Summary
Upon starvation Myxococcus xanthus undergoes multicellular development. Rod-shaped cells move into mounds in which some cells differentiate into spores. Cells begin committing to sporulation at 24-30 h poststarvation, but the mechanisms governing commitment are unknown. FruA and MrpC are transcription factors that are necessary for commitment. They bind cooperatively to promoter regions and activate developmental gene transcription, including that of the dev operon. Leading up to and during the commitment period, dev mRNA increased in wild type, but not in a mutant defective in C-signaling, a short-range signaling interaction between cells that is also necessary for commitment. The C-signaling mutant exhibited ∼20-fold less dev mRNA than wild type at 30 h poststarvation, despite a similar level of MrpC and only twofold less FruA. Boosting the FruA level twofold in the C-signaling mutant had little effect on the dev mRNA level, and dev mRNA was not less stable in the C-signaling mutant. Neither did high cooperativity of MrpC and FruA binding upstream of the dev promoter explain the data. Rather, our systematic experimental and computational analyses support a model in which C-signaling activates FruA at least ninefold posttranslationally in order to commit a cell to spore formation.
Graphical abstract
Abbreviated summary Starvation promotes MrpC accumulation, whereas nutrients favor proteolysis. MrpC activates transcription of fruA, but FruA protein appears to be activated by short-range C-signaling in a cycle leading to mound formation and lysis of some cells. Activated FruA* and MrpC are proposed to cooperatively stimulate transcription of the dev operon and genes that commit starving rod-shaped cells to form spores, while Dev proteins slow commitment, resulting in a spore-filled fruiting body surrounded by peripheral rods.
Introduction
Differentiated cell types are a hallmark of multicellular organisms. Understanding how pluripotent cells become restricted to particular cell fates is a fascinating question and a fundamental challenge in biology. In general, the answer involves a complex interplay between signals and gene regulation. This is true both during development of multicellular eukaryotes (Davidson & Levine, 2008, Frum & Ralston, 2015, Drapek et al., 2017) and during transitions in microbial communities (van Gestel et al., 2015, Norman et al., 2015, Bush et al., 2015, Kroos, 2017). Bacterial cells in microbial communities adopt different fates as gene regulatory networks (GRNs) respond to a variety of signals, including some generated by other cells. Moreover, we now understand that microbial communities or microbiomes profoundly impact eukaryotic organisms, and vice versa (Barratt et al., 2017, Jansson & Hofmockel, 2018). Yet the daunting complexity of microbiomes and multicellular eukaryotes impedes efforts to fully understand their interactions in molecular detail. By studying simpler model systems, paradigms can be discovered that can guide investigations of more complex interactions.
A relatively simple model system is provided by the bacterium Myxococcus xanthus, which undergoes starvation-induced multicellular development (Yang & Higgs, 2014). In response to starvation, cells generate intracellular and extracellular signals that regulate gene expression (Bretl & Kirby, 2016, Kroos, 2017). The rod-shaped cells alter their movements so that thousands form a mound. Within a mound, cells differentiate into ovoid spores that resist stress and remain dormant until nutrients reappear. The spore-filled mound is called a fruiting body. Other cells adopt a different fate and remain outside the fruiting body as peripheral rods (O’Connor & Zusman, 1991). A large proportion of the cells lyse during the developmental process (Lee et al., 2012). What determines whether a given cell in the population forms a spore, remains as a peripheral rod, or undergoes lysis? M. xanthus provides an attractive model system to discover how signaling between cells affects a GRN and determines cell fate. Here, we focus on a circuit that regulates commitment to sporulation.
In a recent study, cells committed to spore formation primarily between 24 and 30 h poststarvation (PS), because addition of nutrients to the starving population prior to 24 h PS blocked subsequent sporulation, addition at 24 h PS allowed a few spores to form subsequently, and addition at 30 h PS allowed many more spores to form (Rajagopalan & Kroos, 2014). At the molecular level, addition of nutrients before or during the commitment period caused rapid proteolysis of MrpC (Rajagopalan & Kroos, 2014), a transcription factor required for fruiting body formation (Sun & Shi, 2001b, Sun & Shi, 2001a).
MrpC appears to directly regulate more than one hundred genes involved in development (Robinson et al., 2014), and one well-characterized MrpC target gene, fruA (Ueki & Inouye, 2003), codes for another transcription factor required for fruiting body formation (Ogawa et al., 1996). FruA and MrpC bind cooperatively to the promoter regions of many genes, and appear to activate transcription (Campbell et al., 2015, Lee et al., 2011, Mittal & Kroos, 2009a, Mittal & Kroos, 2009b, Robinson et al., 2014, Son et al., 2011). In particular, transcription of the dev operon appears to be activated by cooperative binding of the two transcription factors at two sites located upstream of the promoter (Campbell et al., 2015). Because mutations in three genes of the dev operon (devTRS) strongly impair sporulation (Boysen et al., 2002, Thony-Meyer & Kaiser, 1993, Viswanathan et al., 2007a), the feed-forward loop involving MrpC and FruA regulation of the dev operon is an attractive molecular mechanism to control spore formation (Fig. 1). Recent work revealed that products of the dev operon act as a timer for sporulation (Rajagopalan & Kroos, 2017). DevTRS negatively autoregulate expression of DevI, which inhibits sporulation if overproduced, and delays sporulation by about 6 h when produced normally (Rajagopalan & Kroos, 2017, Rajagopalan et al., 2015) (Fig. 1).
Expression of the dev operon and many other developmental genes depends on C-signaling (Kroos & Kaiser, 1987), which has been proposed to activate FruA (Ellehauge et al., 1998) and/or MrpC (Mittal & Kroos, 2009a) (Fig. 1), although the mechanism of C-signal transduction remains a mystery. Null mutations in the csgA gene block C-signaling and sporulation, but the mutants can be rescued by co-development with csgA+ cells which supply the C-signal (Shimkets et al., 1983). C-signaling appears to be a short-range signaling interaction that requires cells to move into alignment (Kim & Kaiser, 1990c, Kim & Kaiser, 1990b, Kroos et al., 1988), as they do during mound formation (Sager & Kaiser, 1993). Two theories about the identity of the C-signal have emerged. One theory states that the C-signal is a 17-kDa fragment of CsgA produced by the specific proteolytic activity of PopC at the cell surface (Kim & Kaiser, 1990a, Lobedanz & Sogaard-Andersen, 2003, Rolbetzki et al., 2008). The other theory is that diacylglycerols released from the inner membrane by cardiolipin phospholipase activity of intact CsgA are the C-signal (Boynton & Shimkets, 2015). However, in neither case has the signal receptor been identified, so our understanding of C-signaling is incomplete. Likewise, how C-signaling impacts recipient cells is unknown.
One way that C-signaling has been proposed to affect recipient cells is to stimulate autophosphorylation of a histidine protein kinase, which would then transfer the phosphate to FruA (Ellehauge et al., 1998). This model was attractive because FruA is similar to response regulators of two-component signal transduction systems (Ellehauge et al., 1998, Ogawa et al., 1996). Typically, a response regulator is phosphorylated by a histidine protein kinase in response to a signal, thus activating the response regulator to perform a function (Stock et al., 2000). The effects of substitutions at the predicted site of phosphorylation in FruA supported the model that FruA is activated by phosphorylation on D59 (Ellehauge et al., 1998). However, a histidine protein kinase capable of phosphorylating FruA has not been identified. Also, several observations suggest that FruA may not be phosphorylated. Most notably, D59 of FruA is present in an atypical receiver domain that lacks a conserved metal-binding residue normally required for phosphorylation to occur, and treatment of FruA with small-molecule phosphodonors did not increase its DNA-binding activity (Mittal & Kroos, 2009a). The receiver domain of FruA was shown to be necessary for cooperative binding with MrpC to DNA, so it was proposed that C-signaling may affect activity of MrpC and/or FruA (Mittal & Kroos, 2009a) (Fig. 1).
