De-regulation of D -3-phosphoglycerate dehydrogenase by domain removal Jessica K. Bell 1 , Paul J. Pease 1 , J. Ellis Bell 2 , Gregory A. Grant 3 and Leonard J. Banaszak 1 1 Department of Biochemistry, Molecular Biology and Biophysics, University of Minnesota, Minneapolis, MN, USA; 2 Department of Chemistry, University of Richmond, Richmond, Virginia, USA; 3 Department of Molecular Biology and Pharmacology and the Department of Medicine, Washington University, St Louis, MO, USA Escherichia coli 3-phosphoglycerate dehydrogenase (PGDH) catalyzes the first step in serine biosynthesis, and is allosterically inhibited by serine. Structural studies revealed a homotetramer in which the quaternary arrangement of subunits formed an elongated ellipsoid. Each subunit consisted of three domains: nucleotide, substrate and regu- latory. In PGDH, extensive interactions are formed between nucleotide binding domains. A second subunit–subunit interaction occurs between regulatory domains creating an extended b sheet. The serine-binding sites overlap this interface. In these studies, the nucleotide and substrate domains (NSDs) were subcloned to identify changes in both catalytic and physical properties upon removal of a subunit– subunit interface. The NSDs did not vary significantly from PGDH with respect to kinetic parameters with the exception that serine no longer had an effect on catalysis. Temperature dependent dynamic light scattering (DLS) revealed the NSDs aggregated > 5 °C before PGDH, indicating de- creased stability. DLS and gel filtration studies showed that the truncated enzyme formed a tetramer. This result negated the hypothesis that the removal of the regulatory domain would create an enzyme mimic of the unregulated, closely related dimeric enzymes. Expression of the regulatory do- main, to study conformational changes induced by serine binding, yielded a product that by CD spectra contained stable secondary structure. DLS and pulsed field gradient NMR studies of the regulatory domain showed the presence of higher oligomers instead of the predicted dimer. We have concluded that the removal of the regulatory domain is sufficient to eliminate serine inhibition but does not have the expected effect on the quaternary structure. Keywords: domains; enzyme regulation; oxidoreductase; 3-phosphoglycerate dehydrogenase; truncation. D -3-Phosphoglycerate dehydrogenase (PGDH) catalyzes the first committed step in the phosphorylated serine biosynthetic pathway. During the PGDH reaction, 3-phos- phoglycerate (GriP), a glycolytic intermediate, is oxidized to 3-phosphohydroxypyruvate (PHP) with the concomitant reduction of NAD. The pathway, as a branch point off the glycolysis pathway, is tightly regulated. In prokaryotes and lower plants, an inhibitory feedback loop utilizes serine to allosterically regulate the initial step of the pathway, the PGDH reaction [1–3]. The serine modulation occurs through rare V max -type effects, and may be contrasted with the more common regulation that directly affects the binding of substrate(s) by altering K m [4]. PGDH belongs to a family of D -2-hydroxyacid dehydrogenases that includes formate dehydrogenase, D -glycerate dehydrogenase, D -lactate dehydrogenase, ery- thronate-4-phosphate dehydrogenase, D -2-isocaproate dehydrogenase and vancomycin resistant protein [4]. The family members share % 22% sequence identity and 50% sequence similarity. Among the D -2-hydroxyacid dehydro- genases all members are dimeric with the exception of PGDH, which forms a homotetramer. Crystallographic studies of four enzymes within this family {2nac (for- mate,dehydrogenase [5]), 1gdh ( D -glycerate dehydrogenase [6]), 2dld ( D -lactate dehydrogenase), 1psd (3-phosphoglycer- ate dehydrogenase [7,8]} have revealed a striking similarity in their conformations, except for the additional regulatory domaininPGDH. The crystal structure of the PGDH:NAD:serine complex [7] is depicted in Fig. 1 and illustrates both the domain and quaternary arrangements. The 222 symmetric tetramer has four binding sites for both serine and NADH. The donut- like appearance of PGDH is similar to the tetrameric form of glycerol kinase [9], another enzyme that is regulated by V max -type kinetic changes. The interface encompassing adjacent nucleotide binding domains is labeled I,andthis subunit:subunit contact is shared among all D -2-hydroxy- acid dehydrogenases. The additional regulatory domain forms an important new subunit interface, labeled II.The two serine-binding sites located at each interface are comprised of residues from both subunits. As will be shown Correspondence to L. J. Banaszak, 6-155 Jackson Hall, Department of Biochemistry, Molecular Biology and Biophysics, University of Minnesota, 321 Church St S.E., Minneapolis, MN 55455, USA. Fax: + 1 612 625 2163, Tel.: + 1 612 626 6597, E-mail: len_b@dcmir.med.umn.edu Abbreviations:PGDH, D -3-phosphoglycerate dehydrogenase; NSD, nucleotide and substrate domains; RBD, regulatory binding domain; IPTG, isopropyl thio-b- D -galactoside; FDH, formate dehy- drogenase; LDH, lactate dehydrogenase; a-KG, a-ketoglutarate; PHP, 3-phosphohydroxypyruvate; 3GriP, 3-phosphoglycerate; DLS, dynamic light scattering; D t , translational diffusion constant; PFG, pulsed-field gradient. Enzymes: D -3-phosphoglycerate dehydrogenase (EC 1.1.1.95). Note: a website can be found at http://biosci.cbs.umn.edu/BMBB/ (Received 8 May 2002, accepted 25 June 2002) Eur. J. Biochem. 