We present a systems biology view on pseudoenzymes that acknowledges that genes are not selfish: the genome is. With network function as the selectable unit, there has been an evolutionary bonus for recombination of functions of and within proteins. Many proteins house a functionality by which they ‘read’ the cell's state, and one by which they ‘write’ and thereby change that state. Should the writer domain lose its cognate function, a ‘pseudoenzyme’ or ‘pseudosignaler’ arises. GlnK involved in Escherichia coli ammonia assimilation may well be a pseudosignaler, associating ‘reading’ the nitrogen state of the cell to ‘writing’ the ammonium uptake activity. We identify functional pseudosignalers in the cyclin-dependent kinase complexes regulating cell-cycle progression. For the mitogen-activated protein kinase pathway, we illustrate how a ‘dead’ pseudosignaler could produce potentially selectable functionalities. Four billion years ago, bioenergetics may have shuffled ‘electron-writers’, producing various networks that all served the same function of anaerobic ATP synthesis and carbon assimilation from hydrogen and carbon dioxide, but at different ATP/acetate ratios. This would have enabled organisms to deal with variable challenges of energy need and substrate supply. The same principle might enable ‘gear-shifting’ in real time, by dynamically generating different pseudo-redox enzymes, reshuffling their coenzymes, and rerouting network fluxes. Non-stationary pH gradients in thermal vents together with similar such shuffling mechanisms may have produced a first selectable proton-motivated pyrophosphate synthase and subsequent ATP synthase. A combination of functionalities into enzymes, signalers, and the pseudo-versions thereof may offer fitness in terms of plasticity, both in real time and in evolution.

Introduction

The accumulation of genome sequence data has led to many surprises. One of these has been the large number of enzyme homologs with imperfect homology at catalytically important residues [1]. In some of these cases, the catalytic activity is indeed absent. These apparent evolutionary remnants of active enzymes are called pseudoenzymes. Most active enzyme families contain such ‘zombielike’ relatives [24]. Pseudoenzymes are often well conserved and expressed which suggests that they are functional even if catalytically inactive.

Ongoing evolutionary selection also diverges protein functions through generating paralogs: some of these may be pseudoenzymes. Examples include proteases, kinases, phosphatases, E2 ubiquitin-conjugating enzymes, and phospholipases. This raises the question why pseudoenzymes are so widespread. Are these ‘zombielike’ enzymes just evolutionary remains? Are they occasional inventions of rare functionalities? Or are they a tip of the iceberg of protein action in a world that selects for network function? Our aim in this review was to examine the third possibility. And we find that a diversity of functionality owes to a plasticity in protein function that also generates pseudoenzymes.

From pseudoenzymes to pseudosignalers

Pseudoenzymes and their definition

Possibly the first case of an inactive cognate enzyme was documented in 1967: not known as an enzyme, α-lactalbumin exhibited a close structural similarity to chicken lysozyme [5]. The authors argued that the similarity between bovine α-lactalbumin and chicken lysozyme had been caused by divergent evolution. Both proteins recognize the β-1,4-sugar linkage, albeit with dissimilar outcomes.

The definition of pseudoenzymes only arose in times where amino acid sequences or structures had become identifiable and when it was recognized that there was substantial sequence and structural similarity between proteins from different organisms. It surfaced that evolutionary divergence is small enough for common origins of proteins to be inferred from their sequences or structures. The reason for this limited divergence was selection pressure on function, i.e. alterations of amino acid sequences that interfered with fitness had been selected against. This led to the expectation that homologous proteins should have the same function. Because organisms are complex, it has been difficult to verify that this expectation is correct in all, or in just a few cases. The observations that genome-wide metabolic network reconstructions make sense biochemically [6] provide support for the expectation. Before this, however, there had been many falsifications of the same expectation. Perhaps, the most convincing one has been the discovery of gene sequences with substantial similarity and likely homology, but with mutations interfering with the molecular function of the corresponding protein. In the clearest cases, these were mutations that did away with an amino acid residue understood to be essential in the enzyme's mechanism of catalysis. In some of these cases, the catalytic inactivity was confirmed experimentally, suggesting that the protein was not an enzyme but a ‘pseudoenzyme’. Lack of confirmation is inconclusive, however, as such catalytic activity may require some regulation that has evolved as a physiological function of the particular organism. Accordingly, pseudoenzymes are more often identified by bioinformatics, i.e. as proteins that are homologous to enzymes, lack amino acid sequence considered essential for the cognate catalytic activity, and are conserved evolutionarily [79]. Biological understanding of many pseudoenzymes is hereby not complete, however, as biological understanding requires elucidation of their alternative molecular functions that support the host organisms' fitness.

Protein phosphorylation conducts information. Upon receipt of a molecular signal, protein kinases transfer phosphoryl groups from ATP to specific amino acid residues like serine, threonine, tyrosine or histidine in proteins, which is then the onset of a signaling process. This protein phosphorylation is also a catalytic activity and hence protein kinases also classify as enzymes, even though their substrates are other proteins rather than metabolites. Canonical amino acids are central to the catalytic activity of protein kinases, their mutation producing inactive cognates of kinases referred to as ‘pseudokinases’. Fifty human protein kinase domains lack canonical catalytic residues such as Lys30, Asp125, and Asp143, suggesting the corresponding protein domains to be inactive [10]. Eyers and Murphy [9] defined active protein kinase domains by using sequence analysis and suggested that pseudokinases, identified by critically incomplete homologies, may represent potential drug targets. Hammaren et al. [11] reinforced this proposition and extended the class of pseudo-protein kinases to homologs that fail to bind the ATP in the first place. Loss of recognized function at incompletely sustained homology would seem to characterize the pseudoenzymes.

Pseudosignalers

Although protein kinases such as cyclic AMP (cAMP)-dependent protein kinase A (PKA) and metabolic kinases such as pyruvate kinase share catalytic identity, they have different functionalities with respect to cell physiology. The latter is a catalyst of a chemical reaction in a metabolic pathway. The former translates a cell-state signal into a signal of a different identity, in our example, a signal in the form of the concentration of cAMP into a signal in the form of the phosphorylation state of a protein. Also because the flux through metabolic enzymes is much higher than the steady-state flux through protein kinases, the latter should perhaps be called signal transducers rather than enzymes, more in line with their function. Pseudo-protein kinases should then be called pseudo signal-transducers or pseudosignalers.

In other examples of signalers, the catalytic activity is lacking in the first place. PII, a protein in the network that regulates ammonia assimilation by E. coli [12], does not have a catalytic function. It regulates whether the ambiguous enzyme [13] ATase (adenylyl transferase) adenylates GS (glutamine synthetase), or de-adenylylates GS-AMP by mere binding. Because regulatory functions of proteins in the absence of catalytic functions are less understood or perhaps less defined mechanistically, it may be difficult to recognize such pseudosignalers as paralogs contributing to fitness but with mutations that eliminate the molecular function of their ortholog.

Networks and pseudoenzymes

Pseudosignaler GlnK regulating ammonium transport?

By way of a first example of a pseudosignaler paralogous of a signaler without catalytic activity, we wish to suggest that the protein GlnK may well be an inactive cognate of the protein GlnB (PII) in the intracellular signal transduction process regulating nitrogen assimilation in E. coli [12]. At low extracellular ammonium concentrations, GS plays a role in the nitrogen assimilation network. Its activity depends on its covalent modification by up to 12 AMP groups. GS adenylylation (GS-(AMP)n + ATP → GS-(AMP)n+1 + PPi) and GS-AMP de-adenylylation (GS-(AMP)n+1 + Pi → GS-(AMP)n + ADP) are catalyzed by the same ambiguous [13] enzyme adenylyl transferase (ATase). The PII protein persuades the ATase (‘PI’) which of the two reactions it should catalyze. When PII is in the uridylylated state, it promotes de-adenylylation; native PII promotes adenylylation. The PII component is also referred to as GlnB due to being encoded by the glnB gene [14]. PII also regulates the phosphorylation state of the protein Nitrogen Regulator (NR)I, which is the transcription factor of the Gln operon [5]. Structurally, the GlnB protein is a homotrimer, which contains three large 20-residue loops (T-loops). The loops are important structures for the interaction of GlnB with ATase, uridylytransferase (UTase), and the protein kinase NRII, through which GlnB regulates NRI. The protein GlnK has emerged as a paralog of PII [15]: glnk is present in the same organisms (e.g. E. coli) as is glnB, also encodes 112 amino acids two-thirds of which are identical with those of PII. GlnK is also homotrimer, and its three-dimensional structure is similar to that of GlnB, including the conformation of three large T-loops, as analyzed by X-ray crystallography [16]. Although glnK is homologous to glnB, PII and GlnK are different functionally. For instance, PII and PII-UMP are more effective than GlnK and GlnK-UMP in activating the adenylylation and de-adenylylation reactions, respectively, and PII is more effective than GlnK in stimulating the phosphatase activity of NRII. Although we initially proposed that GlnK would substitute for PII under special conditions [15], we would here entertain the possibility that GlnK is a pseudosignaler version of the signaler PII. It has retained the ability to be regulated by the nitrogen status of the cell in the same way as PII is regulated, but has lost most of its ability to regulate ATase. Instead, it has gained the possibility to inhibit the AmtB ammonium ion transporter [17] in order to prevent futile ammonium cycling across the cell membrane [5].

Enzymes are combinations of ‘readers’ and ‘writers’ according to systems biology

The smallest genome of living organisms exceeds 300 essential genes [18]. Genome-wide metabolic networks have been mapped and do not just consist of a large number of linear pathways, but are highly branched [19]. The branched nature of these metabolic networks ensures not only flux robustness through redundancy [20], but also multiple alternative substrates for growth, and multiple metabolic exits for surviving environmental limitations such as anaerobiosis and low Gibbs energies in substrates [21]. Metabolic functions alone are not only provided by individual enzymes but also by networks of enzymes [22].

On top of this already complex tier of the metabolic network, there are layers of regulation. Also these are evolutionarily conserved, suggesting that they are essential. The fact that the human genome encodes more regulator proteins than metabolic enzymes [23] further supports our tenet that, in evolution, it is the genome rather than the gene that is selfish [24]: it is the networks more than the individual genes that matter. This linking of metabolic biochemistry to the other tiers of the cell's chemistry that include the layers of signal transduction and the layers of gene expression has been part and parcel of hierarchical regulation and control analyses [25]. The existence of pseudosignalers parallels conceptual integration of metabolism with signaling.

