The Phenotype of Many Independently Isolated +1 Frameshift Suppressor Mutants Supports a Pivotal Role of the P-Site in Reading Frame Maintenance

The main features of translation are similar in all organisms on this planet and one important feature of it is the way the ribosome maintain the reading frame. We have earlier characterized several bacterial mutants defective in tRNA maturation and found that some of them correct a +1 frameshift mutation; i.e. such mutants possess an error in reading frame maintenance. Based on the analysis of the frameshifting phenotype of such mutants we proposed a pivotal role of the ribosomal grip of the peptidyl-tRNA to maintain the correct reading frame. To test the model in an unbiased way we first isolated many (467) independent mutants able to correct a +1 frameshift mutation and thereafter tested whether or not their frameshifting phenotypes were consistent with the model. These 467+1 frameshift suppressor mutants had alterations in 16 different loci of which 15 induced a defective tRNA by hypo- or hypermodifications or altering its primary sequence. All these alterations of tRNAs induce a frameshift error in the P-site to correct a +1 frameshift mutation consistent with the proposed model. Modifications next to and 3′ of the anticodon (position 37), like 1-methylguanosine, are important for proper reading frame maintenance due to their interactions with components of the ribosomal P-site. Interestingly, two mutants had a defect in a locus (rpsI), which encodes ribosomal protein S9. The C-terminal of this protein contacts position 32–34 of the peptidyl-tRNA and is thus part of the P-site environment. The two rpsI mutants had a C-terminal truncated ribosomal protein S9 that destroys its interaction with the peptidyl-tRNA resulting in +1 shift in the reading frame. The isolation and characterization of the S9 mutants gave strong support of our model that the ribosomal grip of the peptidyl-tRNA is pivotal for the reading frame maintenance.


Introduction
Evolution of the translation apparatus involved in transfer of the genetic message stored in mRNA into proteins was an early event [1]. In the beginning of life translation made many missense errors and it was not possible to translate long mRNAs due to difficulties in maintaining the reading frame. Thus, the evolution of how translation avoids missense and reading frame maintenance errors must have occurred early and before the three domains of life emerged. Therefore, its basic mechanism is most likely similar in all organisms [1]. Many missense errors are not harmful and they occur in cells of to-day at a frequency of about 4610 24 per codon [2] although it varies widely at different sites in bacteria and in yeast [3,4]. Even if this error level is low, it would still result in that only 78% of the molecules of a 500 amino acid protein having no missense error [5]. Therefore, in a cell many proteins are not faithfully decoded and contain missense errors. Since many of these errors are in non-critical positions of proteins and influence their activity and stability in only minor ways, such an error level is apparently acceptable for the cell. However, every processivity error, such as a frameshift error, is harmful, since ribosomes shifted into the wrong frame will generally soon encounter a stop codon and terminate and thereby generate a truncated peptide.
Accordingly, the frequency of processivity error should be lower than missense errors although the estimates of such spontaneous frameshift errors have been difficult to assess [5]. Parker suggested that the frameshift errors may be 10 25 or less [2] and thus at least 10-fold less than the level of missense errors.
Although we have learnt much about the mechanism of translation, especially about the 3D structure of the decoding center and how the tRNA is located on the ribosome at the various steps of translation [6][7][8], the mechanism of how the reading frame is maintained is still not known (reviewed in [9]). In some cases there are special sites, the programmed frameshifting sites, at which frameshifting occurs at high frequency due to the presence of various stimulators [10][11][12][13]. However, at a much lower frequency, frameshift errors may also occur at sites with no apparent nearby stimulators. Such frameshift errors seem to require that the ribosome stalls due to imbalances in any of the steps in the translation elongation process. One way to avoid frameshift errors would be that each step in the elongation cycle occurs at a uniform rate. Indeed, the various aminoacylated tRNA combined with elongation factor Tu (EF-Tu) functions equivalently in translation suggesting that tRNA and its cognate amino acid have co-evolved [14]. Moreover, modified nucleosides, which are present in tRNAs in all organisms, uniform the function of the tRNA [15]. Indeed, mutants isolated as deficient in various modified nucleosides with vastly different chemical structures, present in many different positions of tRNA, and in different tRNA species, induce frameshift errors [16] [17][18][19][20][21][22][23]. Structural changes of tRNA as well as alterations in elongation factors and rRNA also induce frameshift errors [9,11]. Moreover, starvation of amino acids, presence of rare or stop codons, or over-expression of tRNA also induce such errors [24][25][26][27] [28,29]. Therefore, imbalances in the supply of aminoacylated tRNA and certain sequences in the mRNA upset the maintenance of the reading frame. Accordingly, changes in the environment may also induce errors in reading frame maintenance and indeed, cells in stationary phase have an intrinsic increased rate of frameshift error rate [30,31].
Transfer RNAs with an extra nucleotide in the anticodon loop suppress certain +1 frameshift mutations [32]. From analysis of Figure 1. The ribosomal grip of the peptidyl-tRNA is pivotal in reading frame maintenance. The figure shows three ways (A, B and C) how certain events may induce slippage by the peptidyl-tRNA and thereby a frameshift error. It is the ternary complex (aa-tRNA*EfTu*GTP) which enters the A-site and interacts with the codon but in the figure we have symbolized it with ''aa-tRNA'' to save space. A. A defective cognate tRNA (red diamond) is slow (broken arrow) entering the A-site allowing a near-cognate aa-tRNA (blue wobble nucleoside) to decode the A-site codon. After a 3 nucleotide translocation the near-cognate peptidyl-tRNA may slip into the +1 frame. B. A defective cognate aa-tRNA (red diamond) decodes efficiently the codon in the A-site. After a 3 nucleotide translocation the defective cognate peptidyl-tRNA may be prone to slip into the +1 frame. C. The defective aa-tRNA (red diamond, yellow tRNA) is slow entering the A-site mediating a pause allowing the cognate wild type peptidyl-tRNA to slip into the +1 frame. Not depicted in the figure, alterations in the ribosomal P-site environment may also induce a frameshift error if the alteration changes the ribosomal grip of the peptidyl-tRNA. The figure is adopted from [36] with permission. Indeed, as shown in this paper a truncation of ribosomal protein S9, which interacts with the peptidyl-tRNA induces an error in reading frame maintenance (See Fig. 6). Moreover, the occupancy of the E-site also improves reading frame maintenance [80,[86][87][88], perhaps by strengthening the ribosomal grip of the peptidyl-tRNA. Therefore, a defective tRNA may also increase frameshifting by altering the dissociation rate of it from the E-site. doi:10.1371/journal.pone.0060246.g001 such altered tRNAs it was inferred that the frame error induced by an inserted nucleotide in the mRNA was corrected by a tRNA having an apparent four nucleotide anticodon. Such an anticodon was suggested to read four bases, allowing a quadruplet translocation and thereby moving the ribosome into the zero frame [32]. This explanation supported the suggestion that the normal tRNA having a three nucleotide anticodon was used as yardstick in reading frame maintenance by monitoring the three nucleotide translocation required for reading frame maintenance [33]. Although the yardstick model was attractive, it was shown not to be valid for the classical frameshift mutations sufA6 and sufB2, which both have an extra G-nucleotide in the anticodon loop of proK tRNA Pro CGG o and proL tRNA Pro GGG , respectively [34]. It was suggested that these tRNAs suppress a +1 frameshift mutation by being defective in A-site entrance and thereby being outcompeted by the near-cognate proM tRNA Pro cmo5UGG . Following a normal three nucleotide translocation, this near-cognate peptidyl-tRNA slips forward one nucleotide thereby moving the A-site codon into the zero frame. In accordance with this model, many base substitutions in the body of proL tRNA Pro GGG , of which the sufB2 tRNA is a derivative, as well as in the anticodon, also suppress certain +1 frameshift mutations and the frameshift event occurs in the P-site [35]. Based on these observations and how deficiency of many different modified nucleosides imposes a +1 frameshift, an explanatory model was suggested [16,36]. In the proposed model ( Figure 1) the ribosomal grip of the peptidyl-tRNA is a key feature in reading frame maintenance [36]. There are several ways that a defective tRNA can induce frameshifting, e.g.: Fig. 1, A; the ternary complex with the defective tRNA is so slow entering the A-site that it allows a ternary complex containing a near-cognate tRNA to decode the A-site codon. After a normal three nucleotide translocation to the P-site, the peptidyl-nearcognate tRNA is prone to slip into an overlapping reading frame. Fig. 1, B; the ternary complex with a defective tRNA decodes the codon in the A-site efficiently, but once the defective tRNA has been translocated into the P-site it may slip on the mRNA. Fig. 1, C; the ternary complex containing a defective tRNA is so slow entering the A-site that it causes a pause which allows the wild type peptidyl-tRNA to slip. Note, that various physiological conditions may reduce the level of charged tRNA or the degree of modification and thereby induce a frameshift error according to this model. Basically, as soon as the kinetics of the entry to the Asite and the translocation is not in balance, the ribosome may stall allowing the peptidyl-tRNA to shift frame. This model of frameshifting is similar in many features to other models suggesting P-site slippage [24,34,35,[37][38][39][40] or models proposing that the aberrant peptidyl-tRNA induces a binding of the aa-tRNA in the A-site to the correct frame [39,41]. In addition we may also expect according to the model that changes in the P-site environment of the ribosome may also induce frameshift errors.
An unbiased test of the proposed model ( Fig. 1) would be to first select many independent mutants able to suppress various +1 frameshift mutations and then genetically and biochemically characterize such mutants to see whether or not their frameshift phenotypes are consistent with the model. Here we address this question by isolating and characterizing many (467) independently isolated +1 frameshift suppressor mutants. These 467+1 frameshift suppressor mutants had alterations in 16 different loci. According to the model we expected that many loci would, in some way, influence the structure and the activity of a tRNA. Indeed 15 of these 16 loci were in this class and the mechanism how they induce frameshifts to correct the consequences of the +1 frameshift are all consistent with the model (Figure 1). Interestingly, two mutants had a defect in a loci not altering any tRNA but changing the P-site environment. These two mutants (rpsI) had a C-terminal truncated ribosomal protein S9, which C-terminal contacts the peptidyl-tRNA and thereby is part of the ribosomal grip of the peptidyl-tRNA. The isolation and characterization of these rpsI mutants gave a strong support of our model that the ribosomal grip of the peptidyl-tRNA is pivotal for the reading frame maintenance.