269, 4176–4184 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03075.x in this report, the tetrameric PGDH belongs to a family of dimeric homologues but the differences in quaternary structure are not explained solely by the presence of the regulatory domain. Finally a third proposed interface across the middle of the PGDH toroid near the region labeled III in Fig. 1 contains essentially no intersubunit contacts except through the visible loops, residues 160–195. These symmet- rically related loops could form relatively close hydrophobic and charge:charge contacts, reinforcing subunit contacts already stabilized by the interface between nucleotide binding domains. The conformational similarity between the family mem- bers is visible in the stereo-drawing shown in Fig. 2 where a PGDH subunit and a formate dehydrogenase subunit have been overlaid by the method of least-squares. The 44-kDa PGDH subunit is divided as follows: nucleotide binding domain (residues 108–292), substrate binding domain (residues 1–102, 304–318), and regulatory or serine-binding domain (residues 336–410). The interconnecting polypep- tide segment, residues 103–108, 293–303, and 319–336, may form hinge-like regions. In fact, the homologous polypep- tide segments connecting the nucleotide- and substrate- binding domains in the dimeric family members have been shown to have conformational variability [5,6,10]. Of equal relevance, but not shown in Fig. 2, the nucleotide binding domains of PGDH associate into a dimer interface entirely homologous with the quaternary structures of dimeric formate [5], glycerate [6] and D -lactate dehydrogenase. Using this well-defined homology, the potential confor- mational changes associated with the inhibited vs. the active form of PGDH were postulated from the crystal structures of apo- and holo-formate dehydrogenase [5]. These crystal complexes revealed that the active site cleft, formed by nucleotide and substrate binding domains, rotated 7.5° into a more closed conformation when ligand was bound. The constraints of the tetrameric nature of PGDH suggest that a similar rotation of the nucleotide and substrate domains into a more closed conformation at the active site would require additional relaxation of interactions at the regula- tory domain interface. The study of proposed domain movements were exam- ined by subcloning portions of PGDH to look at the contribution of the tetrameric structure to catalysis, stability and potential conformational changes at the serine site upon ligand binding. Several chimers consisting of the nucleotide and substrate domains with variable N- and C-termini were made to resemble counterparts in the 2-hydroxyacid dehydrogenase family. The kinetic properties and oligo- meric states of these truncated enzymes were determined and compared to intact PGDH. In addition, the regulatory binding domain, RBD, was subcloned to create a smaller model of the serine-binding pocket that could be manipu- lated for structural study by NMR and evaluated for conformational changes upon ligand binding. MATERIALS AND METHODS The expression vector, pSAWT containing the serA gene was described previously [11]. The plasmids, pTrc99A and pGEX-2T, were from Pharmacia Biotech. PfuDNA polymerase and the SURE cell line were from Stratagene. Fig. 2. Stereoview of PGDH and the homologous formate dehydro- genase. The crystallographic coordinates of formate dehydrogenase and PGDH have been superimposed by the least-squares methods. The resulting overlay of the two subunits is shown in stereo with formate dehydrogenase in red and PGDH in blue. A stick represen- tation NAD bound to FDH (purple) and PGDH (green) is also shown. The regulatory domain of PGDH is at the top followed by the sub- strate binding domain and finally the NAD binding domain at the bottom of the figure. The overlay of the two coordinate sets illustrates the close conformational homology between the two enzymes includ- ing the positioning of the bound coenzyme. 6 Fig. 1. Structure of PGDH: a summary of structure and mutations. The cartoon illustrates the crystal structure of the serine-inhibited form of PGDH. Three of the subunits of the homotetramer are colored gray. The fourth subunit shows the three component domains, nucleotide- binding domain (blue), the substrate-binding domain (red) and a regulatory domain (green). Three arrows mark: (I) the nucleotide- binding domain interface, and (II) the tetramer interface formed by the interactions of two regulatory domains and (III) unobserved contact across the middle of the tetramer. The position of two of the four serine molecules is shown by van der Waal’s surface at the regulatory inter- face on the left. Also shown in van der Waal’s surfaces, the NAD molecule binds within the active site cleft along the top of the nucleo- tide domain. The numbers 1–7 on the left indicate the Ca positions of the truncated enzymes. Numbers 1–4 and number 7 describe the NSD enzymes. Specifically, numbers 1 and 2 show the position of the N terminus, residue 7 (the first ordered residue in the crystal structure) and 10, respectively. The blue carbon atoms 3 and 4 indicate residues 314 and 317 at the C-terminus of two of the NSD proteins. Residue 336,usedinboththeNSDandRBDproteins,isindicatedbythegreen Ca ball. 5 Ó FEBS 2002 D -3-Phosphoglycerate DH: an active, truncated form (Eur. J. Biochem. 