The Majority of Typhoid Toxin-Positive Salmonella Serovars Encode ArtB, an Alternate Binding Subunit

While previous reports had suggested that the typhoid toxin (TT) could potentially use ArtB as an alternate binding subunit, this was thought to play a minor role in the evolution and biology of the toxin. In this study, we establish that both TT genes and artB are widespread among Salmonella enterica subsp. enterica, suggesting that TT likely plays a broader role in Salmonella virulence that extends beyond its proposed role in typhoid fever.

82% identity) and pltB (65% to 68% identity) (see Table S1 in the supplemental material). Therefore, these serovars/strains were regarded as encoding putative pltA and pltB but were not considered encoding all TT genes for the analyses reported below. Proportions of serovars encoding all TT genes were significantly different among S. enterica phylogenetic clades (P , 0.001; Fisher's exact test); clade B had the highest proportion of TT-encoding serovars (80 of 83 serovars), followed by section Typhi (14 of 21 serovars), while TT genes were detected in significantly fewer clade A1 and A2 serovars (P , 0.001 for all pairwise comparisons) with just 17% (8 of 40 serovars) and 4% (3 of 76 serovars) of serovars in these clades encoding TT genes, respectively (Fig. 1B). These results demonstrate a strong phylogenetic relationship between the presence of TT genes and Salmonella clades.
Next, we assessed the genomic location of the TT islet using WGS data for representative serovars (see Materials and Methods for a description of how these were selected) from each TT-positive S. enterica subsp. enterica clade that had a closed genome available. The TT islet was previously reported to be located within Salmonella pathogenicity island (SPI) 11 (5,28); however, the genomic location of SPI-11, and therefore the TT genes, was unknown. While the size of SPI-11 was variable among the serovars examined here, the TT islet was consistently located approximately 3 kb from the terminus of SPI-11. For all genomes evaluated ( , except S. Typhi, SPI-11 was within 5 to 7 kb from phoP-phoQ, which encodes a two-component regulatory system that has been shown previously to be the primary system regulating transcription of TT genes (30). However, in S. Typhi CT18, SPI-11 was located .550 kb upstream of phoP-phoQ. These data suggest that, in general, both the location of SPI-11 and the location of the TT islet within SPI-11 are relatively conserved.
While the overall phylogeny suggests multiple acquisition and/or deletion events of TT genes, the most likely scenario appears to be that TT genes were present in the S. enterica ancestor, and subsequent deletion events in major clades as well as possibly additional independent (re)acquisitions of TT genes (e.g., in some serovars in clade A1 and A2 and section Typhi) occurred.
The majority of TT-encoding serovars also encode artB, which is located on an islet within an operon encoding genes associated with resistance to antimicrobial peptides. Several groups previously established that multiple TT-encoding serovars also encode a homologous binding subunit, artB (28,31,32). Therefore, we also assessed the presence of artB to determine the frequency that TT and ArtB are coencoded in a given serovar. Among all 235 serovars examined, artB was detected in a total of 88 serovars (Fig. 1A). Proportions of artB-encoding serovars were significantly different across the phylogenetic clades (P , 0.001; Fisher's exact test) (Fig. 1C). FIG 1 Legend (Continued) likelihood analysis was computed using a general time-reversible model with gamma distribution in RAxML (66). Bootstrap values for branchpoints between subspecies are shown (based on 100 bootstrap repetitions). Branches are colored to represent the clade of the isolate as reported previously (29) and the detection of TT genes (cdtB, pltA, and pltB; shown in light blue) and artB (shown in purple) among S. enterica isolates using 85% nucleotide identity and 90% query coverage as criteria for detection. S. enterica subsp. arizonae was used to root the phylogeny. Histograms summarizing the proportion of serovars in each clade for which all TT genes (cdtB, pltA, and pltB) (B) or artB (C) was detected with blastn (i.e., the additional 6 S. enterica subsp. enterica serovars having putative pltA and pltB were not included in this analysis); note that clade C is not shown because TT genes and artB were not detected in these serovars. *, P , 0.05; **, P , 0.001; ***, P ,0.0001, FDR corrected for pairwise comparisons of Fisher's exact tests. (D) Comparison of the artB-islet in S. Javiana CFSAN001992 and S. Typhi CT18; ORFs and genes are named according to annotations from NCBI. The artB islet is denoted by the purple horizontal bar. ORFs that are at least 100 nucleotides (nt) in length are shown in gray (those ,100-nt long were classified as pseudogenes and are shown in white), while genes shown in gold represent sap operon genes that flank the artB islet; artB is shown in dark purple. Islets are not drawn to scale. (E) Maximum likelihood reconstruction of phylogeny based on 63,994 core SNPs among 41 S. Javiana isolates and a representative isolate of S. Mississippi (used to root the phylogeny). The S. Javiana strain (FSL S5-0395) used in phenotypic experiments is shown in bold. Circles at nodes represent bootstrap values .90 (based on output from 1,000 bootstrap repetitions). Assembly names are colored to reflect the country from which the isolate originated; the isolation sources (human clinical shown in orange, environmental shown in yellow) as well as the nucleotide identity of TT genes (light blue) and artB (purple) are shown as colored bars. Roman numerals are used to depict the clades of the S. Javiana isolates. Serovars in clades B (66 of 83) and D (3 of 3) and section Typhi (12 of 21) had significantly higher proportions of artB-positive serovars than clade A1 (4 of 48) and A2 serovars (3 of 79) (see Fig. 1C for false-discovery rate [FDR] corrected P values for pairwise comparisons). Among the 105 serovars for which a full TT islet was detected (i.e., excluding the three clade D serovars, which encoded putative pltA and pltB), artB was detected in 85 (i.e., 81% of TT-encoding serovars also encoded artB).
