The Hcp-like protein HilE inhibits homodimerization and DNA binding of the virulence-associated transcriptional regulator HilD in Salmonella

HilD is an AraC-like transcriptional regulator that plays a central role in Salmonella virulence. HilD controls the expression of the genes within the Salmonella pathogenicity island 1 (SPI-1) and of several genes located outside SPI-1, which are mainly required for Salmonella invasion of host cells. The expression, amount, and activity of HilD are tightly controlled by the activities of several factors. The HilE protein represses the expression of the SPI-1 genes through its interaction with HilD; however, the mechanism by which HilE affects HilD is unknown. In this study, we used genetic and biochemical assays revealing how HilE controls the transcriptional activity of HilD. We found that HilD needs to assemble in homodimers to induce expression of its target genes. Our results further indicated that HilE individually interacts with each the central and the C-terminal HilD regions, mediating dimerization and DNA binding, respectively. We also observed that these interactions consistently inhibit HilD dimerization and DNA binding. Interestingly, a computational analysis revealed that HilE shares sequence and structural similarities with Hcp proteins, which act as structural components of type 6 secretion systems in Gram-negative bacteria. In conclusion, our results uncover the molecular mechanism by which the Hcp-like protein HilE controls dimerization and DNA binding of the virulence-promoting transcriptional regulator HilD. Our findings may indicate that HilE's activity represents a functional adaptation during the evolution of Salmonella pathogenicity.

ber of the AraC/XylS transcriptional regulators family, directly induces the expression of HilA, which in turn activates the expression of InvF; HilA and InvF activate the expression of all components of the T3SS-1, their chaperones, and effector proteins (4,14). HilD induces the expression of HilA directly or through a positive feed-forward loop that it forms with HilC and RtsA (15,16). HilC and RtsA are AraC-like transcriptional regulators that bind the same DNA sequence recognized by HilD; HilC is encoded within SPI-1, whereas RtsA is encoded in another island. HilD also controls the expression of many other virulence genes located outside SPI-1, including acquired and ancestral genes, directly, or indirectly through HilA, InvF, or other regulators (12,(17)(18)(19)(20)(21)(22)(23)(24)(25). In agreement with its role as a master transcriptional regulator for a high number of genes, the expression, concentration, and activity of HilD is tightly controlled. Transcription of hilD is positively autoregulated (15,26) and is repressed by H-NS (27), whereas its translation is repressed by CsrA (28). In addition, Fur and FliZ control HilD at post-translational level (29,30); the Lon protease degrades HilD, thus mediating its concentration (31,32); and the CpxR/A two-component system decreases the stability of HilD through Lon-dependent and independent mechanisms (33). Furthermore, several compounds affect HilD: propionate and bile salts decrease its stability (9,10); butyrate and oleate negatively affect its regulatory activity, whereas acetate and formate enhance this (11, 34 -36); and L-arabinose affects HilD expression at post-transcriptional level (37).
The activity of HilD is also negatively controlled by HilE through protein-protein interaction (38). However, the specific effect of this interaction on HilD has remained unknown. The hilE gene resides in a region of the Salmonella chromosome similar to a pathogenicity island (38), supporting that it was acquired by horizontal transfer. The expression of hilE is positively regulated by the PhoPQ and PhoBR two-component systems, as well as by FimZ and LeuO (39,40), whereas this is negatively regulated by Mlc and the small RNA IsmR (41,42), which provides additional inputs controlling the activity of HilD.
Here we show that HilE negatively affects dimerization and DNA binding of HilD, by interacting with both the central and C-terminal regions of HilD, which mediate dimerization and DNA binding, respectively. Therefore, our results demonstrate how HilE regulates HilD activity. Additionally, our results revealed that HilE shares sequence and structural similarities with proteins called Hcp (hemolysin-coregulated protein), which are structural components of type 6 secretion systems in Gram-negative bacteria, supporting the hypothesis that HilE was adapted to act as an important regulatory protein during the Salmonella pathogenicity evolution.

