Natural tuning of restriction endonuclease synthesis by cluster of rare arginine codons

Restriction–modification (R-M) systems are highly widespread among bacteria and archaea, and they appear to play a pivotal role in modulating horizontal gene transfer, as well as in protecting the host organism against viruses and other invasive DNA particles. Type II R-M systems specify two independent enzymes: a restriction endonuclease (REase) and protective DNA methyltransferase (MTase). If the cell is to survive, the counteracting activities as toxin and antitoxin, must be finely balanced in vivo. The molecular basis of this regulatory process remains unclear and current searches for regulatory elements in R-M modules are focused mainly at the transcription step. In this report, we show new aspects of REase control that are linked to translation. We used the EcoVIII R-M system as a model. Both, the REase and MTase genes for this R-M system contain an unusually high number of rare arginine codons (AGA and AGG) when compared to the rest of the E. coli K-12 genome. Clusters of these codons near the N-terminus of the REase greatly affect the translational efficiency. Changing these to higher frequency codons for E. coli (CGC) improves the REase synthesis, making the R-M system more potent to defend its host against bacteriophages. However, this improved efficiency in synthesis reduces host fitness due to increased autorestriction. We hypothesize that expression of the endonuclease gene can be modulated depending on the host genetic context and we propose a novel post-transcriptional mode of R–M system regulation that alleviates the potential lethal action of the restriction enzyme.

Thus far, studies of the regulatory elements involved in Type II R-M system gene expression have been focused at the stage of transcription, and to our knowledge, no report has yet revealed the mechanism(s) related to the post-transcriptional or protein synthesis level. In general, a primary notion has been that the regulatory mechanisms must be relatively host independent to facilitate the interspecies transfer of genes coding for R-M systems. Any reliance on the specific cellular host factor may limit the spread of genes. On the other hand, the successful installation of genes requires an optimal fitting to host gene expression machinery including the recognition of promoter sequences by host RNA polymerase and refactoring to host codon usage. Sequence analysis of numerous R-M systems revealed genes that are different from host genes in features such as GC content, dinucleotide frequencies and codon usage 35 . This led us to focus on the codon usage for a few R-M systems isolated from E. coli. Of the systems evaluated, EcoVIII of E. coli E1585-68, turned out to be especially rich in rare codons for E. coli [36][37][38] .
We show that an abundance of rare codons are clustered within the EcoVIII REase gene and may seriously impact its translation. We hypothesise that the expression of the endonuclease is modulated by host genetic context, and we propose a novel, post-transcriptional mode of R-M system regulation that may help alleviate the lethality of unbalanced restriction enzyme gene expression.

Material and Methods
Bacterial strains and plasmids. The E. coli K-12 strains used in this study are described below. MC1061 [araD139 Δ(ara, leu)7697, ΔlacX74, galU, galK, hsdR, strA] was used in lacZ reporter assays 39 . E. coli DH5α and MM294 were used for all other purposes including the cloning steps. E. coli Rosetta and BL21(DE3) were employed in gene overexpression 40,41 . MP060 and MP064 strains containing the yellow fluorescent protein (YFP) fused to the promoter of sulA (P sulA -yfp) was employed to test the SOS response 12 . The plasmids used are listed in Table 1. Plasmids constructed in this study were deposited in the Collection of Plasmids and Microorganisms, University of Gdansk, Poland.
