Quantification of high-specificity cyclic diguanylate signaling

Abstract

Cyclic di-GMP (c-di-GMP) is a second messenger molecule that regulates the transition between sessile and motile lifestyles in bacteria. Bacteria often encode multiple diguanylate cyclase (DGC) and phosphodiesterase (PDE) enzymes that produce and degrade c-di-GMP, respectively. Because of multiple inputs into the c-di-GMP–signaling network, it is unclear whether this system functions via high or low specificity. High-specificity signaling is characterized by individual DGCs or PDEs that are specifically associated with downstream c-di-GMP–mediated responses. In contrast, low-specificity signaling is characterized by DGCs or PDEs that modulate a general signal pool, which, in turn, controls a global c-di-GMP–mediated response. To determine whether c-di-GMP functions via high or low specificity in Vibrio cholerae, we correlated the in vivo c-di-GMP concentration generated by seven DGCs, each expressed at eight different levels, to the c-di-GMP–mediated induction of biofilm formation and transcription. There was no correlation between total intracellular c-di-GMP levels and biofilm formation or gene expression when considering all states. However, individual DGCs showed a significant correlation between c-di-GMP production and c-di-GMP–mediated responses. Moreover, the rate of phenotypic change versus c-di-GMP concentration was significantly different between DGCs, suggesting that bacteria can optimize phenotypic output to c-di-GMP levels via expression or activation of specific DGCs. Our results conclusively demonstrate that c-di-GMP does not function via a simple, low-specificity signaling pathway in V. cholerae.

The second messenger cyclic di-GMP (c-di-GMP) controls the transition between motile and sessile lifestyles in the majority of bacteria by inducing biofilm formation and repressing motility. c-di-GMP is synthesized from two GTP molecules by diguanylate cyclase (DGC) enzymes encoding a GGDEF domain. Degradation of c-di-GMP is mediated by c-di-GMP–specific phosphodiesterase (PDE) enzymes encoding either an EAL or HD-GYP domain (1). A striking feature of c-di-GMP–signaling systems is the large number of DGCs and PDEs typically encoded in bacterial genomes (2). For example, the human pathogen Vibrio cholerae encodes 40 DGC domains, 20 EAL domains, and 9 HD-GYP domains (2). Therefore, it is unclear whether c-di-GMP signals via a low-specificity central hub or in high-specificity microdomains.

Signaling specificity relates to the degree in which c-di-GMP produced or degraded by individual DGCs or PDEs regulates downstream c-di-GMP–mediated responses. In a high-specificity system, c-di-GMP produced or degraded by individual DGCs or PDEs modulates one or a small subset of c-di-GMP–regulated behaviors. High-specificity signaling would be controlled with molecular mechanisms that segregate signal production with the output response. This could be accomplished with strategies such as signal sequestration in protein complexes, signal localization at distinct cellular sites, or temporal separation of signal production and signal reception (3–6). Furthermore, such mechanisms occur in eukaryotic second messenger–signaling systems (7–12), and mathematical modeling of eukaryotic cells predicts that cAMP-signaling domains as small as 50 nm can be established with precise localization of synthesis and degradation enzymes (13).

Alternatively, in a low-specificity signaling system, c-di-GMP produced or degraded by each DGC or PDE would modulate a common global signal pool. In this situation, the concentration of c-di-GMP is relatively uniform throughout the cell. c-di-GMP–mediated responses would then be regulated at specific concentrations of c-di-GMP determined by the binding parameters of the signal transduction machinery. Production or degradation of c-di-GMP by a DGC or PDE enzyme, respectively, would equally impact each downstream phenotype.

