Molecular Characterization of the Prototrophic Salmonella Mutants Defective for Intraepithelial Replication

Abstract

Three MudJ prototrophs demonstrated that intracellular replication is a Salmonella virulence trait (K. Y. Leung and B. B. Finlay, Proc. Natl. Acad. Sci. USA, 88:11470-11474, 1991). mutS and mutH are disrupted in mutants 3-11 and 12-23, and ssaQ is disrupted in mutant 17-21. Further analysis revealed that loss of Salmonella pathogenicity island 2 function underlies the intracellular replication defect of 3-11 and 17-21.

Three prototrophic Salmonella MudJ mutants defective for intraepithelial replication (Rep⁻) were instrumental in establishing several significant and enduring concepts regarding Salmonella virulence (15). These concepts include the following: intracellular replication is critical for Salmonella disease, discrete molecular determinants distinct from housekeeping genes mediate intracellular replication, and changes to the host cell's endocytic system, typified by Salmonella-induced filament formation (Sif), are associated with wild-type intracellular replication (10, 15). While the Rep⁻ mutants were all severely attenuated in mice and Sif negative (Sif⁻), the Sif⁻ phenotype was found to be unlinked to the MudJ insertional mutations (24). To gain insight into the molecular basis for the Rep⁻ phenotype and possible causes for the unlinked Sif⁻ phenotype, we genetically characterized the three Rep⁻ mutants identified by Leung and Finlay: 12-23, 3-11, and 17-21 (15). Strain 12-23 was originally found to be a rough lipopolysaccharide (LPS) variant refractile to P22 transduction (15). Subsequent analysis of 12-23 LPS revealed that a small amount of smooth LPS (9), known to serve as the P22 phage receptor, was present. Therefore, the MudJ-disrupted allele was backcrossed by P22 transduction (21) to a clean parental background, designated 12-23c. Inverse PCR (11) revealed that 12-23c contains a MudJ insertion within mutH (Table 1). This strain also produces wild-type smooth LPS (data not shown), and thus, the LPS defect of the original isolate did not directly result from MudJ insertion. Salmonella mutH encodes a component that functions along with MutS and MutL to mediate the long-patch DNA mismatch repair (MMR) system (8). While MutL and MutS are widely distributed among bacteria, MutH is found only in very close relatives of Escherichia coli and was selectively acquired by Salmonella (7).

TABLE 1.

Reagents used in this study

Strain or primer	Genotype	Source or reference	Comment
E. coli strains
TOP10	mcrA Δ(mcrCB-hsdSMR-mrr) (φ80lacZΔM15) ΔlacX74 deoR recA1 araD139 Δ (ara-leu)7679 galU galK rpsL endA1 nupG	Invitrogen	TA cloning vector
DH5α	supE44 Δ(lac)U169 hsdR17 recA1 endA1 gyrA96 thi-1 relA1 deoR (φ80lacZΔM15)		Vector propagation
DH5α λ pir	supE44 Δ(lac) U169 hsdR17 recA1 endA1 gyrA96 thi-1 relA1 deoR (φ80lacZΔM15) λ pir		Propagation of pCVD442 allelic exchange vector
SM10 λ pir	thi thr leu tonA lacY supE recA::RP4-2-Tc:Mu Kan pir R6K		Transfer of pCVD442
Serovar Typhimurium strains
SL1344	hisG46		Virulent wild type
ST218	hisD9953::MudJ	24	Virulent, Lac+, for CI determination
3-11	mutS::MudJ, hisD::MudJ	15	Rep−, Sif−, SPI2−, avirulent
3-11c1	mutS::MudJ	Present study	Rep+, Sif+
3-11c2	hisD::MudJ	Present study	Rep+, Sif+
12-23	mutH::MudJ	15	Rep−, Sif−, rough LPS, avirulent
12-23c	mutH::MudJ	Present study	Rep+, Sif+, virulent
17-21	ssaQ::MudJ	15	Rep−, Sif−, SPI2−, avirulent
17-21c	ssaQ::MudJ	Present study	Rep+, Sif+, SPI2+
ST216	ΔssaQ	Present study	Rep−, Sif−, SPI2−, avirulent
ST244	ΔssaV	11	Rep−, Sif−, SPI2−
69F3	ssaP::MudJ	11	Rep−, Sif−
ST319	ΔsseB	28	SPI2−, Sif−
Primer pairsa
Upstream oligonucleotide
Q1: 5′-ATG GAA ACT TTG CTG GAG ATA ATC GC-3′			Amplified downstream portion of ssaQ deletion allele
Downstream oligonucleotide
Q2-PmeI: 5′-GTTTAAAC TCA TTC GCT ATT CTT AAC ATA GAA TAT CTC-3′
Upstream oligonucleotide
Q3: 5′-GTTTAAAC GAT AAA CCT GAT AAG AAA AAT AAT ATG CGA AC-3′			Amplified upstream portion of ssaQ deletion allele
Downstream oligonucleotide
Q4-PmeI: 5′-TCA TGA AAA GCT CTG TAC CAA TTG CG-3′