Given the relative conservation of artB among TT-positive serovars, we next assessed whether the location of the artB islet was conserved among genomes representing the different S. enterica subsp. enterica phylogenetic clades (A1, A2, and B and section Typhi). Although the genomic locations and the numbers of ORFs identified differed slightly ( Fig. 1D; see also Table S2), we found that in all cases, the artB islet was inserted between sapB and sapC (Fig. 1D). The GC content of the artB islet was considerably lower (average, 42.2%; range, 41.2% to 44.7%) than the genome-wide GC content (average, 52.2%; range, 52.1% to 52.3%), suggesting that this islet was likely acquired via horizontal transfer from another source (Table S2). In addition to artB itself, the artB islet includes (i) between 4 to 7 ORFs with no known or predicted functions (Table S2) and (ii) an artA pseudogene, which was identified in all islets examined. The artA pseudogene contains multiple mutations leading to premature stop codons, with the most 59 premature stop codon leading to a predicted peptide with just 4 amino acid (aa) residues. For one serovar, S. Milwaukee, an additional ORF annotated as an IS3 family transposase, was detected in the middle of the artB islet. Together, these results suggested a common site-specific mode of acquisition of the artB islet as well as a selective advantage for maintenance and/or acquisition of the artB islet among most TT-positive serovars.
For serovar S. Javiana, TT genes are highly conserved, while the presence of intact artB is clade dependent. Given that 96% of clade B serovars examined here were TT positive and the majority of studies examining the TT had been conducted with S. Typhi, we characterized the role of artB in the clade B nontyphoidal serovar S. Javiana, which is the most prevalent TT-positive serovar among human clinical salmonellosis cases in the United States (33). Among a collection of 41 S. Javiana isolates selected to represent a range of genotypes (different single nucleotide polymorphism [SNP] clusters), geographical sources, and dates of isolation (see Table S3), we found that sequences of cdtB, pltA, and pltB were highly conserved (99% to 100% nucleotide identity). In contrast, while artB was conserved for most of the isolates (100% identity for 34/41 isolates; clade III) (Fig. 1E), we also detected two small clades for which artB was more variable, including (i) five isolates encoding an artB allele containing a 46-nucleotide deletion resulting in a premature stop codon (clade I) (Fig. 1E), and (ii) two isolates for which artB was not detected with blast (clade II) (Fig. 1E); visual inspection of the sap operon further confirmed the absence of the artB islet at this locus as well as a lack of raw sequence reads mapping to artB. These results suggested that even though artB is also relatively conserved among S. Javiana isolates, loss of ArtB function occurred independently in two clades of S. Javiana, due to either artB deletion or artB pseudogene formation.
artB and pltB are induced when S. Javiana is grown under nutrient-limiting conditions, such as those representing an intracellular host environment. Expression of pltB in S. Typhi was previously shown to occur when cells were grown under conditions that mimic those encountered by intracellular Salmonella (30,32). Therefore, we compared pltB, cdtB, and artB transcript abundances of S. Javiana grown in Luria-Bertani (LB) broth (pH 7) and N-salts minimal medium containing 8 mM Mg 21 (pH 7). When S. Javiana was grown in N-salts minimal medium at pH 7, transcript abundances of pltB, cdtB, and artB were on average 135-, 364-, and 44-fold higher, respectively, than abundances in S. Javiana grown in LB broth ( Fig. 2A). When S. Javiana was grown in N-salts minimal medium acidified to pH 5.8 to simulate intracellular conditions (34), transcript abundances of cdtB and artB were higher (on average 382-and 86-fold compared to growth in LB broth) ( Fig. 2A), while pltB transcript abundances were approximately 2-fold lower than abundances obtained during growth in N-minimal medium at pH 7. Collectively, these results suggested that artB and pltB transcription is induced similarly to that of cdtB when S. Javiana is grown in minimal medium where Mg 21 is limiting. However, the relative transcript abundances (relative to growth in LB broth) of pltB were higher at neutral pH than at acidic pH, while both cdtB and artB had higher relative transcript abundances at acidic pH, suggesting that pH may serve as an environmental cue for modulating expression of pltB and artB.
Binding subunits composed of PltB or ArtB result in active TT. Previous gene knockout experiments suggested that S. Javiana requires either ArtB or PltB to produce active TT, as strains with deletion of either artB or pltB still showed production of active toxin (16). To biochemically confirm this, we expressed various combinations of toxin subunits in a heterologous host and then coincubated lysates containing these toxins with human intestinal epithelial cells (HIEC-6) to assess toxin activity. The accumulation of cells in the G 2 /M phase continues to be the gold standard for assessing TT activity (14,17), as cells that have an activated DNA damage response (DDR) will accumulate at this cell cycle checkpoint (35). Therefore, we analyzed the cell cycle progression of HIEC-6 cells treated with lysates containing holotoxins with PltB or ArtB to confirm the activity of both toxin types. Treatment of HIEC-6 cells with lysates containing CdtB, PltA, and either PltB or ArtB resulted in a significantly higher proportion of HIEC-6 cells in the G 2 /M phase (PltB as binding subunit, P = 0.037, or ArtB as binding subunit, P = 0.006; both compared to cells treated with lysates containing empty vector) (Fig. 2B). There was no significant difference in the proportion of cells in the G 2 /M phase for HIEC-6 cells treated with TT that used ArtB versus PltB as the binding subunit (P = 0.6983), supporting that holotoxins using either monomer for the binding subunit are active. To confirm that the accumulation of cells in G 2 /M phase after treatment with these toxins was due to the activation of a DDR, we also used immunofluorescence (IF) staining to detect two proteins involved in the DDR, namely, phosphorylated histone H2AX (gH2AX), and 53 binding protein 1 (53BP1). Treatment with lysates containing CdtB, PltA, and either PltB or ArtB as the binding subunit resulted in DDR activation in ;78% of HIEC-6 cells ( Fig. 2C and D), confirming that both holotoxins are active. Importantly, treatment with individual toxin subunits did not result in DDR activation among treated cells ( Fig. 2C and D). Together, these data confirm that the TT encoded by S. Javiana can use either PltB or ArtB as the binding subunit.
