Yersinia enterocolitica Type III Secretion: Mutational Analysis of the yopQ Secretion Signal

Abstract

Pathogenic Yersinia spp. secrete Yop proteins via the type III pathway. yopQ codons 1 to 15 were identified as a signal necessary and sufficient for the secretion of a fused reporter protein. Frameshift mutations that alter codons 2 to 15 with little alteration of yopQ mRNA sequence do not abolish type III transport, suggesting a model in which yopQ mRNA may provide a signal for secretion (D. M. Anderson and O. Schneewind, Mol. Microbiol. 31:1139-1148, 2001). In a recent study, the yopE signal was truncated to codons 1 to 12. All frameshift mutations introduced within the first 12 codons of yopE abolished secretion. Also, multiple synonymous mutations that changed the mRNA sequence of yopE codons 1 to 12 without altering the amino acid sequence did not affect secretion. These results favor a model whereby an N-terminal signal peptide initiates YopE into the type III pathway (S. A. Lloyd et al., Mol. Microbiol. 39:520-531, 2001). It is reported here that codons 1 to 10 of yopQ act as a minimal secretion signal. Further truncation of yopQ, either at codon 10 or at codon 2, abolished secretion. Replacement of yopQ AUG with either of two other start codons, UUG or GUG, did not affect secretion. However, replacement of AUG with CUG or AAA and initiating translation at the fusion site with npt did not permit Npt secretion, suggesting that the translation of yopQ codons 1 to 15 is a prerequisite for secretion. Frameshift mutations of yopQ codons 1 to 10, 1 to 11, and 1 to 12 abolished secretion signaling, whereas frameshift mutations of yopQ codons 1 to 13, 1 to 14, and 1 to 15 did not. Codon changes at yopQ positions 2 and 10 affected secretion signaling when placed within the first 10 codons but had no effect when positioned in the larger fusion of yopQ codons 1 to 15. An mRNA mutant of yopQ codons 1 to 10, generated by a combination of nine synonymous mutations, was defective in secretion signaling, suggesting that the YopQ secretion signal is not proteinaceous. A model is discussed whereby the initiation of YopQ polypeptide into the type III pathway is controlled by properties of yopQ mRNA.

Three pathogenic Yersinia species, Y. pestis, Y. pseudotuberculosis, and Y. enterocolitica, use a type III secretion mechanism to inject protein toxins into eukaryotic cells in order to escape phagocytic killing (16, 37, 41). The genes for this mechanism are carried by the 70-kb virulence plasmid (1, 8, 18, 34, 35, 43). Twenty-one ysc genes (Yop secretion genes) encode secretion machinery components that allow transport of 14 Yops (Yersinia outer proteins) across the bacterial double-membrane envelope (14). The open reading frames of several yop genes have been fused to the 5′ ends of coding sequences for reporter genes, e.g., Escherichia coli alkaline phosphatase (phoA) (truncated for its signal peptide) or the alpha fragment of β-galactosidase (lacZ), generating translational hybrids that are transported by the type III pathway (30). Similarly, fusion of yop coding sequences to the 5′ end of the adenylate cyclase-encoding domain of cya (17) results in the secretion of Yop-Cya fusions in a manner that resembles type III transport of native Yops (38, 40). npt encodes neomycin phosphotransferase, a cytoplasmic protein that confers bacterial resistance to aminoglycoside antibiotics (36). Yop-Npt fusions are also secreted by the type III secretion machinery (4, 9).

Fusion of the first 15 codons of yopE, yopH, yopN, or yopQ to the 5′ end of cya or npt leads to the type III secretion of hybrid Yop proteins (2, 4, 6, 38, 39). Sory et al. as well as Schesser et al. proposed that the amino acid sequence generated from the first 15 codons functions as a signal peptide to mediate substrate recognition by the type III machinery (38, 39). Lloyd et al. developed this model further, predicting an amphipathic helical structure as a common substrate property of all Yop signal peptides (27). yop secretion signals (codons 1 to 15) have been altered by frameshift mutations immediately following the AUG, whereupon reporter expression was restored by reciprocal mutations at the fusion site (4). Many of these frame shift mutations do not affect secretion signaling even though the peptide sequence is completely altered (4). It has been concluded that codons 1 to 15 of yop mRNA may function as a signal for the type III secretion of Yop proteins (5). Similar results and conclusions have been reached through the study of AvrBs2, a protein that is transported by the type III machinery of the plant pathogen Xanthomonas campestri (33).

