The Glycerol-Dependent Metabolic Persistence of Pseudomonas putida KT2440 Reflects the Regulatory Logic of the GlpR Repressor

Glycerol consumption pathways in P. putida KT2440 exposed by the genomic organization of the glp locus.Two glycerol dissimilation pathways have been described in bacteria: one begins with a dehydrogenation step, followed by phosphorylation, and the other begins with a phosphorylation reaction, followed by dehydrogenation (28, 29). In either case, the terminal metabolic currency is dihydroxyacetone-P. In the soil bacterium P. putida KT2440, the second pathway is the prevalent (and likely the only) glycerol dissimilatory process (Fig. 1). Uptake of the compound is mediated by the GlpF facilitator, which fosters a diffusion reaction (30). Once inside the cell, glycerol is phosphorylated by an ATP-dependent glycerol kinase (GlpK) to sn-glycerol-3-P (G3P), which cannot diffuse out of the cell. G3P is the substrate for GlpD, a membrane-bound G3P dehydrogenase that yields dihydroxyacetone-P (Fig. 1A). Some microorganisms, such as Escherichia coli and P. aeruginosa, are able to internalize and utilize G3P or dihydroxyacetone (28), intermediates of the biochemical polyol processing (Fig. 1A). The activities are encoded in P. putida by the glp gene cluster (26, 31), which includes glpF, glpK, glpR, and glpD (Fig. 1B). Deep RNA sequencing indicated that glpR is constitutively transcribed at a low level in the whole bacterial population irrespective of the carbon source, while the rest of the glp genes have a significant expression level only in glycerol cultures (27). Coverage plots demonstrated that there are two transcriptional units within the glp gene cluster: one transcriptional unit encompasses glpF and glpK, and glpD is independently transcribed (27). All of these genes are under the transcriptional control of the negative regulator GlpR, a regulatory arrangement previously described in other bacteria (32–34). In silico predictions have suggested that the promoters of the glp regulon share a regulatory motif in Gram-negative bacteria (35), and on the basis of this feature, we were able to recognize a well-defined GlpR-binding sequence upstream of glpD (Fig. 1B). Danilova et al. (35) also hinted at the existence of more GlpR-binding motifs upstream of genes relevant for glycerol uptake and utilization but not upstream of glpR. Against this background, we examined how all these biochemical and genetic features of P. putida KT2440 bring forth a distinctive growth phenotype on glycerol.

• Open in new tab
• Download powerpoint

FIG 1 

(A) Glycerol metabolism in Pseudomonas putida KT2440. The initial steps in the conversion of the polyol into intermediates of the central metabolism are shown along with the enzymes that catalyze these steps. Upon entrance of glycerol into the cytoplasm (mediated by GlpF, the glycerol facilitator), it is used as the substrate of GlpK (glycerol kinase). Glycerol-3-P is then oxidized in an ubiquinone-dependent reaction catalyzed by the membrane-bound GlpD (sn-glycerol-3-P dehydrogenase). Dihydroxyacetone-P (DHAP) formed therein enters into the central metabolic pathways through the activity of TpiA (triosephosphate isomerase, which establishes an equilibrium between DHAP and glyceraldehyde-3-P [GAP]) and Fda (fructose-1,6-P₂ aldolase). Further metabolic steps are indicated in this outline with wide shaded arrows. The question mark below the arrows corresponding to sn-glycerol-3-P transport indicates that no such transporter has been identified in strain KT2440. (B) Genetic organization of the glp locus. The genomic region encompasses glpF (PP1076, a major intrinsic protein [MIP] family channel protein), glpK (PP1075, glycerol kinase), glpR (PP1074, a DeoR family transcriptional regulator), and glpD (PP1073, glycerol-3-P dehydrogenase). The entire cluster is flanked upstream by PP1072, encoding an uncharacterized hypothetical protein, and downstream by PP1077, encoding a YbaK/EbsC-type protein (prolyl-tRNA editing protein). Note that the elements in this outline are not drawn to scale. The 5′ untranslated region preceding the glpD gene is shown along with the predicted glpD promoter (red) and a putative Shine-Dalgarno signal (boldface, underlined).