Construction of ecoVIIIR::lacZ reporter and LacZ activity assay. The translational fusions of ecoV-IIIR fragment carrying arginine codons fused to the lacZ reporter gene were generated as follows. The 280 bp of ecoVIIIR gene fragment containing the natural promoter and ribosome binding site was amplified by PCR (primers: GAGCTCGAGTTAAAGCGTGGG and CAGGATCCCCACTTAATTTGAC introducing XhoI and BamHI site, respectively), cleaved with XhoI and BamHI and cloned into a pLex3B vector 43 that had been linearized with the same restriction enzymes. This yielded pLexRR (arginine codons WT-AGA AGA) in which lacZ is preceded by the ecoVIIIR promoter and 29 codons in the same reading frame. All constructs generated were confirmed by sequencing. The LacZ assays were based on hydrolysis of o-nitrophenyl-β-D-thiogalactoside (ONPG, SIGMA) using exponentially growing cells in M9-glucose minimal medium at 37 °C as described elsewhere 21,44 . Fluorescence assay. Cells were grown with shaking to exponential phase in LB medium 42 , gently pelleted, washed once with PBS buffer and then resuspended again in 500 µl of PBS buffer. Half of the sample was used to monitor the optical density (600 nm) of bacteria and the other half was used to read the YFP intensity (emission at 515 nm with an excitation at 545 nm) in a 96-well plate reader (EnSpire Multimode; Perkin Elmer). Relative fluorescence was corrected by subtracting the level of fluorescence of non-YFP bacterial cells and dividing by the optical density.
Relative restriction activity assay. Restriction activity of E. coli cells carrying EcoVIII/EcoRI R-M system and its variants was measured using the plaque formation efficiency (EOP) of phage λvir. There are 6 EcoVIII (HindIII) sites and 4 EcoRI sites in the λvir genome. The EOP of λvir was calculated as the ratio of plaques formed on E. coli MG1655 containing plasmid with no R-M system to those formed on the same strain containing a plasmid with the EcoVIII/EcoRI R-M system or their variants. Analysis of fitness effect by mixed culture competition experiments. Two comparably sized colonies of E. coli MG1655 picked from fresh transformation on LB-agar plates (with appropriate selective antibiotics) were inoculated into 5 mL of M9-glucose media. One colony carried the pRR plasmid with WT R-M system and was selected on ampicillin. The second colony was selected on chloramphenicol, and contained its variant -the pFFcm plasmid wherein two rare arginine codons at the ecoVIIIR gene were replaced by a higher frequency CGC codon. In addition, the cat gene was cloned to disrupt the bla gene of pFF to change the plasmid selection to chloramphenicol resistance. Control cultures were grown in parallel and contained cells with plasmids bearing restriction-negative and modification-positive variants (pRR0 vs. pFF0cm) or just empty vectors (pT7-3 vs. pT7-3 cm). These cultures were used to test whether the antibiotics-resistance genes alone give some advantage in

Results
The rare arginine codon clusters of the EcoVIII R-M system are features of a horizontally transferred cassette. We analyzed the DNA sequence, for the E. coli E1585-68: restriction endonuclease (REase) and DNA methyltransferase (MTase). We found a significant difference in the overall GC content of the R-M system unit (less than 38%) compared to its natural carrier -plasmid pEC156 (49.3%) and to the E. coli genome (50.8%). We also modelled the codon usage pattern for the EcoVIII R-M system genes using the codon adaptation index (CAI) 45,46 . Genes with low CAI indices reflect the optimal codon usage of their former host 35 . The CAI values for REase and MTase are 0,151 and 0,185 respectively 36 , and indicate that both belong to the group of E. coli genes that were most likely acquired by horizontal gene transfer. The results are consistent with data on codon usage deviation for other R-M systems isolated from E. coli (EcoRI and EcoRV) and from other members of the Enterobacteriaceae family (KpnI, SmaI, SinI, and PvuII) 35 .