In silico modeling suggests that heteropentameric binding subunits of ArtB and PltB may occur. Given that both ArtB and PltB can form active TT and that transcript abundances of artB and pltB are significantly higher when S. Javiana is grown under nutrient-limiting conditions, such as those that might be encountered by intracellular Salmonella, we hypothesized that these subunits likely compete for inclusion in the TT, resulting in toxin binding subunits composed of homo-and heteropentamers. The structures of the TT binding subunits containing homopentamers of PltB (from S. Typhi [18]) or ArtB (from S. Typhimurium DT104 [26]) have been resolved experimentally and were used here to model and predict the energies of interactions between monomers in binding subunits composed of various ratios of ArtB and PltB using the predominant amino acid sequence types for S. Javiana ( Fig. 1E and 3A). While alignment of ArtB from S. Typhimurium DT104 with ArtB from clade III S. Javiana (Fig. 1E) showed only 74% amino acid identity, interface areas and the DG solvation energy gain of the ArtB complex formation for S. Typhimurium DT104 ArtB and S. Javiana ArtB were similar (see Fig. S1), supporting the modeling strategy used. ArtB was shown to interact with more specificity and stronger hydrophobicity with itself (ArtB-ArtB DG average, 211.5 kcal/M) than with PltB (ArtB-PltB DG average, 210.5 kcal/M) (Fig. 3A); however, the theoretical interaction of ArtB-PltB was predicted to be more favorable than PltB-PltB (DG average, 27.5 kcal/M) (Fig. 3A). These data suggested that ArtB and PltB monomers may interact to form heteropentameric binding subunits, which we next tested experimentally.
In vitro interactions between ArtB and PltB monomers and copurification of ArtB and PltB from TTs suggest that heteropentameric binding subunits may occur in vitro. We first assessed the possibility of ArtB-PltB interactions using an adenylate cyclase two-hybrid system in Escherichia coli BTH101. We specifically used a high-throughput screening approach (Fig. 3B) to test for possible interactions of different domains for each toxin component (CdtB, PltA, PltB, and ArtB) including (i) fulllength proteins, (ii) polypeptides without the signal peptide sequence, and (iii) the oligomerization domains predicted from the TT structure (18). Using this technique, we confirmed PltB-PltB, ArtB-ArtB, and CdtB-PltA interactions for both interacting domains and full-length toxin subunits (Fig. 3C), but we were unable to confirm a PltB-ArtB interaction, despite testing interactions for various combinations of interacting domains and full-length PltB and ArtB ( Fig. 3C and Fig. S1).
In a heterologous host two-hybrid system assay, issues arising from protein fusion stability and folding as well as a requirement of basal levels of cAMP to initiate the positive-feedback induction loop may interfere with the ability of the system to detect some protein-protein interactions. Hence, we also constructed a cAMP-null S. Javiana strain (DcyaA) to assess possible interactions in Salmonella. When DcyaA S. Javiana was grown in M63 broth (supplemented with 0.02 mM cAMP, a level that induced basal fusion protein expression but did not support growth of negative controls), we found marginally significant (P , 0.05, uncorrected P value) PltB-PltB, ArtB-ArtB, and ArtB-PltB interactions ( Fig. 3D and Table S4); however, upon correcting for multiple comparisons, only one of the ArtB-PltB interactions was considered statistically significant at an a of 0.1 (corrected P value = 0.0973) ( Fig. 3D and Table S4). These results suggested that (B) Schematic of the method used to screen two-hybrid system clones. In the example shown, DNA fragments encoding the PltB polypeptide or C-terminal oligomerization region, as predicted from TT structure, were cloned at the 59 or 39 ends of regions encoding adenylate cyclase domain T18 or T25. All T18 or T25 clones were pooled into "banks" that were cotransformed into the cAMP-null (DcyaA) E. coli BTH101. Screening was performed by growing the transformants on LB agar supplemented with IPTG and X-Gal. Plasmid DNA was isolated from clones with positive interactions and was retransformed into E. coli BTH101 to exclude constitutively active CAP* mutants. Clones with confirmed interactions were submitted for Sanger sequencing to determine the identity of the interacting domains. (C) E. coli BTH101 two-hybrid system strains grown on LB agar supplemented with IPTG and X-Gal. Clones that showed (Continued on next page) while PltB-PltB and ArtB-ArtB interactions were more readily detected using the twohybrid system, there was some evidence of ArtB-PltB interactions; we therefore conducted additional experiments to determine whether both ArtB and PltB subunits could be incorporated into a heteropentameric configuration of the TT binding subunit.
Previous studies have suggested that the TT binding subunit exists as a stable pentamer (18). Therefore, we also used tandem affinity purification (TAP) of different combinations of CdtB-His, PltA-Strep, PltB-3Â-Flag, and ArtB-c-Myc overexpressed in E. coli BTH101 (Fig. 4A) to assess whether holotoxins can be formed with heteropentameric binding subunits of PltB and ArtB. Purification of His-tagged CdtB followed by purification of PltB-3Â-Flag (Fig. 4B) revealed that ArtB was consistently (results of three independent experiments) (see Fig. S2) copurified with PltB, albeit at lower levels than PltB, in both pulldown steps, supporting that PltB and ArtB can form heteropentameric binding subunits. Additional immunoblotting to control for the possibility of recognition of ArtB-c-Myc by anti-Flag antibodies (used to detect PltB-3Â-Flag) confirmed the specificity of this approach (Fig. 4C and Fig. S2). Taken together, both the weak interaction detected using the two-hybrid system and the copurification of ArtB-c-Myc from PltB-3Â-Flag-containing toxins suggested that binding subunits composed of heteropentamers of ArtB and PltB may occur, further supporting a scenario in which ArtB and PltB monomers may compete for inclusion in the TT binding subunit.
