Although DNA is the ultimate repository of biological information, deployment of its instructions is constrained by the metabolic and physiological status of the cell. To this end, bacteria have evolved intricate devices that connect exogenous signals (e.g. nutrients, physicochemical conditions) with endogenous conditions (metabolic fluxes, biochemical networks) that coordinately influence expression or performance of a large number of cellular functions. The phosphoenolpyruvate:carbohydrate-phosphotransferase system (PTS) is a bacterial multi-protein phosphorylation chain which computes extracellular (e.g. sugars) and intracellular (e.g. phosphoenolpyruvate, nitrogen) signals and translates them into post-translational regulation of target activities through protein-protein interactions. The PTS of Pseudomonas putida KT2440 encompasses one complete sugar (fructose)-related system and the three enzymes that form the so-called nitrogen-related PTS (PTSNtr), which lacks connection to transport of substrates. These two PTS branches cross-talk to each other, as the product of the fruB gene (a polyprotein EI-HPr-EIIA) can phosphorylate PtsN (EIIANtr) in vivo. This gives rise to a complex actuator device where diverse physiological inputs are ultimately translated into phosphorylation or not of PtsN (EIIANtr) which, in turn, checks the activity of key metabolic and regulatory proteins. Such a control of bacterial physiology highlights the prominence of biochemical homeostasis over genetic ruling –and not vice versa.
The term phosphoenolpyruvate:carbohydrate-phosphotransferase system (PTS) deals with a diverse group of enzymes and protein modules that transfer high-energy phosphate (~P) originated in phosphoenolpyruvate (PEP) from one member of the chain to the next one in a stepwise fashion (Cases et al., 2007). The PTS was first described in the mid 1960s by Kundig et al. (1964) in Escherichia coli as a carbohydrate transport system and since then it has been found in many different classes of bacteria (Postma et al., 1993; Saier & Reizer, 1994). However, as time passed it became more evident that different types of PTSs do exist, varying substantially in enzyme composition and function, and that they are not restricted to mediation of carbohydrate transport (Deutscher et al., 2006; Pflüger-Grau & Görke, 2010). A survey of the distribution of PTS domains in various genomes revealed that the main function of PTS proteins as devices for carbohydrate uptake upon phosphorylation is rather an exception found in Enterobacteriaceae, Vibrionales and Firmicutes, than a rule for all bacteria (Cases et al., 2007). Instead there is evidence that such phosphorylation-mediated transport is just one evolutionary outcome originated from a core set of what appears to originally be carbon-sensing phosphotransferases. These gravitated evolutionarily towards fulfilling a multitude of regulatory functions whereby the metabolic status of the cell, in particular the economy of carbon vs. nitrogen pools, is converted into physical actions on target proteins (Cases et al., 2007). In this sense, the PTS of every bacterium acts as a complex biological sensor-actuator device that computes biochemical signals and transduces them into mechanical outputs (De Silva, 2007; Canton et al., 2008). The best studied system is that of E. coli, where c. 60 genes encoding putative PTS proteins have been identified, although only a small fraction has been analyzed in detail (Tchieu et al., 2001). This number exposes the complexity of the PTS in E. coli compared to other bacteria where only very few PTS proteins and even just incomplete systems are found (Cases et al., 2007).