269) 4177 The BLR cell line was from Novagen. Restriction enzymes and ligase came from either Promega or Boehringer Mann- heim. Oligonucleotide primers were synthesized by the Microchemical Facility at the University of Minnesota, or out-sourced via this facility. DNA gel purification chemicals were from the Bio-Rad. PCR Cleanup Kit was from Promega. The Microchemical Facility at the University of Minnesota confirmed the sequences of DNA inserts. All other chemicals were from Sigma unless otherwise noted. Mutagenesis The nucleotide and substrate domain constructs of residues 1–336 (NSD:336) and 1–317 (NSD:317) were subcloned from the pSAWT vector using common PCR techniques into the pTrc99A vector. The NcoIsiteatthe5¢ end of the serA gene was conserved and a stop codon and unique XbaI site were introduced at the new 3¢ terminus at residue 336 or 317. The NSD:10–314 and NSD:10–317 mutants were constructed using the Stratagene Quik Change TM mutagen- esis kit and the NSD:336:pTrc99A vector as the parental DNA. The RBD:336–410 protein, residues 336–410, was made using the same technique as the NSD constructs, but with a BamHI site introduced at the 5¢ end and a HindIII site at the 3¢ end. The PCR product was ligated into the pGEX-2T vector. All mutant sequences were confirmed by DNA sequencing. Expression and purification NSD. NSD vectors were transformed into competent SURE cells. Six 1-L flasks of 2 · YT broth plus 150 lgÆmL )1 ampicillin were grown at 37 °C until the optical density at 600 nm reached 0.6–0.8. Protein expression was induced with 1–1.5 m M isopropyl thio-b- D -galactoside (IPTG). After induction cells were grown for % 14 h at 22 °C. Cell pellets were resuspended in 50 m M KH 2 PO 4 pH 7.0, 2 m M dith- iothreitol, 1 m M EDTA and 0.05% NaN 3 (buffer B) and lysed by sonication. The remainder of the purification protocol has been described previously [11]. Purified protein was concentrated using a Centriprep 10K (Amicon) and dialyzed into buffer B. Protein concentration was deter- mined by Bradford assay and/or UV spectra using an extinction coefficient of 0.67 M )1 Æcm )1 . Protein was stored at 4 °C. RBD:336–410. RBD:336–410 plasmid was transformed into competent BLR cells. Six 1-L flasks of 2 · YT broth plus antibiotic were grown at 37 °CtoD 600 ¼ 0.6–1.0 and then induced with 1 m M IPTG. Cells were grown for % 14 h at 22 °C. The cell pellet was resuspended in STE (10 m M Tris/HCl pH 8.0, 1 m M EDTA, 150 m M NaCl) and incubated on ice with 0.1 mgÆmL )1 lysozyme for 15 min The solution was brought to 5 m M dithiothreitol, 2% (w/v) sarkosyl and sonicated. The mixture was stirred at 4 °Cfor 30 min followed by centrifugation at 10 000 g 1 for 30 min. Polyethyleneamine (0.035%) was added to remove DNA/ RNA, stirred at 4 °C for 30 min and then respun for 30 min at 10 000 g. The supernatant was concentrated using an Amicon concentrator with a PM10 membrane (3 kDa cut- off) and dialyzed into NaCl/P i /EDTA (16 m M Na 2 HPO 4 , 4m M NaH 2 PO 4 , 150 m M NaCl, 1 m M EDTA pH 7.3). The dialyzed lysate was respun to remove particulates and loaded onto a glutathione S-transferase (GST) affinity column (Novagen). The column was washed with 10 column vols NaCl/P i /EDTA and then incubated with 250 U thrombin overnight at room temperature. The cleaved RBD:336–410 was eluted, concentrated using a Centriprep 3K (Amicon), and stored at 4 °C. Protein concentration was calculated from UV spectra using an extinction coefficient of 0.47 M )1 Æcm )1 for RBD:336–410. The identity of the protein was confirmed by N-terminal sequencing of the first 10 residues and amino acid analysis (Microchemical Facility, University of Minnesota, MN, USA). Kinetic analysis The steady-state initial rates were determined by following either the reduction of 3-PHP or a-ketoglutarate (a-KG). Thereactionwassetupwithasaturatingconcentrationof NADH (100–200 l M ) and varied concentrations of PHP (1–100 l M )ora-KG (10.4–5000 l M )at25°C. The enzyme concentration for the a-KG studies was 1 l M and 0.1–0.5 l M for the PHP reactions. The assay buffer for the a-KG reactions was 50 m M Tris pH 7.5, 1 m M EDTA and 2 m M dithiothreitol. For the 3-PHP reactions, the Tris concentration was increased to 500 m M . The reaction was initiated by the addition of substrate and the decrease in D 340 monitored for 10–20 s. The initial rates were deter- mined by fitting a linear regression to the curve and calculating the slope using CARY 50 kinetics software. Assays were repeated a minimum of five times. The data were analyzed by Michaelis–Menten or Lineweaver–Burke plots and kinetic parameters derived using the SIGMA PLOT 5.0 software (Jandel