The flagellar regulon of
Salmonella enterica serovar Typhimurium includes over 60 genes (reviewed in references
11 and
36). The transcripts of the flagellar regulon are organized into a transcriptional hierarchy based on three promoter classes that are temporally regulated in response to assembly (
28) (Fig.
1). At the top of this hierarchy lies the flagellar master operon,
flhDC, where the fundamental decision to produce flagella is controlled. The
flhDC operon is expressed from what is defined as the class 1 promoter (
58). The FlhD and FlhC proteins form a heteromultimeric complex (FlhD
2C
2) that acts as a transcriptional activator to promote σ
70-dependent transcription from the class 2 flagellar promoters (
34,
35). The class 2 promoters direct transcription of genes needed for the structure and assembly of the flagellar motor structure, also known as the hook-basal body (HBB). Upon HBB completion, class 3 promoters are transcribed by σ
28-RNA polymerase, which is specific for flagellar class 3 promoters. Prior to HBB completion, σ
28-RNA polymerase is inhibited by the anti-σ
28 factor, FlgM. Upon HBB completion, FlgM is secreted from the cell, presumably through the completed HBB structure, and σ
28-dependent transcription ensues. In this way, genes whose products are needed only after HBB formation, such as the flagellin filament genes, are transcribed only when there is a functional motor onto which they may be assembled.
DISCUSSION
An extensive search was performed with 150 independent cultures in order to look for novel regulatory mutants affecting FlgM inhibition of σ
28-dependent transcription of the
fliC promoter. The strain in which the selection was set up had an FlgM-LacZ fusion that would inhibit σ
28-dependent transcription even in strains with a functional HBB structure, because the fusion of LacZ to FlgM prevents secretion through the HBB structure (
25). The selection utilized a fusion of the
fliC promoter (P
fliC) to the
cat gene (P
fliC-cat), so that when P
fliC was transcribed, the cells became Cm
r. The selection was done at 30°C and then coupled to a screen for inability to grow at 42°C. In this way, we hoped to identify previously uncharacterized essential genes that might play a role in FlgM inhibition of σ
28-dependent transcription.
The selection yielded some unexpected results. All but one of the Cm
r revertants (non-ts) carry mutations that mapped to either the
flgM or the
fliA region. We expected mutants defective in the
clpXP protease genes. It had been shown that loss of ClpPX protease results in a more stable σ
28 protein that can overcome FlgM inhibition of σ
28-dependent transcription in HBB mutant strains (
3,
53). We propose that the addition of LacZ to the C terminus of FlgM enhances the stability of FlgM, resulting in a selection that is specific for mutations that are specific to FlgM-σ
28 interactions, such that loss of ClpXP protease does not stabilize σ
28 enough to overcome the excess inhibition by the FlgM-LacZ fusion. The fact that 22 of the 150 cultures yielded double mutants carrying mutations that mapped to both the
flgM and
fliA regions suggests that the selection used (Cm
r by P
fliC-
cat in the presence of FlgM-LacZ) required a significant loss in the anti-σ
28 activity of the FlgM-LacZ fusion. However, the mutants were not screened for the presence of mutations in
clpXP protease, so it is possible that many of the mutants that had either an
flgM-linked or an
fliA-linked allele also carried a second mutation in
clpXP.
In the screen for the identification of essential genes involved in FlgM regulation of P
fliC transcription, one ts-lethal mutant from the 150 independent cultures that prevented FlgM-LacZ inhibition of P
fliC-
cat transcription at the permissive temperature and failed to grow on LB medium at 42°C was isolated. This mutant had a G
10:A
10 substitution mutation in the D-stem of an essential tRNA gene,
serT (
4), which encodes
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)(
5). The
serT gene is essential, because only
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)will recognize the UCA codon (
26,
49).
