Linkage between Fitness of Yeast Cells and Adenylate Kinase Catalysis

Enzymes have evolved with highly specific values of their catalytic parameters kcat and KM. This poses fundamental biological questions about the selection pressures responsible for evolutionary tuning of these parameters. Here we are address these questions for the enzyme adenylate kinase (Adk) in eukaryotic yeast cells. A plasmid shuffling system was developed to allow quantification of relative fitness (calculated from growth rates) of yeast in response to perturbations of Adk activity introduced through mutations. Biophysical characterization verified that all variants studied were properly folded and that the mutations did not cause any substantial differences to thermal stability. We found that cytosolic Adk is essential for yeast viability in our strain background and that viability could not be restored with a catalytically dead, although properly folded Adk variant. There exist a massive overcapacity of Adk catalytic activity and only 12% of the wild type kcat is required for optimal growth at the stress condition 20°C. In summary, the approach developed here has provided new insights into the evolutionary tuning of kcat for Adk in a eukaryotic organism. The developed methodology may also become useful for uncovering new aspects of active site dynamics and also in enzyme design since a large library of enzyme variants can be screened rapidly by identifying viable colonies.


Introduction
Enzymes are remarkable bio-catalysts that can tremendously increase rates of otherwise slow cellular chemical reactions, thereby making them significantly faster than global processes, such as cell division, in living organisms Thus, for example, the estimated rate enhancement of the isomerization of (R)-mandelate to (S)-mandelate by mandelate racemase is 1.7x10 15 -fold [1]. Traditionally, catalytic parameters (k cat and K M ) are obtained from data gathered from in vitro experiments, such as spectroscopic observations of substrate depletion and accumulation of product molecules [2]. During the last decade NMR spectroscopy has added significant insights regarding the importance of dynamics (the time-dependent displacement of atomic coordinates) for enzymatic reaction cycles [3][4][5][6][7][8]. Several techniques, including NMR [9][10][11] and fluorescence microscopy [12], have also provided significant advances in analyses of proteins in their native environments inside living cells. This has enabled exploration of fundamental issues regarding, for instance, mechanisms allowing maintenance of enzymes' functionality in the highly complex internal environments of living cells, where numerous variables could potentially affect their activities, such as macromolecular crowding, weak transient interactions and associated effects on translational diffusion [13].
However, the approaches mentioned above cannot address fundamental biological questions regarding the selection pressures responsible for evolutionary tuning of enzymes' catalytic parameters, k cat and K M . To address these questions we have developed an approach where we examine changes in relative fitness (obtained from growth rate constants) of yeast (Saccharomyces cerevisiae) cells expressing Escherichia coli adenylate kinase (Adk eco ) variants with precise perturbations of the enzyme's k cat and K M values for ATP turn-over (K ATP M ). The approach is conceptually related to a previous study where yeast cell growth rates were analyzed in the context of ubiquitin stability [14]. It has been shown in a prokaryotic organism (E. coli) that there exist a large catalytic overcapacity of β-galactosidase activity [15] such that the relative fitness under limiting lactose concentrations is only affected when the catalytic activity of βgalactosidase is significantly impaired. In the present study we investigate the relative fitness of an eukaryotic organism in response to variations of the catalytic activity ofthe essential enzyme adenylate kinase (Adk).
Adk catalyzes the reversible and magnesium-dependent interconversion of ATP and AMP to two ADP molecules (ATP þ AMP ! k f k r 2ADP) and is required for maintenance of the cellular energy balance. The structural basis for Adk eco has been extensively explored and there exist structures of substrate-free open [16] (Fig 1A) and also inhibitor bound closed structures [17] ( Fig 1B). Likewise, the role of dynamics for the catalytic function of Adk eco has been studied extensively. It has been shown with NMR [18,19] and single molecule FRET experiments [20] [18] that the substrate-free enzyme transiently samples a "bound-like" structural state. Adk eco is rate limited by substrate release and the microscopic explanation to this property is slow opening of the substrate binding domains in presence of bound substrate [7,21]. Yeast was chosen as the eukaryotic model organism since robust tools are available for analyzing genes encoding mutated proteins [22]. Yeast cytosolic adenylate kinase (Adk1 yeast ) was selected as a target enzyme for the following reasons: absence of the enzyme is detrimental for yeast growth [23][24][25], crystallographic structures of the enzymes in yeast [26] and E. coli [17] have been determined (Fig 2), abundant information regarding catalytic properties of variants of the E. coli homologue (Adk eco ) is available [19,27], and there is substantial (47%) sequence identity between Adk eco and Adk1 yeast [28], which is reflected in very similar three dimensional structures, with a root mean square deviation of 1.3 Å computed over C α atoms (Fig 2). On basis of the above mentioned features of both yeast as a model organism and Adk as a model enzyme, we developed a yeast cell based approach to address the evolutionary constraints of Adk catalytic parameters in context of fitness of an eukaryotic organism.

Experimental Procedures
Strains, media and genetic procedures The sources and genotypes of yeast and bacteria strains used in this study are listed in S1 Table. The yeast transformations [30], media and genetic procedures applied have been previously described [31]. The heterozygous strain UMY3969 (ADK1/adk1::kanMX) was generated from the diploid strain UMY3387 (ADK1/ADK1) by exchanging one of the ADK1 open reading frames with the KanMX cassette. Strain UMY3969 was allowed to sporulate and tetrad analysis showed that adk1Δ strains were inviable. Diploid strain UMY3969 (ADK1/adk1::kanMX) was transformed with pRS316-ADK1 followed by sporulation to obtain haploid strain UMY3974 (adk1::kanMX + pRS316-ADK1).

Plasmid constructions
A SacI-BamHI fragment containing a wild type yeast ADK1 open reading frame together with 600 bp upstream and 555 bp downstream regions was cloned into corresponding sites of either a LEU2-based low-copy number vector (pRS315) or a URA3-based low-copy number vector (pRS316), generating pRS315-ADK1 and pRS316-ADK1. To clone the E. coli adk gene under control of the yeast ADK1 promoter, we first PCR-amplified the 600 bp upstream region (as a SacI-XbaI fragment) and the 555 bp downstream region (as an XbaI-BamHI fragment) and Structures of yeast and E. coli adenylate kinase in closed and active states. The stereo-view was made by superposition of C α atoms of Adk1 yeast [26] (2AKY) and Adk eco [17] (1AKE.pdb). The inhibitor Ap5A [29] is displayed with a ball and stick representation. Adk1 yeast and the corresponding Ap5A molecule is colored blue while Adk eco with its corresponding Ap5A molecule is colored red. cloned them together into the SacI and BamHI sites of the pRS315 vector, generating pRS315-Up-XbaI-Down. In this construct, the yeast ADK1 open reading frame is exchanged with an XbaI restriction site. To obtain pRS315-adk eco , a DNA fragment encoding the wildtype E. coli adk open reading frame was PCR-amplified and exchanged with the XbaI site in pRS315-Up-XbaI-Down using the infusion cloning procedure (Clontech). To obtain k cat mutant versions of the E. coli adk gene, pRS315-adk eco was used as a template and mutations were introduced by PCR oligonucleotide-directed mutagenesis.

Protein extraction and western blotting
Cells were cultivated logarithmically at 20°C until their optical density reached 0.5 at 600 nm and 5 OD-units of cells were harvested using a previously described TCA protein extraction procedure [32]. Proteins were separated by 12% SDS-PAGE and transferred to nitrocellulose membranes. To detect Adk eco proteins, a rabbit polyclonal anti-Adk eco (Karuso, 14EF 34) antibody was used (Agrisera, Sweden). Actin was detected using a mouse anti-Act1 antibody (Thermo Scientific). Protein levels were quantified using ImageJ software. Act1 protein levels  were used as loading controls and protein levels of the Adk eco variants were normalized according to wild-type Adk 1:00 eco .

Protein expression and purification
Adk variants were produced, 15 N-enriched and purified as previously described [7]. Adk 0:007 eco and Adk 0:0002 eco did not bind to Blue Sepharose so the flow-through was loaded on a Q-Sepharose column and these variants were eluted as previously described [27].

Coupled ATPase assay
Adk activity was quantified at 20°C in the direction of ADP formation with a coupled ATPase assay as outlined previously [33]. The assay couples ADP production to oxidation of NADH through the activity of pyruvate kinase and lactate dehydrogenase. Pyruvate kinase catalyzes the conversion of ADP and phosphoenolpyruvate to pyruvate and ATP. Lactate dehydrogenase in turn catalyzes the conversion of pyruvate and NADH to lactate and NAD+. The assay was performed in a buffer consisting of 80 mM KCl, 2mM MgCl 2 and 100 mM Tris at pH 7.5. The constituents used for the coupled reactions were phosphoenolpyruvate present at 0.4 mM and NADH present at 0.2 mM. The AMP concentration was held constant at 300 μM which is well above the K AMP M value of Adk eco that previously was found to be 26 μM [27]. 1.1-43 nM of Adk variants were used in the reactions. The consumption of NADH was quantified by following the change in absorbance at 340 nm and by using an extinction coefficient of 6220 M -1 cm -1 . The corresponding time-dependent ATP consumption (V in S1 Fig) is related to the half the change in NADH concentration since two ADP molecules is produced for each ATP molecule consumed. Care was taken to verify that sufficient amounts of pyruvate kinase and lactate dehydrogenase were present such that the NADH oxidation was limited by Adk catalysis. Reagents were purchased from Sigma-Aldrich. Catalytic parameters, k cat , and K ATP M , were obtained through fits of initial velocities (V) in response to variation of the ATP concentration ([S] in Eq 1) to the Michaelis-Menten equation: Since AMP is held at a constant concentration the reported K ATP M values should be treated as apparent K M values. The catalytic parameters of the variants I116G and K13Q were taken from the previous studies [34] and [35], respectively.

NMR spectroscopy
NMR spectra were acquired on a Bruker 850 MHz Avance III HD equipped with a 5 mm TCI cryoprobe (Bruker Biospin) or a Bruker 500 MHz Avance III equipped with a 5 mm TBI probe. The samples contained 100-400 μM 15 N-labeled Adk in a buffer consisting of 10% (v/v) 2 H 2 O, 50 mM NaCl and 30 mM MOPS at pH 6.0.

Circular dichroism
Far ultraviolet circular dichroism (CD) experiments were performed on a Jasco J-810 spectropolarimeter. Thermal unfolding was followed by monitoring the CD signal at 220 nm in a 1 mm cuvette with a scan rate of one degree min -1 . Protein concentrations in the CD experiments were 15 μM in a buffer consisting of 10 mM sodium phosphate and 50 mM NaCl at pH 7.0. Melting temperatures (T M ) were quantified with non-linear fits (Microcal Origin) of CDdata to a two-state transition [36].

Design of Adk variants
To enable a test of yeast growth rates in response to variation in the k cat of Adk we designed a set of mutations that has a logarithmic coverage of this crucial catalytic parameter (Fig 3). Activity was determined at 20°C and since the only variant that displayed reduced yeast growth rates at 30°C (see below) was the essentially catalytically dead Adk eco variant K13Q, 20°C is an appropriate temperature for activity measurements. First we determined k cat and K ATP M of yeast Adk1 and compared these values to published values of Adk eco parameters. It was found that the K ATP M values are very similar for the two enzymes (Table 1 and S1 Fig) but that there is a sizable difference in the k cat values that are of 305±12 s -1 for Adk eco and 520±32 s -1 for Adk1 yeast . Thus replacement of Adk1 for Adk eco in the experimental approach described below will in fact represent a data point where k cat is 59% compared to the yeast wild-type. The remaining variation in k cat relative to yeast Adk1 was accomplished through point mutations and in one case an insertion of 11 amino acids into Adk eco . For clarity, the variants of Adk eco are denoted Adk x eco , where "x" indicates the k cat value relative to that of wild-type E. coli Adk eco (hereafter denoted Adk 1:00 eco ). For instance Adk eco , with a T163C amino acid substitution (for which the k cat is 47% of the Adk 1:00 eco value) is denoted Adk 0:47 eco . The identity of the mutations and the k cat values associated with the Adk eco variants used in this study are summarized in Table 1. Displays of the kinetic traces for the unique variants analyzed in this study are displayed in S1 Fig.

Structure and stability of Adk variants
To interpret the results purely in terms of effects of k cat and K ATP M perturbations it was important to confirm that the tested mutations did not cause potentially confounding effects on the structural integrity and stability of the variants. Therefore, we investigated the structural integrity of all Adk variants used in this study with two-dimensional heteronuclear high-resolution . From a stability perspective it was essential for the melting temperatures (T M ) to be well above the cultivation temperatures to be used in the yeast growth experiment (20°C-30°C) to ensure that the proteins were properly folded under the experimental conditions. Circular dichroism spectroscopy demonstrated that the T M of all variants are well above the temperature interval used in the growth experiments (Fig 5). Thus, all variants fulfilled the structural integrity and thermal stability criteria, and the tested mutational perturbations can be considered as clean variations of k cat and K ATP M that do not affect the global structure and stability of the enzyme.

Analysis of yeast fitness
Here we have developed a yeast cell based approach to address the relevance of the catalytic parameters k cat and K ATP M of adenylate kinase (Adk) for the fitness of yeast expressing Adk variants. Depending on the strain background, adk1Δ yeast strains are either very sick or inviable [23][24][25]. In the present study we first showed that in our yeast strain background a knock-out of the ADK1 gene is lethal and can be rescued by a plasmid-borne wild-type yeast ADK1 (see strains, media and genetic procedures in experimental procedures). To study the cell growth responses to perturbations of adenylate kinase k cat and K ATP M parameters in this system, we expressed specifically mutated variants of Adk eco . The rationale for using Adk eco rather than Adk1 yeast for this purpose was that more information is available about Adk eco variants and their catalytic properties. Thus, genes encoding Adk eco variants with perturbations in both k cat and K ATP M were introduced to yeast with a plasmid shuffling system [22] illustrated and explained in Fig 6. According to a coupled ATPase assay, there were no significant differences in K ATP M values between wild-type Adk eco and Adk1 yeast at 20°C, but the k cat of Adk eco (305±12 s -1 ) is 59% relative to that of Adk1 yeast (520±32 s -1 ) ( Table 1). Despite the differences in catalytic turn-over, the growth rates of adk1Δ yeast cells supplemented with wild-type yeast ADK1 or E. coli adk genes were identical, demonstrating that the E. coli adk gene could functionally exchange the yeast ADK1 gene (Table 2). Initially, we compared the growth properties of yeast cells expressing Adk variants in a serial dilution assay, and detected no growth defects in cells expressing the Adk eco variants at 30°C, the optimal growth temperature for yeast (Fig 7). Only the K13Q were slightly and strongly impaired, respectively (Fig 7). The k cat values of these mutants are 19, 11 and 2 s -1 , respectively. The variant with lowest activity and with a growth that cannot be distinguished to that of yeast expressing Adk1 yeast is Adk 0:12 eco that has a k cat of 36 s -1 . Apparently, the stress imposed by the 10°C reduction of growth temperature results in a significant increase in the k cat value required for optimal growth. It should be noted that the k cat values were measured at 20°C and that some of the effect observed may be due to differences in the k cat values at 20 and 30°C. The temperature dependency of k cat of Adk eco has been quantified previously [21] and there exist a 2 fold difference in activity between 20 and 30°C. Assuming that the temperature dependency is similar for the mutant forms of Adk eco studied here it is likely that at least, a part of the overcapacity of adenylate kinase is required for adaptation of yeast to stress conditions. On the basis of the comparison of growth at 20°C and 30°C, subsequent liquid culture experiments were performed at 20°C.
As indicated above, yeast cell growth is sensitive to the perturbations of the catalytic activity of adenylate kinase when the cells are subjected to stress by growth at the sub-optimal temperature 20°C. To quantify the influence of the quantified values of k cat , K ATP M and the specificity constant k cat /K ATP M , the relative fitness of yeast cells expressing all tested Adk variants were quantified at 20°C in liquid cultures ( Table 2). The relative fitness is defined as the ratio between the growth rate of yeast transformed with an Adk variant divided by the growth rate of yeast transformed with yeast Adk1 [15]. Relative fitness were then plotted against k cat , K ATP M and k cat /K ATP M , to identify whether perturbation of k cat or K ATP M was responsible for the growth  Table 1. impairment observed in the serial dilution assay at 20°C. When the relative fitness was displayed against k cat values ( Fig 8A) the first conclusion was that the fitness was identical for cells expressing either yeast Adk1 or E. coli Adk. Since the k cat value of the E. coli enzyme is 59% (Table 1) relative to the yeast enzyme it is immediately evident that there is substantial overcapacity in the Adk catalytic power in yeast cells. Clear impairment of the relative fitness is only observed when the k cat value is below 60 s -1 and further reduction in k cat results in successive slower growth rates. Overall the shape of the plot resembles a saturation curve and the relative fitness of yeast expressing Adk eco variants displaying as little as 7% catalytic activity compared to wild-type yeast Adk1 (12% of wild type Adk 1:00 eco ) were indistinguishable from those expressing Adk1 yeast . In contrast, there was no obvious correlation between the variants' relative fitness and K ATP M values when K ATP M is varied in the interval 34-411 μM (Fig 8B). Not surprisingly a display of relative fitness against k cat /K ATP M ratios showed a functional dependency that is similar to that observed when k cat was analyzed (Fig 8C). Thus, the lack of correlation for K ATP M is superseded by the correlation to k cat when k cat /K ATP M is displayed. Taken together, it is apparent that The yeast ADK1 open reading frame was exchanged with the KanMX cassette. Viability of the resulting strain depends on the presence of a wild-type yeast ADK1 gene in a low-copy number URA3-based vector, pRS316. A second low-copy number LEU2-based vector was used to introduce different alleles of the E. coli adk gene into this strain. Thus, a strain harboring both the URA3 plasmid (wild-type yeast ADK1 gene) and the LEU2 plasmid (mutated E. coli adk gene) can be obtained. If such a strain is plated on medium containing 5-FOA, the URA3 vector will be counter-selected as the URA3 gene product converts 5-FOA to a toxic compound [22]. Thus, this plasmid shuffling procedure can reveal the phenotype conferred by a mutated E. coli adk gene located in the LEU2 plasmid.
doi:10.1371/journal.pone.0163115.g006 the relative fitness in the experiment is predominantly dependent on the k cat values of the Adk variants encoded by the introduced plasmids. An important aspect for the interpretation of the in vivo experiments are the expression levels of the various Adk eco variants since a variation in these levels potentially can affect the conclusions drawn. Protein expression levels were quantified with western blot analysis using a polyclonal antibody raised against Adk eco (Fig 9A). Increased Adk eco protein levels were observed for variants where k cat is below 60 s -1 (Fig 9B). Thus the expression levels were increased for variants that also displayed a growth rate impairment. Hence, the yeast cells seem to compensate the perturbation to the growth rate by an upregulation of the Adk eco expression levels. These increases in expression levels were presumably due to either k cat -dependent upregulation of protein expression or selection of cells with a higher copy number of plasmids harboring the E. coli adk mutant gene. The differences in expression levels do however not change the main observation that cells do have optimal  growth rates even with severe reduction of k cat . This inference is illustrated with a display of relative fitness against apparent k cat values (k app cat ) (Fig 9C), where k app cat corresponds to k cat normalized with the relative expression levels (

Conclusions
Here we report an approach that enables in vivo analysis of effects of perturbations of adenylate kinase catalytic parameters in yeast cells. In the experiments growth rates of cells were quantified in response to Adk eco variants with mutations that "cleanly" perturb the targeted parameters (k cat and K ATP M ) without compromising the enzyme's structural integrity or thermal stability. First we showed that cytosolic Adk is required for yeast viability in our strain background, and that a catalytically dead but properly folded variant (K13Q) could not restore viability. These results supports that the relative fitness of the yeast strains is depending on the level of Adk activity. The intracellular ATP concentrations in yeast cells are in the low millimolar range [37][38][39]. The Michaelis constant is related to the dissociation constant (K d ) for a given substrate, and . Cells were cultivated for growth rate measurements at 20°C. Catalytic parameters (k cat and K ATP M ) were obtained from a coupled ATPase assay [33]. Error bars for growth rate constants indicate standard deviations obtained from three independent biological replicates. Error bars for Adk catalytic parameters k cat and K ATP M indicate standard deviations of three technical replicates.
doi:10.1371/journal.pone.0163115.g008 Act1p was used as a loading control. Protein levels of the Adk eco variants were normalized with respect to wild-type Adk 1:00 eco protein levels ( it has been shown that the K d of ATP binding to Adk eco is 50 μM [40] which is close to the value of K ATP M determined here (~70 μM). With K ATP M values used as a proxy for ATP binding affinity it is, in fact, expected that all Adk variants used in this study should be saturated with ATP inside yeast cells since the K ATP M values ranging from 34 to 417 μM are below the expected cellular ATP levels. This inference was corroborated with the cell growth experiments since no correlation between relative fitness and K ATP M values was observed (Fig 8B). Thus, from a standpoint of K ATP M , the Adk variants are all fully functional in the cellular milieu and the growth defects observed in the experiments can be attributed to variations in k cat (determined at 20°C). It was found that yeast cell growth at the optimal temperature (30°C) was not affected by the mutations (except for Adk 0:0002 eco with a k cat of 0.06 s -1 that was unable to rescue the inviable phenotype), which is remarkable as the k cat of one of the mutants is only 0.4% compared to wild-type Adk1 yeast . In contrast, in response to external stress by growth at a sub-optimal temperature of 20°C the relative fitness of yeast was impaired when the k cat value was less than 7% of the Adk1 yeast value. Taken together the data show that the k cat value of Adk1 yeast is well above the threshold value required for cell growth under optimal and sub-optimal conditions and that there is substantial overcapacity in the catalytic turn over by Adk in yeast cells. A similar functional dependency of relative fitness with a massive catalytic overcapacity has been observed in E. coli for the enzyme β-galactosidase [15,41]. Thus, both the present study in a eukaryotic organism and the cited βgalactosidase studies in a prokaryotic organism indicate that only a fraction of evolved enzymatic activity may be required for optimal cell growth under laboratory conditions. The data presented here show that, at least a part of the overcapacity in catalytic power is required for organisms to survive external stress conditions that may apply to organisms in their natural habitats. There exist other ways of inflicting stress conditions to yeast in laboratory growth experiments and examples thereof are; oxidative stress, and nutritional stress. For the conceptual discovery here temperature was chosen since it is a parameter that can be accurately controlled and no additional variables such as nutritional uptake or intracellular concentrations (of for instance hydrogen peroxide in oxidative stress experiments) needs to be considered. Additionally, the plasmid shuffling system developed here is a useful platform in order to promote novel discoveries in Adk enzymology. This can in principle be performed by searching for intragenic suppressor mutations that can revert/save an inviable phenotype dependant on mutation of key catalytic residues. Intragenic suppressor mutations may bypass the effect of the mutation leading to inviability and this bypass effect can generate novel information on, for instance, the plasticity of active sites. A second application to the method lies within enzyme design, here it is possible to make use of the fact that yeast can survive with a very low k cat value at 30°C (Fig 7). One useful experiment that contains significant information on design would be to evolve adenylate kinase activity from an unrelated ATP binding enzyme. Random mutation of the gene encoding this scaffold enzyme followed by transformation into the yeast plasmid shuffling system would generate viable colonies only if adenylate kinase activity has evolved.  Table. Yeast strains used in this study. (DOCX)