How do kinases regulate enzyme activity

Conformational changes induced by phosphorylation are highly dependent on the structural context of the phosphorylated protein.

Upon phosphorylation, the phosphate group regulates the activity of the protein by creating a network of hydrogen bonds among specific amino acid residues nearby. This network of hydrogen bonds is governed by the three-dimensional structure of the phosphorylated protein and therefore is unique to each protein.

The most notable example of regulation of protein function by phosphorylation-induced conformational changes is glycogen phosphorylase [ 4 ]. Glycogen phosphorylase, made up of two identical subunits, is activated upon phosphorylation of Ser of each subunit by phosphorylase kinase [ 4 ].

Phosphorylation of Ser in one monomer creates a network of hydrogen bonds between the phosphate group and the side chains of Arg of the same monomer as well as Arg of the other monomeric subunit [ 5 ]. This network induces significant intra- and intersubunit configurational changes, allowing access of the substrates to the active sites and appropriately aligning the catalytically critical residues in the active sites for catalysis of the phosphorolysis reaction.

Phosphorylation can also modulate the function of a protein by disrupting the surfaces for protein-ligand interactions without inducing any conformational changes. For example, phosphorylation of Ser of the bacterial isocitrate dehydrogenase almost completely inactivates the enzyme without inducing any significant conformational changes [ 6 , 7 ]. The phosphate group attached to Ser simply blocks binding of the enzyme to isocitrate.

Likewise, phosphorylation can also create ligand-binding surface without inducing conformational changes. For example, tyrosine phosphorylation of some cellular proteins creates the binding sites for SH2 domains and PTB domains [ 8 , 9 ]. The functions of protein kinases and phosphatases are mediated by their target substrates. Understanding how protein kinases and protein phosphatases recognise their respective substrates is one of the methods used by various investigators to elucidate the physiological functions of these important enzymes.

Before completion of the human genome project, most protein kinases were discovered after the discoveries of their physiological protein substrates.

The most notable example is phosphorylase kinase which was discovered after glycogen phosphorylase was discovered to be regulated by phosphorylation. However, in the postgenomic era, the genes encoding protein kinases and phosphatases of an organism are known upon completion of the genome project. The challenge now is to identify their physiological protein substrates.

Protein kinases employ two types of interactions to recognize their physiological substrates in cells: i recognition of the consensus phosphorylation sequence in the protein substrate by the active site of the protein kinase and ii distal interactions between the kinase and the substrate mediated by binding of docking motif spatially separated from the phosphorylation site in the substrate and interaction motif or domain located distally from the active site of the kinase [ 1 , 10 ].

These interactions contribute to the ability of protein kinases to recognize their protein substrates with exquisite specificity. Defining the structural basis of these interactions is expected to benefit identification of potential physiological substrates of protein kinases. Relevant to this, the orientated combinatorial peptide library approach developed in the s and the more recently developed positional scanning peptide library approach allow rapid determination of the optimal phosphorylation sequence of many protein kinases [ 11 , 12 ].

Notably, Mok et al. Scanning the proteomes for proteins that contain motifs similar to the optimal phosphorylation sequence of a protein kinase will assist the identification of potential physiological substrates of the kinase [ 10 ]. Armed with the knowledge of many known three-dimensional structures of protein kinases with the peptide substrate bound to the active site, Brinkworth et al.

Besides the peptide library approaches, researchers can also search for cellular proteins in crude cell or tissue lysates that are preferentially phosphorylated by a protein kinase in vitro. Finally, using specific synthetic small-molecule protein kinase inhibitors, researchers were able to perform large-scale phosphoproteomics analysis to identify physiological protein substrates of a specific protein kinase in cultured cells [ 2 ].

Substrate specificity of protein phosphatases is governed by interactions between interaction motifs or domains located distally from the phosphatase active site and distal docking motifs spatially separated from the target phosphorylation sites in protein substrates [ 17 , 18 ]. The phosphate group may prevent binding of a substrate or ligand. Being strongly negatively charged, the phosphate may disrupt electrostatic interactions between a protein and its ligand.

Alternatively, it may block ligand-binding by steric hindrance. Phosphorylation may cause a dramatic change in the conformation of the protein, as in the case of Src, where the activation loop changes to an open conformation, allowing the substrate to bind.

