Modelling in Escherichia coli of mutations in mitoribosomal protein S12: novel mutant phenotypes of rpsL


Howard T. Jacobs, E-mail; Tel./Fax (+358) 3 215 7731.


The rpsL gene of Escherichia coli encodes the highly conserved rps12 protein of the ribosomal accuracy centre. We have used the E. coli gene to model the phenotypic effects of specific substitutions found in the mitochondrial gene for rps12. Variants created by in vitro mutagenesis were tested in two different plasmid vector systems, in both streptomycin-sensitive and streptomycin-resistant hosts. A substitution with respect to eubacterial rps12 (K87→Q), found in all metazoan and fungal mitochondrial orthologues thus far studied, is associated with low-level resistance to streptomycin and a modest (15%) drop in translational elongation rate, but without significant effects on translational accuracy. An amino-acid replacement at a highly conserved leucine residue (L56→H), associated with the phenotypes of sensitivity to mechanical vibration and hemizygous female lethality in Drosophila, creates a functionally inactive but structurally stable protein that is not assembled into ribosomes. The presence in the cell of the mutant, but not wild-type, rpsL greatly downregulates the level of a prominent polypeptide of ≈ 50 kDa. These results indicate novel structure–function relationships in rps12 genes affecting translational function, ribosome assembly and drug sensitivity, and indicate a novel regulatory pathway that may influence ribosome biogenesis.


The ribosome accuracy centre is a highly conserved component of the cellular apparatus of translation (Brakier-Gingras and Phoenix, 1984; Kurland and Ehrenberg, 1984; Kurland, 1987; Alksne et al., 1993), comprising a ribosomal RNA domain and several polypeptides of the small subunit including S12, S4 and S5 (Escherichia coli nomenclature). A number of mutations in the rpsL gene, encoding the S12 (rps12) polypeptide in E. coli, generate resistance to, and in some cases dependence on, high levels (≥ 25 μg ml−1) of streptomycin (for a review, see Brakier-Gingras and Phoenix, 1984). The ease with which such mutants can be detected has made the gene a favoured subject for mutagenesis studies (Timms et al., 1992; 1995; Timms and Bridges, 1993). Streptomycin-resistant mutants of the rpsL orthologue have also been reported in Mycobacterium, in which they are of clinical significance (Finken et al., 1993; Nair et al., 1993; Honoré and Cole, 1994; Kenney and Churchward, 1994; Meier et al., 1994; Sreevatsan et al., 1996). However, other mutant phenotypes in rpsL are relatively less well characterized.

In E. coli, mutations in ribosomal proteins S4 or S5 (Andersson et al., 1986; Ramakrishnan and White, 1992), or in elongation factor EF-Tu (Tapio and Isaksson, 1988; Tubulekas and Hughes, 1993; Kraal et al., 1995), can lead to translational inaccuracy or hyperaccuracy (Alksne et al., 1993), thus conferring aminoglycoside sensitivity or resistance in combination with appropriate alleles of rpsL. Mutations in several regions of the small subunit (SSU) rRNA, notably the 530 loop (E. coli nomenclature), can also confer streptomycin resistance or dependence on bacterial species harbouring only a single copy of rDNA, e.g. Mycobacterium tuberculosis (Finken et al., 1993; Meier et al., 1994; Honoréet al., 1995; Sreevatsan et al., 1996). Analogous mutations have been demonstrated to increase the frequency of translational errors in E. coli (O'Connor et al., 1992; Pinard et al., 1994). Such studies have defined a fidelity domain of the SSU rRNA comprising the 530 loop and the 915 region, which are in close spatial proximity.

Mutations in components of the ribosome accuracy centre have also been reported in organelles. Streptomycin resistance in Chlamydomonas, Euglena and tobacco is associated with mutations in the fidelity domain of the chloroplast SSU rRNA gene (Montandon et al., 1985; Etzold et al., 1987; Gauthier et al., 1988). Such resistance has also been found in tobacco associated with mutations of the chloroplast rpsL orthologue (Hsu et al., 1993). The fidelity domain of mitochondrial SSU rRNA resembles that of aminoglycoside-resistant strains of bacteria, but a mutation in this region of the human 12S rRNA, associated with aminoglycoside-induced or non-syndromic deafness (Prezant et al., 1993), restores the more conventional, eubacterial type structure. The mutant, but not the wild-type, RNA binds to aminoglycosides (Hamasaki and Rando, 1997), confirming the analogy with aminoglycoside susceptibility in eubacteria, and the mutation confers streptomycin sensitivity on cultured human cells that harbour it (Inoue et al., 1996).

A mutation in the yeast Nam9 gene, encoding the mitochondrial counterpart of E. coli ribosomal protein S4, relaxes mitochondrial translational fidelity and thus acts as a mtDNA nonsense suppressor (Boguta et al., 1992). In Drosophila, a mutation in the nuclear-coded, X-linked gene for mitoribosomal protein S12 (Royden et al., 1987; Shah et al., 1997) results in the phenotypes of ‘bang sensitivity’ (paralysis induced by mechanical vibration), developmental delay and temperature sensitivity (K. M. C. O'Dell and H. T. Jacobs, unpublished), as well as hemizygous female lethality.