Bacteria and Growth Conditions
The bacterial strains used were derivatives of Salmonella enterica serovar typhimurium and Escherichia coli ( Table 1). As rich media Luria-Bertani (LB) was used [42].The minimal solid medium was made from the basal medium [43] with 15g of agar per liter and supplemented with 0.2% glucose and required amino acids and/or vitamins [44]. TYS-agar (10g of Trypticase Peptone, 5g of yeast extract, 5g of NaCl, and 15g of agar per liter) was used as solid rich medium.

Genetic Procedures
Transduction with phage P22 HT105/1 (int-201) [45] was performed as previously described [44]. DNA sequencing was performed on chromosomal DNA or PCR products following the manual of Applied Biosystems ABI PRISM Cycle Sequencing Ready Reaction Kit Big Dye TM .

Systems Used to Isolate +1 Frameshift Suppressor Mutants
We have used different +1 frameshift mutations in the hisC or the hisD genes (Table 2) constructed as described below. The hisC3737 mutation was used earlier to obtain several of the classical frameshift suppressor mutants [46] and apparently only very small amount of the HisC enzyme is required to enable a mutant to form colonies within a day or two without histidine in the growth medium. We also introduced other frameshift mutations at the same site as in hisC3737 to widen our possibilities to obtain various frameshift suppressor mutants. Several frameshift mutations in the hisD gene were constructed, since only 1% of the HisD enzyme is enough to make a cell His + within a day [47]. Thus, only a fraction of a percent suppression of a frameshift mutation in the hisD gene is required to make enough of a functional HisD enzyme to allow a colony to appear within a few days. Monitoring suppression of a frameshift mutation in the hisD gene would be a sensitive way to detect mutations mediating very weak suppressor activity. Indeed, monitoring frameshifting as growth on a plate lacking histidine is a more sensitive way to monitor +1 frameshift suppression than monitor the suppression of the same +1 frameshift mutation in the lacZ gene which encodes bgalactosidase [36].

Constructions of Various +1 Frameshift Mutations in the hisC and hisD Genes
The frameshift mutation hisD10122 was constructed as follows: The tetracycline resistance genes tetA and tetR from Tn10dTc were first inserted into the hisD + gene in strain GT6808 (zdd-2532::cat, hisO1242) generating strain GT7127 (hisD10132::tetRA, zdd-2532::cat, hisO1242) to which plasmid pKD46 from strain GT6315 (LT2, pKD46) was introduced resulting in strain GT7128 (pKD46/ hisD10132::tetRA, zdd-2532::cat, hisO1242). This latter strain was transformed with a 60 nt DNA oligonucleotide designed to replace the tetracycline resistance cassette with the designed frameshift mutation selecting tetracycline-sensitive recombinants [48]; e.g. to construct the frameshifting site CCC-CAA-U present in hisD10122 mutant the codons 13-14 (AGC-CCU) of hisD were replaced by CCC-CAA-U. In order to produce a functional HisD protein from the mRNA of this mutant, the ribosome has to shift to the +1 reading frame before the UGA (stop) codon, which is in the zero frame and placed just after the CCC-CAA sequence. In hisD10122 a +1 shift occurs when tRNA Pro cmo5UGG is in the ribosomal P-site at the CCC codon resulting in a mutant peptide sequence (-Cys-Pro-Asn-Glu-). The hisD10110 and hisD10111 were constructed similarly by replacing the codons 13-14 (AGC-CCU) of hisD with the codons CCC-UAU-U and CCC-AAG-U, respectively, using 60 bp oligonucleotides with the mutated codons in the middle of the pNTR-SD-trmD trmD under IPTG-inducible promoter [54] pNTR-SD-mnmA mnmA under IPTG-inducible promoter [54] pNTR-SD-iscS iscS under IPTG-inducible promoter [54] doi:10.1371/journal.pone.0060246.t001 oligonucleotides. The complete frameshift windows used in this study are listed in Table 2.
The hisC3737 mutation is 31codons from the start codon and it creates a CCC-CAA-sequence upstream of the stop codon UAA in the zero frame (See Table 2). The sequence CCC-CAA of hisC3737 was replaced by CCC-AUG-and CCC-UGG in hisC10106 and hisC10109, respectively, by using the suicide plasmid pDM4 as described [49]. Two complementary 20-mers containing the desired mutation and two primers just outside the hisC gene, were used to generate by crossover PCR a fragment that was cloned into a T/A overhang vector. Then the entire hisC fragment with the mutation was cut out and ligated into the suicide vector pDM4, which contains the sacB gene, and cannot survive on plates with 5% sucrose. The construct was conjugated into S. enterica and a derivative, in which the integrated suicide vector had recombined out of the chromosome, were selected by adding 5% sucrose to the agar plates. Transconjugants surviving on plates with 5% sucrose were chosen and correct frame shift construction was confirmed by PCR and sequencing.