HilE interacts with the central region and with the C-terminal region of HilD
To investigate how HilE negatively affects the activity of HilD, we analyzed the interaction of HilE with different regions of HilD by using the LexA-based genetic system for heterodimerization, which is similar to a two-hybrid system to analyze protein-protein interactions (43,44). It is important to remark that the interaction between HilE and HilD was previously demonstrated with this system (38). In the LexA-based genetic system for heterodimerization, the WT LexA DNAbinding domain (LexA DBDwt ) and a derivative mutant of this (LexA DBDmut ) are expressed from two different plasmids in the Escherichia coli SU202 reporter strain, which carries a chromosomal sulA-lacZ transcriptional fusion containing a LexA hybrid operator (43,44). LexA DBDwt and LexA DBDmut cannot affect the expression of sulA-lacZ; however, when they are fused to proteins that interact between them, an active dimer of LexA DBDwt and LexA DBDmut is formed that is able to bind the hybrid LexA operator on sulA-lacZ and thus to repress the expression of this fusion. Fusion proteins of LexA DBDwt with full-length or distinct regions of HilD and of LexA DBDmut with full-length HilE were generated and assessed. The LexA DBDwt -HilD fusions were named according to the amino acid position for the ends of the HilD fragment carried (Fig. 1A). As expected, the combination LexA DBDwt -HilD 1-309 (full-length HilD) ϩ LexA DBDmut -HilE (full-length HilE), but not LexA DBDwt ϩ LexA DBDmut or LexA DBDwt -HilD ϩ LexA DBDmut , mediated repression of sulA-lacZ (Fig. 1B), further confirming the interaction between HilD and HilE. Interestingly, the fusion proteins LexA DBDwt -HilD 1-220 or LexA DBDwt -HilD 221-309 , in combination with LexA DBDmut -HilE, also repressed the expression of sulA-lacZ (Fig. 1B), indicating that HilE interacts with the regions of HilD spanning amino acids 1-220 and 221-309. In contrast, the combination LexA DBDwt -HilD 1-130 ϩ LexA DBDmut -HilE did not repress the expression of sulA-lacZ, supporting that HilE does not interact with the N-terminal region of HilD from amino acids 1 to 130. However, the lack of interaction between LexA DBDwt -HilD 1-130 and LexA DBDmut -HilE could be due to an incorrect folding of LexA DBDwt -HilD 1-130 . As expected, the combination LexA DBDwt -HilD 1-130 ϩ LexA DBDmut , LexA DBDwt -HilD 1-220 ϩ LexA DBDmut , or LexA DBDwt -HilD 221-309 ϩ LexA DBDmut , used as a negative control, did not affect the expression of sulA-lacZ (Fig. 1B). Western blotting analysis showed expression signals for all the LexA DBDwt -HilD fusion proteins tested (Fig. 1C). To note, the LexA DBDwt -HilD 221-309 fusion repressed the sulA-lacZ fusion, in combination with LexA DBDmut -HilE (Fig. 1B), even when its expression level was lower than that of the LexA DBDwt -HilD 1-309 and LexA DBDwt -HilD 1-220 fusions (Fig. 1C), indicating that differences in the amount of the LexA DBDwt -HilD fusion proteins assessed did not affect the results of these assays. In agreement with a previous report (45), expression of LexA DBD was not detected in our Western blotting analysis (Fig. 1C). These results support that HilE interacts with regions spanning amino acids 130 -220 and 221-309 of HilD.
Pulldown assays were performed to confirm the HilE interactions with HilD. For this, we purified HilE fused to the Trx protein and a His 6 tag (Trx-His-HilE) as described under "Experimental procedures." The Trx-His-HilE fusion protein repressed the SPI-1-mediated protein secretion profile of the WT S. Typhimurium strain ( Fig. 2A), which supports that it is able to negatively affect HilD. Trx-His-HilE was used as the bait protein in the pulldown assays; first, it was immobilized on Ni-NTA resin, and then whole-cell extracts containing HilE controls dimerization and DNA binding of HilD either LexA DBD or LexA DBDwt -HilD 1-309 , LexA DBDwt -HilD 1-130, LexA DBDwt -HilD 1-220 , LexA DBDwt -HilD 221-309 prey proteins were loaded to the Ni-NTA resin carrying Trx-His-HilE. As expected, Trx-His-HilE captured the LexA DBDwt -HilD  and LexA DBDwt -HilD 1-220 prey proteins, but not LexA DBDwt -HilD 1-130 or LexA DBDwt (Fig. 2B), confirming the interaction of HilE with full-length HilD and the central region of HilD. LexA DBDwt -HilD 221-309 showed interaction with HilE in the LexA-based genetic system (Fig. 1B); however, in the pulldown assays, this interaction was no evident (Fig. 2B), which could be explained by the low level of expression showed by LexA DBDwt -HilD 221-309 (Fig. 1C). As a control, parallel pulldown assays were performed using Trx-His as the bait protein, which was unable to capture any of the LexA DBDwt -HilD prey proteins tested (Fig. 2C). Taken together, these results indicate that HilE binds to both the central region and the C-terminal region of HilD.

HilE controls dimerization and DNA binding of HilD
bility of HilD-Myc was drastically increased in the ⌬hilD ⌬lon mutant (Fig. 3, A and B). These results demonstrate that the interaction with HilE does not affect the stability of HilD.