Streptomycin resistance and dependence mutations in eubacterial and chloroplast rps12 genes map to two specific, conserved regions of the polypeptide, centred, respectively, on residues 41–43 (E. coli numbering) and residues 85–91, notably the two lysine residues K42 and K87. A number of different mutations at residue K87 in E. coli rpsL confer either streptomycin resistance, smr (e.g. K87→R), or dependence, smd (e.g. ΔK87; Timms et al., 1992). Enhanced translational accuracy associated with smr mutations in E. coli manifests phenotypically in diverse ways, including nonsense anti-suppression (Gorini, 1971) and altered expression of translationally regulated proteins (Curran, 1995; Redaschi and Bickle, 1996). Streptomycin-dependent strains exhibit hyperaccurate translation with overall growth impairment (Ruusala et al., 1984), reduced translational elongation rate in vivo (Bohman et al., 1984) and reduced ternary complex formation and GTP hydrolysis in vitro (Bilgin et al., 1992).

Analogous mutations engineered in the yeast cytosolic homologue of rps12, encoded by the genes Rps28A and Rps28B, also promote increased translational fidelity and resistance to aminoglycosides (Anthony and Liebman, 1995). By contrast, mutations in the less conserved region on the N-terminal side of the residue corresponding to E. coli K42 show relaxed fidelity. No mutations of E. coli rpsL with reduced accuracy have been reported, although some mutations detected as ancillary changes in smd strains may modulate growth phenotype (Timms et al., 1995).

In an earlier study (Shah et al., 1997), we noted that in all metazoan mitochondrial rpsL orthologues residue K87 (E. coli numbering) is substituted by glutamine. This substitution is found in a streptomycin-resistant strain of Mycobacterium tuberculosis, which also has a SSU rRNA mutation at residue 913 (E. coli numbering) that protects against the bactericidal effect but not the translational misreading induced by streptomycin (Meier et al., 1994). Because, as already noted, human mitochondrial ribosomes have an SSU rRNA substitution (np 1555) that prevents streptomycin binding (Hamasaki and Rando, 1997), the question arises whether the K87→Q substitution could play a part in the natural aminoglycoside insensitivity of metazoan mitochondria.

In the same study, we also detected a novel substitution (L56→H) at a conserved amino acid in the coding sequence of Drosophila mitoribosomal protein S12, associated with the mutant phenotype described above. No other coding region substitutions were evident in the mutant and the mutant allele was expressed at approximately the same level as in wild-type flies, indicating the L56→H substitution as the probable cause of the phenotype. No mutations at this residue have been reported in bacterial rpsL homologues, therefore the molecular effects of the mutation on translational function are unknown.

In an attempt to answer both questions, we have used E. coli as a model by constructing variants of rpsL with the L56→H and K87→Q substitutions, both alone and in combination, and phenotypic testing of these in naturally smr and sms hosts. The results confirm K87→Q as contributing directly to streptomycin insensitivity with a concomitant, modest drop in translation elongation rate, and reveal a defect in incorporation of rpsL into ribosomes associated with the L56→H mutation. The L56→H mutant protein is stably accumulated in cells but provokes a downregulation of an abundantly expressed protein, suggesting a novel regulatory pathway.


Because wild-type rpsL expressed on a high copy number plasmid is moderately toxic to the cell, leading to instability of the dominant sms phenotype, we used two constructs giving a lower level or regulatable expression of the gene. In one case, we used the plasmid pABS12, containing the rpsL gene under the control of its own promoter (see Fig. 1), in the moderate copy number vector pBR322. In the other, we inserted the rpsL coding sequence into the multiple cloning site of pBluescript II SK+, via PCR using a chimeric oligonucleotide, and oriented it such that it was expressed under the control of the lac promoter located in the vector (Fig. 1). In the latter construct, the rpsL coding sequence was positioned such that the stop codon of the residual lacZ peptide overlapped the start codon of rpsL, which we expected would translationally downregulate expression even under inducing conditions.

Figure 1.