Scientific Inc.). Dynamic light scattering experiments Dynamic light scattering (DLS) experiments were conducted in Buffer B for PGDH and the NSD proteins. RBD:336–410 experiments were done in NaCl/P i /EDTA. For each concentration measured, the protein was spun at 14 000 g for 10 min and passed through a 0.1-l M filter. A 12-lL sample was equilibrated by a built-in thermostat at 5 °C increments. Data were collected with a Protein Solutions DLS system and evaluated with the DYNAPRO V 4.0 software. For each temperature 15–20 data points were collected. Mean values were calculated for the DLS parameters. Points that were outside 1 SD were excluded. Data were plotted in SIGMA PLOT 5.0. Gel filtration experiments Gel filtration experiments were performed in Buffer B. PGDH (2 mgÆmL )1 ), NSD:317 (2 mgÆmL )1 ), or 3 D -lactate dehydrogenase ( D -LDH) (2 mgÆmL )1 )wererunovera Sephacryl S200 (Pharmacia) gel filtration column with both low molecular weight standards (ribonuclease A, 13.7 kDa; chymotrypsinogen A, 25 kDa; ovalbumin, 43 kDa; BSA, 67) (Run 1) and high molecular weight standards (aldolase, 158 kDa; catalase, 232 kDa; ferritin, 440 kDa; thyroglobulin, 669 kDa) (Run 2). Chromato- graph profiles were calculated from the absorbance of the fractions at 280 nm for the molecular mass standards and activity measurements for PGDH, NSD:317 and D -LDH. 4178 J. K. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002 The molecular weights of PGDH, NSD:317 and D -LDH were calculated from the linear regression of K av [(V e ) V o )/ (V t ) V o ), where V e is the elution volume, V o is the void volume and V t is the total volume] vs. the log of the molecular weight of the standards. CD RBD:336–410 experiments were performed in NaCl/P i / EDTA. CD spectra were collected on protein (0.722ÆmgÆ mL )1 ) in the presence or absence of 1 m M serine. A buffer blank was completed for both the buffer and buffer plus 1m M serine. The spectra were collected on a Jasco 710 instrument at room temperature using a 0.05-mm quartz cell. Spectra were collected from 250 to % 200 nm with eight accumulations. The data were averaged over the accumu- lations, corrected for the buffer blank and random signals were smoothed using the JASCO software package. Data were exported to SIGMA PLOT 5.0 for analysis. Pulsed-field gradient NMR To corroborate the DLS measurements, pulsed-field gradi- ent (PFG)-NMR [12,13] was used to give an independent determination of the translational diffusion constant (D t ) for the protein RBD:336–410. NMR was carried out in collaboration with the Mayo laboratory at the University of Minnesota. Spectra were collected and analyzed by Shou Lin Chang of the Mayo laboratory. In the PFG-NMR experiment, B o , constant magnetic field, was superimposed twice during a short time interval, d, by an additional inhomogenous gradient (G z ). The result of the two gradient pulses is to create an echo. If no motion or relaxation occurred on the z-axis, the echo would have been identical to the initial signal. However, the observed echo will be attenuated by both relaxation and random motion (diffu- sion), along the z-axis. The attenuation, A(t) can be described by: AðtÞ¼Að0Þ exp ðÀRðtÞÀc 2 G 2 D t d 2 ðd À d=3ÞÞ where R(t) is attenuation due to relaxation, c is the magnetogyric ratio, G is the gradient strength, d is the duration of the gradient pulse, and D is the interval between the start of the two gradient pulses. To determine an accurate measurement of D t , a series of 12 one-dimensional PFG spectra were collected at gradient field strengths, 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55 and 60 GÆcm )1 . The data were then fit to the semi-log of the equation above to determine the value of D t . Experiments were conducted at both 10 and 25 °C. Protein (%1m M )wasin50m M KH 2 PO 4 ,pH6.5,inD 2 O. Serine, when present, was at 1m M . The experiments were carried out on a Varian UNITY 600 MHz NMR with triple resonance probe and triple axis gradient unit (High Magnetic Field Facility, University of Minnesota). RESULTS AND DISCUSSION Nucleotide substrate somains from Gri P DH The preparation of a monodisperse form of PGDH insensitive to the presence of serine but fully active was designed based on a previously determined crystal structure (Fig. 1) [7]. Removal of the serine-binding domain was predicted to eliminate allosteric inhibition by serine and produce a dimeric enzyme. As shown in the data below, manipulation of PGDHs quaternary structure was far more complicated and it was not possible to obtain a dimeric enzyme. Several variations of the NSD’s were developed using the standard PCR technique of introducing a stop codon and unique restriction site at the desired termination point. NSD:336 (residues 1–336), the first two-domain protein to be made, was soluble and yielded % 5–12 mg per 6 L ferment. However, NSD:336 included a segment of the linker sequence between the substrate and the regulatory domains, and tended to form higher oligomeric species (data not shown). This extended linker may have decreased stability and provide a site of aggregation, and was therefore removed in another form NSD:317 (residues 1–317). NSD:317 was monodisperse in solution and relatively stable (see below), and therefore more amenable to study. Two other two-domain enzymes were also created: NSD:10–314 and NSD:10–317. These forms eliminated the N-terminal segment that was disordered in crystalline PGDH with serine. The recombinant products were largely insoluble and further studies were aban- doned. Kinetic evaluation of NSDs As conformational changes had been linked to the catalytic activity of both PGDH and formate dehydrogenase (FDH) [5,14] removal of the regulatory domain was hypothesized to have an effect on the kinetic parameters of PGDH. The steady-state parameters are reported for two of the chimers although our primary focus was NSD:317 because the quaternary structure of this enzyme was definable. The activities of PGDH and NSD:317 were assayed following the reduction of PHP, which occurs % 70-fold faster than the oxidation of GriP [15], or the alternate substrate, a-KG [16]. Although PHP and a-KG are three- and five-carbon substrates, respectively, the fourth carbon and 5-carboxyl of a-KG are similar to the bulky phosphate group in PHP. The results of the steady-state kinetic studies are summa- rized in Table 1. For both the PHP and a-KG assays, substrate inhibition was observed at high concentrations (Fig. 3), possibly due to the slow release of oxidized cofactor and leading to an abortive complex of substrate/ NAD. The data from the reduction of PHP and a-KG, excluding data exhibiting substrate inhibition, were evalu- ated by Michaelis–Menten plots to derive K m and V max . Overall the kinetic parameters of the native and NSD enzymes do not vary significantly (Table 1). The K m(PHP) for PGDH agrees well with the value first published by Pizer, 1.2 ± 1 vs. 1.3 l M [15]. The alternative substrate, a-KG, shows an 18-fold increase in K m over PHP and an order of magnitude decrease in V max /K m .Thelower catalytic efficiency is consistent with the hypothesis that the 5-carboxyl group in a-KG is not a good substitute for the phosphate group of PHP. However, both NSD:317 and PGDH behave similarly with respect to this pseudo- substrate. The effect of serine on NSD:317 was also tested. Using saturating concentrations of both substrate and cofactor in Ó FEBS 2002 D -3-Phosphoglycerate DH: an active, truncated form (Eur. J. Biochem. 269) 4179 the presence and absence of 5 m M serine (IC 50 for native enzyme ¼ 5 l M ; [17]), no change in the initial rate of the catalytic reaction was found (data not shown). Given that the NSD enzymes were not affected by serine, the purity of an enzyme preparation, usually contaminated with wild- type PGDH from Escherichia coli, was routinely determined by assays in the presence and absence of serine. Because the kinetic characteristics of NSD:317 are comparable to those ofthenativeenzyme,thereleaseofthehingedactivesite from the constraints of the regulatory domain have neither increased nor decreased its catalytic capabilities. This reinforces the supposition that the serine-binding domain evolved solely for regulation, and may explain also why the mammalian forms of the enzyme, although no longer regulated by serine [18], have not shed the serine-binding domain. Quaternary structure and stability As shown in Fig. 1, the PGDH tetramer has two major types of subunit interfaces. Removal of the subunit contacts formed by the regulatory domains, as in the NSD enzymes, was predicted to result in a dimeric species. DLS results from solutions of NSD:336 at micromolar subunit concen- trations indicated that this enzyme formed higher oligo- meric species, up to 12-mers (data not shown). The removal of the C-terminal linker region (residues 318–336) in the NSD:317 enzyme alleviated the aggregation problem. An overview of the D t for NSD:317 compared to the native enzyme and the concentration dependence is shown in Fig. 4. In contrast to NSD:336 protein, this truncated form gave reproducible measurements at 0.5, 1.0 and 2.0 mgÆmL )1 (14.7–58.7 l M ). The D t values are slightly larger than those for the native enzyme up to 30 °C, consistent with NSD:317 forming a somewhat smaller molecule. The D t data were analyzed by two different methods, both of which are summarized in the insert to Fig. 4. Using the Stokes–Einstein equation, D t maybeusedtocalculate the equivalent hydrodynamic radius, R h : D t ¼ kT=6pgR h where k is the Boltzman constant, T is the absolute temperature and g is the solvent viscosity. As shown in the inset, D t s of 440 and 520 for PGDH and NSD:317, respectively, lead to R h values of 52 A ˚ and 47 A ˚ .The corresponding molecular weights of PGDH and NSD:317, based upon a spherical model, were 157 and 126 kDa respectively. Given that the subunit molecular mass (m)of NSD:317 is 34 kDa, these results suggested that the truncated enzyme was forming a tetramer instead of the expected dimer. The second method of evaluating D t makes use of the crystallographic model coordinates of PGDH. If the coordinates are used to determine a prolate ellipsoid of equivalent dimensions, R h , of a comparable sphere may be calculated: R h ¼ðab 2 Þ 1=3 where a and b are the half