The phosphorylated residue in the context of the protein may be recognised by another protein. There are very many different protein kinases in eukaryotic cells and many share a common structure for their kinase domain. Variations in amino acid sequence and higher-order structure account for their substrate specificity. Though less numerous than kinases, there are also many phosphatases in eukaryotic cells. Some phosphatases are highly substrate-specific, acting on only one or two phosphoproteins, but there are others that can act on a broad range of substrates.

In the latter, regulatory domains serve to target the enzyme activity to particular substrates. Regulation by phosphorylation is a particularly common mechanism in intracellular signalling. However, other proteins that are not signalling molecules are also regulated in the same way, notably some enzymes in metabolic pathways.

An example is pyruvate dehydrogenase, which catalyses the oxidation of pyruvate, the endproduct of glycolysis, to give acetyl CoA and CO 2. Acetyl CoA then enters the citric acid cycle. Pyruvate dehydrogenase is not actually a single enzyme but is an example of a multienzyme complex see Section 6.

One of these enzymes, pyruvate decarboxylase, is inactivated by phosphorylation of a specific Ser residue. Dephosphorylation reactivates the enzyme. The kinase that catalyses the phosphorylation of pyruvate decarboxylase is subject to allosteric regulation by a number of small molecules, including acetyl CoA, pyruvate and ADP, as indicated in Figure Acetyl CoA is a positive allosteric regulator of pyruvate decarboxylase kinase which, in turn, phosphorylates and hence inactivates pyruvate carboxylase.

The activation time-course demonstrates noticeable phosphorylation of Src targets after only two minutes of light irradiation Figure 2C.

The level of activity of LightR-Src can be regulated by attenuation of light intensity Figure 2—figure supplement 1B. These data support our model for regulation of kinase activity using LightR clamp domain and demonstrate efficient and specific regulation of LightR-Src in living cells. A Analysis of LightR-Src using in vitro kinase assay. LinXE cells transiently expressing the indicated Src constructs bearing an mCherry and a myc tag at the C-terminus were exposed to continuous blue light for 60 min where indicated.

Src constructs were immunoprecipitated and their ability to phosphorylate purified N-terminal fragment of paxillin was assessed using an in vitro kinase assay. LinXE cells transiently expressing indicated LightR-Src construct bearing a tandem mCherry-myc tag at the C-terminus were continuously illuminated with blue light for the specified periods of time.

Cell lysates were probed for phosphorylation of Src substrates, paxillin and pCas. All experiments were repeated at least three times with similar results. Heatmaps represent three categories defined by the initial time of upregulation: Early responders D , 10—30 s , Intermediate responders E , 1—5 min , and Late responders F , 1 hr.

Phosphopeptides are clustered using the correlation distance metric. Columns represent relative abundances of a phosphopeptide at given timepoints normalized to 0 s.

Data shows average of three independent experiments. Supplementary source data for Figure 2D—F and Figure 2—figure supplements 2 — 4.

For this analysis, we used HeLa cell line stably expressing LightR-Src-mCherry-myc construct at levels comparable to that of endogenous Src Figure 2—figure supplement 1C. Phosphorylation of known Src targets, including caveolin1 Y14 , pCas Y12, Y, Y, and Y , paxillin Y88 and Y , p catenin Y and Y , and cortactin Y , was significantly increased in LightR-Src expressing cells, but not in control cells, exposed to blue light Figure 2—source data 1.

Principal component analysis PCA revealed that LightR-Src-expressing cells exhibited a broadly distinct phosphoproteome dynamics from control cells following light exposure Figure 2—figure supplement 2A. Protein interaction network analysis along principal component one revealed that LightR-Src-induced phosphorylation events were enriched for cell migration, cell adherens junctions, and focal adhesions, all of which are processes known to be driven by Src activation Figure 2—figure supplement 2B.

To assess the kinetics of LightR-Src signaling, we analyzed changes in the phosphoproteome at different time points after LightR-Src activation. Several distinct phosphorylation kinetics profiles were identified. Interestingly, we detected Src autophosphorylation on Y at 1 min, indicating that Src phosphorylates some targets before it even undergoes autophosphorylation. Thus, our data indicate that Src only transiently activates specific MAP kinase pathways.