. rpsL constructs and mutagenesis strategy. A. Schematic map of plasmids pABS12 (derived from pBR322) and pBluescript–rpsL (derived from pBluescript II SK+). The rpsL coding sequence is shown as a shaded box. In pABS12, a 525 bp fragment of the str operon, including the promoter (Pstr) and complete coding sequence of rpsL (Post and Nomura, 1980), is inserted at the EcoRI site (R) of pBR322. Also shown is the nearby site for HindIII (H). PCR in vitro mutagenesis of pABS12 to create the L56→H substitution used the strategy illustrated, in which the gene was amplified in two fragments using complementary mutagenic oligonucleotides (L56→H-2 and L56→H-1) each in combination with terminal chimeric primers (rpsL3 and rpsL4 respectively) to add appropriate restriction sites for subsequent cloning. The complete, mutagenized gene was regenerated by a second round of PCR with the two terminal primers using as template the two products, which anneal to one another across the mutation site. The complete gene was recloned into pBR322 cut with EcoRI and HindIII. pBluescript–rpsL was created using pABS12 as template by PCR cloning of a 421 bp segment of DNA containing the entire coding sequence of rpsL, plus 24 bp of 5′ and 22 bp of 3′ flanking DNA, synthesized using chimeric primers rpsL5 and rpsL4, which supply terminal sites for EcoRI and HindIII. The fragment was cloned into the polylinker of pBluescript II SK+ using these sites, and the sequence verified as wild-type by sequencing on both strands. B. The construction places rpsL under the control of the lac promoter of the vector (Plac), which gives rise to a transcript containing a residual open reading frame of 46 amino acids derived from the lacZ translation start, shown using the one-letter code. The stop codon (overlined) of the residual lacZ peptide overlaps the start codon (underlined) of rpsL, expected to confer a degree of translational downregulation. An in-frame stop codon upstream of the rpsL open reading frame, double-underlined, prevents synthesis of any rpsL-related polypeptide from upstream translational start sites. PCR in vitro mutagenesis L56→H was accomplished in the pBluescript–rpsL construct using the same strategy as for pABS12, except that rpsL5 was used as 5′ terminal primer. Mutagenesis K87→Q used a similar strategy, as illustrated for pBluescript–rpsL. Double mutants were created in each construct, via a further round of PCR mutagenesis. All derivatives were sequenced on both strands to exclude unprogrammed mutations, and five independent isolates of each were subjected to phenotypic analysis. C. Amino-acid sequence of the rpsL gene product, numbered with the cleaved N-terminal methionine as residue 0. The residues that were the object of the mutagenesis strategy are arrowed. The conserved motif PKKPNSA, found in all eubacterial and organellar homologues of the gene, is underlined.

Both pABS12 and five independently isolated rpsL constructs in pBluescript II SK+, each with the wild-type rpsL coding sequence, were stable, both phenotypically and at the sequence level, even when grown in IPTG-containing medium in the DS941 smr host, which carries the substitution K42→T in the chromosomal copy of rpsL. Like pABS12, all of the pBluescript rpsL clones showed the dominant sms phenotype characteristic of the wild-type gene, but only if IPTG were present. Prolonged (> 24 h) culture in IPTG- and streptomycin-containing medium did, however, lead to growth of selected cells from which we isolated plasmids containing nonsense or frameshift mutations in rpsL.

Substitutions L56→H and K87→Q were introduced singly or in combination by PCR in vitro mutagenesis using the scheme outlined in Fig. 1. Five independent isolates of each variant in each vector were sequenced on both strands, and were tested to ensure that they had the same growth characteristics on plates as in liquid culture. Growth data and growth curves for representative clones maintained in DS941 are shown in Fig. 2 and Tables 1 and 2. Plasmid DNA from representative clones was transformed into the streptomycin-sensitive host XL-1 blue and phenotypes were retested (see Fig. 3 and Tables 1 and 2[link]).

Figure 2.

. Growth curves of rpsL variants in host DS941 (smr), in media containing various concentrations of streptomycin. Curves are shown for the untransformed host, and for DS941 transformed with wild type (wt) and with K87→Q-substituted rpsL in pBluescript II SK+. Overnight cultures were grown without IPTG or streptomycin, but in the presence of ampicillin for the plasmid-bearing strains, and were inoculated at 1:100 dilution into medium containing IPTG and antibiotics as appropriate. The L56→H and L56→H/K87→Q mutants (not shown) grew with characteristics indistinguishable from the host strain. Cells containing the plasmid vector alone, when grown in ampicillin plus streptomycin, reached a lower final OD600, but otherwise grew at a similar rate.

Table 1. . Phenotypes of rpsL mutants propagated in pBR322 on host DS941. Growth characteristics on plates with or without streptomycin. All except host cells were grown on plates containing 50 μg ml−1 ampicillin. +++, Strong growth; −, no growth.Thumbnail image of
Table 2. . Phenotypes of rpsL mutants propagated in pBluescript II SK+. a. Growth characteristics in each of two hosts, on plates with or without streptomycin at 25 μg ml−1 and IPTG.All except host cells were grown on plates also containing 50 μg ml−1 ampicillin. +++, Strong growth; ++ and +, proportionately weaker growth; −, no growth.Thumbnail image of
Figure 3.

. Growth curves of rpsL variants in host XL-1 blue (sms), in media containing various concentrations of streptomycin. Curves are shown for cells containing wild type (wt) and K87→Q-substituted rpsL in pBluescript II SK+. Cultures were grown as described in the legend to Fig. 2. Host cells, or cells transformed with the vector alone, with the L56→H mutant or the L56→H/K87→Q double mutant were as wild type, i.e. showed no growth at any concentration of streptomycin tested.

The results may be summarized as follows. The variants carrying the L56→H substitution, whether alone or in combination with K87→Q and regardless of the vector used or the presence of IPTG to induce expression from the lac promoter, behaved under all conditions exactly as host cells carrying the vector alone. In other words, DS941 cells remained smr, XL-1 blue cells remained sms and growth rates were unaffected, as if the mutant rps12 protein were simply not present. Furthermore, in contrast to the instability of the wild-type gene, five out of five clones of the L56→H mutant under the control of the str promoter, recloned into the high copy vector pBluescript II SK+, were stably maintained in host DS941, regardless of the presence of streptomycin.