Mutagenesis to Obtain +1 Frameshift Suppressor Mutants
Mutagenesis of the strain GT6588 [pool of EZ-R6 gamma KAN, his-152 (deletion of the his-operon), pSMP24], which contains plasmid pSMP24 harboring dinB, was performed by inducing the expression of DinB [50,51] from plasmid pSMP24 [52]. An overnight culture of strain GT6588 was diluted 2610 6 -fold and inoculated into 500ml of LB+100 mg carbenicillin (Cb)/ml +0.08% L-arabinose. After 24 hours of growth at 376C, phage P22 was added to make phage lysates. The phage lysates were used to infect strains containing the different frame shifting sites and His + clones were collected every day for nine days and saved for further analysis.
To mutagenize with hydroxylamine a phage P22 lysate prepared from strain GT6374 [pool of random Tn10dTc insertions in strain GT6372 (containing a deletion of the hisD and hisC genes (his-644), thus avoiding recombination with wild type his-operon in transductions)] was treated with hydroxylamine as described [53] until approximately 0.1 per cent infectious phage P22 particles remained. This lysate was used to transduce strains with a +1 frameshift mutation as described in Table 2.
Nitrosoguanidine (NG) induced mutations were obtained by placing a crystal of NG on a lawn of the donor strain with either a deletion of the his-operon or with the same +1 frameshift mutation as the recipient strain; e.g. hisD10122 on an agar plate, which was incubated over night at 376C. Around the NG crystal a ring of growing bacteria emerged that contains bacteria with mutations induced by NG. Bacteria were scraped off from several fractions of the bacterial ring and resuspended in 1 ml of LB and allowed to grow for 1-2 hours before phage P22 was added to make lysates. As above the phage lysate was used to infect Hisstrain harboring a suitable +1 frameshift mutation in the his-operon and His + clones were collected as above. To avoid siblings only one mutant from each phage stock was saved. To locate the extragenic +1 suppressor mutation in each His + clone, a Tn10dTet was placed close to the mutation by using a phage P22 stock grown on a random pool of Tn10dTet in strain GT6374 as donor and the His + clone as recipient. Tet R clones were selected and Hisclones were screened. These clones have most likely the +1 frameshift mutation exchanged by the wild type allele from the donor. Following verification that the Tn10dTet was close to the +1 frameshift suppressor mutation, the location of the Tn10dTet was determined by DNA sequencing directly on purified chromosome and with primers binding in the transposon and pointing outwards. The chromosome was purified by Qiagen Tip100/G as described by the manufacturer.

Complementation Analysis
A set of mobile plasmids containing most of the ORFs from E. coli with the expression controlled by P tac /lacI q , was a kind gift from National Institute of Genetics, Mishima, Shizuoka 411-8540, Japan [54]. A plasmid, which contained the wild type copy of the structural gene for the potential mutated gene in the +1 frameshift suppressor mutant, was introduced to this mutant. The frameshift suppressor phenotype was scored as growth on plates lacking His and in several cases the modification of tRNA was established by HPLC. If the plasmid reversed the suppressor phenotype to the wild type phenotype (i. e from His + to His -), the mutant was defective in the gene harbored on the plasmid. This was verified by determining the DNA sequence of the mutated gene on the chromosome.

Analysis of tRNA Levels
Strains were grown in 10 ml LB medium at 37uC to about 4610 8 cells/ml. Following centrifugation the pellet was suspended in 1 ml of ice-cold water. Cells were collected by centrifugation and resuspended in 400 ml of 10 mM Tris-EDTA, pH 7.5. The same amount of acid phenol was added and the mixture was vigorously shaken for 10 sec, incubated for 45 min at 65uC with Table 2. Sequence of the various frameshifts sites in the his-operon used in the selection of 460 independent frameshift suppressor mutants.

Allele number
Sequence of the frameshift window The sequences of (NNN) 124 and (NNN) 31 in hisC3737, C10106 and C10107 are: agc act gaa aac act ctc agc gtc gct gac tta gcc cgt gaa aat gtc cgc aac ctg gta ccg tat cag tct gcc cgc cgt ctg ggc ggt aac ggc gat gtc tgg ctg aac gcg aat gaa ttc ccg aca gcg gtg gag ttt cag ctc acc caa caa acg ctt aac cgc tac ccg gaa tgc cag cca aag gcc gtg att gaa aac tac gcg caa tat gct ggc gta aag ccg gag cag gtg ctg gtc agc cgc ggc gcg gat gaa ggg atc gag ctg gtg atc cgc gcc ttc tgt gaa ccg ggg aaa gac gcc att ctc tac tgt ccg ccc act tac ggt atg tac agc gtc agc gcc gaa acc att ggc and: gta gag cgc cgg acg gtt ccc gcg ctt gaa aac tgg cag ctg gat cta cag ggg att tcc gac aac ctt gac ggc aca aaa gtg gtg ttc gtt tgt agc ccc caa taa [84], respectively. 2. The sequence of (NNN) 4 in hisD10110, hisD10111 and hisD10122 is: AGC-UUC-AAU-ACC. doi:10.1371/journal.pone.0060246.t002 occasional shaking before the phases were separated by centrifugation. To the water phase 400 ml chloroform was added and the mixture was shaken after which the water phase was transferred to a clean test tube. The RNA was precipitated by adding 40 ml of 3M sodium acetate, pH 5.3 and 1 ml of 100% ethanol. The precipitated RNA was washed once with 70% ethanol, centrifuged, and dissolved in 50 ml of water. About 5 mg of RNA was applied to 8% polyacrylamide gel containing 8M Urea in 89 mM Tris-borate buffer pH 8.2 containing 2 mM EDTA. The gel was transferred to a Zeta probe membrane and the RNA was UV cross-linked to the membrane. The tRNAs were detected by Northern hybridization using radioactive oligonucleotides complementary to tRNA Arg or tRNA Gln cmnm5s2UUG tRNA Gln cmnm5s2UUG .
Determination of Aminoacylation of tRNA Gln cmnm5s2UUG in vivo Cells were grown in 30 ml LB medium at 37uC to about 4610 8 cells/ml and cells were collected by centrifugation. Cells were resuspended in 1 ml of water, washed once with 1 ml of water, and finally resuspended in 500 ml of cold 0.1 M NaAc (pH 4-5) containing 10 mM EDTA. To the suspension of cells, 200 ml of glass beads and 500 ml of 25:24:1 phenol-chloroform-isoamylalcohol mixture was added and the mixture was vortexed four times for one minute with a one minute on ice between the shakings. Following centrifugation, the supernatant was transferred to a new tube and RNA was precipitated by adding 3 volumes of ethanol. The RNA was dissolved in 50 ml of 10 mM NaAc pH 4.5 containing 1 mM EDTA. Half of the sample was diluted with equal volume of 0.5M Tris HCl, pH 9.0 for 20 min at 37uC. The deacylated and non-treated samples were run on an acidic gel containing 8% polyacrylamide, 8 M urea, 0.1 M NaAc pH 5.0. RNA was transferred to Zeta probe membrane and tRNA Gln cmnm5s2UUG was detected as above.

Analysis of Modified Nucleosides in tRNA
Bacterial strains were grown over night in LB medium, diluted 100 times in 100 ml of the same medium and grown at 376C to 100 Klett units (approximately 4610 8 cells/ml). Cells were lysed and total RNA was prepared [55] and dissolved in 2 ml buffer R200 (10 mM Tris-H 3 PO 4 , pH 6.3, 15% ethanol, 200 mM KCl) and applied to a NucleobondH AX500 column (Macherey-Nagel Gmbh & Co., Düren, Germany), pre-equilibrated with the same buffer. The column was washed once with 6 ml R200 and once with 2.5 ml R650 (same composition as R200, except for 650 mM KCl instead of 200 mM KCl). Finally, tRNA was eluted with 7 ml R650, precipitated by 0.7 volumes isopropanol, washed twice with 70% and dissolved in water. tRNA was digested to nucleosides by nuclease P1 followed by treatment with bacterial alkaline phosphatase at pH 8.3 [56]. The hydrolysate was analyzed as described earlier [57] using a Supelcosil C-18 column (Supelco) with a Waters Alliance HPLC system.