The regulation of MrpC has been reported to be complex, involving autoregulation, phosphorylation, proteolytic processing, binding to a toxin protein, and stability (Sun & Shi, 2001b, Nariya & Inouye, 2005, Nariya & Inouye, 2006, Nariya & Inouye, 2008, Schramm et al., 2012, Rajagopalan & Kroos, 2014, McLaughlin et al., 2018). Also, since MrpC is similar to CRP family transcription factors that bind cyclic nucleotides (Sun & Shi, 2001b), MrpC activity could be modulated by nucleotide binding, so there are many ways in which C-signaling could affect MrpC activity (Mittal & Kroos, 2009a).
Here, using synergistic experimental and computational approaches, we investigate the impact of C-signaling on a circuit that regulates commitment to sporulation by focusing on the feed-forward loop involving MrpC and FruA control of dev operon transcription (Fig. 1). We describe methods to systematically and quantitatively study the developmental process. Using these methods we measure the levels of GRN components in wild type and in mutants (e.g., a csgA mutant unable to produce C-signal) during the period leading up to and including commitment to spore formation. We then formulate a mathematical model for the steady-state concentration of dev mRNA and use the model to computationally predict the magnitude of potential regulatory effects of C-signaling that would be required to explain our data. By testing the predictions, some potential regulatory mechanisms are ruled out and at least ninefold activation of FruA by C-signaling is supported.
Results
M. xanthus development can be studied systematically
We first established quantitative assays to analyze cellular and molecular changes during M. xanthus development. To facilitate collection of sufficient cell numbers for counting, as well as for RNA and protein measurements, development was induced by starvation under submerged culture conditions. Cells adhere to the bottom of a plastic well or dish, and develop under a layer of buffer. Prior to cell harvest, photos were taken to document phenotypic differences between strains. As expected, wild-type strain DK1622 formed mounds by 18 h poststarvation (PS) and the mounds matured into compact, darkened fruiting bodies at later times (Fig. 2). In contrast, csgA and fruA null mutants failed to progress beyond forming loose aggregates. A devI null mutant was similar to wild type (WT), whereas a devS null mutant formed mounds slowly and they failed to darken. Developing populations were harvested at the times indicated in Figure 2 to measure cellular and molecular changes in the same populations.
To quantify changes at the cellular level, we counted the total number of cells (after fixation and dispersal, so that rod-shaped cells, spores, and cells in transition between the two were counted) and the number of sonication-resistant spores in the developing populations. We also counted the number of rod-shaped cells at the time when development was initiated by starvation (T0). By subtracting the number of sonication-resistant spores from the total cell number, we determined the number of sonication-sensitive cells. About 30% of the wild-type cells present at T0 remained as sonication-sensitive cells at 18 h PS (Fig. S1A), consistent with the suggestion that the majority of cells lyse early during development under submerged culture conditions, which was based on the total protein concentration of developing cultures (Rajagopalan & Kroos, 2014). The number of sonication-sensitive cells continued to decline after 18 h PS, reaching ∼4% of the T0 number by 48 h PS (Fig. S1A). Spores were first observed at 27 h PS and the number rose to ∼1% of the T0 number by 48 h PS (Fig. S1B). The devI mutant was similar to WT, except spores were first observed 6 h earlier at 21 h PS, as reported recently (Rajagopalan & Kroos, 2017). The csgA, fruA, and devS mutants failed to make a detectable number of spores (at a detection limit of 0.01% of the T0 number) and appeared to be slightly delayed relative to WT and the devI mutant in terms of the declining number of sonication-sensitive cells (Fig. S1). We conclude that at the cellular level during the time between 18 and 30 h PS (when we measured RNA and protein levels as described below), the developing populations decline from ∼30-40% to ∼10-20% of the initial rod number and only ∼0.5% (WT, devI) or <0.01% (csgA, fruA, devS) of the cells form sonication-resistant spores (from which the RNAs and proteins we measured would not be recovered based on control experiments). We stopped collecting samples at 30 PS because thereafter the number of sonication-sensitive cells continues to decline and the spore number continues to rise, making RNA and protein more difficult to recover quantitatively, yet many cells are committed at 30 h PS to make spores by 36 h PS even if nutrients are added (Rajagopalan & Kroos, 2014). Hence, we focused on changes at the molecular level between 18 and 30 h PS, the period leading up to and including the time that many cells commit to spore formation.
To measure RNA levels of a large number of samples, we adapted methods described previously (Rajagopalan & Kroos, 2014) to a higher-throughput robotic platform for RT-qPCR analysis. Reproducibility of the analysis was tested among biological replicates and two types of technical replicates as illustrated in Figure S2A, for each RNA to be measured, at 24 h PS, the midpoint of our focal period. No normalization was done in this experiment. Each transcript number was derived from a standard curve of genomic DNA subjected to qPCR. For each RNA, we found that the average transcript number and the standard deviation for three cDNA technical replicates from a single RNA sample, three RNA technical replicates from a single biological replicate, and three biological replicates, was not significantly different (single factor ANOVA, α = 0.05) (Fig. S2B-S2E). These results suggest that biological variation in RNA levels at 24 h PS is comparable to technical variation in preparing RNA and cDNA. In subsequent experiments, we measured RNA for at least three biological replicates and we did not perform RNA or cDNA technical replicates. We also note the high abundance of the mrpC transcript (∼10%) relative to 16S rRNA, and the lower relative abundance of the fruA (∼1%) and dev (∼0.1%) transcripts.
We have typically used 16S rRNA as an internal standard for RT-qPCR analysis during M. xanthus development (Rajagopalan & Kroos, 2014). The high abundance of mrpC transcript relative to 16S rRNA at 24 h PS (Fig. S2B and S2E) raised the possibility that rRNA decreases relative to total RNA at 18 to 30 h PS. To test this possibility, we measured the 16S rRNA level per μg of total RNA from 18 to 30 h PS. Figure S3A shows that the level does not change significantly (single factor ANOVA, α = 0.05), validating 16S rRNA as an internal standard for subsequent experiments. We also found that the total RNA yield per cell does not change significantly from 18 to 30 h PS (single factor ANOVA, α = 0.05) (Fig. S3B), consistent with the finding that the 16S rRNA level does not change significantly, since the majority of total RNA is rRNA.
To measure protein levels, a portion of each well-mixed developing population was immediately added to sample buffer, boiled, and frozen for subsequent semi-quantitative immunoblot analysis (Rajagopalan & Kroos, 2017). The rest of the population was used for cell counting and RNA analysis as described above and in the Experimental Procedures.
Levels of MrpC and FruA fail to account for the low level of dev mRNA in a csgA mutant
By systematically quantifying protein and mRNA levels during the period leading up to and including the time that cells commit to spore formation, we investigated whether the GRN shown in Figure 1 could account for observed changes over time in WT and in mutants. In particular, we were interested in whether changes in the levels of MrpC and/or FruA proteins could account for the observed changes in the level of dev mRNA, since MrpC and FruA bind cooperatively to the dev promoter region and activate transcription (Campbell et al., 2015). In WT, we found that the MrpC level did not change significantly from 18 to 30 h PS (Fig. 3A) and the FruA level rose about 1.5-fold on average (Fig. 3B), whereas the dev mRNA level rose about threefold on average (Fig. 4A). We reasoned that cooperative binding of MrpC and FruA could easily account for the larger rise in dev mRNA. We also measured the levels of mrpC and fruA mRNA. The mrpC mRNA level did not change significantly (Fig. 4B), consistent with the MrpC protein level, but the fruA mRNA level decreased about twofold on average after 18 h PS (Fig. 4C), in contrast to the rise in the FruA protein level (Fig. 3B), suggesting positive posttranscriptional regulation of the FruA level during the period of commitment to spore formation.
