chain interactions in the regulatory domain of ... - Wiley Online Library

JOURNAL OF NEUROCHEMISTRY

| 2015 | 133 | 73–82

doi: 10.1111/jnc.13019

*Faculty of Medicine, Department of Neurology, School of Health Sciences, University of Crete, Heraklion, Crete, Greece †Department of Physiology and Biophysics, Virginia Commonwealth University, School of Medicine, Richmond, Virginia, USA ‡Department of Biology, University of Crete, Heraklion, Crete, Greece

Abstract Glutamate Dehydrogenase (GDH) is central to the metabolism of glutamate, a major excitatory transmitter in mammalian central nervous system (CNS). hGDH1 is activated by ADP and L-leucine and powerfully inhibited by GTP. Besides this housekeeping hGDH1, duplication led to an hGDH2 isoform that is expressed in the human brain dissociating its function from GTP control. The novel enzyme has reduced basal activity (4–6% of capacity) while remaining remarkably responsive to ADP/L-leucine activation. While the molecular basis of this evolutionary adaptation remains unclear, substitution of Ser for Arg443 in hGDH1 is shown to diminish basal activity (< 2% of capacity) and abrogate L-leucine activation. To explore whether the Arg443Ser mutation disrupts hydrogen bonding between Arg443 and Ser409 of adjacent monomers in the regulatory domain (‘antenna’), we replaced Ser409 by Arg or Asp in hGDH1. The Ser409Arg-1 change essentially replicated the Arg443Ser-1 mutation

effects. Molecular dynamics simulation predicted that Ser409 and Arg443 of neighboring monomers come in close proximity in the open conformation and that introduction of Ser443-1 or Arg409-1 causes them to separate with the swap mutation (Arg409/Ser443) reinstating their proximity. A swapped Ser409Arg/Arg443Ser-1 mutant protein, obtained in recombinant form, regained most of the wild-type hGDH1 properties. Also, when Ser443 was replaced by Arg443 in hGDH2 (as occurs in hGDH1), the Ser443Arg-2 mutant acquired most of the hGDH1 properties. Hence, side-chain interactions between 409 and 443 positions in the ‘antenna’ region of hGDHs are crucial for basal catalytic activity, allosteric regulation, and relative resistance to thermal inactivation. Keywords: allosteric regulation, circular dichroism, glutamate dehydrogenase, molecular dynamics simulation, molecular modeling, site-directed mutagenesis. J. Neurochem. (2015) 133, 73–82.

Glutamate dehydrogenase (GDH) (EC 1.4.1.3.) catalyzes the oxidative deamination of glutamate to a-ketoglutarate and ammonia using NAD and/or NADP as cofactors, thus providing a major pathway for the inter-conversion of aamino acids and a-ketoacids. The enzyme is allosterically regulated with GTP and ADP/L-leucine serving as the main endogenous negative and positive modulators, respectively. Studies on rat brain have shown that GDH is mainly expressed in astrocytes, where it is thought to be involved in the metabolism of transmitter glutamate (Aoki et al. 1987). Consistent with this possibility are observations showing that

Received August 6, 2014; revised manuscript received November 16, 2014; accepted December 15, 2014. Address correspondence and reprint requests to Andreas Plaitakis, Professor of Neurology University of Crete, School of Health Sciences Section of Medicine, Heraklion, Crete, Greece. E-mail: [email protected] Abbreviations used: GDH, glutamate dehydrogenase; HC, hill coefficient; hGDH1, human glutamate dehydrogenase encoded by the GLUD1 gene; hGDH2, human glutamate dehydrogenase encoded by the GLUD2 gene; CD, circular dichroism; SVD, singular value decomposition; SC50, Stimulatory Concentration of the agonist giving 50% of maximal stimulation; IC50, half maximal inhibitory concentration; AAT, aspartate amino transferase.

© 2014 International Society for Neurochemistry, J. Neurochem. (2015) 133, 73--82

73

74

V. Mastorodemos et al.

GDH expression increases markedly during postnatal brain development in tandem with terminal growth and synaptogenesis (Rothe et al. 1990). In addition, there is evidence that the GDH flux is activated under conditions of excitatory transmission (McKenna et al.1996). In humans, the enzyme exists in hGDH1 and hGDH2 isoforms that differ in their optimal pH, allosteric regulation by GTP and in their thermal stability, with hGDH1 being heat-stable and hGDH2-heatlabile (Plaitakis et al. 2011). While hGDH1 is encoded by the intron-containing GLUD1 gene (common to all mammals) that is expressed widely, hGDH2 is encoded by the intronless GLUD2 gene expressed in neural, testicular, and renal tissues (Shashidharan et al. 1994). The GLUD2 gene has arisen through a duplication event < 23 million years ago and evolved under positive selection on the line that descended to the human (Burki and Kaessmann 2004). There is evidence that GLUD2 provided a biological advantage to humans, but the reason(s) for this have not been understood. Whereas hGDH1 functions optimally at relatively high pH (7.75–8.0), hGDH2 is also active at lower pH values (7.25–7.75) (Kanavouras et al. 2007). This adaptation may allow efficient hGDH2 function under conditions of intracellular acidification that prevail in astrocytes upon glutamate transport (Azarias et al. 2011). Also, resistance of hGDH2 to GTP control may enable enzyme catalysis under conditions inhibitory to hGDH1 (Plaitakis et al. 2011). However, to prevent unregulated activity from perturbing cell metabolism, hGDH2 adapted by lowering its basal catalytic activity (4–6% of its maximal), while remaining remarkably responsive to activation by ADP and L-leucine. The need for setting hGDH2 activity at relatively low levels is underscored by observations showing that a gain-of-function hGDH2 variant (displaying increased basal activity) accelerates the commencement of Parkinson’s disease (Plaitakis et al. 2010). While resistance to GTP inhibition is shown to result from the Gly456Ala evolutionary change (Zaganas and Plaitakis 2002), the molecular mechanisms by which hGDH2 downregulates its basal activity and remains responsive to activators has not been fully understood. In this regard, we have shown that evolutionary substitution of Ser for Arg443 in hGDH1 diminished basal activity and made the enzyme markedly heat-labile (Zaganas and Plaitakis 2002; Kanavouras et al. 2007). However, the Ser443-hGDH1 mutation abrogated L-leucine activation and decreased ADP affinity. Modeling of the Arg443Ser change revealed that the side chain of Arg443 of one GDH subunit can form one or two H-bonds with Ser409 in the ascending strand of the ‘antenna’ of the adjacent subunit (Zaganas and Plaitakis 2002). We hypothesized that loss of these subunit interactions may lead to a closed state associated with diminished basal activity and abrogation of L-leucine activation. To test this possibility, we replaced here Ser409 in hGDH1 by Arg (S409R-1) or Asp (S409D-1) and studied the obtained mutant enzymes purified