Two general branches of the PTS are known in Gram-negative bacteria. The most widespread is the canonical PTS (or sugar-related PTS; Fig. 1a), which is responsible for the concomitant phosphorylation and transport of specific carbohydrates into the bacterial cell. These PTSs generally share the core EI and HPr phosphotransfer components but they diverge at the EII complexes, the proteins that bring about sugar import specificity (Barabote & Saier, 2005: Fig. 1a). The most ancient canonical PTS seems to be the one evolved to transport fructose, from which the other sugar-transporting devices have originated (Saier et al., 2005). Detailed descriptions of the different classes of sugar-PTS can be found in Postma et al. (1993), Barabote & Saier (2005) and Saier et al. (2005). In addition to the canonical PTS bacteria often possess an abridged version that is, the so-called nitrogen-related PTS (PTSNtr), which lacks EIIB and the membrane-spanning subunit EIIC necessary for carbohydrate transport. PTSNtr is thus not involved in sugar intake but suggested to be involved in regulation of the carbon vs. nitrogen balance (Fig. 1b; Deutscher et al., 2006; Görke & Stülke, 2008; Lengeler & Jahreis, 2009; Pflüger-Grau & Görke, 2010). The controversial designation nitrogen-PTS stems from the fact that two of its encoding genes, ptsO (NPr) and ptsN (EIIANtr) map in the genomes of bacteria that carry them adjacent to rpoN, the gene encoding the σ54 sigma factor of the RNA polymerase (Fig. 2; Reizer et al., 1996; Comas et al., 2008). These two proteins, NPr and EIIANtr, form together with EINtr (encoded by the non-adjacent gene ptsP; Fig. 1b and Fig. 2) an independent chain of PTS proteins that co-exists with various configurations of the sugar-PTS devices. Either PTS type is independently active, but they may communicate under certain conditions by means of an active cross talk via phosphate group transfer (Pflüger & de Lorenzo, 2008; Zimmer et al., 2008; see below). Both PTS branches have in common that the ultimate origin of ~P is PEP. This high-energy phosphate is then transferred through three major phosphotransferase complexes: the EI, HPr/NPr, and EII enzymes (Fig. 1b). These complexes however differ between the two types in subunit composition or special features. EINtr usually carries an additional GAF domain on its N-terminal part (see below), a feature that is characteristic for the nitrogen-PTS and differentiates this EI protein from its counterpart in the sugar-PTS (Reizer et al., 1996). EINtr sequences cluster together phylogenetically and they are distant from any other EI homologues, emphasizing their separation from the sugar-PTS components (Hu & Saier, 2002).
The debated nitrogen connection of PTSNtr
Designation of the second branch of the PTS as the nitrogen-PTS has not been devoid of controversy (Ninfa, 2011), as N-related phenotypes of PTSNtr mutants of E. coli were later suspected of being artefacts (Reaves & Rabinowitz, 2011). Instead, the most clear functional role of this PTS branch in E. coli is the control of potassium uptake by direct interaction of the PtsN (EIIANtr) protein with both the TrkA subunit of the Trk transporter complex, and the sensor kinase KdpD of the two-component system regulating the expression of a second transporter device, KdpFABC (Lee et al., 2007; Lüttmann et al., 2009). Since intracellular K+ is thought to influence sigma factor competition by affecting the binding of σ70 and σS to core RNAP, it could be the case that many of the phenotypes attributed to EIIANtr could just reflect factor selectivity determined by K+ levels (Lee et al., 2010). Many reported nitrogen-related traits observed in E. coli ptsN mutants could therefore be misleading and not directly related to the PTSNtr (Reaves & Rabinowitz, 2011). However, other facts have indicated otherwise. One key observation was that the N-terminal GAF domain of the EINtr is homologous to the sensory domain of the NifA protein of Azotobacter vinelandii (Reizer et al., 1996), the σ54-dependent transcriptional regulator of the nitrogen fixation genes. In turn, the GAF domain of NifA is known to bind α-ketoglutarate (Martinez-Argudo et al., 2004). Since this compound is the immediate precursor of glutamine from Krebs cycle components, this state of affairs prompted Dozot et al. (2010) to hypothesize a consistent model in which PTSNtr could integrate C and N signals by means of the regulation of EINtr by α-ketoglutarate. This somewhat inconspicuous chemical has turned out to be a major metabolic signal that orchestrates carbon catabolite repression in E. coli in respect to nitrogen and sulphur (You et al., 2013). In this context, it came as a surprise that