Behavior of single P. putida KT2440 cells growing on glycerol.When P. putida is grown on glycerol as the sole carbon source, the cultures show an unexplainably prolonged lag phase that typically lasts ≥10 h (26, 31). The conspicuously long time needed to start growth is independent of the carbon source used to grow the inocula and almost disappears (1.5 ± 0.9 h) only when cells are pregrown on the polyol. The lag phase lasted for 17.4 ± 3.9 h and 19.8 ± 2.4 h when cells were transferred from either succinate or glucose cultures into a glycerol-containing medium, respectively. Interestingly, the addition of substrate mixtures (e.g., glucose and glycerol) to the cultures helps reduce the extension of the lag phase. This situation indicates that a metabolic regulatory phenomenon (e.g., the accumulation of a critical metabolic intermediate) could be involved in the delayed growth on glycerol. Since the growth retardation was even more noticeable when P. putida was pregrown in rich LB medium and then passed into M9 minimal medium containing glycerol (the lag phase lasted >20 h), we adopted this culture approach in all the experiments described in this study. A considerable reorganization of metabolic pathways is expected to occur when cells are shifted from rich to minimal growth conditions (36), including changes in gene expression during the lag phase (37). However, the unexpectedly long lag phase of strain KT2440 on glycerol cannot be easily explained. What are the metabolic and regulatory causes behind such delayed growth of P. putida on the polyol?

The first aspect to be addressed was whether (i) the growth retardation applied to all cells of the culture in a more or less synchronous fashion or (ii) what we observed at the population level in reality reflected a stochastic start of growth of individual cells. We monitored the evolution of separate cultures of P. putida in microtiter plates inoculated with a dilution of a bacterial suspension that corresponded to an average of 1 cell per well (Fig. 2). Cells previously cultured in LB medium were used as the inoculum, and after washing them with M9 minimal medium without any carbon source (to remove any traces of the complex medium), we transferred the bacteria into M9 minimal medium supplemented with either glucose (i.e., glycolytic substrate), succinate (i.e., gluconeogenic substrate), or glycerol (in which cells adopt a mixed metabolic regime recruiting both gluconeogenic and glycolytic metabolic activities [26]). Inocula were diluted in such a way that each culture (in 96-well microtiter plates) was calculated to receive 1 cell (see Materials and Methods for details on numerical considerations). Growth in each well was periodically monitored as the change in optical density measured at 600 nm (OD₆₀₀), and results from 1,000 wells were analyzed to ensure a data set large enough to have statistical significance. Figure 2A reveals various informative features. First, the lag phase differed significantly between cultures using different carbon sources (26). Cells started to grow on glucose or succinate in an almost synchronous fashion (data not shown). In contrast, glycerol cultures starting from single cells per well showed a remarkable variation in the extension of the lag phase. Second, once bacterial growth started, the specific growth rate (μ_max) for a given carbon source was virtually identical among all individual wells. Succinate promoted the fastest growth (μ_max = 0.43 ± 0.07 h⁻¹), closely followed by glucose (μ_max = 0.39 ± 0.06 h⁻¹) and then by glycerol (μ_max = 0.24 ± 0.05 h⁻¹). These results qualitatively mirror the behavior of P. putida growing in these carbon sources in shaken-flask cultures and using a larger inoculum (26). Regardless of the extension of the lag phase and the individual μ_max values, all individual samples reached more or less the same maximal OD₆₀₀ values (in the case of glycerol cultures, ca. 0.5 to 0.6 units).

• Open in new tab
• Download powerpoint

FIG 2 

(A) Growth of P. putida KT2440 in single-cell batch cultures. Multiwell microtiter plates were inoculated with a highly diluted preculture previously developed in LB medium in order to start each culture from 1 cell per well. Cells were grown at 30°C in 200 µl of M9 minimal medium containing 40 mM glycerol with rotary agitation. The time needed to reach an optical density measured at 600 nm (OD₆₀₀) of 0.3 (mid-exponential phase of growth) is indicated by a red arrowhead for 50 independent cultures to illustrate the delay in growth initiation among individual wells. This parameter, termed time of metabolic response (t_MR), was further used to quantify the response of single-cell cultures. (B) Probability of inoculating a given number of cells in a particular well, as estimated by the Poisson probability distribution. After dilution and inoculation, and due to stochastic variations (e.g., small differences in the volume of the bacterial suspension transferred into each well), the number of cells per well (n) is not exactly known and may vary from well to well. The probability P of inoculating zero, one, two, or three cells per well is shown, indicating that the probability of inoculating a single cell is the most likely outcome.