to homogeneity. In addition, to test the functional importance of the Arg443Ser evolutionary change in hGDH2, we replaced Ser443-2 by Arg (as this occurs in hGDH1) and studied the mutant hGDH2 enzyme. Results provided important evidence that side-chain interactions between Arg443 and Ser409 of adjacent subunits in the antenna are essential for basal catalytic function, regulation, and relative resistance to thermal denaturation. These are reported below.

Experimental procedures Materials Sf21 cells were obtained from Invitrogen (Carlsbad, CA, USA). The media for the Sf21 insect cells and fetal calf serum were from Life Technologies. Modified baculovirus (BaculoGold) was obtained from BD Pharmigen (San Diego, CA, USA). Ligation to pVL1393 expression vector was performed with the Fast-LinkTM DNA Ligation Kit (Epicentre, Madison, Wisconsin). NADPH, ADP, GTP (lithium salt) were from Roche Molecular Biochemicals (Mannheim, Germany). Phenylsepharose high performance was from Amersham Biosciences (Uppsala, Sweden) and Hydroxyapatite Bio-Gel HT was from Bio-Rad (Hercules, CA, USA). Gel isolation of PCR products was performed with the QIAEX II Gel Extraction Kit (Qiagen Inc., Valencia, CA, USA). Site-directed mutagenesis To obtain the S409R-1 and S409D-1 single mutants, the GLUD1 cDNA (cloned in the pBSKII+ vector) was mutagenized at residue 409 as previously described (Zaganas and Plaitakis 2002). For creating the S409R/R443S-1 (‘Swap’) mutant we used the R443S-1 cDNA as template. To obtain the S443R-2 hGDH2 mutant (has Arg at residue 443 as the wild-type hGDH1 does), we used the wild-type GLUD2 cDNA as template to replace Ser443 by Arg. Following annealing of the mutagenic primers, the plasmids were amplified by T4 DNA polymerase (nicks were ligated by T4 DNA ligase) and used to transform the BMH 71-18 mutS strain of Escherichia coli. The mutated GLUD1 cDNAs were cleaved from the pBSKII+ vector, ligated to the baculovirus transfer vectors pVL1393 or pVL1392 and used to transform the JM109 strain of E.Coli. The subcloned mutated GLUD1 or GLUD2 cDNA was bi-directionally sequenced in its entire length to confirm the presence of the desired mutation and exclude incidental DNA alterations. Expression and Purification of wild type and mutant proteins Wild type and mutant cDNAs were expressed in Sf21 cells as previously described (Shashidharan et al. 1994).The cultured cells were harvested 5 days post infection. GDH was purified from Sf21 cell extracts using a combination of ammonium sulfate fractionation and hydrophobic interaction, and


Antenna-mediated GDH properties

hydroxyapatite chromatography (Kanavouras et al. 2007). Fractions eluted from the hydroxyapatite column containing GDH activity were pooled and used for enzyme assays and for sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). Wild type and mutated GDH proteins were produced and studied in parallel. Enzyme assays and allosteric regulation studies GDH activity was measured spectrophotometrically (at 340 nm) at 250 C in 50 mM Triethanolamine pH 8.0 buffer, 2.6 mM EDTA, 100 mM Ammonium Acetate, 100 lM NADPH, and 8 mM a-ketoglutarate (Zaganas and Plaitakis 2002). Enzyme velocity for the first 30 s after initiation of the reaction was recorded. Regulation of the human recombinant wild type and mutant GDHs by allosteric effectors was studied by adding these to the reaction mixture at various concentrations while keeping the other substrates constant. ADP was varied from 0.025 to 1.0 mM, L-leucine from 0.15 to 10 mM, and GTP from 0.05 to 200 lΜ (final concentrations). To test the synergistic effect of the two activators, we added L-leucine (0.15–10.0 mM final concentrations) to the reaction mixture containing 0.025 mM or 0.05 mM ADP. Kinetic analyses were performed in the presence of 1 mM ADP to determine the Michaelis–Menten constant (Km) for a-ketoglutarate, ammonia, and NADPH. aketoglutarate varied from 0.4 to 10 mM, ammonium acetate from 10 to 100 mM, and NADPH from 10 to 160 lM. Heat inactivation studies For studying heat inactivation, samples containing 40–60 lg/ mL purified enzyme and 4 mg/ml bovine serum albumin were prepared in 100 mM sodium phosphate buffer, pH 6.8. The samples were incubated in a shaking water bath at 47.5°C. Aliquots (10 lL) were removed at specified time intervals and assayed immediately in the direction of reductive amination of a-ketoglutarate in the presence of 1 mM ADP as described above. Molecular dynamic simulation studies The open and closed conformation crystal structure of GDH from the protein data bank (PDB) (Banerjee et al. 2003) was used for Molecular Dynamics (MD) simulation studies. The computational experiments were performed with CHARMM (Brooks et al. 2009) using implicit membrane by means of a Generalized Born model with switching function known as GBSW (Im et al. 2003a,b). GDH is a hexamer composed of two trimers. These were first minimized and further used for MD simulations, which were carried out for 39 ns and included heating (0–300K in increments of 50K), equilibration (3.5 ns), and production runs (35 ns). The production runs were further used for structural analysis. In addition, the S409R-1; R443S-1; S409D-1; S409R/R443S-1 (swap) mutants of hGDH1 were generated from the native crystal form and further minimized and simulated for the open