Determination of the Amino Acid Sequence of the Slippage Junction
To monitor ribosomal slippage and to purify the slippage product, a previously described system was used [58,59]. It employs a fusion protein consisting of maltose-binding protein (MBP) fused to glutathione-S-transferase (GST) at its N-terminus and having six histidine residues (66His) at its carboxy terminus (GST-MBP-66His). The full-length GST-MBP-His 6 fusion proteins were expressed from plasmid pUST290 (CCC-CAA-), pUST292 (UUU-CAA), pUST310 (CCC-CAA) or pUST311 (CCC-AAG). These plasmids were constructed by cloning a DNA fragment containing the frameshift sequence into the BamHI and EcoRI sites of vector pGHM57 [58]. Ligated plasmids were transformed into strain DH5a, analyzed by sequencing the insert, and retransformed by electroporation into different S. enterica strains. The fused GST-MBP-6xHis protein expressed from these plasmids were purified essentially as described by Atkins et al [58], except the Ni-NTA purification was omitted and the MBP-His 6 part of the fusion was released by digestion with PreScission Protease (GE Healthcare), while the GST part was still bound to Glutathione-Sepharose. The MBP-His 6 peptides were separated by SDS polyacrylamide (15%) gel electrophoresis and electroblotted to a Sequi-Blot PVDF membrane (Bio-Rad). The bands corresponding to the MBP-His 6 peptides were excised from the membrane and subjected to N-terminal sequence analysis by Edman degradation. A +1 reading frame shift when tRNA Pro is in the P-site at the CCC codon in pUST310 would result in the sequence GPLGILICPNDK. Unmodified cysteine is too reactive during N-terminal sequencing and is usually only seen indirectly as the absence of an amino acid in one cycle. The confirmation of the presence of cysteine has been described earlier [36].

Nomenclatural Acts
The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ''http://zoobank.org/''. The LSID for this publication is: urn:lsid:zoobank.org:pub: XXXXXXX. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS [author to insert any additional repositories].