Importantly, all these phosphorylation changes were not detected in control HeLa cells that were exposed to blue light but did not express LightR-Src Figure 2—figure supplement 4A—C. Overall, these results demonstrate that LightR-Src phosphorylates known Src substrates, show the fast kinetics of LightR-Src signaling within seconds, and uncover distinct temporal patterns of Src target protein phosphorylation in living cells. Since VVD dimerization is reversible Kawano et al.

To test this, LinXE cells transiently expressing LightR-Src were illuminated with blue light for 30 min and then placed in the dark for different periods of time. Our results show that incubation in the dark led to a significant decrease in phosphorylation of paxillin Figure 3A. However, it took up to 2 hr for phosphorylation to return to basal levels, indicating the slow inactivation kinetics of LightR-Src. This mutation reduces the half-life of VVD dimer in the dark from 18, s to s and thus should facilitate faster LightR-Src inactivation Zoltowski et al.

Within two minutes after the light was switched off, we observed a significant decrease in paxillin phosphorylation Figure 3B. By fifteen minutes, phosphorylation reached basal level. However, we noticed that activation of FastLightR-Src leads to a lower pCas phosphorylation level when compared to the same activation time point of LightR-Src Figure 3—figure supplement 1.

This is potentially due to the fast cycling of I85V mutants between lit and dark state Zoltowski et al. This cycling could happen even when cells are illuminated and thus would reduce the fraction of active LightR-Src molecules at a given time Zoltowski and Crane, Thus, FastLightR-Src may allow researchers to mimic function of Src kinase cycling between activation and inactivation states in living cells Kaimachnikov and Kholodenko, Overall, our results show that the off-kinetics of the LightR switch can be tuned by modifications of the VVD domains.

This provides the flexibility required for mimicking different temporal modes of kinase signaling, a capability that existing optogenetics approaches lack Wang et al. LinXE cells transiently expressing the indicated Src constructs bearing tandem mCherry-myc tag at the C-terminus were continuously exposed to blue light for specified times and then either placed in the dark for different periods of time A, B or repeatedly incubated in the dark for 10 min and in the light for 20 min C.

Cell lysates were collected and probed for phosphorylation of Src substrates. These oscillations of kinase activity can drive a specific biological response Kholodenko, ; Purvis and Lahav, ; Conlon et al. Thus, we wanted to determine whether FastLightR construct can be used to mimic oscillations of kinase activity in living cells. To test this, FastLightR-Src was activated for two periods of twenty minutes each, separated by ten minutes of deactivation.

Our results reveal successful cycles of activation and inactivation as indicated by changes in phosphorylation of pCas Figure 3C. Previous studies show that activation of Src leads to stimulation of cell spreading Klomp et al.

Therefore, we tested whether LightR-Src activation induces a similar response in living cells. Our results show that cells expressing FastLightR-Src start spreading upon irradiation with blue light Figure 4A ; Video 1 and stop immediately when the light is turned off.

Repeated irradiation of cells with blue light induced corresponding cycles of cell spreading; demonstrating again that the LightR tool can be used to mimic oscillation of kinase activity in living cells Figure 4A.

Importantly, illumination of cells expressing catalytically inactive mutant of LightR-Src DR did not induce any cell-spreading Figure 4B. Also, we observed that inactive FastLightR-Src localizes in the perinuclear region and translocates to focal adhesions and cell membrane upon activation Figure 4C ; Video 1.

Notably, this change in localization mimics that of wild type Src Kaplan et al. Overall, our results demonstrate that LightR-Src activation can mediate cell morphodynamic changes and functions similarly to what has been observed for native Src kinase. Yellow arrows point to FastLightR-Src localization at structures resembling focal adhesions.

This video cannot be played in place because your browser does support HTML5 video. You may still download the video for offline viewing.

Cell was globally illuminated with blue light 50 ms pulses every second, 50 pulses per minute as indicated. Time displayed in minutes. Current optogenetic tools enable localized kinase signaling only by re-localizing and sequestering a constitutively active kinase to an organelle or to specific areas in the cell Kerjouan, ; O'Banion et al. Light-mediated regulation of kinase catalytic activity per se has not been achieved at a subcellular level. Localized activation of protein kinases and Src is a critical determinant in the regulation of cell function.