To investigate how C-signaling affects the GRN shown in Figure 1, we measured protein and mRNA levels in the csgA null mutant. In agreement with earlier studies suggesting that C-signaling activates FruA (Ellehauge et al., 1998) and/or MrpC (Mittal & Kroos, 2009a), we found very little dev mRNA in the csgA mutant (Fig. 4A). Notably, the large decrease in the level of dev mRNA in the csgA mutant compared with WT could not be accounted for by a large decrease in the level of MrpC or FruA. The MrpC level was elevated about 1.5-fold on average in the csgA mutant relative to WT (Fig. 3A), but the differences were not statistically significant (p < 0.05 in Student’s unpaired, two-tailed t-tests). The FruA level was diminished in the csgA mutant relative to WT, but only about twofold on average (Fig. 3B). The differences in the FruA level were statistically significant at most time points, but alone the twofold lower levels of FruA in the csgA mutant fail to account for the very low levels of dev mRNA.
The mrpC and fruA mRNA levels were diminished about twofold and 1.5-fold on average, respectively, in the csgA mutant relative to WT (Fig. 4B and 4C), but at most time points the differences were not statistically significant. The lack of significant differences in the level of fruA mRNA is especially noteworthy, since it implies that C-signaling has little or no effect on MrpC activity. The results of our fruA mRNA measurements agree with published reports using fruA-lacZ fusions (Ellehauge et al., 1998, Srinivasan & Kroos, 2004). Furthermore, we found that fruA mRNA stability is similar in the csgA mutant and in WT at 30 h PS (Fig. S4), indicating that the similar steady-state fruA mRNA level we observed (Fig. 4C) reflects a similar rate of synthesis, rather than altered synthesis compensated by altered stability. We conclude that C-signaling does not affect MrpC activity. Therefore, the low level of dev mRNA in a csgA mutant (Fig. 4A) could be due to failure to activate FruA or to dev-specific regulatory mechanisms.
To begin to characterize potential dev-specific regulatory mechanisms during the period leading up to and including commitment to sporulation, we measured protein and mRNA levels in the devS and devI null mutants. The MrpC and FruA levels were similar to WT (Fig. 3). The dev mRNA level ranged from 20-fold higher in the devS mutant than in WT at 18 h PS, to 10-fold higher at 30 h PS (Fig. 4A), consistent with negative autoregulation by DevS (and DevT and DevR) reported previously (Rajagopalan & Kroos, 2017, Rajagopalan et al., 2015). Unexpectedly, the dev mRNA level in the devI mutant was about threefold lower than in WT at 30 h PS (Fig. 4A), suggesting that DevI feeds back positively on accumulation of dev mRNA. The only other statistically significant differences were that the fruA mRNA levels in the devI and devS mutants were about twofold lower than in WT at 27 and 30 h PS (Fig. 4C). Since the FruA levels in these mutants were similar to those in WT (Fig. 3B), positive posttranscriptional regulation of FruA appeared to occur in the mutants, as well as in WT.
To complete our characterization of the GRN shown in Figure 1, we also measured protein and mRNA levels in the fruA and mrpC null mutants. We did not collect samples of the mrpC mutant at as many time points since we expected little or no expression of GRN components. As expected, neither MrpC nor FruA were detected in the mrpC mutant (Fig. S5). In the fruA mutant, the MrpC level was similar to WT and, as expected, FruA was not detected (Fig. 3). Also as expected, in the fruA mutant the fruA mRNA was not detected, the dev mRNA level was very low, and the mrpC mRNA level was similar to WT (Fig. 4). Since the mrpC mutant had an in-frame deletion of codons 74 to 229 (Sun & Shi, 2001b), we were able to design primers for RT-qPCR analysis that should detect the shorter mrpC transcript. Surprisingly, the mrpC mutant exhibited an elevated level of mrpC transcript compared with WT at 18 and 24 h PS (Fig. S6A). The result was surprising since expression of an mrpC-lacZ fusion had been reported to be abolished in the mrpC mutant, which had led to the conclusion that MrpC positively autoregulates (Sun & Shi, 2001b). We considered the possibility that the shorter transcript in the mrpC mutant is more stable than the WT transcript, but the transcript half-lives after addition of rifampicin did not differ significantly (Fig. S7). We conclude that MrpC negatively regulates the mrpC transcript level. While this work was in progress, McLaughlin et al. reached the same conclusion (McLaughlin et al., 2018). In all other respects, the mrpC mutant yielded expected results. The fruA and dev transcripts were very low (Fig. S6B and S6C), consistent with the expectations that MrpC is required to activate fruA transcription (Ueki & Inouye, 2003) and that MrpC and FruA are required to activate dev transcription (Campbell et al., 2015, Ellehauge et al., 1998, Viswanathan et al., 2007b). Also, the mrpC mutant failed to progress beyond forming loose aggregates (Fig. S8), appeared to be slightly delayed relative to WT in terms of the declining number of sonication-sensitive cells (Fig. S9A), and failed to make a detectable number of spores (at a detection limit of 0.01% of the T0 number) (Fig. S9B).
Taken together, our systematic, quantitative measurements of components of the GRN shown in Figure 1 imply that failure to activate FruA and/or dev-specific regulatory mechanisms may account for the low level of dev mRNA in a csgA mutant. Given the complex feedback architecture of dev regulation (i.e., strong negative feedback by DevTRS and weak positive feedback by DevI at 30 h PS), delineating the effects of C-signaling on the dev transcript level requires a mathematical modeling approach.
Mathematical modeling suggests several mechanisms that could explain the low level of dev mRNA in the csgA mutant
The observed small differences in the levels of MrpC and FruA in the csgA mutant relative to WT do not account for the very low level of dev mRNA in the csgA mutant. To evaluate plausible mechanisms that may explain these experimental findings, we quantitatively analyzed transcriptional regulation of dev by formulating a mathematical model that expresses the dev mRNA concentration as a function of the regulators MrpC, FruA, DevI, and DevS. MrpC and FruA bind cooperatively to the dev promoter region and activate transcription (Campbell et al., 2015). Our results suggest that DevI is a weak positive regulator and DevS is a strong negative regulator of dev transcription by 30 h PS (Fig. 4A). Incorporating these effects into a transcriptional regulation model, we express the concentration of dev mRNA as a product of three regulation functions (ΠFM, ΠI, ΠS) divided by the transcript degradation rate δdev (see Experimental Procedures for detailed explanation):
Here, we use a quasi-steady state approximation for the mRNA levels by taking advantage of the fact that mRNA decay (with half-lives typically in minutes) is much faster than our experimental measurement times (in hours). This allows us to assume a rapid equilibrium between the rate of dev transcription and the decay of its mRNA, which leads to the above equation, in which αFM, αI, δdev, a, b, c, KFM, KI and KS are parameters characterizing promoter regulation. We assume that these biochemical parameters are not a function of the genetic background and, therefore, in the strains in which dev mRNA was measured (e.g., the csgA mutant), the concentration of dev mRNA is determined by the concentrations of proteins (indicated by square brackets in the equation), more specifically the concentrations of their transcriptionally active forms (in case there is a posttranslational regulation). To estimate how the different regulation parameters (such as transcription rate, degradation rate, cooperativity constant, etc.) affect the dev mRNA level, we first constrain the model parameters by the experimental result shown in Figure 3B, [FruA]WT/[FruA]csgA ≅ 2, and search for parameters that can result in the observed 22-fold difference in [mRNAdev] in WT relative to the csgA mutant at 30 h PS (Fig. 4A).