Systems and Procedures Used to Isolate Many Independent +1 Frameshift Suppressor Mutants
According to our frameshift model [16,36] the shift in frame occurs not by an error in the A-site as e.g. suggested by the quadruplet translocation model, but by an error in the P-site: i.e it is the peptidyl-tRNA that slips forward one nucleotide resulting in a +1 frameshift (Fig. 1). There are several ways that may induce a shift in frame (see Introduction). To extensively test the model in an unbiased way, we here present the characterization of many independently isolated mutants able to suppress various +1 frameshift mutations. These were, after the initial selection, subjected to a careful analysis of their frameshift suppressing phenotype. Such an analysis would reveal whether or not their frameshifting phenotypes were consistent with the model.
We have used different +1 frameshift mutations in the hisC and hisD genes ( Table 2) to isolate extragenic +1 frameshift suppressors. As described in the Materials and Methods, monitoring frameshifting as growth on a plate lacking histidine is a very sensitive way to isolate weak +1 frameshift suppressor mutants allowing us to isolate a wide range of different +1 frameshift suppressor mutants and thereby extensively test our frameshift model. Furthermore, we placed the frameshift mutations (both the hisC and hisD derivatives) on the chromosome (not on a plasmid!) ensuring a ''wild type balance'' of various factors involved in reading frame maintenance. We feel this is important since overexpression of mRNA, as is the case if the test gene containing the frameshift mutation is residing on a plasmid, or unbalanced tRNA pools may induce translation errors [ [60,61]discussed by Atkins and Björk [9]]. Therefore, we expected these systems to be good tools to extensively test our model.
To isolate many independent frameshift suppressor mutants, a strain having a partial deletion of the his-operon (his-152) was mutagenized by overproduction from the plasmid pSMP24 (dinB + ) of DinB, which induces random mutations of various kinds [50]. Alternatively, we mutagenized cultures of strains harboring one of the indicated his-mutations or a his deletion by nitrosoguanine (NG). Phage P22 were grown on such cultures and used to infect strains having one of the his-mutations shown in Table 2. His + transductants were selected at 37uC following several days of incubation to allow the appearance of weak suppressor mutants. Care was taken to avoid siblings by saving only one unique mutant from each phage stock. Thus, all mutants characterized (Table 3) are of independent origin even if the mutation resulted in the same nucleotide substitution. Next, we placed a Tn10dTc transposon close to each +1 frameshift suppressor mutation by crossing out the suppressor phenotype (the His + phenotype of the suppressor mutant changed to the parent phenotype Hiswhen the transposon is located close to the suppressor mutation). The location of the transposon on the chromosome was determined by DNA sequencing out from the transposon and into the nearby chromosomal region. To link the His + phenotype with the mutated gene, we transduced it back to the parental strain (His -) by selecting Tet R . Frequency of co-transduction between the transposon and the suppressor phenotype (His + ) indicated which gene mediated the +1 frameshift suppressor phenotype. If DNA sequencing showed that the suspected gene contained a mutation, we introduced a plasmid harboring a wild type copy of the mutated gene to complement the mutation and thereby further demonstrate the link between the mutated gene and the +1 frameshift suppressor ability. In this way we obtained and characterized 467 independent mutants harboring an extragenic suppressor to the different +1 frameshift mutations in the hisoperon. These 467+1 frameshift suppressor mutations present in the mutants were distributed in 16 different loci (Table 3).
2. Alterations of the Primary Sequence of tRNA Gln cmnm5s2UUG (106 mutants) or an Alteration of the Gln-tRNA Synthetase (One Mutant) Induce Low Concentration of Charged Gln-tRNA Gln cmnm5s2UUG , which causes Slow Entry into the A-site and thereby Allowing a +1 Frameshift in the P-site A glutamine codon is present in two of the six frameshift sites used ( Table 2). We therefore expected mutations influencing the activity of tRNA Gln cmnm5s2UUG and indeed this was the case. We obtained 93 mutants as extragenic suppressors in the glnU gene to the hisD10122 mutation based on their linkage to a transposon close to the structural gene (glnU) for tRNA Gln cmnm5s2UUG . Of these 11 were verified by determination of the glnU sequence. We also obtained 13 mutants as extragenic suppressors to hisC3737, which were sequenced and further analyzed ( Table 4, Fig. 2). Three of these mutants displayed a temperature sensitive phenotype (glnU1529-30, 1538), nine (glnU1526-28; 1531-33;1535-37) were cold sensitive, and one mutant (glnU1537) had a reduced growth which was similar at all three temperatures tested (Table 4). According to our model (Fig. 1), we suspected that alterations of tRNA Gln cmnm5s2UUG or a defect of Glu-tRNA synthetase should reduce the level of charged Gln-tRNA Gln cmnm5s2UUG . Transfer RNA was prepared from wild type and from the various mutants under conditions which preserve the aminoacylation of tRNA [62]. Although there was no difference in the relative level of glutaminyl-tRNA in the mutants compared to tRNA from the wild type (Fig. 2), a reduced level of tRNA Gln cmnm5s2UUG relative to tRNA Arg was observed in the mutants. We conclude that the alterations in all glnU mutants did not affect the charging of tRNA Gln cmnm5s2UUG , but made the tRNA more unstable resulting in a lower concentration of the Gln-tRNA Gln cmnm5s2UUG than in the wild type mediating +1 frameshifting consistent with our model. Two mutants, glnU1538 (G5C) and glnU1531 (G53T) were chosen to determine the frameshift event at the site of the hisC3737 mutation. For that purpose the plasmid pUST290 was constructed (Fig. 3). The frameshift sequence was inserted between the genes encoding glutathione-S-transferase (GST) and maltose-binding protein (MBP, encoded by the malE gene) with six histidine residues (66His) at the carboxy terminus in the gst-malE fusion gene. malE is in the +1 frame relative to gst, explaining why the complete GST-MBP-66His fusion protein is only synthesized when a +1 frameshift occurs. If +1 frameshifting does not occur, translation terminates at the UAA stop codon present downstream of the gst gene and only GST is produced (Fig. 3). The complete fusion protein was purified from strains containing plasmid pUST290 and glnU1531 or glnU1538 mutations using the GST and 66His affinity tags. To liberate the slippage junction fused to the MBP-66His, the frameshift product was treated with PreScission Protease. This protease cuts the protein at the specific protease site between the GST moiety and the rest of the peptide. The N-terminus of the slippage junction fused to MBP-66His was sequenced. From both mutants the first 15 amino acids of the peptide were determined as GPLGILNP-KANNSQL, where P (proline) was the last amino acid inserted in 0 frame suggesting that the frameshifting tRNA at the frameshift site CCC-CAA-UGA in hisC3737 was a wild type pro-tRNA and not the alter-edtRNA Gln cmnm5s2UUG . In order to analyze the influence the P-site tRNA exerts on the frameshifting event we also constructed a plasmid pUST292 with UUU (Phe) instead of the CCC (Pro) codon (Fig. 3). The sequence of this frameshift peptide revealed that it was the wild type tRNA Phe GAA in the P-site that caused the frameshifting event (Fig. 3). Thus, in both cases low concentration of the Gln-tRNA Gln cmnm5s2UUG caused the wild type peptidyl-tRNA (Pro or Phe) to slip forward one nucleotide and thereby moving the ribosome into the zero frame. Since both wild type Pro-and Phepeptidyl tRNA, which interact with two different codons in the Psite were induced to slip by a ribosomal pause, the identity of the last amino acid in the peptiyl-tRNA and the anticodon-codon interaction in the P-site is not critical. Thus, the frameshift phenotype of these mutants was consistent with our model (Fig. 1, alt. C).
Reduced charging capacity may also occur if the Gln-tRNA synthetase (GlnS) is defective. Indeed we obtained a GlnS mutant with an alteration (N70S) changing the environment where the CCA-end of tRNA Gln cmnm5s2UUG binds to GlnS during the glutaminylation reaction [63]. Since this is an essential gene it was not surprising that we only obtained one mutant.
3. Deficiency of the Wobble Nucleoside cmnm 5 s 2 U in tRNA Gln cmnm5s2UUG causes a +1 Frameshift in the P-site (72 Independently Isolated Mutants) The first step in the synthesis of the side chain present at position 5 of wobble nucleoside (c)mnm 5 s 2 U34, which is present in tRNAs specific for Gln, Lys and Glu, is catalyzed by a heterodimer of MnmG (earlier denoted GidA) and MnmE proteins [64] (Fig. 4). This reaction generates the cmnm 5 -side chain in the presence of glycine or nm 5 -side chain in the presence of ammonia. The MnmC1 activity of MnmC (MnmC enzyme contains two activities, C1 and C2 [65]) converts the cmnm 5 -group to an Alt. C in Fig. 1 - and tRNA Lys mnm5s2UUU chains contain only the mnm 5 -side chain. Therefore, the synthesis of the mnm 5 -side chain depends on four enzymatic activities and any alterations of these proteins encoded by the mnmE, mnmG and mnmC genes might change the extent of modification of the wobble nucleoside (c)mnm 5 s 2 U34 and thereby inducing inefficient decoding.
The first step in the synthesis of the s 2 -group, also present in mnm 5 s 2 U, is catalyzed by the cysteine desulfurase IscS, whose activity is required for the synthesis of all thiolated nucleosides in  bacteria [66,67]. The IscS delivers persulfide sulfur from Cys to TusA, which in turn transfers the sulfur to tRNA in a sulfur relay system consisting of the TusBCDE complex and finally delivers the sulfur to the tRNA by MnmA [68]. Thus, the formation of the s 2group depends on seven proteins (IscS-TusA-TusECDE-MnmA) and alterations in any of these proteins should reduce the formation of the s 2 -group of (c)mnm 5 s 2 U34. Deficiency of the s 2group should negatively influence the coding capacity of the three tRNAs having (c)mnm 5 s 2 U34 as wobble nucleoside. Of the 467 extragenic +1 frameshift suppressor mutants isolated 72 abolished, or reduced, the synthesis of (c)mnm 5 s 2 U34. This large fraction of this kind of mutants depends on the fact that 10 genes are the targets for mutations reducing the synthesis of the (c)mnm 5 s 2 U34. Several of these mutants were analyzed for the level of (c)mnm 5 s 2 U in their tRNA. Both the frameshifting phenotype and the reduced level of (c)mnm 5 s 2 U in the tRNA were returned to that of the wild type by introducing a complementing plasmid (Table 5). According to our model, reduced activity of Gln-tRNA Gln cmnm5s2UUG due to deficiency of the modified nucleoside should induce a shift in frame by the peptidyl-Pro-tRNA and indeed this is the case [16]. Moreover, the entry of tRNA Lys mnm5s2UUU , which also contains the wobble nucleoside mnm 5 s 2 U, to its cognate codon AAG in the A-site should also be reduced. We therefore determined the sequence of the frameshift peptide using plasmids pUST310 (CCC-CAA(Gln)) or pUST311(CCC-AAG(Lys)) which should monitor the slippage of peptidyl-Pro-tRNA upon slow entry of tRNA Gln cmnm5s2UUG and tRNA Lys mnm5s2UUU , respectively (Fig. 3). The amino acid sequences of the frameshift peptide for two different mnmA mutants (lacking the s 2 -group) and one mnmE mutant (lacking the mnm 5 -sidechain) were all consistent with a frameshift error occurring at the sequence CCC-CAA or CCC-AAG by peptidyl-Pro-tRNA (Table 5). Thus, we conclude that the frameshifting phenotype of these 72 mutants lacking (c)mnm 5 s 2 U34 is consistent with our proposed model (Fig. 1, alt. C).