Our analysis of the E. coli codons in the EcoVIII R-M system revealed an overrepresentation of rare arginine (AGG and AGA) (>70%) and rare isoleucine (AUA) codons ( Fig. 1A,C, Table S1). Out of the 20 arginine codons in the REase gene 18 are rare, and 13 of 15 are rare in the MTase gene. A high number of these rare codons cluster close to the C-terminus of REase, whereas the rare codons in the MTase remain in the central part of the gene (Fig. 1A). The rare isoleucine codons for MTase are closer to N-terminus, but distributed evenly across the REase. The arginine codons, however, form a cluster of two rare codons (tandem AGA) and one frequent codon starting at amino acid position 16 near the N-terminus of the REase. In case of MTase the AGG tandem is at position 167-168 (Fig. 1B). It has been shown that the presence of rare codon cluster close to the gene start may cause a translational complication 47 . Effects of rare arginine codon cluster on EcoVIII enzymes synthesis. To test the effects of rare arginine codons on EcoVIII REase and MTase synthesis in E. coli cells, we made transcriptional fusions and used φ10 promoter of T7 phage. In this system, the T7 RNA polymerase gene is under the control of an IPTG inducible lacUV5 promoter that is located on E. coli BL21(DE3) chromosome 41 . In addition, we used the pLysS plasmid to provide a T7 lysozyme, which is a natural inhibitor of T7 RNA polymerase 48 , to decrease the basal expression of the toxic REase gene. EcoVIII MTase and REase synthesis was monitored using a pulse-chase assay carried out in minimal media in the presence of [ 35 S]-methionine after IPTG induction and rifampicin addition to block E. coli RNA polymerase, but not T7 RNA polymerase (Fig. 2). www.nature.com/scientificreports www.nature.com/scientificreports/ The induction of ecoVIIIM expression led to a prominent and exclusive MTase synthesis (36 kDa) ( Fig. 2A). In contrast, EcoVIII REase synthesis was not observed under the same conditions (Fig. 2B). As a control, we used a plasmid, wherein the ecoVIIIR precedes the bla gene in the same orientation (Fig. 2C). Only β-lactamase (precursor and mature form) synthesis was detected, regardless of the presence or absence the preceding EcoVIII REase gene (Fig. 2B). We concluded that the observed translational defect may be due to the presence of the two consecutive rare arginine codons in the REase gene (AGG 16 AGG 17 ), close to the translation initiation site (within 25 first codons). The presence of the AGG/AGA codons in this location may confer a severe effect on initiation of protein synthesis [49][50][51][52] . In addition, no tandem of rare codons was detected within the 25 first codons of ecoVIIIM gene, but there are single rare isoleucin codons (AUA) present at positions 11, 22, 25. It seems these codons do not cause any problem for MTase overproduction in this case ( Fig. 2A). The rare arginine tandems have no severe effect when are present beyond the first 25 triplets 53 . We also found rare arginine codons located in the central part of the EcoVIII MTase gene (AGA 122 AGA 123 ), but still observed that efficient MTase synthesis was achieved (Figs 1 and 2A).
The presence of two consecutive rare arginine codons in ecoVIIIR affects translation efficiency. To test the effects of the tandem rare arginine codons on ecoVIIIR translation efficiency, we used a quantitative lacZ reporter assay. We generated translational fusions of WT ecoVIIIR with the promoter-less lacZ gene in the pLex-3B vector. The lacZ reporter gene in such construct was preceded by the natural promoter region, ribosome binding site and 29 N-terminal ecoVIIIR codons in the same reading frame (pLex-RR). The reporter assay shows the high LacZ activity was achieved only when the plasmid pRARE carrying the genes for tRNAs for the rare codons including tRNA AGG/AGA was present (Fig. 3).
To further investigate the effect of clustered arginine codons, we replaced the rare codons with higher frequency variants and measured whether REase synthesis is overcome after induction of ecoVIIIR gene expression from the T7 promoter. To do this, we constructed variants of the REase gene by site-directed mutagenesis, wherein each rare codon cluster (AGG, pRR) is replaced by a tandem of frequent codons (CGC; pFF/pRF/pFR; Fig. 1B). Next we performed overexpression experiments as shown before (Fig. 2). We compared the total protein  www.nature.com/scientificreports www.nature.com/scientificreports/ content from cells with and without IPTG induction to determine if R.EcoVIII (36,7 kDa) was generated for each of the variants. We observed, that replacing a single rare arginine codon from the tandem is sufficient to unleash REase overproduction (compare pRR versus pFF; pRF and pFR induction positive lanes in Fig. 4A left panel).