What is the consequence of this contention that networks are more important for fitness than individual molecules, for the understanding of pseudoenzymes? In the classical paradigm of one gene–one protein–one function, a pseudoenzyme cannot contribute to fitness. Indeed, in this early paradigm, one could only note that a pseudoenzyme is a ‘dead’ enzyme. It was then incomprehensible, however, how the dead enzyme would persist in evolution. In the systems biology paradigm, an enzyme functions as part of a network and must therefore have properties in addition to being a mere catalyst such as platinum. For one, the enzyme must not only bind and convert its substrate, it must also do this at a certain elasticity [26], such that the rate at which the enzyme contributes to network function depends on the concentration of that substrate. Similarly, that rate should depend on the fitness requirements of the biological network. These requirements of the cellular/organismal metabolic network are exhibited by signals such as the ATP/ADP ratio, cAMP concentration, redox state of NAD(H), level of snRNAs, concentration of other proteins, availability of membranes or DNA, and concentration of the enzyme's substrate. Enzymes read the status of some of these signals directly by binding them. They are home therefore not only to active sites for catalysis, but also to ‘active sites’ for reading aspects of the cell's state. The molecular networks in living organisms thereby essentially consist of proteins that ‘read’ aspects of the cell state and ‘write’ to the cell networks, thereby changing the cell state. The enzymes act only after interpreting what they are reading, the interpretation task explaining part of the complexity of proteins. Enzymes both interact and act; the writing consists of the rate of a chemical reaction. For signalers, the writing consists of managing the state of a cellular component that can be read by other proteins including RNA polymerases and protein kinases.

For the metabolic enzyme pyruvate kinase, the actor/writer part is the part of the enzyme that helps convert phosphoenol-pyruvate (PEP) into pyruvate under concomitant phosphorylation of ADP. The interactor/reader parts are the binding sites of PEP/pyruvate, ADP/ATP, and the regulator fructose 1,6 bis-phosphates. For PKA, the writer part is the catalytic site carrying out the transfer of the phosphoryl group from ATP to its protein substrate, whilst the reader sites are the binding site for that protein substrate, for ATP/ADP, and for its regulator cAMP.

The principle we propose here is one that is not limited to protein kinases: the original or new writer facility may have any catalytic or signaling activity. And the reader facility is not limited to the substrate of the reaction catalyzed by the writer [27]: the reader may read any signal that regulates the catalytic or signaling activity of the writer, including reversible binding of an allosteric modifier, as well as phosphorylation by a protein kinase and binding of the reader subunit to another protein when that is phosphorylated [such as the binding of Grb2 to epidermal growth factor receptor (EGFR), the writing here corresponding to the binding of Grb2 to SOS [28]].

In the two-component regulatory systems of bacteria, the writing itself involves a pathway of signalers/enzymes. The uridylyltransferase, for instance, reads the nitrogen status of the cell and adjusts the uridylylation state of PII accordingly, and PII (-UMP) then testifies this nitrogen status to the histidine kinase NRII, which phosphorylates the transcription factor NRI to a concentration and hence transcription regulation activity that depends on the extent of uridylylation of PII.

Pseudoenzymes: reader–writer permutations?

From this network point of view, the term ‘pseudoenzyme’ obtains an immediate perspective: the pseudoenzyme has lost its original writer's ability but it may still be able to read the original signal. [Alternatively, it still has the original writer's ability, which could be signaling (see below) but has not been identified yet, whereas the writer's ability of a more recent enzyme with the same reader ability has been identified; see below: the meaning of the words ‘original’ and ‘new’ should be tentative here.] Figure 1 illustrates this novel perspective (see also ref. [27]). Figure 1A,B still have the enzyme protein kinase A and its pseudokinase as enzyme units, whilst Figure 1C already decomposes the concept ‘enzyme’ into a ‘reader’ facility (of cAMP concentration) and a ‘writer’ (catalyzing the conversion of E into E-phosphate) facility. Whilst Figure 1B suggests the complete absence of the single, enzyme-as-such, functionality, Figure 1D now suggests that the pseudoenzyme retains the reading functionality. The pseudoenzyme may still be active in the network through its reading activity. In such a case, a non-catalytic signaling may still allow for the writing function that is coupled to it (Figure 1E).

Coupling of reading activity of a protein kinase to the writing activity of a protein kinase, as well as decoupling and recoupling to a signal writer.

Figure 1.
Coupling of reading activity of a protein kinase to the writing activity of a protein kinase, as well as decoupling and recoupling to a signal writer.

(A) Illustrating the phosphorylation of a protein E by a protein kinase A that reads the cAMP concentration in the cell (denoted by ‘A’) (PKA; the black blob) and (B) the corresponding dead pseudo-protein kinase. (C) Recognizing that the PKA contains two active domains, i.e. a ‘reader’ domain examining the cell status as indicated by the cAMP concentration (‘A’) and a ‘writer’ domain phosphorylating the protein E and (D) the corresponding dead pseudokinase with inactivated writing capability but retained reading capability. (E) Illustrating how by genetic re-association (through partial gene duplication of the reader domain) a signaling protein driven by the cAMP state of the cell [A] may result. The role of ‘A’ may also be taken by the protein substrate of the kinase [E], which is also ‘read’ by the PKA.

Figure 1.
Coupling of reading activity of a protein kinase to the writing activity of a protein kinase, as well as decoupling and recoupling to a signal writer.

(A) Illustrating the phosphorylation of a protein E by a protein kinase A that reads the cAMP concentration in the cell (denoted by ‘A’) (PKA; the black blob) and (B) the corresponding dead pseudo-protein kinase. (C) Recognizing that the PKA contains two active domains, i.e. a ‘reader’ domain examining the cell status as indicated by the cAMP concentration (‘A’) and a ‘writer’ domain phosphorylating the protein E and (D) the corresponding dead pseudokinase with inactivated writing capability but retained reading capability. (E) Illustrating how by genetic re-association (through partial gene duplication of the reader domain) a signaling protein driven by the cAMP state of the cell [A] may result. The role of ‘A’ may also be taken by the protein substrate of the kinase [E], which is also ‘read’ by the PKA.

How does this principle help in understanding contributions to fitness of ‘dead enzymes’? [27] Progress toward understanding the functions of pseudoenzymes that are declared ‘dead’ is enlivening. We here suggest that the first answer to the above question is that pseudoenzymes are not dead, but only appear to be so: the reader domain has become associated with a different catalytic writer domain the catalytic activity of which has not been recognized. The second answer we propose is that the function of the new writer domain is not catalysis but signaling and hence not recognized as catalytic.

Multiplexing by pseudoenzymes, signal integration, and scaffolding

The reader part of the protein may consist of multiple reader domains. One of these might enable the binding of the substrate for catalysis, a second might enable binding of an allosteric modifier such as AMP, a third might enable association with a copy of the same protein, i.e. dimerization, and a fourth the binding of a different protein. In this example, the reader part of the protein would measure the state of the cell in terms of four concentrations. It might activate the writer only if all four of these were high, or if the first two were low and the third and fourth concentrations were high, or in some other combination defined by the structure of the protein. Hereby, the protein would engage in signal integration.

The ErbB (HER) proteins constitute examples. One of these is the EGFR (EGFR = HER1 = ErbB1). In our reader–writer terminology, the EGFR would have a dual reader in its extracellular part, i.e. one for EGF and a second for a second molecule of EGFR that has EGF bound. The writer domain of EGFR is a protein kinase residing in its intracellular part. In this case, the mechanism through which the reader activates the writer in the proteins is clear: activation of the reader domain causes dimerization of EGFR [28]. The writing then consists of transphosphorylation of the two monomers by each other [28]. The protein Grb2 has a reader of the phosphorylation state of EGFR and a writer that binds SOS hence increases the SOS concentrations close to the membrane and hence close to Ras, which has an SOS reader [42]. In this case, a pseudoenzyme could be a homolog of EGFR that still has the reader-dimerization domain, but of which the writer domain is defunct, lacking the transkinase activity. This pseudoenzyme would function as a negative regulator of signal transduction or as a limiter of the extent of signaling. ErbB3 appears to be such a pseudokinase [27,29]. Because of its dimerization potential, it should bind to EGFR when both EGF and neuregulin-1 (the ligand to ErbB3) [29] are around, thereby possibly tempering EGF-induced signal transduction when neuregulin-1 is present.

Their paralog ErbB2 is what one might call an ‘inverse-pseudo kinase’ in that it is ‘pseudo’ at the reader site: it lacks the ligand reading activity of EGFR but does harbor the kinase activity. Conceivably, this leads to integration of the signal input into ErbB3 and ErbB2, the activation of the former causing heterodimerization of ErbB3 with ErbB2, the ErbB2 kinase then phosphorylating its ErbB3 partner [30]. ErbB2 does not bind extracellular ligand, but its level is regulated by translation regulation such as induced by α6β4 integrin [30].

Multiplicity of reading can bring an even larger number of proteins together. A case in point may be the KSR1 protein, in part homologous to Raf and [31] involved in heterodimerization with Rafs. A lysine residue in its catalytic site is lacking, so it is suspected to lack its cognate kinase activity. The protein appears to have readers of the concentration of other relevant proteins as well, including Raf, MEK, and ERK. This property enables it to bind some of these proteins simultaneously or alternatively [32]. This is proposed to lead to constitutive binding to ERK and MEK, and conditional (depending on Raf activation by Ras) binding to Raf. The consequence would be a signal-induced meta-scaffolding bringing the ‘bottom’ of the mitogen-activated protein (MAP) kinase cascade in close contact with its top (plasma membrane-bound Ras and associated Raf). The classification of KSR1 as a pseudokinase is not obvious, however, as it appears to have a kinase activity perhaps even versus RAF (understandable as heterodimeric transphosphorylation) and MEK (understandable as a remnant of its original RAF-like kinase activity) [32]. KSR1 may correspond not to a mutilated but to a modulated kinase with additional scaffolding function as a consequence of its multiple reader activity.

Further down the pathway, there is yet another role for a pseudoenzyme, this time a pseudophosphatase (STYX), as shown by in silico discovery with experimental validation [33]. The pseudophosphatase has retained an affinity for the substrate of its non-pseudo-counterpart: it reads the nuclear ERK concentration. Because STYX localizes to the nucleus (as was shown experimentally in ref. [33]), it works as nuclear anchor to the ERK, thereby reducing the cytosolic activity of the latter [33]. Here, the one reader unit of STYX may note the presence of a physical substrate in the nucleus, the second reader unit recognizing ERK. The protein thereby localizes ERK to the nucleus [33].