PltB and ArtB may compete for inclusion in the holotoxin. Collectively, our data suggested that artB and pltB are both induced when Salmonella is grown in medium mimicking intracellular conditions during Salmonella infection of epithelial cells, and both ArtB and PltB can be used to form biologically active toxins. Therefore, we hypothesized that ArtB and PltB monomers likely compete for inclusion in the holotoxin binding subunit. We thus used a modified competition assay to assess whether addition of PltB to lysates containing CdtB, PltA, and ArtB resulted in replacement of ArtB in the holotoxin, and vice versa. Cell lysate from E. coli BTH101 overexpressing CdtB, PltA, and either ArtB or PltB was challenged with various concentrations of PltBor ArtB-containing lysate, respectively, prior to TAP and subsequent Western blot detection. Upon challenging cell lysates containing CdtB, PltA, and ArtB with increasing amounts of lysates containing PltB, we found that PltB replaced ArtB, thereby reducing the total amount of ArtB bound in the holotoxin ( Fig. 4D and Fig. S2). This result was not reciprocal, however, as addition of ArtB only marginally reduced the amount of PltB in the holotoxin ( Fig. 4D and Fig. S2). Together, this suggests that PltB and ArtB may compete for inclusion in the binding subunit.

DISCUSSION
Multiple investigations into the role of ArtB, a homolog of PltB, have suggested that ArtB can form a functional TT binding subunit (16,26,32). Our results suggest a role for ArtB in the diversification of the TT in NTS serovars, as (i) both toxin subunits are coencoded by serovars spanning multiple lineages of S. enterica subsp. enterica, (ii) both artB and pltB are expressed under conditions mimicking the intracellular environment, with differences in transcript abundances at neutral and acidic pH potentially favoring assembly of TT with different binding subunits (ArtB or PltB) under different environmental conditions, and (iii) ArtB and PltB may compete for inclusion in the binding subunit. Constructs with the C terminus are denoted with "ct." *, P , 0.05 versus the negative control before (blue) and after (pink) Dunnett's test for multiple comparisons adjustment; the PltB T25 -ArtB T18 interaction was only marginally significant after multiple comparison corrections (corrected P value = 0.0973).
TT genes are considerably more widespread among S. enterica subsp. enterica serovars than previously thought. Owing to the initial discovery of cdtB in S. Typhi strain CT18 in 2000, the majority of research efforts have, and continue to, focus primarily on the potential contributions of this toxin to typhoid fever (17,18,24,30,32,(36)(37)(38)(39)(40). Several studies have, however, demonstrated that TT is encoded by many nontyphoidal serovars (27)(28)(29)31), suggesting that this toxin is not the sole contributing factor to the acute onset of typhoid fever, an observation which was recently confirmed in a human challenge study that demonstrated that onset of acute typhoid fever occurred even in subjects challenged with a TT-null strain (24). Importantly, our  Fig. S2 in the supplemental material. (C) Immunoblots demonstrating the specificity of the pulldown assays and antibody detection used for TAP (additional blots for additional replicates can be found in Fig. S2). (D) Results of the binding subunit exchange assay. PltB-3Â-Flag was added at various concentrations to lysates containing CdtB-His, PltA-Flag, and ArtB-c-Myc (left), or ArtB-c-Myc was added to lysates containing CdtB-His, PltA-Flag, and PltB-3Â-Flag (right). Holotoxins were purified by pulling down with a-His antibodies to detect CdtB-His, and subunits were detected using tag-specific antibodies (PltA-Flag, CdtB-His, and PltB-3Â-Flag). Results from one representative experiment are shown; the assay was performed as three independent experiments. Images for additional replicates can be found in Fig. S2. results support that TT genes are common in clade B serovars and are rare among clade A1 and A2 serovars. As the number of WGS data available for NTS serovars continues to increase, the number of serovars that are known to encode TT will likely increase. Regardless, our results suggest that studies examining the role of the TT in NTS serovars are warranted given the conservation of this toxin in nearly all clade B serovars, as reported here and in previous studies (28,29).
Maintenance of the artB islet in TT-encoding serovars representing multiple lineages of S. enterica subsp. enterica suggests an important role for ArtB. Previous studies had suggested that some serovars encode both TT genes and artB (27,31,32). Our analyses demonstrate that 81% of TT-positive serovars also encode artB. More importantly, our results indicate that the location of the artB islet within the sap operon is consistent across representative serovars in clades A1, A2, and B and section Typhi. sap operon genes have been shown to play a role in resistance to host-derived antimicrobial peptides for S. Typhimurium (41), uropathogenic E. coli (42), and Haemophilus ducreyi (43). In 2010, Rodas et al. demonstrated that integration of the genomic islet GICT18/1 (putative artB islet) into S. Typhi sapB resulted in an increased sensitivity to the antimicrobial peptide protamine (44). Furthermore, the authors showed that excision of GICT18/1 restored transcription of sapD and sapF and restored resistance to protamine to levels similar to that observed for S. Typhimurium, which lacks the GICT18/1 islet (44). As insertion of the artB islet is expected to result in sensitization to antimicrobial peptides, these results suggest that there is likely an advantage for the selective maintenance of artB by most TT-encoding serovars.
This begs the question of why the majority of TT-positive Salmonella serovars would evolve to encode multiple binding subunits, especially if acquisition of one (ArtB) may result in an increased susceptibility to host-derived antimicrobial peptides. Our current understanding is that PltB and ArtB have adapted to bind to different cell surfaces (26,32), thereby expanding the variety of cell types that TT can bind to. While PltB has been reported to preferentially bind Neu5Ac-terminated glycans (26), ArtB can bind both Neu5Ac-and Neu5Gc-terminated glycans (26), suggesting that PltB and ArtB allow TT to bind to different cell types. Furthermore, Lee et al. recently established that specific amino acid residues in PltB produced by S. Javiana (called Javiana toxin in that study) were associated with toxin binding to intestinal epithelial cells but not to endothelial cells of brain arterioles (45), therefore supporting the idea that PltB produced by S. Typhi and S. Javiana may have evolved to enable binding to different tissues. Given that single amino acid changes have been found to significantly dampen binding affinity to cell surface glycans (45), collectively, this suggests that encoding two different subunits may provide an evolutionary advantage, as it likely enables TT-positive Salmonella to affect a wider range of cell types, tissues, and possibly hosts. Therefore, future studies examining the sequence variants of ArtB and PltB among a broader range of Salmonella serovars, as well as the binding affinities for these variants for a panel of different cells representing different tissues and hosts, will provide a more complete picture of the potential role of the TT binding subunit in facilitating tissue and or host tropism.
