Sequestration of histidine kinases by non-cognate response regulators establishes a threshold level of stimulation for bacterial two-component signaling

Abstract

Bacterial two-component systems (TCSs) consist of a sensor histidine kinase (HK) that perceives a specific signal, and a cognate response regulator (RR) that modulates the expression of target genes. Positive autoregulation improves TCS sensitivity to stimuli, but may trigger disproportionately large responses to weak signals, compromising bacterial fitness. Here, we combine experiments and mathematical modelling to reveal a general design that prevents such disproportionate responses: phosphorylated HKs (HK~Ps) can be sequestered by non-cognate RRs. We study five TCSs of Mycobacterium tuberculosis and find, for all of them, non-cognate RRs that show higher affinity than cognate RRs for HK~Ps. Indeed, in vitro assays show that HK~Ps preferentially bind higher affinity non-cognate RRs and get sequestered. Mathematical modelling indicates that this sequestration would introduce a 'threshold' stimulus strength for eliciting responses, thereby preventing responses to weak signals. Finally, we construct tunable expression systems in Mycobacterium bovis BCG to show that higher affinity non-cognate RRs suppress responses in vivo.

Fig. 1. Binding affinities of phosphorylated MtrB for cognate and non-cognate RRs.

Normalized fluorescence intensity obtained from microscale thermophoresis (see “Methods” section) of 50 nM of fluorescently tagged MtrB post autophosphorylation, P~MtrB-GFP, as a function of the concentration of the titrant RR (concentration range): A MtrA (0.45 nM to 15 μM), B NarL (0.46 nM to 15 μM), C TcrX (0.61 nM to 10 μM), D KdpE (3.1 nM to 50 μM), E PhoP (0.76 nM to 25 μM), and F TcrA (0.31 nM to 10 μM). The resulting K_D values are indicated. Curves are best-fits and symbols are mean ± S.E.M (n = 4 independent experiments for MtrB~P with TcrX and n = 3 independent experiments for the remaining plots).

Fig. 2. Binding affinities of several HKs for their cognate and non-cognate RRs.

Affinities measured as in Fig. 1 for phosphorylated (A) PhoR, (B) KdpD, (C) DevS, and (D) PrrB, for their respective cognate (gray) and some non-cognate (green) RRs implicated in crosstalk with the HKs. The affinities as mean ± S.E.M. from at least three repeats are indicated. Detailed measurements leading to the affinity estimates are in Figs. S8 and S9, including for any non-cognate RRs with weaker affinities than the cognate ones. The error bars represent mean ± S.E.M (n = 4 independent experiments for KdpD~P with KdpE and DevS~P with DevR interaction, n = 3 independent experiments for remaining bar plots).

Fig. 3. Phosphotransfer kinetics from MtrB to MtrA with and without sequestration in vitro.

Time course assay of the phosphotransfer from the phosphorylated HK MtrB~P to the cognate RR MtrA in the (A) absence or (B) presence of the non-cognate RR NarL-mRuby. The concentrations used were 50 pmol for MtrB and NarL and 100 pmol for MtrA. C The same as (B) but in the absence of MtrA. M represents marker and A represents autophosphorylation control of MtrB~P. Top panels in A–C are autoradiograms and bottom panels are the corresponding Coomassie Brilliant Blue (CBB) stained gels. D–F Densitometric analysis of the time course assays in A–C, respectively, performed using autoradiograph band intensities normalized by the same band in the CBB stained gel. The autophosphorylation control was used to normalize the intensities of the individual bands. Blue symbols represent MtrB~P and red symbols MtrA~P. Lines represent best-fits of our model (“Methods” section). The error bars represent mean ± S.E.M (n = 3).

Fig. 4. Schematic of the mathematical model.

The model considers an extracellular stimulus triggering the autophosphorylation of HK and activating a TCS pathway. The phosphorylated HK can transfer the phosphoryl group to its cognate RR, which can bind DNA and trigger a response including the synthesis of the HK and RR proteins, marking positive autoregulation. The phosphorylated HK could bind non-cognate RRs (red) preferentially, when the latter have higher affinity for the HK than its cognate RR, resulting in HK sequestration and the suppression of cognate signaling. Only with a sufficiently strong stimulus does sufficient HK autophosphorylation result leading to cognate RR binding despite sequestration and the mounting of a response through the cognate pathway. Equations (1–19) list the reaction events in this model. The rate equations are in “Methods” section.

Fig. 5. Model predictions of TCS signal transduction and the impact of sequestration.

A Representative inputs, I, indicating strong but short-lived (black) and weak but extended (red) stimuli. B The corresponding outputs without (solid lines) and with (dashed lines) sequestration by a non-cognate RR. C The peak of the response (O_max) as a function of the maximum input, I₀, for different extents of sequestration, determined by the ratios of the non-cognate RRs, RR_nc, to the cognate RR, RR_c, indicated. D O_max as a function of the ratio RR_nc/RR_c for different I₀. E O_max as a function of the signal half-life, τ, for different values of RR_nc/RR_c. F O_max as a function of RR_nc/RR_c for different values of τ. (τ is in minutes throughout.) Heatmaps showing O_max as functions of I₀ and τ in the (G) absence or (H) presence of non-cognate RRs, indicating the threshold stimulation for response shifting to higher I₀ and τ with sequestration. Corresponding calculations for the total response, O_total, are in Fig. S13. Model predictions were obtained by solving Eqs. (20–47) (“Methods” section) using parameter values listed in Table S4.

Fig. 6. Advantage of sequestration over reduction in phospotransfer rate.

A The peak of the response (O_max) as a function of the maximum input, I₀, for different extents of sequestration, determined by the ratios of the non-cognate RRs, RR_nc, to the cognate RR, RR_c, indicated. B O_max as a function of I₀ for different phosphotransfer rates indicated. C O_max as a function of I₀ for different ratios of the non-cognate RR_nc/RR_c indicated, in the absence of positive autoregulation. Here, we also set protein degradation rates to zero, to eliminate the threshold introduced by the lack of proteins. Model predictions were obtained by solving Eqs. (20–47) (“Methods” section) using parameter values listed in Table S4.

Fig. 7. Effect of HK sequestration on the cognate TCS response in vivo.

Effect of MtrB sequestration by two non-cognate RRs, a strong binder NarL and a weak binder KdpE examined by monitoring expression changes for MtrA-specific target dnaA in vivo. Overexpression of (A) narL in M. bovis BCG strain under tetracycline-inducible promoter using 10 ng/ml or 50 ng/ml aTC at OD₆₀₀ ~ 0.5 for 8 h and (B) the corresponding expression level of dnaA. Analogous experiments with the induction of (C) kdpE using 50 ng/ml aTC at OD₆₀₀ ~ 0.5 for 8 h and (D) the associated dnaA expression. Gene expression was normalized to the expression levels of 16 s rRNA, followed by the expression levels of the respective genes in the strain carrying only the vector pTic6 (vector control). Data represent n ≥ 3 biologically independent experiments. P values evaluated using one-tailed Student’s t test).