An incoherent feedforward loop formed by SirA/BarA, HilE and HilD is involved in controlling the growth cost of virulence factor expression by Salmonella Typhimurium

An intricate regulatory network controls the expression of Salmonella virulence genes. The transcriptional regulator HilD plays a central role in this network by controlling the expression of tens of genes mainly required for intestinal colonization. Accordingly, the expression/activity of HilD is highly regulated by multiple factors, such as the SirA/BarA two-component system and the Hcp-like protein HilE. SirA/BarA positively regulates translation of hilD mRNA through a regulatory cascade involving the small RNAs CsrB and CsrC, and the RNA-binding protein CsrA, whereas HilE inhibits HilD activity by protein-protein interaction. In this study, we show that SirA/BarA also positively regulates translation of hilE mRNA through the same mentioned regulatory cascade. Thus, our results reveal a paradoxical regulation exerted by SirA/BarA-Csr on HilD, which involves simultaneous opposite effects, direct positive control and indirect negative control through HilE. This kind of regulation is called an incoherent type-1 feedforward loop (I1-FFL), which is a motif present in certain regulatory networks and represents a complex biological problem to decipher. Interestingly, our results, together with those from a previous study, indicate that HilE, the repressor component of the I1-FFL reported here (I1-FFLSirA/BarA-HilE-HilD), is required to reduce the growth cost imposed by the expression of the genes regulated by HilD. Moreover, we and others found that HilE is necessary for successful intestinal colonization by Salmonella. Thus, these findings support that I1-FFLSirA/BarA-HilE-HilD cooperates to control the precise amount and activity of HilD, for an appropriate balance between the growth cost and the virulence benefit generated by the expression of the genes induced by this regulator. I1-FFLSirA/BarA-HilE-HilD represents a complex regulatory I1-FFL that involves multiple regulators acting at distinct levels of gene expression, as well as showing different connections to the rest of the regulatory network governing Salmonella virulence.