HilD acts as a dimer, and its central region mediates the interaction itself
Several AraC-like transcriptional regulators act as dimers (46 -49). Therefore, we thought that the interaction of HilE with the central region of HilD could affect dimerization of HilD. To investigate this possibility, we first analyzed whether HilD indeed dimerizes by using the LexA-based genetic system for homodimerization (43,44). In this case, LexA DBDwt is expressed from a plasmid in the E. coli SU101 reporter strain, which carries a chromosomal sulA-lacZ transcriptional fusion containing the LexA WT operator (43,44). LexA DBDwt does not affect the expression of sulA-lacZ; however, when it is fused to a protein that itself interacts, an active dimer of LexA DBDwt is formed, which is able to bind the WT LexA operator on sulA-lacZ and thus represses the expression of this fusion. The same LexA DBDwt -HilD fusion proteins used before in the heterodimerization assay were now tested in the homodimerization assay. LexA DBDwt alone and the fusion protein LexA DBDwt -H-NS, whose dimerization capacity has been tested before (50), were used as negative and positive controls, respectively. Both LexA DBDwt -H-NS and LexA DBDwt -HilD 1-309 , but not LexA DBDwt , repressed the expression of sulA-lacZ (Fig. 4A), indicating that HilD dimerizes. A Western blotting analysis showed expression signals for the LexA DBDwt -HilD 1-309 and LexA DBDwt -H-NS fusion proteins (Fig. 4B). To confirm the dimerization of HilD, the size of the maltose-binding protein (MBP)-HilD fusion, purified in native conditions, was analyzed by gel-filtration chromatography. The MBP-HilD fusion protein eluted from the gel-filtration chromatography as a ϳ178-kDa product (Fig. S1), which corresponds to a size similar to that expected for a dimer of this protein.

Figure 2. Pulldown assays showing the interaction between HilE and HilD.
A, secretion of the SPI-1-encoded proteins SipA, SipB, SipC, and SipD was tested in the WT S. Typhimurium strain and its isogenic ⌬hilD mutant, as well as in the WT S. Typhimurium strain carrying the pET32-HilE plasmid expressing Trx-His-HilE or the pET32b(ϩ) vector expressing Trx-His. The bacterial cultures were grown for 9 h in LB at 37°C, in the presence (ϩ) or absence (Ϫ) of 1 mM IPTG to induce or not the expression of Trx-His-HilE or Trx-His. Supernatants of the cultures were analyzed in SDS-PAGE at 12%. FliC is a flagellar protein whose secretion is SPI-1-independent. B and C, bait proteins Trx-His-HilE (B) or Trx-His (C) immobilized on Ni-NTA resin were incubated with whole-cell extracts containing the LexA DBDwt , LexA DBDwt -HilD 1-309 , LexA DBDwt -HilD 1-130 , LexA DBDwt -HilD 1-220 , or LexA DBDwt -HilD 221-309 prey proteins. After washing, the proteins captured by the Trx-His-HilE or Trx-His bait proteins were analyzed by Western blotting using polyclonal anti-LexA antibodies. The Trx-His-HilE or Trx-His bait proteins were also detected with monoclonal anti-His 6 antibodies. MW, protein molecular weight standards (Precision Plus Protein TM ; Bio-Rad). The arrowheads indicate the expected bands. Asterisks indicate bands showing cross-reaction with the polyclonal anti-LexA antibodies.