The multiplexing opens many more regulatory possibilities. For instance, if one of the signals that are read by the reader is the presence of a gene, the copy number of the signal molecule (i.e. the gene) is limited. Here, the presence of a pseudoenzyme with a reader that binds the same gene but has no writer may compete with the binding of a transcription factor that has the same reading domain but an active writer domain such as a histone acetylase or an RNA polymerase. The multiplex signals read by the pseudoenzyme might here regulate histone modification or transcription, providing the pseudokinase, which is dead in terms of its writer activity, with a selectable function [30]. That also the writer domains may be multiplex will be discussed below.

With all these examples, one might surmise that all enzymes should be recombinations of readers and writers. We would agree that they may not be: Below in the section ‘Multiplex Writer Domains’), we shall give examples of enzymes that are combinations of writers, where pseudoenzymes are recombinations of writers with writers. Furthermore, more and different types of pseudoenzymes [27] remain to be discovered.

Pseudoenzymes: what's in a name?

In the present paper, we use the word ‘pseudoenzyme’ and introduce the word ‘pseudosignaler’, even though both of these may be misnomers. We use the former to refer to an enzyme with reader and writer domains where the writer domain has been replaced by a different writer domain; this is with reference to a purportedly original paralogous enzyme. In actual practice, the function of the writer domain may or may not yet have been recognized. In case that function has been recognized and it is catalysis of some sort, then the name ‘pseudoenzyme’ may not be appropriate: it is just an enzyme with a different combination of reader and writer domains when compared with the supposed original. In fact, the enzyme proposed to be the pseudoenzyme may have been the progenitor of what one considers the ‘original’ enzyme. Similarly, if the writer function has been identified as signaling, the protein should perhaps be called a signaler and neither a pseudoenzyme nor a pseudosignaler. Should the writer function not have been identified, then it may still not be appropriate to call the protein a pseudoenzyme as it might still turn out to be a true enzyme or a true signaler. As much as it did not matter to Juliet that Romeo was called Montague [34], we here use the words pseudoenzyme and pseudosignaler merely reflect evolutionary family ties as well as continuity with the existing literature [1,7,9].

Challenging the cell cycle: dynamic recoupling of readers and writers

We will next sketch an even further diversity of possibilities offered by our reader–writer concept. We do this by discussing three examples in detail, one describing known regulation of the cell cycle in terms of ‘readers’ and ‘writers’, a second explicating a proposed case of a viral pseudoenzyme [35] by emphasizing network interactions, and a third in which we model curious functional effects pseudoenzymes might have on signal transduction by the MAP kinase (MAPK) pathway.

A most inspiring example of the reader–writer concept is that of the cyclin-dependent kinases (CDKs). Some of these are known to regulate the cell cycle, others manage transcription, epigenetic regulation, or metabolism [36,37]. CDKs ‘write’ by phosphorylating target proteins that then co-ordinate many cell activities, such as ones that are necessary for a particular phase of the cell cycle, such as the S (DNA synthesis) phase, to proceed. CDKs require the binding of a reader unit to become active. The reader units are called cyclins. The cyclins not only activate CDK but also determine which of the possible protein substrates it will phosphorylate: they determine the specificity of the protein kinase. Possibly, the fair diversity of CDKs involved in the mammalian cell cycle originates in a single such yeast reader [38], which is directed to particular moments in the cell cycle by the four most abundant cyclin pairs [39,40].

In our view, each of these cyclins is a ‘reader’ of the cell state. Each reads whether the corresponding next phase of the cell cycle is appropriate for the cell to engage in. An additional cell-state aspect to be assessed is whether the cell has witnessed the activity of the cyclin that drives the preceding phase of the cell cycle. This aspect may exist in an up-regulation of the concentration of the next cyclin by the previous cyclin. Similarly, a cyclin-reader for a phase in the cell cycle should read whether the cyclin for the next phase is already present. If so, then this should be one reason for the earlier cyclin to stop activating the CDK into its specificity: this is achieved by the later cyclin suppressing the concentration of the earlier cyclin [41].

Also, the cyclins may have a single evolutionary origin [42]. Deletion of each cyclin alone or in various combinations still allows the cell cycle to proceed, albeit at a slower speed and somewhat irregularly [43,44]. In a case where all but one cyclin was deleted, the remaining cyclin continued to increase in expression level with time, as it was no longer suppressed by the subsequent cyclins (see the previous paragraph) and the onset of subsequent phases of the cell cycle seemed to be determined by the concentration of the remaining cyclin exceeding a threshold level that was higher for each next phase [45]. This one reader–one writer system may have been the original, single, cell-cycle clock. The coupling of three clocks into a super clock, as proposed by Barberis and colleagues [46], may be a more recent evolutionary invention. It may have resulted from gene duplication of the one cyclin with subsequent specialization making the system somewhat, but not essentially more robust.

In the case of the cell cycle, reader domains and writer domains operate independently in real time: as the cell cycle is running, the central CDK first associates with one, then with a second, and then with the subsequent cyclins. This requires a dynamic association between CDKs and cyclins at a time scale of minutes if not seconds, as well as changes in the concentrations (or activities) of the various cyclins due to variation in gene expression or regulated proteolysis. This results in apparent switches between cell-cycle phases during cell cycling.

Reader and writer domains may also be stuck together for much longer, i.e. for more than the life time of the integral protein or even longer than the life time of its host individual organism. In this case, genetic recombination between readers and writers results in new genes, the mutations becoming fixed by selection for fitness. We shall call this ‘static recombination’ or ‘static shuffling’. Below, these static and dynamic modes of shuffling will contribute to new insights into bioenergetics.

Networks revitalizing ‘dead’ pseudoenzymes for parasites

Viruses often develop complicated mechanisms to divert host immune responses. This may include the use of a pseudoenzyme strategy. The example we wish to cite here [35], and where our explication owes much to clarifications by Kolakofsky and Garcin [47], is that of the ORF75 protein of gammaherpesvirus 68, which is homologous with phosphoribosyl-formylglycinamidine synthetase (PFAS), a cellular glutamine phosphoribosyl-formylglycine amidotransferase. It is energized by ATP hydrolysis. This cGAT resides in the purine synthesis pathway and is active as an oligomer. The ORF75 protein lacks the catalytic activity of PFAS and was therefore baptized a pseudoenzyme, i.e. vGAT [35]. RIG-1 is a cytosolic PRR (pattern recognition receptor) of the innate immune system, composed of (i) caspase recruitment domains (CARDs) critical for signal transmission toward innate host defense against pathogens, (ii) a central helicase domain that binds dsRNA, and (iii) a C-terminal domain of CARD that binds 5′ppp-dsRNA. Active RIG-1 contains a glutamine residue (Q10) in its CARD domain and asparagines (N245 and N445) within its helicase 1 domain. It was a bit of an enigma how pseudoenzyme vGAT could interfere with RIG-1 function. The enigma is resolved by acknowledging that as a potential oligomer, cGAT has a reader of the status of other cGAT molecules, which it may use before writing, i.e. before engaging in transamidation with glutamine (Q) as an amide source. Though lacking the transamidation activity, vGAT may still avail of this reading activity of dimerizing with cGAT. By gene-module shuffling, vGAT may have acquired an additional reader domain that enables it to read the presence of RIG-1 by enabling it to bind, either directly or indirectly through the dsRNA. The consequence is that the catalytic activity of cGAT is brought into the vicinity of RIG-1. This activity then senses a high local concentration of glutamines (Q) and asparagines (N), which although part of the protein backbone of RIG-1 may still be de-amidated to glutamate and aspartate residues, respectively. This process leads to inactivation of RIG-1, dissolution of the antiviral innate immune response, and a positive contribution to fitness not of the host but of the parasite.

Remarkable potential network effects of ‘dead’ pseudoenzymes

We shall now examine whether pseudoenzyme homologs of the MAPK pathway, solely by being catalytically inactive, could have unexpected effects on the dynamic performance of the pathway. The MAPK regulates cell proliferation, cell differentiation, and cell apoptosis. Dysregulated MAPK signaling has been implicated in a wide range of cancers and occurs by multiple mechanisms including inappropriate activation of receptors and activating K-Ras and B-Raf mutations. The variation with time of the phosphorylation state of the bottom kinase of the pathway (i.e. ERKPP) is important for the proliferative versus differentiation effect of activation of the pathway [48,49]. The MAPK pathway has been a frequent subject for systems biology investigations. The achievements of such studies have been to design kinetic models that are operational in silico [50,51] and verifiable in vitro or in vivo, to enable the discovery of principles underpinning cell signaling [52,53], and to support attempts at network-based drug design vis-à-vis cancer [54].

The MAPK pathway begins with the binding of ligands such as EGF to the extracellular portion of a so-called tyrosine kinase receptor. Signaling molecules like Grb2 and SOS are then recruited to the internal docking site of the receptor, proximity to the membrane resulting in the activation of the membrane-bound G-protein Ras [55]. Each MAPK pathway is then mediated by a three-tiered kinase cascade, consisting of an MAPK kinase kinase (MAPKKK, e.g. RAF), an MAPK kinase (MAPKK, e.g. MEK), and an MAPK (e.g. ERK) [56]: Ras-GTP triggers a phosphorylation cascade that involves RAF, MEK, and ERK proteins, leading to ERK activation through phosphorylation and translocation to the nucleus. Phosphorylated ERK activates several transcription factors, again by phosphorylation.

We shall here model the MAPK pathway in silico so as to search for effects a pseudokinase might have on the dynamics of signal transduction. Could such a pseudokinase defect reduce or disintegrate cellular signaling, or would it offer an opportunity of having a more complex and useful signal? To address these questions, we formulated 11 kinetic rate equations, 7 of which corresponded to one of our earlier [57] kinetic models for the pathway (Table 1). The remaining four equations enable a pseudokinase version of the MAPKKK that is supposed to be present in addition to the kinase of MAPKK, to be phosphorylated upon activation of regulator R (Ras), and to then bind to, but not phosphorylate MAPKK (Figure 2). We also examined various alternative models by adjusting parameter values (Table 2).

Scheme of the MAPK pathway with additional pseudoenzyme (pseudo_E) paralogue of MAPKKK in SBGN.

Figure 2.
Scheme of the MAPK pathway with additional pseudoenzyme (pseudo_E) paralogue of MAPKKK in SBGN.