Moreover, if E. coli cells with pRR plasmid carrying WT EcoVIII R-M system were supplemented with tRNA Arg from the pRARE plasmid then inducible REase overproduction occurs indicating that the translational suppression is relieved (Fig. 4A right panel). To confirm these observations, we also performed a pulse-chase assay for cells carrying plasmids with the REase gene, having the rare arginine codons substituted. We used the same conditions as in the experiment shown in Fig. 2B. The experiments revealed the exclusive presence of a comparably intense band for the EcoVIII REase for each variant when at least one rare arginine codon was replaced (Fig. 4B). We also tested whether the REase mRNA stability is affected by the presence of genes for rare codons carried by pRARE plasmid. We used the approach, where the cultures of E. coli MG1655 carrying pRR plasmid with and without pRARE, were treated with the rifampicin, which inhibits the initiation of transcription. The results clearly show that no significant change in mRNA level for REase gene has been found. The determined half-life of REase mRNA was about 2 min. (Supplementary Fig. S2).
Overall, these results indicate that the inhibition of R.EcoVIII synthesis under overexpression conditions is primarily caused by the presence of tandem rare arginine codons at the N-terminus of the ecoVIIIR gene (positions 16 and 17).

Substitution of the arginine tandem increases REase restriction activity and induces the cell filamentation.
To further understand the role of the rare arginine cluster in EcoVIII R-M expression, we examined how their presence or absence affects restriction activity under steady-state levels, when gene expression signals come from a constitutive promoter and other natural genetic elements. We used λvir bacteriophage to compare plaque formation for the WT EcoVIII R-M system and the three variants with codon substitutions in the tandem arginine codons (Fig. 1B). The rare arginine codons (AGG_AGG; pRR, WT), when substituted with CGC_AGG (pFR); AGG_CGC (pRF) or CGC_CGC (pFF) conferred comparable and almost seven-fold higher restriction than the wild-type (Table 2). We also tested the activity of REase variants in vitro using crude cell extracts and substrate DNA to calculate the enzyme activity in units per amount of total cellular protein content. The results were similar to the in vivo observations (not shown) and taken together clearly indicate the biological significance of the rare arginine clusters present at the 5′ end of ecoVIIIR gene. We also supplemented the WT EcoVIII R-M system with pRARE carrying the genes for tRNA of rare codons (arginine, leucine, isoleucine, glycine, proline; Table 1) and observed significantly higher levels of restriction (38-fold) ( Table 2). In this case, we concluded, although the rare arginine codon cluster seems to be important for the initiation of translation, www.nature.com/scientificreports www.nature.com/scientificreports/ however the rare arginine codons outside the cluster, as well as other rare codons, such as isoleucine and glycine present in a high number throughout both REase and MTase genes also affect their expression (Fig. 1A, Supplementary Table S1).
We examined the cell morphology for the R-M system variants and the microscopic images revealed the presence of cell filamentation, a typical manifestation of SOS response induction, for the variants with at least one rare codon replaced with a higher frequency variant (pRF, pFR or pFF), but not for WT R-M system alone  The arrows indicate the band corresponding to the REase product (37 kDa) obtained only if at least one rare codon was replaced (pRF; pFR; pFF) or pRARE plasmid carrying the genes for rare tRNAs was present (B). The exquisite REase production (indicated by an arrow) was also confirmed by the pulse chase assay, as described in Fig. 2. The full blot is presented at Figure S1 of the Supplementary Material. www.nature.com/scientificreports www.nature.com/scientificreports/ (pRR) (Fig. 5C vs. D). However, when the plasmid carrying the WT EcoVIII R-M system was supplemented with pRARE plasmid extensive cell filamentation occurred (Fig. 5B vs. A,C). These data are in agreement with the restriction activity test, as the elevated restriction of phage DNA correlated with SOS response induction.

Lack of rare arginine codon cluster in REase gene impairs the cell fitness.