HilE controls dimerization and DNA binding of HilD
LexA DBDwt -HilD 1-130 , LexA DBDwt -HilD 221-309 , and the negative control LexA DBDwt did not affect the expression of sulA-lacZ ( Fig. 4C), supporting that the regions between amino acids 1 and 130 and amino acids 221 and 309 are not involved in the dimerization of HilD. These results suggest that the dimerization of HilD seems to be mediated by the region comprising amino acids 130 -220. To further investigate this asseveration, we constructed and tested the LexA DBDwt -HilD 130 -309 fusion protein, containing an extended N-terminal HilD region with respect to LexA DBDwt -HilD 221-309. As could be expected, LexA DBDwt -HilD 130 -309 was able to represses the expression of sulA-lacZ (Fig. 4C), indicating that it interacts itself. Together, these results indicate that HilD dimerizes through its central region ranging from amino acids 130 to 220.
Next, we sought to determine whether dimerization is required for the regulatory activity of HilD. For this, the ability to induce the expression of a hilA-cat transcriptional fusion of LexA DBDwt -HilD 1-309 , LexA DBDwt -HilD 221-309 , and LexA DBDwt -HilD 130 -309 , which carry the DNA-binding domain of HilD, was tested in an E. coli K-12 strain. Because hilA is a gene directly and positively controlled by HilD (4, 51) and HilD is not present in E. coli K-12. As a negative control, LexA DBDwt was also assessed; additionally, as a reference for the expression of hilA-cat in the presence or absence of HilD, the activity of this fusion was also determined in the WT S. Typhmurium strain and its isogenic ⌬hilD mutant. The expression of hilA-cat was induced in the presence of LexA DBDwt -HilD 1-309 or LexA DBDwt -HilD 130 -309 , which show dimerization, but it was not induced by LexA DBDwt -HilD 221-309 that does not dimerize (Fig. 5). As expected, the expression of hilAcat was also induced in the WT S. Typhimurium strain, but not in the ⌬hilD mutant or in the presence of LexA DBDwt (Fig. 5). These results support that HilD needs to form dimers to induce the expression of target genes. To confirm this, we generated and analyzed the chimeric protein LexA DBDwt -LZ-HilD 221-309 , which is a fusion of LexA DBDwt , the leucine zipper (LZ) motif of the GCN4 transcriptional factor of Saccharomyces cerevisiae, and the DNA-binding domain of HilD (amino acids 221-309) (Fig. 6A). The LZ motif has been used before as a dimerization module (52). The ability of the LexA DBDwt -LZ-HilD 221-309 fusion protein to undergo dimerization and to induce the expression of hilA was assessed and compared with that of the LexA DBDwt -HilD 221-309 and LexA DBDwt -HilD 1-309 fusion proteins. Both LexA DBDwt -LZ-HilD 221-309 and LexA DBDwt -HilD 1-309 , but not LexA DBDwt -HilD 221-309 , repressed the expression of sulA-lacZ in the E. coli SU101 reporter strain (Fig. 6B), indicating that LexA DBDwt -LZ-HilD 221-309 dimerizes through the heterologous LZ motif. Notably, LexA DBDwt -LZ-

HilE controls dimerization and DNA binding of HilD
HilD 221-309 and LexA DBDwt -HilD 1-309 , but not LexA DBDwt -HilD 221-309 , induced the expression of the hilA-cat fusion in E. coli K-12 (Fig. 5), showing that the LZ motif generates dimers of the DNA-binding domain of HilD, which are able to induce expression of target genes. Overall, these results demonstrate that HilD dimerizes through its central region spanning amino acids 130 -220, which is essential for the regulatory activity of HilD.

HilE negatively affects the dimerization of HilD
Based on the results described above, we now analyzed whether HilE affects the dimerization of HilD. For this, the LexA DBDwt -HilD 1-309 fusion protein, which carries the dimerization domain of HilD, was tested for homodimerization in the presence of the pA6 -HilE1 plasmid expressing HilE or in the presence of the pMPM-A6⍀ vector. Expression of HilE from pA6 -HilE1 drastically reduced the SPI-1-mediated protein secretion profile of the WT S. Typhimurium strain (Fig. 7A), supporting that the amount of HilE reached from this plasmid can negatively affect HilD. The effect of HilE on LexA DBDwt -LZ-HilD 221-309 and LexA DBDwt -H-NS, which dimerize through the LZ motif and the H-NS dimerization domain, respectively, was also assessed as negative controls. Interestingly, HilE affected the repression of sulA-lacZ mediated by the dimerization of LexA DBDwt -HilD 1-309 ; in contrast, it did not affect the repression of sulA-lacZ mediated by the dimerization of LexA DBDwt -LZ-HilD 221-309 or LexA DBDwt -H-NS (Fig. 7B). These results indicate that HilE negatively affects the dimerization of HilD but not that of H-NS or the LZ motif tested.

HilE directly affects the DNA-binding domain of HilD
The results described above indicate that HilE affects the dimerization of HilD by interacting with the central region of this regulator, which in turn would indirectly inhibit its regulatory activity. However, our results revealed that HilE also interacts with the C-terminal region of HilD (Fig. 1B), carrying the DNA-binding domain, suggesting that HilE could also directly affect the DNA binding of HilD. To investigate this, we ana-  (Fig. 8). These results, together with the results indicating that HilE does not affect the dimerization mediated by the LZ motif (Fig. 7B), show that HilE can directly affect the DNA-binding domain of HilD, independently of its effect on the dimerization of this regulator. Therefore, the higher negative effect of HilE on the regulatory activity of LexA DBDwt -HilD 1-309 than that of LexA DBDwt -LZ-HilD 221-309 (Fig. 8)