Introduction
Pathogenic bacteria have developed diverse regulatory mechanisms to control the appropriated spatiotemporal expression of virulence genes. Both activation and repression regulatory mechanisms are essential for bacteria to colonize different niches of hosts.
Salmonella enterica serovar Typhimurium (S. Typhimurium) contains a large number of virulence genes and a very complex regulatory network that control their expression. Two groups of genes mainly govern Salmonella virulence; those located in Salmonella Pathogenicity Island 1 (SPI-1) and those located in SPI-2. Both SPI-1 and SPI-2 encode a Type III Secretion System (T3SS), several effector proteins and chaperones, as well as transcriptional regulators [1,2]. The T3SS is a syringe-like multiprotein complex through which bacteria inject effector proteins directly into the cytoplasm of host cells. Effector proteins have distinct biological activities that alter different signal transduction pathways of eukaryotic host cells [3]. The SPI-1 genes mediate Salmonella invasion of cells at the intestinal epithelium, which generates enteritis, whereas the SPI-2 genes primarily mediate Salmonella replication within host cells [1,2]. Replication/survival inside macrophages within a membrane-bound compartment called the Salmonella-containing vacuole (SCV), leads to a systemic infection like typhoid fever [1,2]. Many other Salmonella virulence genes are located in distinct genomic regions, with their biological function being related to that of SPI-1 or SPI-2 [2,4,5].
Consistent with their role in intestinal infection, expression of the SPI-1 genes is induced when Salmonella resides in the intestinal lumen and in the cytosol of epithelial cells [6,7]. In addition, expression of the SPI-1 genes is regulated by different environmental cues commonly found in the intestine of hosts, such as short-and long-chain fatty acids, bile, oxygen level, osmolarity and pH [8][9][10][11][12][13]. In vitro, these genes are expressed in nutrient-rich media, such as lysogeny broth (LB), during the late exponential and early stationary growth phases [14][15][16]. Interestingly, different studies have reported that the SPI-1 genes show a bistable expression both in vitro [17][18][19][20][21] and in the gut lumen of mice [22], where only 10-50% of cells from clonal Salmonella populations express SPI-1. Moreover, it was shown that the two subpopulations of cells generated by this bistable expression (SPI-1 ON and SPI-1 OFF ) cooperate for the successful invasion of host cells and intestinal colonization by Salmonella [20,21].
Expression of the SPI-1 and related genes is controlled by a very complex regulatory network involving many positive and negative regulators that act at the transcriptional, translational or post-translational level [2,23,24]. The AraC-like regulator HilD, encoded in SPI-1, is the apex of different regulatory cascades through which HilD controls the expression of the SPI-1 genes and many other genes located outside of SPI-1. HilD induces expression of: 1) the SPI-1 genes and other genes located outside of SPI-1, through the HilA, InvF or SprB regulators, encoded in SPI-1 [2,[23][24][25][26]; 2) the SPI-2 genes and other genes located outside SPI-2, through the SsrA/SsrB two-component system, encoded in SPI-2 [15,27,28]; 3) the flagellar and chemotaxis genes, through the FlhD 4 C 2 transcriptional complex [29,30]; and 4) several other genes located in distinct genomic regions, by direct interaction [26,31,32]. Furthermore, HilD forms a positive feedforward regulatory loop with the AraC-like regulators HilC and RtsA, which recognize the same DNA motif as HilD and form heterodimers with HilD; however, HilD has a dominant function over HilC and RtsA [33][34][35].
Consistently with its role as a master regulator for Salmonella virulence, the expression, concentration and activity of HilD are tightly controlled, representing the central point for regulating SPI-1 genes and many other related genes [2,23,24]. For instance, the Hcp-like protein HilE, encoded in a genomic island other than the SPIs, inhibits HilD activity by protein-protein interaction, thereby affecting homodimerization and DNA binding of HilD [36][37][38]. In addition, the SirA/BarA two-component system induces expression of hilD at the translational level [39]. In this two-component system, SirA and BarA are the response regulator and the sensor kinase, respectively [40][41][42].
Orthologs of SirA/BarA are present in many other bacteria. In Escherichia coli (UvrY/ BarA), Pseudomonas spp. (GacA/GacS), and Vibrio cholerae (VarA/VarS), this system controls expression of numerous genes encoding different cellular activities, including virulence, motility, biofilm formation and metabolism [43,44]. This system has been better characterized in E. coli where UvrY/BarA was shown to form a regulatory cascade with the non-translated small RNAs (sRNAs) CsrB and CsrC, and the RNA binding protein CsrA [44,45]. CsrA represses numerous genes at the translational level by interacting with sequences overlapping the ribosome binding sites on target mRNAs [44,45]. In response to its phosphorylation by BarA, UvrY directly activates transcription of CsrB and CsrC, which contain several binding sites for CsrA [44,45]. Thus, CsrB/CsrC sequester CsrA and antagonize its ability to regulate expression of target transcripts [44,45]. Several studies have shown that the counterparts of the UvrY/Bar-A-CsrB/CsrC-CsrA global regulatory network function similarly in other bacteria [43,44]. For instance, the SirA/BarA system induces expression of the SPI-1 genes through CsrB and CsrC, which antagonize the CsrA-mediated translational repression of the hilD transcript [39,46,47].
Interestingly, expression of SPI-1 genes causes growth retardation of S. Typhimurium in laboratory conditions. This effect becomes even more pronounced when the SPI-1 genes are overexpressed, such as in the absence of HilE [19]. However, it remains poorly understood how S. Typhimurium fine-tunes the expression level of the SPI-1 genes to maintain an appropriate balance between the penalty on growth and the benefit for virulence.
In this study, we report that the SirA/BarA-CsrB/CsrC regulatory cascade induces the expression of hilE by counteracting CsrA-mediated translational repression on the hilE mRNA, which reveals that SirA/BarA, HilD and HilE form an incoherent type-1 feedforward loop. Additionally, we demonstrate that HilE is necessary to reduce the growth cost imposed by the expression of genes regulated by HilD, in laboratory conditions and in the intestine of mice. Thus, our results support that the control of HilD expression by the feedforward loop formed by SirA/BarA, HilD and HilE, plays a role in the fitness of S. Typhimurium during intestinal infection of hosts.