α-ketoglutarate, which accumulates under nitrogen limitation, directly blocks glucose uptake by inhibiting the EI enzyme of the canonical PTS (Doucette et al., 2011). Since such a default EI enzyme lacks the GAF domain of the EINtr counterpart, these data were used as an argument that the sugar-related PTS is able by itself to respond to N/C ratios and the nitrogen-branch of the system is in fact unrelated to N. Subsequent work, however, leaves this question still open. In particular, Lee et al. (2013) have suggested that α-ketoglutarate and glutamine reciprocally regulate the autophosphorylation activity of EINtr (and thereby the phosphate flow through the PTSNtr) and that the GAF domain is necessary for such regulation. From this, it follows that sensing the α-ketoglutarate/glutamine ratio by EINtr could in fact be the primary signal that triggers the phosphotransfer chain PEP→EINtr→NPr→EIIANtr, thereby making a bona fide connection between the PTSNtr and nitrogen availability (note that the canonical EI enzyme activity inhibited by α-ketoglutarate was not affected by glutamine; Doucette et al., 2011). It is thus possible that one or more metabolic inputs signalling N and/or the N/C ratio can be computed by a merged PTS, although many details on how this could be remain elusive at the time of writing this mini-review. It is also possible that the sensing and actuation mechanism through which physiological signals are converted by the PTS into regulatory actions may differ among bacteria (see below).
What is PtsN (EIIANtr) good for?
PTS proteins exert their regulatory action by physical contact with their partners in a fashion that depends on their phosphorylation state and which results in either repression or activation of the corresponding biological function. A large variety of techniques have been applied for identifying such partners, typically pull-downs and two-hybrid systems (Pflüger-Grau et al., 2011; Karstens et al., 2014). While both canonical HPr and EIIA/B proteins have been found to interact with a variety of regulatory, structural and enzymatic associates (Deutscher et al., 2006), the repertoire of proteins known to be influenced by the PTSNtr is thus far limited to those bound by PtsN (EIIANtr). In most, if not all, cases the non-phosphorylated EIIANtr is the one species competent in protein–protein interaction. The best characterized interactions are those related to K+ metabolism through binding to the low affinity potassium transporter TrkA (Lee et al., 2007). In addition, EIIANtr regulates expression of the high-affinity K+ transporter KdpFABC by interacting with sensor kinase KdpD, which results in increased phosphorylation of the cognate response regulator KdpE (Lüttmann et al., 2009). As mentioned above, changes in intracellular K+ concentration may result in different cell-wide transcriptomic regimes owing to redistribution of sigma factors in the population of RNAP molecules. This makes it difficult to ascertain whether the diverse effects of knocking-out ptsN in different bacteria are direct or indirect. It is also intriguing that EIIANtr modulates the activity of the histidine kinase PhoR which controls the phosphate starvation response of E. coli (Lüttmann et al., 2012). On the other hand, EIIANtr inhibits the binding of the response regulator SsrB to its target promoter of the SPI-2 pathogenicity island of Salmonella (Choi et al., 2010). It is thus plausible that PTSNtr modulates a number of two-component systems by either targeting the sensor kinase or the response regulator. Note that neither the mechanisms nor the sign of the effect of EIIANtr on such regulatory systems are necessarily conserved: while dephosphorylated PtsN stimulates the kinase activity of KdpD (Lüttmann et al., 2009), it is the phosphatase activity that is inhibited in the sensor kinase PhoR (Lüttmann et al., 2012). In other cases EIIANtr interacts with specific metabolic enzymes such as the pyruvate dehydrogenase (PDH) complex (see below; Pflüger-Grau et al., 2011). One additional target of EIIANtr turns out to be the SpoT1 enzyme of Ralstonia eutropha, a bifunctional ppGpp synthase/hydrolase that is a key enzyme of the stringent response in this bacterium. Interplay of this enzyme with EIIANtr was not, however, observed in E. coli (Karstens et al., 2014). It is also possible that virtually all ABC transporters of Rhizobium leguminosarum (not only KdpABC) are regulated by PTSNtr in response to the cellular energy charge (Prell et al., 2012) in order to save ATP for essential processes (including K+ homeostasis). This indicates the plasticity and stickiness of the PTS for connecting dissimilar functions in different bacterial hosts.