HilE inhibits the DNA binding of HilD
Our results strongly support that HilE inhibits the DNAbinding activity of HilD. To confirm this, we performed competitive electrophoretic mobility shift assays (EMSAs) with purified Trx-His-HilE and MBP-HilD proteins and a 50-bp fragment of the regulatory region of hilC carrying a HilD-binding site. As shown above, purified Trx-His-HilE interacts with HilD (Fig. 2B), and on the other hand, in a previous study we demonstrated that purified MBP-HilD binds to target genes in EMSAs (12). The hilC fragment was incubated with a constant concentration of MBP-HilD (0.5 M) without or with increasing concentrations of Trx-His-HilE (0.3, 0.5, 1.0, 1.5, and 2 M). Parallel binding reactions were performed with purified Trx-His instead Trx-His-HilE, as negative controls. From the concentration of 1.0 M, Trx-His-HilE, but not Trx-His, drastically reduced the formation of the DNA/MBP-HilD complex (Fig. 9, A and B), indicating that Trx-His-HilE interferes with the DNA binding of MBP-HilD. Additional binding reactions ruled out a possible DNA-binding activity of Trx-His-HilE, even at the highest concentration tested (2.0 M) (Fig. 9C).

HilE controls dimerization and DNA binding of HilD
Thus, these results further confirm the interaction of HilE with HilD and demonstrate that this interaction inhibits the HilD binding to DNA.

HilE shares sequence and structure similarities with Hcp proteins from T6SSs
HilE was reported several years ago as a protein that does not present homology with any other protein in the databases (38). However, our recent BLASTp analysis revealed that the HilE sequence present an identity of ϳ30% with Hcp proteins from type 6 secretion systems (T6SSs) of different Gram-negative bacteria ( Fig. 10A and data not shown). To determine whether HilE also presents structural analogy with the Hcp proteins, it was modeled by I-TASSER server (53,54). HilE was successfully modeled yielding a structure with a C-score of 1.07 and a TMscore of 0.86. The predicted structure is composed by a tight ␤-barrel domain with two ␤-sheets, with four and five ␤-strands each, flanked in one side by an ␣-helix (Fig. 10B). The overall modeled structure of HilE highly resembles that described for the monomeric subunits of Hcp proteins: Hcp1 of Pseudomonas aeruginosa (55), Hcp1 of Acinetobacter baumannii (56), EvpC of Edwardsiella tarda (57), a T6SS effector (T6SSe) of Yersinia pestis, 4 and Hcp2 of S. Typhimurium (58) (Fig. 10C)

HilE controls dimerization and DNA binding of HilD
structures sharing the same topology. These results are consistent with structural similarity shared between HilE and Hcp proteins.

Discussion
Negative regulation of HilE on the SPI-1 virulence genes is important for Salmonella fitness (59). Previously, it was shown that HilE negatively controls the expression of the SPI-1 genes by interacting with HilD, a central positive regulator for SPI-1 (38). In this study, we show that HilE specifically regulates the DNA-binding activity of HilD by inhibiting the dimerization and by directly acting on the DNA-binding domain of this regulator. In agreement with this, a recent study also found that HilE blocks the DNA binding of HilD (60). Our results indicate that HilD requires dimerization to induce the expression of target genes. Consistently, the HilD-binding sites on different genes have two direct repeat sequences (21,27,51). Several other AraC-like transcriptional regulators act as dimers, such as ToxT, ExsA, UreR, and AggR, which control expression of virulence genes in Vibrio cholerae, P. aeruginosa, Proteus mirabilis, and E. coli (45,48,49,61,62). HilE interacts with the central region of HilD and thus inhibits its dimerization, which indirectly would affect the DNA-binding activity of this regulator. Additionally, HilE interacts with the C-terminal region of HilD containing the DNA-binding domain, which negatively affects the transcriptional activity of HilD independently of its dimerization. This double interaction and effect supports a tight control of the HilD activity by HilE. In enteroagregative E. coli, the Aar protein interacts with the central region of AggR, containing the dimerization domain, inhibiting the dimerization and the DNA binding of this AraC-like regulator (62). Similarly, in P. aeruginosa, the ExsD protein interacts with the N-terminal region of ExsA, containing the dimerization domain, thus preventing the dimerization and DNA binding of this regulator (45,61,63). Neither Aar nor ExsD seem to directly compromise the DNA-binding domain of their target regulators. One example of a protein that directly acts on the DNA-binding domain of a transcriptional regulator is the control exerted by CarS on the CarA repressor in Myxococcus xanthus. CarS binds to the DNA-binding domain of CarA and thus inhibits its interaction with target genes; interestingly, the structure of CarS mimics the operator DNA recognized by CarA (64). Further studies, such as three-dimensional analysis, are required to determine the precise stoichiometry and the amino acids mediating the interaction between HilE and HilD, which is a matter of our current investigation.
Protein-protein interaction has been shown to regulate the stability of transcriptional regulators. For instance, FliT interacts with the FlhD 4 C 2 complex, the central positive regulator of the flagellar genes, which enhances the degradation of FlhC subunit by the ClpXP protease (65). Our results discarded a negative effect of HilE on the stability of HilD in the growth conditions tested, further supporting that HilE only controls the regulatory activity of HilD. However, because HilD positively autoregulates (15,26), HilE indirectly controls the expression and thus the amount of HilD.
Interestingly, we found that HilE shares sequence identity and structure analogy with Hcp proteins from different Gramnegative bacteria; thus, HilE could be postulated as a Hcp-like protein. Important to note, the Hcp-like proteins, including HilE, conserve high structure similarity but low sequence identity among them (55) (Fig. 10A). The Hcp proteins are structural components of the T6SSs, which are translocation machines that resemble an inverted phage tail, involved in different functions such as antibacterial activity and virulence (66,67). Interestingly, the T6SSs components, including the Hcp proteins, are closely related to those of the phages tail, which work as channels for DNA delivery into bacteria (66,67). Specifically, the Hcp proteins form a hexameric ring structure that stacks in a head-to-tail fashion, forming the tube through which T6SS effector proteins are either injected from the bacterial cytoplasm into the prey cell or released to the extracellular milieu (66 -68). Furthermore, the Hcp proteins can also act as chaperones that help to the translocation of the T6SS effector proteins (69). S. Typhimurium possesses a functional T6SS, which is encoded in SPI-6 (70). Whether HilE has a structural/ chaperone role in this T6SS remains to be determined.
The hilE gene is located in a genomic region that shows characteristics to be acquired by Salmonella (Fig. S2) (38); BLASTp analysis indicated that the other proteins also encoded in this Salmonella genomic island do not show homology with any protein related to the T6SSs. It is tempting to speculate that