CsrA directly represses the expression of hilE
In a current project to analyze the global effect of CsrA on Salmonella we obtained circumstantial evidence that CsrA could negatively regulate the expression of hilE. To further investigate this phenomenon, we analyzed the chromosomal expression of FLAG-tagged HilE (HilE-FLAG) in the wild-type (WT) S. Typhimurium SL1344 strain carrying the pK3-CsrA plasmid expressing CsrA, when grown in conditions that favor the expression of SPI-1 genes (LB medium, at 37˚C, with shaking; SPI-1-inducing conditions). Expression of CsrA from pK3-CsrA completely inhibited the amount of HilE-FLAG ( Fig 1A). Similar results were obtained by assessing the expression of a hilE-lacZ translational fusion ( Fig 1B). We were unable to evaluate the expression of HilE-FLAG and hilE-lacZ in the absence of CsrA because csrA mutants of S. Typhimurium exhibit severe growth defects [39,48]. These results indicate that CsrA negatively controls the expression of hilE.
To determine whether CsrA regulates hilE directly, quantitative electrophoretic mobility shift assays (EMSAs) were performed using purified CsrA and the 5'-end-labelled leader RNA of hilE. A band with lower mobility was detected with concentrations of CsrA between 19 and 75 nM, indicating that CsrA formed a complex with the hilE transcript; at 150 and 300 nM CsrA a second complex with even lower mobility was also observed (Fig 2A). These data support that CsrA binds two sites on the hilE transcript. Nonlinear least-squares analysis of these EMSAs data yielded an apparent K d value of 37 ± 13 nM CsrA for hilE mRNA. The specificity of the CsrA-hilE RNA interaction was evaluated by performing competition experiments with specific (hilE) and non-specific (phoB) unlabeled RNA competitors. Whereas unlabeled hilE RNA was an effective competitor, phoB RNA was not ( Fig 2B). In agreement with these results, data from a previous global analysis by CLIP-seq showed that CsrA binds in vivo to a sequence located near the translation start codon of the hilE mRNA [49]. Thus, we conclude that CsrA binds specifically to the hilE leader transcript. Together, our results indicate that CsrA directly represses the expression of hilE.

SirA/BarA induces the expression of hilE through CsrB/C
CsrA-mediated repression of target genes is counteracted by the two-component system SirA (UvrY)/BarA through the sRNAs CsrB and CsrC, which bind and sequester CsrA [44,45]. To define the complete regulatory cascade involving CsrA that controls the expression of hilE, the expression of HilE-FLAG and that of the lacZ-hilE fusion was monitored in the WT S. Typhimurium strain and its ΔsirA, ΔcsrB, ΔcsrC and ΔcsrB ΔcsrC derivative mutants, grown in SPI-1-inducing conditions. Both the amount of HilE-FLAG and the expression of hilE-lacZ were reduced in the ΔsirA mutant, as well as in the ΔcsrB ΔcsrC double mutant, compared with the WT strain (Fig 3A and 3B). Additionally, as shown in Fig 3B, activity of the hilE-lacZ fusion was reduced in the ΔbarA mutant lacking BarA, the cognate sensor kinase of the SirA response regulator [41,42]. In contrast, the ΔcsrB and ΔcsrC single mutants showed WT levels of HilE-FLAG and hilE-lacZ expression (Fig 3A and 3B), which is consistent with previous reports indicating that in S. Typhimurium, the absence of both CsrB and CsrC is required for observable effects on the expression of other target genes of the SirA/BarA-Csr cascade [39,50]. As expected, complementation of the ΔsirA mutant with the pK3-SirA plasmid expressing SirA, restored the activity of the hilE-lacZ fusion to levels even slightly higher than those of the WT strain ( Fig 3C). Moreover, complementation of the ΔsirA mutant with the pK3-CsrB plasmid expressing CsrB, also restored the activity of the hilE-lacZ fusion ( Fig 3C). Collectively, these

HilE represses HilD-mediated expression of SPI-1, SPI-2 and SPI-5 genes
HilE negatively affects the activity of the transcriptional regulator HilD [36][37][38]. Consequently, HilE would decrease the expression of the large number of genes controlled by HilD. To further investigate this repressor role of HilE, we quantified the expression of two genes each from SPI-1 (hilA and invF), SPI-2 (ssrAB and ssaG) and SPI-5 (sopB and pipB), as cat transcriptional fusions. HilD positively regulates the expression of all of these genes directly or indirectly [15,23,27,33,[51][52][53]. Experiments were performed in the WT S. Typhimurium strain and its ΔhilE derivative mutant grown in SPI-1-inducing conditions. As shown in Fig 4A, the expression of every gene increased significantly in the ΔhilE mutant compared with the WT strain. As expected, due to the absence of HilD, only low expression level was observed for these genes in the ΔhilD and ΔhilE ΔhilD mutants (Fig 4A). Consistent with our results, previous studies also indicate that the expression of hilA increases in the absence of HilE [36,54]. Additionally, we examined and compared the expression level of the hilA-cat fusion in the WT S. Typhimurium strain and its ΔhilE, sirA::Tn10d (transposon insertion in sirA) and ΔhilE sirA::Tn10d derivative mutants in the absence or presence of the pK3-SirA plasmid expressing SirA, the pMPM-K3/K6 vector, or the pK6-HilE plasmid expressing HilE. In agreement with previous studies indicating that SirA positively regulates SPI-1 genes [39,40,42,55], the hilAcat fusion showed reduced expression levels in the sirA::Tn10d mutant containing the pMPM-K3 vector, with respect to the WT strain ( Fig 4B). Surprisingly, the hilA-cat fusion was expressed in the ΔhilE sirA::Tn10d double mutant containing the pMPM-K3 vector at a level only slightly lower than in the WT strain (Fig 4B), revealing than in the absence of HilE, hilA is expressed independently of SirA. It is important to consider that the HilE expression is positively controlled by SirA (this study) but also by other regulators [56][57][58][59] as discussed later. In fact, our results show that a partial amount of HilE is present in the absence of SirA (Fig 3). Thus, our results suggest that in the absence of SirA the expression of hilA is repressed by two mechanisms: 1) CsrA inhibits translation of the hilD mRNA and 2) the diminished amount of HilE is able to inactivate background levels of HilD, which would avoid the positive autoregulation of HilD. It is reasonable to suggest that when SirA counteracts the CsrA-mediated repression of hilD, higher levels of HilE would be required to negatively control the activity of HilD, which can be reached with the positive regulation of the hilE expression by SirA.
The presence of the pK3-SirA plasmid further increased the expression of hilA-cat in the ΔhilE sirA::Tn10d mutant and, as could be expected, the presence of the pK6-HilE plasmid completely inhibited the expression of hilA-cat in the ΔhilE sirA::Tn10d mutant ( Fig 4B). To note, the hilA-cat fusion showed similar expression levels in the ΔhilE and ΔhilE sirA::Tn10d + pK3-SirA mutants, as well as in the ΔhilD mutant complemented with the pK6-HilD plasmid CAT activity of the hilA-cat, invF-cat, ssaG-cat, ssrAB-cat, pipB-cat and sopB-cat transcriptional fusions, contained in plasmids philA-cat, pinvF-cat, pssaG-cat, pssrAB-cat, ppipB-cat and psopB-cat, respectively, was determined in the WT S. Typhimurium strain and its derivative ΔhilE, ΔhilD and ΔhilE ΔhilD mutants. (B) CAT activity of the hilA-cat transcriptional fusion, contained in the philA-cat plasmid, was determined in the WT S. Typhimurium strain and its derivative ΔhilE, sirA::Tn10d and ΔhilE sirA::Tn10d mutants in the absence or presence of pMPM-K3 or pMPM-K6 vectors, the pK3-SirA plasmid expressing SirA from a constitutive promoter or the pK6-HilE plasmid expressing HilE from an arabinose-inducible promoter. Expression of HilE from pK6-HilE was induced by adding 0.001% L-arabinose to the medium. (C) CAT activity of the hilA-cat transcriptional fusion, contained in the philA-cat plasmid, was determined in the WT S. Typhimurium strain and its derivative ΔhilD mutant carrying the pMPM-K6 vector or the pK6-HilD plasmid, which expresses HilD from an arabinoseinducible promoter. Expression of HilD from pK6-HilD was induced with 0%, 0.0001% or 0.001% L-arabinose. CAT-specific activity was obtained from samples collected from bacterial cultures grown in LB at 37˚C. Data represent the average of three independent experiments done in duplicate. Error bars indicate the standard deviations. Statistically different values are indicated ( �� , p-value < 0.01; ��� , p-value < 0.001; ���� , p-value < 0.0001). https://doi.org/10.1371/journal.ppat.1009630.g004 overexpressing HilD (Fig 4B and 4C), which could suggest that in the absence of HilE (or overexpression of HilD) maximal expression levels of hilA are reached.
We next aimed to analyze the effect of HilE on the fraction of cells expressing SPI-1 genes in the Salmonella population. To do this, GFP fluorescence expression from an invF-gfp transcriptional fusion was quantified by flow cytometry in cultures of the WT S. Typhimurium strain and its ΔhilE and ΔhilE ΔhilD derivative mutants, as well as in cultures of the sirA:: Tn10d and ΔhilE sirA::Tn10d mutants carrying the pK4-SirA plasmid expressing SirA or the pMPM-K4O vector, grown in SPI-1-inducing conditions. Similar to other reports [17-21], only 28% of the cells from cultures of the WT strain expressed invF-gfp, whereas 62% of the cells expressed this fusion in cultures of the ΔhilE mutant ( Fig 5A). As expected, expression of invF-gfp in the ΔhilE ΔhilD mutant or the WT strain carrying the gfp reporter gene without a promoter showed nearly undetectable expression of GFP fluorescence (Fig 5A). In addition, the fraction of cells expressing the invF-gfp fusion was higher in the ΔhilE sirA::Tn10d double mutant than in the sirA::Tn10d single mutant, both in the presence of the vector pMPM-K4O or the pK4-SirA plasmid ( Fig 5B). Intriguingly, complementation of the sirA::Tn10d and ΔhilE sirA::Tn10d mutants yielded a subpopulation of cells with an invF-gfp expression pattern somewhat different to that of cells from the WT strain and the ΔhilE mutant (Fig 5A and 5B), which seems to be an effect of the expression of SirA from a multicopy plasmid. Nevertheless, these results clearly show that HilE reduces the fraction of cells expressing SPI-1 genes in the Salmonella population.
Collectively, these results reinforce that HilE plays a major role in regulating the SPI-1 genes by acting as a repressor of HilD. It is important to note that the presence of HilE only decreases, but does not eliminate, the HilD-mediated activation of target genes, or the fraction of cells from the Salmonella population that expresses these genes, in SPI-1-inducing conditions.

HilE reduces the deleterious effect on growth caused by expression of genes controlled by HilD
Interestingly, SirA/BarA induces the expression of both HilD [39] and HilE (this study), which would seem counterintuitive. What would be the role of the regulation of hilE by SirA/BarA?
Overexpression of HilD or the absence of HilE retard growth of S. Typhimurium in competitive assays performed in LB, which was partially attributable to the overexpression of some genes regulated by HilD [19]. To further investigate this phenomenon, we monitored the growth of the WT S. Typhimurium strain and the ΔhilE, ΔhilD and ΔhilE ΔhilD mutants, in mixed cultures grown in SPI-1-inducing conditions. In the WT/ΔhilE mixed cultures, the WT strain outcompeted the ΔhilE mutant, representing 80% of the viable cells at the end of these assays (Fig 6A). In contrast, in the WT/ΔhilD mixed cultures, the WT strain was outcompeted by the ΔhilD mutant; only 14% of the viable cells at the end of these assays were WT (Fig 6B). These results are in agreement with those obtained previously, showing that the absence of HilE and HilD has negative and positive effects, respectively, on the growth of S. Typhimurium cultures [19]. Furthermore, we observed that the WT strain was also outcompeted by the ΔhilE ΔhilD mutant to a similar extent as the ΔhilD single mutant (Fig 6B and 6C), suggesting that the negative effect caused by the absence of HilE on the S. Typhimurium growth requires the presence of HilD.
Additionally, we analyzed the competitive growth between the ΔsirA and ΔhilE sirA::Tn10d mutants carrying the pMPM-K3 vector or the pK3-SirA plasmid expressing SirA. Interestingly, in the presence of the pMPM-K3 vector, the ΔsirA and ΔhilE sirA::Tn10d mutants grew similarly through the time assessed (Fig 6D). In contrast, in the presence of the pK3-SirA plasmid, the ΔhilE sirA::Tn10d mutant was outcompeted by the ΔsirA mutant; only 18% of the viable cells at the end of these assays were from the ΔhilE sirA::Tn10d mutant strain (Fig 6E). These results reveal that SirA is required for HilE to play a role in the fitness of S. Typhimurium.
Together, these results show that HilE reduces the deleterious effect on the S. Typhimurium growth caused by the expression of genes regulated by HilD in laboratory conditions. SPI-1-inducing conditions are considered to somehow mimic the intestinal environment found by Salmonella in hosts. Therefore, on the basis of our results described above, we investigated whether HilE is important for successful intestinal colonization by S. Typhimurium. For this analysis, we determined and compared the survival of the WT S. Typhimurium strain to that of the ΔhilE, ΔhilD and ΔhilE ΔhilD mutants, in the intestine of streptomycin-pretreated mice, a model used to study the intestinal colonization by S. Typhimurium [60]. Groups of eight mice were orally infected with a mix of an equal number of cells of the WT strain and each mutant strain. Three days post-infection, bacteria from feces and cecum of mice were counted, and a competitive index was obtained to determine the proportion between each mutant and the WT strain. The ΔhilE mutant showed a~30-fold and~250-fold reduction in survival with respect to the WT strain, in feces and cecum, respectively. In contrast, the survival of the ΔhilD and ΔhilE ΔhilD mutants was similar or slightly higher than that of the WT strain in both feces and cecum (Fig 7). Thus, the absence of HilE attenuates the intestinal colonization by S. Typhimurium in a HilD-dependent way. These results are in agreement with those from a previous study showing that HilE is important for the intestinal disease caused by long term infections of S. Typhimurium in mice (i.e. at 10 days post-infection), including colonization and induction of enteritis [20]. Furthermore, our results are also consistent with reports indicating that SPI-1 or HilD function is required to invade cells from the intestinal epithelium, and thus to induce enteritis, but not to colonize the cecum lumen [2,20,[60][61][62]. Replication of Salmonella in the intestinal lumen has been shown to require genes for the use of specific nutrients present in the inflamed intestine, such as the eut, ttr and pdu genes [63][64][65], which, according to previous transcriptomic analyses, are not regulated by HilD [66]. These data reveal a crucial role for HilE during intestinal infection by S. Typhimurium. In all, together with previous reports, our data suggest that by inducing the expression of HilD and its negative regulator HilE, SirA/BarA exerts a fine-tuning regulatory control that cooperates to lessen the growth penalty produced by HilD-mediated expression of several tens of virulence factors, which is required for the successful intestinal colonization by S. Typhimiurium.

Discussion
Regulatory networks are built up by recurring patterns of interactions between regulatory factors that are designed as network motifs, which are present in diverse organisms including bacteria, yeast, plants, animals, and humans [67,68]. One family of network motifs is the feedforward loop (FFL), where a factor X regulates factors Y and Z, and Y regulates Z. As the three regulatory interactions between X, Y, and Z can be either positive or negative, eight structural types of FFLs are possible [68,69]. One FFL commonly present in the best-studied transcriptional networks, those from E. coli and yeast, is the incoherent type-1 FFL (I1-FFL), where a factor X positively regulates factors Y and Z, while Y negatively regulates Z. Thus, X exerts an "incoherent" (or "paradoxical") regulation on Z through two opposite effects [69,70] (Fig 8A), which represents a complex biological problem to decipher.
Our results from this study reveal a novel I1-FFL involved in gene expression in bacteria. We previously determined that in S. Typhimurium, the SirA/BarA two-component system induces the expression of the transcriptional regulator HilD, through the sRNAs CsrB and CsrC, which counteracts the CsrA-mediated repression of hilD translation [39]. HilD induces the expression of a large number of Salmonella virulence genes [2,15,26,29,31,32,66]. In this work, we show that the SirA/BarA/Csr system also regulates expression of HilE by a similar way. HilE negatively controls HilD activity by protein-protein interaction [36][37][38]. It is important to note that HilD is the only known target of HilE. Hence, the SirA/BarA-CsrB/C cascade exerts two opposite regulatory effects on HilD; the activation of the hilD mRNA translation, and the inhibition of HilD activity through HilE, thus constituting an I1-FFL (I1-FFL SirA/BarA-HilE-HilD ) that involves regulation at the transcriptional, translational and post-translational levels ( Fig 8B).
I1-FFLs can influence fold-change detection [71], adaptive tuning [72], response time [69,70] and response amplitude of gene expression [73]. Our results and those from a previous study [19] show that in the absence of HilE, the repressor factor of I1-FFL SirA/BarA-HilE-HilD ( Fig  8B), the growth of S. Typhimurium is slowed down in laboratory conditions, which can be attributed to the overexpression of genes regulated by HilD that has a deleterious effect by a still unknown mechanism. We found that the absence of HilE negatively affects the S. Typhimurium growth only in the presence of SirA and HilD, the two other components of I1-FFL SirA/BarA-HilE-HilD . Additionally, our results and a prior report [20] indicate that the SPI-1 genes are expressed in vitro in a bistable fashion that involves a control by HilE. In the presence of HilE only about one third fraction of the S. Typhimurium population expresses SPI-1, while in the absence of HilE the fraction of cells expressing SPI-1 increases more than twofold. Moreover, our results and those reported previously [20] indicate that the absence of HilE attenuates the intestinal colonization by S. Typhimurium, an effect also mediated through HilD. Interestingly, in the intestine of mice, the SPI-1 genes are also expressed in a bistable fashion, and in the absence of HilE, the proportion of S. Typhimurium cells expressing SPI-1 increases, imposing a growth pressure that leads to the selection of avirulent hilD mutants [20]. Furthermore, the lack of HilE increases the invasion of S. Typhimurium into epithelial culture cells by 2-5-fold [36,54]. Altogether, these findings suggest that I1-FFL SirA/BarA-HilE-HilD is involved, together with other regulatory pathways, in the control of the precise amount and activity of HilD, resulting in an appropriate balance between the growth cost and the virulence benefit generated by the expression of the genes induced by this regulator (Fig 8B).
The SirA/BarA system seems to be directly activated by short-chain fatty acids such as acetate and formate, which are known to induce the expression of SPI-1 genes [8,[74][75][76]. Therefore, these molecules commonly present in the mammalian intestines, probably in conjunction with other environmental cues, are expected to turn on I1-FFL SirA/BarA-HilE-HilD .
However, several other positive and negative regulatory mechanisms acting on HilD or HilE provide additional inputs to I1-FFL SirA/BarA-HilE-HilD for fine tuning the expression level of the genes regulated by HilD (Fig 8B). Expression of HilD is controlled positively by transcriptional autoregulation, directly or through the feedforward loop that HilD forms with HilC and RtsA [32][33][34]77,78]. HilD is also controlled by Hfq, ArcA, LoiA, Fis and Hu at the transcriptional initiation level [79][80][81][82], by Gre factors acting during transcriptional elongation [83], as well as by Fur and FliZ acting at the transcriptional/post-translational and post-translational levels, respectively [84,85]. Whereas the expression or amount of HilD is controlled negatively by

PLOS PATHOGENS
H-NS, IscR, SsrB/SsrA and PhoP/PhoQ at the transcriptional level [79,[86][87][88][89], by StdE/StdF at the post-transcriptional level [90], by FnrS and ArcZ at the translational level [13], by CRP-cAMP at the post-translational level [91], as well as by the Lon protease and CpxR/CpxA that decrease the stability of HilD [78,92]. Additionally, controversial reports indicate that the acetyltransferase enzyme Pat decreases or increases the stability and controls the DNA-binding activity of HilD [93][94][95]. On another hand, expression of HilE is controlled positively by FimYZ, PhoP/PhoQ, PhoB/PhoR and LeuO at the transcriptional level [56][57][58][59], whereas it is controlled negatively by MIc and IsrM at the transcriptional and translational levels, respectively [96]. Thus, the three main components of I1-FFL SirA/BarA-HilE-HilD could connect in a dynamic way to this regulatory motif with the rest of the regulatory network controlling physiology and virulence in Salmonella.
Only a few I1-FFLs that control gene expression have been well-defined experimentally in bacteria. One of the best characterized is that involved in galactose metabolism in E. coli, where CRP-cAMP controls the expression of the gal and mgl genes by two opposite pathways, via direct positive regulation and negative regulation through GalS [70,73]. Additionally, in Pseudomonas aeruginosa, a quorum sensing system controlling virulence-related phenotypes is constituted by an I1-FFL, where LasR activates the expression of LasI, as well as that of RsaL, which in turn represses expression of LasI [97]. Furthermore, also in P. aeruginosa, the alternative sigma factor σ 22 activates the expression of the virulence-associated enzyme AlgC, as well as that of the sRNA ErsA that represses translation of algC [98]. In addition, in Rhodobacter sphaeroides, two I1-FFLs are involved in the expression of photosynthesis-related genes. In one case, MppG exerts opposite transcriptional regulation on these genes, direct activation and indirect repression through AppA and PpsR, respectively [99], whereas in the other case, PrrA directly activates transcription of these genes but also represses their translation through the sRNA PcrZ [100,101]. Finally, in S. Typhimurium, the PhoP/PhoQ two-component system activates the expression of mgtC, encoding a virulence-associated protein, and that of the sRNA AmgR, while AmgR destabilizes mgtC mRNA [102], which also constitutes an I1-FFL.
Our study reveals a novel function for an I1-FFL, that is, fine tuning regulation to reduce the growth cost imposed by simultaneous expression of a high number of virulence genes. Our results and those from a previous study [20] provide evidence indicating that I1-FFL SirA/BarA-HilE-HilD is necessary for intestinal colonization by Salmonella. Moreover, our findings illustrate the integration of ancestral (e.g. SirA/BarA) and acquired regulators (e.g. HilD and HilE) into a specific regulatory motif, which can lead to the expansion of regulatory networks during evolution.

Ethics statement
Animal experiments were conducted according to the standard operating protocols approved by the International Committee for Animal Care and Use from CICUAL-UNAM and by the Official Mexican Norm NOM-062-Z00-1999.

Enzymatic assays
β-galactosidase and chloramphenicol acetyltransferase (CAT) assays, as well as protein quantification to calculate the respective enzymatic specific activity, were performed as described previously [107,108].

Flow cytometry analysis
Samples containing~10 7 cells were taken from bacterial cultures grown as described above.
After washing with 0.22-μm-pore-size filtered 1X phosphate-buffered saline (PBS), cells were fixed for 20 min at room temperature in 400 μl of 2% (wt/vol) paraformaldehyde (Sigma) in 1X PBS. Fixed cells were centrifuged and resuspended in one ml of 1X PBS. GFP fluorescence was assessed on a FACSCanto II cytometer (BD Biosciences) equipped with the FACSDiva software (BD Biosciences). Data from 100 000 events were analyzed with the FlowJo v10 software (Tree Star Inc). Cells carrying the gfp reporter gene without a promoter were used to define the background level of fluorescence. Data shown represent the frequency of GFP positive cells.

Electrophoretic mobility shift assays
The electrophoretic mobility shift assays (EMSAs) followed published procedures [109,110]. His-tagged CsrA (CsrA-H6) from E. coli was purified as described previously [111]. Note that CsrA from E. coli and S. Typhimurium are identical. RNA was synthesized in vitro using the RNA Maxx Transcription Kit (Agilent Technologies). PCR fragments used as templates in transcription reactions contained a T7 promoter and hilE sequence extending from +153 to +280 relative to the start of transcription. Gel-purified RNA was 5 0 -end labeled with [γ-32 P]-ATP (7,000 Ci/mmol). RNA suspended in Tris-EDTA (TE) buffer was heated to 85˚C for 3 min followed by slow cooling at room temperature for 10 min. Binding reactions (10 μl) contained 10 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 100 mM KCl, 200 ng/μl yeast RNA, 0.2 mg/ml bovine serum albumin (BSA), 7.5% glycerol, 20 mM dithiothreitol (DTT), 0.2 nM RNA, CsrA-H6 (various concentrations), and 0.1 mg/ml xylene cyanol. Competition assay mixtures also contained unlabeled competitor RNA. Reaction mixtures were incubated for 30 min at 37˚C to allow CsrA-RNA complex formation. Samples were then fractionated through native 10% polyacrylamide gels using 0.5X Tris-borate-EDTA (TBE) as the gel running buffer. Radioactive bands were visualized with a Typhoon 9410 phosphorimager (GE Healthcare) and quantified using ImageQuant 5.2 software. Apparent equilibrium binding constants (K d ) of CsrA-hilE RNA interaction were calculated as described previously [112].

Competitive growth experiment
The WT S. Typhimurium SL1344 strain and its isogenic mutants were grown in 5 ml of LB at 37˚C with shaking, until the cultures reached an OD 600nm of 0.6. Next, mixed cultures were started by inoculating an equal amount of the initial culture of the respective strains (500 μl + 500 μl) in 250-ml flasks containing 50 ml of fresh LB without antibiotics, which were incubated at 37˚C with shaking for 8 h.

Competitive index assay
Pathogen-free BALB/c female mice (6-to 7-week-old) were obtained from the Experimental Medicine Research Unit, School of Medicine, UNAM, México. Maintenance, streptomycin treatment (50 mg) and euthanasia of the mice were performed as described previously [113].
Overnight cultures of the WT S. Typhimurium SL1344 strain and its isogenic ΔhilE, ΔhilD and ΔhilE ΔhilD mutants were diluted 1:100 in 5 ml of fresh LB and incubated at 37˚C with shaking for about 3 h. Then, bacterial suspensions containing 0.5 x 10 7 CFUs/ml of each the WT strain and the respective mutant (WT + ΔhilE, WT + ΔhilD or WT + ΔhilE ΔhilD) were prepared in 1X PBS. Next, groups of eight streptomycin-pretreated mice were infected by orogastric route with 100 μl of the corresponding bacterial suspension. At 3 days post-infection, cecum and feces were harvested aseptically and homogenized in 1 ml of sterile and cold 1X PBS, solubilized in 2% Triton X-100, serially diluted and differentially plated on LB agar plates containing streptomycin or kanamycin to determine CFUs/mg. The number of CFUs for each strain was obtained as described in Competitive growth experiment. The competitive index was calculated by dividing the total CFUs of mutant strain by the total CFUs of the WT strain.

Statistical analysis
Data analyses were performed using GraphPad Prism version 6.0c for Mac OS X, GraphPad Software, La Jolla, California USA, using One-way ANOVA followed by Tukey's multiple comparisons posttest. P-values < 0.05 were considered statistically significant.