Deciphering Parameter Sensitivity in the BvgAS Signal Transduction

Tarunendu Mapder; Srijeeta Talukder; Sudip Chattopadhyay; Suman K. Banik

doi:10.1371/journal.pone.0147281

Abstract

To understand the switching of different phenotypic phases of Bordetella pertussis, we propose an optimized mathematical framework for signal transduction through BvgAS two-component system. The response of the network output to the sensory input has been demonstrated in steady state. An analysis in terms of local sensitivity amplification characterizes the nature of the molecular switch. The sensitivity analysis of the model parameters within the framework of various correlation coefficients helps to decipher the contribution of the modular structure in signal propagation. Once classified, the model parameters are tuned to generate the behavior of some novel strains using simulated annealing, a stochastic optimization technique.

Citation: Mapder T, Talukder S, Chattopadhyay S, Banik SK (2016) Deciphering Parameter Sensitivity in the BvgAS Signal Transduction. PLoS ONE 11(1): e0147281. https://doi.org/10.1371/journal.pone.0147281

Editor: Attila Csikász-Nagy, King’s College London, UNITED KINGDOM

Received: October 13, 2015; Accepted: January 2, 2016; Published: January 26, 2016

Copyright: © 2016 Mapder et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and its Supporting Information files.

Funding: TM, SC, and SKB acknowledge financial support from CSIR (Council of Scientific & Industrial Research), India [01(2771)/14/EMR-II]. SKB is thankful to Bose Institute, Kolkata, India for research fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Living systems sustain in a diverse and dynamically changing niche that they have to cope with to survive. As a result, every organism adopts specialized communication machinery that helps in responding to alteration of the immediate environment. Any rapid or slow change in the surroundings is taken care of through intracellular signal transduction pathways and a number of genetic switches [1]. The signal transduction pathways comprise of some specialized motifs to carry out the process of intracellular communication. Two-component system is one such signaling motif prevalently found in bacteria [2–6]. A typical bacterial two-component system comprises of a trans-membrane sensor protein along with a cognate cytoplasmic response regulator protein. Any change in the immediate surroundings is sensed by the sensor protein, which then communicates the information downstream to its cognate partner through the mechanism of phosphorelay. The response regulator then regulates one or several downstream genes in response to the change in the environment.

The diversity of extracellular environment is very high for the microorganisms that invade into the host as pathogens and proliferate. They have to deal with both the environments: inside and outside the host. The microorganism of the current study, Bordetella pertussis—a gram-negative human pathogen, shows dormancy in the atmospheric environment (∼25°C) and becomes virulent inside host (∼37°C) [7–9]. Like many other prokaryotes, B. pertussis adopts the environmental distortion through a prominent two-component signal transduction cascade, the BvgAS. The key components of the BvgAS two-component system are BvgS and BvgA, the sensor and the response regulator protein, respectively. The expression of several virulent factors, like toxins and adhesins, are mediated by the BvgAS two-component system. The membrane bound sensor kinase, BvgS, encounters temperature dependent activation and down-modulation only through the presence of modulators like MgSO₄, nicotinic acid or reduced temperature. The BvgAS sensory transduction system shows three different phases: Bvg⁺, Bvg⁻ and Bvgⁱ through gene regulation in response to stimuli [7–9]. The regime of the intermediate (Bvgⁱ) phase is narrow than the repressed (Bvg⁻) phase or activated (Bvg⁺) phase due to the sharp switch of the BvgA population. Each of the phases has a unique pattern of gene regulation. At Bvg⁺ phase, the virulence-activated genes (vags) fhaB, ptxA and bvgAS itself show maximum expression. In contrast, at Bvg⁻ phase, the virulence-repressed genes (vrgs) flaA, frlAB are expressed at the maximum level, but no vags show good expression. The Bvgⁱ phase is characterized by the maximal expression of bvgAS, bipA and fhaB, and nominal expression of vrgs and ptxA. In the present study, we focus on bipA and fhaB along with bvgAS to observe the temperature mediated switch from dormant to virulent phase.

Depending on the transcription factor binding affinity, the variety of the Bvg-regulated genes are classified into four classes [7–9]. Class 1 genes include cyaA and ptxA, containing low-affinity binding sites and are activated at high level of phosphorylated BvgA. Class 2 gene fhaB possesses high-affinity BvgA binding site and produces transcripts at a very small level of BvgA-P. The unique class 3 gene bipA starts transcription at a moderately low level of BvgA-P bound at the low-affinity binding site, but get repressed at a high level of BvgA-P. frlAB belongs to class 4 gene and shows repression at high BvgA-P. The class 3 and class 4 gene expression are not observed in B. pertussis through temperature elevation. Only class 3 gene are expressed in B. pertussis for intermediate level of MgSO₄. Thus, at a low level of BvgA, B. pertiussis expresses class 1, 2 and 3 minimally but class 4 genes maximally. Class 2 and 3 genes show a high level of transcription in contrast with the low level of class 1 and 4 expression at the moderate level of BvgA. Finally, at a high level of BvgA, class 1 and 2 genes show maximum and class 4 genes minimum level of expression with low level of class 3 gene expression.

The well known BvgAS motif is studied here in accord with the network parameter sensitivity. The functionality of the BvgAS signal transduction motif and its effect on the downstream gene regulation is well reported through experimental work [7–9], thus providing a scope to analyze the network from sensitivity analysis and optimization point of view. In this connection it is important to mention that several theoretical approaches have been undertaken to explore the signal transduction mechanism in typical bacterial two-component system [10–21]. However, few theoretical formalism have been developed to address the underlying signaling mechanism in B. pertussis and its effect on the downstream differential gene regulation. In one of our earlier communications, we have theoretically identified the temperature mediated molecular switch that controls the signaling mechanism [22]. In addition, theoretical analysis have been made to understand the role of positive feedback on target gene regulation [22, 23]. A preliminary level of sensitivity analysis suggests that the rate constants associated with the kinase and phosphatase activities are most sensitive among all the rate parameters [22]. To extend these earlier reports further and to understand the role of individual rate parameters on the model output we undertake a theoretical approach in the present study that incorporates correlation coefficient based sensitivity analysis and stochastic optimization. The modular approach we have adopted in the present work is the following. In the first module, we classify the kinetic rate parameters associated with the model according to their sensitivity. We identify the set of sensitive model parameters as a function of the input signal. In the next module, we invoke stochastic optimization technique to generate the sharp molecular switch. Finally, the optimized parameter set has been used to reproduce the features of some novel experimental results [24].

Methods

The bvg operon

Many of the pathogenic secretions of B. pertussis are controlled by phosphorylated BvgA, a member of the BvgAS two-component system and is encoded by the bvg operon [7, 8]. In bvg operon, four promoters (P₁, P₂, P₃, and P₄) together control the production of BvgS and BvgA [25–27] (see Fig 1). Out of the four promoters, P₂ shows constitutive behavior in absence of any external stimulus. However, as B. pertussis experiences temperature rise in the surroundings, activity of P₂ goes down. Under the same condition, the rest of the promoters get activated. Activity of P₃ is very low under induction [27] and has been excluded from our model. Similarly, contribution of P₄ has been excluded from our model as P₄ produces anti-sense RNA whose target has not been yet identified. In this connection, recent work by Hot et al. [28] is worth mentioning, where the authors have identified an anti-sense RNA bprJ2, regulated by BvgAS TCS, with unknown functionality.

Download:

Fig 1. BvgAS signal transduction motif.