The minimalist PTS of Pseudomonas putida
The Gram-negative soil and rhizosphere bacterium P. putida was one of the first organisms in which a PTS, specifically one for fructose was described (Sawyer et al., 1977). Identification of its molecular components had to wait until a draft sequence of the genome of P. putida KT2440 was available in the late 1990s. The sequence was finally completed and published by Nelson et al. (2002). An experimental match between the early biochemical characterization of the PTSFru and the genomic data was not produced, however until much later (Velázquez et al., 2007). This sugar-PTS includes the full set of components for fructose intake distributed in two co-expressed proteins, FruA and FruB (Fig. 2). Disruption of fruB leads to complete loss of the ability of P. putida to grow on fructose, thereby showing that the PTSFru is essential for growth on this carbohydrate and that no other fructose transporters are present in this bacterium (Velázquez et al., 2007). The multiphosphoryl transfer protein FruB carries an EIIAFru domain (140 aa), a HPr domain (86 aa), and an EI domain (548 aa), whereas the sugar transporting component (FruA) bears a EIIBFru domain (83 aa) and an EIICFru domain (340 aa). The genes fruB and fruA cluster together in the fruR/fruBKA operon (fruK encodes fructose-1-P kinase; Fig. 2), which is under the control of the fruB promoter. When cells are grown in a gluconeogenic regime (e.g. succinate as sole C source), expression of the fruBKA genes is strongly repressed by FruR, a regulator also known as the catabolite repressor/activator or the Cra protein (Chavarria et al., 2011, 2013). Such repression can be lifted only by endogenous accumulation of fructose-1-phosphate (F1P), which acts as the one and only effector of Cra/FruR (Chavarria et al., 2011; Chavarría et al., 2014). This is an important detail, as discussed below.
The fructose-PTS of P. putida is accompanied by a complete PTSNtr formed by PtsP (EINtr), PtsO (NPr) and PtsN (EIIANtr), as shown in Fig. 2. While the ptsO and ptsN genes cluster together adjacent to the rpoN gene, they seem to belong to a different transcriptional unit (Kim et al., 2013) that also includes the somewhat elusive genes rpoX and yhbJ. The RpoX protein, first thought to affect the activity or expression of rpoN (Nelson et al., 2002), turns out in reality to promote the phenomenon known as ribosome hibernation [thereby the alternative name hibernation promoting factor (HPF)], through which 70S ribosomes dimerize in stationary phase to form inactive 100S variants (Ueta et al., 2008; Polikanov et al., 2012). On the other hand, the protein YhbJ (also called RapZ) seems to be a RNA chaperone specific for some sRNAs. In E. coli, RapZ recruits the major endoribonuclease RNase E to degrade the sRNA called GlmZ, which activates translation of the glmS mRNA (encoding glucosamine-6-phosphate synthase; Kalamorz et al., 2007; Gopel et al., 2013) through RNA-RNA base-pairing. Other activities of the RapZ protein (which has homologues in bacteria lacking the PTSNtr for example YvcJ of Bacillus subtilis; Luciano et al., 2009) might occur as well. The evolutionary pressure that has resulted in such different genes as hpf (rpoX) and rapZ being co-transcribed with ptsO and ptsN at a site immediately adjacent to rpoN in P. putida (Fig. 2) is quite a mystery. Yet, the synteny of the gene cluster in many other bacteria does suggest some type of coordination of the regulatory actions of each component of the group. In reality, all genes of this cluster are expressed (with some fluctuations; Yuste et al., 2006; Kim et al., 2013) under all growth conditions tested, the same being true for the stand-alone ptsP gene encoding the EI enzyme of the PTSNtr. It is thus likely that the performance of the system relies only on the signalling between pre-existing proteins rather than in the control of their expression.