Recent work has aimed at generating experimental evidence for a clear distinction between the perceived modes of substrate recognition, i.e., signal peptide or RNA signal hypothesis. Lloyd et al. studied a yopE signal that was truncated to codons 1 to 12 (27). While the wild-type secretion signal consisting of codons 1 to 12 is functional (38), frameshift mutations that alter codons 1 to 12 are defective in secretion signaling (27). Further, multiple synonymous mutations that change the mRNA sequence of codons 1 to 12 without altering the signal peptide did not affect secretion (27). These results suggest that yopE mRNA may not be involved in signaling type III transport, thereby favoring the signal peptide model (26).

We wondered whether the reported properties of yopE are universal for all yop signals. It is reported here that codons 1 to 10 of yopQ act as a minimal secretion signal. All further truncation of yopQ, at either codon 10 or 2, abolished secretion signaling. Replacement of yopQ AUG with two other start codons, UUG or GUG, did not affect secretion. Replacement of AUG with CUG or AAA and initiation of translation with npt at the fusion site did not permit Npt secretion, suggesting that the translation of the yopQ secretion signal is a prerequisite for the type III transport of YopQ. Frameshift mutations that altered yopQ codons 1 to 10, 1 to 11, and 1 to 12 abolished secretion signaling, whereas frameshift mutations of yopQ codons 1 to 13, 1 to 14, and 1 to 15 did not. Codon changes at yopQ positions 2 and 10 affected secretion when placed within the first 10 codons but showed no effect when positioned within a larger fusion of the first 15 codons of yopQ. An mRNA mutant, generated by a combination of nine synonymous mutations of yopQ codons 1 to 10, was defective in secretion signaling and in the regulation of yopQ expression. These data are discussed in the context of a model whereby the posttranscriptional regulation of yopQ mRNA translation may lead to the initiation of YopQ polypeptide into the type III pathway.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Y. enterocolitica W22703 (15), MC3 [Δ(yopQ)] (6), and VTL2 [Δ(yopD)] (24) have been described previously. E. coli strains MM294 and DH5α served as hosts for DNA manipulations (20, 32).

DNA methods and plasmid construction.

yopQ fusions to npt were constructed using the single-copy-number vector pHSG576 as a backbone. Briefly, expression of all constructs was driven by a 500-nucleotide segment upstream of the yopQ open reading frame. Introduced into this segment by PCR was a BglII site that was engineered 19 nucleotides upstream of the translation start site by the replacement of the guanine nucleotide at −17 relative to the start codon with adenine and the cytosine at −15 with thymine. These mutations (which gave rise to the engineered BglII site) did not affect the expression or secretion of wild-type YopQ (data not shown). Annealed oligonucleotides specifying the desired insertion were synthesized and ligated between the BglII site and a KpnI site fusing the insertion to Npt. When +1 or −2 frameshifts were introduced into the yopQ minimal secretion signal, two UAA nonsense codons were encountered. These codons were changed to CAA.

Protein electrophoresis and immunodetection.

Overnight cultures of Yersinia strains were grown in tryptic soy broth (Difco) at 26°C and diluted 1:50 into 4 ml of tryptic soy broth containing 5 mM EGTA. Antibiotics (30 μg of chloramphenicol/ml) were added to the media for plasmid maintenance as necessary. Cultures were grown at 26°C for 2 h and then shifted to 37°C for 3 h. Aliquots (1.4 ml) of cultures were removed and centrifuged at 15,000 ×g for 15 min. One milliliter of the culture supernatant was removed and precipitated with 75 μl of 100% trichloroacetic acid. Cell pellets were suspended in 700 μl of water, and 500 μl of this suspension was precipitated with 500 μl of 10% trichloroacetic acid. Precipitated proteins were washed with acetone, solubilized in sample buffer, and analyzed by sodium dodecyl sulfate-15% polyacrylamide gel electrophoresis. Following electrotransfer to polyvinylidene fluoride membrane, proteins were immunoblotted with α-Npt, α-YopE, and anti-chloramphenicol acetyltransferase (CAT) (purified polyclonal rabbit antibodies and anti-rabbit horseradish peroxidase conjugates). The chemiluminescent signal was visualized on a Fluorchem 8800 Imaging System (Alpha Innotech) and quantified.