75

conformation of the structure. Interactions between Ser409 and Arg443 of adjacent monomers were analyzed. Circular dichroism Circular dichroism (CD) spectra were acquired using a J-815 CD spectropolarimeter with a 1-mm path length quartz cuvette, at protein concentrations of 0.18 mg/mL for hGDH1, hGDH2, and S409D-1, and 0.07 mg/mL for S409R-1, in 10 mM sodium sulfate. Far-UV spectra were measured with 50 nm/min scanning speed, 1 min response time, and three accumulations. Thermal denaturation data were collected monitoring the CD signal at 222 nm in the range 10–70°C with a temperature increase of 80°C/h and a waiting time of 3 s for stabilization. Full far-UV CD spectra (260–195 nm) were also recorded in the range 10–70°C in steps of 5°C for hGDH1, hGDH2, and S409D-1 and 2°C in the case of S409R-1. Singular value decomposition of the thermal denaturation spectra was performed to determine significant independent states in the unfolding transition (Amprazi et al. 2014). Structural modeling Homology based 3D-molecular modeling of wild-type hGDH-1 and its mutants was performed using the atomic coordinates of apo Human GDH (PDB 1LIF) (Smith et al. 2002) in the PyMOL Molecular Graphics System, Version 1.4 (Schr€ odinger, LLC, New York, NY, USA). Studies of the structural models of bovine GDH were performed with the use of the RasMol (version 2.7.1.1) (Herbert J. Bernstein, Bernstein + Sons, Bellport, NY, USA), the Swiss-PDBviewer (version 3.7.b2) (SIB Swiss Institute of Bioinformatics. Lausanne, Switzerland), and the RIBBONS programs, as previously described (Zaganas et al. 2002). Statistical analyses Differences in kinetic and allosteric behavior were evaluated using Student’s t-test. IC50 and SC50 values were determined graphically. The Hill plot coefficients for GTP inhibition were calculated according to the method discussed by Cornish-Bowden (1979).

Results Production of wild type and mutant hGDHs Expression of the wild type and mutant (S409R-1; S409D-1; R443S-1; S409R/R443S-1; S443R-2) cDNAs in Sf21 cells produced catalytically active enzymes when assayed at 1.0 mM ADP. As the endogenous GDH of the Sf21 cells is NAD(H) specific (Shashidharan et al. 1994), all assays were carried out in the presence of NADP(H), thus eliminating all background activity. SDS-PAGE analysis of purified fractions revealed that the recombinant enzymes were more than 95% pure, forming a single band (~ 56– 58 KDa) (Figure S1). Also, as described earlier (Zaganas and


76


Plaitakis 2002), the wild-type hGDH2 migrated on SDSPAGE slightly higher than hGDH1 owing to evolutionary substitution of Ser for Arg443. Indeed, all mutants containing the R443S change migrated on SDS-PAGE to the same level as the wild-type hGDH2 (Figure S1). Characterization of the S409R-1 mutant Introduction of Arg in place of Ser409 in hGDH1 diminished basal activity (< 2% of maximal) and abrogated activation by L-leucine in the absence of ADP (Figs. 1 and 2). These results are similar to those previously obtained for the R443S-1 hGDH1 mutant (Zaganas et al. 2002). The S409R1 mutant was fully activated by 1.0 mM ADP (Fig. 3). However, the ADP SC50 ( SEΜ) for the S409R-1 mutant (110.3 7.6 lM) was higher than that for the wild-type hGDH1 (SC50 = 20.5 2.4 lM), but closer to that of the wild-type hGDH2 (78.3 7.9 lM) (Fig. 3.) Whereas the S409R-1 mutation essentially abrogated L-leucine activation (Fig. 2), low concentrations of ADP (10–50 lM) permitted activation of the S409R-1 mutant by L-leucine, as previously reported for the R443S-1 mutant (Zaganas et al. 2002). Thus, at 25 lM ADP, L-leucine (9.0 mM) enhanced the activity of the S409R-1 mutant by 1519%. This activation is proportionally higher than that of the wild-type hGDH1 (by 127%), being close to that of the wild-type hGDH2 (by 1043%) (Table S1). Addition of increasing concentrations of ADP, reduced the estimated L-leucine SC50 for the S409R-1 mutant as previously reported for the wild-type hGDH1 and hGDH2 (Kanavouras et al. 2007). Compared to the wildtype hGDH1, the S409R-1 mutant was more sensitive to GTP inhibition, with its inhibitory curve lacking the characteristic sigmoidal pattern of the wild-type hGDH1 (Fig. 4). Hill plot analyses confirmed that GTP binding to the S409R-1 mutant was not co-operative (hill coefficient = 1.0– 1.35). This is similar to that of the wild-type hGDH2 and contrasts the co-operative behavior of the wild-type hGDH1 (hill coefficient = 1.5–2.5). Substitution of Ser409 for Arg in hGDH1 made the mutant enzyme more sensitive to thermal inactivation at 47.5 C (t = 262 min) than the wild-type hGDH1 (t = 434 min) (Fig. 5), but this effect was less dramatic than that induced by the R443S-1 mutation (Fig. 5). Molecular dynamics simulation In the native hGDH1 structure, hydrogen bonding interaction between the O–H group of Ser409 and Arg443 side-chain N– H was found to exist for over 60% of the production run. Hydrogen bonds between Ser409 and Arg443 were seen between adjacent monomers (A, B and C) of the trimetric structure. Study of inter-monomer distances in the native structure revealed that residues 409 and 443 come closer in the open than in the closed conformation (Table S2). Simulation studies were also performed on hGDH1 mutants, involving the critical residues Ser409 and Arg443. Specifically, four mutant structures (S409R-1, R443S-1, S409D-1,

and ‘swap’ S409R/R443S-1) of hGDH1 were simulated. Results of these MD simulations predicted that Ser409 and Arg443 of neighboring monomers in the native form come in close proximity in the open conformation and that introduction of either Ser443-1 or Arg409-1 mutation made these residues (409 and 443) to move away from each other while their proximity was largely restored in the swap mutant (S409R/R443S-1) (Table 1). Characterization of the ‘swap’ S409R/R443S-1 mutant As described in the Methods, a R409/S443-1 ‘swap’ was created by introducing the S409R-1 change to the R443S-1 mutant. While each of these two single amino acid replacements made the enzyme essentially inactive (catalytic activity ~ 2% of maximal), the ‘swap’ double mutant (S409R/R443S-1) regained most of the properties of the wild-type hGDH1 (Figs. 1–5). Hence, when Ser409 and Arg443 were mutually exchanged, the double mutant essentially reverted to the wild-type hGDH1 concerning its basal activity and L-leucine activation (Figs. 1–2). Also, unlike the R443S-1 single mutant that was rapidly denatured by heat, the ‘swap’ structure acquired the heat stability of the wild-type hGDH1 (Fig. 5). The ADP SC50 for the ‘swap’ (S409R/R443S-1) mutant (90.4 5.4 lM) was close to that of the wild-type hGDH2 (78.3 7.9 lM) (Fig. 3), being substantially lower than that of the R443S-1 mutant (405.6 14.8 lM) (Fig. 3). However, the S409R/R443S-1 (‘swap’) mutant was relatively resistant to GTP inhibition in the presence of ADP (Fig. 4). Characterization of the S409D-1 mutant Replacement of Ser409-1 by Asp decreased basal activity to 10–15% of maximal (Fig. 6). While this is higher than that of the S409R-1 mutant, it is substantially lower than that of the wild-type hGDH1 (35–40% of maximal). In contrast, to S409R-1 mutant, the S409D-1 mutant was robustly activated by L-leucine in the absence of ADP (Fig. 6). Addition of ADP (up to 1.0 mM final concentration) resulted in full activation of the S409D-1 hGDH1 mutant, with the ADP SC50 (16.2 1.9 lM) being similar to that of the wild-type hGDH1 (20.5 4.1 lM) (Fig. 6). However, the S409D-1 hGDH1 mutant proved resistant to GTP inhibition as compared to the wild-type hGDH1 or hGDH2 (Fig. 6). Also, replacement of Asp by Ser409-1 rendered the mutant enzyme sensitive to thermal inactivation (half-life at 47.5°C: 13.2 min) (Fig. 6). Dependence of basal activity on protein concentration As previous studies revealed that the basal activity of the wild-type hGDH2 is concentration-dependent (Kanavouras et al. 2007), we performed here enzyme assays by adding increasing enzyme amounts to the reaction mixture (0.4– 20 lg protein/mL). Results revealed that the basal activity of the S409R-1 and the S409D-1 hGDH1 mutants correlated significantly with enzyme protein concentration (Table S3);



77

Fig. 1 Structural models and functional consequences of amino acid replacements at residues 409 and 443. All cartoon diagrams were created €dinger, using the Pymol Molecular Graphics System, Version 1.4 (Schro LLC) using the apo form of human Glutamate Dehydrogenase (hGDH)1 (PDB entry 1LIF; Smith et al. 2002) as a template. For simplicity, only parts of two of the six subunits that compose the GDH hexamer are depicted. Wild-type hGDH1: Ser409 and Arg443; R443S-1 mutant: Ser409 and Ser443; S409R-1 mutant: Arg409 and Arg443; S409R/ R443S-1 (‘Swap’) mutant: Arg409 and Ser443. The histogram next to each cartoon shows the enzyme’s activities, measured as basal: (no

allosteric effectors), L-leucine: activated by 4.5 mM L-leucine and ADP: activated by 1 mM ADP. Enzyme activity was measured (at 3.0–5.0 lg of enzyme protein/1 ml reaction buffer) in the direction of reductive amination of a-ketoglutarate in TRA buffer, pH 8.0 and is expressed as percentage of maximal activity (determined at 1.0 mM ADP). This maximal activity corresponds to the following specific activities (lmoles of NADPH oxidized/mg protein/minute): WT-hGDH1 = 197.02, S409R1 = 185.11, R443S-1 = 111.19, S409R/R443S-1 (‘swap’) = 165.71, WT-hGDH2 = 130.59. Data (columns) represent mean values of at least three experimental determinations and bars correspond to the SE.

however, the slope of the regression line for S409D-1 mutant was significantly steeper than that of S409R-1 mutant or of the wild-type hGDH2. On the other hand the wild-type hGDH1 and the ‘swap’ S409R/R443S-1 double mutant showed a negative correlation between enzyme concentration and activity (Table S3).

studied was consistent with a two-state transition (Holtzer and Holtzer 1995) between the native and the unfolded state (Figure S2B). The singular value decomposition analysis (Amprazi et al. 2014) of the far-UV CD spectra confirms this transition, as it produces only two significant species of linearly independent CD spectra (data not shown). The double-wavelength plot (Holtzer and Holtzer 1995) [h]200 versus [h]222 of molar residual ellipticities at 200 nm and at 222 nm (Figure S2C) reflects a propensity for structural disorder above 50°C, with characteristics of molten globule/ pre-molten globule states (Uversky 2003). The native statelike characteristics and the higher stability of hGDH1 are also reflected in this plot. The folding states adopted by the S409R-1 mutant closely resemble those of hGDH1.

Kinetic analyses Kinetic analyses (obtained at 1.0 mM ADP) revealed that the Km for a-ketoglutarate, ammonia or NADPH for the S409R1 and S409D-1 hGDH1 mutants were comparable to those for the wild-type hGDH1 and hGDH2 (data not shown). Circular dichroism analysis The secondary structure content of hGDH1, hGDH2, S409R1, and S409D-1 was estimated from the far-UV CD spectra and found to be consistent with the crystal structure of hGDH1. The structural stability of these proteins was characterized by monitoring the temperature dependence of the CD signal at 222 nm (Figure S2A) upon thermal unfolding. The 1st derivative of the thermal unfolding data revealed a transition temperature of 53°C for the wild-type hGDH1, 49°C for S409R-1, 48°C for wild-type hGDH2, and 41°C for S409D-1. The 203 nm isodichroic point of the CD spectra collected at different temperatures for all proteins

The wild-type hGDH2 essentially reverts to wild-type hGDH1 via a single amino acid substitution To further test the role of the evolutionary substitution of Arg443 by Ser in the functional properties of hGDH2, we replaced Ser443 by Arg in hGDH2 (as is in its ancestor hGDH1). The obtained S443R-2 mutant displayed under baseline conditions much greater catalytic activity than the wild-type hGDH2, exceeding even that of wild-type hGDH1 (Fig. 7). Also, the S443R-2 mutant became similar to the wildtype hGDH1 with respect to its activation by ADP and


78


Fig. 2 L-leucine activation curves of mutant and wild-type glutamate dehydrogenase (hGDH)s. Highly purified preparations of the S409R-1 and S443R-1 single mutants and of the S409R/R443S-1 double mutant (‘Swap’) were studied. GDH activity was measured in the direction of reductive amination of a-ketoglutarate in TRA buffer, pH 8.0, in the presence of increasing concentrations of L-leucine (in the absence of ADP) as described in Methods. These are representative activation curves deriving from at least three independent experiments. Data points are expressed as percentage of maximal activity, which was obtained at 1.0 mM ADP and bars correspond to the SE.

L-leucine and its relative resistance to thermal inactivation (Figure S3A and C). Thus, through this single amino replacement, the wild-type hGDH2 essentially reverted to the wild-type hGDH1; however, resistance to GTP inhibition was retained, with the S443R-2 mutant displaying a GTP IC50 similar to that of the parental wild-type hGDH2 (Figure S3B). These results are consistent with previous findings showing that the evolutionary substitution of Ala for Gly456 made the wild-type hGDH2 resistant to GTP (Zaganas and Plaitakis 2002).

Discussion GDH is a mitochondrial enzyme that is mainly expressed in CNS astrocytes (Aoki et al. 1987; Rothe et al. 1995), the cells responsible for clearing and metabolizing > 80% of glutamate released from glutamatergic terminals during excitatory transmission. Glutamate taken up by astrocytes is in part transported into the mitochondria where it is converted (via GDH or aspartate amino transferase 2) to aketoglutarate, which in turn can enter the Krebs cycle leading to synthesis of ATP. Although GDH can attain high levels in astrocytic mitochondria (up to 10 mg/ml mitochondrial matrix) (Rothe et al. 1995), the enzyme is potently inhibited by GTP (IC50 = 0.1–0.3 lM) generated by the Krebs cycle. This may prevent glutamate from fuelling this cycle under conditions of adequate cellular energy charge (Plaitakis et al. 2011). On the other hand, dissociation of hGDH2 from GTP

Fig. 3 ADP activation curves of purified mutant and the wild-type glutamate dehydrogenase (hGDH)s. GDH activity was measured TRA buffer, pH 8.0, in the presence of increasing concentrations of ADP as described in the Methods. Data points are mean values from at least three experimental determinations and are expressed as percentage of maximal activity. Error bars correspond to the SE. The activator SC50 SE (ADP concentration that elicits 50% of the maximal activation) was calculated from the corresponding activation curves and are shown below for each expressed enzyme: SC50 = 20.5 2.4 lM for the wildtype hGDH1; 78.3 7.9 lM for the wild-type hGDH2; 110.4 7.6 lM for S409R-1 and 90.4 5.4 lM for the double S409R/R443S-1 mutant (‘swap’). The ADP SC50 for the R443S-1 is 405.6 14.8.

control, a major adaptive evolution for the novel enzyme, may permit glutamate flux even when the Krebs cycle generates GTP levels sufficient to completely inactivate hGDH1. However, as noted above, hGDH2 adapted by reducing its basal activity, probably to prevent unregulated enzyme function from perturbing cell metabolism. To better understand the molecular mechanisms by which hGDH2 down-regulates its basal activity, we mutagenized hGDH1 at position 409, a residue thought to interact with Arg443, and found that replacement of Ser409 by Arg diminished basal catalytic activity and abrogated L-leucine activation (probably by preventing L-leucine from entering the active site). These functional consequences are similar to those of the R443S-1 evolutionary substitution. As such, present observations are consistent with the thesis that mutations of Arg443 and Ser409 disrupt hydrogen bonds between adjacent subunits in the antenna leading to closed conformation. This possibility was further explored by Molecular Dynamics simulation of the native structure, which predicted that residues 409 and 443 come closer in the open than in the closed conformation and that the S409R-1 and R443S-1 replacements caused neighboring monomers to move away from each other. Importantly, the proximity of these monomers was largely reinstated in the swap mutant (S409R/R443S-1), a prediction substantiated by the study of the swap mutant protein, which revealed that the recombinant



79

Table 1 Inter-monomer distances between residues 409 and 443 for the native and mutant (S409R-, S409D, R433S, and the S409R/R433S swap) conformers

Structure – ID

Monomers involveda

Shortest distance after MD (35 ns) in Ab

1nr1 (native form)

Ser409(A)-Arg443(C) Ser409(B)-Arg443(A) Ser409(C)-Arg443(B) Arg409(A)-Arg443(C) Arg409(B)-Arg443(A) Arg409(C)-Arg443(B) Asp409(A)-Arg443(C) Asp409(B)-Arg443(A) Asp409(C)-Arg443(B) Ser409(A)-Ser443(C) Ser409(B)-Ser443(A) Ser409(C)-Ser443(B) Arg409(A)-Ser443(C) Arg409(B)-Ser443(A) Arg409(C)-Ser443(B)

4.6 5.9 5.4 8.7 8.4 13.4 7.8 4.1 15.1 8.1 6.1 12.6 5.3 6.0 6.3

S409R-1 mutant

S409D-1 mutant

Fig. 4 GTP inhibition curves of purified mutant and wild-type hGDHs. The GTP inhibition curve of the S409R-1 single hGDH1 mutant and the S409R/R443S-1 double mutant (‘swap’) were studied along with the wild-type hGDH1 and hGDH2. GDH activity was measured in the presence of increasing concentrations of GTP in the presence of 1.0 mM ADP. Values represent percentage of baseline activity (no GTP added) and are means of at least two experimental determinations. Error bars correspond to the SE. The initial specific activities (100%) of these iso-enzymes are described in the Legend of Figure 1.

Fig. 5 Thermal stability of purified mutant and wild-type hGDHs. About 40–60 lg/mL of purified S409R-1 and R443S-1 single mutants, of S409R/R443S-1 (‘swap’) double mutant and of the type hGDH1 and hGDH2 were mixed 1 : 1 (vol/vol) with bovine serum albumin (4 mg/ mL) in 100 mM sodium phosphate pH 6.8 buffer. The enzyme mixture was incubated at 47.5°C in polypropylene tubes. Aliquots were removed at the intervals specified and assayed in the direction of reductive amination of a-ketoglutarate in TRA pH 8.0 buffer in the presence of 1.0 mM ADP. Data are mean SE (bars) values of three determinations and represent percentage of baseline activity obtained at 0 time.

R443S-1 mutant

S409R/R443S-1 (swap mutant)

a

The letters in parenthesis indicate the monomer that the residue belongs. b Shortest distance is picked from the array of distances between every atom of Ser 409 to every atom of Arg 443. For MD simulations, we used the crystal native form to model the open (1nr1) conformation; the mutant conformers were also modeled from 1nr1 as described in the Methods section. The structures obtained after 35 ns revealed the shortest distances between Ser 409 and Arg 443 for the three monomers.

protein regained most of the functional properties of the wild-type hGDH1. As the present results showed, replacement of Ser409 by Arg diminished basal activity and abrogated L-leucine activation without substantially affecting thermal stability. On the other hand, substitution of Asp for Ser409 had lesser effects on basal activity and permitted activation by Lleucine while rendering the enzyme markedly heat-labile. Whereas electrostatic repulsion between Arg443 and the introduced Arg409 in the antenna is expected to be disruptive, introduction of Asp409, which places a negative charge on N-terminus of the ascending a-helix, could stabilize the helix through favorable dipole moment interaction on top of potential Asp409-Arg443 electrostatic interactions. This could account for the observed differences in the basal activity of the S409R-1 and S409D-1 mutants. Regarding the observed marked differences in thermal inactivation, CD analysis revealed the gradual loss of the secondary structure of S409D-1 upon thermal unfolding. In this regard, repulsive electrostatic interactions between Asp409 and Glu406 and/or Glu439 may render the S409D1 mutant markedly heat-labile. On the other hand, our model


80


(a)

(b)

(c)

(d)

Fig. 6 Functional properties of the S409D-1 mutant. GDH activity was measured in cell extracts in the direction of reductive amination of aketoglutarate in TRA buffer, pH 8.0 as described in the Methods. For L-leucine activation (a) GDH activity was measured in the presence of increasing concentrations of L-leucine. For GTP inhibition (b) GDH activity was measured at increasing concentrations of GTP in the presence of 1.0 mM ADP, with values represent percentage of baseline activity (no GTP added). ADP activation (c) was studied in the presence of increasing concentrations of ADP. Thermal stability (d) was evaluated as described in the Legend of Fig. 5. Data points are mean values from at least two experimental determinations and are expressed as percentage of maximal activity (determined at 1.0 mM ADP) and bars correspond to the SE. The maximal activity of the S409D-1 mutant corresponds to a specific activity of 147.93 lmoles of NADPH oxidized/mg protein/minute.