There is currently a lack of consensus on the degree of c-di-GMP–signaling specificity because evidence supports both high- and low-specificity models. A number of studies have concluded that c-di-GMP functions via high specificity based on the observation that the intracellular concentration of c-di-GMP does not directly correlate with phenotypic expression (14–17), an observation inconsistent with a low-specificity system. These studies relied on mutagenesis of respective DGCs. Although informative, these observations are qualitative in nature because deletion of a DGC only produces two states in which the intracellular concentration of c-di-GMP can be analyzed: the native and mutant states. This limitation makes it impossible to quantitatively describe the relationship between the intracellular concentration of c-di-GMP and phenotypic output for different DGCs. Furthermore, in many cases, the concentrations of c-di-GMP and expression of phenotypic outputs were not determined using identical growth conditions.

In contrast, evidence for low-specificity signaling has also been documented. DGC enzymes from divergent species showing no sequence conservation in the variable N-terminal signal recognition domain are able to cross-complement mutations of one another (18). These findings are not consistent with high-specificity signaling models that predict other properties of these enzymes, such as localization signals or protein interaction domains, would be important for activity. In these situations, the complementing DGCs are not expressed at native levels, and it is possible that overproduction of c-di-GMP “floods” the normal cellular pools, although rigorous quantification of the intracellular concentration of c-di-GMP has generally not been performed to examine this possibility. A genetic reductionist approach was used to study signaling specificity in Salmonella Enteritidis (19). In this study, a complete DGC mutant of this species was created, and each individual DGC was individually complemented at natural expression levels. Phenotypic analysis determined that active DGCs functioned redundantly, supporting a low-specificity system, although the intracellular concentration of c-di-GMP in the various DGC strains was not determined.

A rigorous examination of c-di-GMP–signaling specificity requires a determination of the intracellular concentration of c-di-GMP and expression of associated phenotypic outputs generated by different DGCs at multiple states. Importantly, these measurements must be taken from bacteria grown in identical conditions because subtle environmental changes can influence the intracellular concentration of c-di-GMP. Here, we report our studies that incorporate all of these elements to quantify c-di-GMP–signaling specificity in the human pathogen V. cholerae. Our experiments used a systematic analysis to determine the intracellular concentration ranges of c-di-GMP at which multiple DGCs induce biofilm formation and expression of the c-di-GMP–inducible gene aphA. When considering all data points generated by multiple DGCs, no correlation of the total c-di-GMP concentration with biofilm formation and transcription was observed. However, a significant correlation between c-di-GMP production and phenotypic output was observed when examining individual DGCs. These results conclusively show that c-di-GMP functions via high-specificity signaling in V. cholerae. In addition, we observed that the rate of change in c-di-GMP at which different DGCs impact phenotypic output is distinct, suggesting that different DGCs can be used by V. cholerae to generate a fast or slow induction of biofilm formation and transcription in response to changing c-di-GMP.

Results

Rationale.

To differentiate between low- or high-specificity signaling in c-di-GMP–mediated regulation of V. cholerae biofilm formation and transcription, we quantified these phenotypic responses at multiple in vivo c-di-GMP levels generated by seven distinct DGC enzymes. Low-specificity signaling predicts that a significant correlation will exist between total intracellular c-di-GMP levels and downstream effects independent of the DGC examined because c-di-GMP is uniform throughout the cell. However, high-specificity signaling predicts that little correlation will exist between the total in vivo c-di-GMP level and downstream phenotypic regulation by DGCs because individual DGCs contribute to a segregated c-di-GMP microdomain. However, significant correlation will exist between c-di-GMP and phenotypic outputs generated by a single DGC domain. Importantly, our approach allows the quantification of c-di-GMP–signaling specificity without an a priori understanding of the molecular mechanisms involved in maintaining this specificity by directly measuring the input to the system (i.e., c-di-GMP synthesized by different DGCs) and the output (i.e., phenotypic response).

Determination of the Enzymatically Active DGCs in V. cholerae.