RESULTS

Alternate start codons do not abolish the secretion of YopQ.

One prediction of the RNA signal hypothesis is that translational control is imposed upon secretion substrates. We observed that the only commonalities in the known or presumed secretion signals (i.e., the first 15 codons) of all 14 Yop proteins are that the initial amino acid is invariably a methionine and that the initiating codon is invariably an AUG. We wondered if alternate start codons are tolerated by type III secretion substrates. To address this question, we fused various truncations of yopQ to the 5′ end of the npt reporter gene. Reporter fusions have been successfully used in the study of various secretion pathways in prokaryotes and eukaryotes (28, 29). Indeed, the only way to discover the sufficiency of a protein secretion signal is to demonstrate that its capability for secretion signaling can be conferred to a protein that is otherwise not a secretion substrate (7, 21). Researchers in our laboratory have used the gene encoding neomycin phosphotransferase II (npt) for the analysis of type III secretion signals (9). Recently, it has been stated that Npt may be transported by the Salmonella flagellar type III secretion pathway and as such is unfit for studies of Salmonella or Yersinia type III secretion (26, 27). Although the experiments that led to this conclusion have apparently not yet been published (13), we wondered whether previous analysis of the Npt reporter (2, 4, 6, 9-12, 22, 24, 25) overlooked the possibility that the protein is transported nonspecifically by the Yersinia type III pathway. Expression of npt alone from the yopQ promoter Y. enterocolitica W22703 (pDA183) did not result in Npt secretion when bacteria were grown at 37°C in the presence of 5 mM EGTA, a condition that led to the secretion of YopE and YopQ (Fig. 1). These data suggest that Npt does not enter the type III pathway in a nonspecific manner and may be used as a reporter protein for the mapping of secretion signals.

FIG. 1.

Translational initiation of codons 1 to 15 of yopQ is required for YopQ_1-15-Npt secretion. Yersinia type III secretion was measured by separating the medium (S) and the bacterial pellet (P) of centrifuged cultures that had been induced by temperature shift and low calcium. Proteins in both fractions were separated on sodium dodecyl sulfate-15% polyacrylamide gel electrophoresis, electroblotted onto polyvinylidene dichloride membrane, and detected with antisera raised against purified YopQ (αYopQ) and YopE (αYopE) (both type III secretion substrates), Npt (αNPT) (reporter protein for fusion experiments), and CAT (αCAT) (cytoplasmic protein). Y. enterocolitica MC3 (yopQ1) expresses wild-type yopQ from the yopQ promoter on the single-copy-number plasmid pDA218. Y. enterocolitica W22703 (wild type) expresses npt or yopQ-npt translational fusions from the yopQ promoter cloned on single-copy-number plasmids. Chemiluminescent signals were scanned and quantified and are reported as the signal intensity of protein in the culture supernatant as a percentage of the total amount of protein in both the culture supernatant and the bacterial pellet.

Wild-type yopQ was cloned on a single-copy-number plasmid and expressed from its native promoter in a Yersinia yopQ mutant. YopQ was secreted into the extracellular medium (96% secretion) (Fig. 1). Fusion of npt to the 3′ end of a yopQ variant with the yopQ stop codon deleted generated a hybrid protein that was also secreted by the type III pathway (78% secretion). As demonstrated previously, fusion of the first 15 codons of yopQ to npt resulted in 61% secretion of the YopQ_1-15-Npt hybrid (6). Substituting the AUG start codon of yopQ for UUG within the yopQ_1-15-npt fusion did not affect secretion, as 63% of the fusion protein was found in the culture medium (Fig. 1). GUG, a start codon that is far rarer than AUG or UUG (19), also supported the secretion of the YopQ_1-15-Npt hybrid. As a control, CUG, an arginine codon that is not known to initiate translation in bacteria (19), as well as AAA, a lysine codon, did not promote secretion of Npt. Synthesis of Npt in pKR25 (CUG)- and pKR27 (AAA)-transformed yersiniae was greatly reduced, indicating that the translational initiation signals (Shine-Dalgarno sequence) of yopQ are inappropriately spaced for promotion of efficient expression of npt. As internal controls for type III secretion and proper fractionation, the expression of yopQ-npt hybrids did not interfere with the secretion of YopE and CAT was found only in the bacterial cell pellet (Fig. 1). It can thus be concluded that substitution of the AUG start codon of yopQ does not interfere with the ability of YopQ to be secreted. Further, translation of yopQ codons 1 to 15 is a prerequisite for secretion signaling and type III transport of fused reporter proteins.