^aIntroduced PmeI site is underlined.

Southern blot analysis (data not shown) and identification of MudJ lesions within strain 3-11 and backcrossed variants demonstrated that two independent insertions had occurred. As indicated in Table 2, one insertion occured within mutS (strain 3-11c1), while the other was within hisD (strain 3-11c2). The junction of the hisD::MudJ insertion in 3-11c2 was identical to that of the TT10289 hisD9953::MudJ donor strain (14) used to generate the mutant pool in the study by Leung and Finlay (15). Therefore, the hisD::MudJ allele was most likely introduced into 3-11 during MudJ mutagenesis, as commonly occurs when the recipient strain contains the histidine operon (14). In contrast, the mutS::MudJ insertion resulted from the desired random transposition event.

TABLE 2.

Intracellular replication mutants^g

Straina	Gene at MudJ junctionb	Rifrc	Ctxrd	CIe	Sif formationf
3-11c1	mutS; base 1749 of M18965	2.4 × 10−4	2.7	ND	Wild type
12-23c	mutH; base 568 of U16361	1.14 × 10−4	2.7	0.80 ± 0.28	Wild type
17-21c	ssaQ; base 7725 of YO9357	2.4 × 10−8	1.3	ND	Wild type
ST216 (ΔssaQ)	NA	ND	22.8	≤0.001	Negative

^aMutant designations are those of Leung and Finlay (15) with the suffix “c” appended to represent variants backcrossed to parental SL1344 backgrounds.

^bThe base number that abuts the left arm of MudJ and the accession number is given.

^cSpontaneous mutation frequency to rifampin (Rif) resistance (50 μg/ml) was calculated as the total number of CFU per milliliter divided by the number of Rif^r CFU per milliliter. The mutation frequency shown for 17-21c is in the range reported for wild-type SL1344 (24). The data shown are from an experiment representative of four, and in each experiment, several independent cultures were examined.

^dThe increase in cefotaxime (Ctx) resistance relative to the parental strain is given (11). The data shown are representative of three experiments performed.

^eThe CI for the mutant strain indicated was calculated as follows: the number of mutants was divided by the number of Lac⁺ reference variants obtained from spleen homogenates from mice coinfected 5 days earlier by the oral route. The values given are averages from 5 (ST216) or 12 (12-23c) different BALB/c mice with the error reflecting 1 standard deviation from the mean. When lactose differentiation was used, the lower limit of our assay was 0.001.

^fThe presence of Sif was determined within infected HeLa cells fluorescently stained for lysosomal membrane glycoproteins as previously described (11).

^gNA, not applicable; ND, not determined.

We reasoned that the mutS::MudJ mutation within 3-11c1 was probably the more relevant of the two insertions to the Rep⁻ phenotype for several reasons. Introduction of hisD9953::MudJ does not result in a new phenotype since Salmonella enterica serovar Typhimurium SL1344 is already a histidine auxotroph (Table 1). Unlike 3-11 (15), SL1344 hisD9953::MudJ is fully virulent. In fact, this strain is used as a wild-type reference in assays for virulence (Table 2) (24). Finally, two of the three Rep⁻ mutants contained MudJ within enzymes required for MMR (e.g., MutH and MutS), and this seemed to suggest a relationship between this repair system and the Rep⁻ mutant phenotypes. Therefore, 3-11c1, and not 3-11c2, was characterized further.