We also questioned whether or not the WT EcoVIII REase or its variant with cluster of arginine codons replaced by high frequency codons for E. coli, would show any differences in viability or fitness under long-term culture conditions. New plasmid constructs were prepared carrying the R-M system variants or vectors control, where the chloramphenicol resistance cassette was inserted into bla gene in order to restore the ampicillin sensitivity (pT7-3cm -vector, pFF0cm -R − M + and pFFcm -R + M + ) making antibiotic resistance the cell marker to monitor the numbers of competing strains over the entire growth course. For the competitions, cells carrying the appropriate plasmids (pRR WT, R + M + ampR vs. pFFcm, R + M + , cmR) were mixed 1:1 and inoculated into the same flask (with replicates). The cultures were prepared in minimal media and kept over 220 generations without the antibiotic pressure with sub-culturing every 20/21 generations (CFU were counted at these time points). For controls, parallel flasks containing the restriction negative variants (pRR0, R − M + ampR vs. pFF0cm; R − M + cmR) and compatible vectors to normalize the data (pT7-3; R − M − ampR vs. pT7-3cm; R − M − cmR) were grown under the same conditions. We observed a significant and continuous loss of cells with the frequent codon clusters (pFFcm) when cultured with the strain bearing the WT version of the rare codon cluster (pRR) indicating a loss of fitness that likely due to higher REase levels (Fig. 6).
Moreover, we tested also the presence of restriction activity for cells derived from the last counting (about 220 generation) and found that some fraction of cells from pFFcm pool, but not pRR, had changed their initial highly-restrictive phenotype into restriction-negative. We isolated these plasmids and observed some changes using fragmentation of plasmid DNA with various restriction enzymes. For the control co-cultures with restriction-negative variants, a difference in relative fitness was not detected (pRR0 vs pFF0cm). The number of the competing cells stayed comparable for the entire course of the experiment (Fig. 6).

Rare codon clusters at other R-M system units.
To determine if what we observed in the EcoVIII R-M system occurs in the other Type II systems, we surveyed the REBASE data, for "true" (not putative) Type II R-M systems, whose genes are cloned, sequenced and isolated from an E. coli host 54 . Next, we analyzed the codons for the REase and MTase genes to determine if the rare codon clusters are also formed at the N-terminus of other similar REases. There are single rare codons in several R-M systems, but out of 22 R-M systems surveyed, we found only two examples, where a cluster of rare codons was detected: EcoRI and Eco57I/Eco9272I. These codons were at positon of 9 and10, and coded for different amino-acids (arginine and leucine) in both cases: AGG_CUA (Table 3). In most cases, one to three rarest codons within 25 first codons were detected at REase gene, and comparable number at MTase gene (Table 3).
Finally, we used the well-studied EcoRI R-M system as a test to see if changing of the rare codon cluster would significantly alter the gene expression and restriction activity in a manner similar to EcoVIII R-M system. We changed the rare codons AGG_CUA (pIM-RM) into the more frequent CGC_CUG (pIMEcoRI-FF) via site-directed mutagenesis. When compared to WT, the relative restriction level for the EcoRI R-M variant system (pIMRM vs. pIMEcoRIFF) are indeed higher (phage assay, Fig. 7A). However, the two-fold increase over WT is modest compared to the change observed for the EcoVIII R-M system (8-fold, Table 2) using the same phage assay with a comparable number of restriction sites on phage genome. This may be explained by the fact the EcoRI R-M system has cluster of two codons (arginine and leucine), that are delivered by two different tRNAs, and this affects translation of REase gene to lesser extent than in EcoVIII case. In addition, we also tested if the www.nature.com/scientificreports www.nature.com/scientificreports/ elevated restriction activity of EcoRI REase variant can induce the SOS response quantitativly using a E. coli YFP reporter strain (P sulA -yfp). The results show that the altered EcoRI R-M system is comparable to the parental (WT) in inducing some level of SOS signal (Fig. 7B).