HilE controls dimerization and DNA binding of HilD
HilE diverged from an ancestral T6SS or phage Hcp protein.
Thus, our results could illustrate the adaptation of a structural protein, during the evolution of Salmonella, to act as a regulator of virulence genes expression.

Bacterial strains and growth conditions
Bacterial strains used in this study are listed in Table S1. Bacterial cultures were grown in LB containing 1% tryptone, 0.5% yeast extract, and 1% NaCl at pH 7.5. When necessary, the media were supplemented with the following antibiotics: ampicillin (200 g/ml), streptomycin (100 g/ml), tetracycline (12 g/ml), or kanamycin (20 g/ml). Cultures for the determination of chloramphenicol acetyltransferase activity (CAT) and for the LexA-based genetic system assays were performed as we described previously (12,28,50).

Construction of plasmids
Plasmids and primers used in this study are listed in Table S1 and Table S2, respectively. To construct the plasmids pSR658 -HilD1, pSR658 -HilD2, pSR658 -HilD3, pSR658 -HilD4, and pSR658 -HilD5, fragments of the hilD gene were amplified by PCR using the primers pairs HilD-SacI/HilDexR-PstI, HilD-SacI/HilD-130, HilD130 -2/HilDexR-PstI, HilD-SacI/HilD-220, and HilD-221/HilDexR-PstI, respectively. The resulting PCR products were digested with SacI and PstI and cloned into the vector pSR658 digested with the same restriction enzymes. To construct the pSR659-HilE1 plasmid, the hilE gene was amplified by PCR using the primers HilE-SacI and HilE-HindIII-3Ј. The resulting PCR product was digested with SacI and HindIII and cloned into the vector pSR659 digested with the same restriction enzymes. To construct the plasmid pSR658 -HilD6, the DNA fragment encoding 35 amino acids of the LZ motif from GCN4 of S. cerevisiae was amplified by PCR using the primers LZ-F and HilD221LZR and DNA of the pUT18C-zip plasmid as template. The fragment encoding the region of HilD from codons 221 to 309 (HilD 221-309 ) was also amplified by PCR using the LZ-HilD221F and HilDexR-PstI primers. Then the LZ and HilD 221-309 fragments were fused by overlap extension PCR using the LZ-F and HilDexR-PstI primers. The resulting PCR product was digested with XhoI and PstI and cloned into the pSR658 vector digested with the same restriction enzymes. To construct the pA6 -HilE1 and pK6 -HilE1 plasmids, the hilE gene was amplified by PCR using the HilE-NcoI-2 and HilE-His 6 primers. The resulting PCR product was digested with NcoI and HindIII and cloned into the pMPM-A6⍀ or pMPM-K6⍀ vectors digested with the same restriction enzymes. To construct the plasmid pET32-HilE expressing the Trx-His-HilE fusion protein, the hilE gene was amplified by PCR using the HilE-NcoI-2 and HilE-PUT-BamHI primers. The resulting PCR product was digested with NcoI and BamHI and cloned into the pET32b(ϩ) vector digested with the same restriction enzymes.