In various versions of the model, the pseudokinase interacts with MAPKK (or MAPKKP for the ‘converse pseudokinase’ case), preventing their activities (i.e. their abilities to phosphorylate, be phosphorylated, or be dephosphorylated, as specified in Tables 1 and 2).

Figure 2.
Scheme of the MAPK pathway with additional pseudoenzyme (pseudo_E) paralogue of MAPKKK in SBGN.

In various versions of the model, the pseudokinase interacts with MAPKK (or MAPKKP for the ‘converse pseudokinase’ case), preventing their activities (i.e. their abilities to phosphorylate, be phosphorylated, or be dephosphorylated, as specified in Tables 1 and 2).

Table 1
Reaction kinetic equations for the model (Figure 2) of the MAPK pathway with additional pseudo-MAPKKK

Free pseudo-MAPKKK was supposed to have retained MAPKKK's abilities to be phosphorylated by activator R, to dephosphorylate spontaneously, and, if itself phosphorylated, to bind reversibly to MAPKK. It is no longer able to phosphorylate the latter, however. In the first column, each process considered has been given a number which follows the symbol ‘Re’ for reaction in Figure 2. A reaction number is given in the second column that corresponds to the rate or equilibrium equation for the process that is named in the third column. The fourth column describes the process by a chemical reaction equation; where substances occur behind a semicolon, they are mere influences on the process without themselves being consumed.

SBGN map number Reaction number Name Reaction equation Rate equation 
re 12 R1 MAPKKK phosphorylation MAPKKK → MAPKKKP; R v1 = Vm1 * R * MAPKKK/(Km1 + MAPKKK) 
re 13 R2 MAPKKKP dephosphorylation MAPKKKP → MAPKKK v2 = Vm2 * MAPKKKP/(Km2 + MAPKKKP) 
re 14 R3 MAPKK phosphorylation MAPKK → MAPKKP; MAPKKKP v3 = Vm3 * MAPKKKP * MAPKK/(Km3 + MAPKK) 
re 15 R4 MAPKKP dephosphorylation MAPKKP → MAPKK v4 = Vm4 * MAPKKP/(Km4 + MAPKKP) 
re 16 R5 MAPK phosphorylation MAPK → MAPKP; MAPKKP v5 = Vm5 * MAPKP * MAPK/(Km5 + MAPK) 
re 17 R6 MAPKP dephosphorylation MAPKP → MAPK v6 = Vm6 * MAPKP/(Km6 + MAPKP) 
re 10 R7 Activator degradation R → R0 v7 = k7 * R 
re 4 R8 Pseudo_E phosphorylation Pseudo_E → Pseudo_E_P; R v8 = Vm8 * R * Pseudo_E/(Km8 + Pseudo_E) 
re 20 R9 Pseudo_E_P dephosphorylation Pseudo_E_P → Pseudo_E v9 = Vm9 * Pseudo_E_P/(Km9 + Pseudo_E_P) 
re 6 and re 7 R10 Pseudo_E_P binding to MAPKK Pseudo_E_P + MAPKK = Pseudo_E_P _MAPKK v10 = k10 * Pseudo_E_P * MAPKK −  k−10 * Pseudo_E_P_MAPKK 
re 8 and re 9 R11 Pseudo_E_P binding to MAPKKP Pseudo_E_P + MAPKKP = Pseudo_E_P _MAPKKP v11 = k11 * Pseudo_E_P * MAPKKP − k−11 * Pseudo_E_P_MAPKKP 
SBGN map number Reaction number Name Reaction equation Rate equation 
re 12 R1 MAPKKK phosphorylation MAPKKK → MAPKKKP; R v1 = Vm1 * R * MAPKKK/(Km1 + MAPKKK) 
re 13 R2 MAPKKKP dephosphorylation MAPKKKP → MAPKKK v2 = Vm2 * MAPKKKP/(Km2 + MAPKKKP) 
re 14 R3 MAPKK phosphorylation MAPKK → MAPKKP; MAPKKKP v3 = Vm3 * MAPKKKP * MAPKK/(Km3 + MAPKK) 
re 15 R4 MAPKKP dephosphorylation MAPKKP → MAPKK v4 = Vm4 * MAPKKP/(Km4 + MAPKKP) 
re 16 R5 MAPK phosphorylation MAPK → MAPKP; MAPKKP v5 = Vm5 * MAPKP * MAPK/(Km5 + MAPK) 
re 17 R6 MAPKP dephosphorylation MAPKP → MAPK v6 = Vm6 * MAPKP/(Km6 + MAPKP) 
re 10 R7 Activator degradation R → R0 v7 = k7 * R 
re 4 R8 Pseudo_E phosphorylation Pseudo_E → Pseudo_E_P; R v8 = Vm8 * R * Pseudo_E/(Km8 + Pseudo_E) 
re 20 R9 Pseudo_E_P dephosphorylation Pseudo_E_P → Pseudo_E v9 = Vm9 * Pseudo_E_P/(Km9 + Pseudo_E_P) 
re 6 and re 7 R10 Pseudo_E_P binding to MAPKK Pseudo_E_P + MAPKK = Pseudo_E_P _MAPKK v10 = k10 * Pseudo_E_P * MAPKK −  k−10 * Pseudo_E_P_MAPKK 
re 8 and re 9 R11 Pseudo_E_P binding to MAPKKP Pseudo_E_P + MAPKKP = Pseudo_E_P _MAPKKP v11 = k11 * Pseudo_E_P * MAPKKP − k−11 * Pseudo_E_P_MAPKKP 

Abbreviations: =, reversible reaction; →, irreversible reaction; X → Y, R reaction converting X into Y with regulator R. In the final column, the rate equation used for the process is shown with the rate constant numbered in accordance with the reaction number, negative if in the reverse direction. v=reaction rate; V=Vmax, maximal velocity; k=rate constant; Km=Michaelis constant.

Table 2
Reaction kinetic parameters used in the simulations

Each row corresponds to a process, the first column indicating reaction number, the second the reaction, and the other columns the parameter values in the various models (A–E). The second row describes what case the model is meant to simulate. Rates are expressed in nM/min or/min, time in minutes. Initially R = 1 nM; all three MAPKs were in the unphosphorylated state at 2 nM, pseudokinase in the unphosphorylated form at 6 nM in Models B–E, whilst all other initial concentrations were zero. Differences between any model and Model A are highlighted by bold face italic type. All KM values were taken as 0.1 nM.

Reaction number ↓ Reaction type ↓ Model A Model B Model C Model D* Model E 
Model type →  Normal pathway: no pseudoenzyme Standard pseudoenzyme Pseudoenzyme
Not phosphorylated 
Converse pseudoenzyme* Both pseudo and converse pseudoenzyme 
 Initial Pseudo_E 0 nM 6 nM 6 nM 6 nM 6 nM 
 R1 MAPKKK phosphorylation kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min 
 R2 MAPKKKP dephosphorylation kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min 
 R3 MAPKK phosphorylation kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min 
 R4 MAPKKP dephosphorylation kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min 
 R5 MAPK phosphorylation kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min 
 R6 MAPKP dephosphorylation kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min 
 R7 Activator degradation k1 = 10/min k1 = 10/min k1 = 10/min k1 = 10/min k1 = 10/min 
 R8 Pseudo_E phosphorylation kcat = 100 nM/min kcat = 100 nM/min kcat=0 nM/min kcat = 100 nM/min kcat = 100 nM/min 
 R9 Pseudo_E_P dephosphorylation kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min 
 R10 Pseudo_E_P binding to MAPKK k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 0.0/nM/mink2=0.0/min k1 = 5.0/nM/min
k2 = 0.1/min 
 R11 Pseudo_E_P binding to MAPKKP k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 5.0/nM/mink2 = 0.1/min k1 = 5.0/nM/mink2 = 0.1/min 
Reaction number ↓ Reaction type ↓ Model A Model B Model C Model D* Model E 
Model type →  Normal pathway: no pseudoenzyme Standard pseudoenzyme Pseudoenzyme
Not phosphorylated 
Converse pseudoenzyme* Both pseudo and converse pseudoenzyme 
 Initial Pseudo_E 0 nM 6 nM 6 nM 6 nM 6 nM 
 R1 MAPKKK phosphorylation kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min 
 R2 MAPKKKP dephosphorylation kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min 
 R3 MAPKK phosphorylation kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min kcat = 2.0/min 
 R4 MAPKKP dephosphorylation kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min kcat = 0.05 nM/min 
 R5 MAPK phosphorylation kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min kcat = 100/min 
 R6 MAPKP dephosphorylation kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min kcat = 10.0 nM/min 
 R7 Activator degradation k1 = 10/min k1 = 10/min k1 = 10/min k1 = 10/min k1 = 10/min 
 R8 Pseudo_E phosphorylation kcat = 100 nM/min kcat = 100 nM/min kcat=0 nM/min kcat = 100 nM/min kcat = 100 nM/min 
 R9 Pseudo_E_P dephosphorylation kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min kcat = 5.0 nM/min 
 R10 Pseudo_E_P binding to MAPKK k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 5.0/nM/min
k2 = 0.1/min 
k1 = 0.0/nM/mink2=0.0/min k1 = 5.0/nM/min
k2 = 0.1/min 
 R11 Pseudo_E_P binding to MAPKKP k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 0.0/nM/min
k2 = 0.0/min 
k1 = 5.0/nM/mink2 = 0.1/min k1 = 5.0/nM/mink2 = 0.1/min 

*Converse pseudoenzyme (binding to MAPKKP instead of MAPKK).

The dotted line in Figure 3A shows the normal behavior of this MAPK pathway in silico in terms of the transient nature of ERK phosphorylation subsequent to a fast transient pulse of activator R (active R decayed within a minute; not shown): MAPK became fully phosphorylated within a minute, remained fully phosphorylated for some 40 min, and then decayed almost as rapidly as it had arisen. We took the pseudoenzyme as homologous to MAPKKK and as still being able to bind to MAPKK but unable to phosphorylate it. We expected that the pseudokinase would reduce the level of activation of MAPK, but that the activation would last equally long. We found the complete opposite (see the full line in Figure 3A): the activation level of MAPK was unaltered, i.e. complete, but the duration of activation was much reduced by the mere presence of the, inactive, pseudokinase.

A pseudokinase could have a major effect on the temporal characteristics of signal transduction.

Figure 3.
A pseudokinase could have a major effect on the temporal characteristics of signal transduction.