For continuity, in vitro phenotypes of 12-23c and 3-11c1 MMR mutants are reported together. Both are strong hypermutators acquiring forward mutations imparting rifampin resistance at a rate over 3 orders of magnitude greater than that of their parent (Table 2). This is as expected for MMR loss-of-function mutants (8). Neither MMR mutant displayed significant replication deficiencies (Fig. 1A) in Madin-Darby canine kidney (MDCK) cells when a gentamicin protection assay (26) was performed as was detailed earlier (24). They also proliferated within RAW 264.7 macrophage (Fig. 1B) when methods that were described previously were used (11). While a slight increase in cefotaxime resistance was displayed (Table 2), the 2.7-fold increase in survival relative to their parent strain was less then the fivefold increase considered to portend an intraepithelial replication deficit (11, 15) (ST216 [Table 2]). Both MMR mutants make Sif at wild-type frequencies, as was reported before (11) and is confirmed in this study (Table 2).

FIG. 1.

Intracellular proliferation of serovar Typhimurium, backcrossed, prototrophic replication mutants over 20 h within MDCK epithelial cells (A) and over 24 or 25.5 h in RAW 264.7 macrophages (B) as described previously (11). Error bars represent 1 standard deviation from the mean of triplicate time points. They are visible only if the error was larger then the area displaced by the data point symbol. Data shown are from a representative of three independent experiments.

Both 3-11 and 12-23 were found to be strongly attenuated for virulence (15), yet MMR is dispensable for Salmonella pathogenesis, since mutS mutants were fully virulent in mice (4, 26). It is therefore unlikely that the MMR lesions within 3-11 or 12-23 caused the attenuation of these Rep⁻ mutants. In fact, the secondary mutation resulting in the rough LPS of 12-23 rendered this strain serum sensitive, which is sufficient to account for its attenuation. Nevertheless, the contribution of the selectively acquired MutH MMR component to virulence has never been directly elucidated. 12-23c virulence was evaluated by determining competitive indices (CI). BALB/c mice were coinfected with equal numbers of CFU of 12-23c and a Lac+ wild-type variant (ST218 [Table 2]). A total volume of 200 μl of phosphate-buffered saline, containing 10⁶ CFU of both strains, was delivered orally by using an 18-gauge feeding needle. After 5 days, mice were killed and spleen homogenates were plated on media differential for lactose utilization (24a). As shown in Table 2, nearly identical numbers of CFU of 12-23c and the SL1344 parental strain were obtained. Therefore, the loss of MutH, like the loss of MutS (4, 26), does not diminish Salmonella virulence. This demonstrates that neither conserved nor S. enterica-specific MMR factors serve a role in systemic Salmonella virulence.

Collectively, it seems that both the Rep⁻ and Sif⁻ phenotypes and the murine attenuation displayed by 3-11 and 12-23 (10, 15) were the result of mutations other than the MudJ insertions. In the case of 3-11, we identified a defect that can account for all the mutant phenotypes displayed by 3-11 (below).

The 17-21c mutant was disrupted within ssaQ, located in the structural II region (Fig. 2A) of the Salmonella pathogenicity island 2 (SPI2) (20, 23). SPI2 encodes a type III secretion system (TTSS) required for intramacrophage survival (5, 12, 20), intraepithelial replication (5, 11), and Sif formation (1, 2, 11). The ssaQ::MudJ insertion occurred near the beginning of the ssaQ open reading frame at base position +100 disrupting the V34 codon of the 322-amino-acid protein. SsaQ shares significant similarity with the C terminus of YscQ of the Yersinia TTSS (13). As depicted in Fig. 2B, there is now a family of TTSS proteins that share this C-terminal surface presentation of antigens (SpoA; Pfam 01052) motif. The archetype is FliN, which is part of the flagellar C ring (17) located within the cytoplasmic membrane. SpoA proteins of virulence-associated TTSS are components of a surface-exposed (e.g., references 6 and 16) adapter complex predicted to allow the translocator to interact with the secretion apparatus (22). Therefore, we anticipate that SsaQ is also exported to the bacterial surface and is directly involved in effector traffic, but this theory awaits experimental demonstration.

