer on the complementary strand 13 . The sgRNA consists of a Cas9-binding RNA structure and a target-specific complementary region, rendering it easily reprogrammable to target virtually any genomic site. However, the stringent PAM requirement is a major constraint that limits the targeting scope of (d)Cas9 and thus its wide applications, especially when precise positioning is required.
The most direct solution to ease the PAM restriction would be to engineer Cas9 variants with altered or broader PAM specificities. Although mining Cas9 orthologs from a multitude of microbial resources could potentially identify Cas9s with alternate or minimal PAMs, protein characterization as well as application validation in various genetic contexts is still laborious [14][15][16][17] . Instead, engineering well-characterized Cas9s to recognize target or expanded PAMs would be more straightforward and purposeful 18,19 . Among known Cas9s, Streptococcus pyogenes Cas9 (SpCas9) has been most extensively engineered due to its wide and robust in vivo applications. Structure-activity relationship investigations on the SpCas9-sgRNA-DNA complex have revealed its PAM recognition mechanism, laying the structural basis for rewiring the PAM preferences via protein engineering [20][21][22][23] . SpCas9 recognizes the 5′-NGG-3′ PAM by a pair of arginine residues (R1333/R1335) within the PAMinteracting (PI) domain inserted into the major groove of the PAM DNA duplex 20 . In order to modify the PAM specificity, earlier efforts utilized directed evolution to generate the VQR, EQR, and VRER variants that recognize altered 5′-NGA-3′, 5′-NGAG-3′, and 5′-NGCG-3′ PAMs, respectively [24][25][26] . Since then, intensive efforts have been focused on creating SpCas9 variants with relaxed PAM specificities, ranging from xCas9 and SpCas9-NG targeting 5′-NG-3′ PAM to SpCas9-NRRH, SpCas9-NRTH, and SpCas9-NRCH that collectively recognize 5′-NRNH-3′ PAMs 19,27,28 . Very recently, a variant SpRY has been developed with almost no PAM constraint (5′-NRN-3′ > 5′-NYN-3′ PAMs) 29 . These endeavors significantly expanded the targetable space of Cas9 and demonstrated the feasibility of engineering Cas9 to accommodate noncanonical PAMs in a purpose-driven manner.
In this work, we propose to create SpdCas9 variants capable of recognizing and binding to gene start codons, which could potentially serve as universal gene repressors. In biological systems, AUG is the most commonly used start codon, whereas non-AUG start codons are rare in eukaryotic genomes, while prokaryotes permit frequent use of alternate start codons 30,31 . In Escherichia coli, 83% of all genes (3542/4284) start with AUG and the other 17% initiate with alternate non-AUG codons like GUG and UUG 32 . CRISPRi targeting DNA sites close to start codons could achieve a comparable inhibition efficiency as opposed to targeting the -35/-10 boxes on promoter or ribosome binding site    The -35 and -10 boxes, and ribosome binding site (RBS), are shown in gray. +1, transcription initiation site. b PAM-interacting (red) or proximal residues (blue) in SpCas9wt (PDB ID: 4UN3) and SpNG (PDB ID: 6AI6). PAM sequence is shown in yellow. c The dual-plasmid eGFP repression system for SpdCas9 variant characterization: pCS27 containing Plpp1-controlled SpdCas9wt or variants and pZE12-luc containing P L lacO1-controlled eGFP and sgegfp-TGG or sgegfp-CAT. d Impact of mutating PAM interaction residues (R1333/R(V)1335/T(R)1337) in SpdCas9wt, dxCas9-3.7, and SpdNG on 5′-TGG-3′ (blue) and 5′-CAT-3′ (orange) PAM recognition. The residues in shaded boxes correspond to residues in unmutated SpdCas9wt, dxCas9-3.7, and SpdNG, respectively. e Impact of combinatorial mutations of R1333/V1335 in SpdNG on 5′-TGG-3′ (blue) and 5′-CAT-3′ (orange) PAM recognition. NC, E. coli BW25113(F′) cotransformed with the empty pCS27 plasmid and pZE-eGFP-sgegfp-TGG or pZE-eGFP-sgegfp-CAT. Data indicated the mean ± standard deviation (n = 3 independent biological replicates). Source data are provided as a Source Data file.