HilD dimerization assays
To test dimerization of HilD, the plasmids pSR658, pSR658 -HilD1, or pSR658 -HNS were transformed into the E. coli SU101 reporter strain for homodimerization assays, which carries the chromosomal sulA-lacZ transcriptional fusion (43,44). Likewise, to determine the region of HilD containing the dimerization domain, the plasmids pSR658, pSR658 -HilD1, pSR658 -HilD2, pSR658 -HilD3, pSR658 -HilD4, or pSR658 -HilD5 were transformed into the same reporter strain. Transformants were grown in LB with tetracycline and 1 mM IPTG to induce expression of LexA DBD and LexA DBD fusion proteins. The samples were collected at an A 600 of 1.0 and used for the determination of ␤-gal activity.
To test whether HilE affects the dimerization of HilD, the E. coli SU101 reporter strain was first transformed with the plasmids pSR658, pSR658 -HilD1, pSR658 -HNS, or pSR658 -HilD6 and then transformed with the pMPM-A6⍀ vector or the pA6 -HilE1 plasmid expressing HilE from the arabinoseinducible promoter. Transformants were grown in LB with tetracycline and ampicillin, as well as with 1 mM IPTG to induce the expression of LexA DBD and LexA DBD fusion proteins and with 0.1% L-arabinose to induce the expression of HilE. The samples were collected at an A 600 of 1.0 and used for the determination of ␤-gal activity.

HilD-HilE heterodimerization assays
To test the interaction between HilD and HilE, the E. coli SU202 reporter strain for heterodimerization assays, which carries the sulA-lacZ transcriptional fusion with a hybrid LexA operator (43,44) was first transformed with the plasmids pSR658, pSR658 -HilD1, pSR658 -HilD2, pSR658 -HilD4, or pSR658 -HilD5 and then transformed with the pSR659 vector or the pSR659-HilE1 plasmid. Transformants were grown in LB with tetracycline and ampicillin and with 1 mM IPTG to induce expression of LexA DBD and LexA DBD fusion proteins. The samples were collected at an A 600 of 1.0 and used for the determination of ␤-gal activity.

␤-gal assays
The ␤-gal assay and protein quantification to calculate specific activities were performed as previously described (71).

CAT assays
The CAT assays and protein quantification to calculate CAT-specific activities were performed as previously described (72).

HilE controls dimerization and DNA binding of HilD
alog no. G6532) polyclonal antibodies at 1:10,000 and 1:100,000 dilutions, respectively. Horseradish peroxidase-conjugated anti-mouse (Sigma, catalog no. A9044) or anti-rabbit (Rockland, catalog no. 611-1302) at a dilution of 1:10,000 were used as the secondary antibodies. The bands on the blotted membranes were developed by incubation with the Western Lightning Chemiluminescence reagent plus (PerkinElmer) and exposed to Kodak X-Omat films or Amersham Imager 600 (GE Healthcare).

Expression and purification of MBP-HilD
Expression and purification of MBP-HilD were performed as we previously described (12).

Expression and purification of Trx-His and Trx-His-HilE
The Trx-His and Trx-His-HilE proteins were expressed in E. coli BL21/DE3 carrying the pET32b(ϩ) or pET32-HilE plasmids, respectively, and purified from soluble bacterial extracts by using Ni-NTA-agarose columns (Qiagen). Bacterial cultures were grown in 250 ml of LB with ampicillin up to an A 600 of 0.6, at 37°C. Then the expression of Trx-His or Trx-His-HilE was induced by adding 1 mM IPTG, and the cultures were incubated at 30°C for 4 h. Bacteria were collected by centrifugation at 4°C, and the pellets were washed once with ice-cold binding buffer (5 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 8.0) and resuspended in 30 ml of the same buffer. The cells were lysed with French press, and bacterial debris were removed by centrifugation at 4°C. The soluble bacterial extracts were loaded into Ni-NTA-agarose columns previously equilibrated with 50 ml of binding buffer. The columns were washed with 100 ml of binding buffer and then with 100 ml of wash buffer (20 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 8.0). The proteins were eluted with elution buffer (250 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 8.0). The collected fractions were analyzed in a 12% SDS-PAGE. Those fractions containing the purified Trx-His and Trx-His-HilE proteins were loaded into a Slide-A-Lyzer 7K cassette (Thermo) and dialyzed at 4°C in a buffer containing 20 mM Tris-HCl, pH 8.0, 40 mM KCl, 1 mM EDTA, and 20% (v/v) glycerol. Protein concentration was determined by the Bradford procedure. Aliquots of the purified Trx-His and Trx-His-HilE proteins were stored at Ϫ70°C.