Discussion
Horizontal gene transfer is a major force driving the microbial evolution resulting in genome plasticity and mosaic pattern of their genes [55][56][57][58] . In general, the genes to be acquired need to successfully overcome the wide range of barriers to sustain the viability of the accepting host and to be retained in the genome 59,60 . The newly acquired DNA fragments with genes carrying the beneficiary function for the host have the highest chance to be integrated into the genome 61 . R-M systems certainly deliver the advantageous efficient anti-phage tools to their host 5 , but also once taken, remain utterly linked to the host as the addictive modules 62 . Losing the R-M system has  . Relative fitness of cells with a variant of EcoVIII R-M system with high frequency codon cluster at its REase is heavily impaired. Mixed cultures were prepared by adding equal number of two types of competing E. coli cells into medium without any antibiotic (Material and Methods). Each type carried a plasmid with a specific R-M system variant and its distinct antibiotic marker (chloramphenicol or ampicillin), as indicated below the diagram. One flask co-cultures were diluted every 24 generations into fresh media and CFUs of competing cells were measured. Relative competitive fitness (W) was estimated individually for each mixed culture represented on the diagram as a single symbol, calculated as W = log(CFU cm /CFU amp ) and normalized to vector control (pT7-3 vs. pT7-3cm) (Materials and Methods). Black diamonds represent four separate cocultures, where the WT R-M system in ampicillin resistant cells (pRR) competed with its R-M system variant with frequent codon cluster in REase gene (pFFcm) in chloramphenicol resistant cells. In control, parallel cocultures of cells with plasmids with restriction-negative and modification-positive variants (pRR0 vs. pFF0cm; white circles) were used.
www.nature.com/scientificreports www.nature.com/scientificreports/ the serious death consequences by mechanism of the post-segregational killing also typical for toxin-antitoxin systems 23 . EcoVIII R-M genes, carried by a natural ColE1-type plasmid, displayed the features of the horizontally transferred genes 36,37 by conferring the atypical codon usage and biased GC content 63 . It is known that codon usage reflects an adaptation of genes to the translational machinery of their host, unlike the most recently acquired, alien genes 64 . It also seems that the successful lateral gene transfer is most likely to occur if the codon usage of entering genes is compatible with the accepting genome [65][66][67] . Moreover, if accepted, the genes also tend to adapt to the host tRNA pool for their better expression to maintain their function 68 . However, this process may not occur for certain genes, as for tested here EcoVIII R-M system, which maintained the unusual overrepresentation of rare codons for E. coli. Both EcoVIII R-M genes display a strong codon bias, which seems to be favored by the host because there is a strong selective pressure to keep the rare codons and provide an adequately low amount of protein to perform its function satisfactorily, for fitness and to minimize the R-M system toxicity, as reported by others 65,69,70 . Many rare codon cluster positions are conserved within homologous coding sequences across diverse bacteria, suggesting they result from a positive selection and have a functional role 71 . In particular, we tested the role of rare arginine cluster near the start codon of REase and found its potential regulatory function in protein synthesis, R-M system maintenance and impact on the host cell fitness. We found that the cluster of rare arginine codon alone negatively affects the REase synthesis, as the delivery of tRNAs for the rare arginine codons highly increased its expression, in accord with other reports related to production of heterologous proteins 72,73 . Indeed, the tRNAs' availability seems to be the limiting factor for the protein synthesis, when the rare codon cluster is located close to translational start of the gene. We tested whether such cluster affects EcoVIII REase (positions 16 and 17) and MTase (positions 122 and 123) synthesis by pulse chase assay with a T7 promoter system. Our results confirmed the cluster position is detrimental for REase synthesis due to its proximity to translational start, but not for MTase, whose synthesis occurred efficiently. The REase translational suppression could Type II R-M system Position of rare codons within first 25 codons of gene coding for:

REase MTase
EcoVIII AGGAGG (16,17) (5) CCC (21) CCC (17) AGA (22) Eco31I GGA (4) M1 AUA (4) M2 no  www.nature.com/scientificreports www.nature.com/scientificreports/ be relieved in two ways: if more tRNA Arg was delivered or at least one rare arginine codon within the cluster was replaced by a high frequency codon. The data from the REase overproduction in T7 promoter system matched well the natural promoter context, where we measured the level of relative restriction, which is the function of two active enzymes: REase and MTase. We also show that if only one rare arginine codon was replaced for the high frequency codon, the relative restriction due to REase synthesis was significantly augmented (by 7-fold). Such exchange seems to be deleterious to cell fitness. Using direct competition fitness assay, we show that the cells carrying the REase gene variant with cluster of high frequency arginine codons were outcompeted by the cells with WT REase. Outcompeted cells with elevated relative restriction (about 7-fold; REase with CGC cluster) were eliminated from co-cultures due to possible autorestriction of its host genomes. Clearly, loss of rare arginine codon cluster in REase gene also shifted the balance in REase/MTase activities and subsequently triggered the SOS response, manifested in cell filamentation. This phenomenon is frequently observed whenever R-M system balance is lost [12][13][14]74,75 .