Furthermore, there is also some evidence to support the role of TT as an immunotoxin. Previous reports showed that injection of TT containing PltB or ArtB (called PltC in S. Typhi) into C57BL/6 mice resulted in significantly lower levels of total white blood cells, lymphocytes, and monocytes, than injection with phosphate-buffered saline (PBS) (32). In a human challenge study, subjects that were administered TT-null S. Typhi had higher fold changes of IgA and IgG against O9:lipopolysaccharide (LPS) than subjects that were administered wild-type TT (although not explicitly addressed in the study, this strain would also encode artB) (24). Therefore, TT may play a role in dampening the adaptive immune response, which may have implications for reinfection with TT-producing Salmonella. While these studies suggest that TTs containing ArtB and/or PltB may act as an immunotoxin, additional experimentation to assess the expression of and activity of ArtB-and PltB-containing TTs will be essential for addressing TT's potential role in modulating the immune response during infection with typhoidal and nontyphoidal TT-producing serovars.
While copurification of ArtB and PltB suggest the formation of heteropentameric forms of the TT binding subunit, homopentamers appear to be the predominant form of the binding subunit. Overall, our data, combined with previous studies, suggest that multiple mechanisms likely fine-tune ArtB and PltB expression and their assembly into different TT forms (i.e., toxins having binding subunits containing just PltB or ArtB or a mix of both PltB and ArtB). At the transcriptional level, both artB and pltB are induced when cells are grown under low-Mg 21 conditions (this study and reference 30). However, our results also indicate that pH may serve as an additional environmental cue modulating expression of artB and TT genes. For example, artB is more highly induced under acidified conditions (pH 5.8), while induction of pltB is higher at neutral pH. In S. Typhi, artB and pltB are regulated by different two-component systems; artB (called pltC in S. Typhi) expression is primarily regulated by SsrA-SsrB (32), while TT genes are primarily regulated by PhoP-PhoQ (30). However, both SsrA-SsrB and PhoP-PhoQ are induced under low-Mg 21 conditions (34,46). As SsrB DNA binding is enhanced at low pH (47), while binding of PhoP (predicted regulator of TT genes) is not pH sensitive, our study provides additional support for a role of pH in fine-tuning transcription of TT genes. While this also suggests that regulation of artB and pltB transcription is conserved across S. enterica subspecies additional studies are needed to confirm the regulatory networks that control artB and pltB transcription in a broader range of NTS species.
At the level of toxin assembly, our experimental data suggest that the formation of heteropentameric TT binding subunits may occur, at least under some environmental conditions. The possible occurrence of PltB-ArtB interactions was confirmed by (i) twohybrid system experiments, which showed evidence for some ArtB-PltB interactions for S. Javiana, and (ii) pulldown experiments, which showed consistent copurification of ArtB with PltB. However, our data also indicate that homopentamers may be the favored configuration of the binding subunit, as (i) PltB-PltB and ArtB-ArtB interactions were readily detected even in a heterologous host, while weak evidence for ArtB-PltB interactions (significant at a = 0.1) was only detected for some constructs in S. Javiana grown in broth, (ii) Western blot detection of PltB and ArtB after TAP suggested that low levels of ArtB and PltB are present when TTs are coimmunoprecipitated in a manner targeting the other binding units (i.e., ArtB or PltB, respectively), and (iii) competition assays demonstrate that PltB efficiently replaces ArtB in the holotoxin at neutral pH, which was in contrast to the results of the in silico modeling suggesting that ArtB-PltB interactions were more favorable than PltB-PltB interactions. Similar to our experimental findings, a recent study suggested that TT produced by S. Typhi is predominantly composed of either PltB or ArtB homopentamers, although the possibility of heteropentameric configurations could not be excluded in that study (32). As both the study described here and a previous study (32) overexpressed toxin subunits in a heterologous host, future studies examining toxin expression and assembly using untagged subunits (as tagged subunits were used in this study) under native conditions in Salmonella will be important for understanding the putative role of heteropentameric configurations of the TT binding subunit. Furthermore, differences in the amino acid sequences of PltA and ArtB encoded by different serovars (such as S. Typhi and S. Javiana [33]) may also need to be considered for studying the assembly of this toxin among a broader range of S. enterica serovars.
It is important to note that the experimental methods used here cannot distinguish configurations of the TT binding subunit with differing ratios of PltB and ArtB; additional structural evidence will be necessary to confirm the formation of heteropentameric configurations of the binding subunit and to assess the ratios of holotoxins with homo-and heteropentameric configurations of the binding subunit. In addition, whether the formation of ArtB-PltB heteropentameric binding subunits represents transitional forms of the TT or whether they have specific functional or virulence benefits also remains to be elucidated. One possibility is that differences in transcriptional regulation of artB and TT genes (cdtB, pltA, and pltB) resulting from changes in environmental cues (e.g., acidification of the Salmonella-containing vacuole [SCV]) could lead to a transitory state in which TTs with heteropentameric binding subunits represent an intermediate step that occurs as the TT transitions into a different homopentameric binding subunit (i.e., ArtB only or PltB only) (Fig. 5). Assessment of TT assembly under different conditions such as those mimicking intracellular conditions (e.g., acidified SCV) will therefore be important for procuring a more complete understanding of the potential complementary roles that ArtB and PltB play in the function of the TT. Additional experiments examining the delicate balance of transcriptional, as well as possible translational and posttranslational, regulatory mechanisms driving the formation of TT will be important for understanding the in vivo occurrence and any potential biological relevance of TTs with different binding subunits.