FIG. 2.

(A) Schematic diagram of the SPI2 virulence locus. The inverted triangle depicts the locations of the MudJ insertion within 17-21c. The internal deletion of ssaQ in ST216 is schematically shown below the SPI2 diagram. (B) Alignment of the C-terminal portion of SsaQ with components of TTSS proteins possessing an SpoA motif. Proteins aligned are as follows: SsaQ (GenBank accession no. P74860), serovar Typhimurium FliN (GenBank accession no. P26419), serovar Typhimurium SPI1-encoded SpaO (GenBank accession no. AAC43938.1), Yersinia enterocolitica plasmid-encoded YscQ (GenBank accession no. AAD16827), Shigella flexneri plasmid-encoded Spa33 (GenBank accession no. 49846), and enteropathogenic E. coli SepQ (GenBank accession no. AAC38386.1).

As reported previously (24), we found that 17-21c still generates Sif at a wild-type frequency (Table 2). 17-21c also replicates within epithelial cells based on by its wild-type cefotaxime sensitivity (Table 2) and its proliferation within MDCK cells (Fig. 1A). However, a consequence of the MudJ insertion is revealed as a partial defect when 17-21c resides within the RAW 264.7 macrophage (Fig. 1B). 17-21c does not proliferate like wild-type strains, but a net reduction in bacterial numbers does not occur as it does with SPI2 loss-of-function mutants. For example, 69F3 (ssaP::MudJ) (11) is reduced by 75% within the macrophage when the earliest and latest time points are compared (Fig. 1B).

The wild-type Sif and intraepithelial replication phenotypes displayed by 17-21c contrast with the Sif⁻ (2, 11) and Rep⁻ phenotypes displayed by previously characterized SPI2 mutants (5, 11). Therefore, either SsaQ is not essential for SPI2 function or the particular ssaQ::MudJ lesion allows production of a truncated SsaQ able to mediate most SPI2 phenotypes. To discriminate between these two possible reasons for the nearly wild-type behavior of 17-21c, a ΔssaQ strain was generated. As detailed previously (11), an internal deletion allele was generated by joining regions upstream and downstream of ssaQ with PmeI sites introduced during PCR amplification of these regions. Oligonucleotides Q1 and Q2-PmeI amplified the upstream region, and oligonucleotides Q3-PmeI and Q4 amplified the downstream region (Table 1). The resultant deletion allele was used for positive-selection allelic exchange (11). The resulting ΔssaQ strain (ST216) has an in-frame, internal deletion that removes nearly the entire ssaQ open reading frame (Fig. 2A; Table 1).

The ΔssaQ strain was severely attenuated in mice as indicated by its CI (Table 2) and was completely unable to generate Sif within HeLa epithelial cells (Table 2). It is also defective for intraepithelial replication, as demonstrated by a 22.8-fold increase in cefotaxime resistance (Table 2) and a single doubling in MDCK cells, while 17-21c undergoes nearly three doublings over the same time course (Fig. 1A). The ΔssaQ strain also displays an intracellular survival deficiency within RAW 264.7, with nearly 75% of the internalized bacteria killed after 24 h (Fig. 1B). Thus, SsaQ is essential for SPI2 function, and it is likely that a truncated SsaQ variant is produced by 17-21c, which is sufficient for nearly normal SPI2 function. We speculate that an internal restart occurs within the ssaQ::MudJ allele, since a polar effect is not exerted upon the downstream ssaU (Suvarnapunya and Stein, unpublished data) and ssaR (2) genes that encode proteins essential for SPI2 function.

The implication of the SPI2 regulon by the ssaQ::MudJ insertion, albeit insufficient to abolish SPI2-mediated phenotypes such as Sif formation (11), prompted us to evaluate SPI2 function within the Rep⁻ mutants. To this end, we assessed SPI2 function by the export of the SPI2 translocon component, SseB (e.g., reference 19) to the bacterial surface. Equal numbers of whole bacteria were extracted with n-hexadecane. This selectively removes exported SseB by using methodologies that were detailed recently (28). As shown in Fig. 3, 17-21 and 3-11 are SPI2 secretion mutants. The ΔssaQ strain also fails to export SseB to the bacterial surface. However, 17-21c does export SseB, confirming that the particular ssaQ::MudJ lesion within 17-21c does not abolish SPI2 function. In fact, we consistently observed that n-hexadecane extracts from 17-21c contain more SseB than was obtained from the parent. This may suggest that certain alterations in SsaQ, as documented for the related FliN (17), may also influence the length of type III assembled organelles.