Defining the minimal yopQ secretion signal.

In an effort to define the absolute minimal secretion signal, we successively deleted codons from the 3′ end of the yopQ signal (the first 10 codons) (Fig. 2). The first 10 codons of yopQ have previously been shown to function as a secretion signal for the npt reporter (51% secretion). Further 3′ truncations of this signal led to the abolishment of secretion but did not significantly reduce the synthesis of YopQ-Npt hybrids (Fig. 2). To define the 5′ boundary of the yopQ secretion signal, we generated deletions immediately following the AUG start codon. Deletion of codon 2 (UUU, phenylalanine at position 2) as well as deletion of codons 2 and 3 (AUU, isoleucine at position 3) abolished the secretion of YopQ-Npt hybrids (Fig. 3). These observations as well as previously published data allow us to surmise that the first 10 codons of yopQ are necessary and sufficient for secretion signaling. Furthermore, codons 2 and 10 are essential in order for yopQ to be recognized as a secretion substrate.

FIG. 2.

Mapping the yopQ secretion signal by truncation from the 3′ end of the coding sequence. (A) The nucleotide sequence of the yopQ UTR and codons 1 to 15 and the fusion site with the coding sequence for npt. yopQ-npt fusions were generated by inserting annealed oligonucleotides between the BglII and KpnI sites of pKR63 and by transformation of recombinant plasmids into Y. enterocolitica W22703. (B) yopQ codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

FIG. 3.

Further mapping the yopQ secretion signal by deletion of coding sequence. The secretion signal of yopQ_1-10 was mapped by deleting codon(s) 2 (pKR64), 2 to 3 (pKR65), 2 to 4 (pKR66), 2 to 5 (pKR67), 2 to 6 (pKR68), 2 to 7 (pKR69), and 2 to 8 (pKR70). yopQ codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

The yopQ UTR is dispensable for substrate recognition.

The yopQ-npt fusion constructs described above are under the transcriptional control of yopQ promoter DNA and under the translational control of the yopQ mRNA upstream untranslated region (UTR), respectively (3, 6). Previous work reported that yopQ promoter DNA is not essential for YopQ secretion, as its replacement with the lacZ promoter sequence did not affect type III secretion (6). We wondered whether the yopQ UTR is required for type III secretion and therefore cloned full-length yopQ fused to npt under the control of the npt promoter and 5′ UTR in pEC78 (Fig. 4A). Although the expression of yopQ-npt in pEC78 was greatly reduced compared to the expression of yopQ-npt from the yopQ promoter and 5′ UTR, type III secretion was not affected (100% secretion for YopQ-Npt when expressed from pEC78 compared to 86% from cells harboring pDA243) (Fig. 4A). We wished to determine whether the yopQ promoter and 5′ UTR are required for YopQ-Npt secretion when signaled by the first 10 codons of yopQ. yopQ_1-10-npt was cloned under the transcriptional and translational control of the npt promoter and 5′ UTR in pKR93 (Fig. 4B). As with that of full-length yopQ, the secretion signaling of yopQ_1-10-npt was not abolished by its fusion to the npt transcriptional and translational control elements, as 31% of the YopQ_1-10-Npt was found secreted into the extracellular medium. We therefore conclude that the minimal secretion signal of yopQ is located within the first 10 codons (30 nucleotides) of yopQ.

FIG. 4.

The yopQ UTR is not specifically required for type III secretion of YopQ_1-182-Npt or YopQ_1-10-Npt. (A) Y. enterocolitica W22703 was transformed with either pDA243 or pEC78 carrying yopQ_1-182-npt, which is expressed via fusion to either the wild-type yopQ promoter and UTR (pDA243) or the npt promoter and UTR (pEC78), respectively. (B) Y. enterocolitica W22703 was transformed with either pKR63 or pKR93 carrying yopQ_1-10-npt, which is expressed via fusion to either the yopQ promoter and UTR (pKR63) or the npt promoter and UTR (pKR93), respectively. yopQ codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

Some mutations in codons 2 and 10 of the minimal secretion signal of yopQ abolish secretion.