The signal transduction motif is composed of the phosphotransfer and autoregulation modules. The temperature acts as inducer of the autophosphorylation of the dimer of the sensor kinase protein BvgS (blue). Similarly, dimerized BvgA represents for response regulator in red. P in black sphere stands for phosphate group. Note that, sensor kinase acts both as source and sink for the phosphate group.

https://doi.org/10.1371/journal.pone.0147281.g001

The BvgAS two-component system

The phosphorylated dimer of the cytosolic response regulator protein BvgA binds at the promoter site as a transcription factor and catalyzes the promoter activation. Upon activation, RNA polymerase transcribes the polycistronic mRNA m, which on translation accumulates the pool of the monomeric cognate pair proteins BvgS and BvgA. BvgS, after dimerization, spans in the transmembrane region to sense the external stimulus. In the present work, we denote the dimers of BvgS and BvgA as S₂ and A₂, respectively. Although the sensor and the response regulator proteins first get expressed as a monomer and then dimerize, we do not consider the dimerization kinetics in our model as only the dimer forms of the proteins are of functional interest here. As the temperature in the surroundings increases, S₂ gets autophosphorylated at the histidine residue to form S_2P. S_2P then transfers the phosphate group to the aspertate residue of the cognate A₂ through kinase activity, thus producing A_2P. A_2P on the other hand, gets dephosphorylated to A₂ due to the phosphatase activity of S₂. As a result, the bifunctional sensor protein BvgS helps in producing the pool of phosphorylated response regulator BvgA which in turn autoregulates its operon as well as regulates the expression of several downstream genes (see Fig 1). For detailed kinetic scheme and parameter set, we refer to Table 1. These parameter set was used earlier to understand the mechanism of molecular switch in B. pertussis [22]. Based on the above information the signal transduction module in BvgAS TCS can thus be divided into two parts: the autoregulation motif and the phosphotransfer motif, typical characteristics of bacterial TCS [11, 22]. These two motifs together generate a sharp molecular switch under the induction of temperature increase in the surroundings [22].

Download:

Table 1. List of kinetics schemes and the values of rate parameters used in the model.

https://doi.org/10.1371/journal.pone.0147281.t001

Experimentally, it has been observed that the ratio of total BvgS (S_T) to total BvgA (A_T) is ≈6 [27] which can be utilized to employ quasi-steady state approximation in analyzing the behavior of the key components of the signal transduction motif at steady state [11, 22]. Considering this, we define (1) Furthermore, we define two dimensionless quantities (2) to analyze our results.

Sensitivity Analysis

To decipher the sensitivity of the rate parameters on the output of the model we use the tools of correlation coefficient in the present study. The Pearson correlation coefficient (CC) is defined as the covariance between the input and the output parameters with a normalization through division by the product of their standard deviations, (3) where r_{k_i β} is the CC of the input parameter k_i and output β. 〈k_i〉 and 〈β〉 are the mean (ensemble average) of k_i and β, respectively, and N is the number of random sampling. If the output rises or falls with the increment of a particular input parameter, it is said to be positively or negatively correlated. The magnitude of the correlation coefficient, which implies the strength of dependency, spans upto ±1 for maximum +ve or -ve association.

The CC can not explain the sensitivity precisely in the case of nonlinear input-output dependencies. For nonlinear but monotonic increasing relation, one can use the Spearman rank correlation coefficient (RCC) which is the calculation of the correlation coefficient after a rank transformation. The partial rank correlation coefficient (PRCC) takes care of the association of an individual input parameter k_i with the output β provided that the dependency of all the other k_j-s have been eliminated. This makes PRCC the most reliable among all the sampling based sensitivity indices. PRCC can be calculated from the rank correlation matrix (), where is the RCC between the i-th and j-th element. The co-factor P_ij of is utilized to calculate PRCC of an input parameter k_i with the output β as (4)

As correlation coefficient measures how much the output of a network is dependent on the particular input parameter, the correlation coefficient is used as an index of sensitivity in this study. The Pearson correlation coefficient (CC), the Spearman rank correlation coefficient (RCC), and the partial rank correlation coefficient (PRCC) have been calculated for a range of signal k_ps and analyzed. For this purpose, we have perturbed each of the input parameters simultaneously and solved the set of coupled rate equations (see S1 Text) to calculate the output β. The distribution of the random perturbations are of Gaussian type whose mean is the base value and the variance is ±5% of the base value. In addition, we have used the data set obtained from 10⁵ independent runs to calculate the correlation coefficients.

Sensitivity analysis of any chemical kinetic network suggests an insight about the priority of the cascade inputs. The utility of parameter sensitivity analysis is to classify the input parameters according to their relative impact on the output [29–32]. In the present study, the phosphorylated fraction of the response regulator (β = A_2P/A_T) is taken as the output variable. As the temperature mediated autophosphorylation rate constant k_ps alters, the dynamics of β changes. In the present work, we perform sensitivity analysis over all the rate parameters (except k_ps) to check the robustness and the relative importance of the parameter set considered for the reaction kinetics involved in the signal transduction.

Stochastic Optimization

In the present work, we implement simulated annealing (SA) [33, 34] to decipher the correct parameter set to reproduce results of some in vitro experiments reported by Jones et al. [24]. At first, we consider the autophosphorylation of the sensor protein (S₂) and phosphotransfer to the response regulator (A₂) as done in the in vitro phosphorylation experiment by Jones et al. [24]. Here, it is important to mention that SA simulation is an algorithmic replica of thermodynamical annealing process [33, 34]. In metallurgical annealing, the metal alloy is taken at a very high temperature and then slowly cooled down to get the most thermodynamically stable state. Similarly, in SA, an algorithmic temperature, called the annealing temperature, is defined. The annealing temperature controls the extent of the search space (or the solution space) that is being sampled. During the simulation a randomly chosen variable is allowed to take a move for each sampling. The maximum step length taken in our simulation is of 5 (minimum) −15 (maximum) % with respect to the value of the particular variable at the previous sampling step. To be explicit, for any parameter k, the update using SA is done by the rule k′ = k + k × (−1)ⁿ × δ × r_n, where k′ is the updated value of k, n is a random integer, δ is the amplitude of allowed change (kept between 0.05 and 0.15), and r_n is a random number between 0 and 1. Using these information a cost or objective function is calculated for the new set of variables after each iteration. With the progress of iteration, the cost profile goes down as the output becomes close to the desired value. The cost function at the i-th step of the iteration is calculated as (5) where β_ex(j) is the experimental value of relative phosphorylation and β_{T_at}(j) denotes the relative phosphorylation at the algorithmic (annealing) temperature T_at for j-th time at the i-th step of the simulation. Similarly, to generate the transcript profile we use the following expression (6) with m_ex(j) and m_{T_at}(j) being the experimental value of transcript and the simulated value of the same at the algorithmic (annealing) temperature T_at, respectively. While going from i-th to i + 1-th SA step, the cost function may increase or decrease. If the cost function decreases, we accept the move. On the other hand, if it increases we do not discard the move outright. Instead, we subject it to the Metropolis test [35]. If the quantity Δ = cost_i − cost_i−1 has a positive value, the probability for accepting the move is determined by the function (7) For positive Δ, F is always between 0 and 1. For each evaluation of F, we invoke a random number ξ (say) between 0 and 1. If F > ξ, we accept the move. Otherwise, the move is rejected. Thus, at very high T_at, F will be close to 1 and most moves will be accepted, such that a greater region of the search space will be sampled. As the simulation proceeds, T_at is decreased by the annealing schedule. Once the correct path towards the global minimum is attained, we need not search the entire space and concentrate on a small region, which will guide us correctly to the global minimum. In other words, as T_at is lowered, a decreasing number of moves pass the Metropolis test. Finally, we recover the correct set of parameter to reproduce the experimental results. In addition, the optimization is carried out until the cost reaches to zero or a low enough steady value and the corresponding data set is taken as optimized set.