To estimate the contribution of autoregulation by Dev proteins to their own transcription (i.e., the terms ΠI, ΠS) in WT and the csgA mutant, we employ the data from the devI and devS mutants (Fig. 4A). Specifically, we take the ratio of the dev mRNA level in WT to that in devI and devS mutants to estimate the feedback regulation from DevI and DevS, respectively (see Experimental Procedures for details). We find the contribution from DevI and DevS feedback regulation in WT to be ΠI,WT = 2.9 and ΠS,WT = 0.091, respectively. Using these values, we find the contribution from FruA and MrpC regulation to be ΠFM,WT/δdev,WT = 11. In the csgA mutant, since the dev mRNA level is very low, we assume the DevI and DevS protein levels to be low. This gives the contribution of different regulation functions as ΠI,csgA ≈ 1, ΠS,csgA ≈ 1, and ΠFM,csgA/δdev,csgA = 0.13. In summary, this analysis reveals that the twofold reduction of FruA protein observed in the csgA mutant (Fig. 3B) leads to a change of (ΠFM,WT/ΠFM,csgA)(δdev,csgA/δdev,WT) ≈ 84-fold in the FruA-and MrpC-dependent transcript regulation term. We reasoned that the observed 22-fold reduction in dev transcript in the csgA mutant relative to WT at 30 h PS (Fig. 4A) could result from a reduction in the FruA-and MrpC-dependent activation rate ΠFM and/or an increase in the transcript degradation rate δdev. In what follows we use the mathematical model to predict the magnitude of these effects that would be necessary to explain the observed 22-fold difference in [mRNAdev].
Hypothesis 1: Increase in dev transcript degradation rate in the csgA mutant
First, we estimate the difference in dev transcript degradation rate necessary to explain the observed difference in transcript level between WT and the csgA mutant. For this, we make two assumptions. First, we assume that MrpC and FruA bind to the dev promoter region with a Hill cooperativity coefficient a = 2 (i.e., the maximum for a single cooperative binding site). Second, we assume that the observed twofold difference in FruA protein level results in a twofold difference in transcriptionally active FruA. Under these assumptions, we vary the remaining unknown parameters to compute the required fold difference in transcript degradation rate for different values of promoter saturation. Our results plotted in Figure 5A show that at least a 20-fold difference in transcript degradation rate is required to explain the transcript data. This experimentally testable prediction will be assessed in a subsequent section. If the results are inconsistent with this prediction, we must conclude that at least one of the two assumptions above is invalid, resulting in the following two alternative hypotheses: the Hill coefficient of MrpC and FruA binding to the dev promoter region is much higher than a = 2 and/or the amount of transcriptionally active FruA does not scale with the measured FruA protein level (e.g., if csgA-dependent C-signaling is also involved in posttranslational activation of FruA).
Hypothesis 2: High cooperativity of MrpC and FruA binding to the dev promoter region Next, we test if a higher binding cooperativity can explain the difference in dev transcript level between WT and the csgA mutant. We compute the required cooperativity coefficient by assuming the degradation rate does not change between the two strains. Our results plotted in Figure 5B show that the minimum cooperativity coefficient required to explain the experimental results is six for low promoter saturation. In biologically realistic conditions, where promoter saturation is higher; the required cooperativity is even higher. Such a large cooperativity can only be explained if there is more than one site in the promoter region where MrpC and FruA bind with high cooperativity. We know that the dev promoter region has at least two MrpC and FruA cooperative binding sites; one is proximal upstream, whereas the other is distal upstream (Campbell et al., 2015). The distal upstream binding site appeared to boost dev promoter activity after 24 h PS, based on β-galactosidase activity from a lacZ reporter. Hence, in a subsequent section, we study the impact of a distal site deletion on different transcripts (mrpC, fruA, dev) and proteins (MrpC, FruA) to test if presence of the distal site contributes to higher cooperativity. If the results are not consistent with the model predictions, we must conclude that the fold difference in active FruA exceeds that observed for the total concentration of each protein (i.e., csgA-dependent C-signaling is involved in posttranslational activation of FruA).
Hypothesis 3: Posttranslational regulation of FruA activity
To assess the difference in active FruA level required to explain the observed difference in dev transcript level, in the absence of other effects, we fix the cooperativity coefficient at a = 2 and assume the transcript degradation rate to be unchanged between WT and the csgA mutant. We then use our model to compute the fold difference in active FruA required to achieve a 22-fold reduction in dev transcript in the csgA mutant relative to WT. Our results plotted in Figure 5C show that at least a ninefold reduction in active FruA is needed in the csgA mutant. The reduction in active FruA in the csgA mutant would presumably be due to the absence of C-signal-dependent posttranslational activation of FruA, not due to the twofold lower level of FruA protein we observed in the csgA mutant relative to WT (Fig. 3B). The reduction in active FruA may be considerably greater than ninefold if the dev promoter region approaches saturation (e.g., 20-fold at 80% saturation in Fig. 5C). Also, mathematical modeling of our data at each time point from 18 to 30 h PS yields a similar result (Fig. S10), suggesting that in WT, FruA has already been activated by C-signaling at least ninefold by 18 h PS, and perhaps as much as 30-fold if the dev promoter region approaches saturation (righthand panel in Fig. S10).
Stability of the dev transcript is unchanged in a csgA mutant
To measure the dev transcript degradation rate in WT and the csgA mutant, we compared the dev transcript levels after addition of rifampicin to block transcription at 30 h PS. The average half-life of the dev transcript in three biological replicates was 11 ± 6 min in WT and 7 ± 1 min in the csgA mutant (Fig. 6), which is not a statistically significant difference (p = 0.36 in a Student’s unpaired, two-tailed t-test). We conclude that elevated turnover does not account for the low level of dev transcript in the csgA mutant. These results allow us to rule out Hypothesis 1.
The distal upstream binding site for MrpC and FruA has little impact on the dev transcript level
In a previous study, weak cooperative binding of MrpC and FruA to a site located between positions - 254 and -229 upstream of the dev promoter appeared to boost β-galactosidase activity from a lacZ transcriptional fusion about twofold between 24 and 30 h PS, but deletion of the distal upstream site did not impair spore formation (Campbell et al., 2015). These findings suggested that the distal site has a modest impact on dev transcription that is inconsequential for sporulation. However, β-galactosidase activity from lacZ fused to dev promoter segments with different amounts of upstream DNA and integrated ectopically may not accurately reflect the contribution of the distal site to the dev transcript level. Therefore, we measured the dev transcript level in a mutant lacking the distal site (i.e., DNA between positions -254 and -228 was deleted from the M. xanthus chromosome). The level of dev transcript in the distal site mutant was similar to WT measured in the same experiment, in this case increasing about twofold from 18 to 30 h PS (Fig. 7). Likewise, there were no significant differences between the distal site mutant and WT in the levels of mrpC or fruA transcripts (Fig. S6) or the corresponding proteins (Fig. S5), with the exception that the average MrpC level was twofold lower in the mutant than in WT at 30 PS. The distal site mutant formed mounds by 18 h PS, which matured into compact, darkened fruiting bodies at later times, similar to WT (Fig. S8), and the percentages of sonication-sensitive cells and sonication-resistant spores observed for the distal site mutant were similar to WT (Fig. S9). We conclude that the distal site has little or no impact on the developmental process. In particular, the distal site does not contribute to high cooperativity of MrpC and FruA binding to the dev promoter region that could explain the higher level of dev transcript in WT than in the csgA mutant. These results allow us to rule out Hypothesis 2.