Concentration versus time for the phosphorylated form of MAPK (ERK) obtained for a simplified kinetic model of the MAPK route to which a pseudokinase version of MAPKKK similarly activated by the upstream signaling as the MAPKKK had been added. Analysis of the proposed network with 11 sets of kinetic reactions as detailed in Tables 1 and 2. (A) Performance of the MAPK cascade in the absence (Model A of Table 2: dotted line) and presence (Model B: full line) of active pseudoenzyme. (B) Performance of a standard pseudoenzyme if it cannot be phosphorylated (Model C: dashed line), of a converse pseudoenzyme that binds and sequesters MAPPKKP rather than MAPKK (Model D: full line), and of a pseudoenzyme that binds to both MAPKK and MAPKKP (Model E). (C and D) Time dependence of various components of the cascade in the absence (C; Model A) and presence (D; Model B) of standard pseudoenzyme.

Figure 3.
A pseudokinase could have a major effect on the temporal characteristics of signal transduction.

Concentration versus time for the phosphorylated form of MAPK (ERK) obtained for a simplified kinetic model of the MAPK route to which a pseudokinase version of MAPKKK similarly activated by the upstream signaling as the MAPKKK had been added. Analysis of the proposed network with 11 sets of kinetic reactions as detailed in Tables 1 and 2. (A) Performance of the MAPK cascade in the absence (Model A of Table 2: dotted line) and presence (Model B: full line) of active pseudoenzyme. (B) Performance of a standard pseudoenzyme if it cannot be phosphorylated (Model C: dashed line), of a converse pseudoenzyme that binds and sequesters MAPPKKP rather than MAPKK (Model D: full line), and of a pseudoenzyme that binds to both MAPKK and MAPKKP (Model E). (C and D) Time dependence of various components of the cascade in the absence (C; Model A) and presence (D; Model B) of standard pseudoenzyme.

Figure 3C,D compares the time dependencies of the various players in the cascade between the absence (Figure 3C) and presence (Figure 3D) of the pseudokinase, using a logarithmic time scale. Not plotted is the MAPKKKP, which was produced and again decayed within a minute in both cases. Figure 3D shows that much of the MAPKK is complexed by the phosphorylated pseudoenzyme, thereby reducing the concentration of the MAPKK that could be phosphorylated by MAPKKKP in the short time span the latter was available. The result is that the extent of phosphorylation of MAPKK achieved was reduced by a factor of 3. Because, after its synthesis, MAPKKP disappears by effectively a zero-order process, its disappearance thereby took three times less time when pseudoenzyme was present (Figure 3D) than when it was absent (Figure 3C), as did the phosphorylation state of MAPK (MAPKP).

When compared with the dotted line in Figure 3A, the dashed line in Figure 3B confirms the expectation that if the pseudoenzyme cannot be phosphorylated, the MAPK pathway should function as before. The full line in Figure 3B shows the results for the case where the pseudoenzyme interacted with the phosphorylated form of the MAPKK, rather than with its unphosphorylated form. We call this the ‘converse pseudoenzyme’. Intriguingly, now two peaks of MAPKP appeared, with a deep trough in between. The dashed line shows the case of a pseudokinase binding both the MAPK and MAPKP, again leading to quite a different time stamp and fitness. We conclude that pseudosignalers in a signal transduction cascade could have substantial and counterintuitive effects on the signal transduction by that cascade.

Network evolution

Shuffling, decoupling, and recoupling as a possible origin of pseudoenzymes and pseudosignalers

By attributing both reader and writer functions to proteins, we have introduced the potential that each of these have led independent evolutionary lives. In part, this hinges on the definition of ‘protein’. In its original biochemical definition, a protein was a unit that could be isolated as such. It might perform a molecular activity of some sort. Many such proteins consisted of a quaternary structure of protein subunits each with an independent amino- and carboxy-terminal end. In the DNA sequence, each of these subunits would correspond to an identifiable open reading frame (called ‘gene’ in molecular biology’). Multiple such open reading frames together would therewith encode a single inheritable molecular function (called ‘gene’ in classical biology). With the recognition of protein domains in proteins that appeared to operate as evolutionary units [58], the concept of subunit became extended to parts of a protein that were mostly identifiable in terms of homology of tertiary or even primary structure. The ‘reader ‘and ‘writer’ function of proteins that we introduced above may correspond to classical subunits or to protein domains. If they do not, then the likelihood that a mutation destroying the writer function of a protein leaves the reader function intact, may be smaller, but perhaps not zero.

The organization of proteins in terms of subunits and domains has greatly accelerated evolution, because it allowed for new functions to be generated by recombination of compatible and quasi-autonomous subfunctions. Conceivably, a metabolic kinase able to phosphorylate a hexose could quickly evolve to a kinase phosphorylating a triose, by exchanging its binding site for the hexose with that of a triose. Similarly, proteins such as the EGF receptor could obtain ATP-binding sites by evolutionary reshuffling of parts of genes that correspond to the function of binding ATP. An additional requirement for this is that the binding of the ATP should couple somehow to the ‘writing’ activity of the protein. This could be the exposure of the terminal pyrophosphate bond to oscillatory movements induced by the reading (see below) in the case of a kinase or the induction of a rotating protein state in a proton-motive ATPase, or stabilization of a higher Gibbs energy protein conformation in the case of allosteric regulation.

Figure 1 illustrates our proposal that through gene shuffling, a signaler's writer domain may be coupled to the reader domain of a metabolic enzyme or a protein kinase. That additional domain may still have its original writer domain associated, but its functionality and hence evolutionary conservation would no longer depend on its activity, explaining why deleterious mutations in that domain would persist. The new signaler in Figure 1E would read the information in [A] but would now not regulate the phosphorylation of the protein E but some other signaling function. Its paralog would still be carrying out that [A]-regulated protein kinase function.

The enzymes involved in a metabolic pathway may also be organized in terms of writer and reader domains. Figure 4 illustrates how such shuffling of reader domains could generate new regulatory networks around a metabolic pathway, which may again also transpose to signalers. Figure 4A illustrates a case where the middle enzyme is regulated by the level of cAMP by having the corresponding reader domain. In Figure 4B, the third enzyme has acquired the same reader domain, whilst the first enzyme has acquired a domain that assesses the concentration of the pathway product. Figure 4D shows that by gene shuffling, the reading domain may associate with a single-writer domain elsewhere in the genome (Figure 4C). Because different ecological niches differ in metabolic requirements, reshuffling should help speciation.

Transfers of reader domains between metabolic enzymes enable diverse pathway regulation.

Figure 4.
Transfers of reader domains between metabolic enzymes enable diverse pathway regulation.

(A) Three metabolic enzymes in a reaction. One does have a ‘reader’ and a ‘writer’ domain, whilst the other two lack ‘reader’ domains; they are not regulated by the cAMP status of the cell and are writing constitutively. (B) Two metabolic enzymes are regulated by the same ‘reader’, which reads the cAMP status of the cell, whilst the first enzyme is regulated through an attached reader domain that reads the concentration of the product of the pathway. (C) The same as (A) but emphasizes the presence of a signaler writer domain (the lightning arrow) elsewhere in the genome, whilst (D) illustrates how the [cAMP] reader domain may have become associated genetically (through partial gene duplication) to a writer domain that does not write by catalyzing a chemical reaction but by signaling. This is no longer a (regulated) enzyme but a pseudoenzyme, as it has lost its catalytic activity. The writer may here be a transcription factor with increased activity for its promoter or a protein kinase such as in Figure 1.

Figure 4.
Transfers of reader domains between metabolic enzymes enable diverse pathway regulation.

(A) Three metabolic enzymes in a reaction. One does have a ‘reader’ and a ‘writer’ domain, whilst the other two lack ‘reader’ domains; they are not regulated by the cAMP status of the cell and are writing constitutively. (B) Two metabolic enzymes are regulated by the same ‘reader’, which reads the cAMP status of the cell, whilst the first enzyme is regulated through an attached reader domain that reads the concentration of the product of the pathway. (C) The same as (A) but emphasizes the presence of a signaler writer domain (the lightning arrow) elsewhere in the genome, whilst (D) illustrates how the [cAMP] reader domain may have become associated genetically (through partial gene duplication) to a writer domain that does not write by catalyzing a chemical reaction but by signaling. This is no longer a (regulated) enzyme but a pseudoenzyme, as it has lost its catalytic activity. The writer may here be a transcription factor with increased activity for its promoter or a protein kinase such as in Figure 1.

We contend that many pseudoenzymes and pseudosignalers correspond to results of genetic reshuffling where enzymes inclusive of reader and writer domains have been associated to different writer domains. The original writer domains have thereby become inessential for the protein's contribution to fitness, and the new writer domains have become useful, explaining the evolutionary persistence of the new protein. This phenomenon would constitute an explanation for the fact that our planet has so many different microorganisms that all in essence harbor much the same genome-wide metabolic map, a hardware that is hard to change by evolution. What is easier to change, i.e. the regulatory software, sits in the reader domains, in the writer domains of the signalers, or in the connections between readers and writers. Together with the reader multiplicity (see above), this also may help explain the old conundrum why enzymes are so big [59].

Multiplex writer domains, their coupling, and gear-shifting in bioenergetics

Above we noted that an enzyme may have multiple reader domains, one for each substrate, one for each allosteric regulator, and potentially another one if the enzyme dimerizes. But will it have multiple writer domains? Isomerases may only have a single such domain, but other metabolic enzymes may have two. Oxidoreductases perform one writing action by which they extract electrons from an electron–donor molecule and engage in a second such action when donating the electron(s) to an electron acceptor. Most often either the electron donor or the electron acceptor is a redox coenzyme such as NAD(H), ferredoxin, a quinone, or NADP(H). If the oxidoreductase contains a metal center that can store the electrons, the two writing activities can function independently to some extent. Similarly, ATP hydrolysis and Na+/K+ transport can be two writer actions coupled through a common high-energy intermediate in the form of a phosphorylated enzyme state [60].