The components of the P. putida's PTS are shown in Fig. 2. As is the case in most known instances, PtsP consists of an EI domain fused to an N-terminal 137 aa GAF module. Also similarly to other bacteria, PtsO and PtsN are single-module enzymes, carrying a single NPr (83 aa) or EIIANtr (146 aa) domain, respectively (Fig. 2). Under metabolic conditions where the FruB protein is expressed, the EIIAFru-HPr-EI polyprotein can hand over high-energy phosphate from the PTSFru to the PTSNtr, thereby enabling a physiologically relevant cross-talk between the two systems in vivo (Pflüger & de Lorenzo, 2008; Fig. 3a). Since fruB expression is under the control of the FruR/Cra repressor that allows fruBKA transcription only when cells produce intracellular F1P (see above), such cross-talk is most evident when cells are grown in fructose (Chavarria et al., 2013) and not detectable at all when consuming succinate or other carbon sources that enter the central metabolism below the upper glycolytic pathways. This means that in fructose-grown cells, the PtsN protein is phosphorylated by FruB regardless of any other ~P signal coming from PtsP. In fact, the only way to supress altogether PtsN phosphorylation is to mutate both ptsP and fruB (Pflüger & de Lorenzo, 2008; Chavarria et al., 2013), as either of the two enzymes does the job equally well. Since non-phosphorylated EIIANtr protein is the form believed to generally interact with the functional targets of the PTSNtr (see above) this state of affairs predicts that some phenotypes linked to a ptsN mutation should be manifested also when P. putida cells grow on fructose. This is because virtually the whole pool of the protein in these conditions is phosphorylated and thus unable to interplay with cognate companions. In fact, when one compares core metabolic fluxes of a ptsN mutant of P. putida with a wild-type strain growing on fructose: the outcome of the two scenarios is quite similar (Chavarría et al., 2012). The practical consequence of the transfer of ~P from PTSFru to PTSNtr is that any signal that could originate in PtsP is altogether superseded by those coming from FruB in fructose-grown cells (or other conditions that allow expression of the fruBKA operon). To further examine this issue, a mathematical model describing the phosphorylation state of PtsN under different physiological states allowed determination of the phosphate flux distribution during growth on fructose (Kremling et al., 2012). Interestingly, the model predicts that during growth on this sugar the great majority of phosphate molecules (c. 80%) are trafficked through the PTSNtr. This raises new questions: on one hand, metabolic signals can be entered into the PTSNtr through either of the two EI enzymes available in P. putida. On the other hand, EIIANtr~P may not be able to back-phosphorylate the EIIBCFru transporter, as suggested by the fact that a fruB−/fruA+ strain fails to grow on fructose as sole carbon source (Velázquez et al., 2007). Regardless of details, the model of Kremling et al. (2012) suggests that the PEP/pyruvate ratio, a gross proxy of carbon fluxes through the central metabolism (Hogema et al., 1998) controls the extent of phosphorylation of PtsN.
The metabolic links of the P. putida's PTSNtr
The tight connection between the biochemical status of the cell and phosphorylation state of PtsN is also reflected by the fact that the ratio EIIANtr~P/EIIANtr changes with the growth phase of P. putida. In the wild-type strain, PtsN becomes increasingly phosphorylated during growth, to the point that the P-free protein species disappears completely in stationary cells (Pflüger & de Lorenzo, 2007). PtsN phosphorylation is thus influenced not only by the immediate carbon source and N status (provided that P. putida's PtsP behaves as in E. coli, see above), but also by the growth state of the cells. The N source also plays a role in PtsN phosphorylation: while cells grown with a rich nitrogen source show both PtsN species, a poor nitrogen source results in only the phosphorylated form being detectable (Pflüger & de Lorenzo, 2007). Since the non-phosphorylated PtsN species is the one able to contact (most often inhibiting) its targets, it seems that a general role of the protein is to post-translationally hinder their activities during C-rich and N-rich growth but unleash their action when cells face nutritional limitations. Consistent with this notion, we found that cells lacking PtsN have significantly higher carbon fluxes in the reactions of the pyruvate shunt, which could be traced to increased activity of the corresponding metabolic enzymes (Chavarría et al., 2012).