Boosting the FruA level in the csgA mutant has no effect on the dev transcript level
Having ruled out the first two hypotheses, our modeling predicts that the only viable option to explain the effect of the csgA null mutation on the dev transcript level is Hypothesis 3: at least a ninefold reduction in active FruA is needed in the csgA mutant as compared with WT. Specifically, our model showed that the low dev transcript level in the csgA mutant is not due to its twofold lower FruA level (Fig. 3B), but rather due to a failure to activate FruA in the absence of C-signaling (Fig. 5C and S10). As a result, the model predicts that in the csgA mutant most of the FruA remains inactive. To test this prediction, we integrated fruA transcriptionally fused to a vanillate-inducible promoter ectopically in the csgA mutant. Upon induction the csgA Pvan-fruA strain accumulated a similar level of FruA as WT (Fig. 8A), but the dev transcript level remained as low as in the csgA mutant (Fig. 8B). Hence, boosting the FruA level in the csgA mutant had no effect on the dev transcript level, consistent with our prediction and supporting the hypothesis that C-signaling activates FruA at least ninefold.
The boost in FruA level correlated with a boost in fruA transcript level in the csgA Pvan-fruA strain at 24 and 30 h PS (Fig. S11A). As expected, the mrpC transcript (Fig. S11B) and MrpC protein (Fig. S12) levels were similar in the csgA Pvan-fruA strain as in the csgA mutant. Induction of the csgA Pvan-fruA strain did not rescue its development since it failed to progress beyond forming loose aggregates (Fig. S13), failed to make a detectable number of spores by 48 h PS (at a detection limit of 0.01% of the T0 number; data not shown), and appeared to be slightly delayed relative to WT in terms of the declining number of sonication-sensitive cells, like the csgA mutant (Fig. S14).
As a control, Pvan-fruA was integrated ectopically in the fruA mutant. Upon induction the fruA Pvan-fruA strain formed mounds by 18 h PS and the mounds matured into compact, darkened fruiting bodies at later times, similar to WT without or with vanillate added (Fig. S15). Also, the induced fruA Pvan-fruA strain exhibited a similar number of sonication-resistant spores as WT at 36 h PS. These results show that induction of the fruA Pvan-fruA strain rescued its development, presumably because C-signaling activated FruA produced from Pvan-fruA.
Discussion
Our systematic, quantitative analysis of a key circuit in the GRN governing M. xanthus fruiting body formation implicates posttranslational regulation of FruA by C-signaling as primarily responsible for dev transcript accumulation during the period leading up to and including commitment to spore formation. Mathematical modeling of the dev transcript level allowed us to predict the magnitude of potential regulatory mechanisms. Experiments ruled out C-signal-dependent stabilization of dev mRNA or highly cooperative binding of FruA and MrpC to two sites in the dev promoter region as the explanation for the much higher dev transcript level in WT than in the csgA mutant. Although the FruA level was twofold lower in the csgA mutant than in WT (Fig. 3B and 8A), boosting the FruA level in the csgA mutant had no effect on the dev transcript level (Fig. 8B). Taken together, our experimental and computational analyses provide evidence that C-signaling activates FruA at least ninefold posttranslationally during M. xanthus development (Fig. 9). The activation of FruA may be considerably greater than ninefold if the dev promoter region approaches saturation (Fig. 5C and S10). Since efficient C-signaling requires cells to move into close proximity (Kim & Kaiser, 1990c, Kim & Kaiser, 1990b, Kroos et al., 1988), we propose that activation of FruA by C-signaling acts as a checkpoint for mound formation during the developmental process (Fig. 9).
Regulation of FruA by C-signaling
If activation of FruA by C-signaling acts as a checkpoint for mound formation, then active FruA should be present at 18 h PS since mound formation is well underway (Fig. 2). In agreement, mathematical modeling of our data using the assumptions of hypothesis 3 at each time point from 18 to 30 h PS yields a similar result (Fig. S10). This analysis implies that FruA has already been activated by C-signaling at least ninefold by 18 h PS, if the assumptions of hypothesis 3 apply. The assumption that the distal site does not contribute to high cooperativity of MrpC and FruA binding to the dev promoter region applies since the dev transcript level did not differ significantly in the distal site mutant as compared with WT at 18 or 24 h PS (Fig. 7). We did not measure dev transcript stability at 18 to 27 h PS, but at 30 h PS there was no significant difference between WT and the csgA mutant (Fig. 6). Therefore, C-signaling may have already activated FruA at least ninefold by 18 h PS, and perhaps as much as 30-fold if the dev promoter region approaches saturation (90% saturation in the righthand panel of Fig. S10). We note that during the period from 18 to 30 h PS, the dev transcript level rises, but the rise is due to positive autoregulation by DevI (Fig. 4A). Hence, active FruA may not be the limiting factor for dev transcription during this period (i.e., the dev promoter region may indeed approach saturation binding of active FruA and MrpC). The proximal upstream site in the dev promoter region, which is crucial for transcriptional activation, exhibits a higher affinity for cooperative binding of FruA and MrpC than the distal upstream site (Campbell et al., 2015) or several other sites (Robinson et al., 2014, Son et al., 2011), perhaps conferring on dev transcription a relatively low threshold for active FruA.
The mechanism of FruA activation by C-signaling is unknown. Since FruA is similar to response regulators of two-component signal transduction systems, phosphorylation by a histidine protein kinase was initially proposed to control FruA activity (Ellehauge et al., 1998, Ogawa et al., 1996). While this potential mechanism of posttranslational control cannot be ruled out, a kinase capable of phosphorylating FruA has not been identified despite considerable effort. Moreover, the atypical receiver domain of FruA and the inability of small-molecule phosphodonors to increase its DNA-binding activity suggest that FruA may not be phosphorylated (Mittal & Kroos, 2009a).
Several atypical response regulators have been shown to be active without phosphorylation and a few are regulated by ligand binding (Bourret, 2010, Desai et al., 2016). For example, the atypical receiver domain of Streptomyces venezuelae JadR1 is bound by jadomycin B, causing JadR1 to dissociate from DNA, and the acylated antibiotic undecylprodigiosin of Streptomyces coelicolor may use a similar mechanism to modulate DNA-binding activity of the atypical response regulator RedZ (Wang et al., 2009). Conceivably, FruA activity could likewise be regulated by binding of M. xanthus diacylglycerols, which have been implicated in C-signaling (Boynton & Shimkets, 2015). Alternatively, FruA could be regulated by a posttranslational modification other than phosphorylation or by binding to another protein (i.e., sequestration).
In addition to regulating FruA activity posttranslationally, C-signaling appears to regulate the FruA level posttranscriptionally. The FruA level was reproducibly twofold lower in the csgA mutant than in WT (Fig. 3B and 8A), but the fruA transcript level was not significantly different (Fig. 4C and S11A). These results suggest that positive posttranscriptional regulation of the FruA level requires C-signaling. C-signaling may increase synthesis (i.e., increase fruA mRNA accumulation slightly and also increase translation of fruA mRNA) and/or decrease turnover of FruA. We did not investigate this further because the FruA deficit in the csgA mutant could be overcome with Pvan-fruA, yet there was very little effect on the dev transcript level (Fig. 8). This demonstrates that the activity of FruA, rather than its level, primarily controls the level of dev transcript.
Regulation by Dev proteins
DevI inhibits sporulation if overexpressed, as in the devS mutant (Rajagopalan et al., 2015) (Fig. 2 and S1). Deletion of devI or the entire dev operon allows spores to begin forming about 6 h earlier than normal, but does not increase the final number of spores (Rajagopalan & Kroos, 2017) (Fig. S1). The level of MrpC was about twofold higher on average in the devI mutant than in WT at 15 h PS, perhaps accounting for the observed earlier sporulation, although the difference diminished at 18-24 h PS (Rajagopalan & Kroos, 2017), as reported here (Fig. 3A). It was concluded that DevI may transiently and weakly inhibit translation of mrpC transcripts during the period leading up to commitment, delaying sporulation (Rajagopalan & Kroos, 2017). As noted above, DevI positively autoregulates, causing a small rise in the dev transcript level by 30 h PS (Fig. 4A, 7, and 8B). Although the mechanism of this feedback loop is unknown, one possibility is that DevI inhibits negative autoregulation by DevTRS (Fig. 9).
