Induction and relaxation dynamics of the regulatory network controlling the type III secretion system encoded within Salmonella pathogenicity island 1

Abstract

Bacterial pathogenesis requires the precise spatial and temporal control of gene expression, the dynamics of which are controlled by regulatory networks. A network encoded within Salmonella Pathogenicity Island 1 controls the expression of a type III protein secretion system involved in the invasion of host cells. The dynamics of this network are measured in single cells using promoter-green fluorescent protein (gfp) reporters and flow cytometry. During induction, there is a temporal order of gene expression, with transcriptional inputs turning on first, followed by structural and effector genes. The promoters show varying stochastic properties, where graded inputs are converted into all-or-none and hybrid responses. The relaxation dynamics are measured by shifting cells from inducing to noninducing conditions and by measuring fluorescence decay. The gfp expressed from promoters controlling the transcriptional inputs (hilC and hilD) and structural genes (prgH) decay exponentially, with a characteristic time of 50-55 min. In contrast, the gfp expressed from a promoter controlling the expression of effectors (sicA) persists for 110+/-9 min. This promoter is controlled by a genetic circuit, formed by a transcription factor (InvF), a chaperone (SicA), and a secreted protein (SipC), that regulates effector expression in response to the secretion capacity of the cell. A mathematical model of this circuit demonstrates that the delay is due to a split positive feedback loop. This model is tested in a DeltasicA knockout strain, where sicA is complemented with and without the feedback loop. The delay is eliminated when the feedback loop is deleted. Furthermore, a robustness analysis of the model predicts that the delay time can be tuned by changing the affinity of SicA:InvF multimers for an operator in the sicA promoter. This prediction is used to construct a targeted library, which contains mutants with both longer and shorter delays. This combination of theory and experiments provides a platform for predicting how genetic perturbations lead to changes in the global dynamics of a regulatory network.

Figure 1

The regulatory network encoded within Salmonella Pathogeneity Island 1 (SPI-1) is shown. The SPI-1 T3SS is induced by a number of environmental and cell state signals, such as osmolarity, Mg2+, and extracellular phosphate, which differentially affect the regulators hilD, hilC, and hilA (Ellermeier and Slauch, 2007). These transcription factors co-regulate each other and control different operons internal to SPI-1. Only HilA regulates the expression of genes forming the inner membrane ring and shaft (prg), but all three converge at two promoters controlling a long operon that begins with the invF gene. There is a promoter internal to this operon that is activated by InvF when bound to the SicA chaperone(Darwin and Miller, 1999, 2000, 2001). The chaperone is sequestered by the SipC protein until it is exported, after which SicA can bind to InvF, leading to the upregulation of additional effector proteins. The colors correspond to the temporal data shown in Figure 2. The cryo EM structure of the needle complex is shown (Marlovits et al., 2004).

Figure 2

The SPI-1 promoters are induced in a temporal order. The hilD (orange), hilC (red), prgH (purple) and sicA (blue) promoters are transcriptionally fused to green fluorescent protein and introduced into cells on a plasmid. Cells are grown in non-inducing media (L Broth) and diluted into inducing media and fluorescence is measured in single cells as a function of density (OD₆₀₀). The average gated fluorescence of the population is calculated for the fraction of the population that turn on at the end of the induction experiment. The average is then normalized by the maximum (Kalir et al., 2001). The data represents four replicate experiments performed on different days for each promoter. Each data point is an aliquot from these 16 growth experiments. The lines are drawn for each promoter using a polynomial averaging algorithm. The dashed line is drawn as a reference for when the promoter reaches 20% of its maximum.