The observations that codons 1 to 10 of yopQ serve as a minimal secretion signal and that deletions of codons 2 and 10 abolish secretion suggested to us that codons 2 and 10 may play a critical role in signaling to the type III pathway. In wild-type yopQ, codon 2 (UUU) encodes phenylalanine. The homopolymeric UUU run was replaced with CCC (proline), GGG (glycine), or AAA (lysine) and fused to npt with the remainder of the yopQ_1-10 signal. Substituting CCC or GGG (encoding proline or glycine, respectively) for UUU in codon 2 resulted in the nearly complete loss of synthesis of the fusion protein (Fig. 5A). Previous work reported that yopQ is negatively regulated by a posttranscriptional mechanism requiring the gene products of yopD, lcrH, and yscM1 or yscM2 when the type III pathway is shut off (3, 42). This regulation is also effective for yopQ variants that are not secreted even under type III-inducing conditions (37°C and 5 mM EGTA without calcium) (3). Transformation of pKR65 and pKR66 into the yopD mutant strain VTL2 allowed expression of yopQ_1-10-npt but failed to restore secretion signaling of codon 2 CCC and GGG substitutions (Fig. 5B). Thus, although a complete nucleotide and codon substitution (phenylalanine to lysine) at position 2 can be tolerated without loss of function, two other codon 2 substitutions do abolish secretion signaling.

FIG. 5.

Codons 2 and 10 of the yopQ_1-10 secretion signal are essential for its function. yopQ_1-10-npt was mutagenized by replacement of codon 2 (CCC, GGG, or AAA replacing UUU) as shown in panel A or of codon 10 (GGU, GGA, GGC, or GGG replacing CGU) as shown in panel C. Type III secretion of YopQ_1-15-Npt in Y. enterocolitica wild-type (WT) strain W22703 was examined as shown in panels A and C. As shown in panel B, the codon 2 substitutions are expressed in a yopD mutant. yopQ_1-10 codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

The deletion analysis represented in Fig. 2 revealed that codon 10 plays an essential role in yopQ secretion signaling. Truncation of the yopQ signal to nine codons results in a replacement of CGU (arginine) with GGU (glycine) [KpnI fusion site (GGUACC) between yopQ and npt]. To determine whether the GGU substitution at codon 10 within the yopQ_1-10 signal (followed by the KpnI site GGUACC) affected secretion, the appropriate nucleotide changes were engineered and secretion of YopQ_1-10-Npt was examined (Fig. 5C). Indeed, replacement of codon 10 (CGU) with GGU (glycine) abolished type III secretion of the hybrid protein. All possible glycine codon substitutions inserted at position 10 (GGU, GGA, GGC, and GGG) displayed the same phenotype. It should be noted that these are the first yopQ mutants that display an uncoupled phenotype, separating YopQ synthesis defects from YopQ secretion defects.

The minimal yopQ_1-10 secretion signal does not tolerate frameshift mutations.

The mRNA signal hypothesis was born of the observation that the secretion signal within the first 15 codons of many type III secretion substrates could tolerate frameshift mutations that largely preserve the RNA sequence but thoroughly disrupt the encoded amino acid sequence. Recently, Lloyd and colleagues (27) demonstrated that codons 1 to 12 of yopE function as a secretion signal; however, frameshift mutations of this nucleotide sequence abolish secretion signaling. We wondered whether the minimal secretion signal of yopQ codons 1 to 10 could tolerate frameshift mutations. To that end, we fused the first 10 codons of yopQ to npt, introducing one of various frameshifts to the secretion signal by deleting or inserting adenine nucleotides immediately following the start codon. The reading frame was restored at the fusion site with the reporter protein, and any stop codons that were encountered were corrected with single nucleotide substitutions. The −3 frameshift was not investigated here, since the deletion of codon 2 had already been shown to abolish secretion (Fig. 3). In stark contrast to results obtained while frameshifting codons 1 to 15 of yopQ (6), introduction of five different frameshift mutations in the yopQ codon 1 to 10 minimal secretion signal completely abolished secretion of the hybrid protein (Fig. B). Thus, our results corroborate the data of Lloyd et al. (27), who observed that frameshift mutations of the secretion signal of yopE codons 1 to 12 abolished transport of YopE.