In previous studies, mutations in devT, devR, or devS relieved negative autoregulation, resulting in ∼10-fold higher dev transcript accumulation at 24 h PS (Rajagopalan & Kroos, 2017, Rajagopalan et al., 2015). In this study, a devS mutant likewise accumulated ∼10-fold more dev transcript than WT at 24-30 h PS, and the difference was ∼20-fold at 18 and 21 h PS (Fig. 4A), suggesting that negative autoregulation mediated by DevS has a stronger effect leading up to the commitment period than during commitment. Strong negative autoregulation may promote commitment to sporulation by lowering the level of DevI, which would raise the MrpC level by relieving inhibition of translation of mrpC transcripts (Rajagopalan & Kroos, 2017). Our data suggest that negative autoregulation by DevTRS weakens during the commitment period, perhaps accounting for the observed small rise in the dev transcript level (Fig. 4A, 7, and 8B). If the elevated dev transcript level is accompanied by a small increase in the level of DevI, then DevI may inhibit translation of mrpC transcripts, causing the MrpC level to decrease slightly by 30 h PS in WT (Fig. 3A). DevI is predicted to be a 40-residue polypeptide (Rajagopalan et al., 2015) and currently no method has been devised to measure the DevI level. This is a worthwhile goal of future research, as is understanding how cells overcome DevI-mediated inhibition of sporulation (depicted in Fig. 9 as inhibition of cellular shape change).
In addition to regulating the timing of commitment to spore formation, Dev proteins appear to play a role in maturation of spores. Mutations in dev genes strongly impair expression of the exo operon (Licking et al., 2000, Rajagopalan & Kroos, 2017), which encodes proteins that help form the polysaccharide spore coat necessary to maintain cellular shape change and form mature spores (Muller et al., 2012, Ueki & Inouye, 2005).
The role of MrpC
Our results add to a growing list of observations that indicate MrpC functions differently during M. xanthus development than originally proposed. We found that MrpC negatively autoregulates accumulation of mrpC mRNA about twofold at 18 and 24 h PS (Fig. S6A), and it does so at 18 h PS without significantly altering transcript stability (Fig. S7). This contradicts an earlier study that concluded MrpC positively autoregulates, based on finding that expression of an mrpC-lacZ fusion was abolished in an mrpC mutant (Sun & Shi, 2001b). Recently, and in agreement with our result, it was reported that MrpC is a negative autoregulator that competes with MrpB for binding to the mrpC promoter region (McLaughlin et al., 2018). MrpB, likely when phosphorylated, binds to two sites upstream of the mrpC promoter and activates transcription. MrpC binds to multiple sites upstream of the mrpC promoter (Nariya & Inouye, 2006, McLaughlin et al., 2018), including two that overlap the MrpB binding sites (McLaughlin et al., 2018). Purified MrpC competes with the MrpB DNA-binding domain for binding to the overlapping sites, supporting a model in which MrpC negatively autoregulates by directly competing with phosphorylated MrpB for binding to overlapping sites (McLaughlin et al., 2018) (Fig. 9).
The role of MrpC in cellular lysis during development appears to be less prominent than originally proposed. MrpC was reported to function as an antitoxin by binding to and inhibiting activity of the MazF toxin protein, an mRNA interferase shown to be important for developmental programmed cell death (Nariya & Inouye, 2008). However, the effect of a null mutation in mazF on developmental lysis depends on the presence of a pilQ1 mutation (Boynton et al., 2013, Lee et al., 2012). In pilQ+ backgrounds such as our WT strain DK1622, MazF is dispensable for lysis. Here, we found only a slight delay of the mrpC mutant relative to WT in terms of the declining number of sonication-sensitive cells at 18-48 h PS (Fig. S9A), comparable to other mutants (csgA, fruA, devS, csgA Pvan-fruA) that were unable to form spores (Fig. S1 and S13; data not shown). Under our conditions, MrpC appears to play no special role in modulating the cell number during development.
Both the synthesis and the degradation of MrpC are regulated. Synthesis is regulated by phosphorylated MrpB and MrpC acting positively and negatively, respectively, at the level of transcription initiation as described above (McLaughlin et al., 2018) (Fig. 9). Degradation is regulated by the complex Esp signal transduction system (Cho & Zusman, 1999, Higgs et al., 2008, Schramm et al., 2012), which presumably senses a signal and controls the activity of an unidentified protease involved in MrpC turnover, thus ensuring that development proceeds at the appropriate pace (Fig. 9). Interestingly, preliminary results suggest that the Esp system does not govern the proteolysis of MrpC observed when nutrients are added at 18 h PS (Rajagopalan & Kroos, 2014) (Y. Hoang, R. Rajagopalan, and L. Kroos; unpublished data). This implies that another system senses nutrients and degrades MrpC to halt development (Fig. 9).
Combinatorial control by MrpC and FruA
Nutrient-regulated proteolysis of MrpC provides a checkpoint for starvation during the period leading up to and including commitment to sporulation (Rajagopalan & Kroos, 2014) (Fig. 9). If activation of FruA by C-signaling acts as a checkpoint for mound formation as we propose (Fig. 9), then combinatorial control by MrpC and activated FruA could ensure that only starving cells in mounds express genes that commit them to spore formation.
MrpC and FruA bind cooperatively to the promoter regions of five C-signal-dependent genes (Lee et al., 2011, Mittal & Kroos, 2009a, Mittal & Kroos, 2009b, Son et al., 2011, Campbell et al., 2015). In each case, cooperative binding to a site located just upstream of the promoter appears to activate transcription. Hence, MrpC and FruA form a type 1 coherent feed-forward loop with AND logic (Mangan & Alon, 2003). This type of loop is abundant in GRNs and can serve as a sign-sensitive delay element (Mangan & Alon, 2003, Mangan et al., 2003). The sign sensitivity refers to a difference in the network response to stimuli in the “OFF to ON” direction versus the “ON to OFF” direction. What this means for the feed-forward loop created by MrpC, FruA, and their target genes is that target gene expression is delayed as MrpC accumulates, awaiting FruA activated by C-signaling (i.e., the “OFF to ON” direction) (Fig. 9). As cells move into mounds and engage in short-range C-signaling, activated FruA would bind cooperatively with MrpC, stimulating transcription of target genes that eventually commit cells to spore formation (depicted in Fig. 9 as cellular shape change). However, if nutrients reappear prior to commitment, MrpC is degraded and transcription of target genes rapidly ceases, halting commitment to sporulation (i.e., the “ON to OFF” direction). The number of target genes may be large since MrpC binds to the promoter regions of hundreds of developmental genes based on ChIP-seq analysis, and in 13 of 15 cases cooperative binding of MrpC and FruA was observed (Robinson et al., 2014).
In addition to the feed-forward loop involving cooperative binding of MrpC and FruA to a site located just upstream of the promoter, the promoter regions of some genes have more complex architectures that confer greater dependence on C-signaling for transcriptional activation. For example, in the fmgD promoter region, binding of MrpC to an additional site that overlaps the promoter and the FruA binding site appears to repress transcription, and it has been proposed that a high level of active FruA produced by C-signaling is necessary to outcompete MrpC for binding and result in transcriptional activation (Lee et al., 2011) (Fig. S16A). In the fmgE promoter region, a distal upstream site with higher affinity for cooperative binding of MrpC and FruA appears to act negatively by competing for binding with the lower affinity site just upstream of the promoter (Son et al., 2011) (Fig. S16B). In addition to fmgD and fmgE, other genes depend more strongly on C-signaling and are expressed later during development than dev (Kroos & Kaiser, 1987). We infer that such genes require a higher level of active FruA than dev in order to be transcribed. In contrast to the dev promoter region, which may have a relatively low threshold for active FruA and therefore approach saturation binding of active FruA and MrpC at 18 h PS (Fig. S10), we predict that the promoter regions of genes essential for commitment to sporulation have more complex architectures and a higher threshold for active FruA. According to this model, C-signal-dependent activation of FruA continues after 18 h PS and the rising level of active FruA triggers commitment beginning at 24 h PS. We speculate that genes governing cellular shape change are under combinatorial control of MrpC and FruA (Fig. 9), and have a high threshold for active FruA.
Experimental Procedures
Bacterial strains, plasmids and primers
The strains, plasmids, and primers used in this study are listed in Table S1. Escherichia coli strain DH5α was used for cloning. To construct pET1, primers FruA-F-NdeI-Gibson and FruA-R-EcoRI-Gibson were used to generate PCR products using chromosomal DNA from M. xanthus strain DK1622 as template. The products were combined with NdeI-EcoRI-digested pMR3691 in a Gibson assembly reaction to enzymatically join the overlapping DNA fragments (Gibson et al., 2009). The cloned PCR product was verified by DNA sequencing. M. xanthus strains with Pvan-fruA integrated ectopically were constructed by electroporation (Kashefi& Hartzell, 1995) of pET1, selection of transformants on CTT agar containing 15 µg/ml of tetracycline (Iniesta et al., 2012), and verification by colony PCR using primers pMR3691 MCS G-F and pMR3691 MCS G-R.
Growth and development of M. xanthus
Strains of M. xanthus were grown at 32°C in CTTYE liquid medium (1% Casitone, 0.2% yeast extract, 10 mMTris-HCl [pH 8.0], 1 mM KH2PO4-K2HPO4, 8 mM MgSO4 [final pH 7.6]) with shaking at 350 rpm. CTT agar (CTTYE lacking yeast extract and solidified with 1.5% agar) was used for growth on solid medium and was supplemented with 40 µg/ml of kanamycin sulfate or 15 µg/ml of tetracycline as required. Fruiting body development under submerged culture conditions was performed using MC7 (10 mM morpholinepropanesulfonic acid [MOPS; pH 7.0], 1 mM CaCl2) as the starvation buffer as described previously (Rajagopalan & Kroos, 2014). Briefly, log-phase CTTYE cultures were centrifuged and cells were resuspended in MC7 at a density of approximately 1,000 Klett units. A 100 μl sample (designated T0) was removed, glutaraldehyde (2% final concentration) was added to fix cells, and the sample was stored at 4°C at least 24 h before total cells were quantified as described below. For each developmental sample, 1.5 ml of the cell suspension plus 10.5 ml of MC7 was added to an 8.5-cm-diameter plastic petri plate. Upon incubation at 32°C, cells adhere to the bottom of the plate and undergo development. At the indicated times developing populations were photographed through a microscope and collected as described below.
Microscopy
Images of fruiting bodies were obtained using a Leica Wild M8 microscope equipped with an Olympus E-620 digital camera. In order to quantify cells in samples collected and dispersed as described below, high resolution images were obtained with an Olympus BX51 microscope using a differential interference contrast filter and a 40× objective lens, and equipped with an Olympus DP30BW digital camera.
Sample collection
At the indicated times the submerged culture supernatant was replaced with 5 ml of fresh MC7 starvation buffer with or without inhibitors as required. Developing cells were scraped from the plate bottom using a sterile cell scraper and the entire contents were collected in a 15-ml centrifuge tube. Samples were mixed thoroughly by repeatedly (three times total) vortexing for 15 s followed by pipetting up and down 15 times. For quantification of total cells, 100 μl of the mixture was removed, glutaraldehyde (2% final concentration) was added to fix cells, and the sample was stored at 4°C for at least 24 h before counting as described below. For measurement of sonication-resistant spores, 400 μl of the mixture was removed and stored at -20°C. For immunoblot analysis, 100 μl of the mixture was added to an equal volume of 2× sample buffer (0.125 M Tris-HCl [pH 6.8], 20% glycerol, 4% sodium dodecyl sulfate [SDS], 0.2% bromophenol blue, 0.2 M dithiothreitol), boiled for 5 min, and stored at - 20°C. Immediately after collecting the three samples just described, the remaining 4.4 ml of the developing population was mixed with 0.5 ml of RNase stop solution (5% phenol [pH < 7] in ethanol), followed by rapid cooling in liquid nitrogen until almost frozen, centrifugation at 8,700 × g for 10 min at 4°C, removal of the supernatant, freezing of the cell pellet in liquid nitrogen, and storage at -80°C until RNA extraction. Control experiments with a sample collected at 30 h PS indicated that the majority of spores remain intact after boiling in 2× sample buffer or RNA extraction as described below, so the proteins and RNAs analyzed are from developing cells that have not yet formed spores.
Quantification of total cells and sonication-resistant spores
During development a small percentage of the rod-shaped cells transition to ovoid spores that become sonication-resistant. The number of sonication-resistant spores in developmental samples was quantified as described previously (Rajagopalan & Kroos, 2014). Briefly, each 400-μl sample collected as described above was thawed and sonicated for 10-s intervals three times with cooling on ice in between. A 60 μl sample was removed and ovoid spores were counted microscopically using a Neubauer counting chamber. A remaining portion of the sample was used to determine total protein concentration as described below. The total cell number, including rod-shaped cells, ovoid spores, and cells in transition between the two, was determined using the glutaraldehyde-fixed samples collected as described above. Each sample was thawed and mixed by vortexing and pipetting, then 10 or 20 μl was diluted with MC7 to 400 μl, sonicated once for 10 s, and all cells were counted microscopically. The total cell number minus the number of sonication-resistant cells was designated the number of sonication-sensitive cells (consisting primarily of rod-shaped cells) and was expressed as a percentage of the total cell number in the corresponding T0 sample (consisting only of rod-shaped cells).
RNA extraction and analysis
RNA was extracted using the hot-phenol method and the RNA was digested with DNase I (Roche) as described previously (Higgs et al., 2008). One μg of total RNA was subjected to cDNA synthesis using Superscript III reverse transcriptase (InVitrogen) and random primers (Promega), according to the instructions provided by the manufacturers. Control reactions were not subjected to cDNA synthesis. One μl of cDNA at the appropriate dilution (as determined empirically) and 20 pmol of each primer were subjected to qPCR in a 25 μl reaction using 2× reaction buffer (20 mM Tris-HCl [pH 8.3], 13 mM MgCl2, 100 mM KCl, 400 μM dNTPs, 4% DMSO, 2× SYBR Green I [Molecular Probes], 0.01% Tween 20, 0.01% NP40, and 0.01 μg/μl of Taq polymerase) as described previously (Bryant et al., 2008). qPCR was done in quadruplicate for each cDNA using a LightCycler® 480 System (Roche). A standard curve was generated for each set of qPCRs using M. xanthus wild-type strain DK1622 genomic DNA and gene expression was quantified using the relative standard curve method (user bulletin 2; Applied Biosystems). 16S rRNA was used as the internal standard for each sample. Relative transcript levels for mutants are the average of three biological replicates after each replicate was normalized to the transcript level observed for one replicate of wild type at 18 h PS in the same experiment. Transcript levels for wild type at other times PS were likewise normalized to that observed for wild type at 18 h PS in the same experiment. Since each experiment had one replicate of wild type, the relative transcript levels for wild type at times other than 18 h PS are the average of at least three biological replicates from different experiments, yet the standard deviations of these measurements were comparable to those of mutants, for which three biological replicates were measured in the same experiment. The standard deviation of the measurements for wild type at 18 h were also comparable, but in this case the transcript levels of at least three biological replicates from different experiments were normalized to their average, which was set as 1.