lengths of the long and short axis of the crystallographic prolate ellipsoid, respectively. The proposed structure of NSD as either a dimer, as expected, or Table 1. Steady state properties of NSD:336, NSD:317 and D -3-phosphoglycerate dehydrogenase. Rates of NADH oxidation were determined by measuring the decrease in OD at 340 nm. The a-KG assays were completed in 50 m M Tris, pH 8.0, 2 m M dithiothreitol, 1 m M EDTA with saturating cofactor, 200 l M ,anda-KG concentrations from 10.4 to 5000 l M at 25 °C. The 3-phosphohydroxypyruvate assays were carried out with a 10-fold higher concentration of Tris, 500 m M , and 3-phosphohydroxypyruvate concentrations from 1 to 100 l M . Enzyme form Assay K m a l M V max a s )1 V max /K m a s )1 Æ M )1 PGDH 3-PHP 1.2 ± 1 2.6 ± 0.07 2.2 · 10 6 NSD:336 3-PHP 0.6 ± 0.07 2 ± 0.03 3.3 · 10 6 NSD:317 3-PHP 1.7 ± 0.2 2.3 ± 0.05 1.4 · 10 6 PGDH a-KG 18.5 ± 1 3.5 ± 0.03 1.9 · 10 5 NSD:336 a-KG 21.4 ± 1.8 2.3 ± 0.03 1.1 · 10 5 NSD:317 a-KG 28.3 ± 3.7 2.5 ± 0.05 8.8 · 10 4 a Parameters derived from fitting the velocity vs. substrate concentration plot to the Michaelis–Menten equation. Fig. 3. Michaelis–Menten plot of PGDH, NSD:317 and NSD:336 ki- netic data for the a-KG substrate. The velocity vs. substrate concen- tration plots of the kinetic data for PGDH (d), NSD:317 (s) and NSD:336 (m) clearly show that no significant differences between kinetic parameters are distinguishable. The largest difference occurs in the value of V max but this is less than a twofold difference between native and truncated enzymes. At a-KG concentrations > 2–3 m M , substrate inhibition was observed. Data points exhibiting inhibition (shaded in grey) were excluded from calculation of the kinetic parameters. Experimental conditions are given in Table 1. Similar data were collected with PHP as the substrate, not shown. The y-axis, v,is defined as [NADH]/[enzyme] with units of s )1 . 4180 J. K. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002 a tetramer, utilizing contacts of the extended loops across the ellipsoid, were modeled from the PGDH coordinates by removing the regulatory domain. The inset of Fig. 4 summarizes the results of these approximations. The agreement between the observed R h and the R h calculated from the crystallographic ellipsoid is consistent with a tetrameric form of NSD:317. The unexpected results of the DLS experiments suggest- ing a tetrameric form of the NSD:317 enzyme was confirmed by gel filtration. The chromatographs of PGDH (predicted m 176 kDa) and NSD:317 (predicted dimeric m 68 kDa, predicted tetrameric m 136 kDa) revealed that both enzymes were eluting before the molecular mass standard aldolase (m 158 kDa) (Fig. 5). In fact, PGDH coeluted with the molecular mass standard catalase (m 232 kDa) at a higher than predicted molecular mass, indicating that the ellipsoidal quaternary structure has affected its elution pattern. To evaluate the oligomeric state of NSD:317 while allowing for the overall shape of the molecule, we compared its elution pattern with that of a known dimeric D -2-hydroxyacid dehydrogenase of similar fold, D -LDH (predicted dimeric m 74 kDa) [19] (1ldh). The D -LDH elution profile indicates that this enzyme forms both a dimer (majority) and a tetramer [19], with predicted molecular masses of 74 and 148 kDa, respectively. NSD:317 elutes slightly after the tetrameric form D -LDH but significantly before the dimeric form of D -LDH. The differences in tetrameric molecular mass of D -LDH and NSD:317 may result from D -LDH being slightly larger (subunit m of 37 kDa vs. 34 kDa) or reflect a tighter packing of the tetramer form of NSD:317 leading to a more compact and thus ÔsmallerÕ species. If the elution profiles of the well characterized PGDH, the tetrameric D -LDH and dimeric D -LDH are used to determine a molecular mass standard curve, the mass of NSD:317 would be calculated as 141.8 kDa compared to the predicted tetrameric mass of 136 kDa. Therefore, gel filtration results of nonspherical proteins greatly benefit from evaluation with respect to proteins of known similar folds and quaternary structure. The results of the gel filtration studies are consistent with the DLS data in support of a tetrameric form for NSD:317. The DLS measurements were also used to evaluate the stability of NSD:317 in comparison to PGDH by monitor- ing D t as a function of temperature. The D t values for the NSD:317 dropped dramatically above 30 °C compared to native enzyme, indicative of formation of a larger species. In addition, the polydispersity, that was negligible below 30 °C, rises considerably. The decreased stability of NSD:317 and the length dependence of the C-terminus to determine monodispersity are consistent with the now exposed substrate:regulatory domain contact potentially offering a site of aggregation or preliminary unfolding. As mentioned above, mammalian PGDH retains its regulatory binding domain although it no longer allosterically regulat- ed by serine. Perhaps, the RBD has been retained to increase protein stability and limit aggregation. Fig. 4. DLS of NSD:317. The DLS experiments were conducted as a function of both temperature