Geranylation of (c)mnm 5 s 2 U34 Results in a Decreased
Charging of (c)mnm 5 s 2 U34 Containing tRNAs and thereby Inducing a +1 Frameshift Phenotype (One Mutant) One mutant was found to have an altered YbbB protein (G67R). The mutation is dominant and it induces the ability of the altered YbbB protein to add a geranyl-group (a C 10 H 17 -fragment abbreviated ''ge'') to the sulfur of the wobble nucleoside cmnm 5 s 2 U34 of tRNAs specific for Gln [69,70]. This generates the presence of mnm 5 ges 2 U34 in a fraction of tRNA Gln cmnm5s2UUG which in turn reduces the level of glutaminylated tRNA Gln cmnm5s2UUG . Apparently, only certain amino acid substitutions induce this activity explaining that we only found one mutant among the 467+1 frameshift mutants isolated. Interestingly, the alteration of YbbB found here (ybbB181,G67R) is an amino acid substitution at the same position of YbbB as a mutant (denoted ybbB204,G67E) isolated in 1966 and earlier characterized by us [69]. In the case of tRNA Gln cmnm5s2UUG this alteration results in a decreased level of charged tRNA Gln cmnm5s2UUG [69] nicely explaining its ability to induce +1 frameshifting. A reduced level of charged tRNA Gln cmnm5s2UUG induces a pause and allows the peptidyl-Pro-tRNA Pro cmo5UGG to shift frame. The resulting frameshift peptide is consistent with this interpretation [69]. Thus, the frameshifting phenotype of the YbbB (G67R) mutant was found to be consistent with the proposed model (Fig. 1, alt C).  [72]). Such reduced rate of A-site selection results in a frameshift in the P-site [71]. Therefore, Y38 deficient tRNA Gln cmnm5s2UUG o enters the A-site slowly resulting in a ribosomal pause and thereby allowing the tRNA Pro (most likely the tRNA Pro cmo5UGG ) in the P-site to shift frame according to our model. We characterized 10 hisT mutants and found that their frameshifting phenotype is consistent with our model (Fig. 1, alt. C).
6. Alterations in the Activity of the Three tRNA Pro :s Induce +1 Frameshifts (253 proL, 1 proM and 2 proK Mutants) The four proline codons are read by proL (tRNA Pro GGG ), proM (tRNA Pro cmo5UGG ) and proK (tRNA Pro CGG ) and their coding capacities are shown in Figure 5. The proM tRNA Pro cmo5UGG reads all four proline codons and is the only tRNA Pro that read CCA [73]. Accordingly, the proM tRNA is essential for viability and no frameshift suppressor mutant has earlier been isolated as having an altered proM tRNA. The proL tRNA Pro GGG has anticodon GGG and reads the codons CCC and CCU and the proK tRNA Pro CGG with its anticodon CGG reads only CCG. The classical dominant +1 frameshift suppressors are derivatives of proL (sufB2) and of proK (sufA6). Both these frameshift suppressors have an extra base inserted in the anticodon loop [34,74]. We isolated 253+1 frameshift suppressor mutants, which were in some way defective in the synthesis or activity of proL tRNA Pro GGG (Table 5). This large amount of proL mutants was expected, since the proL gene, is not essential for viability and deletion of this gene induces +1 frameshifting [75]. Table 6 shows that deletions, duplications, base substitutions and promoter mutations were obtained. Only three mutants (two in proK and one in proM) were obtained that affected the other two tRNA Pro :s. Since the proM is essential we did not expect many mutants defective in this gene especially as no +1 frameshift mutant has earlier been characterized (See Discussion). However, the proK is not essential [73] and we expected an equally large amount of mutations in this gene as in proL. However, this was not the case and moreover, the two independently isolated proK mutants have an insertion of a G in the anticodon similar to the classical sufA6 tRNA Pro [34]. A deletion of the proK gene does not induce +1 frameshifts of either the hisD10122 nor the hisC3737 mutations ( [76] and unpublished results). The small number of proK mutants suggests that only specific alterations of the proK gene induce frameshifting and this aspect is discussed below (See Discussion). The 256 characterized mutations altering a proline tRNA induce +1 frameshift errors. For proL mutants it has been shown earlier that the frameshift occurs in the P-site [35] and this is the case also for mutants in proK [34] and in proM [36]. Thus, the frameshifting phenotypes of all these 256 mutants defective in any of the three tRNA Pro are all consistent with the frameshifting model [ Fig. 1, Alt. A (proL and proK) and B (proK and proM)].

m 1 G37 Deficiency Induces +1 Frameshift in the P-site (19 Mutants)
The trmD gene codes for the enzyme (tRNA(m 1 G37)methyltransferase, which synthesizes m 1 G37 in all proline tRNAs, all leucine tRNAs reading the CNN codons, and the arginine tRNA reading codon CGG. We obtained 19 trmD mutants of which 6 were sequenced and tested for the level of m 1 G in their tRNAs. As expected, very low levels of m 1 G were present in their tRNA ( Table 6). Lack of m 1 G37 was the first modification deficiency shown to induce frameshift errors [18] and such deficiency induces . Schematic picture of the synthesis of (c)mnm 5 s 2 U34, mnm5ges 2 U34, and se 2 (c)mnm 5 U. (''ge'' is a geranylgroup abbreviated ''ge''; GPP is geranylpyrophosphate). The sulfur relay from Cys to the s 2 -group of the nucleoside is shown in red and the different enzymes involved in the synthesis of these thiolated derivatives are shown in green denoted as protein with their genetic symbols starting with a capital letter. A geranylgroup from GPP is transferred to cmnm 5 s 2 U of tRNA Gln cmnm5s2UUG by YbbB to generate the hypermodified ges 2 cmnm 5 U34 [69] and to mnm5s2U of Lys-and Glu-tRNA to generate ges 2 mnm 5 U [77]. YbbB is also responsible for the exchange of s 2 by Se forming mnm 5 Se 2 U if selenium phosphate is available [89]. doi:10.1371/journal.pone.0060246.g004 Table 5. Analysis of some typical mutations in genes inducing suppression of frameshift mutations.  a)Monitored as the ability to suppress the his-allele the mutant was selected to suppress. Growth of mutants on a plate lacking His following and incubation at 37uC for 4-6 days. The parental strain, which has no suppressor mutation but the indicated his-allele, did not grow on plates without His.
b)The suppression was monitored as the ability to suppress the CCC-CAA-UAG sequence placed in front of the lacZ gene (See M-M). c)Plasmids (See figure 4) used to determine the amino acid sequence at the frameshifting site and the last amino acid in the zero frame is indicated in parenthesis. d)P-site according to Qian et al 1998 [34]. e)P-site according to [35]. f)According to [36]. g)According to [69]. h)According to [16].
i)P-site since these mutants also lack the s 2 -group of mnm 5 s 2 U similar to the mnmA mutants.
j)P-site since mnmG mutants like mnmE mutants lack the mnm 5 -side chain of mnm5s2U.

Physical Alterations in the Ribosomal P-site Induce +1 Frameshifts (Two Mutants)
Alterations of the ribosomal P-site might also induce +1 frameshifts if the ribosomal grip of the peptidyl-tRNA is weakened. The C-terminal end of ribosomal protein S9 penetrates the ribosome like a tentacle and the two last amino acids make a contact with the 59 phosphate of nucleotide 32 (R130) and the 59phosphates of positions 33 and 34 (K129) of peptidyl-tRNA ( Fig. 6; [6]). Thus, ribosomal protein S9 might be an important feature of the ribosomal grip of the peptidyl-tRNA in order to maintain the reading frame. Indeed, two of the 467 independently isolated +1 frameshift mutants (rpsI2 and 3) had a 20 amino acids or a 33 amino acids, respectively, truncated C-terminal of ribosomal protein S9. Analysis of these mutants as well as two mutants isolated by direct substitution of amino acid R130 and K129 revealed that the frameshift occurred in the P-site as shown both by peptide sequencing of the frameshift product and by reduced frameshift by overexpression of the aa-tRNA predicted to read the A-site codon [36]. The unexpected isolation of these two +1 frameshift mutants having a defective ribosomal protein S9 among all our 467 independently isolated +1 frameshift suppressor mutants strongly support the fundamental importance of the ribosomal P-site in maintaining the reading frame.