What is the minimum secretion signal of yopQ that could tolerate frameshift mutations? Previously, the secretion signal encoded in the first 15 codons of yopQ was shown to tolerate +1, −1, and −2 frameshifts (3, 6). yopQ codons 1 to 15, 1 to 14, 1 to 13, 1 to 12, and 1 to 11, each containing a +1 frameshift, were fused to the 5′ end of npt coding sequence and examined for type III secretion (Fig. 6C). It should be noted that these variants contain two stop codon suppressor mutations (codon 4 UA[A/C]AA and codon 8 UA[A/C]AA) as well as a suppressor of the frameshift at the npt fusion site. The results depicted in Fig. 6C demonstrated that the secretion signals of yopQ codons 1 to 15, 1 to 14, and 1 to 13 were capable of tolerating the +1 frameshift mutation, whereas the secretion signal within codons 1 to 11 was not. The yopQ codon 1 to 12 signal displayed a significant drop in function upon introduction of the +1 frameshift mutation. Thus, it appears that the ability to tolerate frameshift mutations is dependent on the length of the type III secretion signal. In the case of yopQ, secretion signals of 12 or fewer codons failed to promote the secretion of reporter proteins upon introduction of the frameshift mutation. Preliminary and yet unpublished data for frameshift mutations of the yopE secretion signal produced similar results (K. S. Ramamurthi, unpublished observations), corroborating the observations of Lloyd et al. (27).

FIG. 6.

The yopQ_1-10 secretion signal does not tolerate frameshift mutations. (A) mRNA sequences for codons yopQ_1-10 in various plasmid constructs. The secretion signal was mutagenized by inserting adenine nucleotides (+1, +2, or +3) or deleting nucleotides (−1 or −2) immediately following the AUG start codon and by correcting the reading frame at the fusion site with npt by the introduction of reciprocal deletions or adenine insertions. The +1 and −2 reading frame mutants also harbor suppressor mutations of stop codons at positions 4 and 3 and positions 8 and 7, respectively. (B) The yopQ_1-10 codon fusions to npt described for panel A were analyzed for type III secretion as described in the legend to Fig. 1. (C) The effect of a +1 frameshift mutation was analyzed within different size secretion signals: yopQ codons 1 to 15 (15, pKR98), 1 to 14 (14, pKR97), 1 to 13 (13, pKR96), 1 to 12 (12, pKR103), or 1 to 11 (11, pKR107).

The minimal yopQ secretion signal does not tolerate all synonymous mRNA substitutions.

Lloyd et al. sought to test the mRNA signal hypothesis by generating synonymous mutations that altered the mRNA sequence without changing the amino acid sequence (27). A mutant construct encompassing 17 nucleotide changes within the first 12 codons of yopE (36 nucleotides) was not defective in secretion signaling (27). The authors concluded that the amino acid sequence, but not mRNA, must therefore be involved in secretion signaling. In fact, we too observed that fusion of the first 11 codons of yopE, with all wobble nucleotides altered, to npt promoted the secretion of YopE_1-11-Npt (data not shown). To test the universality of the Lloyd et al. observation, we altered yopQ codons 2 to 10 and measured the secretion of YopQ_1-10-Npt. Figure 7A shows the mRNA sequence of yopQ_1-10 in pKR63 (wild-type signal) and pKR92 (synonymous mRNA signal). When plasmids were transformed into Y. enterocolitica W22703 (wild type), pKR63 carrying yopQ_1-10-npt was efficiently expressed and 44% of YopQ_1-10-Npt was found secreted into the extracellular medium. In contrast, pKR92 carrying yopQ_1-10-npt was not expressed. Presumably, translation of the yopQ_1-10-npt transcript in pKR92-transformed yersiniae is blocked by the YopD-, LcrH-, and YscM1/YscM2-mediated mechanism (3). To test this, plasmids pKR63 and pKR92 were transformed into Y. enterocolitica VTL2 (yopD1). The pKR63-encoded YopQ_1-10-Npt was efficiently secreted into the extracellular medium, as 53% of the polypeptide was found in the supernatant of centrifuged cultures. In contrast, the pKR92-encoded YopQ_1-10-Npt was not secreted into the extracellular medium (Fig. 7B). To determine whether single synonymous codon substitutions in the nucleotide sequence are sufficient to abolish type III secretion, eight mutant variants of the yopQ_1-10 signal were analyzed. None of the single codon substitutions abolished secretion, indicating that some, but not all, synonymous mutations of the yopQ_1-10 signal affect signaling (Ramamurthi, unpublished). Further, the data also suggest that recognition of YopQ as a secretion substrate cannot solely occur in an amino acid-mediated fashion, supporting at least in part the previously reported RNA signal hypothesis (4).