Immunoblot analysis
A semi-quantitative method of immunoblot analysis was devised to measure the relative levels of MrpC and FruA in many samples collected in different experiments. Equal volumes (10 μl for measurement of MrpC and 15 μl for measurement of FruA) of samples prepared for immunoblot analysis as described above were subjected to SDS-PAGE and immunoblotting as described previously (Rajagopalan & Kroos, 2014, Yoder-Himes & Kroos, 2006). On each immunoblot, a sample of the wild-type strain DK1622 at 18 h PS served as an internal control for normalization of signal intensities across immunoblots. Signals were detected using a ChemiDoc MP imaging system (Bio-Rad), with exposure times short enough to ensure signals were not saturated, and signal intensities were quantified using Image Lab 5.1 (Bio-Rad) software. After normalization to the internal control, each signal intensity was divided by the total protein concentration of a corresponding sample that had been sonicated for 10-s intervals three times as described above. After removal of a sample for spore quantification, the remaining portion was centrifuged at 10,000 × g for 1 min and the total protein concentration of the supernatant was determined using a Bradford (Bradford, 1976) assay kit (Bio-Rad). The resulting values of normalized signal intensity/total protein concentration were further normalized to the average value for all biological replicates of wild type at 18 h PS, which was set as 1. The normalized values for at least three biological replicates were used to compute the relative protein level (average and standard deviation). As observed for the relative transcript levels, the standard deviations of the relative protein levels were comparable for mutants (three biological replicates in the same experiment) and wild type (at least three biological replicates from different experiments).
Mathematical modeling
Activation of dev transcription by FruA and MrpC
FruA and MrpC bind cooperatively to the dev promoter region and activate transcription (Campbell et al., 2015). In agreement, no dev mRNA was detected in either the fruA mutant (Fig. 4A) or the mrpC mutant (Fig. 7). We represent the activation of dev transcript by FruA and MrpC using a phenomenological Hill’s function,
where αFM denotes the maximal dev transcription rate, KFM is the half-saturation constant, and a denotes the cooperativity of binding. Note that this expression will give ΠFM = 0 when [FruA] = 0 or [MrpC] = 0 (i.e., we have neglected any basal transcription rate as we did not detect dev mRNA in the fruA or mrpC mutant. The expression in brackets can be thought as the promoter occupancy probability (P in the equation below), a dimensional parameter telling what fraction of the promoters will be occupied by the transcription factors for a given value of KFM.
Note that the sensitivity of this expression to changes in the concentrations of FruA and MrpC are maximal when P∼0 and minimal near saturation when P∼1. In Figure 5 we assess how different hypotheses about the role of C-signaling in dev regulation play out at different levels of KFM. To facilitate the biological interpretation of the findings, we plot these as a function of dev promoter saturation.
Feedback regulation by Dev proteins
The dev mRNA level is further regulated by Dev proteins DevI and DevS. Our finding that the dev transcript level is lower in the devI mutant than in WT (Fig. 4A) indicates that DevI is a positive regulator of dev mRNA accumulation. In contrast, the dev transcript level in the devS mutant is significantly higher than in WT (Fig. 4A), indicating that DevS is a negative regulator of dev mRNA accumulation. Since the exact mechanisms of regulation by DevI and DevS are unclear, we assume for simplicity that these proteins regulate the dev transcript level through independent mechanisms. We model these regulation functions as follows:
Note that these functions are normalized so that ΠI = 1 for the devI mutant and ΠS = 1 for the devS mutant (i.e., when [DevI] = 0 or [DevS] = 0).
We assume that regulation by the Dev proteins is independent of that by FruA and MrpC, and the effects will be multiplicative: where, KFM, KI, and KS are the saturation constants for regulation by [FruA][MrpC], [DevI], and [DevS], respectively.
Numerical procedure to estimate unknown regulation parameters
To explain the difference in the dev mRNA level in the csgA mutant as compared with WT, in terms of perturbation of potential regulatory mechanisms, we use a mathematical approach where we constrain the FruA ratio ([FruA]WT/[FruA]csgA ≅ 2) and find the regulation parameters that can result in the observed 22-fold difference in [mRNAdev]. Specifically, we use the expression of dev transcript ratio between WT and the csgA mutant below: where,
First, we estimate the contribution from Dev protein regulation terms (ΠI, ΠS) in determining the dev transcript level in WT and the csgA mutant. Since we did not measure the Dev proteins explicitly in our experiments, we estimate their contribution in regulating dev transcription in WT by comparing the changes in transcript level in their absence (i.e., in the devI and devS mutants). Based on our transcript data for WT, and the devI and devS mutants (Fig. 4A), we have the following relations between the regulation functions; and . Using these relations, we obtain ΠI,WT = 2.9, ΠS,WT = 0.091. For the csgA mutant, assuming regulation by Dev proteins is absent due to the low dev transcript level, we have ΠI,csgA ≈ 1 and ΠS,csgA ≈ 1. With these estimates, the above expression for dev transcript ratio has three unknown parameters δR, a, PWT.
Next, we determine the required fold change in degradation rate δR for different promoter saturation probability PWT values that explains the observed 22-fold difference in dev transcript. To estimate this, we set the cooperativity constant (a) to 2 and take the fold change in FruA from the experiments, while assuming MrpC is unchanged between WT and the csgA mutant. The result is plotted in Fig. 5A. Then, we determine the required cooperativity a for different PWT values with the FruA fold change from the experiments and assuming no change in the degradation rate (δR = 1). The result is plotted in Fig. 5B. Finally, we compute the fold change in FruA with δR = 1 and a = 2 for different PWT values. The result is shown in Fig. 5C.
RNA stability
At the indicated time the submerged culture supernatant was replaced with fresh MC7 starvation buffer supplemented with 50 μg/ml of rifampicin to inhibit RNA synthesis. Samples were collected immediately (designated t0) and 8 and 16 min later for RNA extraction and analysis as described above, except for each biological replicate the transcript levels after 8 and 16 min were normalized to the transcript level at t0, which was set as 1, and the natural log of the resulting values was plotted versus minutes after rifampicin treatment and the slope of a linear fit of the data was used to compute the mRNA half-life.
Induction of Pvan-fruA
To induce expression of fruA fused to a vanillate-inducible promoter in M. xanthus, the CTTYE growth medium was supplemented with 0.5 mM vanillate when the culture reached 50 Klett units. Growth was continued until the culture reached 100 Klett units, then the culture was centrifuged and cells were resuspended at a density of approximately 1,000 Klett units in MC7 supplemented with 0.5 mM vanillate, followed by submerged culture development as described previously (Rajagopalan & Kroos, 2014).
Author contributions
Conception or design of the study: LK, OI, SS, PP
Acquisition of the data: SS, PP
Analysis or interpretation of the data: SS, PP, LK, OI
Writing of the manuscript: LK, SS, PP, OI
Acknowledgements
We thank Monique Floer for advice about high-throughput qPCR and for use of the LightCycler® 480 System. We thank Emily Titus for constructing pET1 and M. xanthus strain MET1. We thank Montserrat Elias-Arnanz and Penelope Higgs for sharing strains. This work was supported by the National Science Foundation (award MCB-1411272) and by salary support for L.K. from Michigan State University AgBioResearch.