Microorganisms called acetogens may have been important for the origin of life on this planet [61,62]. They can fix carbon dioxide anaerobically using hydrogen as the electron donor in processes coupled to the synthesis of ATP [63,64]. Acetogens are home to redox enzymes that interconvert coenzymes such as the ones mentioned above. These include so-called electron bifurcating enzymes [65], which have three writer domains, enabling them to carry out two different redox reactions starting from the same electron donor, one thermodynamically uphill and the other thermodynamically downhill; the third theoretical possibility should be forbidden by some mechanism: they do this in a coupled way, enabling the first of the three reactions to proceed using the driving force of the second. An example is the non-membrane-bound hydrogenase HydABCD of Acetobacterium woodii, which has electron centers (cofactors or prosthetic groups that can store electrons) on board, i.e. iron–sulfur centers and flavins, and three writing domains, i.e. for the oxidation of hydrogen, ferredoxin, and NADH. By proper coupling of these writings, the enzyme can reduce ferredoxin at the same time as NAD whilst oxidizing hydrogen. Two-thirds of the reduced ferredoxin and NAD are then used to reduce carbon dioxide to acetate, therewith providing the carbon building blocks necessary for growth biochemistry. One-third of the reduced ferredoxin is oxidized by RnfB, a writer in the membrane-bound Rnf complex. A second writer in this complex (RnfC) uses the electrons to reduce NAD. Most importantly, possibly also for the earliest bioenergetics of this planet, these writings are coupled to the action of a third writer (RnfD), i.e. one that enables the outward movement of sodium ions across the membrane. This generates an electrochemical potential difference for Na+ across the bacterial membrane, which can then be used by the sodium motive ATPase of the organism to drive the synthesis of ATP [66]. The Rnf complex lacks the cytochromes of the better known, more ‘modern’, electron transfer chains and may thereby constitute one of the earliest mechanisms for the generation of an electrochemical potential difference able to drive the synthesis of ATP.

The genera Acetobacterium and Clostridium, to which most acetogens belong, are highly versatile in how they bring about the synthesis of acetate coupled to the phosphorylation of ADP. They appear to avail of a redox protein construction kit [68] of enzymes that are similar, except that writer domains of one specificity (e.g. for NAD) have been exchanged with ones of a different specificity (e.g. for Fd). This may enable C. ljungdahlii to oxidize hydrogen either with ferredoxin plus NAD or ferredoxin plus NADP as an electron acceptor, and to reduce carbon dioxide to formate with ferredoxin, hydrogen, or ferredoxin plus NADPH as the electron donor. The consequence is that acetogens may function at a variety of yields of ATP per acetate produced [63].

A formate dehydrogenase that uses ferredoxin rather than hydrogen as the electron donor would lack a functional writer that can accept electrons from hydrogen. In terms of its ability to oxidize hydrogen, it would appear to be a dead enzyme, a pseudoenzyme indeed. Rephrasing what we wrote above, we here propose that many pseudoenzymes are cases of the enzyme construction kit [68] of evolution, where we simply have not yet recognized the writer or reader domain by which an original reader or writer domain has been replaced.

The complex formate dehydrogenase of C. ljungdahlii that was mentioned above exists in a complex gene cluster (Cju_c06990-07080), but the same organism contains another gene cluster (Clju_c20030–20040) that appears specific for ferredoxin as the electron donor. This suggests that the diversity of redox enzymes with the corresponding diversity in ATP/acetate ratios is not just a diversity of species within genera, but may constitute a more dynamic diversity within a single species C. ljungdahlii. Such diversity would enable the organism to function at a variety of ATP/acetate production ratios. This would correspond to dynamic gear-shifting, which should be useful when the going gets tough for the organism in the sense of an increased ATP/ADP ratio to work against. The organism would then continue to be able to synthesize ATP but at a lower rate by shifting to a lower gear [21].

Analysis of the genome-wide network of C. ljungdahlii has shown that due to the redox enzyme variety and the number of possible ways, the redox enzyme can be networked, a combinatorial explosion emerges, leading to no fewer than 24 possible ATP/acetate stoichiometries [21]. After eliminating some of these because they are smaller or equal to zero and others because they are inconsistent with the second law of thermodynamics, 15 feasible gear states and hence ATP/acetate production ratios remained, possibly enabling almost seamless gear-shifting (Figure 5). This degree of dynamics in intracellular bioenergetics because of the shuffling of writer domains between enzymes remains speculative, however: it is unclear whether, indeed, enzymes can re-associate different writer domains dynamically at a time scale of less than a cell-cycle time. Less extensive examples of such gear-shifting may have arisen later in evolution, where different cytochrome oxidases with different H+/e stoichiometry absorb electrons from the bc1 complex of the cytochrome-containing electron transfer chain [69], leading to different growth yields.

Proposed gear-shifting for C. ljungdahlii: the optimal ratio between ATP and acetate production flux for various Gibbs energies of ATP hydrolysis.

Figure 5.
Proposed gear-shifting for C. ljungdahlii: the optimal ratio between ATP and acetate production flux for various Gibbs energies of ATP hydrolysis.

A genome-wide metabolic map for C. ljungdahlii was extended with redox reactions identified by Schuchmann and Müller [67] and for a fixed lower bound of carbon dioxide [2] and hydrogen influx [4], requiring the acetate flux to be 1 or higher (which set that flux at 1); the flux pattern was optimized for ATP output flux. This was done for each combination of electron donors and electron acceptors for hydrogenase, formate dehydrogenase, MTHFR, and MTHFD, as present in the extended genome-wide metabolic map, thus fixing the redox route for each calculation. Calling the ratio of ATP synthesis flux to acetate flux n, the ATP synthesis flux was calculated as JATP = n·(1 − n·(ΔGATP/ΔGA)), with ΔGA representing the Gibbs energy released in acetate synthesis from CO2 and hydrogen (2CO2 + 4H2 → CH3COOH + 2H2O) of 40 kJ/mol taken from Schuchmann and Müller [67]. For each ΔGATP, the n with the highest JATP was selected as the optimum n and plotted as the ordinate with ΔGATP as the abscissa (see further ref. [21]).

Figure 5.
Proposed gear-shifting for C. ljungdahlii: the optimal ratio between ATP and acetate production flux for various Gibbs energies of ATP hydrolysis.

A genome-wide metabolic map for C. ljungdahlii was extended with redox reactions identified by Schuchmann and Müller [67] and for a fixed lower bound of carbon dioxide [2] and hydrogen influx [4], requiring the acetate flux to be 1 or higher (which set that flux at 1); the flux pattern was optimized for ATP output flux. This was done for each combination of electron donors and electron acceptors for hydrogenase, formate dehydrogenase, MTHFR, and MTHFD, as present in the extended genome-wide metabolic map, thus fixing the redox route for each calculation. Calling the ratio of ATP synthesis flux to acetate flux n, the ATP synthesis flux was calculated as JATP = n·(1 − n·(ΔGATP/ΔGA)), with ΔGA representing the Gibbs energy released in acetate synthesis from CO2 and hydrogen (2CO2 + 4H2 → CH3COOH + 2H2O) of 40 kJ/mol taken from Schuchmann and Müller [67]. For each ΔGATP, the n with the highest JATP was selected as the optimum n and plotted as the ordinate with ΔGATP as the abscissa (see further ref. [21]).

The coupling of readers and primitive membrane-associated ATP synthesis

Above we have already noted that ion-motive ATPases may have a similar modular origin, having arisen from the integration of a writer of ion passage across a membrane and a reader-binding ATP. Perhaps, pH gradients sustained by flow from geothermal vents across early membranes, which should then be less proton than ion permeable in order to dissipate the electric potential [70], could provide the Gibbs energy for early proton-motivated pyrophosphate synthesis. The mechanism of this would still be unclear if these pH gradients were steady and complex proton-motive proteins were still absent. If such proton gradients oscillated or fluctuated, however, or if the membrane would be more proton than ion permeable such that the electric potential would oscillate or be noisy, fairly primitive protein structures should be able to extract Gibbs energy from the oscillations [71]. Should that structure already contain two reading sites for phosphate, for instance, an oscillating electric field across one of the two bound phosphates could make the two bumps into each other and then cause the endergonic synthesis of pyrophosphate. This pyrophosphate could have played the role of ATP as central Gibbs energy coenzyme in early bioenergetics, until it was replaced by ATP in organisms that were fitter because they used AMP as the carrier of the coenzyme pyrophosphate.

Gear-shifting might also occur within the Na+- or H+-ATPase, through the possibly variable number and type of c-subunits that determine the number of ions to be translocated. Müller [72] has proposed that the A. woodii F0F1 ATPase may adjust its c-subunit composition and thereby H+/ATP stoichiometry, enabling it to shift between an ATP synthesizing and an ATP-hydrolyzing mode.

Concluding remarks

Pseudoenzymes and pseudosignalers may be tips of the iceberg of an evolution that moved forward through shuffling of networks into optimality rather than by evolving proteins to their individual optimality. The reader–writer concept proposed here may help understand the corresponding plasticity, which may also explain the phenomenon of pseudoenzymes and pseudosignalers. It should be worth our while to examine further than we were able to do here whether this concept applies to the emerging plethora of pseudoenzymes [27]. The focus on interactions between functionalities that has been introduced by systems biology may improve our understanding of molecular biology, where the word biology is then related to function and fitness.

Abbreviations

     
  • ATase

    adenylyl transferase

  •  
  • cAMP

    cyclic AMP

  •  
  • CARDs

    caspase recruitment domains

  •  
  • CDK

    cyclin-dependent kinase

  •  
  • EGFR

    epidermal growth factor receptor

  •  
  • ERK

    bottom protein kinase in the MAP kinase/phosphatase cascade

  •  
  • ERK-PP

    dual phosphorylated form of ERK

    activator of transcription

  •  
  • GS

    glutamine synthetase

  •  
  • MAP

    mitogen-activated protein

  •  
  • MAPK

    MAP kinase

  •  
  • MAPKK

    MAPK kinase

  •  
  • MAPKKK

    MAPK kinase kinase

  •  
  • MEK

    middle protein kinase in the MAP kinase/phosphatase cascade

    phosphorylated and activated by phosphorylated RAF

    kinase of ERK

  •  
  • NR

    Nitrogen Regulator

  •  
  • PEP

    phosphoenol-pyruvate

  •  
  • PFAS

    phosphoribosyl-formylglycinamidine synthetase

  •  
  • PKA

    protein kinase A

  •  
  • RAF

    top kinase in the MAP kinase kinase/phosphatase pathway

    acivated by active RAS

    kinase of MEK

  •  
  • Ras

    regulator R

  •  
  • RIG

    RIG-1 is a cytosolic PRR (pattern recognition receptor) of the innate immune system

  •  
  • SBGN

    Systems Biology Graphical Notation

    lingua franca for network diagrams that are automatically translatable into dynamic models

  •  
  • SOS

    protein connecting Grb2

    which binds the EGF receptor upon phosphorylation of the latter

    to Ras

    thereby activating the latter

Funding

This work was financially supported by the Netherlands Organization for Scientific Research (NWO) in the integrated program of WOTRO [W01.65.324.00/project 4] Science for Global Development as well as by various systems biology grants, including Synpol: EU-FP7 [KBBE.2012.3.4-02 #311815], Corbel: EU-H2020 [NFRADEV-4-2014-2015#654248], Epipredict: EU-H2020 MSCA-ITN-2014-ETN: Marie Skłodowska-Curie Innovative Training Networks (ITN-ETN) [#642691], and BBSRC China [BB/J020060/1].