Previous studies have suggested that local activation of Src at the cell periphery should stimulate the formation of local membrane protrusions Cary et al. However, this hypothesis has only been indirectly supported and has not been rigorously evaluated due to the limitations of existing methods. The LightR-Src approach would allow us to define the effects of local activation of Src in living cells. Indeed, we observed that local illumination of HeLa cells transiently expressing FastLightR-Src induced the formation of membrane protrusions within the illuminated area and caused polarization of the cell towards the light Figure 5A—C ; Figure 5—figure supplement 1A ; Video 2.

This effect was reversed as soon as the light was switched off Figure 5A ; Video 2. Notably, FastLightR-Src translocated to focal adhesions only in the area illuminated with blue light Figure 5D ; Figure 5—figure supplement 1B ; Video 2 and relocated back once the light was switched off Figure 5D ; Video 2. This is again consistent with known activation-dependent changes in localization of wild type Src Dhar and Shukla, ; Weernink and Rijksen, These results reveal that local activation of Src is sufficient to induce local protrusions and demonstrate that LightR approach can be used for the regulation of kinase activity at a subcellular level.

HeLa cells transiently co-expressing FastLightR-Src-mCherry-myc and Stargazin-iRFP plasma membrane marker were imaged every minute and illuminated with blue light for indicated periods of time. A Representative cell projection images showing protrusions formed between indicated time points. Blue circle outlines illuminated area. Box indicates the range percentile 25, 75 ; whiskers indicate the range outlier 1. Black dots represent individual cells.

Vertical lines indicate the region illuminated with blue light. Horizontal lines indicate the time of blue light illumination. H-J Mean velocity of the cell membrane region closest to the center of illumination spot.

Vertical lines indicate the time of blue light illumination. Images were taken every minute using total internal reflection fluorescence microscopy 60X objective. Time is displayed in minutes. Images are shown as inverted contrast image representation. Local regulation of protein kinase allows us to assess the dynamics of local morphological changes. A previous study using Src family kinases SFK biosensor suggested strong correlation between SFK activity at the cell periphery and cell-edge velocity Gulyani et al.

LightR-Src allows us to determine the direct effect of local Src activity on protrusion dynamics. To achieve this, we evaluated temporal changes in cell edge velocity upon continuous local activation of LightR-Src in HeLa cells. Our analysis revealed that Src induced local waves of increased cell-edge velocity Figure 5E,H ; Figure 5—figure supplement 2A ; suggesting the induction of recurrent local contractions that slow down membrane protrusion.

To verify that the LightR approach is applicable to other kinases, we set out to engineer LightR variants of tyrosine kinase Abl and a dual specificity kinase bRaf.

Since the majority of kinases share a conserved catalytic domain structure Fabbro et al. This variant exhibited significantly faster deactivation kinetics, with a half-life time around 15 min compared to approximately 3 hr half-life of LightR-bRaf Figure 6—figure supplement 1. We also assessed whether FastLightR-bRaf can undergo cyclic activation and deactivation by monitoring ERK2 kinase translocation into the nucleus, a known outcome of bRaf activation Burack and Shaw, Overall, our data show that the LightR approach can be applied to achieve light-mediated regulation of different protein kinases.

Yellow arrows indicate LightR insertion site. Cell lysates were probed for the phosphorylation of indicated proteins. Representative images of mCherry-ERK2 were taken at the indicated time points.

All experiments were done at least three times with similar results. Images were taken every minute and cell was globally stimulated with blue light ten seconds every minute as indicated. To demonstrate the broad applicability of the LightR tool to other types of enzymes, beyond kinases, we engineered LightR-Cre recombinase.

Cre-recombinase has become an essential tool in biomedical research because it allows for genetic recombination and induced activation or deletion of genes Abremski and Hoess, ; Abremski et al. The variant with LightR domain inserted at D residue in Cre showed activation in response to blue light Figure 6F ; Figure 6—figure supplement 2. This demonstrates that the LightR approach can be used broadly for the precise regulation of several types of enzymes.

Our study describes an optogenetic approach that provides several advantages for the interrogation of signaling pathways and demonstrates its broad applicability to address important biological questions.

The key features of this method include: 1 allosteric regulation of the enzymatic activity, 2 tight temporal control of activity with tunable kinetics, 3 local regulation of activity at a subcellular level, and 4 broad applicability to different enzymes. Importantly, unlike other approaches, LightR combines all these advantages in one tool, thus, simplifying application of optogenetics in biological research. We achieved direct regulation of enzymatic activity via an allosteric control by inserting the LightR switch into small loops within the protein structure.