REase/MTase fused into one polypeptide
There are many reports that link the position of translationally non-optimal codons near the start of the gene with modulation of its gene expression at the translational stage [76][77][78][79] . The presence of consecutive rare codons may lead to translational pausing, frameshifting and/or amino-acid misincorporation in the growing polypeptide chain due to delay in appropriate tRNA delivery 47,[49][50][51][52]80,81 . The local pauses in translation may be beneficial for some protein synthesis and co-translational folding [82][83][84][85] . In particular, the NGG codons located near the initiation codon (+2 -+5 of codons) have a severe reducing effect on gene expression, but if shifted slightly downstream, the expression proceeds at normal level 86 . However, some reports also indicate that the effect of rare codons' cluster is related rather to the mRNA secondary structure and its stability, and sometimes it may even enhance gene expression [87][88][89][90] . Genome-wide analysis also supports this observation, i.e. the relaxed mRNA structure at the beginning of gene is favored, not the codon usage 91 and the non-optimal codons tend to destabilize mRNA 92 . When we made the RNA structure prediction for different arginine cluster variants using the RNAfold software, the result showed comparable structures (Supplementary Fig. S3).
The regulatory role of rare codon cluster at 5′ gene termini has been shown for bacteriophage lambda integrase 93 . In addition, even a single synonymous change for rare codon for certain genes could exert a strong phenotypic effect. Such a change at the hfq gene, a global RNA regulator of S. typhimurium, caused only a two-fold decrease of hfq expression, but showed a distinct phenotype of the strain virulence with reduced biofilm formation, motility and survival in macrophages 94 . For some genes, the presence of several rare codons in regulatory genes is required to keep their low expression and sustain the optimal operon coordination, like for FimB recombinase 95 or global regulator BldA in Streptomyces 96,97 . These schemes of regulation rely on the accessibility of tRNAs for rare codons as a limiting factor for gene expression, as it also seems true for the reported here EcoVIII restriction endonuclease, whose upregulation may be toxic for the host strain. We screened more Type II R-M systems isolated from E. coli to find whether the rare codon-dependent regulation may be a more general phenomenon. We found several examples of REase genes with only a single rare codon present near the 5′ terminus, on average. In two cases, the cluster of two rare codons for different amino-acids was detected, for the EcoRI and Eco57I R-M system. We tested the effect of a synonymous change for the EcoRI REase, and found the relative restriction changed only by two-fold, much less than for the EcoVIII R-M system (7-fold).
For mobilizable R-M genes with toxic potency (also toxin and antitoxin systems), the flexible level of expression should be avoided, thus such a wide repertoire of regulatory mechanisms exist, where REase and MTase expression is controlled at the transcription step 23 . However, in this report, we indicate a novel aspect, never raised before, which is associated with translation of the REase. Its expression seems to be dependent on the pool of host tRNAs for arginine, which may be limiting in some bacteria, such as in E. coli, or not at all in bacteria which use the AGG codon frequently. In the latter case, the installation and optimal maintenance of EcoVIII R-M www.nature.com/scientificreports www.nature.com/scientificreports/ system may be problematic. During bacterial growth the tRNAs' level also slightly changes 98 , so the R-M system is challenged to adapt to such conditions and not to kill its host. The example of the Type II EcoVIII R-M system shows the potency of the toxic gene expression, the restriction endonuclease, to be tuned to the host's genetic context, that may help to alleviate the lethality of unbalanced R-M system gene expression.