The K87→Q mutation alone produced a more complex phenotype. In the pBR322 construct, transfected DS941 cells retained streptomycin resistance. DS941 cells carrying the K87→Q variant in pBluescript also showed no growth impairment on streptomycin in the absence of induction, and IPTG induction in streptomycin-free medium caused no reduction in growth rate. IPTG induction in streptomycin-containing medium, however, gave a growth phenotype intermediate between that of the host alone and that of cells transformed by the wild-type construct (which did not grow on streptomycin plates, and showed only very slow growth in liquid culture at 1 μg ml−1 streptomycin with no growth at all at higher concentrations of the drug). By contrast, upon IPTG induction, cells harbouring the K87→Q pBluescript construct grew relatively well at low concentrations of streptomycin, and showed some growth even at 25 μg ml−1 both in liquid culture and on plates.

In the streptomycin-sensitive host XL-1 blue, the K87→Q pBluescript construct conferred effective resistance to the drug at 1 μg ml−1 and enabled some growth at 2 μg ml−1, whereas cells carrying the wild-type rpsL construct were unable to grow. These results indicate that the K87→Q substitution by itself contributes low-level resistance to streptomycin. However, the different resistance levels on the two hosts indicate incomplete dominance, implying that the structurally altered protein may be less stable or less able to be incorporated into ribosomes.

To investigate these issues further, SDS–PAGE was used to analyse the levels of rps12 protein accumulated under inducing conditions in cells expressing these rpsL variants (Fig. 4). All four variants (wild type, K87→Q, L56→H and the double mutant) accumulated to high levels, with that of the mutant proteins consistently higher than that of the wild type. The hypothesis that the K87→Q and L56→H substitutions destabilize the protein is therefore clearly incorrect. The fact that the L56→H variant is stably accumulated yet has no apparent effects on translation in two different hosts of opposite aminoglycoside sensitivity implies that the mutant protein cannot be incorporated into ribosomes.

Figure 4.

. SDS–PAGE of bacterial cell proteins from DS941 transformed with pBluescript-derived plasmids containing wild type and mutant forms of rpsL as shown, alongside a host cell control. Cultures were grown in medium containing IPTG and/or antibiotics as indicated, after inoculation at 1:100 dilution from overnight cultures grown without streptomycin or IPTG. A. IPTG induction of rpsL expression, 5 h. B. Overnight growth in IPTG in absence or presence of 25 μg ml−1 streptomycin. At this streptomycin concentration, cells harbouring plasmids encoding either wild type or K87→Q-mutant rpsL are unable to grow. The cells that grow up overnight from such cultures have lost all detectable expression of the plasmid-encoded rps12 protein. The positions of molecular weight markers are indicated, as are the presumed polypeptides corresponding with rps12, EF-Tu, β-lactamase and the two proteins (X and Y) downregulated in response, respectively, to mutant and to high levels of all forms of rps12.

Streptomycin had no effect on the expression of the mutant rpsL polypeptides although prolonged culture in the presence of IPTG and streptomycin of DS941 cells transformed with the ‘wild-type’rpsL gene completely failed to show induced rps12 expression, consistent with the recovery of only nonsense and frameshift mutants of rpsL from such cultures (Fig. 4).

Careful examination of steady-state proteins in the SDS gels shown in Fig. 4 reveals that a prominently expressed, unidentified polypeptide (‘X’), migrating with an apparent molecular weight of ≈ 50 kDa is markedly downregulated in cells transformed with the mutant, but not the wild-type rpsL constructs. This downregulation occurs even without IPTG induction of the plasmid-borne rpsL gene, probably due to leakiness of the repression. The affected polypeptide is clearly not EF-Tu (43 kDa), which can be easily identified on gels as the most abundant polypeptide in exponentially growing cells. Another polypeptide (‘Y’) of molecular weight ≈46 kDa appeared to be downregulated by IPTG-induced expression of all rps12 variants tested. The expression of EF-Tu appeared to be unresponsive to rps12 overexpression.

Structural predictions using the GCG program peptidestructure (Program Manual for the Wisconsin Package, 1994) suggested little change as a result of the L56→H substitution. Residue L56 is predicted to be at the edge of a β-stranded segment of the polypeptide, in both E. coli and Drosophila. Substitution by histidine leaves this structure largely intact.

Next, we examined the properties of additional ‘unprogrammed’ mutants of rpsL recovered from PCR in vitro mutagenesis (Table 3). Most of these had additional amino-acid replacements as well as one or both of the programmed mutations at residues 56 and 87, none of which had any significant further effects on phenotype. One missense mutant (A47→V), falling within the conserved motif PKKPNSA that contains the lysine residue most frequently altered in streptomycin-resistant mutants, showed slow growth in DS941 (but not XL-1 blue) in the presence or absence of streptomycin. The mutation may create a structurally altered but inactive protein mildly toxic to E. coli. All of the nonsense and frameshift mutants examined had no effects on cell phenotype in either host, and were thus deemed to be rpsL nulls. One mutant carried a different substitution at residue 56 (L56→P) as well as the K87→Q replacement, and this also appeared to create a functionally inactive protein.