Discussion
Maintaining the reading frame is a vital feature of translation in all organisms. Here we extensively test a model for how +1 frameshift errors may occur (Fig. 1). A key feature of the model is the critical role of interactions between constituents of the ribosomal P-site and the peptidyl-tRNA in maintaining the reading frame. If a ternary complex consisting of aa-tRNA, EF-Tu and GTP, is slow reading the codon in the A-site, the probability of destructions of interactions between the peptidyl-tRNA and some key components of the ribosomal P-site increases and a frameshift error may occur by the peptidyl-tRNA. We have characterized many (467) independently isolated mutants able to suppress a +1 frameshift mutation; i.e mutants with defects in maintaining the reading frame. Analyses of the frameshifting phenotype of all these mutants can be explained by our model. The mutants were defective in 16 different loci of which 15 in various ways reduced the activity of a tRNA and thereby changed either the rate with which the ternary complex enters the A-site or its interaction as part of the peptidyl-tRNA with some P-site component(s). The model also predicts that some alterations of the P-site environment might mediate a slippage of wild type peptidyl-tRNA. Indeed, we obtained two such +1 frameshifting mutants which possessed an altered ribosomal protein S9, whose Cterminal reaches into the P-site and interacts with the anticodon of the peptidyl-tRNA (Fig. 6). The fact that we obtained these S9 mutants as able to suppress +1 frameshift mutations is a strong support for the pivotal role of the P-site in maintaining the reading frame. In the proline coding box there are three tRNAs reading the four proline codons and they are encoded by proK, proL, and proM (One copy of each gene is present in Salmonella). proM tRNA has cmo 5 U34 as wobble base and decode all four proline codons [73]. A circle corresponds to a codon read by a tRNA and the line between circles denotes that the same tRNA read those codons. Note also that the proM tRNA is essential, since it is the only tRNA reading the CCA codon. The proL tRNA having G34 as wobble nucleoside reads U and C ending codons and proK tRNA, which has C34 as wobble nucleoside, should read only CCG codon. The Gln codons CAA/G are read by two tRNAs having mnm 5 s 2 U34 and C34 as their wobble nucleoside. The C34 containing tRNA reads only CAG whereas the mnm 5 s 2 U containing tRNA (glnU tRNA) decodes both CAA and CAG although less efficient CAG (Unfilled circle). Note that the latter tRNA (glnU tRNA) is essential, since it is the only tRNA reading the CAA codon. In the Lys and Glu codon boxes one tRNA having mnm 5 s 2 U as wobble nucleoside reads AAA (Lys)/GAA (Gln) and less efficient AAG (Lys)/GAG (Glu) (Unfilled circle). doi:10.1371/journal.pone.0060246.g005 At certain frameshifting sites ( Table 2) the glutamine codon CAA was present. This codon is read by tRNA Gln cmnm5s2UUG encoded by the glnU gene, which is found in only one copy in Salmonella. It is thus essential for viability. Using such sites to select for frameshift suppressor mutants, we expected to find only mutants with a reduced level, charging, or stability of this tRNA, but no deletion of its only structural gene glnU. This was indeed the case and we obtained 106 mutants of which we sequenced 23 and analyzed 12 in detail ( Fig. 2; Table 3 and 4). Using the two frameshift mutations hisC3737 and hisD10122, both having the frameshift site CCC-CAA, these kinds of mutants were frequent, suggesting that deficiency of a fully active tRNA Gln cmnm5s2UUG results in efficient frameshifting, explaining the ease with which we obtained them. The 12 mutants analyzed in detail were either temperature or cold sensitive. All of them charged the tRNA Gln cmnm5s2UUG efficiently, but the level of Gln-tRNA Gln cmnm5s2UUG was decreased due to instability of the tRNA (Fig. 2). According to our model a reduced availability of Gln-tRNA Gln cmnm5s2UUG should induce a shift in frame by tRNA Pro in the P-site. Indeed this was shown by sequencing the frameshift product and the expected amino acid was consistent with the tRNA Pro in the P-site having shifted frame (Fig. 3, plasmid pUST290). Moreover, to show that a lower level of tRNA Gln cmnm5s2UUG could also induce another peptidyl-tRNA to shift frame, we determined the frameshift product of a UUU-CAA site (Fig. 3, plasmid  pUST292). Such an analysis showed that the tRNA Phe shifted frame. Thus, we conclude that the frameshift phenotype of all these glnU mutants is consistent with our model (Fig. 1). The tRNA Gln cmnm5s2UUG is charged by Gln-tRNA synthetase (GlnRS), which is encoded by the glnS gene. Defective GlnRS results in reduced charging of tRNA Gln cmnm5s2UUG but we expected to find only a few mutations in this gene since it is essential explaining why we find only one such mutant.  The tRNA Gln cmnm5s2UUG tRNA Lys mnm5s2UUU and tRNA Glu mnm5s2UUC have as wobble nucleoside (c)mnm 5 s 2 U34, the complex synthesis of which is shown in Figure 4. Neither the s 2 -nor the (c)mnm 5group are essential for growth although deficiency of any of them severely reduces the growth. We have earlier shown that lack of this wobble modification induces +1 frameshifting [16]. Since synthesis of this modification requires as many as 10 genes, we expected many mutants defective in the synthesis of (c)mnm 5 s 2 U as was the case. The 72 independently isolated mutants were obtained using either the hisD10122/C3737 (CCC-CAA) or the hisD10111 (CCC-AAG) frameshift mutations. Deficiency of the s 2group or the mnm 5 -group induces frameshifting irrespectively of which gene involved in their synthesis is affected. This shows that it is the undermodified tRNA that is responsible for the frameshift phenotype and not lack of any of the biosynthetic proteins involved in the synthesis of these modifications. Determination of the frameshift products encoded both at a CCC-CAA and CCC-AAG site, showed that it was the P-site tRNA Pro that slipped into the +1 frame ( Fig. 3; pUST310 and pUST311; Table 6). Thus, these 72 independently isolated mutants defective in the synthesis of mnm 5 s 2 U induce a +1 frameshift by peptidyl-tRNA slippage caused by slow entry of the mnm 5 s 2 U34 deficient tRNA specific for Gln or Lys consistent with the model presented in Figure 1 (Alt. C). We also obtained one mutant with an amino acid alteration at position 67 of the YbbB protein and such alteration increases the activity of the YbbB protein to transfer a geranyl (''ge'') group to the s 2 -group of cmnm 5 s 2 U in tRNA Gln cmnm5s2UUG generating a ges 2 cmnm 5 U hypermodified tRNA Gln cmnm5s2UUG [69,70]. This hypermodification decreases the glutaminylation of the geranylated tRNA Gln cmnm5s2UUG [69,77]. As the sequence of the trans-frame encoded peptide revealed, it induces +1 frameshifting in the P-site [69]. Such alteration of the YbbB protein also mediates geranylation of the Glu-and Lys-tRNA, which contain the mnm 5 s 2 U as wobble nucleoside [77]. Unlike the hypermodified Gln-tRNA, the hypermodified tRNA Lys and tRNA Glu are aminoacylated to the same level as the respective wild type tRNA and still a +1 frameshift occurs at a +1 frameshift mutation containing lysine codons [77]. Since the frameshift peptide was not established it is not known which tRNA makes the frameshift error. However, a ternary complex having a tRNA with such a large hydrophobic compound as part of the wobble nucleoside is not likely to be efficiently accepted, if at all, in the A-site. This causes a ribosomal pause and induces the peptidyl-tRNA to make a +1 frameshift error. Thus, the +1 frameshifting induced by geranylated Lys-tRNA may be consistent with our model (Fig. 1  alt C).
The proL gene is not essential for viability and deletion of this gene also mediates +1 frameshift suppression [36,75]. Therefore, we expected to obtain many mutations in the proL gene, as we indeed did (Table 3). Since we obtained deletions, duplications, promoter mutations as well as various base substitutions in this gene, it appears that any alteration that reduces the activity of tRNA Pro GGG mediated +1 frameshifting (Table 5) consistent with the observation that a deletion of the proL gene induces +1 frameshifting. Note, that one of the proL mutations had an extra G in the anticodon and thus was identical to the classical sufB2 mutation obtained in 1970 [46,74]. Genetically it is possible to remove the cmo 5 U modification, which is only present in the proM tRNA Pro cmo5UGG among the tRNA Pro :s. Such manipulation reduces, by almost 100%, the proL mediated +1 frameshift suppression, demonstrating that it is the wild type proM tRNA Pro cmo5UGG that makes the frameshift error and not the altered proL tRNA Pro GGG [34]. Although tRNA Pro cmo5UGG is able to read the CCC codon [73], its interaction with the near-cognate codon CCC in the P-site may not be optimal thus inducing a +1 slippage upon a ribosomal pause. Since the frameshifting event occurs in the P-site [35], the many mutants obtained with defects in the proL gene is consistent with our model (Fig. 1, alt A).