Fig. 7 Basal activity of the S443R-2 mutant (reverse mutant of hGDH2). GDH activity was measured in cell extracts in the direction of reductive amination of a-ketoglutarate in TRA buffer, pH 8.0 as described in the Methods. Basal activities are expressed as percentage of maximal activity (determined at 1.0 mM ADP). Columns are mean values from three determinations and bars correspond to the SE.

suggests that the observed heat stability of the S409R-1 mutant could be due to new attractive electrostatic interactions between the introduced Arg409 residue and Glu439 and/or Glu406, as these side chains are positioned favorably for potential formation of inter-subunit salt bridges. We also observed that the S409R-1 mutant was markedly sensitive to GTP inhibition. With this regard, it is likely that the closed conformation induced by the S409R change facilitates GTP binding by opening the space in the back of the NAD(P) domain (where GTP binds). However, previous observations showed the R443S-1 hGDH1 mutation that also promotes the closed state does not alter GTP inhibition (Zaganas and Plaitakis 2002). Because the R443S-1 mutant required substantially higher concentrations of ADP for its activation (SC50 = 474 lM) than the S409R-1 mutant (SC50 = 110 lM) and because the GTP inhibition of these mutants could only be studied in the presence of ADP, the obtained results may reflect the opposing interactions of ADP and GTP with these enzymes. Residue 443 is located in a short helix of the descending strand of the antenna that undergoes marked conformational



changes during catalysis (Smith et al. 2002). As such, substitution of this residue may alter enzyme function by affecting these dynamics. A similar mechanism has been previously suggested for hGDH1 mutations that cause the hyperinsulinism/hyperammonemia (HI/HA) syndrome (Fang et al. 2002). Specifically, HI/HA mutations Phe440Leu, Gln441Arg, Ser445Leu, and Gly446Ser/Arg/Asp/Cys, located in the short helix of the antenna (flanking residue 443), are thought to affect enzyme activity and regulation by interfering with the ability of the short helix to act as molecular spring upon opening and closing of the catalytic cleft (Smith et al. 2002). Two other HI/HA mutations (Asn410Thr/Tyr, Leu413Val) that lie in the ascending strand of the antenna (adjacent to the 409 residue, substitution of which was shown here to affect basal activity and regulation) may interfere with antenna-mediated inter-subunit communication, a process of importance for GDH allostery (Smith et al. 2002). Our data also showed that the specific activity of the S409R-1 and S409D-1 mutants was dependent on the concentration of the enzyme protein used during assay. Similar findings have been previously shown for the R443S-1 mutant and for the wild-type hGDH2 (Kanavouras et al. 2007). Increasing GDH protein concentration is expected to stabilize the enzyme and this may be physiologically relevant for wild-type hGDH2, which may attain high levels in the mitochondrial matrix as noted above (Rothe et al. 1995). In contrast, the ‘swap’ double mutation studied here, which restored basal activity and regulation, showed a negative correlation between enzyme concentration and specific activity as is also the case for the wild-type hGDH1. As such, these data argue that interaction between 409 and 443 in the antenna is involved in the observed protein dependency. That the Arg443-Ser409 bonding in the antenna of hGDH1 is essential for maintaining the substantial basal activity of this housekeeping enzyme and that substitution of Ser for Arg443 was the single evolutionary change that allowed hGDH2 to set its activity at low levels, was confirmed by changing Ser443-hGDH2 back to Arg443. Via this change (S443R-2), hGDH2 essentially reverted to the wild-type hGDH1. The only exception was that the S443R-2 mutant maintained the resistance of hGDH2 to GTP inhibition. This, however, was expected, as hGDH2 dissociated its function from GTP control via the Gly456Ala evolutionary change. Hence, these data underscore the importance of side-chain interaction network (between the ascending and the descending helices of adjacent ‘antennae’) for the catalytic function, regulation, and thermal stability of human GDHs.

Acknowledgments and conflict of interest disclosure This work was supported by the European Union (European Social Fund – ESF) and Greek national funds through the Operational

81

Program ‘Education and Lifelong Learning’ of the National Strategic Reference Framework (NSRF) – Research Funding Program: THALIS – UOC, Title ‘Mitochondrial dysfunction in neurodegenerative diseases’ (Grant Code 377226). We are indebted to Nikolas Borompokas for his help in the enzyme’s purification. The authors have no conflict of interest to declare. All experiments were conducted in compliance with the ARRIVE guidelines.