The underlying reason to employ stochastic optimization is to optimize the model parameters in a more reliable and efficient manner. The primary target of any optimization method is to minimize a scalar valued objective function or cost function. This makes stochastic optimization technique the most effective method in developing and fabricating a complex system with large number of components. For systems with high nonlinearity, such an approach gets favor over any deterministic method of optimization. This happens due to the implementation of a Markov Chain Monte Carlo search direction in such a way that one can distinguish the global optima from many local optima. To utilize the principles of stochastic optimization, we implement SA technique [33, 34] which has been successfully applied recently to understand the role of different bonding (stacking and hydrogen) interactions on the breathing dynamics of DNA [36, 37].

Results and Discussion

Amplification and Switch

In presence of a stimulus, regulatory networks in all living systems need a switching from off to on state or vice versa. Genetic switch, a typical regulatory system, sometimes gets controlled by the output protein through a feedback loop. The presence of a positive feedback motif in the network makes the switching phenomena sharp. As a result, the network response in terms of the population of the network output shows a sharp growth curve even for a small change in the input signal. Typically, amplification in the signal can be classified into two categories, magnitude amplification and sensitivity amplification [38]. Magnitude amplification occurs in ligand-gated ion channels, where ∼10⁴ ions flow as a single ligand molecule binds to the channel protein. Here, the response is much greater than the stimulus. On the other hand, sensitivity amplification is defined as the fractional change in response with respect to a fractional change in the stimulus. The sensitivity amplification can be calculated locally as [38–40] (8) where the change in the input signal (Δsignal) is infinitesimally small. The sensitivity amplification can be calculated globally from the signal-response characteristic curve (see Fig 1 of [38]). Following Koshland et al, the order of sensitivity can be classified as ultrasensitive, hyperbolic sensitive and subsensitive [38]. A curve is said to be ultrasensitive if a 10%–90% change in the response can be obtained for a very narrow range (4–5 fold) of the signal [38]. In the case of hyperbolic Michaelian response, the range of the signal is ∼81 [38] and for subsensitive amplification, it is about a few thousand [38].

In the present model, the pool of A_2P is developed when both the phosphotransfer motif and the autoregulation motif are functional. In Fig 2A, we show the amplification of the output response β, the fraction of phosphorylated BvgA, as a function of the autophosphorylation rate k_ps. The local sensitivity for this model is (9) Here, and with i and f being the initial and final value, respectively. At lower range of k_ps, one can observe a first order ultrasensitive response that reduces to zero order at a high value of the signal. The quantity S_local approaches 1 for small stimulus and sharply falls to 0 for large stimulus (Fig 2B). This happens as the pool of β gets saturated at the high value of k_ps. If we focus on the global sensitivity with reference to Fig 2A, the amplification of β from 10% to 90% occurs in response to a 10 fold increment of the signal. Thus, the sensitivity amplification does not show an ultrasensitive switch. Absence of ultrasensitivity implies that the molecular switch in B. pertussis lacks co-operativity in the positive feedback operative at the bvg operon.

Download:

Fig 2. Amplification.

A. The % amplification and B. the sensitivity amplification of the output response β with respect to the signal k_ps. Note the logarithmic scale in the abscissae.

https://doi.org/10.1371/journal.pone.0147281.g002

Parameter sensitivity analysis

To understand the significance of model parameters on the generation of the molecular switch, we perform sensitivity analysis for all the rate parameters except k_ps (which is treated as the signal) with respect to β as the output parameter. The positive feedback network shows a sharp switch with respect to k_ps. Thus, sensitivity analysis is performed for a broad range of k_ps values (k_ps = 10⁻³ − 10⁻¹) that signify the different regions of the amplification profile shown in Fig 2A. The results thus obtained help us to comment not only on the sensitivity of the rate parameters but also show how the order of sensitivity gets modified with the switch.

CC, RCC and PRCC calculations are carried out as a measure of sensitivity index. The numerical values of the same for three different values of k_ps (10⁻³, 10⁻² and 10⁻¹) are presented in Table 2. At 10⁻³, β is very low; at 10⁻², β starts to increase and at 10⁻¹, it reaches a high value after a sharp change. The trend of correlation is same in all the three correlation coefficient calculations (CC, RCC, PRCC). In magnitude, CC and RCC are very close representing the linear nature of the output with respect to the input. Since in the calculation of PRCC for a parameter excludes the effect of other variables, the PRCC values are quite higher than that of CC and RCC values.

Download:

Table 2. CC, RCC, and PRCC values for all the input parameters with output β for low, medium and high values of k_ps.

https://doi.org/10.1371/journal.pone.0147281.t002

At k_ps = 10⁻³, k_tp1, k_dm, k_sa, k_dp and k_pf have considerably high correlation coefficient values, where k_tp1, k_sa and k_pf are negatively correlated and the others show positive correlation. All the rate parameters show quite similar trends at k_ps = 10⁻², but at k_ps = 10⁻¹, the order as well as the nature of correlation for some parameters get modified. The sensitive parameter at lower k_ps value remains sensitive, however, the nature of correlation becomes opposite for k_tp1, k_dm and k_dp. At this k_ps value, k_ss, k_dps and k_tf also show high correlation which are practically insensitive at a low value of k_ps. The change in the nature of sensitivity before and after the on state for all the rate parameters are presented in a tabular form in Table 2. With this gross nature of the sensitivity of the parameters, we analyze the nature as well as the order of sensitivity of the rate parameters in the following.

The nature of correlation of k_sa does not change much with the change of k_ps value. k_sa is the rate constant for generation of A₂ from mRNA. As the output β for the calculation of correlation coefficients is inversely proportional to the A₂ concentration, k_sa shows negative correlation (see Fig 3D). Although it remains negatively correlated, the magnitude of its correlation coefficients (CC, RCC and PRCC) get reduced at a high value of k_ps. At high k_ps, concentration of S_2P is abundant and phosphotransfer rate increases. Thus, enhanced production of A₂ implies enhanced phosphotransfer to A₂, which in fact, increases the amount of A_2P. In a way, if one increases the value of k_sa, not only the A₂ concentration increases but a gain in the A_2P concentration also takes place. With these two opposing factors, effectively the order of sensitivity of this parameter reduces at the high value of k_ps.

Download:

Fig 3. Correlation coefficients.

The correlation coefficients CC (solid line), RCC (dashed line) and PRCC (dotted line) as a function of input signal k_ps. A, B, C, D, E, F, G and H are for k_tp1, k_dm, k_ss, k_sa, k_dps, k_tf, k_pf and k_dp, respectively. Note the logarithmic scale in the abscissae.

https://doi.org/10.1371/journal.pone.0147281.g003

If we consider mRNA generation, the system output β varies inversely at low value of k_ps. With the increase of mRNA concentration, both the protein concentrations (S₂ and A₂) increase. Increase in A₂ show direct effect on β. The rise in S₂ concentration also reduces β, as A_2P is dephosphorylated to A₂ by S₂ along with low phosphotransfer (generation of A_2P) due to low concentration of S_2P. Thus, both the rate constants for mRNA generation (k_tp0 and k_tp1) show negative correlation at low k_ps value, with k_tp1 showing high correlation value than k_tp0. In the present work, k_tp1 is the rate of mRNA generation from the active state of a promoter (P_a) and transformation of P_i → P_a involves loss of A_2P. Hence, the effect of k_tp1 on the output would be much greater than k_tp0. At high k_ps, the phosphotransfer motif dominates the whole reaction system thus affecting the mRNA generation. Thus, k_tp1 show positive correlation at high k_ps (Fig 3A). The parameter k_dm deals with degradation of mRNA and shows a trend opposite to that of k_tp1. Hence, it is positively correlated at low k_ps, while showing negative correlation at high k_ps (Fig 3B).