Acknowledgments

We are indebted to an anonymous reviewer who alerted us to the issue discussed in the section ‘Pseudoenzymes: What's in a name?’ and to both reviewers for additional suggestions.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Leslie
,
M
. (
2013
)
‘Dead’ enzymes show signs of life
.
Science
340
,
25
27
doi:
2
Todd
,
A.E.
,
Orengo
,
C.A.
and
Thornton
,
J.M.
(
2002
)
Sequence and structural differences between enzyme and nonenzyme homologs
.
Structure
10
,
1435
1451
doi:
3
Puente
,
X.S.
,
Sánchez
,
L.M.
,
Overall
,
C.M.
and
López-Otín
,
C
. (
2003
)
Human and mouse proteases: a comparative genomic approach
.
Nat. Rev. Genet.
4
,
544
558
doi:
4
Pils
,
B.
and
Schultz
,
J
. (
2004
)
Inactive enzyme-homologues find new function in regulatory processes
.
J. Mol. Biol.
340
,
399
404
doi:
5
Brew
,
K.
,
Vanaman
,
T.C.
and
Hill
,
R.L.
(
1967
)
Comparison of the amino acid sequence of bovine alpha-lactalbumin and hens egg white lysozyme
.
J. Biol. Chem.
242
,
3747
3749
PMID:
[PubMed]
6
Thiele
,
I.
,
Swainston
,
N.
,
Fleming
,
R.M.T.
,
Hoppe
,
A.
,
Sahoo
,
S.
,
Aurich
,
M.K.
et al. 
(
2013
)
A community-driven global reconstruction of human metabolism
.
Nat. Biotechnol.
31
,
419
425
doi:
7
Östberg
,
L.J.
,
Persson
,
B.
and
Höög
,
J.-O.
(
2015
)
The mammalian alcohol dehydrogenase genome shows several gene duplications and gene losses resulting in a large set of different enzymes including pseudoenzymes
.
Chem.-Biol. Interact.
234
,
80
84
doi:
8
Östberg
,
L.J.
,
Persson
,
B.
and
Höög
,
J.-O.
(
2016
)
Computational studies of human class V alcohol dehydrogenase — the odd sibling
.
BMC Biochem.
17
,
16
doi:
9
Eyers
,
P.A.
and
Murphy
,
J.M.
(
2013
)
Dawn of the dead: protein pseudokinases signal new adventures in cell biology
.
Biochem. Soc. Trans.
41
,
969
974
doi:
10
Manning
,
G.
,
Whyte
,
D.B.
,
Martinez
,
R.
,
Hunter
,
T.
and
Sudarsanam
,
S
. (
2002
)
The protein kinase complement of the human genome
.
Science
298
,
1912
1934
doi:
11
Hammaren
,
H.M.
,
Virtanen
,
A.T.
and
Silvennoinen
,
O.
(
2015
)
Nucleotide-binding mechanisms in pseudokinases
.
Biosci. Rep.
36
,
e00282
doi:
12
van Heeswijk
,
W.C.
,
Westerhoff
,
H.V.
and
Boogerd
,
F.C.
(
2013
)
Nitrogen assimilation in Escherichia coli: putting molecular data into a systems perspective
.
Microbiol. Mol. Biol. Rev.
77
,
628
695
doi:
13
Ortega
,
F.
,
Acerenza
,
L.
,
Westerhoff
,
H.V.
,
Mas
,
F.
and
Cascante
,
M
. (
2002
)
Product dependence and bifunctionality compromise the ultrasensitivity of signal transduction cascades
.
Proc. Natl Acad. Sci. U.S.A.
99
,
1170
1175
doi:
14
Adler
,
S.P.
,
Purich
,
D.
and
Stadtman
,
E.R.
(
1975
)
Cascade control of Escherichia coli glutamine synthetase. Properties of the PII regulatory protein and the uridylyltransferase–uridylyl-removing enzyme
.
J. Biol. Chem.
250
,
6264
6272
PMID 239942
15
van Heeswijk
,
W.C.
,
Hoving
,
S.
,
Molenaar
,
D.
,
Stegeman
,
B.
,
Kahn
,
D.
and
Westerhoff
,
H.V.
(
1996
)
An alternative PII protein in the regulation of glutamine synthetase in Escherichia coli
.
Mol. Microbiol.
21
,
133
146
doi:
16
Xu
,
Y.
,
Cheah
,
E.
,
Carr
,
P.D.
,
van Heeswijk
,
W.C.
,
Westerhoff
,
H.V.
,
Vasudevan
,
S.G.
et al. 
(
1998
)
Glnk, a PII-homologue: structure reveals ATP binding site and indicates how the T-loops may be involved in molecular recognition
.
J. Mol. Biol.
282
,
149
165
doi:
17
Boogerd
,
F.C.
,
Ma
,
H.
,
Bruggeman
,
F.J.
,
van Heeswijk
,
W.C.
,
García-Contreras
,
R.
,
Molenaar
,
D.
et al. 
(
2011
)
AmtB-mediated NH3 transport in prokaryotes must be active and as a consequence regulation of transport by GlnK is mandatory to limit futile cycling of NH4+/NH3
.
FEBS Lett.
585
,
23
28
doi:
18
Glass
,
J.I.
,
Assad-Garcia
,
N.
,
Alperovich
,
N.
,
Yooseph
,
S.
,
Lewis
,
M.R.
,
Maruf
,
M.
et al. 
(
2006
)
Essential genes of a minimal bacterium
.
Proc. Natl Acad. Sci. U.S.A.
103
,
425
430
doi:
19
Swainston
,
N.
,
Currin
,
A.
,
Day
,
P.J.
and
Kell
,
D.B.
(
2014
)
Genegenie: optimized oligomer design for directed evolution
.
Nucleic Acids Res.
42
,
W395
W400
doi:
20
Raamsdonk
,
L.M.
,
Teusink
,
B.
,
Broadhurst
,
D.
,
Zhang
,
N.
,
Hayes
,
A.
,
Walsh
,
M.C.
et al. 
(
2001
)
A functional genomics strategy that uses metabolome data to reveal the phenotype of silent mutations
.
Nat. Biotechnol.
19
,
45
50
doi:
21
Mondeel
,
D.G.A.
,
Rehman
,
S.
,
Zhang
,
Y.
,
Verma
,
M.
,
Durre
,
P.
,
Barberis
,
M.
and
Westerhoff
,
H.V.
(
2017
)
Maps for when the living gets tough: maneuvering through a hostile energy landscape
.
Proceedings 6th IFAC Conference on Foundations of Systems Biology in Engineering, TuPMP.1.
22
Groen
,
A.K.
,
Wanders
,
R.J.
,
Westerhoff
,
H.V.
,
van der Meer
,
R.
and
Tager
,
J.M.
(
1982
)
Quantification of the contribution of various steps to the control of mitochondrial respiration
.
J. Biol. Chem.
257
,
2754
2757
PMID:
[PubMed]
23
Venters
,
B.J.
,
Wachi
,
S.
,
Mavrich
,
T.N.
,
Andersen
,
B.E.
,
Jena
,
P.
,
Sinnamon
,
A.J.
et al. 
(
2011
)
A comprehensive genomic binding map of gene and chromatin regulatory proteins in Saccharomyces
.
Mol. Cell
41
,
480
492
doi:
24
Dawkins
,
R.
(
1976
)
The Selfish Gene
,
Vol. 11
, p.
224
,
Oxford University Press
,
New York
25
ter Kuile
,
B.H.
and
Westerhoff
,
H.V.
(
2001
)
Transcriptome meets metabolome: hierarchical and metabolic regulation of the glycolytic pathway
.
FEBS Lett.
500
,
169
171
doi:
26
Brown
,
G.C.
,
Westerhoff
,
H.V.
and
Kholodenko
,
B.N.
(
1996
)
Molecular control analysis: control within proteins and molecular processes
.
J. Theor. Biol.
182
,
389
396
doi:
27
Reiterer
,
V.
,
Eyers
,
P.A.
and
Farhan
,
H
. (
2014
)
Day of the dead: pseudokinases and pseudophosphatases in physiology and disease
.
Trends Cell Biol.
24
,
489
505
doi:
28
Ullrich
,
A.
and
Schlessinger
,
J
. (
1990
)
Signal transduction by receptors with tyrosine kinase activity
.
Cell
61
,
203
212
doi:
29
Holbro
,
T.
,
Beerli
,
R.R.
,
Maurer
,
F.
,
Koziczak
,
M.
,
Barbas
, III,
C.F.
and
Hynes
,
N.E.
(
2003
)
The ErbB2/ErbB3 heterodimer functions as an oncogenic unit: ErbB2 requires ErbB3 to drive breast tumor cell proliferation
.
Proc. Natl Acad. Sci. U.S.A.
100
,
8933
8938
doi:
30
Wieduwilt
,
M.J.
and
Moasser
,
M.M.
(
2008
)
The epidermal growth factor receptor family: biology driving targeted therapeutics
.
Cell. Mol. Life Sci.
65
,
1566
1584
doi:
31
Rajakulendran
,
T.
,
Sahmi
,
M.
,
Lefrancois
,
M.
,
Sicheri
,
F.
and
Therrien
,
M
. (
2009
)
A dimerization-dependent mechanism drives RAF catalytic activation
.
Nature
461
,
542
545
doi:
32
Zhang
,
H.
,
Koo
,
C.Y.
,
Stebbing
,
J.
and
Giamas
,
G
. (
2013
)
The dual function of KSR1: a pseudokinase and beyond
.
Biochem. Soc. Trans.
41
,
1078
1082
doi:
33
Reiterer
,
V.
,
Fey
,
D.
,
Kolch
,
W.
,
Kholodenko
,
B.N.
and
Farhan
,
H
. (
2013
)
Pseudophosphatase STYX modulates cell-fate decisions and cell migration by spatiotemporal regulation of ERK1/2
.
Proc. Natl Acad. Sci. U.S.A.
110
,
E2934
E2943
doi:
34
Shakespeare
,
W.H.
(
1597
)
An EXCELLENT conceited Tragedie of Romeo and Jullet
.
John Danter, London
.
35
Zhao
,
J.
,
Li
,