Figure 3

SPI-1 promoters follow graded, all-or-none, or hybrid induction dynamics. Gated cytometry data is shown for the hilC, hilD, prgH, and sicA promoters transcriptionally fused to green fluorescent protein. Cells are grown in non-inducing media (L broth with 0.9 M NaCl) and then shifted into inducing media (LB broth with 3 M NaCl) (Materials and Methods). Data is shown for each promoter pre-induction (first row), at intermediate induction (second row), and fully induced (third row). The OD₆₀₀ of these samples are: hilC (0.13, 0.80, 1.9), hilD (0.11, 0.75, 1.8), prgH (0.12, 1.5, 2.06), and sicA (0.12, 1.4, 2.07). Both the hilC and hilD promoters have a graded response to the shift in condition, where the entire population is induced uniformly. In contrast, the prgH promoter has an all-or-none response and the sicA promoter has a hybrid response with features of both all-or-none and graded dynamics. After the cells are fully induced, they are shifted back into non-inducing conditions. Fluorescence from the hilD, hilC, and prgH promoters decays rapidly to the pre-induced level. However, expression from the sicA promoter persists strongly. After dilution into L Broth, time points are shown after 50 and 140 minutes. The dashed lines are drawn as a guide for the eye.

Figure 4

Expression from the sicA promoter persists after cells are shifted into non-inducing conditions. Cells are grown in inducing media (3M NaCl LB Broth) until the network is fully activated after 400 minutes (Materials and Methods). Cells are then shifted into non-inducing L media (0.3M NaCl LB Broth) and grown for 140 minutes. The decay of fluorescence is shown after the cells are shifted into non-inducing conditions (t = 0 min). The fluorescence values are normalized by the maximum fluorescence prior to dilution. On a semi-log plot, exponential decay appears as a straight line, as is the case for the hilC (orange circles), hilD (red +), and prgH (pink triangles) promoters. There is a significant delay in the decay of the sicA promoter (white diamonds). The straight lines are the best fit to an exponential equation, as determined using linear regression algorithm. It is noteworthy that the sicA and hilC promoters have almost equal fluorescence prior to the shift to non-inducing conditions. Each data point represents the average of four experiments performed on different days and the standard deviation of these experiments is shown as error bars.

Figure 5

An inducible sicA plasmid was constructed and tested for complementation in a ΔsicA knockout strain. (A) To reconstitute the feedback loop, the sicA promoter and gene were cloned together from the Salmonella genome and ligated into PBAD30 at an introduced XhoI site that eliminates the araC gene and P_BAD promoter (top). To create an inducible system without the feedback loop, the sicA gene was cloned into the PBAD30 plasmid (bottom). Different ribosome binding sites were tested to obtain an inducible construct. (B) Cytometry data is shown for the ΔsicA kockout containing the psicA_gfp reporter on a ColE1 plasmid. Data is shown for (bottom to top) 2, 4, 6, and 8 hours after shifting cells into inducing conditions. No activity from the sicA promoter is observed when ΔsicA cells carry the empty PBAD30 plasmid as a control (top left). When the sicA gene is driven by the wild-type sicA promoter (P30.psicA.sicA), induction is recovered (bottom left). When the sicA gene is fused to an arabinose-inducible promoter (PBAD30.sicA) the sicA promoter is induced in the presence of 1.33 mM arabinose (bottom right), but not in the absence of arabinose (top right).

Figure 5

An inducible sicA plasmid was constructed and tested for complementation in a ΔsicA knockout strain. (A) To reconstitute the feedback loop, the sicA promoter and gene were cloned together from the Salmonella genome and ligated into PBAD30 at an introduced XhoI site that eliminates the araC gene and P_BAD promoter (top). To create an inducible system without the feedback loop, the sicA gene was cloned into the PBAD30 plasmid (bottom). Different ribosome binding sites were tested to obtain an inducible construct. (B) Cytometry data is shown for the ΔsicA kockout containing the psicA_gfp reporter on a ColE1 plasmid. Data is shown for (bottom to top) 2, 4, 6, and 8 hours after shifting cells into inducing conditions. No activity from the sicA promoter is observed when ΔsicA cells carry the empty PBAD30 plasmid as a control (top left). When the sicA gene is driven by the wild-type sicA promoter (P30.psicA.sicA), induction is recovered (bottom left). When the sicA gene is fused to an arabinose-inducible promoter (PBAD30.sicA) the sicA promoter is induced in the presence of 1.33 mM arabinose (bottom right), but not in the absence of arabinose (top right).