Still, the most direct proof of interaction of the PTSNtr with central metabolism of P. putida is the physical contact of PtsN with the E1 subunit of the PDH complex of this bacterium (Pflüger-Grau et al., 2011). PDH controls inter alia the flux of carbon from catabolism of carbohydrates into the Krebs cycle and thus operates as a divider of metabolic resources towards distinct physiological fates (Fig. 3b). The PDH complex is formed by multiple copies of three enzymatic subunits, the decarboxylase E1 (encoded by aceE), the acetyltransferase domain E2 (aceF) and the dihydrolipoamid dehydrogenase E3 (lpdG). These three enzymes together are responsible for the decarboxylation of pyruvate to acetyl-CoA. As this is an irreversible process, the PDH complex thereby serves as sort of a gatekeeper of the TCA cycle, as the amount of acetyl-CoA introduced into the cycle fuels its performance. From detailed analysis of PDH activities in cell extracts of various ptsN mutant backgrounds, it has been demonstrated that (1) the non-phosphorylated form of PtsN is the inhibitor of PDH activity and that (2) this mechanism does not involve phosphate transfer to/from the PDH complex. Thus, the phosphorylation state of PtsN seems to be directly communicated to PDH, thereby influencing its activity, and thus modulating the amount of acetyl-CoA produced in the cell.
The link between PtsN and acetyl-CoA could account for why a P. putida ptsN mutant accumulated about 70% more polyhydroxyalkanoates (PHAs) from octanoates than the wild type, whereas ptsO or ptsP mutants were almost totally incapable of PHAs production (Velázquez et al., 2007). At that time, we argued that the loss of PtsN is sensed by the PHA synthesis machinery as an indicator of carbon surplus with respect to other limiting nutrients. This could then be the signal that channels much of the available octanoate to the synthesis of PHAs. Conversely, the PHA polymerization machinery of cells lacking either PtsP or PtsO would detect a shortage of carbon, thereby directing octanoate to other destinations. In view of the more recent data concerning PDH (Pflüger-Grau et al., 2011) it could well happen that the extra acetyl-CoA predicted to be produced in a ptsN mutant strain is either diverted directly into the accumulation of the PHA or serves as a signal for a more intense PHA synthesis. This is consistent with the fact that a ptsO mutant produces less PHAs: since the loss of NPr results in a greater concentration of non-phosphorylated PtsN, the PDH complex is inhibited and less acetyl-CoA is produced. The same reasoning can be applied for a ptsP mutant, which also fails to accumulate PHAs (Velázquez et al., 2007). In summary, it seems that the PTSNtr of P. putida has connections to both upper (e.g. glycolysis) and lower (PDH, Krebs cycle) central metabolic routes as well as to peripheral biochemical activities such as PHA synthesis.
A more complicated scenario is posed by the role of PtsN in mediating the carbon-source repression of m-xylene catabolism encoded by the pWW0 plasmid of P. putida mt-2 (Cases et al., 1999, 2001). While the phenomenon is clear in that ptsN mutants do not display the distinct repression by glucose of the σ54-dependent and m-xylene responsive promoter Pu that drives transcription of the xyl genes for degradation of the aromatic substrate, the mechanism behind it is still uncertain. This is because not only PtsN but also other regulatory components such as the RNA-binding Crc protein (Moreno et al., 2010) and other unknown actors (Silva-Rocha & de Lorenzo, 2011) influence the expression of the xyl genes as well. This makes it difficult to distinguish direct from indirect effects while one wonders why a simple metabolic trait is subject to such an intricate regulation (Silva-Rocha et al., 2013). Sorting out the interplay between PtsN, Crc, small RNAs and the omnipresent Hfq RNA-binding factor in the regulation of catabolic systems of P. putida is currently under investigation by different research groups (Moreno et al., 2014).