Electrophoretic mobility assays
EMSAs were performed using the purified MBP-HilD, Trx-His-HilE, or Trx-His proteins and a 50-bp DNA fragment of hilC containing a binding site of HilD. The 50-bp hilC fragment was generated by annealing the complementary primers HilCRR-F and HilCRR-R. For this, the primers, each at a final concentration of 15 M, were boiled together at 95°C for 10 min and then slowly cooling to room temperature. Competitive binding reactions were performed by mixing ϳ100 ng of the hilC fragment with MBP-HilD (0.5 M) and increasing amounts of Trx-His-HilE or Trx-His (0.3, 0.5, 1.0, 1.5, and 2.0 M), in a total volume of 20 l of binding buffer containing 100 g/ml BSA, 30 mM HEPES, pH 7.5, 5 mM EDTA, 3 mM DTT, 200 mM KCl, 25 mM MgCl 2 , and 5% glycerol. For noncompetitive EMSAs, only increasing amounts of Trx-His-HilE purified protein (0.3, 0.5, 1.0, 1.5, and 2.0 M) were used. Protein-DNA binding reactions were incubated at 37°C for 20 min and then electrophoretically separated in 6% nondenaturing polyacrylamide gels in 0.5ϫ Tris borate-EDTA buffer, at room temperature. The DNA fragments were stained with ethidium bromide and visualized with an Alpha-Imager UV Transilluminator (Alpha Innotech Corp.).

HilD stability assays
HilD stability assays were performed as we described previously (33). The values for HilD-Myc bands were normalized with respect to those for GroEL bands, and then the relative percentage of HilD-Myc at each indicated time, with respect to time 0, was calculated. The half-life time (t 1/2 ) of HilD was calculated by one phase decay equation.

Protein secretion analysis
Protein secretion assays were performed as we described previously (28). The samples were analyzed in a 12% SDS-PAGE and stained with Coomassie Brilliant Blue R-250.

Gel-filtration assay
The purified protein MBP-HilD was subjected to gel filtration chromatography analysis by using AKTA-FPLC system (Superdex 200 HiLoad TM 26/60 column; GE Healthcare Life Sciences), with a flow rate of 0.5 ml/min and a pressure limit of 0.3 MPa, in a buffer containing 200 mM Tris-HCl, pH 8.0, 150 mM NaCl at 20°C. The column was precalibrated by using a gel filtration molecular weight markers kit (Sigma-Aldrich) including cytochrome c from horse heart (12.4 kDa), carbonic anhydrase from bovine erythrocytes (29 kDa), albumin bovine serum (66 kDa), alcohol dehydrogenase from yeast (150 kDa), and ␤-amylase from sweet potato (200 kDa). The relative molecular mass of MBP-HilD was determined by comparison to the five-point molecular weight calibration curve.

Pulldown assays
Pulldown assays were performed with purified Trx-His-HilE or Trx-His fusion proteins and whole-cell extracts from E. coli SU101 expressing LexA DBDwt , LexA DBDwt -HilD 1-309 , LexA DBDwt -HilD 1-130 , LexA DBDwt -HilD 1-220 , or LexA DBDwt -HilD 221-309 . To immobilize the bait protein, 15 g of purified Trx-His-HilE or Trx-His were incubated for 1 h at 10°C with 80 l of Ni-NTA resin (Qiagen) previously equilibrated with lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 5 mM imidazole, pH 8.0) in a 1.5-ml microcentrifuge tube. The resin containing the immobilized bait protein was washed with 1 ml of wash buffer (70 mM NaH 2 PO 4 , 300 mM NaCl, 35 mM imidazole, pH 8.0) by centrifugation at 4000 ϫ g for 1 min. The supernatant was carefully removed, and then the resin was incubated for 1 h at 10°C with 80 l of the whole-cell extract containing the respective prey protein. To remove the unbound proteins, the resin was washed five times with 1 ml of wash buffer by centrifugation at 4000 ϫ g for 1 min. Finally, the resin was resuspended with 20 l of SDS loading buffer containing 1.5% of ␤-mercaptoethanol and boiled for 5 min at 99°C. After this, the samples were analyzed by Western blotting.

HilE controls dimerization and DNA binding of HilD Sequence alignment and structure prediction of HilE
The sequence of HilE and that of some Hcp proteins were aligned using the Clustal Omega server (73). The sequence of HilE was submitted to I-TASSER server for structural modeling (53,54), and a final model with a C score of 1.07, a TM score of 0.86, and a root mean square deviation of 2.7 Å was selected. All molecular graphics were done in PyMOL version 1.8 (PyMOL Molecular Graphics System, version 1.8).

Statistical analysis
Data from CAT and ␤-gal assays were analyzed using oneway analysis of variance with the Tukey's multiple comparison test. A p value of Ͻ 0.05 was considered significant. This statistical analysis was performed using Prism 6 program version 6.01 (GraphPad Software, San Diego, CA).