Our approach relies on generation of multiple c-di-GMP input states by distinct DGCs. To accomplish this, we constructed expression vectors for all 40 V. cholerae DGCs under the control of the Ptac promoter on the plasmid pEVS143. This promoter is a derivative of the trp and lac promoters and is inducible with the synthetic inducer isopropyl β-D1-thiogalactopyranoside (IPTG). Induction of the Ptac promoter in V. cholerae occurs in a graded response (Fig. S1). Importantly, the expression of each DGC is driven by the same promoter and translation initiation sequence.

DGC enzymes are modular proteins encoding a variety of signal reception domains in the N terminus. Because these signaling domains control the activity of DGC enzymes, typically only a subset of DGCs is active in the cell at any given condition. To determine which DGCs are enzymatically active in the conditions examined here, each of these expression plasmids was conjugated into the wild type (WT) strain of V. cholerae, and biofilm formation was assessed using a minimum biofilm-eliminating concentration (MBEC) biofilm assay. WT V. cholerae grows to the high–cell-density quorum-sensing state, which represses biofilm formation (20, 21). Indeed, the WT strain containing an empty vector control produced minimal biofilms (Fig. 1). Furthermore, biofilm formation of a V. cholerae ΔvpsL mutant, which is unable to synthesize extracellular polysaccharide (22), formed the lowest level of biofilm formation. Overexpression of QrgB, an active DGC enzyme from the bacterium V. harveyi, induced biofilm formation as reported previously (20). Expression of each of the DGCs was induced by addition of 0.1 mM IPTG, and biofilm formation was measured at 8 h (Fig. 1). Biofilm formation was expressed as a fold increase relative to the ΔvpsL mutant. The DGC proteins showed a wide margin of activity ranging from no induction of biofilms to over 26-fold induction of biofilm formation (VC1599). We selected 18 of the DGC enzymes that showed approximately 5-fold or greater induction of biofilms for further examination.

Fig. 1.

Determining active biofilm forming DGCs. Each DGC from V. cholerae was induced with 0.1 mM IPTG, and biofilm formation was determined using the MBEC biofilm assay. The ΔvpsL mutant is unable to produce VPS and form significant biofilms. A strain containing an empty vector control and expressing a known active DGC (qrgB) are also indicated. Biofilm fold change is expressed relative to the ΔvpsL mutant. Three biological replicates were performed, and the error bars show the SD.

Quantification of Biofilm Formation Using Flow Cytometry.

Because the intracellular levels of c-di-GMP are constantly in flux and rapidly respond to changes in environmental conditions (23), to rigorously examine signaling specificity, c-di-GMP must be quantified from the identical conditions at which the phenotypic outputs are measured. Because it is difficult to reproducibly measure c-di-GMP from a surface attached biofilm, we developed a method that allowed us to quantify biofilm formation and extract c-di-GMP concurrently from cultures of V. cholerae grown planktonically. This assay is based on our observation that hyper–biofilm-forming strains of V. cholerae aggregate in culture, even when grown with vigorous shaking. We reasoned that the number and size of these aggregates could be quantified using the forward-scatter and side-scatter parameters of a bacterial population analyzed by flow cytometry. These parameters measure the relative size and granularity of a particle, respectively. A similar approach has been used to quantify biofilm formation in Pseudomonas aeruginosa (24). To test whether flow cytometry was an appropriate method to quantify biofilms, we induced expression of the DGC VC1599 using eight different concentrations of IPTG consisting of a threefold dilution series from a maximum amount of 1 mM inducer and measured aggregate formation with flow cytometry. Induction of VC1599 showed a dose-dependent increase of the aggregate population present in the P1 gate (Fig. 2). To quantify aggregation, we multiplied the percentage of the total population present in the P1 gate by the average forward scatter of the P1 population to generate a biofilm formation index. We confirmed that aggregate formation as measured by this assay is dependent on expression of the major extracellular polysaccharide matrix in V. cholerae, vibrio polysaccharide (VPS), because induction of DGCs in a vpsL mutant did not produce aggregates (Fig. S2).

Fig. 2.

Analysis of biofilm formation using flow cytometry. Flow cytometry plots for the induction of VC1599 are shown. The x axis represents side scatter (granularity), and the y axis represents forward scatter (size). The gate P1 brackets the bacterial aggregates in the population. The concentrations (in mM) of IPTG used were: 0.0005 (A); 0.0014 (B); 0.0041 (C); 0.0123 (D); 0.037 (E); 0.11 (F); 0.33 (G); and 1.0 (H).

The 18 DGCs able to induce biofilm formation we identified using the MBEC assay (Fig. 1) were analyzed for dose-dependent biofilm induction at 8 IPTG concentrations by flow cytometry. From these experiments, nine DGCs yielding suitable biofilm-induction curves were selected for further analysis. Two of the 18 active enzymes, VC2285 and VCA0557, were discarded because these DGCs produced robust biofilms when uninduced. This is presumably attributable to high activity of these DGCs such that the low level of transcription driven by the Ptac promoter in the absence of IPTG was sufficient for full induction of biofilm formation. Seven DGCs (VC1370, VC1372, VC1376, VCA0697, VCA0939, VCA0956, and VCA0965) did not show increased biofilm induction in response to IPTG when analyzed by flow cytometry. This finding represents a discrepancy with the MBEC biofilm assay because these DGCs increased biofilm formation in that condition (Fig. 1). We hypothesize that this discrepancy between analysis of biofilm formation by the MBEC assay and analysis of biofilms using flow cytometry arises from differential activity of these DGCs in these two conditions. This differential activity could be attributable to a number of factors, including growth on a surface, density of the culture, oxygen concentration, or degree of agitation. Surface activation of a DGC has been described previously because the DGC WspR from P. aeruginosa forms active clusters when grown on a solid surface (25). This observation further highlights the importance of quantifying c-di-GMP and phenotypic output from identical growth conditions.

The biofilm formation of each of the nine suitable DGCs was then measured by flow cytometry at eight IPTG induction states in triplicate (Fig. S3). Additionally, to test whether the induction of biofilm formation by these DGCs is dependent on c-di-GMP synthesis, we mutated the Gly-Gly residues in the active site of each DGC to Ala-Ala. Indeed, mutation of the active site of each of these DGCs abolished their ability to induce biofilm formation, showing that c-di-GMP synthesis is required for the observed activity of these DGCs (Fig. S2).

Measurement of the Levels of c-di-GMP and Correlation to Biofilm Formation.

In addition to measuring biofilm formation for the nine specified DGCs, we also measured the in vivo c-di-GMP concentration in triplicate from the same cultures at each of the eight IPTG expression states using liquid chromatography combined with tandem mass spectrometry (LC-MS/MS) (Fig. S3). We observed that induction of the DGCs VC1067 and VC2224 at the highest concentrations of IPTG produced c-di-GMP levels in excess of 100 μM (Fig. S3). Because this concentration is much greater than the typical concentrations of c-di-GMP in bacteria (26), and high levels of c-di-GMP can inhibit growth of bacteria (27), these two DGCs were excluded from further analysis.

The correlation between biofilm formation and the concentration of c-di-GMP for the remaining seven DGCs at eight expression states was determined. The best-fit line of the dataset that we were able to obtain, giving an r² value of 0.10, indicated there was little correlation between global c-di-GMP levels and biofilm formation when analyzing the data points in toto (Fig. 3A). However, a reanalysis of the data to determine the correlation between c-di-GMP levels and biofilm formation separately for each individual DGC found significant correlations ranging from r² values of 0.67 to 0.99 with an average r² value for all seven DGCs of 0.78 (Fig. 3B). These data suggest that DGCs do not contribute to a global c-di-GMP pool, indicative of low-specificity signaling, but, rather, that each uniquely induces biofilm formation, indicative of high-specificity signaling.

Fig. 3.

Correlation of c-di-GMP concentrations with biofilm formation and aphA expression. The amount of biofilm formation is plotted versus the intracellular concentration of c-di-GMP for each of the eight induction states of the seven DGCs that were analyzed (A and B). Each data point represents the average of three independent measurements. Error bars are not shown for clarity. The best-fit line for the entire dataset indicates no significant correlation (A), whereas analysis of the individual DGCs shows significant correlation (B). The r² values for each DGC in B are indicated in the legend at right. Likewise, aphA expression versus c-di-GMP is shown (C and D). The best-fit line for the complete dataset (C) and individual DGCs (D) are shown. Each DGC is indicated by a different color, as displayed in the legend at right.

Interestingly, the slopes of the best fit lines for the analysis of individual DGCs were highly divergent (Fig. 3B and Fig. S4). The DGCs VC1599, VC2454, and VCA0165 had the largest slopes, indicating that these DGCs are capable of producing a large induction of biofilm formation over a relatively small change in in vivo c-di-GMP levels. We name these “fast response” DGCs. Alternatively, the DGCs VC1104, VC1216, and VC1353, exhibited intermediate slope values, termed “intermediate response,” whereas VCA0074 had the smallest slope exhibiting a gradual increase in biofilm formation from 2.5 to 25 μM c-di-GMP, termed “slow response.” These divergent responses of signal production to phenotypic output exhibited by different DGCs suggest that each enzyme can specifically impact the concentration and rate of change in c-di-GMP at which biofilm formation is induced in V. cholerae.

Correlation of c-di-GMP Concentration with aphA Gene Expression.

To determine whether high-specificity signaling exists in V. cholerae for other c-di-GMP–mediated responses besides biofilm formation, we examined the c-di-GMP–mediated induction of transcription. Two c-di-GMP–dependent transcription factors, VpsR and VpsT, have been identified in V. cholerae. VpsR directly binds to c-di-GMP to increase transcription of vpsT (28). VpsT itself then binds to c-di-GMP (29), and both c-di-GMP–bound VpsR and VpsT are thought to induce expression of the vps genes to generate the extracellular matrix necessary for biofilm formation (30). In addition to vpsT induction, c-di-GMP–bound VpsR also induces expression of the gene aphA (28), which encodes a transcription factor that activates a subset of the low–cell-density quorum sensing regulon and virulence gene cascade (31, 32). To study the impact of c-di-GMP on transcriptional control, we measured induction of a VpsR target gene. However, because multiple positive-feedback loops exist between VpsR and VpsT (30), and both of these transcription factors are c-di-GMP–dependent, we reasoned a more direct assay for VpsR activation by c-di-GMP was quantification of aphA induction.

aphA gene expression was determined upon induction of the seven DGCs described above at eight IPTG concentrations using quantitative real-time PCR (Q-PCR) in triplicate. The correlation between the in vivo c-di-GMP concentration and relative abundance (RQ) of aphA mRNA was determined (Fig. 3C). Similar to our analysis of biofilm formation, no significant correlation (r² = 0.08) between total c-di-GMP levels and aphA expression was observed when examining all data points. However, analysis of the individual DGCs revealed significant correlation between the c-di-GMP produced by specific DGCs and the subsequent expression of aphA with r² values ranging from 0.45 to 0.87 and an average r² of 0.64 (Fig. 3D). Similar to our analysis for biofilm formation, these results indicate that c-di-GMP induction of transcription occurs via high-specificity signaling.

Again, varying slopes of the best-fit correlation lines for individual DGCs were observed, revealing different DGCs generate distinct input/output response rates (Fig. 3D and Fig. S4). In the case of aphA induction, the DGCs VC1104 and VCA0165 displayed the fast response profile with rapid induction of transcription over relatively small changes in c-di-GMP levels. In contrast to these fast response DGCs, VCA0074 again showed a slow response profile with a relatively small change in aphA transcription over a much wider range of c-di-GMP (about 3–25 μM).

Discussion

One of the major unresolved questions in the field of c-di-GMP signaling is the degree of signaling specificity. Our analysis of seven DGCs each expressed at eight expression levels indicates that no significant correlation exists between the global intracellular c-di-GMP concentration and biofilm formation or induction of aphA expression. However, for both of these phenotypes, significant correlations were observed when DGCs were individually examined. These data suggest that different DGCs contribute to segregated subsets of intracellular c-di-GMP in the cell. Therefore, our findings strongly support that c-di-GMP functions via high-specificity signaling in the control of biofilm formation and VpsR-mediated transcription in V. cholerae.

Our results contrast the conclusion of low-specificity signaling obtained in Salmonella Enteritidis in which the ability of the 12 native DGCs to control c-di-GMP–mediated phenotypes was determined in a complete DGC mutant strain (19). In that study, four DGCs were found to be functionally redundant to induce cellulose synthesis. However, the levels of c-di-GMP produced by the single DGCs reinserted into the genome of the DGC-null strain were not measured, and the relationship of c-di-GMP concentration to phenotypic output was not determined as reported here. A closer examination of this relationship in Salmonella Enteritidis might reveal evidence for high-specificity signaling, or, alternatively, the c-di-GMP–signaling system of this bacterium could be inherently different from that of V. cholerae.

An important question is how the observed ranges of in vivo c-di-GMP levels relate to the normal physiological levels of c-di-GMP in the cell. We typically observe the concentration of c-di-GMP in the WT strain grown in Luria–Bertani (LB) media to a density of OD₆₀₀ 1.0, the optical density of the cultures examined in this study, to be 0.5–2 μM. Mutation of the master high–cell-density quorum-sensing regulator of V. cholerae, hapR, locks the cells into the low–cell-density state, increasing both c-di-GMP and biofilm formation (20, 33). We have found that the concentration of c-di-GMP in a ΔhapR mutant is ∼8–10 μM. However, it is important to note that we do not know whether 10 μM is the maximum physiologically relevant concentration of c-di-GMP experienced by V. cholerae because other natural environments could stimulate higher intracellular concentrations. All of the expression states examined here produced c-di-GMP at concentrations under 15 μM, with the exception of VCA0074, which produced c-di-GMP up to 25 μM. Thus, the intracellular concentration of c-di-GMP examined here are, for the most part, within the expected natural variation of the intracellular concentration of c-di-GMP in V. cholerae.

Interestingly, the rate of input to output response varied greatly between the different DGCs. We hypothesize these different rates provide V. cholerae with flexibility to appropriately modulate and adapt the regulation of behaviors by c-di-GMP in different environments. For example, the biofilm fast-response DGCs VC2454, VCA0165, and VC1599 might function in environments where a rapid increase in biofilm formation versus c-di-GMP concentration is required. Perhaps the environmental cues recognized by these DGCs are indicative of environments where biofilm formation is the most highly adapted lifestyle. In contrast, the slow-response DGC VCA0074 affects biofilm formation over a much larger range of c-di-GMP. Slow-response DGCs could gradually alter biofilm formation when fast responses are not needed or could function as fine-tuning mechanisms to cause smaller phenotypic changes. The subset of active DGCs specific for different environments would be selected either through regulation of protein expression or activation of DGCs through binding to specific cues indicative of the local environment.

The next question then becomes: what are the molecular mechanisms that lead to high-specificity signaling? One possibility is that DGCs and PDEs could be differentially localized to microdomains within the bacterial cell. The DGC PleD localizes to the stalk cell pole in the bacterium Caulobacter crescentus (3, 4). Another means of signaling specificity is association of DGCs with protein complexes. This mechanism was recently observed by the interaction of the DGC DosC and PDE DosP with the RNA modifying enzyme PNPase (6) and the interaction of DGCs with an HD-GYP protein in Xanthomonas campestris (34). Specific localization of PDEs in the cell likely plays an important role in segregating c-di-GMP–signaling pools as has been shown for cAMP signaling in eukaryotic organisms (12). Finally, it is possible that certain c-di-GMP pools are sequestered in the cell. For example, the slow response of VCA0074 for aphA induction might be attributable to the inability of the c-di-GMP that it synthesizes to interact with VpsR. In support of this hypothesis, it was shown that 80% of the cellular c-di-GMP in Gluconacetobacter xylinus is associated with the membrane (35).