The selfish metabolism – or who rules here?
The data discussed above verify that in P. putida (1) the phosphorylated form of PtsN is present in all growth stages, (2) PtsN~P accumulates in the stationary phase, (3) the non-phosphorylated PtsN appears under growth conditions not limited by C or N, (4) the lack of PtsN leads to changes of the metabolic fluxes of central carbon metabolism and (5) PtsN decreases the activity of the PDH complex. The information encrypted in the phosphorylation ratio of PtsN/PtsN~P is then decoded through protein–protein interactions into instructions for the functioning of different metabolic modules. All these results point to a role of the PTSNtr as a non-genetic, protein-based actuator of the physiological situation of the cell in response to endogenous and exogenous signals. The question is whether this is just one more element of the repertoire of devices that cells use to maintain their homeostasis or, in fact, that it unveils a different dimension of a key biological problem. Since the late 1970s the metaphor of the selfish gene as proposed by Dawkins (2006) has guided a view of evolution that takes for granted that the sole agenda of biological systems is to ensure self-perpetuation of the genes they contain. This notion is however growingly challenged by a more integrated view in which the highly interconnected network of genes, proteins and metabolites, with metabolism holding the leading role, becomes the actual driver of both short-term functioning and evolutionary adaptation on the longer run (de Lorenzo, 2014). While the flow of information in any live object is DNA→RNA→proteins→metabolism, the chain of command is just the opposite: metabolism controls gene expression and often rules directly the functioning of RNAs and proteins. Metabolism seems to have in fact its own evolutionary drive towards both conquering of new chemical landscapes (de Lorenzo, 2014) and sorting out biochemical conflicts. The rest of the cell machinery could well be submitted to this drive, an occurrence that we have termed selfish metabolism (as opposed to the selfish DNA metaphor).
Apart from well-known regulatory duties in allosteric control of enzyme activities, intermediate metabolites and small molecules are increasingly being identified as trigger agents in transcriptional attenuation, regulating riboswiches and ribozymes. Metabolites thus deploy their regulatory role not only by interacting with proteins but also with RNA for example determining translational activity. Genome, proteome and metabolome are not three distinct entities, separated from each other, but should be perceived as a highly interconnected whole (Fig. 4), in which none can exist and fulfil its duties without the control and feed-back of the other, but all subordinated to metabolism. These features account for the high efficiency and robustness of cellular physiology. In many cases, the loss of one gene can be compensated by the joint activity of the whole, leading only to minor deficiencies, many still not detectable with today's methods. However, changing the level of a core metabolite leads to drastic effects on metabolic fluxes, which in turn have an influence on transcriptional and translational activity of the whole cell (see de Lorenzo, 2014 for a thorough discussion). The interplay between these three levels, however, requires molecular mediators and we suggest that the PTSs are key go-betweens that make connections between genome, proteome, and metabolism. In P. putida we have shown that various physiological factors such as C source, N source, or growth rate are reflected in the phosphorylation state of PtsN which in turn modulates the activity of the PDH complex, thereby changing the levels of the core metabolite acetyl-CoA. In E. coli the PTSNtr was recently shown to connect carbon metabolism to potassium homeostasis, as outlined above, and this same phenomenon in P. putida is currently under investigation in our laboratories. Nevertheless, metabolic changes both sensed and transmitted through the PTSNtr are likely to have greater consequences than the mere resetting of the transcriptome and the proteome of individual cells in that they can also propagate into population behaviour for example swimming and formation of biofilms (Fig. 5). In sum, we argue that non-genetic regulation of cell homeostasis as endowed by the PTS will be key not only to understanding the functioning of extant cells but also to guiding the engineering of new traits of biotechnological interest in bacteria like P. putida (Nikel et al., 2014).
This study was supported by the BIO Program of the Spanish Ministry of Science and Innovation, the ST-FLOW and ARYSIS Contracts of the EU, the ERANET-IB Program and the PROMT Project of the CAM. The authors have no conflict of interest to declare.