These results made us wonder whether the asynchronous distribution of growth curves on glycerol (Fig. 2A) could be just the spurious consequence of differences in the initial number of cells per well. In other words, we questioned whether the observed phenomenon might be related to the manipulation of the cells rather than arising from physiological or genetic traits. Due to stochastic variations in the experimental procedure (e.g., differences in the inoculum volume transferred into the wells), the number of cells per well is not exactly known and may vary from well to well. Assuming that (i) the number of cells per well is proportional to its volume, (ii) the probability of having two cells in an infinitesimal volume is negligible, and (iii) inoculating a given number of cells is independent from the particular well considered, the Poisson probability distribution (38) can be used to theoretically assess the number of cells which can be expected per well (see Materials and Methods). Figure 2B shows that the probability of inoculating zero, one, two, or three cells per well is appreciable, whereas the probability of inoculating more than four cells is negligible. Inoculating a single cell per well is the most likely outcome (P = 0.343), and consequently, the growth pattern observed in each well can be associated with a sole P. putida cell. The possibility of inoculating more than 1 cell is not to be neglected, but we show that our observations are not produced by the initial variability in the number of cells per well (see below).

Single P. putida KT2440 cells grown on glycerol undergo different metabolic regimes.The results described above raised the question of how individual cells within a P. putida KT2440 population react metabolically to exposure to different carbon sources. The gross physiological state of single cells in shaken-flask cultures was examined to study this issue. Since a large number of bacterial oxidases and reductases are involved in central enzymatic functions, the merged redox activity of a cell can be considered a proxy of its overall physiological vitality. On this background, the BacLight RedoxSensor Green vitality stain, 3,8-diamino-5-[3-(diethylmethylammonio)propyl]-6-phenyl di-iodide (RSG reagent), was used to diagnose the metabolic state of individual cells. This reagent can be transformed by cell reductases into a highly fluorescent compound, a reliable descriptor of changes in the electron transport chain activity and other vital cellular processes (39, 40). As such, the metabolic heftiness of individual cells could be directly related to the intensity of the RSG signal (17). We stained P. putida cells harvested during mid-exponential growth from shaken-flask cultures with the same carbon sources described above. Bacteria were precultured in LB medium, washed with M9 minimal medium without any carbon source, and then passed into fresh M9 minimal medium supplemented with either succinate, glucose, or glycerol. Mid-exponential cultures were harvested at an OD₆₀₀ of 0.5 (note that the duration of the lag phase was different for each carbon source [26]), and cells were stained with the RSG reagent and inspected under the fluorescence microscope. The distribution of RSG-stained cells revealed different patterns depending on the substrate used (Fig. 3). On one hand, the relative intensity of RSG staining increased in the order glycerol < glucose << succinate. On the other hand, the same cell-bound signals appeared to be homogeneously distributed in the population of bacteria grown on glucose and succinate but not so in the glycerol-grown counterpart. No significant signal was observed in cells not treated with the RSG reagent, ensuring that the observed fluorescence stems from redox activity and not from autofluorescence or siderophore production (41). While small cell-to-cell differences in the RSG output of the glucose- and succinate-grown cultures can be considered normal, the clear divide between different metabolic states exposed by the data in Fig. 3 (bottom panel) prompted us to reexamine the case in more detail by means of flow cytometry. This method not only allows for the quantitation of the fluorescent signal in single cells but also enables the recognition of distinct subpopulations (42). Quantitative image analysis of the fluorescence microscopy pictures of Fig. 3 further exposes the cell-to-cell variability observed in terms of RSG staining in glycerol cultures but not in glucose or succinate cultures (see also Fig. S1 in the supplemental material).