Table 3. . Phenotypes of ‘unprogrammed’ mutants of rpsL. R, resistant to streptomycin at 25 μg ml−1; S, sensitive to streptomycin at 25 μg ml−1; NT, not tested; *, stop codon.Thumbnail image of

We extended this analysis by carrying out a random mutagenesis of residue L56 in the pBluescript II SK+ construction (see Fig. 1) to determine which other substitutions could generate a similar phenotype to L56→H. Of 63 insert-containing colonies analysed, 25 grew very poorly or not at all in the DS941 host on plates containing IPTG plus streptomycin at 25 μg ml−1, although all grew well on streptomycin plates lacking IPTG. All 63 were sequenced, giving the results shown in Table 4. Only the 28 clones that showed a mutation uniquely at residue 56 are included; the others being either wild type or mixed sequence, or else containing other unprogrammed mutations including frameshifts. Three relatively conservative substitutions (leucine to isoleucine, valine or threonine) were tolerated at residue 56 with little phenotypic effect. A set of other, less conservative substitutions (serine, alanine or phenylalanine) enabled very slow growth on streptomycin, i.e. showed partial dominance over the host allele. Most substitutions gave a similar phenotype to that of L56→H, i.e. loss of dominance. Three substitutions (to glutamate, asparagine and tryptophan) were not picked up, and the frequencies of others were highly non-random. All variants studied showed accumulation of detectable rps12 protein, although some, notably those with highly polar amino acids at position 56, did not accumulate the polypeptide to very high levels under inducing conditions, indicating a possible impairment to stability. However, many were recovered only as single clones, and although they were sequenced on both strands firm conclusions should probably not be drawn. Most others (e.g. alanine, phenylalanine, cysteine or tyrosine) clearly accumulated the mutant polypeptide to high levels, indicating that, as for L56→H, a failure of assembly rather than a destabilization of the protein was responsible for the partial or complete loss of dominance. A nonsense mutation at residue 56 caused loss of dominance and of detectable polypeptide.

Table 4. . Phenotypes of rpsL mutants at residue L56. a. Explanation of phenotype symbols: −, no detectable growth on streptomycin plates (i.e. dominant, as wild-type rpsL); ±, very slow growth on streptomycin plates, normal growth in absence of streptomycin/IPTG; +, good growth on streptomycin plates.b. These clones, like wild-type, were able to grow on streptomycin plates lacking IPTG, although slightly less well than the others.c. Scored either as (+) detectable rps12 protein or (−) no detectable protein. Various amounts of protein were detected (see text).d. Other clones carrying these substitutions, plus additional unprogrammed mutations, were obtained. All of these (L56→H/L73→M, L56→D/R13→H and L56→R/R49→H) grew on streptomycin.Not obtained: N, E and W. An unprogrammed mutant in the rpsL Shine–Dalgarno sequence also grew well on streptomycin.*, Stop codon.Thumbnail image of

Finally, we examined the question of whether the variants tested (both K87→Q and L56→H) bring about a detectable change in the translational properties of ribosomes in vivo. Two such properties were examined: translational fidelity and ribosome elongation rate. Translational fidelity was studied using a nonsense suppression assay, in which the rate of synthesis of β-galactosidase is measured in an E. coli strain containing a nonsense mutation in the lacZ coding sequence compared with an otherwise isogenic strain that lacks this mutation. Ribosomal elongation rate was studied using a different E. coli strain, carrying an inducible β-galactosidase gene, in which the time of first appearance of product was extrapolated, based on the accumulation of activity after induction. To avoid complications arising from the simultaneous induction of rpsL under the lac promoter, we used only the variants created in pBR322 constructs, which were each transformed into the relevant host strains. In every case, multiple colonies were picked from each plate, and assayed for these translational properties, and the plasmids resequenced to check that the relevant rpsL mutation was still present, and that no other changes had occurred in the coding sequence or promoter region. Translational data were discarded for any clone in which such changes were found.

In the host strain for measuring ribosomal elongation rates we noticed a consistent difference in growth rate between the wild type and K87→Q-substituted rpsL-containing clones. The latter grew much more slowly and any cultures which appeared to have recovered to the wild-type growth rate invariably contained a secondary mutation, typically an IS1 or IS5 insertion into the promoter region. From bona fide clones, we extrapolated ribosome elongation rates for the wild-type and for the K87→Q and L56→H variants grown continuously in 100 μg ml−1 ampicillin, which revealed that the K87→Q substitution results in a modest, but significant, increase in ribosome transit time compared with wild-type rpsL (Fig. 5 and Table 5[link]). The result of the elongation rate assay for cells containing the L56→H variant (Table 5) was not significantly different from wild type, consistent with our earlier inference that this variant is not incorporated into ribosomes and hence should not alter translational phenotype. Nonsense suppression assays for the same three variants revealed only very small differences (within the range of experimental error), indicating that the K87→Q and L56→H substitutions do not influence translational fidelity.

Table 5. . Effects of rpsL variants on translational parameters rpsL variant.Thumbnail image of
Figure 5.

. Translational elongation rate assays of cells expressing wild-type and K87→Q-substituted rpsL. Induction kinetics of wt and K87→Q mutant cells after addition of IPTG. Time required for synthesis of the first β-galactosidase monomers can be estimated from the intercepts of the lines with x-axis (wt = 76.6 ± 3.3 s, K87→Q = 88.9 ± 4.8 s). Values are means of several experiments (nWT = 18, nK87→Q = 12). Results are plotted as mean ± SD.


The analyses reported here demonstrate that the K87→Q substitution in E. coli rps12, such as found in a streptomycin-resistant strain of M. tuberculosis (Meier et al., 1994) as well as in metazoan mitochondria (Shah et al., 1997), is associated with low-level resistance to streptomycin, even in the absence of alterations in the fidelity domain of the SSU rRNA that accompany it in both contexts (Meier et al., 1994; Hamasaki and Rando, 1997). This result is consistent with the idea that this substitution contributes to the natural aminoglycoside resistance of the mitochondrial translational apparatus in fungi and metazoans. Some land plants, for example the liverwort Marchantia polymorpha (Oda et al., 1992), also have glutamine at this position. A different substitution K87→R, found in streptomycin-resistant strains of a number of different eubacteria, has been reported in the mitochondrial rps12 gene of Paramecium (Pritchard et al., 1990). The inference from our data that the K87→Q variant is only rather poorly assembled into ribosomes in E. coli, taken together with the observation that this substitution occurs with a SSU rRNA substitution in both mitochondria and streptomycin-resistant mycobacteria, suggests co-evolution of the protein and RNA components of the ribosomal accuracy centre. It will be interesting to measure the effects on drug resistance and ribosome assembly of combining the K87→Q variant with 16S rRNA mutants in E. coli. The ancillary mutations picked up earlier in strains carrying the smd K87→R mutation (Timms et al., 1992; 1995; Timms and Bridges, 1993) may also compensate for such an assembly defect, and perhaps warrant further investigation.

Measurements of the translational properties of cells expressing the K87→Q variant indicated a drop of ≈ 15% in translational elongation rate, but no significant change in accuracy. This suggests that the mitochondrial ribosome has evolved aminoglycoside resistance, conferred by this substitution, without any significant sacrifice in translational efficiency. Such a modest drop in elongation rate is unlikely to be rate limiting for any of the 13 mtDNA-encoded polypeptides in, for example, metazoans. Even if it were, the mitochondrial translation system appears to be endowed with considerable excess capacity, given that loss of up to 70% of mitochondrial protein synthesis can be tolerated with no impairment to respiration or oxidative phosphorylation (Spelbrink et al., 1994). This is perhaps not surprising, given that the mitochondrial translation products are encoded by a multicopy genome (>1000 copies per cell in mammals) yet need to be co-assembled with the products of numerous single-copy nuclear genes.

The L56→H mutation creates an apparently non-functional but, nevertheless, stable polypeptide. Because the presence in the cell of large amounts of the mutant protein has no effect on translational function in either the sms or smr host, it may be inferred that the mutant protein is assembled very poorly, if at all, into ribosomes. The predicted effects of the mutation on secondary structure are minimal, and almost all other non-conservative substitutions at this residue lead to a similar phenotype. The likeliest explanation is, therefore, that the conserved leucine residue 56 is involved in contacts with other components of the ribosome, or with an assembly factor or chaperone involved in ribosome biogenesis. Such a partner is presumably a protein because only the extreme N-terminal region of the rps12 polypeptide of E. coli can be cross-linked to rRNA. Non-functionality, or greatly reduced functionality, of the L56→H-substituted rps12 protein in E. coli is consistent with the genetic properties of the analogous substitution in the Drosophila mitochondrial orthologue (tko). The tko(25t) allele that carries this amino-acid replacement is recessive and, as expected for a severe hypomorphic allele of an essential X-linked gene, is lethal in the female hemizygote.

A novel finding in this work is the observation that the steady-state level of at least two abundantly expressed proteins is downregulated in response to the presence in the cell of plasmid-encoded rps12. In one case, the effect is specific for rpsL mutants encoding polypeptides that are assembled poorly into ribosomes. The effects are visible even under non-inducing conditions, probably due to leakiness of the promoter although other explanations are possible. The effect is completely absent in the case of wild-type rpsL, even when expressed at a high level. Because the amount of wild-type rps12 produced under inducing conditions obviously exceeds the requirements of ribosome biogenesis, the mutant-specific phenotype (downregulation of protein ‘X’; see Fig. 4) cannot be due to the presence of unassembled rps12 as such. Instead, it must be due, either directly or indirectly, to the mutant rps12 that is incompetent to be assembled, or at least to the gene that encodes it. In contrast, the second protein downregulated by plasmid-encoded rps12 (protein ‘Y’; see Fig. 4) responds to overexpression of rps12 per se, rather than to the particular protein variant.

Translational feedback regulation of ribosomal protein operons by one of the proteins they encode is well established (for a review, see Nomura et al., 1982; Lindahl and Zengel, 1986). The str operon is regulated by ribosomal protein S7 (the rpsG gene product), acting as the translational repressor of itself, rps12, EF-G and, to a lesser extent, EF-Tu (Dean et al., 1981; Saito et al., 1994). Unincorporated S7 binds to an intergenic site between rpsL and rpsG (Saito and Nomura, 1994), probably destabilizing the entire mRNA. However, an involvement of rps12 in translational regulation has not been reported previously. Moreover, none of the proteins downregulated in response to excess rps12 is of an appropriate size to be one of those encoded by the str operon. Therefore, this represents a novel example of trans-regulation by a ribosomal protein, although its targets are unidentified and we cannot say yet whether it is direct or indirect or whether it is affected at the RNA or protein level. The process of ribosome assembly may, thus, be regulated by a feedback mechanism that monitors the availability of key components in the pathway, of which rps12 is implied to be one.

These findings indicate novel structure–function relationships in the rpsL gene, regarding translational function, ribosome assembly and drug sensitivity. They furthermore reveal a novel regulatory pathway relating rps12 and other cellular proteins, which may influence ribosome biogenesis. Clearly, this phenomenon warrants further experimental investigation, and raises the question of whether any similar kind of feedback regulation operates in mitochondria, whose ribosomal proteins are encoded, synthesized and assembled for function in three separate cellular compartments.

Experimental procedures

Hosts and plasmid constructs

Plasmid pABS12, containing a 525 bp segment of the str operon (Post and Nomura, 1980) cloned into the EcoRI site of pBR322 using a synthetic linker (see Fig. 1), was maintained in the streptomycin-resistant E. coli host DS941 (thr1 leu6 hisG4 thi1 ara14 proA2 argE3 galK2 sup37 xyl15 mtl1 tsx33 str31 recF143 supE44 lacIqZΔM15 ). The segment commences at np 175 of GenBank database entry ECOSTR1 (accession number J01688), and includes the entire coding sequence of rpsL, 128 bp of DNA 5′ of the coding region that contains the complete promoter, and 22 bp downstream of rpsL cloned in the same orientation as the downstream tcr gene. Host XL-1 blue {recA1 endA1 gyrA96 thi1 hsdR17 supE44 relA1 lac F′[proAB lacIqZΔM15 Tn10 (Tetr)]} and the vector pBluescript II SK+ were obtained from Stratagene. Strains used for nonsense suppression and ribosomal elongation rate assays, kindly supplied by Dr Diarmaid Hughes (Uppsala University, Sweden), were as follows. Strains UD1D2 {Δ(lac proAB) Ara argE malA rpoB(Rifr) thi F′[lacI14 proA+,B+]} and UD5B6 {Δ(lac proAB) Ara argE malA rpoB(Rifr) thi F′[lacIZΔ14 UGA189 proA+,B+]} were used in nonsense suppression analysis, and strain UD8C1{Δ(lac proAB) Ara argE malA rpoB(Rifr) thi F′23[lac+ proA+,B+]} was used for elongation rate assays.

Bacterial cultures

Except when stated, cultures were grown on L-agar plates or in L-broth (Sambrook et al., 1989) containing 50 μg ml−1 ampicillin plus, when indicated, 1 mM isopropylthiogalactoside (IPTG) and streptomycin at concentrations shown in legends to figures and tables. Host DS941 was maintained on plates or broth containing 25 μg ml−1 streptomycin. For growth rate analysis, cultures in L-broth, supplemented with antibiotics and/or IPTG as indicated, were inoculated at 1:100 dilution from overnight cultures grown in L-ampicillin only. Bacterial growth was measured by optical density at 600 nm. Strains UD1D2, UD5B6 and UD8C1 were grown on M9 minimal medium (Sambrook et al., 1989) containing 0.2% glucose and supplemented with 0.4 mM arginine and 0.2 mM thiamine. After transformation, selection for pBR322–plasmid constructs was provided by adding ampicillin to 100 μg ml−1.


All oligonucleotides were purchased from DNA Technology. Primers for amplification of the chromosomal rpsL gene were rpsL1 (ACGTGTTTACGAAGCAAAAGCTAAAACC) and rpsL2 (ACATTTAAGTTAAAACGTTTGGCCTTACT). Primers for the PCR in vitro mutagenesis strategy outlined in Fig. 1 were rpsL3 (CGGAATTCTGATGGCGGGATCGTTGTATATT) as chimeric 5′ primer introducing an EcoRI site, underlined, rpsL4 (CCAAGCTTGGCCTTACTTAACGGAGAACCA) as 3′ chimeric primer and introducing a HindIII site, underlined, plus mutagenic primers L56→H-1 (GCCGTGTTCGTCACACTAACGGTTTCGA, mutated nucleotides underlined) and its exact complement L56→H-2, or K87→Q-1 (TGGCGGTCGTGTTCAAGACCTCCCGG, mutated nucleotide underlined) and its exact complement K87→Q-2. For PCR cloning into pBluescript II SK+, primer rpsL4 was used together with the alternate chimeric 5′ primer rpsL5 (CGGAATTCTAAGCTAAAACCAGGAGCTATTTA, EcoRI site underlined). PCR in vitro mutagenesis of rpsL in pBluescript II SK+ used the mutagenic primers listed above, plus terminal primers rpsL5 and rpsL4. Random mutagenesis of residue L56 used alternate primers L56→Z-1 (GCCGTGTTCGTNNNACTAACGGTTTCGA) and its exact complement L56→Z-2. Primers for sequencing from pBR322 constructs were: pBR-F (CATGACATTAACCTATAAAAATA) and pBR-R (GTGCCTGACTGCGTTAGCAA), located just beyond the vector sites used for cloning (EcoRI and HindIII respectively).

PCR and molecular cloning

Except when stated, all steps were carried out using standard procedures (Sambrook et al., 1989). PCR reactions containing 20 pmol of each of the appropriate primers used Pfu thermostable DNA polymerase (Stratagene) in the manufacturer's buffer, over 30 cycles of annealing for 45 s at 60°C followed by extension for 2 min. PCR products created using the mutagenic primers (see scheme outlined in Fig. 1) were purified on QIAquick spin columns, diluted 100-fold, mixed together and used as templates in a second-round PCR reaction to generate full-length products terminated with the restriction sites for cloning. Products for cloning were purified on spin columns, digested with 10 units each of EcoRI and HindIII (New England Biolabs, manufacturer's buffer) in a total reaction volume of 100 μl, phenol-extracted, then co-precipitated in a fivefold molar excess with 350 ng EcoRI/HindIII-digested vector. For cloning into pBR322, the cut vector DNA was first gel purified to eliminate the short stuffer fragment between the cloning sites. Ligation was carried out using 1 unit of T4 DNA ligase (New England Biolabs) in the manufacturer's buffer at room temperature for 30 min, in a total volume of 10 μl. Ligation products were transformed into competent cells of E. coli DS941, and plated on L-agar plates containing 50 μg ml−1 ampicillin. Colonies containing appropriate recombinant plasmids were identified by single-colony PCR using vector-specific primers (55°C annealing temperature), and plasmid DNA was then isolated, using appropriate Promega Wizard kits, from small (2 ml) or large (100 ml) cultures grown overnight in the presence of ampicillin. Plasmid DNAs, after characterization, were retransformed into XL-1 blue, and single colonies were picked for analysis.

DNA sequencing

Plasmid DNAs and PCR products were sequenced using dye-terminator chemistry on the Perkin Elmer ABI 310 Genetic Analyser, with kit reagents supplied by the manufacturer plus standard (M13R, T7) and customized primer oligonucleotides as listed above.

DNA sequences were analysed using the GCG package (Program Manual for the Wisconsin Package, 1994).


Various volumes of growing cultures, typically 1–1.5 ml adjusted for cell density based on OD600, were centrifuged in a microfuge for 30 s, washed once in 100 mM Tris-HCl, pH 7.4, resuspended in 200 μl SDS sample buffer (Sambrook et al., 1989) and boiled for 15 min. Aliquots (20 μl) were run on SDS–PAGE (15% polyacrylamide gels) and stained using standard procedures (Sambrook et al., 1989).

Ribosomal elongation rate assays

For analysis of elongation rates in vivo, colonies were picked as fresh transformants, to avoid additional mutations in plasmid constructs, and plasmids were also resequenced after growth of cells to check their identity. Elongation rate determinations were made essentially as described by Schleif et al. (1973) and Andersson et al. (1982), but using ampicillin selection. Samples were taken at 15 s intervals after induction with IPTG, starting from 1 min. Two parallel 0.2 ml samples from cold chloramphenicol solution were taken to subsequent steps, and average values of nitrophenol formation of these were calculated. The data were transformed by subtracting average values of parallel zero-time samples and by normalizing cell densities at the start of the assay to OD600 = 1.0 and ONPG incubation times to 1 h. Data points were then plotted as square roots of corrected OD420 values.

Nonsense suppression in vivo

Freshly transformed clones (4–5 single colonies for each rpsL variant) were grown to mid-log phase. Growth was stopped on ice and parallel samples (0.1 ml from UD1D2 clones and 1.0 ml from UD5B6 clones) were pelleted and resuspended in 1 ml of ice-cold Z buffer (Miller, 1972). Measurements of β-galactosidase activity were made as described by Miller (1972), with typical incubation times of 2 h and 15–16 h for UD1D2 and UD5B6 clones respectively. OD420 values were normalized to the amount of cells and time of incubation, and averaged for each rpsL variant grown in a given host. Suppression values for each rpsL variant were then derived by taking the ratio of averaged, normalized OD420 values for a given variant in the two host strains.


Financial support for this work was provided by the Finnish Academy, European Union, Juselius Foundation and Tampere University Hospital Medical Research Fund. We are grateful to Anja Rovio for technical assistance. We thank Diarmaid Hughes and Charles Kurland for advice and bacterial strains, as well as Bryn Bridges, Alan Lehmann and David Sherratt for useful discussions.