(RBS) that are common targets for efficient gene repression 9,33 . Thus, instead of searching for the randomly distributed 5′-NGG-3′ PAM within the 5′ untranslated region, dCas9 variants recognizing the featured nucleotide triplet (ATG) could be readily repurposed as ideal gene-specific repressors with customizable sgRNA design. CRISPRi targeting start codons also minimizes any undesirable interference of upstream genes, especially considering the compact nature of prokaryotic genomes 34 .
Since RNA-directed dCas9 binding to the nontemplate DNA strand of gene coding sequence could afford effective gene repression 9,35,36 , we herein aim to mutagenize SpdCas9 to recognize the noncanonical 5′-CAT-3′ PAM for binding to the ATG start codons. By employing structure-based mutagenesis and an eGFP repression assay, we obtain one SpdCas9-NG derived variant named SpdNG-LWQT that exhibited expanded compatibility toward 5′-NRN-3′ and some 5′-NYN-3′ PAMs. We further validate its 5′-CAT-3′ PAM recognition by restoring and substantiating its nuclease activity (SpNG-LWQT) both in vivo and in vitro. Finally, with eGFP repression or mevalonate production enhancement, we demonstrate the application of SpdNG-LWQT as a universal gene repressor in both prokaryotic and eukaryotic cells. This work generates a dCas9 variant with relaxed and desired PAM specificity, which could serve as a programmable transcriptional repressor covering any gene. More broadly, the variant described here could also extend to other Cas9 or dCas9-based applications.
Structure-guided improvement of the activity of SpdNG-QT. To further improve the activity of SpdNG-QT, we attempted to promote the stacking interaction between Cas9 and PAM by mutating the key residues around the PAM-Cas9 duplex, including G1104, D1135, S1136, S1216, and E1219 (Fig. 2a). We hypothesized that mutating these PAM adjacent residues could afford expanded PAM tolerance. In SpCas9wt, E1219 forms a salt bridge with R1335, stabilizing its PAM specificity to 5′-NGG-3′ 20 . E1219 mutations have been frequently observed in evolved SpCas9 variants and demonstrated to be critical in relaxing PAM stringency 19,27,38,39 . Thus, we first focused on investigating the impact of the E1219 mutation on PAM recognition. Since SpdNG-QT harbors an E1219F mutation, we replaced F1219 with smaller residues including A, Q and V. The eGFP repression assay showed that only F1219V could further improve its activity toward 5′-CAT-3′ PAM, achieving 77.3% repression (Fig. 2b).

PAM interaction analysis via molecular dynamics (MD) simulations.
To support the evidence that mutated Cas9 can recognize 5′-CAT-3′ PAM, we conducted MD simulations to predict molecular interactions between the PI domain and the PAMs. Four independent simulations were performed by using SpNG (PDB ID: 6AI6) or its mutant SpNG-LWQT binding to 5′-TGG-3′ or 5′-CAT-3′ PAM (i.e., SpNG-TGG, SpNG-CAT, SpNG-LWQT-TGG, and SpNG-LWQT-CAT) (Fig. 4a). Based on trajectory files, we calculated the minimum distances between PAMs and the residues on the PI domain. As a result, we found that the minimum distance between R1333 and dC1 (or dA2) on SpNG-CAT is longer than 6.0 Å and also longer than other structures after 45 ns, suggesting that R1333 can scarcely interact with 5′-CAT-3′ PAM (Fig. 4b, c). A similar scenario occurred between V1335 and dC1 (or dA2) (Fig. 4d, e). Notably, after we mutated four amino acids on the PI domain (V1135L/S1136W/R1333Q/V1335T), we found that the minimum distance between Q1333 and dC1 (or dA2) is shorter than that of R1333 and dC1 (or dA2) on SpNG-CAT after 45 ns (Fig. 4b, c). For T1335 and dC1 (or dA2), the minimum distance is shorter than V1335 and dC1 (or dA2) after about 5 ns (Fig. 4d, e). These results demonstrated that the mutated PI domain in SpNG-LWQT may have slightly stronger interaction with PAM than that on SpNG. In addition, the mutated SpNG-LWQT also interacts with 5′-TGG-3′ PAM. Except that the distance between R1333 and dG2 is significantly shorter than the distance between Q1333 and dA2 (Fig. 4c), we found that the minimum distances between SpNG  TTG  ATG  GTG  CTG  TAG  AAG  GAG  CAG  TGG  AGG  GGG  