Figure 4 shows the results obtained by flow cytometry of cells from the same shaken-flask cultures described above. In order to validate the positive correlation between RSG staining and metabolic activity of individual cells, samples from each culture condition were either treated with NaN₃ or heated at 85°C for 25 min to inactivate the cells. In either case, P. putida KT2440 showed no RSG signal in flow cytometry and cells were indistinguishable from the unstained control irrespective of the carbon source (data not shown). Various types of RSG-staining patterns quickly developed in live cells. First, as observed under the fluorescence microscope (Fig. 3), no significant differences were noticed in either the apparent size or the complexity of individual cells irrespective of the carbon source (i.e., forward scatter versus side scatter plots; Fig. 4, left panels). Second, bacteria grown on glucose or succinate were all positive for RSG staining and thus originated a single peak of fluorescence that was clearly displaced from the negative (unstained) control. This situation indicates that all cells were actively metabolizing the substrates on which they were growing. The unimodal distribution of RSG⁺ signals and their intensity matched the fluorescence microscopy data of Fig. 3, with P. putida cells from succinate cultures giving the highest output. In contrast to cultures developed under entirely glycolytic or gluconeogenic regimes, when cells grew on glycerol a sharp divide of the bacterial population into two groups became apparent. The two subpopulations differed in their degree of staining: 62% of the cells were RSG⁺, while the remaining 38% exhibited almost no fluorescence (Fig. 4, bottom panel). One of the subpopulations virtually overlapped with the negative (i.e., unstained) control experiment, meaning that these cells lacked any significant metabolic activity and can be considered metabolically dormant. The divide was consistently observed throughout the mid-exponential phase of growth but tended to gradually merge as a single bacterial pool of metabolically active cells (i.e., most of the cells became RSG⁺) as the growth progressed and the cultures moved into the late exponential phase (see Fig. S2 in the supplemental material).

• Open in new tab
• Download powerpoint

FIG 3 

Fluorescence microscopy exploration and image analysis of the BacLight RedoxSensor Green (RSG) signal in P. putida KT2440 grown on different carbon sources. Bacteria were grown in shaken-flask cultures using M9 minimal medium supplemented with the indicated carbon sources, and cells were harvested during mid-exponential growth (corresponding to an optical density measured at 600 nm of 0.5). Bacteria were stained with the RSG reagent as recommended by the manufacturer and observed under the microscope as detailed in Materials and Methods. The first two columns at the left of the figure show the microscopy pictures. The top panels indicate a control experiment (C), in which cells were not stained (note that only glucose-grown cells are shown in these panels; the same control experiment performed with cells grown on succinate or glycerol gave similar results in terms of background fluorescence). A closeup of a group of glycerol-grown P. putida cells is shown in the bottom panel to illustrate the differences in the level of fluorescence among individual bacteria. The accompanying image analysis shows two-dimensional plots of the RSG staining intensity as quantified in the output of the green (G) channel. Groups of cells were randomly selected from the fluorescence microscopy pictures for each carbon source (as indicated in the insets). In the case of cells from glycerol cultures, the cluster shown in the closeup inset of the fluorescence microscopy picture was subjected to quantitative image analysis. The background green fluorescence was subtracted from each two-dimensional plot.

• Open in new tab
• Download powerpoint

FIG 4 

Flow cytometry analysis of the BacLight RedoxSensor Green (RSG) signal in P. putida KT2440 growing on different carbon sources. Cells were grown in shaken flasks until they reached the mid-exponential phase of growth and were treated with RSG as indicated in the legend to Fig. 3, and the fluorescence intensity was quantified by flow cytometry. The gated population, in which the RSG signal was investigated, is indicated in the forward scatter (FSC) versus side scatter (SSC) plots (left column). The histograms in the right column show the distribution of RSG fluorescence in cells harvested from glucose, succinate, and glycerol cultures. The gray rectangle in each histogram plot identifies the region considered negative for the fluorescence signal (as assessed with unstained cells).

These results provide a direct link between the time-dependent metabolic bimodality of single cells grown in glycerol and the macroscopically observed lag phase of the corresponding batch cultures. Since the bimodal behavior of RSG⁺ P. putida cells was observed only in glycerol cultures and not in the presence of glucose or succinate, we hypothesized that the bifurcation was connected to the metabolism of the polyol itself. However, such plausible connection was by no means evident in the physiological experiments conducted so far, and we therefore set out to explore it with a suite of genetic and biochemical approaches.