It was fortuitous that we used P
fliC-
cat and not P
fliC-
lac as our reporter with some other nonsecreted FlgM fusion. If we compare strains TH9784 (
hpaB::Tn
10dTc Δ
flgG-L2157
fliC5050::Mu
dJ
fljBenx vh2) and TH9785 (TH9784
serT), we find that they are isogenic
serT+ and
serT strains, respectively. The β-galactosidase assays shown in Table
2 show that TH9784 has 10-fold-higher levels of activity and that the cells are darker blue on X-Gal (not shown). This all fits with the
serT mutants having reduced FlgM levels corresponding to increased transcription of P
fliC-
lac. However, if we compare the strains on MacConkey lactose indicator plates, we find that TH9784 is pink in color after overnight incubation, indicating low levels of lactose utilization, while TH9785 is completely white, indicating no lactose fermentation (not shown). How can this be if TH9785 has 10-fold-higher levels of β-galactosidase? It takes many more genes than just the β-galactosidase gene to ferment lactose to achieve the Lac
+ phenotype. It is likely that many genes required to ferment lactose are defective in the presence of the SerT-G
10:A
10 tRNA, resulting in the Lac
− phenotype, and that the expression of only one gene is required for the LacZ
+ phenotype. Because
lacZ mRNA has 8 UCA codons, the 10-fold increase in β-galactosidase in TH9785 probably represents a lower limit of increased transcription of P
fliC-
lac due to reduced translation of
lacZ mRNA in the presence of the SerT-G
10:A
10 tRNA.
The textbook picture of a tRNA is the secondary cloverleaf structure: an amino acid acceptor stem on top; an anticodon stem and loop on the bottom; a left arm known as the D-stem and D-loop, due to the presence of many dihydrouridine modifications in the D-loop; and a right arm known as the TψC stem and TψC loop, due to the presence of a conserved pseudouridine (ψ) residue (
5). When folded into the tertiary structure, the D-stem and D-loop become fused with the anticodon stem and loop (
29,
43). The base substitution of A for G would result in the loss of a G:C base pair in the D-stem that might allow a conformational change in the tRNA tertiary structure at 42°C. Such a conformational change probably does not result in a nonfunctional tRNA, because several proteins containing UCA codon products are normally expressed in an
Escherichia coli serT mutant strain with the identical temperature-sensitive lethal G
10:A
10 allele (
1).
The identical (G
10:A
10) substitution mutation in the
serT gene had been isolated previously in
Escherichia coli and named
divE42 because it affected cell cycle-dependent enzymes (
39). When shifted to 42°C, the cells grew until they doubled in volume, but when shifted down to low temperature, they began to divide synchronously. The mutant also exhibited low β-galactosidase levels at the nonpermissive temperature (42°C) (
1,
40). Synthesis of a number of cell cycle-dependent proteins was decreased significantly in the
divE42 mutant at 42°C (
47). DNA sequence analysis revealed
divE42 to be a mutation in
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \), and it was renamed
serT (
50). The
divE42 allele could be complemented in
trans by a wild-type copy of the
serT+ gene expressed from an F′ plasmid, indicating that it was a recessive allele (
47). Remarkably, a pBR322-derived plasmid expressing the
divE42 allele could also complement the chromosomal
divE42 allele for growth at 42°C, implying that stability or a concentration-dependent interaction required for translation of essential genes is lost at 42°C (
57). Furthermore, the
E. coli divE42 allele is more labile at 44°C than at 42°C, suggesting that the mutant phenotype shows a gradient effect with temperature (
1). Presumably, the translation of essential genes is reduced at elevated temperatures to the degree that the cell can no longer grow. It was fortuitous that the
Salmonella serT mutant was defective in
flgM translation at 30°C, since this enabled it to be picked up in our mutant screen.