This modular design provides significant flexibility in the selection of the insertion site and allows for specific regulation of catalytic activity without compromising key functions of the protein such as its interactions with binding partners or its native localization in the cell Karginov et al.

Furthermore, variation of the flexible linkers, both between the kinase and VVD as well as between the two VVDs, or changing the insertion site for LightR in the targeted enzyme could result in reversing or altering the regulatory mechanism of LightR, as was previously demonstrated for insertion of LOV2 domain in DHFR Reynolds et al.

While we implemented only one linker type and only one insertion site in LightR-Src, this is a venue for future work to generate more modular tools. Application of LightR approach to different classes of enzymes suggests its broad applicability for the regulation of a wide variety of protein functions in living cells. Thus, unlike other light-regulated allosteric switches Wu et al. We showed that regulation by LightR domain is tunable, enabling different modes of regulation.

LightR switch with slow off kinetics will be useful for long-term activation of LightR-enzymes; since brief periodic pulses of light will be sufficient to maintain activity while avoiding phototoxicity caused by long exposure to blue light.

The FastLightR switch, on the other hand, is more suitable for studies that mimic transient, oscillatory or localized activation of a protein. It could also be used to study the kinetics of negative regulators of signaling pathways immediately after a signaling input is turned off.

Activation of a LightR-enzyme requires low intensity light Figure 2—figure supplement 1B , 0. This level is lower than the intensity used in other optogenetic studies using light-sensitive switches to regulate enzymes Dagliyan et al. The activation of LightR is limited to the blue light spectrum, thus enabling its multiplexing with other red-shifted optogenetic tools or FRET biosensors.

Several optogenetic approaches for the regulation of protein kinases have been described previously; however, all had specific shortcomings. While altered growth phenotypes strongly suggest functionality for the majority of investigated phosphosites, the absence of growth phenotypes does not exclude the possibility of metabolic compensation in response to altered enzyme activity.

Metabolite extracts were obtained from exponentially growing cultures on the same carbon sources used for phenotypic screening. Ions detected in the Dalton range were annotated to metabolites using a genome-wide model of E.

The majority of the 89 phosphomutants analyzed exhibited metabolic changes in at least one of the tested conditions Supplementary Data 3. While phosphomutants exhibited generally similar metabolic responses in both conditions with a median of four changing metabolites, some phosphosite mutations had more substantial consequences when flux through the reaction was higher than on glucose; i.

Assuming that de phosphorylation can also inhibit enzymatic activity, the metabolic response to a phosphosite perturbation and gene knockout should be similar. Local metabolic changes and concordant metabolic signatures between a phosphomutant and a knockout provided evidence of functionality for 38 phosphosites, including 14 that were phenotypically silent Fig. By comparing the phenotypic and metabolic outcomes of abolishing or mimicking phosphorylation, and where available the enzyme deletion, we hypothesized positive or negative consequences of phosphorylation on enzymatic activity for the majority of the investigated sites Fig.

Only 13 phosphosites did not exhibit a metabolic phenotype, eight of which also did not have a growth phenotype. Bright and shaded colors represent glucose and alternative growth conditions, respectively. Negative correlation values are not shown. Albeit not used for ranking, global metabolic changes defined as at least eight metabolites twice the median amount of changing metabolites for all phosphomutants in all conditions are indicated in light green. Additionally, we included Gnd S that displayed opposing growth effects for mimicking and abolishing mutations.

These phosphosite mutations were introduced into the seven genes encoded on ASKA library plasmids with His-tags To exclude protein misfolding as the cause of mutant phenotypes, we performed thermal shift assays with wild-type and phosphomutant enzymes purified from E.

For GpmA and AcnB we observed considerably decreased T m , unexpectedly, also in the abolishing mutants mimicking the unphosphorylated wild-type enzyme, suggesting that any mutation of the investigated residues can affect protein folding and stability. For further validation, we focused on the enzymes with unchanged T m , except SucB that functions only as part of the 2-oxoglutarate dehydrogenase complex.