The other two rate parameters that show significant sensitivity after the amplification switch is on are k_ss and k_dps. The first one is related to the generation of S₂ from mRNA and the other one is the rate constant for the autodephosphorylation of S_2P. At high k_ps value, the phosphotransfer process from S_2P to A₂ becomes significant and eventually the output β varies with the concentration of S_2P. As with the increase in k_ss and decrease in k_dps amount of S_2P increases, k_ss and k_dps show positive and negative correlation, respectively (Fig 3C and 3E).

At low k_ps value, the kinase reaction does not show any significant effect on the phosphotransfer kinetics as the amount of S_2P is negligible. However, at the high value of k_ps, the amount of S_2P increases, thus increasing the phosphorylation of A₂. This leads to the generation of A_2P which, in turn, increases the level of β. Hence, it is justified for the rate parameter k_tf to show high positive correlation at the higher value of k_ps (Fig 3F). On the other hand, k_pf is associated with the phosphatase reaction that takes care of the dephosphorylation of A_2P; thus showing negative correlation irrespective of the value of k_ps (Fig 3G).

The degradation of all the proteins (S₂, A₂, S_2P and A_2P) is controlled by the rate parameter k_dp. Hence, it gets correlated with the output (β) with a decreasing trend. In the low range of k_ps, where the protein pool is minuscule, k_dp shows high positive correlation. However, at the saturation of the protein pool at high k_ps value, the associated correlation of k_dp becomes sharply negative (Fig 3H).

As mentioned earlier, here we have used a Gaussian perturbation of ±5% to calculate the sensitivity of the individual parameters. To check whether the parameter set can withstand any larger perturbation (>±5%) we have systematically increased the magnitude of perturbation by increasing the variance of the Gaussian distribution up to ±20%. The list of resultant data are given in S1 Table for k_ps = 10⁻² and shows that the CC, RCC and PRCC values remain consistent for ±5%, ±10%, ±15% and ±20% perturbation.

In vitro assay and stochastic optimization: The mutants

The previous subsection describes how efficiently one can decipher the set of reaction rate constants (k_i-s) on the basis of their sensitivity towards output β. This elucidates the degree of priority of the reactions given in S1 Text. Once the sensitivity of the parameter set is determined, one can target simple motifs present within the complex signaling circuit. One such simple motif is the phosphorylation of BvgA, which has been studied experimentally using some novel mutants [24]. As discussed earlier, activation of the signaling cascade is triggered by the phosphorylation of the transcription factor BvgA by the sensor BvgS. Any alteration through site-directed mutagenesis at the phosphorylation domain of BvgA may influence the phosphorylation kinetics. Two such mutants, T194M and R152H, were employed by Jones et al. in studying the in vitro phosphorylation assay [24]. The wild type and the two mutant BvgA were incubated with GST-′BvgS in the presence of [γ−³²P]-ATP and the phosphotransfer kinetics was monitored for 5 min, and the relative amount of phosphorylation was noted at different time points. The relative amount of [γ−³²P]-ATP thus incorporated in BvgA was quantified using phosphoimager. The total protein concentration of BvgS and BvgA used were 0.8 μM and 2.1 μM, respectively. The experimental results suggest that the mutant R152H behaves almost like the WT strain whereas the mutant T194M is heavily impaired in its ability to get phosphorylated. This happens as both arginine (R) and histidine (H) are positively charged and are good acceptor of the negatively charged phosphate group. On the other hand, when the polar threonine (T) residue is replaced by methionine (M) it heavily impairs the phosphorylation capacity.

To examine the performance of the phosphorylated BvgA, one can construct a transcription assay as BvgA on phosphorylation acts as a transcription factor for its promoter and the promoters of the downstream genes. Among the four classes of downstream genes, we opted for fhaB (class 2) and bipA (class 3) gene as their promoters have high affinity binding sites and show quick response even under low levels of induction [24]. Also, results of in vitro transcription assay are available for these classes of genes [24]. High affinity binding site in these two mutants suggests that affinity of phosphorylated transcription factor (BvgA-P) towards the DNA of these two mutants will be high [41]. This information on the other hand suggests that phosphorylated R152H will have higher affinity compared to phosphorylated T194M. Surprisingly, electrophoretic mobility shift assay (EMSA) exhibits reverse result [24], i.e., binding affinity of T194M towards high affinity binding site is higher compared to R152H. One of the plausible mechanism could be the interaction between the negatively charged backbone of the double helix and the positively charged R152H reduces drastically due to histidine (H). Keeping this in mind, we only consider the experimental results of these two genes for optimization purpose.

Since we have deciphered the sensitivity of the model parameters, we now focus on reproducing the behavior of some novel mutants mentioned earlier. In the present work, we use simulated annealing, a stochastic optimization technique, to estimate the optimal set of the rate parameters involved in the in vitro phosphorylation assay (Table 3). At this point it is important to mention that the process of optimization is computationally expensive as the optimized set of variables is far away from the parameter space from where the sampling is started. The parameter set have to travel a long way to reach the optimized value. Thus, we allowed the parameters to take long step and keep the size up to 15% and carried out 500 independent SA runs. The uncertainty in the output of the SA runs are given in the form of standard deviation in Table 3. The nature of standard deviation suggests that the uncertainty lies within ∼20% of the optimized parameter value. In addition, we have considered only those SA runs where the cost function asymptotically moves towards zero. Otherwise, we do not consider the output of a SA run in our calculation. Keeping this in mind we first generate the profiles of the in vitro phosphorylation assay experiment reported by Jones et. al [24]. The kinetic rate constants, optimized using SA, reproduces the experimental profile (the symbols in Fig 4). The associated cost function and the evolution of rate parameters (only 5 out of 500 trajectories) as a function of SA steps are shown in S1 Fig. Fig 4 shows that the phosphorylation ability of R152H is higher than that of T194M. On a relative scale, R152H and T194M could be phosphorylated ∼80% and ∼30%, respectively, compared to the WT strain.

Download:

Fig 4. In vitro phosphorylation assay.

Profiles of in vitro phosphorylation assay generated using stochastic optimization. The solid, dashed and dotted lines are for WT, R152H and T194M, respectively. The symbols are experimental results due to Jones el al. [24]. In the figure legend Th and Expt stand for theoretical and experimental data, respectively.

https://doi.org/10.1371/journal.pone.0147281.g004

Download:

Table 3. List of optimized parameters used for the simulation of in vitro phosphorylation assay.

Here, x±y stand for the value of optimized parameter x with standard deviation y. The standard deviation is calculated using the data of 500 independent SA runs.

https://doi.org/10.1371/journal.pone.0147281.t003

As discussed earlier, on mutation, the successive binding of phosphorylated BvgA (A_2P) at different promoters gets affected. The three strains WT, R152H and T194M have been used to observe the effect on DNA-protein interaction through in vitro transcription assay. At this point, it is important to mention that the experimental data does not have any error bar that gives an estimation of the uncertainty in the experimental results. The snapshots of the profile of cost function and the parameter optimization with respect to SA steps are shown in S2 and S3 Figs. The optimized parameter set thus obtained could reproduce the qualitative behavior (the solid, dashed and dotted lines in Fig 5) of the in vitro experimental results (the open squares, circles and triangles in Fig 5). As mentioned in the previous paragraph, the mutant R152H activates the target genes later compared to the mutant T194M due to its weak interaction with the promoter region of the target gene.

Download:

Fig 5. In vitro transcription assay.

Profiles of in vitro transcription assay generated using stochastic optimization for A. fhaB and B. bipA. The solid, dashed and dotted lines are for WT, R152H and T194M, respectively. The symbols are experimental results due to Jones el al. [24]. In the figure legend Th and Expt stand for theoretical and experimental data, respectively.

https://doi.org/10.1371/journal.pone.0147281.g005

Download:

Table 4. List of optimized parameters used for the simulation of in vitro transcription assay of fhaB.

Here, x ± y stand for the value of optimized parameter x with standard deviation y. The standard deviation is calculated using the data of 500 independent SA runs.

https://doi.org/10.1371/journal.pone.0147281.t004

Download:

Table 5. List of optimized parameters used for the simulation of in vitro transcription assay of bipA.

Here, x ± y stand for the value of optimized parameter x with standard deviation y. The standard deviation is calculated using the data of 500 independent SA runs.

https://doi.org/10.1371/journal.pone.0147281.t005

Conclusion

The present study undertakes the temperature mediated activation and virulence of BvgAS cascade in the light of sensitivity based optimization. The kinetic model of BvgAS has been simulated for a broad range of sensor protein autophosphorylation that mimics the kinetics of the same quantity due to temperature elevation. The sharp switch in the phosphorylated response regulator is a consequence of positive feedback operative on the bvgAS operon. The BvgA-P sharp switch arising due to the integration of the modular structure (the autoregulation and the phosphotransfer module) shows maximum response to the extra-cellular stimulus.

The development of sharp switch has been extensively investigated in the light of sensitivity analysis coupled with stochastic optimization. Our analysis shows the phosphotransfer module to be the more sensitive compared to the autoregulation module. The parameter sensitivity opens up the avenue to classify and explain the role of model parameters in accord with their influence on the steady state dynamics. Once classified, it is possible to tune the most sensitive parameters to regenerate the experimental profiles computationally. The simulated annealing based stochastic optimization performed on three different strains of BvgA (WT, R152H and T194M) could successfully reproduce the characteristics of in vitro experimental results. In addition, it helps in the understanding of different nature of attenuation or delay during activation of the mutants. Quantification of such delay is biologically important as a delay in the switch due to temperature elevation brings in obstruction in the virulence to be triggered within the host. The obstruction might become operative in the level of protein-protein interaction or the protein-DNA interaction. Although, in vivo fabrication of such synthetic network is difficult, it may be a good starting point in understanding the functionality of stimulus mediated in vivo systems. Future experiments leading to target characterization and quantitative measurements of such interactions will help one to build more efficient models.

Supporting Information

S1 Text. General supplementary information.

The text contains the detailed kinetic mechanism of BvgAS two-component system, the kinetics of in vitro phosphorylation assay and the kinetics of in vitro transcription assay.

https://doi.org/10.1371/journal.pone.0147281.s001

(PDF)

S1 Fig. Optimization of kinetic parameters associated with in vitro phosphorylation assay.

The cost function and the optimization profiles of the kinetic parameters associated with the simulation of in vitro phosphorylation assay results (Fig 4 and Table 3) as a function of SA steps. The colored (red, green, blue, cyan and magenta) lines are representatives of five different SA runs. The black horizontal line represents the base parameter value given in Table 1. Note the logarithmic scale in the ordinates.

https://doi.org/10.1371/journal.pone.0147281.s002

(TIF)

S2 Fig. Optimization of kinetic parameters associated with in vitro transcription assay of fhaB.

The cost function and the optimization profiles of the kinetic parameters (Table 4) associated with the simulation of in vitro transcription assay results of fhaB (Fig 5A) as a function of SA steps. The colored (red, green, blue, cyan and magenta) lines are representatives of five different SA runs. The black horizontal line represents the base parameter value given in Table 1. Note the logarithmic scale in the ordinates.

https://doi.org/10.1371/journal.pone.0147281.s003

(TIF)

S3 Fig. Optimization of kinetic parameters associated with in vitro transcription assay of bipA.

The cost function and the optimization profiles of the kinetic parameters (Table 5) associated with the simulation of in vitro transcription assay results of bipA (Fig 5B) as a function of SA steps. The colored (red, green, blue, cyan and magenta) lines are representatives of five different SA runs. The black horizontal line represents the base parameter value given in Table 1. Note the logarithmic scale in the ordinates.

https://doi.org/10.1371/journal.pone.0147281.s004

(TIF)

S1 Table. CC, RCC, and PRCC values for all the input parameters with output β for k_ps = 10⁻² at different range of perturbation.

https://doi.org/10.1371/journal.pone.0147281.s005

(PDF)

Acknowledgments

We are thankful to Debi Banerjee for critical reading of the manuscript. TM, SC and SKB acknowledge financial support from CSIR, India [01(2771)/14/EMR-II]. SKB is thankful to Bose Institute, Kolkata, India for research fund.

Author Contributions

Conceived and designed the experiments: TM SKB. Performed the experiments: TM ST. Analyzed the data: TM ST SC SKB. Contributed reagents/materials/analysis tools: SC SKB. Wrote the paper: TM ST SC SKB.

References

1. Alon U (2007) An Introduction to Systems Biology: Design Principles of Biological Circuits. CRC Press, New York.
2. Appleby JL, Parkinson JS, Bourret RB (1996) Signal transduction via the multi-step phosphorelay: not necessarily a road less traveled. Cell 86: 845–848. pmid:8808618
- View Article
- PubMed/NCBI
- Google Scholar
3. Hoch JA (2000) Two-component and phosphorelay signal transduction. Curr Opin Microbiol 3: 165–170. pmid:10745001
- View Article
- PubMed/NCBI
- Google Scholar
4. Stock AM, Robinson VL, Goudreau PN (2000) Two-component signal transduction. Annu Rev Biochem 69: 183–215. pmid:10966457
- View Article
- PubMed/NCBI
- Google Scholar
5. Laub MT, Goulian M (2007) Specificity in two-component signal transduction pathways. Annu Rev Genet 41: 121–145. pmid:18076326
- View Article
- PubMed/NCBI
- Google Scholar
6. Mitrophanov AY, Groisman EA (2008) Signal integration in bacterial two-component regulatory systems. Genes Dev 22: 2601–2611. pmid:18832064
- View Article
- PubMed/NCBI
- Google Scholar
7. Beier D, Gross R (2008) The bvgs/bvga phosphorelay system of pathogenic bordetellae: structure, function and evolution. Adv Exp Med Biol 631: 149–160. pmid:18792687
- View Article
- PubMed/NCBI
- Google Scholar
8. Cotter PA, Jones AM (2003) Phosphorelay control of virulence gene expression in bordetella. Trends Microbiol 11: 367–373. pmid:12915094
- View Article
- PubMed/NCBI
- Google Scholar
9. Melvin JA, Scheller EV, Miller JF, Cotter PA (2014) Bordetella pertussis pathogenesis: current and future challenges. Nat Rev Microbiol 12: 274–288. pmid:24608338
- View Article
- PubMed/NCBI
- Google Scholar
10. Alves R, Savageau MA (2003) Comparative analysis of prototype two-component systems with either bifunctional or monofunctional sensors: differences in molecular structure and physiological function. Mol Microbiol 48: 25–51. pmid:12657043
- View Article
- PubMed/NCBI
- Google Scholar
11. Batchelor E, Goulian M (2003) Robustness and the cycle of phosphorylation and dephosphorylation in a two-component regulatory system. Proc Natl Acad Sci U S A 100: 691–696. pmid:12522261
- View Article
- PubMed/NCBI
- Google Scholar
12. Kato A, Mitrophanov AY, Groisman EA (2007) A connector of two-component regulatory systems promotes signal amplification and persistence of expression. Proc Natl Acad Sci U S A 104: 12063–12068. pmid:17615238
- View Article
- PubMed/NCBI
- Google Scholar
13. Shinar G, Milo R, Martínez MR, Alon U (2007) Input output robustness in simple bacterial signaling systems. Proc Natl Acad Sci U S A 104: 19931–19935. pmid:18077424
- View Article
- PubMed/NCBI
- Google Scholar
14. Miyashiro T, Goulian M (2008) High stimulus unmasks positive feedback in an autoregulated bacterial signaling circuit. Proc Natl Acad Sci USA 105: 17457–17462. pmid:18987315
- View Article
- PubMed/NCBI
- Google Scholar
15. Sureka K, Ghosh B, Dasgupta A, Basu J, Kundu M, Bose I (2008) Positive feedback and noise activate the stringent response regulator rel in mycobacteria. PLoS One 3: e1771–e1771. pmid:18335046
- View Article
- PubMed/NCBI
- Google Scholar
16. Igoshin OA, Alves R, Savageau MA (2008) Hysteretic and graded responses in bacterial two-component signal transduction. Mol Microbiol 68: 1196–1215. pmid:18363790
- View Article
- PubMed/NCBI
- Google Scholar
17. Banik SK, Fenley AT, Kulkarni RV (2009) A model for signal transduction during quorum sensing in Vibrio harveyi. Phys Biol 6: 046008–046008. pmid:19843985
- View Article
- PubMed/NCBI
- Google Scholar
18. Kierzek AM, Zhou L, Wanner BL (2010) Stochastic kinetic model of two component system signalling reveals all-or-none, graded and mixed mode stochastic switching responses. Mol Biosyst 6: 531–542. pmid:20174681
- View Article
- PubMed/NCBI
- Google Scholar
19. Tiwari A, Balazsi G, Gennaro ML, Igoshin OA (2010) The interplay of multiple feedback loops with post-translational kinetics results in bistability of mycobacterial stress response. Phys Biol 7: 036005. pmid:20733247
- View Article
- PubMed/NCBI
- Google Scholar
20. Csikász-Nagy A, Cardelli L, Soyer OS (2011) Response dynamics of phosphorelays suggest their potential utility in cell signalling. J R Soc Interface 8: 480–488. pmid:20702449
- View Article
- PubMed/NCBI
- Google Scholar
21. Wei K, Moinat M, Maarleveld TR, Bruggeman FJ (2014) Stochastic simulation of prokaryotic two-component signalling indicates stochasticity-induced active-state locking and growth-rate dependent bistability. Mol Biosyst 10: 2338–2346. pmid:24955938
- View Article
- PubMed/NCBI
- Google Scholar
22. Bandyopadhyay A, Banik SK (2012) Positive feedback and temperature mediated molecular switch controls differential gene regulation in Bordetella pertussis. BioSystems 110: 107–118. pmid:22960292
- View Article
- PubMed/NCBI
- Google Scholar
23. Prajapat MK, Saini S (2013) Role of feedback and network architecture in controlling virulence gene expression in Bordetella. Mol Biosyst 9: 2635–2644. pmid:24056999
- View Article
- PubMed/NCBI
- Google Scholar
24. Jones AM, Boucher PE, Williams CL, Stibitz S, Cotter PA (2005) Role of bvga phosphorylation and dna binding affinity in control of bvg-mediated phenotypic phase transition in bordetella pertussis. Mol Microbiol 58: 700–713. pmid:16238621
- View Article
- PubMed/NCBI
- Google Scholar
25. Roy CR, Miller JF, Falkow S (1990) Autogenous regulation of the bordetella pertussis bvgabc operon. Proc Natl Acad Sci U S A 87: 3763–3767. pmid:2111016
- View Article
- PubMed/NCBI
- Google Scholar
26. Scarlato V, Prugnola A, Aricó B, Rappuoli R (1990) Positive transcriptional feedback at the bvg locus controls expression of virulence factors in bordetella pertussis. Proc Natl Acad Sci U S A 87: 10067–10067. pmid:2263607
- View Article
- PubMed/NCBI
- Google Scholar
27. Scarlato V, Aricò B, Prugnola A, Rappuoli R (1991) Sequential activation and environmental regulation of virulence genes in bordetella pertussis. EMBO J 10: 3971–3975. pmid:1718746
- View Article
- PubMed/NCBI
- Google Scholar
28. Hot D, Slupek S, Wulbrecht B, D’Hondt A, Hubans C, Antoine R, et al. (2011) Detection of small RNAs in Bordetella pertussis and identification of a novel repeated genetic element. BMC Genomics 12: 207. pmid:21524285
- View Article
- PubMed/NCBI
- Google Scholar
29. Salteli A (2004) Sensitivity Analysis in Practice: A Guide to Assessing Scientific Models. Wiley, New Jersy.
30. Saltelli A, Ratto M, Tarantola S, Campolongo F (2005) Sensitivity analysis for chemical models. Chem Rev 105: 2811–2828. pmid:16011325
- View Article
- PubMed/NCBI
- Google Scholar
31. Marino S, Hogue IB, Ray CJ, Kirschner DE (2008) A methodology for performing global uncertainty and sensitivity analysis in systems biology. J Theor Biol 254: 178–196. pmid:18572196
- View Article
- PubMed/NCBI
- Google Scholar
32. Talukder S, Sen S, Metzler R, Banik SK, Chaudhury P (2013) Stochastic optimization-based study of dimerization kinetics. J Chem Sci 125: 1619–1627.
- View Article
- Google Scholar
33. Kirkpatrick S, Gelatt CD, Vecchi MP (1983) Optimization by simulated annealing. Science 220: 671–680.
- View Article
- Google Scholar
34. Kirkpatrick S (1984) Optimization by simulated annealing: Quantitative studies. J Stat Phys 34: 975–986.
- View Article
- Google Scholar
35. Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E (1953) Equation of state calculations by fast computing machines. J Chem Phys 21: 1087–1092.
- View Article
- Google Scholar
36. Talukder S, Chaudhury P, Metzler R, Banik SK (2011) Determining the DNA stability parameters for the breathing dynamics of heterogeneous DNA by stochastic optimization. J Chem Phys 135: 165103. pmid:22047268
- View Article
- PubMed/NCBI
- Google Scholar
37. Talukder S, Sen S, Chakraborti P, Metzler R, Banik SK, Chaudhury P (2014) Breathing dynamics based parameter sensitivity analysis of hetero-polymeric DNA. J Chem Phys 140: 125101. pmid:24697480
- View Article
- PubMed/NCBI
- Google Scholar
38. Koshland DE, Goldbeter A, Stock JB (1982) Amplification and adaptation in regulatory and sensory systems. Science 217: 220–225.
- View Article
- Google Scholar
39. Savageau MA (1976) Biochemical Systems Analysis: A Study of Function and Design in Molecular Biology. Addison-Wesley, Reading, Massachusetts.
40. Goldbeter A, Koshland DE (1981) An amplified sensitivity arising from covalent modification in biological systems. Proc Natl Acad Sci U S A 78: 6840–6844. pmid:6947258
- View Article
- PubMed/NCBI
- Google Scholar
41. Boucher PE, Murakami K, Ishihama A, Stibitz S (1997) Nature of dna binding and rna polymerase interaction of the bordetella pertussis bvga transcriptional activator at the fha promoter. J Bacteriol 179: 1755–1763. pmid:9045838
- View Article
- PubMed/NCBI
- Google Scholar

[ref1] 1. Alon U (2007) An Introduction to Systems Biology: Design Principles of Biological Circuits. CRC Press, New York.

[ref2] 2. Appleby JL, Parkinson JS, Bourret RB (1996) Signal transduction via the multi-step phosphorelay: not necessarily a road less traveled. Cell 86: 845–848. pmid:8808618
View Article
PubMed/NCBI
Google Scholar

[3] View Article

[4] PubMed/NCBI

[5] Google Scholar

[ref3] 3. Hoch JA (2000) Two-component and phosphorelay signal transduction. Curr Opin Microbiol 3: 165–170. pmid:10745001
View Article
PubMed/NCBI
Google Scholar

[7] View Article

[8] PubMed/NCBI

[9] Google Scholar

[ref4] 4. Stock AM, Robinson VL, Goudreau PN (2000) Two-component signal transduction. Annu Rev Biochem 69: 183–215. pmid:10966457
View Article
PubMed/NCBI
Google Scholar

[11] View Article

[12] PubMed/NCBI

[13] Google Scholar

[ref5] 5. Laub MT, Goulian M (2007) Specificity in two-component signal transduction pathways. Annu Rev Genet 41: 121–145. pmid:18076326
View Article
PubMed/NCBI
Google Scholar

[15] View Article

[16] PubMed/NCBI

[17] Google Scholar

[ref6] 6. Mitrophanov AY, Groisman EA (2008) Signal integration in bacterial two-component regulatory systems. Genes Dev 22: 2601–2611. pmid:18832064
View Article
PubMed/NCBI
Google Scholar

[19] View Article

[20] PubMed/NCBI

[21] Google Scholar

[ref7] 7. Beier D, Gross R (2008) The bvgs/bvga phosphorelay system of pathogenic bordetellae: structure, function and evolution. Adv Exp Med Biol 631: 149–160. pmid:18792687
View Article
PubMed/NCBI
Google Scholar

[23] View Article

[24] PubMed/NCBI

[25] Google Scholar

[ref8] 8. Cotter PA, Jones AM (2003) Phosphorelay control of virulence gene expression in bordetella. Trends Microbiol 11: 367–373. pmid:12915094
View Article
PubMed/NCBI
Google Scholar

[27] View Article

[28] PubMed/NCBI

[29] Google Scholar

[ref9] 9. Melvin JA, Scheller EV, Miller JF, Cotter PA (2014) Bordetella pertussis pathogenesis: current and future challenges. Nat Rev Microbiol 12: 274–288. pmid:24608338
View Article
PubMed/NCBI
Google Scholar

[31] View Article

[32] PubMed/NCBI

[33] Google Scholar

[ref10] 10. Alves R, Savageau MA (2003) Comparative analysis of prototype two-component systems with either bifunctional or monofunctional sensors: differences in molecular structure and physiological function. Mol Microbiol 48: 25–51. pmid:12657043
View Article
PubMed/NCBI
Google Scholar

[35] View Article

[36] PubMed/NCBI

[37] Google Scholar

[ref11] 11. Batchelor E, Goulian M (2003) Robustness and the cycle of phosphorylation and dephosphorylation in a two-component regulatory system. Proc Natl Acad Sci U S A 100: 691–696. pmid:12522261
View Article
PubMed/NCBI
Google Scholar

[39] View Article

[40] PubMed/NCBI

[41] Google Scholar

[ref12] 12. Kato A, Mitrophanov AY, Groisman EA (2007) A connector of two-component regulatory systems promotes signal amplification and persistence of expression. Proc Natl Acad Sci U S A 104: 12063–12068. pmid:17615238
View Article
PubMed/NCBI
Google Scholar

[43] View Article

[44] PubMed/NCBI

[45] Google Scholar

[ref13] 13. Shinar G, Milo R, Martínez MR, Alon U (2007) Input output robustness in simple bacterial signaling systems. Proc Natl Acad Sci U S A 104: 19931–19935. pmid:18077424
View Article
PubMed/NCBI
Google Scholar

[47] View Article

[48] PubMed/NCBI

[49] Google Scholar

[ref14] 14. Miyashiro T, Goulian M (2008) High stimulus unmasks positive feedback in an autoregulated bacterial signaling circuit. Proc Natl Acad Sci USA 105: 17457–17462. pmid:18987315
View Article
PubMed/NCBI
Google Scholar

[51] View Article

[52] PubMed/NCBI

[53] Google Scholar

[ref15] 15. Sureka K, Ghosh B, Dasgupta A, Basu J, Kundu M, Bose I (2008) Positive feedback and noise activate the stringent response regulator rel in mycobacteria. PLoS One 3: e1771–e1771. pmid:18335046
View Article
PubMed/NCBI
Google Scholar

[55] View Article

[56] PubMed/NCBI

[57] Google Scholar

[ref16] 16. Igoshin OA, Alves R, Savageau MA (2008) Hysteretic and graded responses in bacterial two-component signal transduction. Mol Microbiol 68: 1196–1215. pmid:18363790
View Article
PubMed/NCBI
Google Scholar

[59] View Article

[60] PubMed/NCBI

[61] Google Scholar

[ref17] 17. Banik SK, Fenley AT, Kulkarni RV (2009) A model for signal transduction during quorum sensing in Vibrio harveyi. Phys Biol 6: 046008–046008. pmid:19843985
View Article
PubMed/NCBI
Google Scholar

[63] View Article

[64] PubMed/NCBI

[65] Google Scholar

[ref18] 18. Kierzek AM, Zhou L, Wanner BL (2010) Stochastic kinetic model of two component system signalling reveals all-or-none, graded and mixed mode stochastic switching responses. Mol Biosyst 6: 531–542. pmid:20174681
View Article
PubMed/NCBI
Google Scholar

[67] View Article

[68] PubMed/NCBI

[69] Google Scholar

[ref19] 19. Tiwari A, Balazsi G, Gennaro ML, Igoshin OA (2010) The interplay of multiple feedback loops with post-translational kinetics results in bistability of mycobacterial stress response. Phys Biol 7: 036005. pmid:20733247
View Article
PubMed/NCBI
Google Scholar

[71] View Article

[72] PubMed/NCBI

[73] Google Scholar

[ref20] 20. Csikász-Nagy A, Cardelli L, Soyer OS (2011) Response dynamics of phosphorelays suggest their potential utility in cell signalling. J R Soc Interface 8: 480–488. pmid:20702449
View Article
PubMed/NCBI
Google Scholar

[75] View Article

[76] PubMed/NCBI

[77] Google Scholar

[ref21] 21. Wei K, Moinat M, Maarleveld TR, Bruggeman FJ (2014) Stochastic simulation of prokaryotic two-component signalling indicates stochasticity-induced active-state locking and growth-rate dependent bistability. Mol Biosyst 10: 2338–2346. pmid:24955938
View Article
PubMed/NCBI
Google Scholar

[79] View Article

[80] PubMed/NCBI

[81] Google Scholar

[ref22] 22. Bandyopadhyay A, Banik SK (2012) Positive feedback and temperature mediated molecular switch controls differential gene regulation in Bordetella pertussis. BioSystems 110: 107–118. pmid:22960292
View Article
PubMed/NCBI
Google Scholar

[83] View Article

[84] PubMed/NCBI

[85] Google Scholar

[ref23] 23. Prajapat MK, Saini S (2013) Role of feedback and network architecture in controlling virulence gene expression in Bordetella. Mol Biosyst 9: 2635–2644. pmid:24056999
View Article
PubMed/NCBI
Google Scholar

[87] View Article

[88] PubMed/NCBI

[89] Google Scholar

[ref24] 24. Jones AM, Boucher PE, Williams CL, Stibitz S, Cotter PA (2005) Role of bvga phosphorylation and dna binding affinity in control of bvg-mediated phenotypic phase transition in bordetella pertussis. Mol Microbiol 58: 700–713. pmid:16238621
View Article
PubMed/NCBI
Google Scholar

[91] View Article

[92] PubMed/NCBI

[93] Google Scholar

[ref25] 25. Roy CR, Miller JF, Falkow S (1990) Autogenous regulation of the bordetella pertussis bvgabc operon. Proc Natl Acad Sci U S A 87: 3763–3767. pmid:2111016
View Article
PubMed/NCBI
Google Scholar

[95] View Article

[96] PubMed/NCBI

[97] Google Scholar

[ref26] 26. Scarlato V, Prugnola A, Aricó B, Rappuoli R (1990) Positive transcriptional feedback at the bvg locus controls expression of virulence factors in bordetella pertussis. Proc Natl Acad Sci U S A 87: 10067–10067. pmid:2263607
View Article
PubMed/NCBI
Google Scholar

[99] View Article

[100] PubMed/NCBI

[101] Google Scholar

[ref27] 27. Scarlato V, Aricò B, Prugnola A, Rappuoli R (1991) Sequential activation and environmental regulation of virulence genes in bordetella pertussis. EMBO J 10: 3971–3975. pmid:1718746
View Article
PubMed/NCBI
Google Scholar

[103] View Article

[104] PubMed/NCBI

[105] Google Scholar

[ref28] 28. Hot D, Slupek S, Wulbrecht B, D’Hondt A, Hubans C, Antoine R, et al. (2011) Detection of small RNAs in Bordetella pertussis and identification of a novel repeated genetic element. BMC Genomics 12: 207. pmid:21524285
View Article
PubMed/NCBI
Google Scholar

[107] View Article

[108] PubMed/NCBI

[109] Google Scholar

[ref29] 29. Salteli A (2004) Sensitivity Analysis in Practice: A Guide to Assessing Scientific Models. Wiley, New Jersy.

[ref30] 30. Saltelli A, Ratto M, Tarantola S, Campolongo F (2005) Sensitivity analysis for chemical models. Chem Rev 105: 2811–2828. pmid:16011325
View Article
PubMed/NCBI
Google Scholar

[112] View Article

[113] PubMed/NCBI

[114] Google Scholar

[ref31] 31. Marino S, Hogue IB, Ray CJ, Kirschner DE (2008) A methodology for performing global uncertainty and sensitivity analysis in systems biology. J Theor Biol 254: 178–196. pmid:18572196
View Article
PubMed/NCBI
Google Scholar

[116] View Article

[117] PubMed/NCBI

[118] Google Scholar

[ref32] 32. Talukder S, Sen S, Metzler R, Banik SK, Chaudhury P (2013) Stochastic optimization-based study of dimerization kinetics. J Chem Sci 125: 1619–1627.
View Article
Google Scholar

[120] View Article

[121] Google Scholar

[ref33] 33. Kirkpatrick S, Gelatt CD, Vecchi MP (1983) Optimization by simulated annealing. Science 220: 671–680.
View Article
Google Scholar

[123] View Article

[124] Google Scholar

[ref34] 34. Kirkpatrick S (1984) Optimization by simulated annealing: Quantitative studies. J Stat Phys 34: 975–986.
View Article
Google Scholar

[126] View Article

[127] Google Scholar

[ref35] 35. Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH, Teller E (1953) Equation of state calculations by fast computing machines. J Chem Phys 21: 1087–1092.
View Article
Google Scholar

[129] View Article

[130] Google Scholar

[ref36] 36. Talukder S, Chaudhury P, Metzler R, Banik SK (2011) Determining the DNA stability parameters for the breathing dynamics of heterogeneous DNA by stochastic optimization. J Chem Phys 135: 165103. pmid:22047268
View Article
PubMed/NCBI
Google Scholar

[132] View Article

[133] PubMed/NCBI

[134] Google Scholar

[ref37] 37. Talukder S, Sen S, Chakraborti P, Metzler R, Banik SK, Chaudhury P (2014) Breathing dynamics based parameter sensitivity analysis of hetero-polymeric DNA. J Chem Phys 140: 125101. pmid:24697480
View Article
PubMed/NCBI
Google Scholar

[136] View Article

[137] PubMed/NCBI

[138] Google Scholar

[ref38] 38. Koshland DE, Goldbeter A, Stock JB (1982) Amplification and adaptation in regulatory and sensory systems. Science 217: 220–225.
View Article
Google Scholar

[140] View Article

[141] Google Scholar

[ref39] 39. Savageau MA (1976) Biochemical Systems Analysis: A Study of Function and Design in Molecular Biology. Addison-Wesley, Reading, Massachusetts.

[ref40] 40. Goldbeter A, Koshland DE (1981) An amplified sensitivity arising from covalent modification in biological systems. Proc Natl Acad Sci U S A 78: 6840–6844. pmid:6947258
View Article
PubMed/NCBI
Google Scholar

[144] View Article

[145] PubMed/NCBI

[146] Google Scholar

[ref41] 41. Boucher PE, Murakami K, Ishihama A, Stibitz S (1997) Nature of dna binding and rna polymerase interaction of the bordetella pertussis bvga transcriptional activator at the fha promoter. J Bacteriol 179: 1755–1763. pmid:9045838
View Article
PubMed/NCBI
Google Scholar

[148] View Article

[149] PubMed/NCBI

[150] Google Scholar

Figures

Abstract

Introduction

Methods

The bvg operon

The BvgAS two-component system

Sensitivity Analysis

Stochastic Optimization

Results and Discussion

Amplification and Switch

Parameter sensitivity analysis

In vitro assay and stochastic optimization: The mutants

Conclusion

Supporting Information

S1 Text. General supplementary information.

S1 Fig. Optimization of kinetic parameters associated with in vitro phosphorylation assay.

S2 Fig. Optimization of kinetic parameters associated with in vitro transcription assay of fhaB.

S3 Fig. Optimization of kinetic parameters associated with in vitro transcription assay of bipA.

S1 Table. CC, RCC, and PRCC values for all the input parameters with output β for kps = 10−2 at different range of perturbation.

Acknowledgments

Author Contributions

References

S1 Table. CC, RCC, and PRCC values for all the input parameters with output β for k_ps = 10⁻² at different range of perturbation.