and concentration. D t ,increases,as predicted by the Stokes–Einstein equation, with temperature to 30 °C. At 35 °CtheD t value decreases by approximately one-third, sug- gesting that the protein has begun to aggregate. Native enzyme is shown as closed circles, mutant as open symbols. The increase in D t for NSD:317 does not appear to be concentration dependent over this concentration range, 0.5 mgÆmL )1 (s), 1 mgÆmL )1 (h)and 2mgÆmL )1 (n). The inset compares the calculation R h ,fromthe experimental D t and the Stokes–Einstein equation vs. calculation from the crystallographic structure and a prolate ellipsoid. The values of a and b are the length of the two axes of the ellipsoid measured from the crystal structure, 1psd. DLS measurements were conducted in 50 m M KH 2 PO 4 pH 7.0, 2 m M dithiothreitol, 1 m M EDTA, 0.05% NaN 3 . At a given temperature the values for each parameter were averaged for the 0.5, 1.0 and 2.0 mgÆmL )1 measurements. Fig. 5. Gel filtration chromatograph of PGDH and NSD:317. The elution profiles of PGDH (m,176 kDa;d), NSD:317 (j), and D -LDH (m,74 kDadimeric;m, 148 kDa tetrameric; m) are shown with respect to the profile of molecular weight standards, catalase (m, 232 kDa), aldolase (m, 158 kDa) and ovalbumin (m,43kDa)depictedbythe gray line. PGDH elutes with catalase suggesting that the ellipsoidal shape of the enzyme increases the apparent molecular mass. D -LDH appears to run as a dimer, D -LDH 1, and tetramer, D -LDH 2, with the majority seen as a dimer. Both D -LDH species elute at a higher than predicted molecular mass (100 kDa and 220 kDa), again this observed increase in molecular mass can be attributed to the elongated shape of the enzyme. The comparison of the NSD:317 elution with the D -LDH pattern suggests that the truncated enzyme is forming a tetramer with a molecular mass of 196 kDa (predicted m, 136 kDa). Gel filtration studies were completed in Buffer B on a Sephacryl S200 matrix with each protein sample at a concentration of 2 mgÆmL )1 . Note that the elution of PGDH, NSD:317 and D -LDH were determined by activity measurements to remove ambiguity of elution profiles from absor- bance measurements at 280 nm. Ó FEBS 2002 D -3-Phosphoglycerate DH: an active, truncated form (Eur. J. Biochem. 269) 4181 Regulatory domain The regulatory or serine-binding domain of PGDH consists of 76 residues (residues 336–410). In the crystal structure, the subunit–subunit interface at the regulatory domains (II in Fig. 1) was shown to consist of an extended b sheet created by adjacent subunits [7]. Serine binding was proposed to increase the interactions at this interface thereby locking the active site into a more open and inactive conformation. The uninhibited form of the enzyme would be more flexible at the interface formed by the regulatory domains allowing more motion at the hinge regions and permitting the active site to close. To allow this conforma- tional flexibility, changes at the interface formed by the regulatory domains were proposed to involve the disruption of the extended b sheet. To study the effect of serine binding at this subunit interface, we attempted to develop a simple dimer of the regulatory domains. The small size of this domain, 76 residues, would allow for structural studies by NMR or crystallography. However, polypeptides of this molecular weight proved difficult to purify from E. coli extracts, so RBD:336–410 was expressed as a GST fusion protein. After cell lysis, SDS gels indicated that the majority of the target protein was in the resulting insoluble pellet. Addition of a detergent, sarkosyl, solubilized much of the GST- RBD:336–410. RBD:336–410 could be obtained in pure form by chromatography on a glutathione column followed by proteolysis with thrombin to remove the GST tag (data not shown). Unlike the NSD proteins, RBD:336–410 could not be characterized by a catalytic assay. The chemical identity of this small, purified protein was verified by both amino acid analysis and N-terminal sequencing of the first 10 residues. As the protein was solubilized with detergent, CD measurements were conducted to determine whether stable secondary structure had formed. The CD measurements were completed in the presence and absence of serine. Fig. 6 shows that RBD:336–410 had minima for both b structure (217 nm) and a helix (222 and 208 nm). The addition of serine had no significant effect on the secondary structure. The CD spectra show the presence of secondary structural elements consistent with the intact enzyme. To determine the oligomeric nature of RBD:336–410, both DLS experiments at 18 and 23 °C(0.5mgÆmL )1 ) and PFG-NMR studies in collaboration with the Mayo laboratory at the University of Minnesota were carried out. If RBD:336–410 was dimeric, this would be apparent in the D t and the corresponding R h .The NMR studies would also be useful for determining whether the structure of RBD:336–410 could be solved by NMR. The results of these experiments, shown in Table 2, indicated that the new protein formed not the expected dimeric species, but a higher oligomeric mole- cule. In Table 2, values of R h are based on the Stokes–Einstein relationship