Quantitative analysis of growth patterns.To analytically study the growth pattern under different culture conditions, we defined a kinetic parameter, termed time of metabolic response (t_MR). This parameter is the time needed by a given single-cell culture to attain an OD₆₀₀ of 0.3, which is roughly half the maximal cell density observed in glycerol cultures in microtiter plates (Fig. 2A). Note that the very definition of t_MR merges both the kinetics of cell growth and the extent of the lag phase. This concept is reminiscent of the parameters used for describing the dynamic behavior of regulatory transcriptional networks (43, 44). The characteristic t_MR values were recorded for each well in 1,000 independent cultures, and their absolute frequency was plotted against time, resulting in Gaussian-like distributions that were clearly unimodal for cells growing on either succinate or glucose as the carbon source (Fig. 5A). In sharp contrast, glycerol cultures had a multimodal distribution of t_MR values, hinted at by a scattered distribution of frequencies. Yet, is this distribution composed by more than one subpopulation, as intimated by visual inspection of Fig. 5A? To answer this question, we resorted to model-based unsupervised clustering to identify possible subpopulations. According to Table S2 in the supplemental material, two different subpopulations can be discerned according to their characteristic t_MR values. A first subpopulation comprised 89% of the entire population and exhibited an average t_MR of 23.7 ± 2.2 h, whereas a second subpopulation comprised 11% of the entire population and had a t_MR of 29.8 ± 0.7 h (Fig. 5B). Furthermore, the application of these statistical methodologies to data from succinate and glucose cultures revealed a single population for each condition, as expected from the results in Fig. 5A. In other words, our statistical analysis identified two subpopulations in glycerol cultures, clearly distinguishable by their characteristic t_MR values (Fig. 5B, inset). These results, in turn, correlate well with the bimodal distribution of cells differing in their metabolic activity (Fig. 4). We simulated a scenario where the only source of stochasticity was the initial number of cells and obtained a unimodal distribution of t_MR values (see Fig. S3 in the supplemental material). Once the phenomenon of metabolic bimodality on glycerol (which results in a stochastic distribution of growth initiation of single cells and a population-level prolonged lag phase) was well documented, the next question related to the underlying molecular mechanism.

• Open in new tab
• Download powerpoint

FIG 5 

(A) Frequency distribution of the time of metabolic response (t_MR, defined as the time needed for a given culture to attain an optical density measured at 600 nm [OD₆₀₀] of 0.3) in 1,000 single-cell P. putida KT2440 cultures carried out in microtiter plates with different carbon sources as described in the legend to Fig. 2. The bars indicate the number of independent wells in which the cultures attained an OD₆₀₀ of 0.3 at a given t_MR. (B) Closeup of the frequency of t_MR values observed in glycerol cultures. The calculated Gaussian distributions are superimposed on the bar graph to identify the possible subpopulations under these conditions (see also Table S2 in the supplemental material). The inset indicates the absolute values of the t_MR derived from the distribution shown in panel A for succinate (S), glucose (G), and glycerol cultures. Note that in the case of glycerol cultures, there are two values corresponding to the subpopulations observed under these conditions. In the inset, each bar represents the mean value of the t_MR parameter ± SD.

The GlpR repressor imposes a bistable growth pattern of cells grown on glycerol.Inspection of the biochemical and regulatory components underlying glycerol utilization in P. putida (Fig. 1A and B) and comparison of the same elements in other bacteria reveal important information about the use of this polyol by Gram-negative bacteria (45). It can be seen that (i) the glp structural genes are under the transcriptional control of a single regulatory protein (i.e., GlpR), known to act as a transcriptional repressor (34, 35); (ii) G3P seems to be the key metabolite that relieves the GlpR-dependent repression of the glp gene cluster (28); (iii) since G3P is not a substrate of the GlpF facilitator (30, 46), it can neither be taken up by P. putida cells nor leaked to the surrounding medium; and (iv) the very driving force for glycerol uptake thus arises from substrate phosphorylation by GlpK. Not surprisingly, this particular arrangement of regulatory elements is reminiscent of other genetic systems known to be submitted to stochastic fluctuations, such as the lac operon in E. coli (see below).

A typical bistable behavior at the transcriptional level is imposed by a repressor protein on the transcription of regulated genes (16, 18). The underlying mechanism proposed to explain bistability assumes a specific feedback that acts in combination with a nonlinear response within a transcriptional network (14). The archetypal example in this sense is the lac operon of E. coli, in which the LacI repressor sets a bimodal expression pattern on the cognate genes (47); the same goes for the structure of the glp genes in P. putida. Within this framework, it seems likely that the GlpR regulator and the glp genes comprise a genetic system prone to exhibiting a bistable regime. To explore this possibility, we eliminated the glpR gene in P. putida KT2440 (which would determine the constitutive expression of the glp genes) and we studied the phenotypic traits discussed in the sections above in the mutant strain.