Supporting information Additional supporting information may be found in the online version of this article at the publisher's web-site: Figure S1. 8.5% SDS-PAGE analysis of purified human GDHs. Figure S2. Thermal unfolding of wild type and mutant human GDHs monitored by CD. Figure S3. Functional properties of the S443R-2 mutant (reverse mutant of hGDH2). Table S1. Effect of low concentrations ADP on Leucine activation of mutants and wild-type hGDHs. Table S2. Inter-monomer distances between residues 409 and 443 for the open and closed conformation from MD simulations. Table S3. Regression analysis of enzyme concentration vs basal specific activities of mutant and wild-type hGDHs.

References Amprazi M., Kotsifaki D., Providaki M., Kapetaniou E. G., Fellas G., Kyriazidis I., Perez J. and Kokkinidis M. (2014) Structural plasticity of 4-a-helical bundles exemplified by the puzzle-like molecular assembly of the Rop protein. PNAS 111, 11049–11054. Aoki C., Milner T. A., Berger S. B., Sheu K. F., Blass J. P. and Pickel V. M. (1987) Glial glutamate dehydrogenase: ultrastructural localization and regional distribution in relation to the mitochondrial enzyme, cytochrome oxidase. J. Neurosci. Res. 18, 305–318. Azarias G., Perreten H., Lengacher S., Poburko D., Demaurex N., Magistretti P. J. and Chatton J. Y. (2011) Glutamate transport decreases mitochondrial pH and modulates oxidative metabolism in astrocytes. J. Neurosci. 31, 3550–3559. Banerjee S., Schmidt T., Fang J., Stanley C. A. and Smith T. J. (2003) Structural studies on ADP activation of mammalian glutamate dehydrogenase and the evolution of regulation. Biochemistry 42, 3446–3456. Brooks B. R., Brooks C. L., 3rd, Mackerell A. D., Jr et al. (2009) CHARMM: the biomolecular simulation program. J. Comput. Chem. 30, 1545–1614. Burki F. and Kaessmann H. (2004) Birth and adaptive evolution of a hominoid gene that supports high neurotransmitter flux. Nat. Genet. 36, 1061–1063. Cornish-Bowden A. (1979) Fundamentals of Enzyme Kinetics, pp. 147– 176. Butterworth, London Ltd, UK. Fang J., Hsu B., MacMullen C., Poncz M., Smith T. and Stanley C. (2002) Expression, purification and characterization of human glutamate dehydrogenase (GDH) allosteric regulatory mutations. Biochem. J. 363, 81–87. Holtzer M. E. and Holtzer A. (1995) The use of spectral decomposition via the convex constraint algorithm in interpreting the CD-observed unfolding transitions of coiled coils. Biopolymers 36, 365–379. Im W., Feig M. and Brooks C. L. (2003a) An implicit membrane generalized born theory for the study of structure, stability, and interactions of membrane proteins. Biophys. J . 85, 2900–2918.


82


Im W., Lee M. S. and Brooks C. L. (2003b) Generalized born model with a simple smoothing function. J. Comp. Chem. 24, 1691–1702. Kanavouras K., Mastorodemos V., Borompokas N., Spanaki C. and Plaitakis A. (2007) Properties and molecular evolution of human GLUD2 (neural and testicular tissue-specific) glutamate dehydrogenase. J. Neurosci. Res. 85, 3398–3406. McKenna M. C., Sonnewald U., Huang X., Stevenson J. and Zielke H. R. (1996) Exogenous concentration regulates the metabolic fate of glutamate in astrocytes. J. Neurochem. 66, 386–393. Plaitakis A., Latsoudis H., Kanavouras K. et al. (2010) Gain-of-function variant in GLUD2 glutamate dehydrogenase modifies Parkinson’s disease onset. Eur. J. Hum. Genet. 18, 336–341. Plaitakis A., Latsoudis H. and Spanaki C. (2011) The human GLUD2 glutamate dehydrogenase and its regulation in health and disease. Neurochem. Int. 59, 495–509. Rothe F., Wolf G. and Schunzel G. (1990) Immunohistochemical demonstration of glutamate dehydrogenase in the postnatally developing rat hippocampal formation and cerebellar cortex: comparison to activity staining. Neuroscience 39, 419–429. Rothe F., Brosz M. and Storm-Mathisen J. (1995) Quantitative ultrastructural localization of glutamate dehydrogenase in the rat cerebellar cortex. Neuroscience 64, 3–17.

Shashidharan P., Michaelides T. M., Robakis N. K., Kretsovali A., Papamatheakis J. and Plaitakis A. (1994) Novel human glutamate dehydrogenase expressed in neural and testicular tissues and encoded by an X-linked intronless gene. J. Biol. Chem. 269, 16971–16976. Smith T., Schmidt T., Fang J., Wu J., Siuzdak G. and Stanley C. (2002) The structure of apo human glutamate dehydrogenase details subunit communication and allostery. J. Mol. Biol. 318, 765–777. Uversky V. N. (2003) Protein folding revisited. A polypeptide chain at the folding-misfolding-nonfolding cross-roads: which way to go? Cell. Mol. Life Sci. 60, 1852–1871. Zaganas I. and Plaitakis A. (2002) Single amino acid substitution (G456A) in the vicinity of the GTP binding domain of human housekeeping glutamate dehydrogenase markedly attenuates GTP inhibition and abolishes the cooperative behavior of the enzyme. J. Biol. Chem. 277, 26422–26428. Zaganas I., Spanaki C., Karpusas M. and Plaitakis A. (2002) Substitution of Ser for Arg443 in the regulatory domain of human housekeeping (GLUD1) glutamate dehydrogenase virtually abolishes basal activity and markedly alters the activation of the enzyme by ADP and L-leucine. J. Biol. Chem. 277, 46552–46558.