Figure 6

A generalized schematic of the secretion-control circuit is shown for the topology that appears in Salmonella and two alternative topologies (2 and 3). Topology 1 shows the split positive feedback loop motif, where the transcription factor x (InvF) appears external of the feedback loop, but the activating chaperone y (SicA) is internal to the loop. The parameters used in the model are shown (Equations 1–3 and Table 1). Topology 2 is a simple cascade, where both the transcription factor and chaperone are outside of the internal promoter, thus eliminating any feedback. This topology is roughly recreated when sicA is placed under inducible control and inducer is added at the time when invF is expressed (Figures 5A and 7). Topology 3 is a complete feedback loop, where all of the necessary components are internal to the loop.

Figure 7

The persistence of expression from the sicA promoter is eliminated by a disruption of the positive feedback loop. Data is shown for the ΔsicA deletion mutant, where the sicA gene is complemented either by an inducible construct (PBAD30.sicA, Figure 5A) or under the control of the sicA promoter (P30.psicA.sicA). Cells are grown in inducing conditions and then shifted into non-inducing conditions (time = 0 minutes) and the decay in fluorescence is measured. When the sicA promoter is under the control of an inducible promoter, the total fluorescence from the reporter reaches approximately the same maximum (Figure 5B), but the delay is eliminated (triangles). When the feedback loop is reconstituted on the plasmid, the delay is recovered (squares). The lines represent the fit to the model (Equations 1–3 and Table 1). Each data point is the average of six experiments performed on different days and the error bars represent the standard deviation. The data is normalized by the maximum fluorescence under inducing conditions.

Figure 8

A robustness analysis of the model is performed. Each parameter is independently varied from the nominal set (Table 1). For each set of parameters, Equations 1–3 are solved and three values are recorded from the simulation: the maximum fluorescence without secretion (top graph, red line), the maximum fluorescence with secretion (top graph, green line), and the half-life for the decay of fluorescence (τ_½, bottom graph). Very few of the parameters have the capacity to affect the decay half-life without resulting in a circuit that cannot be induced by secretion. Note that when there is no secretion (k_Z = 0), the positive feedback loop is broken and the delay is eliminated. The units for the parameters are provided in Table 1.

Figure 9

Mutations to the SicA:InvF binding site in the internal promoter vary the strength of the positive feedback loop, resulting in different relaxation times. (A) The −50 to −30 region of the sicA promoter is shown, which includes the previously identified SicA:InvF binding site (Darwin and Miller, 2001). For each sequence, the resulting empirical half-life (τ_½) is shown in addition to the binding constant (K) derived from fitting the mutant data to the model. The errors are reported as the standard deviation of six experiments performed on different days. When the binding site from the sicA promoter is mutated to the sopE site, there is no change in the observed half-life. A small library was then created, where the differences between the sicA and sopE operators were randomly mutated (library). The resulting mutants have a significant diversity of decay times (A.1 to A.7). (B) The decay data is shown for three representative mutants. The wild-type sicA promoter (top black line) and the PBAD30.sicA (lower black line) data are shown as a reference (Figure 7). The mutant with the longest decay time (A.1, red triangles) and two with the fastest decay times (A.6, blue diamonds and A.7, red squares) are shown. The error bars are the standard deviation from six experiments performed on different days.

Figure 9

Mutations to the SicA:InvF binding site in the internal promoter vary the strength of the positive feedback loop, resulting in different relaxation times. (A) The −50 to −30 region of the sicA promoter is shown, which includes the previously identified SicA:InvF binding site (Darwin and Miller, 2001). For each sequence, the resulting empirical half-life (τ_½) is shown in addition to the binding constant (K) derived from fitting the mutant data to the model. The errors are reported as the standard deviation of six experiments performed on different days. When the binding site from the sicA promoter is mutated to the sopE site, there is no change in the observed half-life. A small library was then created, where the differences between the sicA and sopE operators were randomly mutated (library). The resulting mutants have a significant diversity of decay times (A.1 to A.7). (B) The decay data is shown for three representative mutants. The wild-type sicA promoter (top black line) and the PBAD30.sicA (lower black line) data are shown as a reference (Figure 7). The mutant with the longest decay time (A.1, red triangles) and two with the fastest decay times (A.6, blue diamonds and A.7, red squares) are shown. The error bars are the standard deviation from six experiments performed on different days.