Overall, our data contribute to a growing body of evidence that suggests that ArtB plays an important and currently underappreciated role in the activity of the TT. As additional serovars are found to encode both ArtB and PltB, a more complete understanding of the TT will require a diversified approach to better assess the role that TT plays in infection with nontyphoidal serovars.

MATERIALS AND METHODS
Bacterial strains, plasmids, primers, and media. Bacterial strains used in this study are listed in Table S5 in the supplemental material, vectors and recombinant constructs are listed in Table S6, and primers are listed in Table S7. Bacterial strains were routinely grown in Difco Luria-Bertani (LB) broth, pH 7, containing 5 g/liter NaCl (Becton, Dickinson [BD], Franklin Lakes, NJ). For two-hybrid system interactions, E. coli and S. Javiana strains were grown in M63 medium with maltose as the sole carbon source (48). N-salts minimal medium (pH 5.8 or 7) containing 8 mM MgSO 4 (34) was used for experiments assessing transcript abundances of TT genes and artB. Unless otherwise indicated, ampicillin and kanamycin were used at 100 mg/ml and 50 mg/ml, respectively, in complex medium (LB broth) and 50 mg/ml and 25 mg/ml, respectively, in chemically defined media (M9 or N-salts minimal medium).
Cloning and expression of toxin proteins in E. coli. Toxin subunits were expressed in E. coli BTH101 cells (50) using (i) pltB-3Â-Flag-artB-c-Myc cloned into the high-copy-number pUT18 vector or (ii) cdtB-His-pltA-Flag and cdtB-His-pltA-Strep, each cloned into the low-copy-number pKNT25 plasmid. All PCRs were performed using the high-fidelity polymerase Q5 (New England Biolabs [NEB], Ipswich, MA) and the primers listed in Table S7. All constructs were designed to contain identical ribosome binding sites (AGGAGG) 5 to 7 bases upstream of the ATG start codons. The DNA sequences of all constructs were confirmed with Sanger sequencing.
For construction of pUT18::pltB-3Â-Flag-artB-c-Myc, PCR products representing (i) pltB-3Â-Flag and  (30), which may lead to the preferential inclusion of PltB in the TT binding subunit. Conversely, at slightly acidic pH, SsrB induces transcription of artB (32) in a pH-sensitive manner (47), which may favor preferential inclusion of ArtB in the TT binding subunit. The copurification of ArtB and PltB suggests the potential formation of heteropentameric forms of the binding subunit, although a role for these heteropentameric subunits will require further investigation.
Assessment of two-hybrid system interactions. Interactions of CdtB, PltA, PltB, and ArtB subunits were first assessed using the bacterial adenylate cyclase two-hybrid (BACTH) system in E. coli BTH101 cells (50). The full-length polypeptides of the toxin subunits (including signal peptides) were fused to the N termini of T25 or T18 subunits in pKNT25 and pUT18, respectively (Fig. 3C). To ensure cytoplasmic CyaA activity, we also cloned the coding regions of CdtB, PltA, PltB, and ArtB lacking the signal peptide into BACTH plasmids for fusion at either the N terminus (pKNT25 and pUT18) or C terminus (pKT25 and pUT18C). Regions encoding the C-terminal interacting domains of PltA, PltB, and ArtB, as predicted from the TT structure (18,26), were also cloned into pKNT25, pUT18, pKT25, and pUT18C plasmids (Tables S6  and S7). As 324 theoretical combinations exist, we used a high-throughput screening method. Screening was performed by cotransforming T18 and T25 vectors into E. coli BTH101 cells, selecting on LB agar plus 1% (wt/vol) glucose with ampicillin and kanamycin, and subsequently substreaking colonies onto LB agar with ampicillin and kanamycin plates containing 1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) and 40 mg/ml of 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-Gal) for blue-white screening. To ensure that positive interactions were not false positives arising from cAMP-independent CAP* spontaneous mutations, plasmid DNA was purified from blue colonies and retransformed into E. coli BTH101. Plasmids were isolated from clones that resulted in blue colonies after the second transformation and were analyzed by Sanger sequencing to determine the identity of the interacting domains.
Construction of DcyaA S. Javiana. Deletion of cyaA was performed to enable screening for maltose utilization resulting from an interaction of T18-and T25-fused proteins. Whole-genome sequence analysis identified a single full-length cyaA in the S. Javiana strain used in this study (FSL S5-0395). Construction of a DcyaA S. Javiana strain was performed using the l-Red recombinase system as described previously (16). Plasmids and primers used to generate the DcyaA strain are listed in Tables S6 and S7, respectively. The in-frame deletion was confirmed by Sanger sequencing. The mutant strain was also phenotypically confirmed by a lack of growth in M63 medium containing maltose as the sole carbon source; growth was restored upon addition of 0.05 mM cAMP.
Expression of the two-hybrid system constructs in M63 minimal medium. Plasmids containing ArtB and PltB constructs were transformed into the DcyaA S. Javiana strain via electroporation. Transformed cells were selected on LB agar plates supplemented with ampicillin and kanamycin. Overnight cultures (12 to 14 h) were grown shaking at 30°C in LB broth with ampicillin and kanamycin. Subsequently, M63 broth containing ampicillin, kanamycin, and 0.02 mM cAMP was inoculated with an overnight culture (diluted 1:100), followed by incubation with shaking at 30°C for 24 h. The addition of 0.02 mM cAMP (a concentration that does not support the growth of negative controls) (Fig. 3D) was necessary for the basal expression of the fusion proteins in S. Javiana. Growth was assessed as optical density at 600 nm (OD 600 ), which was measured after 24 h with a BioTek synergy plate reader (BioTek Instruments, Inc., Winooski, VT). Each growth assay was performed as three independent experiments.