FIG. 7.

The yopQ_1-10 secretion signal does not tolerate synonymous mutations. (A) The secretion signal was altered by substituting nine nucleotides that changed the nucleotide sequence but not the amino acid specificity of eight codons. WT, wild type. Plasmids carrying the wild-type yopQ_1-10-npt (pKR63) and the altered but synonymous mRNA yopQ_1-10-npt mutant (pKR92) were transformed into Y. enterocolitica W22703 (wild-type strain) (B) or Y. enterocolitica VTL2 (yopD) (C). yopQ_1-10 codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

The length of a mutant yopQ signal determines its type III secretion phenotype.

One interpretation of the secretion defect observed for the frameshift mutations in the yopQ 1 to 10 signal is that some codon 2 changes abolish secretion signaling. The phenotypic defect of these mutations is restored when larger nucleotide segments (codons 1 to 15) are fused to the npt reporter, presumably because yopQ contains two functionally redundant secretion signals. In this model, the second signal must be located within yopQ codons 11 to 15. We sought to test this prediction by analyzing the phenotype of codon 2 and 10 mutations within the yopQ codon 1 to 15 signal. Replacement of codon 2 UUU with GGG within yopQ codons 1 to 15 greatly reduced the synthesis of YopQ_1-15-Npt (pKR147) and did not allow an accurate analysis of the secretion phenotype (Fig 8A). However, when pKR147 was transformed into the yopD mutant, YopQ_1-15-Npt was efficiently expressed and secreted (Fig. 8B). Replacement of codon 10 CGU (arginine) with GGU (glycine) abolished secretion signaling of yopQ codons 1 to 10 without affecting the function of the larger yopQ codon 1 to 15 signal in pKR148 (Fig. 8A and B). Together, these observations corroborate the notion that yopQ contains at least two functionally redundant secretion signals.

FIG. 8.

Mutations in codon 2 or in codon 10 do not abolish the function of the yopQ_1-15 signal. The yopQ_1-15 secretion signal (wild-type [WT] sequence in pKR23) was altered by introducing a codon 2 substitution (UUU to GGG, pKR147), or a codon 10 substitution (CGU to GGU, pKR148). Plasmids were transformed into Y. enterocolitica W22703 (wild-type strain) (A) or VTL2 (yopD mutant strain) (B), and yopQ_1-15 codon fusions to npt were analyzed for type III secretion as described in the legend to Fig. 1.

DISCUSSION

Since the discovery of type III secretion in Yersinia species about 10 years ago (31, 37), the answer to the deceptively simple question of how only 14 proteins are recognized as substrates for secretion by the type III machinery remains largely a mystery. Currently, there seems to be widespread agreement that the signal for secretion resides in either the 5′ or N-terminal end of Yops and that displacement of the signal to other regions of the protein is not tolerated. In addition, it is clear that the signal consists of approximately 15 codons or amino acids. The difficulty in assessing substrate recognition arises when one compares the presumed type III signal peptides from various secretion substrates, as there is no easily discernible similarity in sequence between them, certainly not at an amino acid level, and not obviously at the level of mRNA sequence. One of these conclusions is not shared by Lloyd and colleagues, who propose that an amphipathic peptide helix, resembling the targeting signal of mitochondrial precursor proteins (21), is responsible for the initiation of secretion substrates into the type III pathway (26).

Two apparently conflicting observations have clouded the issue. First, the observation that YopE can tolerate drastic changes in its mRNA sequence in the first 11 codons that preserve its amino acid sequence certainly suggests that the type III signal sequence is proteinaceous. In an effort to define the nature of this proteinaceous signal, it was suggested that a strictly alternating hydrophobic and polar amino acid sequence could promote the secretion of YopE (27). This was demonstrated by mutating three nucleotides in the mRNA encoding the native YopE secretion signal, which does not normally adhere to the alternating hydrophobic-polar residue rule, such that the first 8 codons consisted of serines and isoleucines. This mutant YopE was, indeed, a substrate for type III secretion (27). It should be noted that an alternating serine-isoleucine motif is not found in the wild-type secretion signal of Yop proteins, nor is there any Yop protein which contains a strictly alternating hydrophobic-polar residue motif in its secretion signal. Moreover, since all 20 common amino acids are either polar or hydrophobic, it should not be surprising that the secretion signal contains these residues. Furthermore, in the absence of a pattern to which residues of the secretion signal of all Yop proteins comply, it is difficult to understand how a simple mix of hydrophobic and polar amino acids could confer onto a polypeptide the ability to become a type III secretion substrate. Indeed, it could be argued that any number of cytosolic proteins that are not secreted contain a simple mix of hydrophobic and polar residues in their N termini, thereby making this motif an unlikely candidate for identification as the type III secretion signal.