Nevertheless, we concluded that phosphorylation at S has an inhibitory effect on the SucB reaction based on growth inhibition, metabolic correlation with the knockout, strong accumulation of substrate, no change in T m in the mimicking SucB SE mutant and almost no effects in the abolishing SucB SA mutant Figs. Four independent replicates six for TpiA wt were measured.

For a — d statistical significance was calculated with a two-tailed unpaired t -test and p -values in b , d Benjamini—Hochberg adjusted are indicated. Source data are provided as a Source Data file. To assess whether the physiological consequences of phosphosite mutations were indeed caused by altered enzyme activities, we determined in vitro activities for TpiA, Pta, Gnd and PykF phosphomutants Fig.

Phosphomimetic mutation of TpiA SE resulted in a barely active enzyme, while TpiA TE had milder effects, consistent with the more extreme changes in neighboring metabolite abundances seen in vivo for these mutations Supplementary Data 3.

In accordance with our previous hypothesis, mutating the catalytic residue S of Pta to mimic or abolish phosphorylation reduced enzyme activity significantly, likely by interfering with substrate binding Fig. While phosphoinhibition of Pta S is most likely achieved through interference with the active site 44 , phosphoinhibition of TpiA at S and T is presumably allosteric in nature by hindering the structural dynamics required for catalysis The physiological and metabolic data suggests that phosphorylation of Gnd S and PykF S activates enzyme activity Fig.

Does the above identified phosphoregulation control in vivo pathway usage? During growth on glucose, inhibition of TpiA via phosphomimetic mutation increased the concentration of the methylglyoxal pathway intermediate R -S-lactoylglutathione Fig.

Since phosphorylation at S directly inhibits TpiA activity, likely via limiting the crucial loop-6 movement upon substrate binding 45 , we provide evidence for a flux redirection from glycolysis to the methylglyoxal pathway during glucose catabolism. Inhibition via phosphomimetic mutation at T has a similar effect but since the inhibition of in vitro enzyme activity was much lower Fig. To prove that the phosphomimetic Pta SE is indeed inactive in vivo, the inactive genomic Pta SA mutant was supplemented with either a wild-type or SE Pta expressing plasmid, demonstrating that only the wild-type but not the phosphomimetic Pta enzyme could rescue SA lethality on acetate Fig.

Overall, our results suggest that acetate secretion can be regulated via inhibitory phosphorylation. Data interpretation was straightforward for TpiA and Pta with altered in vitro activities, leaving us with Gnd and PykF.

In both cases, growth rates of phosphoabolishing mutants were reduced Fig. Since the melting temperatures of the mutants were effectively unchanged Fig.

Moreover, we found no difference in PTMs between overexpressed wild-type and mutant enzymes, suggesting absence of compensatory modifications Supplementary Fig. Since Gnd and PykF function in vivo as a homodimer and homotetramer, respectively 46 , 47 , we next checked whether phosphosite mutations affected oligomerization of overexpressed enzymes. While the majority occurred as monomers, the ratio of monomers to multimers was unaltered in the phosphomutants Supplementary Fig.

Thus, the most parsimonious explanation is that phosphorylation allosterically modulates in vivo activity of the multimeric forms. Alternatively, phosphorylation might be relevant for other in vivo properties, such as preventing aggregation during exponential growth as in the case of yeast pyruvate kinase Phenotypic similarity between gnd knockout and the abolishing Gnd SA mutant Fig. Although reaction substrate and product levels were unaltered in knockout and Gnd SA mutant, the phosphomimetic mutant Gnd SE not only grew faster than the wild-type Fig.

As expected, the wild-type catabolized glucose mainly through glycolysis and the PP pathway with lower ED pathway flux 50 Fig. Deletion of gnd blocked utilization of 6-phosphogluconate through the PP pathway and rerouted flux into the ED pathway, as was described before 51 , 52 , Consistent with the hypothesis that phosphorylation at S activates Gnd, fluxes in the phosphoabolishing Gnd SA mutant were similar to the knockout with an active ED pathway and almost no flux through the Gnd reaction, while fluxes in the phosphomimetic Gnd SE were similar to the wild-type Fig.

Thus, even though phosphorylation at non-catalytic S does not change in vitro Gnd activity, it is crucial for in vivo activity and regulates the PP pathway flux.

Flux through the Gnd-catalyzed reaction is highlighted in dark blue. Despite the recent expansion to more than detected phosphosites in E. An additional 38 phosphosites on enzymes at key flux-regulating nodes were reported to affect fitness upon perturbation Since serine and threonine are also subject to other modifications, we cannot exclude that some phenotypic changes following replacement with alanine might at least in part be influenced by other PTMs.

Overall, our results demonstrate that the fraction of functional phosphosites is higher in E. The in vivo occupancy of phosphosites varies with conditions and is typically lower in prokaryotes 5 , with a median occupancy of 6. For AcnB and Pta, inhibition occurs probably through phosphorylation of catalytic residues.

Since Pta is likely to be a structural homolog of Icd whose substrate-binding is inhibited by phosphorylation through the kinase AceK 44 , 61 , it is tempting to speculate that acetate node and TCA cycle flux are both inhibited by AceK to minimize futile cycling in acetate utilization and subsequent use of the glyoxylate shunt, by blocking the Pta and Icd reactions, respectively.

While some phosphosite mutations affect cellular fitness by causing misfolded proteins i. GpmA and AcnB , unaltered thermal stability of five of the seven mutated enzymes suggests their phenotypes to originate from modulated protein activity rather than misfolding. We demonstrated reduced glycolytic flux and increased activity of the methylglyoxal pathway upon allosteric phosphoinhibition of TpiA at S and T, likely by limiting the structural movement required for catalysis. The recent demonstration of increased phosphorylation at S and T in a YeaG kinase knockout indicates that YeaG downregulates another kinase that phosphorylates TpiA For PykF and Gnd, our results suggest more complex phosphoregulation such as promoting protein-protein or protein-metabolite interactions, since phosphorylation was required in vivo but did not affect in vitro enzyme activity.

For Gnd, 13 C-flux analysis demonstrated that phosphorylation at S was required for in vivo PP pathway flux because preventing phosphorylation caused knockout-like fluxes and phenotype. The fact that only phosphorylated Gnd is active in vivo may contribute to previous observations that the PP pathway operates below its maximum in vitro capacity As 36 of the 44 here identified functional phosphosites were never described before, we nearly double the number of phosphosites with known regulatory function in E.

We provide evidence that protein phosphorylation is a major regulation network in bacterial metabolism, capable of controlling metabolic fluxes by various mechanisms, such as shielding the substrate binding site, allosterically limiting structural dynamics, and disrupting interactions relevant for in vivo activity.

Genomic point mutations of E. To abolish phosphorylation, serine and threonine were mutated to alanine, and tyrosine to phenylalanine; to mimic phosphorylation, serine and threonine were substituted with glutamic acid. Cell-oligo mixture was transferred in a prechilled 0. Overall, 3—4 MAGE cycles were performed for each mutation. Cultures at an OD 0.

Each mutant and the wild-type were grown in triplicates and the experiment was repeated on a different day, resulting in six growth curves for each mutant. A linear fit on the log-transformed growth curve was used to determine the growth rate. Metabolite extracts were analyzed by direct flow double injection on an Agilent series iFunnel quadrupole time-of-flight mass spectrometer Agilent, Santa Clara, CA, U.

Mass spectrometry data processing and analysis were performed in Matlab The Mathworks, Natick. Deprotonated ions were annotated based on mass using 0.

For every ion, the abundances from all replicates of a given mutant were pooled and compared to the pooled abundances of the wild-type sample. The log 2 fold change of an ion abundance in the mutant compared to the wild-type was determined, and a two-sided 2-sample t -test with unequal variance was performed. The obtained p -values were corrected for multiple testing using the Benjamini—Hochberg procedure.

Metabolites that are not included in the E. The phosphomutant plasmids were introduced into chemically competent E. Gel densitometry analysis was performed in ImageJ 1.

Protein T m was determined as a maximum of the melt curve derivative. For each enzyme, melt curves were recorded in six replicates. The initial reaction rates of mutants were divided by the wild-type values to obtain relative enzymatic activities.

Protein was resolved across a linear gradient of buffer B Full MS data was acquired to examine sample purity and intact protein mass. To increase sequence coverage, each enzyme was subjected to higher energy collisional dissociation with normalized collisional energy values ranging from Summed full scan spectra were deconvoluted to neutral masses using UniDec 4.

MS2 data was deconvoluted using the Xtract algorithm as part of FreeStyle 1.