J.
,
Xu
,
S.
and
Feng
,
P
. (
2016
)
Emerging roles of protein deamidation in innate immune signaling
.
J. Virol.
90
,
4262
4268
doi:
36
Malumbres
,
M.
(
2014
)
Cyclin-dependent kinases
.
Genome Biol.
15
,
122
doi:
37
Lim
,
S.
and
Kaldis
,
P
. (
2013
)
Cdks, cyclins and CKIs: roles beyond cell cycle regulation
.
Development
140
,
3079
3093
doi:
38
Kõivomägi
,
M.
,
Valk
,
E.
,
Venta
,
R.
,
Iofik
,
A.
,
Lepiku
,
M.
,
Balog
,
E.R.M.
et al. 
(
2011
)
Cascades of multisite phosphorylation control Sic1 destruction at the onset of S phase
.
Nature
480
,
128
131
doi:
39
Gerard
,
C.
and
Goldbeter
,
A
. (
2009
)
Temporal self-organization of the cyclin/Cdk network driving the mammalian cell cycle
.
Proc. Natl Acad. Sci. U.S.A.
106
,
21643
21648
doi:
40
Bloom
,
J.
and
Cross
,
F.R.
(
2007
)
Multiple levels of cyclin specificity in cell-cycle control
.
Nat. Rev. Mol. Cell Biol.
8
,
149
160
doi:
41
Barberis
,
M.
,
Linke
,
C.
,
Adrover
,
M.À.
,
González-Novo
,
A.
,
Lehrach
,
H.
,
Krobitsch
,
S.
et al. 
(
2012
)
Sic1 plays a role in timing and oscillatory behaviour of B-type cyclins
.
Biotechnol. Adv.
30
,
108
130
doi:
42
Schwob
,
E.
and
Nasmyth
,
K
. (
1993
)
CLB5 and CLB6, a new pair of B cyclins involved in DNA replication in Saccharomyces cerevisiae
.
Genes Dev.
7
,
1160
1175
doi:
43
Epstein
,
C.B.
and
Cross
,
F.R.
(
1992
)
CLB5: a novel B cyclin from budding yeast with a role in S phase
.
Genes Dev.
6
,
1695
1706
doi:
44
Cross
,
F.R.
,
Schroeder
,
L.
and
Bean
,
J.M.
(
2007
)
Phosphorylation of the Sic1 inhibitor of B-type cyclins in Saccharomyces cerevisiae is not essential but contributes to cell cycle robustness
.
Genetics
176
,
1541
1555
doi:
45
Gutiérrez-Escribano
,
P.
and
Nurse
,
P.
(
2015
)
A single cyclin–CDK complex is sufficient for both mitotic and meiotic progression in fission yeast
.
Nat. Commun.
6
,
6871
doi:
46
Linke
,
C.C.A.
,
González-Novo
,
A.
,
Al Sawad
,
I.
,
Tognetti
,
S.
,
Klipp
,
E.
,
Loog
,
M.
et al. 
(
2017
)
A Clb/Cdk1-mediated regulation of Fkh2 synchronizes CLB expression in the budding yeast cell cycle
.
NPJ Syst. Biol. Appl.
,
in press
47
Kolakofsky
,
D.
and
Garcin
,
D
. (
2015
)
γHV68 vGAT: a viral pseudoenzyme pimping for PAMPs
.
Mol. Cell
58
,
3
4
doi:
48
McCubrey
,
J.A.
,
Steelman
,
L.S.
,
Chappell
,
W.H.
,
Abrams
,
S.L.
,
Wong
,
E.W.T.
,
Chang
,
F.
et al. 
(
2007
)
Roles of the Raf/MEK/ERK pathway in cell growth, malignant transformation and drug resistance
.
Biochim. Biophys. Acta
1773
,
1263
1284
doi:
49
Munshi
,
A.
and
Ramesh
,
R
. (
2013
)
Mitogen-activated protein kinases and their role in radiation response
.
Genes Cancer
4
,
401
408
doi:
50
Nguyen
,
L.K.
,
Kolch
,
W.
and
Kholodenko
,
B.N.
(
2013
)
When ubiquitination meets phosphorylation: a systems biology perspective of EGFR/MAPK signalling
.
Cell Commun. Signal
11
,
52
doi:
51
Sachs
,
K.
,
Gifford
,
D.
,
Jaakkola
,
T.
,
Sorger
,
P.
and
Lauffenburger
,
D.A.
(
2002
)
Bayesian network approach to cell signaling pathway modeling
.
Sci. STKE
2002
,
pe38
PMID:
[PubMed]
52
Huang
,
C.Y.
and
Ferrell
, Jr,
J.E.
(
1996
)
Ultrasensitivity in the mitogen-activated protein kinase cascade
.
Proc. Natl Acad. Sci. U.S.A.
93
,
10078
10083
PMID:
[PubMed]
53
Hornberg
,
J.J.
,
Bruggeman
,
F.J.
,
Binder
,
B.
,
Geest
,
C.R.
,
de Vaate
,
A.J.M.B.
,
Lankelma
,
J.
et al. 
(
2005
)
Principles behind the multifarious control of signal transduction. ERK phosphorylation and kinase/phosphatase control
.
FEBS J.
272
,
244
258
doi:
54
Schoeberl
,
B.
,
Eichler-Jonsson
,
C.
,
Gilles
,
E.D.
and
Müller
,
G
. (
2002
)
Computational modeling of the dynamics of the MAP kinase cascade activated by surface and internalized EGF receptors
.
Nat. Biotechnol.
20
,
370
375
doi:
55
Kholodenko
,
B.N.
,
Hoek
,
J.B.
and
Westerhoff
,
H.V.
(
2000
)
Why cytoplasmic signalling proteins should be recruited to cell membranes
.
Trends Cell Biol.
10
,
173
178
doi:
56
Ferrell
, Jr,
J.E.
(
1996
)
Tripping the switch fantastic: how a protein kinase cascade can convert graded inputs into switch-like outputs
.
Trends Biochem. Sci.
21
,
460
466
doi:
57
Hornberg
,
J.J.
,
Binder
,
B.
,
Bruggeman
,
F.J.
,
Schoeberl
,
B.
,
Heinrich
,
R.
and
Westerhoff
,
H.V.
(
2005
)
Control of MAPK signalling: from complexity to what really matters
.
Oncogene
24
,
5533
5542
doi:
58
Sonnhammer
,
E.L.L.
and
Kahn
,
D
. (
1994
)
Modular arrangement of proteins as inferred from analysis of homology
.
Protein Sci.
3
,
482
492
doi:
59
Sumegi
,
B.
and
Srere
,
P.A.
(
1984
)
Binding of the enzymes of fatty acid beta-oxidation and some related enzymes to pig heart inner mitochondrial membrane
.
J. Biol. Chem.
259
,
8748
8752
PMID:
[PubMed]
60
Fukushima
,
Y.
and
Tonomura
,
Y
. (
1973
)
Two kinds of high energy phosphorylated intermediate, with and without bound ADP, in the reaction of Na+-K-dependent ATPase
.
J. Biochem.
74
,
135
142
doi:
61
Drake
,
H.L.
,
Gößner
,
A.S.
and
Daniel
,
S.L.
(
2008
)
Old acetogens, new light
.
Ann. N. Y. Acad. Sci.
1125
,
100
128
doi:
62
Martin
,
W.F.
,
Sousa
,
F.L.
and
Lane
,
N
. (
2014
)
Energy at life's origin
.
Science
344
,
1092
1093
doi:
63
Schuchmann
,
K.
and
Müller
,
V
. (
2016
)
Energetics and application of heterotrophy in acetogenic bacteria
.
Appl. Environ. Microbiol.
82
,
4056
4069
doi:
64
Schuchmann
,
K.
,
Vonck
,
J.
and
Müller
,
V
. (
2016
)
A bacterial hydrogen-dependent CO2 reductase forms filamentous structures
.
FEBS J.
283
,
1311
1322
doi:
65
Buckel
,
W.
and
Thauer
,
R.K.
(
2013
)
Energy conservation via electron bifurcating ferredoxin reduction and proton/Na+ translocating ferredoxin oxidation
.
Biochim. Biophys. Acta
1827
,
94
113
doi:
66
Trifunović
,
D.
,
Schuchmann
,
K.
and
Müller
,
V
. (
2016
)
Ethylene glycol metabolism in the acetogen Acetobacterium woodii
.
J. Bacteriol.
198
,
1058
1065
doi:
67
Schuchmann
,
K.
and
Müller
,
V.
(
2014
)
Autotrophy at the thermodynamic limit of life: a model for energy conservation in acetogenic bacteria
.
Nat. Rev. Microbiol.
12
,
809
821
68
Baymann
,
F.
,
Lebrun
,
E.
,
Brugna
,
M.
,
Schoepp-Cothenet
,
B.
,
Giudici-Orticoni
,
M.-T.
and
Nitschke
,
W
. (
2003
)
The redox protein construction kit: pre-last universal common ancestor evolution of energy-conserving enzymes
.
Philos. Trans. R. Soc. Lond. B. Biol. Sci.
358
,
267
274
doi:
69
Brown
,
W.W.
,
Roggemann
,
M.C.
,
Schulz
,
T.J.
,
Havens
,
T.C.
,
Beyer
,
J.T.
and
Otten
,
L.J.
(
2001
)
Measurement and data-processing approach for estimating the spatial statistics of turbulence-induced index of refraction fluctuations in the upper atmosphere
.
Appl. Opt.
40
,
1863
1871
doi:
70
Sojo
,
V.
,
Pomiankowski
,
A.
and
Lane
,
N.
(
2014
)
A bioenergetic basis for membrane divergence in archaea and bacteria
.
PLoS Biol.
12
,
e1001926
doi:
71
Westerhoff
,
H.V.
,
Tsong
,
T.Y.
,
Chock
,
P.B.
,
Chen
,
Y.D.
and
Astumian
,
R.D.
(
1986
)
How enzymes can capture and transmit free energy from an oscillating electric field
.
Proc. Natl Acad. Sci. U.S.A.
83
,
4734
4738
doi:
72
Müller
,
V
. (
2003
)
Energy conservation in acetogenic bacteria
.
Appl. Environ. Microbiol.
69
,
6345
6353
doi: