Noise in a phosphorelay drives stochastic entry into sporulation in Bacillus subtilis

Abstract

Entry into sporulation in Bacillus subtilis is governed by a phosphorelay in which phosphoryl groups from a histidine kinase are successively transferred via relay proteins to the response regulator Spo0A. Spo0A~P, in turn, sets in motion events that lead to asymmetric division and activation of the cell-specific transcription factor σ^F, a hallmark for entry into sporulation. Here, we have used a microfluidics-based platform to investigate the activation of Spo0A and σ^F in individual cells held under constant, sporulation-inducing conditions. The principal conclusions were that: (i) activation of σ^F occurs with an approximately constant probability after adaptation to conditions of nutrient limitation; (ii) activation of σ^F is tightly correlated with, and preceded by, Spo0A~P reaching a high threshold level; (iii) activation of Spo0A takes place abruptly just prior to asymmetric division; and (iv) the primary source of noise in the activation of Spo0A is the phosphorelay. We propose that cells exhibit a constant probability of attaining a high threshold level of Spo0A~P due to fluctuations in the flux of phosphoryl groups through the phosphorelay.

Keywords: cell fate; constant probability; phosphorelay; sporulation.

Figure 1. Tracking early transcriptional activation events in individual sporulating cells by fluorescence time‐lapse microscopy using a microfluidic platform

1. A schematic of the phosphorelay governing entry into sporulation. An auto‐phosphorylating sensor kinase feeds single phosphoryl groups through a series of phosphotransfer reactions constituting the phosphorelay, ultimately leading to the phosphorylation of Spo0A (0A˜P). Phosphate is drained from the relay by the action of the phosphatase Spo0E (0E) acting on Spo0A˜P.

2. The cartoon depicts the construction and geometry of the microfluidic device used to image single‐cell lineages over time. Small, 1‐μm‐wide growth channels are arrayed off a main feeding channel, and growth medium bathes cells via a 5‐μm‐wide shallow overlay. Growth channels are loaded with a culture of isogenic cells, which grow and divide, pushing daughter cells out into the waste stream. After a period of steady‐state growth, cell behaviors are visualized after a switch to starvation medium. All of the cell lineages in a field of view were used for analysis; lineages that exited the field of view due to growth or cell lysis were truncated at the point of loss, and only lineages that were trackable past a threshold time (200 min) were used for analysis.

3. A representative kymograph of results for reporters of low and high levels of Spo0A˜P and of σ^F in cells of strain JRR368 grown in the microfluidic device. A kymograph shows images of a single cell taken every 30 min after a switch to starvation conditions. The cell divides symmetrically multiple times before activating low levels of Spo0A˜P (green), high levels of Spo0A˜P (red), and σ^F (cyan), and ultimately forming a phase‐bright spore (white; Phase). Time‐lapse sequences of JRR368 sporulating in the microfluidic device are also shown in Movies EV1 and EV2.

4. Cartoon depicting the pattern, localization, and colors of the three fluorescent reporters shown in panel (C).

5. Single‐cell lineages were constructed using a combination of SuperSegger (Stylianidou et al, 2016) and custom MATLAB scripts. An individual cell's response to a switch to sporulation conditions is plotted for each reporter channel: P_{low_0A˜P} (P_sdp‐mTurquoise2, green curve), P_{high_0A˜P} (P_spoIIG‐mNeonGreen, red curve), and P_σF (P_spoIIQ‐mNeptune, cyan curve). After a delay following the switch, activation of a low‐threshold Spo0A˜P promoter (P_sdp) precedes the activation of a high‐threshold Spo0A˜P promoter (P_spoIIG), which in turn precedes the activation of a σ^F‐directed promoter (P_spoIIQ).

Figure 2. The timing of σ^F and Spo0A~P activation is heterogeneous but shows a constant probability after a switch to constant sporulation‐inducing conditions

1. A–C

   The profiles of fluorescent intensities for P_{low_0A˜P} (A) P_{high_0A˜P} (B) and P_σF (C) reporters are plotted for individual cell lineages of a strain (JRR368) bearing all three fluorescent reporters. Lineage behaviors are shown from 5 h before a switch into sporulation medium until approximately 15 h after the switch. For simplicity of visualization, the reporter data are smoothed by a box filter with a size corresponding to 30 min. In cases where activation occurs, lineages are plotted from the beginning of the experiment until the frame 1 h after the maximal slope of their activation. Fluorescence values are corrected for background fluorescence during the course of the experiment.

2. D

   After a delay, beginning when cells experienced the switch to sporulation medium, growing cells achieved a stable probability that a division would result in activation of σ^F. The probability that a division at time (t) resulted in asymmetric division and σ^F activation was calculated by dividing the number of σ^F activation events by the total number of divisions, both asymmetric and symmetric.

3. E

   During the period that cells exhibited a constant probability of sporulation (hours 9–16), the distribution of waiting times prior to σ^F activation fit well to an exponential distribution (red curve), a characteristic of a memoryless process. The single‐exponential fit was calculated from a distribution of 80 events using the fit function in MATLAB. Events are plotted in bins of width corresponding to 50 min. This fit was reproduced qualitatively in an analysis of mother cell (MC) lineages where 38 events were fit to a similar exponential distribution (inset).

Figure 3. Cell lineage sorting by σ^F activation reveals two classes of behavior, with stereotyped timing between high‐threshold Spo0A activation and σ^F activation

1. A–F

   Lineage data from a triple‐labeled strain (JRR368) was sorted into two classes: those that activated σ^F (A–C) and those that did not (D–F). While low‐threshold activation of Spo0A was broadly distributed between these two classes (A and D), high‐threshold activation occurred only in cells that went on to activate σ^F (B and E). Activation of σ^F was defined as the signal crossing a threshold 5σ above the background variation prior to the switch. Lineages were plotted from 5 h before a switch to starvation medium until 15 h after the switch. For simplicity of visualization, reporter data are smoothed by a box filter with a size corresponding to 30 min. In cases where activation occurred, lineages were plotted from the beginning of the experiment until the frame 1 h after the maximal slope of their activation.

2. G–I

   Cell lineages (JRR368) were aligned in time to the point at which they crossed the threshold for σ^F activation. The data corresponding to the fluorescent channels for each reporter were then plotted on the same time axis. (G) σ^F activation events were well‐aligned in their profiles of activation. (H) Activation of a high‐threshold reporter for Spo0A correlated with and preceded activation of σ^F by a mean time (μ) of 1.12 h (CV ≈ 0.53). (I) Activation of a low‐threshold reporter was poorly correlated with but tended to precede activation of σ^F (μ = 3.03 h, CV ≈ 0.75). Activation timing was calculated as when a reporter signal crossed a threshold 5σ above the background variation prior to the switch. Mean activation timing (μ) was calculated as an average of the time differences between activation events with a standard deviation of σ. The coefficient of variation (CV) was calculated as CV = σ/μ.

Figure 4. Spo0A activity rarely exhibits cell cycle‐dependent pulsing prior to activation of σ^F

1. Profiles of low‐threshold P_{low_0A˜P} (green) and P_σF (cyan) fluorescence intensity are plotted for single‐cell lineages (JRR424) after a medium switch. Vegetative cell divisions during growth are marked in time with dotted lines. There was no consistent evidence of pulsatile Spo0A activation preceding activation of σ^F. Some cells did exhibit behavior in which Spo0A activity increased between cell cycles (bottom plot), but this pattern was exceptional. Reporter data were smoothed with a box filter with a size corresponding to 30 min.

2. Cell cycles were extracted from lineages of cells that ultimately activated σ^F. The profile of low‐threshold Spo0A reporter activity for the two cell cycles (G_‐2 and G_‐1) preceding the cycle in which σ^F becomes activated are plotted (top row). The lifetimes of each cell cycle (as defined by its preceding and subsequent divisions) are normalized to a unit vector length. The majority of cells did not exhibit progressively higher levels of Spo0A activity, even as measured by the most sensitive reporter we tested (P_sdp). The derivative of average reporter intensity with respect to time is a rough proxy for the concentration of Spo0A˜P in a given cell. The derivative is plotted (bottom row) over the same normalized time for the cell cycles preceding σ^F activation. Potential pulses of Spo0A˜P were detected in a few (3) cells in generation G_‐1, but the remaining cells observed (41) did not exhibit pulsing in the two cell cycles preceding σ^F activation. Data are presented as raw traces corresponding to individual cell lineages.

Figure 5. KinA‐GFP levels do not predict σ^F activation

1. KinA‐GFP levels produced under the control of the native P_kinA promoter were low and stable in time after a medium switch to starvation conditions. Highlighted (cyan) are cell lineages that ended in σ^F activation. Cell lineages were plotted for a strain (JRR425) harboring a construct for inducible synthesis of KinA‐GFP (amyE::P_hyperspank‐kinA‐gfp) in addition to a reporter for P_σF (P_spoIIQ‐mNeptune).

2. The maximum value of KinA‐GFP achieved in each intact cell lineage in (A) is plotted against the corresponding maximal value of the σ^F reporter reached in that cell lineage. Activation of σ^F (cyan) is defined as signal 5σ above the background variation prior to the medium switch (dotted line). Data shown are from a single experiment with 121 intact lineages from the same strain as in (A). The result was reproduced qualitatively in at least three other experiments.

3. Under the control of an inducible P_hyperspank promoter, KinA‐GFP levels were tuned across a range of concentrations by the addition of various concentrations of IPTG (0–10 μM) concomitant with the switch to sporulation medium. KinA‐GFP levels reached a maximum after induction before decreasing, presumably as some cells began to enter sporulation.

4. Despite variation in induction, KinA‐GFP levels were uncorrelated to the activation of σ^F, although high levels of induction did result in predominantly σ^F‐activating cell lineages (10 μM, dark green). Activation of σ^F is defined as signal 5σ above the background variation prior to the medium switch (dotted line). Data shown are from single experiments for at least 100 intact lineages from each induction condition. The result was reproduced qualitatively in two separate experiments.

Figure 6. Growth medium affects the level of KinA required to activate σ^F but not heterogeneity

1. A, B

   Expression of KinA‐GFP was induced by the addition of IPTG to JRR425 cells bearing a construct for inducible synthesis of KinA‐GFP (kinA::P_hyperspank‐kinA‐gfp). Cells were either switched to sporulation medium (A, green) or maintained in growth medium (B, blue). Higher levels of KinA were needed to activate σ^F in cells that did not experience a concomitant medium switch. The maximal value of KinA‐GFP achieved in each intact cell lineage is plotted against the corresponding maximal value of the σ^F reporter reached in that cell lineage. Data shown are from single experiments for at least 100 intact lineages from each induction condition. The result was reproduced qualitatively in two separate experiments.

Figure 7. Bypass of the phosphorelay reduces heterogeneity in σ^F activation over a range of KinA levels

1. A–C

   A strain with a point mutation in the spo0A gene (spo0A ^E14A, also known as sof‐3) was assessed for its response to a range of induced KinA‐GFP levels in the presence (JRR497) or absence of spo0F (JRR500). The maximal value of KinA‐GFP achieved in each intact cell lineage is plotted against the corresponding maximal value of the σ^F reporter in that cell lineage. Over a range of KinA‐GFP levels, activation of σ^F was more switch‐like in the relay mutant (B) than in cells with an intact phosphorelay (A). Representative time‐lapse sequences of KinA‐GFP cells at different induction levels are shown in Movies EV7, EV8 and EV9 (JRR497, relay‐intact cells) and in Movies EV10, EV11 and EV12 (JRR500, relay‐bypassed cells). Activation of σ^F is defined as a signal 5σ above the background variation prior to the medium switch (dotted line). The data shown are from a single experiment with at least 100 intact lineages per induction condition. The result was qualitatively reproduced in at least two experiments. The proportion of cells activating σ^F is plotted for cells expressing a range of KinA‐GFP levels (C). Data are shown for cells with an intact relay (JRR497, black) and in the absence of an intact relay (JRR500, red). The transition between KinA‐GFP levels yielding no activation and levels yielding uniform activation is sharper in the absence of an intact phosphorelay (red points). The proportion was calculated as a ratio of the number of cells activating σ^F to the total number of cells observed during the observation window.

Figure 8. Kymographs showing the effect of phosphorelay bypass on the activation of σ^F

A strain with a point mutation in the spo0A gene (spo0A ^E14A, also known as sof‐3) was visualized in the microfluidic device as it responded to a range of KinA‐GFP levels in the presence (JRR497) or absence (JRR500) of spo0F. In the top row, heterogeneous behavior can be seen in spo0F ⁺ cells across a range of inducer (IPTG) concentrations. In the bottom row, Δspo0F cells failed to activate σ^F at low levels of inducer, but activated σ^F in almost all cells at high levels of inducer. Kymographs were constructed from images captured at 30‐min intervals for 8 h after a medium switch. Images are a merge of CFP fluorescence (σ^F reporter) images and phase‐contrast images. The figure is a representative example of lineages on which the results shown in Fig 7 are based.