In
Salmonella serovar Typhimurium,
serT was identified by DNA sequence analysis as flanking
Salmonella pathogenicity island 5 (
54). The
serT gene is located basically at the same positions in both the
E. coli and
Salmonella serovar Typhimurium chromosomes. Further studies with the
divE42 allele from
E. coli suggested that many proteins with serine residues expressed from UCA codons are made at normal levels in the presence of the
divE42 allele, suggesting that the context of translation may be critical to the effect of
divE42 on protein expression. The temperature-sensitive phenotype of the
divE42 mutant could be suppressed by a double mutation in
rne (RNase E) and
pnp (polynucleotide phosphorylase) (
31), members of the mRNA degradosome (
8). The
lacZ mRNA, containing eight UCA codons, is unstable in the
divE42 mutant, but stability is restored in a strain that also carries the
rne-1 and
pnp-7 mutations (
1). This suggests that a defect in translation of UCA by the G
10:A
10 substitution in
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)could lead to mRNA instability.
Other suppressors of the temperature-sensitive growth of the
E. coli divE42 allele were isolated. Intragenic suppressors include C
25:U
25, which can now form a base pair with A
10, and G
67:A
67, which forms a base pair with the acceptor stem (
41). One unusual mutant, resulting from a single amino acid substitution of amino acid 275 of trigger factor (TF), actually reversed the phenotype: the
tig divE42 double mutant grew at 42°C but not at 30°C, and the growth defect at 30°C could not be complemented by
divE+ but could be complemented by
tig+ (
38). TF is a ribosome-associated chaperone that associates with the DnaK protein-folding system to catalyze proper folding of newly synthesized cytosolic proteins. TF is nonessential except in
dnaK mutants (
17,
52). TF consists of an N-terminal domain that mediates ribosome binding and by itself is sufficient to complement for chaperone activity (
30). A central domain with peptidyl-prolyl
cis/trans isomerase activity, substrate binding activity, and a C-terminal domain contributes to the chaperone activity (
30). Amino acid 275 is located at the beginning of the C-terminal domain (amino acids 248 to 432) (
30). In their model, Nakano et al. propose that reduced TF at 42°C compensates for the reduction of an essential cell protein (X) produced at low levels in the
divE42 mutant at 42°C. They propose that the ratio of TF to X determines lethality. In
divE42 tig+ cells, the TF
+:X ratio is too high at 42°C, and cell division is inhibited. In the
divE42 tig double mutant at 42°C, low TF activity resulting from the
tig mutation and low X levels resulting from the
divE42 mutation result in normal TF:X ratios, and the cells grow. At 30°C, X levels are wild type in the
divE42 mutant, but TF is low, so the TF:X ratios are too low.
The serine codon usages in Salmonella serovar Typhimurium per thousand are 8.4 (UCU), 10.6 (UCC), 8.0 (UCA), 9.5 (UCG), 8.6 (AGU), and 17.7 (AGC), and in E. coli are 8.4 (UCU), 8.6 (UCC), 7.1 (UCA), 8.9 (UCG), 8.8 (AGU), and 16.0 (AGC). While UCA usage is the lowest, UCA is far from rare. Suppression of the G10:A10 substitution mutation in \( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)by mutants in the mRNA degradosome (rne-1 and pnp-7 mutations) suggests that stalling plays a critical role in the phenotype. The temperature-sensitive nature of the mutant is as follows: defective at 30°C (evidenced by reduction in flgM translation), more defective at 42°C (evidenced by inhibition of cell division and significant reduction in levels of some proteins), and even more defective at 44°C (evidenced by further reduction in affected proteins at 44°C). Self-complementation of divE42 by the divE42 allele present in high copy numbers is consistent with the loss of a concentration-dependent interaction with G10:A10-\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)that is dependent on the G10:A10-\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)tertiary structure.
We rule out the possibility that the UCA codon early (amino acid codon 7) in the flgM gene accounts for sensitivity to the serT mutation. It was possible that G10:A10-\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)- dependent stalling of translation early in the flgM gene transcript resulted in enhanced flgM mRNA degradation. We examined the 12 flg-linked flagellar hook-basal body structural genes for numbers and locations of UCA codons and obtained the following results (with gene, number of amino acids, number of serine residues, and number of UCA codons listed in that order and followed by the UCA codon position[s], if any, in parentheses): flgA, 219, 13, 1 (49); flgB, 138, 8, 0; flgC, 134, 9, 1 (107); flgD, 232, 15, 0; flgE, 400, 34, 3 (18, 173, 180); flgF, 250, 15, 2 (30, 198); flgG, 260, 20, 2 (4, 64); flgH, 232, 20, 0; flgI, 365, 28, 3 (37, 109, 281); flgJ, 316, 26, 0; flgK, 553, 46, 4 (61, 105, 440, 490); and flgL, 317, 31, 1 (261). We observed 17 UCA codons/265 serine codons (0.64%) that were located throughout the genes. One gene, flgG, had a UCA at the position corresponding to amino acid 4, a position similar to that in the flgM gene. Given that the serT mutant shows normal motility at 30°C, it seems likely that the flagellar genes are translated at normal efficiencies.
As mentioned above, certain proteins with serine residues encoded by UCA are present at lower levels in the divE42 mutant strain, and the number of proteins affected increases with increasing temperature. One interpretation of the observed effect of the serT mutant tRNA is that there is less-efficient charging of it. This should give a lower concentration of the ternary complex and thereby stalling of the ribosome and less-efficient translation. This could significantly affect mRNA instability of some transcripts but not that of others.
Another interpretation is that the context of the UCA codon in the mRNA is critical for translation by G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \). Such a context effect could be due to an adjacent base influencing the ability to read the UCA codon or the fitting of an adjacent tRNA to the UCA codon. The binding efficiency of tRNA
fMet for the initiation codon, AUG, of the Qβ coat gene was shown to increase more than threefold when the G residue adjacent to the 3′ side of the AUG was replaced by an A residue (
51). Structures of tRNA species can be altered to read four codons, consistent with the possibility that even a given triplet-specific tRNA could be influenced by adjacent bases (
44). There are several lines of evidence to suggest that efficient translation of UCA codons by G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)might be dependent on which tRNAs read adjacent codons. The efficiency of the
supE amber suppressor in translating the UAG codon varies over an order of magnitude depending on the nucleotide adjacent to the 3′ side of the codon (
7). The frameshift suppressor tRNA SufJ will read four codons where the first three are ACC and the fourth base can be anything, suggesting that the SufJ tRNA sterically prevents another tRNA from reading the fourth codon (
6). The influence of reading context on the efficiency of nonsense suppression has been observed by a number of labs (
2,
13,
14,
18-
20,
45,
56). Thus, context is critical for translational efficiency, and it is likely that the fact that some UCA-containing mRNAs are translated more efficiently than others represents an example of context effects. If the G
10:A
10 substitution in
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)affects translation in a context-dependent manner, the suppression of
divE42 by itself in high copy numbers would be consistent with an effect on the 5′ side. This would mean that the ability of G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)to interact with the tRNA in the P site of the ribosome is more likely to be sensitive to G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)concentration than the ability of a tRNA to enter the A site with G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)present in the P site. It remains to be determined if the ability of G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)to read the UCA codon is affected by context. To our knowledge, the involvement of position 10 in the D-stem of tRNA in tRNA-tRNA interactions during translation has not been demonstrated. However, tRNA-tRNA interactions during translation have not been characterized to any significant extent. The finding that some proteins with UCA codons are translated normally, while others are not, argues against a simple loss-of-function type of mutant. It is possible that G
10:A
10-
\( \(\mathrm{tRNA}_{\mathrm{cmo}5\mathrm{UGA}}^{\mathrm{Ser}}\) \)results in a conformational change that affects its ability to sterically fit adjacent to specific tRNA species in the ribosome during translation. The results presented here remind us that context effects can be critical for gene expression and could play an important role in gene regulation by determining the level at which a given gene is expressed in the cell and also in gene evolution.