The systematic approach we have undertaken is, to our knowledge, the most detailed characterization of the intracellular levels of c-di-GMP generated by multiple DGCs. This approach is widely applicable to study signaling specificity in other bacterial systems. We conclude that c-di-GMP functions via high-specificity signaling pathways in V. cholerae and our results will serve as a framework to probe molecular mechanisms by which high-specificity signaling is mediated.

Materials and Methods

DNA Manipulation and Growth Conditions.

The protocols for manipulation of DNA were performed as described previously (36). A description of construction of the DGC expression vectors can be found in SI Materials and Methods. Bacteria were grown in LB media with kanamycin (Sigma) at 100 μg/mL at 37 °C with shaking.

Quantification of Biofilm Formation and c-di-GMP.

Static biofilm formation was determined using MBEC plates as described previously (37). MBEC plates were grown for 8 h with gentle rocking before biofilms of V. cholerae were determined by staining with crystal violet. To measure biofilm formation of V. cholerae grown planktonically, 3 mL of LB containing kanamycin in 18 × 150 mm glass tubes was diluted 1/100 with an overnight culture and the inducer isopropyl IPTG was added at a concentration of 1 mM followed by a threefold dilution series seven times to yield eight IPTG-inducer concentrations. This mixture was analyzed by a Becton Dickinson LSR flow cytometer with settings of forward scatter (FSC) and side scatter (SSC) in log at voltages of 548 and 359 V, respectively, with an SSC threshold of 300 V. Biofilm formation at these eight induction states was measured independently in triplicate. Analysis of biofilm formation in Fig. S2 was done similarly, except cultures were grown in 1 mL of LB with kanamycin in 5 mL of polystyrene round-bottom tubes (BD Falcon) and induced with 0.1 mM IPTG. The intracellular concentration of c-di-GMP was determined as described in SI Materials and Methods.

Quantification of aphA Gene Expression.

RNA was harvested from the cultures described above using an RNeasy protocol (Qiagen) according to the instruction of the manufacturer. DNA was removed from extracted RNA using a DNase digestion assay supplied by Ambion according to the instructions of the manufacturer. Subsequent conversion of RNA to cDNA for Q-PCR was performed using reverse transcriptase GoScript supplied by Promega according to the instructions of the manufacturer. Q-PCR was then performed with aphA gene–specific primers and a TaqMan probe on an Applied Biosystems StepOne real-time PCR system. gyrA gene expression was used as an internal reference. Primer sequences are indicated in Table S1. Reagents and protocols for Q-PCR were conducted in accordance with those specified by Applied Biosystems.

Correlating the c-di-GMP Concentration with Phenotypic Output.

Biofilm formation and aphA transcription data were correlated to c-di-GMP levels using a number of different data transformations. When fitting a model to a set of data, we were most concerned whether or not the residuals, defined as the difference of each point from the best-fit model, were normally distributed. To test normal distribution of residuals, we used a Box–Cox transformation (38). For the logarithmic transformation reported in Fig. 3, we found that the residuals are normally distributed using a type one error rate of 0.01. The logarithmic transformation fell within the 95% confidence limit using a Box–Cox analysis; therefore, the transformation reported here was the one deemed best to normalize the residuals and yet still ensure the highest observed r² values. We attempted a number of other data transformations not reported here. Regardless of the transformation used, the r² values were similar and the r² values obtained when fitting individual DGCs were always significantly higher than that of the r² models obtained when examining all data points. Transformation and model generation were generated using the computer program Prism version 5 (GraphPad).