The proM tRNA Pro cmo5UGG is essential for growth and no alterations of this tRNA have been described earlier. We therefore did not expect many mutants with altered proM tRNA Pro cmo5UGG . The one we obtained (G31A) disrupted the last base-pair of the anticodon stem creating a 9 member anticodon loop. Using localized mutagenesis we have isolated additional 108 mutants all with point mutations in the proM gene [36]. Among these 109 mutants, none had an extra base in the anticodon as the classical +1 frameshift derivatives of proL (sufB2) and proK (sufA6) have. The G31A mutation was found in 32% of these proM mutants most likely because it is the strongest frameshift suppressor among the 108 proM mutants characterized. Interestingly, most of the alterations in the proM tRNA Pro cmo5UGG are in close proximity to components in the ribosomal P-site further supporting the pivotal role of the P-site in reading frame maintenance [36]. Accordingly, the frameshift error induced by these altered proM tRNA Pro cmo5UGG : s occurs in the P-site as determined by amino acid sequence of the frameshift product and from overexpression of the tRNA reading the A-site codon [36]. Clearly, our extensive selection of +1 frameshift suppressor mutants identified novel alterations in the tRNA Pro family and identified, for the first time, an alteration of the proM tRNA Pro cmo5UGG that mediates a frameshift error with a mechanism consistent with our model (Fig. 1, alt. B).
In Salmonella tRNA Pro CGG is encoded by a single gene, proK, and with its C34 as wobble nucleoside it should only read CCG (Fig. 5). Although a mutant with a proK deletion is viable and grows as wild type such a strain does not suppress the hisC3737 or the hisD10122 mutations (unpublished observation). Apparently only specific changes of the proK tRNA Pro CGG induce +1 frameshift suppression consistent with a proK mutation (sufA6) being dominant [78]. We therefore did not expect many +1 frameshift suppressor mutants with alterations in this tRNA. Indeed, this was the case, since only two independently isolated mutants (proK2236 and proK2237) with an altered proK tRNA Pro CGG were obtained. Interestingly, both of them had the same alteration as the classical sufA6 mutant; i.e an extra G in the anticodon loop. The sufA6 tRNA, and therefore the two altered proK tRNA Pro CGG :s characterized here, have a normal sized anticodon bordered by m 1 G37 and U33 and the insertion of the extra G is 39 to the m 1 G37 [34]. Such altered proK tRNA induces frameshifting in the P-site according to our model [34]. The three-nucleotide size of the anticodon of these proK mutants and their induction of +1 slippage in the P-site are not consistent with the quadruplet translocation model. Such proK mutants have three tRNA Pro :s (wild type proL tRNA Pro GGG and proM tRNA Pro cmo5UGG in conjunction with the altered proK tRNA Pro CGG ) any of which may cause the +1 frameshift error and thereby suppress the +1 frameshift mutation. Of these, the proL tRNA Pro GGG is the cognate tRNA reading the CCC codon and we therefore find it unlikely that this tRNA makes the frameshift error in the Psite, since its structural interaction with the P-site environment and potential anticodon-codon interaction, is normal. Thus, either the wild type proM tRNA Pro cmo5UGG or the altered proK tRNA Pro CGG causes the frameshift error. Removing the cmo 5 U modification genetically reduces the frameshift suppression by about 50% suggesting that this part of the frameshifting is due to wild type proM tRNA Pro cmo5UGG , since it is the only tRNA Pro containing this modified nucleoside [34] (Fig. 1, alt A). As the sufA6 (a derivative of proK) mediated frameshifting occurs in the P-site [34], it is apparently a mixed population of peptidyl-Pro-tRNA Pro (wild type proM and mutant proK (sufA6, proK2236, or proK2237) tRNAs) that slips into the +1 frame in the P-site. The proM tRNA Pro cmo5UGG is less abundant in the cell compared to proL and proK tRNAs (68% of proL and 95% of proK at growth rate of 2.5 doublings [79]). In the proK mediated frameshifting, proM tRNA Pro cmo5UGG must therefore compete efficiently with the wild type proL tRNA Pro proL tRNA Pro GGG which makes a slippage in the P-site due to the lack of m 1 G although we cannot exclude that also the m 1 G deficient proM tRNA Pro cmo5UGG makes such an error. The m 1 G37 may interact in the P-site with some constituents of rRNA/rprotein(s), and lack of it in proL tRNA Pro GGG or in proM tRNA Pro cmo5UGG , may disrupt such a stabilizing interaction of the peptidyl-tRNA. Lack of the ms 2 -group or of the ms 2 io 6 -groups of the ms 2 io 6 A37, which is, like m 1 G37, present in position 37 but in tRNA Phe and tRNA Tyr , also induces +1 slippage in the P-site [16]. Crystal studies of tRNA Phe show that the ms 2 i 6 A37 is part of a network of interactions with the first base of the P-site codon, third base of the E-site codon and anchors the tRNA in the P-site with A790 and U789 of the 16S rRNA [80]. Thus, these structural data support a pivotal role of the P-site environment to maintain the reading frame and more specifically point to some fundamental interactions between position 37 of the peptidyl-tRNA and 16S rRNA. Interestingly, lack of two other modified nucleosides (t 6 A37 and yW37) present in position 37 of the tRNA also mediate increased frameshifting [21][22][23]. Apparently, the modification in position 37 of several tRNAs is important to maintain the reading frame. In summary, the fact that m 1 G deficiency induces frameshift errors in the P-site further supports the model in Figure 1 and the structural data presented for another modified nucleoside present in the same position as m 1 G, strengthen the importance of the P-site environment in reading frames maintenance and the role of modification in position 37 of tRNA in this process.
If the ribosomal P-site is important for reading frame maintenance, one would expect that a nonessential alteration in this part of the ribosome would also induce +1 frameshifting. Of course we did not expect many mutants defective in this part of the ribosome since many of such alterations would be lethal due to the important function of the P-site in translation. Still, we obtained two mutants among the 467+1 frameshift mutants both of which lacked several amino acids of the C-terminal end of ribosomal protein S9. The S9 protein is part of the common core shared by bacteria and eukaryotes and is one of the 34 conserved ribosomal proteins (15 in the small subunit and 19 in the large subunit) [81].
The C-terminal part of the S9 protein penetrates like a tentacle into the P-site of the ribosome and the two last amino acids of the protein make contacts with the 59-phosphate of nucleotide 32 (R130) and the 59-phosphates of positions 33 and 34 (K129) of peptidyl-tRNA [6] (Fig. 6). The existence of these mutants prompted us to change the two most C-terminal amino acids of S9 protein and monitor the ability to induce +1 frameshifts [36]. Indeed all these rpsI (S9) mutants induce +1 frameshifts. The isolation and characterization of these mutants defective in ribosomal protein S9 is a strong support that the frameshift event occurs in the P-site and makes the ribosomal grip of the peptidyl-tRNA one of the important features to maintain the reading frame.

Conclusion
All our results support the frameshifting model presented in Figure 1 and demonstrate the pivotal role of the ribosomal grip of the peptidyl-tRNA. Moreover, the results demonstrate an intricate competition between the three Pro-tRNAs to read the four codons in the Pro-box and highlight the importance of the modified nucleosides in positions 37 (next to and 39 of the anticodon), 34, and 38 in maintaining the reading frame.