FIG. 3.

Immunoblot analysis of SPI2 TTSS function. Bacteria were grown in MgM at pH 5.0 and identical numbers were extracted with n-hexadecane to selectively remove the SseB translocon component exported to the bacterial surface by SPI2. Proteins were resolved on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (10.5% Tris-tricine) gels and were immunoblotted with anti-SseB mouse polyclonal antibody as detailed elsewhere (28). SL1344 is the wild-type parent. Controls for nonspecificity are ΔssaV (ST244), deficient for SPI2 secretion, and ΔsseB (ST319). The apparent molecular mass of standards is given at right.

In contrast to the other original Rep⁻ mutants, 12-23 appears to possess a functional SPI2 secretion system at least to the point of SseB export. However, SPI2 mutations have been identified that abolish SPI2 translocation function without interfering with SseB export (19). It is unlikely that 12-23 is defective for production of SifA, which is also required for Sif formation (24). This is because the associated hyperreplication phenotype displayed by SifA loss-of-function mutants (24) is not displayed by 12-23 (data not shown). Alternatively, we have recently identified several Salmonella loci distinct from SPI2 and SifA that also result in an abolition of Sif formation (Suvarnapunya and Stein, unpublished). It is therefore conceivable that 12-23 sustained a lesion within one of these loci. We also cannot exclude the possibility that the unidentified secondary lesion that rendered 12-23 rough may also underlie the Sif⁻ and/or Rep⁻ phenotype. Additional studies are required should the cause for the 12-23 Rep⁻ Sif⁻ phenotypes be elucidated.

The molecular characterization of the prototrophic replication mutants of Leung and Finlay (15) yields the unanticipated conclusion that mutations secondary to the MudJ insertion underlie phenotypes attributed to the original isolates. The means by which unlinked, secondary mutations were acquired by 12-23 and 3-11 is readily suggested by their strong hypermutability combined with the potent selection exerted during cefotaxime enrichment. However, the possibility that a point mutation was acquired prior to MudJ insertion cannot be eliminated.

In the case of 17-21, possible causes for its SPI2 secretion deficit (Fig. 3) are harder to tender, especially with the finding that the actual ssaQ::MudJ lesion is not sufficient to prevent SPI2 export of SseB (Fig. 3). Nevertheless, it appears that either a spontaneous secondary mutation occurred within SPI2 or that the original MudJ insertion exerted a SPI2 loss-of-function effect in a manner not retained when backcrossed.

This study confirms and extends several previous findings. The virulence characterization of 12-23c verified (4, 27) that MMR is dispensable for Salmonella virulence, despite the contribution of MMR to the virulence of other intracellular pathogens (18). Nevertheless, DNA damage does occur during the course of Salmonella infection. The pleiotropic DNA repair system regulator, RecA, is required for Salmonella intramacrophage proliferation and for full virulence (3), and the base excision repair system may also facilitate repair of macrophage-mediated damage to DNA (25).

The in vitro phenotypes (10, 15) (Fig. 1 and 3) and in vivo behavior (15) of 3-11 and 17-21 are consistent with those ultimately attributed to SPI2 mutants (1, 2, 5, 11, 12, 20, 23). While it cannot be excluded that additional secondary mutations unrelated to SPI2 are present in these Rep⁻ mutants, their behavior in our hands is indistinguishable from that of targeted SPI2 deletion or insertion mutants (11, 25, 27). In retrospect, it appears that the identification and descriptions of these particular Rep⁻ mutants were the first characterization of SPI2 function.

The present finding that cefotaxime selected for hypermutable strains also highlights an important consideration when evaluating mutants obtained after multiple enrichment steps, as occurs in cefotaxime selection assays. The realization that point mutants may be obtained in this manner could be a useful strategy for molecular analysis via second-site suppressor screens.