In striking contrast to these observations as noted for YopE are the puzzling observations that have been made for the Yersinia proteins YopE, YopN, and YopQ, for the Pseudomonas syringae protein AvrPto, and for Xanthomonas campestri AvrBs2 (2, 4-6, 33). After introducing various frameshift mutations to the first 15 codons of these proteins, the proteins were all still substrates for type III secretion. These experiments, of course, suggested that the secretion signal may not be proteinaceous at all and could be at least partially contained within the yop mRNA. Recently, seemingly contradictory evidence was reported, in which frameshift mutations introduced to the first 12 codons of YopE did not promote the secretion of that protein in chaperone-independent manner (27). The data from the present study, however, seem to have reconciled the differences, since it is clear in the case of YopQ that frameshift mutations are tolerated only when at least 13 codons of the type III signal sequence are present. It is conceivable, therefore, that YopE also requires a minimum number of frameshifted codons to serve as a functional secretion signal and that 12 codons do not fulfill this requirement. In fact, our preliminary data corroborate this assertion (Ramamurthi, unpublished).

In an effort to further understand how proteins are recognized as substrates for type III secretion, the present study focused on characterizing the minimal secretion signal for yopQ, i.e., 30 nucleotides or 10 amino acids. The data show that mutations in the second codon of the secretion signal may abolish synthesis of YopQ and that mutations in the tenth codon may abolish secretion of YopQ without affecting its synthesis. This second phenotype was particularly curious, since it was the first time that we have observed any mutation in a type III signal sequence that uncoupled the synthesis of a polypeptide from its ability to be secreted. We hope to use this mutation in further studies that will help us understand what role the tenth codon plays in the recognition of YopQ as a type III secreted substrate.

The experiments depicted in Fig. 7 inextricably link the codons of yopQ mRNA to the ability of that protein to be synthesized and, ultimately, its ability to be secreted. By changing nine nucleotides at wobble positions of eight codons within the yopQ minimal secretion signal and conserving the amino acids that are encoded therein, it was possible to repress the synthesis of YopQ. Further analysis revealed that this repression was indeed mediated by the type III secretion system. YopD has previously been shown to be a negative regulator of synthesis of Yop proteins under circumstances that are not conducive to type III secretion (23, 42), including mutations in the type III signal sequence (3). Accordingly, when the construct containing the wobble nucleotide substitutions in the yopQ secretion signal was expressed in a yopD null background, its synthesis was restored. Although both wild-type and wobble secretion signals are predicted to synthesize the same YopQ_1-10-Npt polypeptide, the latter signal is nonfunctional and causes the encoded polypeptide to remain in the bacterial cytoplasm. Together these data suggest that all of the information necessary for secreting a protein by the type III pathway cannot be contained within the amino acid sequence of the secretion substrate.

The results depicted in Fig. 8 lead us to speculate that yop mRNAs harbors multiple, redundant secretion signals. We have been unable to identify either a common structure in yop mRNAs or a conserved nucleotide sequence that could be detected by BLAST searches. Nonetheless, we think one likely model for the initiation of proteins into the type III pathway may involve the binding of machinery components to discrete features of yop mRNA. This event may not even stall translation but could rather serve as a mechanism that directs yop transcripts to ribosomes that are sequestered within the type III pathway. Genetic experiments can now reveal the presumed (discrete) features of mRNAs, as yopQ codons 1 to 10 represent the minimal secretion signals whose function is sensitive to mutagenesis, unlike the yopQ codons 1 to 15 signal. A critical test for the RNA signal hypothesis is the identification of machinery components that recognize the secretion signal, presumably by binding to specific nucleotide segments of yop mRNA while failing to interact with mutant secretion signals.