Molecular modeling of different forms of the TT binding subunit. In silico modeling of TT subunits was done using the Phyre2 server (51). Construction of the TT with different ratios of PltB and ArtB was performed using the MatchMaker tool in Chimera software (52). The stability and biological relevance of homo-and heteropentamer formation were assessed by calculating free energies from individual subunit-subunit interactions within the pentamer, using the PDBe Protein Interfaces, Surfaces and Assemblies (PDBe PISA) server (53).
Tandem affinity purification. Beveled flasks containing 250 ml of LB broth with ampicillin and kanamycin were inoculated with overnight cultures (12 to 14 h, 1:500 dilution) of E. coli BTH101 harboring plasmids pAG37 and pAG43, which express PltB-3Â-Flag ArtB-c-Myc and CdtB-His PltA-Strep, respectively. Cells were grown shaking at 200 rpm for 4.5 h at 37°C, followed by induction with IPTG and cAMP (both 1 mM final concentrations) for an additional 1.5 h. Cells were collected by centrifugation and stored at 280°C. Cells were thawed on ice, resuspended in 2 ml of NTA buffer (50 mM NaH 2 PO 4 [pH 8.0], 0.3 M NaCl) containing 2 mg lysozyme, and incubated at 37°C for 30 min. Cell lysis was achieved with three rounds of freeze-thaw cycles (submersion in liquid nitrogen and incubation at 37°C) followed by sonication using a Branson Sonifier 250 sonicator (80% duty cycle, 7 output control) for 30 s on ice (performed twice). Cell debris was removed by centrifugation at 15,000 Â g for 10 min. For CdtB-His purification, 50 ml of Dynabeads (Thermo Fisher Scientific; Waltham, MA) was washed twice with 500 ml of cold NTA buffer in preparation for TAP. Cell lysates were added to beads, followed by incubation overnight (12 to 14 h) at 4°C in a tube rotator. Beads were collected on a magnetic stand and were washed three times at 4°C with 500 ml of cold NTA buffer with incubation periods of 10 min. Proteins were eluted using 300 ml of NTA buffer containing 250 mM imidazole. Eluted proteins were then dialyzed against TBS buffer (50 mM Tris-Cl, 150 mM NaCl, pH 7.5) containing 10% (vol/vol) glycerol. PltB-3Â-Flag was pulled down using Flag magnetic beads (EMD Millipore) according to the manufacturer's instructions. Proteins were eluted from beads by boiling in the presence of 100 ml of 1Â SDS-loading dye. PltB-3Â-Flag and ArtB-c-Myc were detected in subfractions using Western blot analyses performed with rabbit anti-Flag (Sigma number [no.] F7425, diluted 1:500) and rabbit anti-c-Myc (Sigma no. SAB4301136, diluted 1:4,000) antibodies with goat anti-rabbit-horseradish peroxidase (HRP) antibodies (Thermo Fisher Scientific no. 65-6120, diluted 1:3,000). Horseradish peroxidase-conjugated secondary antibodies were detected using Clarity Western ECL substrate (Bio-Rad). Blots were visualized using the Bio-Rad ChemiDoc MP imaging system.
Western blot detection of proteins. Protein samples were resolved on 4% to 20% Mini-Protean TGX precast protein SDS-PAGE gels (Bio-Rad Laboratories; Hercules, CA) and blotted on polyvinylidene difluoride (PVDF) membranes using the Trans-Blot Turbo transfer system (Bio-Rad). Membranes were incubated in TBS buffer containing 0.1% (vol/vol) Tween 20 (TTBS) and 5% (wt/vol) blocking reagent (Bio-Rad) with gentle shaking at room temperature for 30 min. Primary antibodies were added as described above with mouse anti-strep (Millipore Sigma no. 71590-M, 1:1,000 dilution) and mouse anti-His antibodies (Santa Cruz Biotechnology, Dallas, TX; no. SC-53073, 1:400 dilution) diluted in TTBS with 0.5% (wt/vol) blocking reagent, followed by incubation with gentle shaking at room temperature for 12 to 14 h. Membranes were washed three times with TTBS, followed by incubation (2 h at room temperature with gentle shaking) with secondary antibodies (goat anti-rabbit-HRP [Thermo Fisher Scientific no. 65-6120, diluted 1:3,000] and donkey anti-mouse-Alexa 647 [Thermo Fisher Scientific no. A-31571, diluted 1:1,000]) diluted in TTBS with 0.5% (vol/vol) blocking reagent. Membranes were washed twice in TTBS and once in TBS before Western blot detection. Blots were visualized using the Bio-Rad ChemiDoc MP imaging system.
Binding subunit exchange assay. Protein expression and cell lysis of E. coli BTH101 strains (FSL G4-0035 to G4-0038) were performed as described above for TAP with cell pellets resuspended in PBS containing 10% (vol/vol) glycerol. Final total protein concentrations were determined spectrophotometrically (Nanodrop 2000c) (54,55). The PltB-3Â-Flag-containing lysate was added at various protein concentrations to 100 mg of total cell lysate containing CdtB-His, PltA-Flag, and ArtB-c-Myc for a final reaction volume of 50 ml of PBS with 10% (vol/vol) glycerol. The reaction mixture was incubated at 37°C for 30 min, followed by 2 h of incubation at 4°C to ensure equilibrium. A 200-ml aliquot of NTA buffer was added, and protein complexes were purified using Ni-nitrilotriacetic acid (NTA) beads, as described above. The assay was also performed with lysates containing CdtB-His, PltA-Flag, and PltB-3Â-Flag as target and ArtB-c-Myc as the competitor. All proteins were detected by Western blotting using antibodies that recognize the corresponding protein tag (i.e., Flag, anti-His, and anti-c-Myc); three independent experiments were performed.