The first indication of the involvement of GlpR in the bistable growth phenotype of P. putida came from shaken-flask culture experiments with the ΔglpR mutant using glycerol as the carbon source. Cells used as the inoculum were grown in LB medium. The duration of the lag phase was reduced to 6.5 ± 1.7 h (i.e., approximately 16 h shorter than the lag phase observed for the wild-type strain under the same culture conditions). We then focused on the catabolic enzymes involved in glycerol utilization (Fig. 6A), and we measured the GlpK activity in vitro to study the effect of eliminating GlpR in strain KT2440. Figure 6B shows that the GlpK activity is highly responsive to the carbon source. Wild-type cells growing under a glycolytic metabolic regime had the lowest GlpK activity, although the level of enzymatic activity was distinguishable from the background (attaining 7.5 ± 2.5 nmol ⋅ min⁻¹ ⋅ mg protein⁻¹). The in vitro GlpK activity increased ca. 5-fold when P. putida cells were cultured on glycerol, as reported previously (26). However, the specific GlpK activity in the ΔglpR strain did not react to the carbon source used to grow the cells. The GlpK activity was detectable even in mutant cells grown in LB medium (data not shown), a culture condition under which the same enzymatic activity was almost completely repressed in the wild-type strain. These results indicate that the elimination of the GlpR repressor protein renders the expression of glpK (and therefore the cognate enzymatic activity) constitutive. Moreover, elimination of GlpR completely abolished the bistable growth phenotype, and the entire subpopulation now showed a unimodal behavior (Fig. 6C). This evidence would in principle suffice to explain the observed phenotype of P. putida KT2440 on glycerol, and it also provides a biochemical clue to its origin. As an additional control experiment designed to ascribe the bistable growth behavior to the GlpR repressor, we cloned the cognate gene in the low-copy-number pSEVA224 expression vector (giving rise to plasmid pE-glpR [Table 1]), and the ΔglpR strain was complemented back. The expression of glpR in trans restored the stochastic profile of growth (Fig. 6C and D), causing a significant decrease in the in vitro GlpK activity (8.7 ± 1.9 nmol ⋅ min⁻¹ ⋅ mg protein⁻¹, comparable to the activity observed in the wild-type strain under similar growth conditions). The results accredit the role of GlpR as a transcriptional repressor of the cognate glp genes and suggest that the delayed growth of part of the bacterial cells is related to a limited activity of the catabolic enzymes encoded therein.

• Open in new tab
• Download powerpoint

FIG 6 

(A) Glycerol catabolic pathway in P. putida KT2440. Upon uptake of the substrate by the GlpF facilitator, it is phosphorylated by GlpK (glycerol kinase) and the resulting intermediate is further oxidized by GlpD (sn-glycerol-3-P dehydrogenase), which uses ubiquinone (UQ₈) as the cofactor (26) to yield dihydroxyacetone-P. This metabolite finally enters into the central catabolism, as indicated by a shaded arrow. (B) In vitro quantitation of the specific (Sp) GlpK enzymatic activity of cells grown in shaken-flask cultures on M9 minimal medium supplemented with either glucose or glycerol. Cells were harvested in the mid-exponential phase of growth (OD₆₀₀ of 0.5), and the activity of GlpK was determined in the cell extract as detailed in Materials and Methods. Each bar represents the mean value of the corresponding enzymatic activity ± standard deviation of duplicate measurements from at least three independent experiments. (C) Distribution of the time of metabolic response (t_MR) in 1,000 single-cell cultures (carried out as described in the legend to Fig. 2 and using glycerol as the sole carbon source) for P. putida KT2440 ΔglpR and its complemented derivative (termed the ΔglpR glpR⁺ strain, which carries plasmid pE-glpR). The calculated Gaussian distributions are superimposed on the bar graph to identify possible subpopulations under these culture conditions. (D) Absolute values of the t_MR parameter in glycerol cultures, as derived from the distribution shown in panel C. Note that in the case of the cultures of the complemented strain, there are two values corresponding to the subpopulations observed under these culture conditions. Each bar represents the mean value of the t_MR value ± standard deviation.

Time-dependent transcription of the glp genes as a descriptor of the macroscopic phenotype on glycerol.In an attempt to correlate the regulation of the glp genes and the phenotype observed in glycerol cultures, we constructed a transcriptional Φ(glpD-gfp) fusion based on the green fluorescent protein (GFP) to assess the glpD expression level (Fig. 7A). In this construct, the native promoter and the Shine-Dalgarno sequence of glpD were placed before gfp (in the pSEVA637 vector) to drive its expression. In these experiments, P. putida KT2440 cells carrying the Φ(glpD-gfp) reporter (plasmid pTF-glpD [Table 1]) were grown on glycerol in shaken-flask cultures, and broth samples were taken at selected time points to study gene expression by flow cytometry. Note that, besides a negative control (Fig. 7B), the data points at which cells were collected correspond to the lag phase (8 h, Fig. 7C), mid-exponential phase (18 h, Fig. 7D), and stationary phase (36 h, Fig. 7E), reflecting different physiological situations as the bacteria proceed through the growth curve. Figure 7C to E summarizes the evolution of the glpD transcriptional activity in strain KT2440 carrying pTF-glpD over time. At 8 h, before cells had started noticeable growth, most of the bacteria were negative for Φ(glpD-gfp) transcription, although a small subpopulation of GFP⁺ cells was consistently detected (accounting for <20% of the total population; Fig. 7C). After 18 h of glycerol-dependent growth, when cells were actively growing, the separation of two subpopulations in terms of GFP content was still evident (Fig. 7D), although the percentage of GFP⁺ cells under these circumstances increased to 57%. Moreover, the analytical deconvolution of the flow cytometry histogram indicated that the whole signal was the result of two overlapped Gaussian distributions (Fig. 7D, inset). This analysis indicates that the GFP⁻ and GFP⁺ populations represent 35% and 65% of the total fluorescence signal in the histogram, respectively. When cells attained the early stationary phase (Fig. 7E), all of them were GFP⁺ in a unimodal distribution. Interestingly, the glpD expression pattern qualitatively mirrored the distribution of metabolically active and inactive cells (Fig. 4; see also Fig. S2 in the supplemental material).

• Open in new tab
• Download powerpoint

FIG 7 

Transcriptional analysis of glpD expression in P. putida KT2440. (A) Schematic representation of the transcriptional Φ(glpD-gfp) fusion used to explore the expression of glpD. The 5′ untranslated region preceding glpD (which spans the P_glpD promoter and a Shine-Dalgarno [SD] sequence [Fig. 1B]) was cloned into the pSEVA637 vector, which carries a promoter-less gfp, generating plasmid pTF-glpD. The elements in this outline are not drawn to scale. (B to E) Time-lapse of GFP fluorescence in P. putida KT2440 cells carrying the Φ(glpD-gfp) transcriptional fusion (plasmid pTF-glpD) and grown in shaken-flask cultures in M9 minimal medium containing glycerol as the sole carbon source. The gray rectangle in each plot identifies the region considered negative for the fluorescence signal (as assessed with cells carrying the empty pSEVA637 vector). The inset in panel D identifies the percentages of GFP⁻ and GFP⁺ cells as identified by analytical deconvolution of the raw flow cytometry data.

The experimental evidence so far indicated that GlpR is ultimately responsible for the bistable growth phenotype of P. putida KT2440 on glycerol. The next question was whether the transcription of the glp genes in a ΔglpR background also reflects the macroscopic effects observed in the mutant strain. We repeated the transcriptional assessment of the Φ(glpD-gfp) expression in P. putida ΔglpR and in the same strain complemented in trans with glpR (Fig. 8), using cells transformed with the empty pSEVA637 vector as the negative control (Fig. 8A). The transcription of glpD at 18 h in the mutant strain showed an entirely unimodal distribution (Fig. 8B), indicating that the transcription of Φ(glpD-gfp) becomes constitutive in the absence of the repressor protein in a fashion similar to what was observed in terms of the overall metabolic activity (Fig. 4) and the in vitro GlpK activity (Fig. 6B). This behavior was reverted by supplying glpR in an expression plasmid (Fig. 8C), and the resulting bimodal distribution mirrored the one observed for the wild-type strain under the same growth conditions (Fig. 7D). These results accredit the role of GlpR-dependent repression of the glp genes as the main cause of the phenotypic and growth properties of P. putida KT2440, drawing a direct correlation between the transcription of the glp genes, the enzymatic activity of the cognate catabolic pathway, and the overall physiology (as exposed by the duration of the lag phase) when cells are grown en masse on glycerol.

• Open in new tab
• Download powerpoint

FIG 8 

Role of GlpR repressor and of availability of glycerol-3-P in the expression of the glp genes in P. putida. (A to C) Transcriptional analysis of the expression of Φ(glpD-gfp) (carried on plasmid pTF-glpD) in the ΔglpR background at 18 h. Cells were grown in shaken-flask cultures using glycerol as the sole carbon source. The gray rectangle in each plot identifies the region considered negative for the fluorescence signal (as assessed with cells carrying the empty vector pSEVA637). (D) In vitro quantitation of the specific (Sp) GlpK enzymatic activity in the wild-type strain grown on glycerol and carrying the empty expression vector (pSEVA224) or overexpressing both glpF and glpK from plasmid pE-glpFK. Each bar represents the mean value of the corresponding enzymatic activity ± the standard deviation of duplicate measurements from at least three independent experiments. (E) GFP fluorescence in P. putida cells carrying the Φ(glpD-gfp) transcriptional fusion in the wild-type strain overexpressing both glpF and glpK after 18 h of growth in shaken-flask cultures using glycerol as the sole carbon source.

Increasing the intracellular glycerol-3-P content relieves the bimodal growth pattern on glycerol.The driving force for the uptake of glycerol is the substrate phosphorylation to G3P by GlpK (Fig. 1A and 6A). G3P, in turn, relieves the GlpR-associated repression of the glp genes in a positive feedback loop. We reasoned that this repression could also be lifted by artificially increasing the G3P pool. Is it possible for P. putida cells to transport G3P directly from the extracellular environment? A genomic analysis of the P. putida KT2440 catalogue of glycerol-processing genes suggests that there are not specific transporters for this small molecule (26, 48), and when cells were incubated on M9 minimal medium containing G3P as the sole carbon source, no growth was observed (data not shown). An alternative possibility would be to boost the GlpK activity alone, but in this case, glycerol (the substrate for the kinase) might become limiting if the GlpF transporter (also dependent on GlpR) is not sufficiently active (49).

We attempted to increase the G3P input by overexpressing both glpF and glpK from strain KT2440 by cloning these genes in an expression plasmid (pSEVA224), giving rise to pE-glpFK (Table 1). Upon induction of the P_trc promoter, which drives the expression of glpF and glpK with isopropyl-β-d-1-thiogalactopyranoside (IPTG), we measured the specific GlpK activity in vitro in glycerol cultures of the recombinants (Fig. 8D). The overexpression of glpF and glpK from plasmid pE-glpFK resulted in a 3-fold increase in the in vitro GlpK activity compared to a control experiment with P. putida KT2440 cells carrying the empty pSEVA224 vector. The activity values obtained for the control strain in this experiment are similar to those reported in Fig. 6B, indicating that the presence of the plasmid does not alter the in vitro GlpK activity. Although we did not measure the actual G3P concentration in these cells, it can be safely assumed that the turnover of this metabolite should be also increased in the strain overexpressing glpK, as observed in Ralstonia eutropha (49). We then explored the transcription level of the glpD gene using the transcriptional fusion described above (Fig. 7A) in glycerol cultures of the wild-type KT2440 strain overexpressing glpF and glpK. At 18 h, all the cells gave a positive signal for GFP in a unimodal fashion (Fig. 8E), similarly to the distribution observed in the ΔglpR mutant under the same growth conditions (Fig. 8B). This result indicates that the GlpR-dependent repression of the glp genes is relieved by increasing the intracellular G3P availability—demonstrating that G3P is the key metabolite that mediates the bistable phenotype of P. putida KT2440 on glycerol. When the analysis of single-cell batch cultures in microtiter plates was performed with these recombinant cells, the two subpopulations previously seen for the wild-type strain were not observed (data not shown). These results are in line with the notion that if cells are pregrown on glycerol (ensuring a sufficiently high G3P supply), the stochastic growth phenomenon is no longer observed.