CGG  TCG  ACG  GCG  CCG  TTA  ATA  GTA  CTA  TAA  AAA  GAA  CAA  TGA  AGA  GGA  CGA  TCA  ACA  GCA  CCA  TTT  ATT  GTT  CTT  TAT  AAT  GAT  CAT  TGT  AGT  GGT  CGT  TCT  ACT  GCT  CCT  TTC  ATC  GTC  CTC  TAC  AAC  GAC  CAC  TGC  AGC  GGC  CGC  TCC  ACC  GCC  Q1333 (or T1335) and dT1 (or dG2) on SpNG-LWQT-TGG are similar or even shorter compared to the distance between R1333 (or V1335) and dT1 (or dG2) (Fig. 4b, d, e). For R1337 and dG3, the minimum distance on SpNG-LWQT-TGG is shorter than that on SpNG-TGG (Fig. 4f). These results showed that SpNG-LWQT can interact with not only 5′-TGG-3′ PAM but also 5′-CAT-3′ PAM, suggesting that the ability of Cas9 to interact with diverse PAM sequences is improved.

Discussion
The RNA-directed Cas9 nucleases have been adapted for multiple applications with programmability, covering genome recombination, base editing, DNA transposition and gene regulation, etc. The stringent PAM requirement is crucial for Cas9 DNA specificity, which however, also restricts the targeting scope and thus the wide applications of the Cas9 toolkits. Although the most widely used SpCas9 has been overwhelmingly engineered to rewire its PAM specificity from the cognate 5′-NGG-3′ to altered or expanded PAMs, most of them are G-containing PAMs 19,25,27 . Very recently, SpCas9 variants recognizing non-G PAMs, including the near-PAMless variant SpRY (5′-NRN-3′ > 5′-NYN-3′ PAMs), have been created via directed evolution or systematic mutagenesis 28,29 . In this research, to generate a gene-specific repressor that blocks transcription elongation at the start codon ATG, we created a PAM-expanded SpdCas9 variant SpdNG-LWQT with compatibility to the counterpart non-G 5′-CAT-3′ PAM. This endeavor would enrich the SpCas9 toolboxes with expanded targeting scopes and offer a unique variant suitable as a universal gene repressor. The ability of evolved SpCas9 variants xCas9-3.7 and SpNG to recognize expanded 5′-NG-3′ PAMs provides potential molecular scaffolds for PAM rewiring. To facilitate PAM profiling, we utilized an eGFP repression assay with the nuclease-inactive SpdCas9 or its variants, where PAM recognition relies on DNA target binding. Considering that Gln could form hydrogen bonds with adenine, we introduced R1333Q to increase potential contact with the second-position adenine of 5′-CAT-3′ PAM. Our initial efforts identified that R1333Q-harboring SpdNG (SpdNG-Q) showed partial recognition on 5′-CAT-3′ and retained recognition on 5′-TGG-3′. To further improve its recognition, scanning mutagenesis of the neighboring V1335 (R1335 in SpCas9wt) yielded SpdNG-QT (R1333Q/V1335T) showing improved recognition on both 5′-CAT-3′ and 5′-TGG-3′ PAMs. Based on that, we then introduced mutations on PAM-proximal residues that could potentially influence PAM recognition, either by increasing DNA contacts or by pushing the sugar-phosphate backbone toward PI residues. The final mutant, SpdNG-LWQT (V1135L/S1136W/R1333Q/V1335T), afforded the highest eGFP repression toward both 5′-TGG-3′ and 5′-CAT-3′ PAMs, which even outperformed the near-PAMless SpdRY variant ( Fig. 2b and Supplementary Fig. 1). Further examination of PAM profile confirmed that SpdNG-LWQT exhibited a similar PAM range as The dCas9s were expressed on pCS27, and sgRNAs were expressed on pZE12-luc. All sgRNAs were targeting coding sequence on nontemplate DNA strand. SpdCas9wt was coexpressed with sgRNAs targeting 5′-NGG-3′ PAM while SpdRY and SpdNG-LWQT were coexpressed with sgRNAs targeting 5′-CAT-3′ PAM at the ATG start codons. Cells were plated on Luria-Bertani (LB) plates containing 0.5 mM IPTG and appropriate antibiotics. Representative image from two independent repeats. c The effects of SpdCas9wt or its variants mediated CRISPRi on mevalonate production in E. coli host BW25113(F′)::MVA. The host strain with empty pCS27 and pZE12-luc was applied as a negative control (NC). All shake flasks were performed in M9 minimal medium containing 20 g/L glucose and 5 g/L yeast extract, and samples were taken at 48 h. Data indicated the mean ± standard deviation (n = 3 independent biological replicates). Source data are provided as a Source Data file.
SpdRY, and showed higher activities toward 5′-NGG-3′ and most 5′-NRT-3′ PAMs (Fig. 2c). The PAM recognition of SpdNG-LWQT was corroborated by restoring its nuclease activity toward the eGFP target both in vitro and in E. coli cells (Fig. 3). The following MD simulations predicted that SpNG-LWQT retained stable hydrogen bonding between R1337 and the third G of 5′-TGG-3′, and showed shorter minimum distances between PI residues and 5′-CAT-3′ than SpNG, which potentially explains the PAM tolerance of SpdNG-LWQT toward both 5′-TGG-3′ and 5′-CAT-3′. By targeting to start codons, SpdNG-LWQT could serve as a promising dCas9-based universal gene repressor with programmability and tunability. SpdCas9 has been demonstrated with high efficacy and wide applications in gene repression at the transcriptional level 11,49,50 . However, the near-knockout repression of target genes, especially the essential ones, may cause cell fitness defects and limitations for dCas9-based genetic circuits 36,43,51 . SpdNG-LWQT could instead afford relaxed gene repression at the start codons, and achieve tunability simply by modulating the sgRNA copies (Fig. 5c). When needed, targeting 5′-NGG-3′ PAM via SpdNG-LWQT could further increase repression efficiency to more than 95%. As showcased in mevalonate pathway, SpdNG-LWQT mediated CRISPRi of fabD in E. coli achieved 40.1% increase of mevalonate titer, in stark contrast to growth defects with SpdCas9wt mediated one (Fig. 6). Noteworthy, SpdNG-LWQT outperformed SpdRY toward 5′-NGG-3′ and 5′-CAT-3′ PAMs in both E. coli and S. cerevisiae, indicating its host flexibility (Fig. 5). In addition, SpdNG-LWQT would make the sgRNA design customizable, with the spacer complementary to the nontemplate DNA strand adjacent to start codons. All of these features render SpdNG-LWQT a promising tunable gene repressor.
In conclusion, to create a universal gene repressor targeting gene start codons, we engineered an SpdCas9 variant with an expanded PAM range including a signature 5′-CAT-3′ PAM. This exemplified the viability of harnessing PAM-flexible Cas9s for purposeful PAM alteration. The resultant SpdNG-LWQT adds to the SpdCas9-based repressors with customizable sgRNA design, tunability and host flexibility, three features that could readily render implementation into dCas9-based genetic circuits or gene control systems in a broad range of organisms. More broadly, SpdNG-LWQT could complement SpdRY to eventually permit unrestrictive access of genome targets for related genetic manipulations.
Plasmids and bacterial strains construction. DNA manipulations were conducted following the standard molecular cloning protocols 52 . The cas9 from S. pyogenes was amplified and inserted into pCS27 in between Acc65I and BamHI under control of Plpp1 promoter 37 , yielding plasmid pCS-Plpp1-SpCas9wt. Sitedirected mutagenesis to obtain SpdCas9wt (D10A/H840A) and its derived variants were performed using the method described by Chiu et al. 53 . The reporter plasmid pZE-eGFP was constructed by inserting eGFP into pZE12-luc in between Acc65I and XbaI. The synthesized sgegfp was inserted into pCS27 under the control of the P L lacO1 promoter. The P L lacO1-sgegfp cassette was then amplified from pCSsgegfp and inserted into pZE-eGFP in between SpeI and SacI, yielding pZE-eGFPsgegfp. To generate an NNN PAM library containing eGFP, the NNN nucleotides were placed adjacent to the start codon ATG of eGFP during primer design and introduced into the plasmid pZE-eGFP-sgegfp. For NNN insertions that introduce stop codons (TAA, TAG, and TGA), sgRNA spacers were instead mutated to target corresponding PAMs near the start codon of eGFP. SpCas9 or its variants were cloned into pETDuet-1 under the control of the T7 promoter. The mevalonate pathway plasmid pCS-thl-mvaS-mvaA was constructed by amplifying and inserting thl from C. difficile, mvaS from L. casei, and mvaA from R. pomeroyi into pCS27. The P L lacO1-thl-mvaS-mvaA cassette was then amplified and integrated into E. coli BW25113(F′) via λ-Red recombination 54 , resulting in E. coli BW25113(F′)::MVA. To repress competing genes for mevalonate pathway, sgRNA targeting gltA, accA and fabD were constructed into pZE12-luc, yielding pZE-sggltA, pZE-sgaccA, and pZE-sgfabD, respectively. P GAP -controlled yeGFP cassette was amplified from pZ_P-GAP-eGFP and inserted into pSP571 using SpeI and BamHI, yielding pSP571-yeGFP. The tRNA Tyr promoter controlled sgRNA module was amplified from pCAS and inserted into pSP571-yeGFP between SphI and SacI, yielding pSP571-yeGFP-sgyegfp. Yeast codon-optimized SpdCas9wt under control of P GAP promoter was amplified from pSP571 and inserted into pSP571-yeGFP-sgyegfp by HindIII and SpeI, generating pSP571-SpdCas9wt-yeGFP-sgyegfp. Yeast SpdCas9 variants and sgRNA spacers were mutated using the method described by Chiu et al. 53 . All plasmids or bacterial strains involved in this study were listed in Supplementary Table 1.
The eGFP repression assay. To determine the repression efficiency of SpCas9 or its variants, reporter plasmids (pZE-eGFP-sgegfp or pZE-NNN-eGFP-sgegfp) were cotransferred with the pCS27 plasmids carrying SpdCas9wt or its variants into E. coli BW25113(F′) cells. Reporter plasmids with the empty pCS27 were cotransferred into E. coli BW25113(F′) cells as controls. Single colonies were picked and inoculated in 3 mL LB tubes with appropriate antibiotics and IPTG. For yeast eGFP repression assay, pSP571-dCas9-yeGFP-sgyegfp or its derivative plasmids were transformed into S. cerevisiae BY4741. pSP571-yeGFP-sgyegfp was used as the control. Single yeast colonies were picked and inoculated in 3 mL histidine dropout YNB medium. After 24 h, 20 μL culture was sampled and diluted with 180 μL distilled H 2 O in a black 96-well plate. Cell optical density at 600 nm (OD 600 ) was measured and eGFP fluorescence was detected using an excitation filter of 485/ 20 nm and an emission filter of 528/20 nm with a Synergy microplate reader (BioTek, Winooski, VT). The relative eGFP expression was the ratio of the normalized eGFP fluorescence per OD600 (RFU/OD 600 ) of cells with dCas9s to those without dCas9s.
Shake flask experiments. For mevalonate production, E. coli BW25113(F′)::MVA was cotransformed with pZE12-luc derived plasmids harboring sgRNAs and pCS27 derived plasmids harboring SpdCas9wt or its variants. Empty pZE12-luc and pCS27 were used as the negative control. The shake flask experiment was conducted in a rotary shaker (New Brunswick Scientific, Edison, NJ) at 30°C with a speed of 270 rpm. Transformants of E. coli BW25113(F′)::MVA were inoculated in 3 mL LB medium and grown at 37°C for 8-10 h. The seed cultures were then transferred to 20 mL fresh M9 minimal medium containing 20 g/L glucose and 5 g/ L yeast extract in 125-mL shake flasks as 2% (v/v) inoculum and grown at 30°C for 48 h. IPTG was added with a final concentration of 0.5 mM during initial inoculation.
Protein purification and in vitro cleavage assay. For protein purification of SpCas9 and its variants, the pETDuet-1 derived plasmids were transferred into E. coli BL21 Star(DE3). The transformants were inoculated in 3 mL LB tubes at a 37°C shaker at 270 rpm. Two hundred microliters of overnight cultures were transferred into a 250-mL shaker with 50 mL fresh LB medium. When the OD 600 reached around 0.6, 0.5 mM IPTG was added and cells were transferred to a 30°C shaker for production induction and expression overnight. The cells were collected by centrifuging at 9391 × g for 10 min and then lysed using Mini Bead Beater (Biospec). Protein purification was performed using His-Spin Protein Miniprep Kit (Zymo Research, Irvine, CA) following manufacturers' instructions. The purified protein was verified by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) using 12% protein gel and the protein concentration was measured using a Pierce BCA Protein Assay Kit (Thermo Scientific, Waltham, MA) as manufacturers' instructions.
For in vitro DNA cleavage assay, sgRNA was prepared by in vitro transcription using T7 RNAP, the products were digested with DNase I and purified with Monarch ® RNA Cleanup kit following manufacturers' instructions. The sgRNA was quantified using NanoDrop 2000c spectrophotometer (Thermo Scientific, Waltham, MA). 100 nM of purified His 6 -tagged SpCas9 or its variants and 30 µM sgRNA were incubated with SacI-linearized reporter plasmid pZE-eGFP-sgegfp (15 µg, 500 nM) for in vitro cleavage. The reaction was conducted at 37°C for 30 min, in 15 μL of reaction buffer containing 20 mM HEPES-NaOH, pH 7.5,100 mM KCl, 2 mM MgCl 2 , 1 mM DTT, and 5% glycerol, and stopped by heating to 72°C for 10 min. Cleavage products were resolved by electrophoresis on 1% agarose gel and visualized by GelDoc.
HPLC analysis. Metabolites from shake flask cultivations were analyzed by Dionex Ultimate 3000 HPLC equipped with a Coregel-64H column (Transgenomic, Omaha, NE). One milliliter of sample from each shake flask culture was centrifuged at 21,130 × g for 10 min, and the supernatant was filtered through 0.22 μm membrane filter before HPLC analysis. The mobile solution was 4 mN sulfuric acid setting at a flow rate of 0.4 mL/min. The column oven temperature was set at 45°C.
MD simulation. To further compare SpCas9 or its variants accommodation with different PAM sequences, we conducted MD simulation by using GROMACS version 2018 and CHARMM36 force field 55 . First, the crystal structure of SpNG (PDB ID: 6AI6) was retrieved from RCSB's Protein Data Bank (www.rcsb.org) 56 .
The PAM sequence (5′-TGG-3′) of SpNG was virtually mutated into 5′-CAT-3′ to generate NG-CAT in software Maestro (Schrodinger, version 12.4). LWQT-NGG and LWQT-CAT were generated by mutating four amino acids (V1135L/S1136W/ R1333Q/V1335T) in a PDB viewer. Then, we used CHARMM-GUI to build the MD simulation solution box, which was a cubic box with a length of 138 Å and filled with water molecules 57,58 . After energy minimization, the structures were equilibrated using an NVT ensemble (constant number of particles, volume, and temperature) and NPT ensemble (the number of particles, pressure, and temperature). The target equilibration temperature was 300 K. Finally, MD simulations were performed for 100 ns. After the MD simulations, we calculated the number of hydrogen bonds and the minimum distance between the native/mutated amino acids and nucleotides. The protein structures were visualized via PyMOL 59 .
Statistics. No statistical method was used to predetermine sample size. All data for eGFP repression assay and shake flask experiments were presented as the mean ± standard deviation of biological triplicates (n = 3), which were also reported in the corresponding figure legends. The colonies used for data collection were randomly selected from the agar plates. For two-tailed t-test analysis of repression efficiency of SpdNG-LWQT and SpdRY, ten independent biological replicates (n = 10) were applied. Data were analyzed using Microsoft Excel and the two-tailed t-test was performed with JMP Pro 16 software. No data were excluded from the analyses and the investigators were not blinded to allocation during experiments and outcome assessment.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Information. Plasmids from Addgene (#139498, #126717, and #60847) were used in this study. Structural information from PDB (ID: 6AI6 and 4UN3) was used in this study. The raw and/or processed data underlying the bar charts and uncropped gels generated in this study are provided in the Source Data file. Source Data are provided with this paper.