Intoxication of HIEC-6 cells with crude lysates of toxin components expressed in E. coli BTH101 cells. Fresh Opti-MEM (containing 10 ng/ml recombinant epidermal growth factor [r-EGF] and 10% [vol/ vol] FBS) medium supplemented with 10 mg/ml gentamicin was added to HIEC-6 cells grown to confluence on 12-mm coverslips (Thermo Fisher Scientific) in 24-well plates (Corning Inc., Corning, NY). Lysates containing various combinations of toxin subunits were then added to HIEC-6 cells at a final concentration of 400 mg total protein per ml, and the HIEC-6 cells were subsequently incubated at 37°C with 4.5% CO 2 . Immunofluorescence (IF) detection of DDR foci and cell cycle analyses were performed at 24 6 2 h after inoculation. Three independent experiments were performed for both cell cycle and IF detection experiments.
Cell cycle analyses. Staining with propidium iodide for cell cycle analysis determination was performed as described previously (16,33). DNA content (for cell cycle analysis) was assessed using the FACSAria flow cytometer (BD). Gating was performed to exclude multiplets as described previously (16,57).
Whole-genome sequence data analyses for 235 Salmonella serovars. Assemblies for the 235 S. enterica subsp. enterica serovars and five other S. enterica subspecies were downloaded from NCBI (see Data Set S1 for S. enterica data set). For the S. enterica data set, isolates were selected to include serovars commonly associated with human clinical disease (7) and those representing novel clades of S. enterica (29). Sequences from a convenience sample of 40 S. Javiana isolates, representing 28 unique SNP clusters were downloaded from the NCBI Pathogen Detection browser (https://www.ncbi.nlm.nih.gov/ pathogens/isolates/#/search/) (see Table S3 for details); WGS data for S. Javiana FSL S5-0395 was also included. S. Mississippi isolate SRR1960042 was included as an outgroup for phylogenetic analyses (Table S3). For these data, Illumina adapters from sequence reads were trimmed, and low-quality bases were removed using Trimmomatic 0.33 with default settings (58). Determination of the quality of trimmed reads was performed using FastQC v0.11.7 (59). De novo assembly of all genomes was performed using SPAdes 3.6.0 (60). To assess the qualities of draft genomes, QUAST 3.2 (61) was used, followed by BBmap 35.49 (62) and SAMtools 1.3.1 (63) to calculate average coverage. Serotypes were confirmed using SISTR (64). kSNP3 was used to identify core SNPs in (i) the S. enterica data set and (ii) 40 S. Javiana genomes and strain FSL S5-0395; a k-mer size of 19 was used for both data sets, as determined using kSNP3's Kchooser function (65). Phylogenetic trees were constructed with RAxML (66) using a general time-reversible model with gamma-distributed sites constructed from either 100 (S. enterica data set) or 1,000 (S. Javiana data set) bootstrap repetitions. iTOL v4 was used for editing of phylogenetic trees (67,68).
Detection of TT genes and artAB. The presence of artA, artB, and TT genes was queried for all isolates using nucleotide BLAST (blastn) version 2.3.0 with default settings to query artA, artB, pltA, cdtB, and pltB sequences from S. Javiana strain CFSAN001992 (69); an E value of 1e220 was used for querying S. Javiana genomes, and an E value of 1e210 was used for querying the 235 serovars. In addition, closed genomes were downloaded to visually assess the location of TT genes and artB in serovars representing different phylogenetic clades (see Table S2 for assembly and strain information). Genomes were selected if they (i) represented commonly used strains from serovars associated previously with outbreaks of human clinical illness (e.g., CT18 for S. Typhi, AKU_12601 for S. Paratyphi A, CFSAN001992 for S. Javiana, and ATCC BAA-1673 for S. Poona) or (ii) were among the few closed genomes available for serovars that were positive for both TT and artB (serovars S. Goldcoast [clade A1] and S. Milwaukee [clade A2]); serovar S. Indiana was selected as an additional serovar for assessing the genomic location of toxin genes in section Typhi serovars, as S. Paratyphi A and S. Typhi represented a unique phylogenetic clade in our core genome analysis (Fig. 1A). In addition, to confirm the absence of artB in select S. Javiana assemblies, mapping of raw reads to artB from CFSAN001992 was performed using BWA version 0.7.17.
RT-qPCR quantification of artB and pltB transcript levels. Overnight cultures (16 to 18 h) of FSL S5-0395 grown in LB broth were subcultured 1:1,000 into LB or N-salts minimal medium (either pH 7 or pH 5.8), and subcultured samples were grown at 37°C with shaking at 200 rpm until cells reached midexponential phase (3 h for LB broth, 5 h for N-salts minimal medium). RNA was stabilized with RNA protect (Qiagen) and was collected using the RNeasy kit (Qiagen). DNA was depleted with Ambion DNase I (Life Technologies), and DNA depletion was confirmed with quantitative PCR (qPCR); a cycle threshold (C T ) of 34 for rpoB in DNase-treated RNA samples was used for judging successful DNA depletion. cDNA libraries were prepared with the Superscript reverse transcription kit (Thermo Fisher), according to manufacturer's instructions. qPCR was performed with SYBR green 2Â master mix (Applied Biosystems) in a reaction mixture containing 0.4 mM each primer (see Table S7) and 1 ml of cDNA (approximately 15 ng of cDNA) as the template. Fold expression was calculated by raising the DDC T to the power of the efficiency calculated for each primer pair (70). Results are the averages from three independent experiments, each performed in technical duplicates.
Statistical analyses. Statistical differences were assessed using R studio version 3.4.2. using packages lme4 (71)

ACKNOWLEDGMENTS
We thank Pete Chandrangsu for his guidance with the in silico modeling and the staff at the Cornell Statistical Consulting Unit and the Biotechnology Resource Center Imaging Facility for their guidance with statistical analyses and flow cytometry data collection, respectively.
A.G., A.S.H., M.W., and R.A.C. designed the study. A.G., A.S.H., A.R.C., and R.A.C. performed experimental, bioinformatical, and/or statistical analyses. All authors wrote and assisted with the revision of the manuscript.
Confocal microscopy data were acquired through the Cornell University Biotechnology Resource Center, with NIH 1S10RR025502 funding for the shared Zeiss LSM 710 Confocal Microscope. R.A.C. was partially supported by USDA 2020-67034-31905.