mentioned earlier. The results from the two experimentally independent methods, NMR and DLS, agree within 10%. Furthermore, the data in Table 2 indicate that the addition of serine had no significant effect on the D t values. Given the consistency of the data, taking the overall average appeared justified resulting in an R h of 37 A ˚ .Using a partial specific volume of 0.73 mLÆg )1 , the molecular mass of the new aggregate would be 42 kDa. With a monomeric molecular mass for RBD:336–410 of 8.1 kDa, the regula- tory domain by itself behaves like either a pentamer or a Fig. 6. CD spectra of RBD:336–410 in the presence/absence of 1 m M serine. CD was performed in 16 m M Na 2 HPO 4 ,4m M NaH 2 PO 4 , 150 m M NaCl, 1 m M EDTA, pH 7.3 at 0.72 mgÆmL )1 (% 0.1 m M ) protein. Serine, when present, was at 1 m M . The RBD:336–410 spectra are shown as black lines: RBD:336–410 + 1 m M Serine are shown as gray lines. RBD:336–410 contains two minima at 217–222 and 206– 208 nm corresponding to a-helical and b strand content, respectively. The addition of serine to RBD:336–410 does not have a significant effect on the secondary structure. Table 2. DT for RBD:336–410 calculated from DLS and PFG-NMR data. DLS experiments were conducted in 16 m M Na 2 HPO 4 ,4 m M NaH 2 PO 4 , 150 m M NaCl, 1 m M EDTA pH 7.3 in the presence and absence of 1 m M serine as indicated. NMR studies were completed on protein at % 1m M under identical conditions. D t s are reported in cm 2 Æs )1 · 10 9 . R h (equivalent sphere) was calculated using the Stokes–Einstein model. DLS NMR 18 °C23°C10°C5°C D t R h (A ˚ )D t R h (A ˚ )D t R h (A ˚ )D t R h (A ˚ ) – Ser 641 ± 39 37 698 ± 34 33 426 ± 7 58 615 ± 16 40 +Ser 558 ± 16 43 624 ± 38 37 412 ± 9 60 645 ± 14 38 4182 J. K. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002 hexamer. Because of the inherent shape uncertainties in extrapolating molecular masses from DT, and in spite of the close agreement between the two independent methods, the exact oligomeric state and nature of subunit interactions remains unresolved. The results, however, do clearly indicate that the RBD does not form the expected dimer. CONCLUSIONS The NSD enzymes were developed as an alternative to the serine regulated native PGDH. Removal of the regulatory domain had little influence on the enzyme’s catalytic reaction and kinetic parameters determined from steady- state studies. The largest differences between the native enzyme and the NSD proteins occurred in the stability of the enzymes. The native tetramer retained its predicted DT upto40°C whereas the NSD protein began to aggregate and/or denature between 30 and 35 °C. These results are not surprising as removal of the regulatory or serine-binding domain exposes a surface area that is partially buried in the native tetramer and may be susceptible to aggregation or unfolding. In mammals, serine feedback regulation has been replaced by transcriptional control [18] yet the alignment of sequences from a variety of species indicate that the regulatory domain has been conserved. The retention of the regulatory domain and thus the subunit:subunit inter- face may provide additional stability to the quaternary structure as is observed in the differences between E. coli PGDH and the NSD enzymes. Mutations within this domain of the human PGDH lead to loss of or lowered serine production without a significant decrease in mRNA production [20]. The work presented here would suggest that stability studies of these clinically characterized muta- tions may give insight as to the role of the regulatory domain in higher eukaryotes. We predicted that the NSDs would more closely resemble other dimeric D -2-hydroxyacid dehydrogenases. The oligo- meric structure of NSD:317 was, instead, a tetramer. From the crystallographic structure of the serine-inhibited en- zyme, and some preliminary structural results with a mutant form of PGDH, a model has been formulated. Figure 7A,B reiterate the subunit contacts of the PGDH–NAD–serine structure and the proposed conformational change upon catalysis or release of inhibition. Given that the tetrameric interface, labeled II in Fig. 7, had been removed, NSD:317 must have formed a new subunit–subunit interface to remain a tetramer. New structural results from a point mutation, W139G PGDH, have shown the collapse of the ellipsoid with extensive interactions being made between the extended loops (residues 165–190) and the subunits across the toroid [21]. Based upon this new structural data we propose that the NSD:317 enzyme has formed a new, or as yet structurally uncharacterized, tetrameric interface through the interaction of the extended loops (residues 165–190) (Fig. 7C). Perhaps similar subunit:subunit inter- actions are important in the uninhibited form of PGDH, in which the active site cleft has adopted a closed conforma- tion. Structural studies to investigate that possibility are currently underway. The subcloning of the regulatory binding domain offered a unique opportunity to look at conformational changes induced by serine as a subset of the whole enzyme. However, the construct proved poorly soluble unless it was coupled with a fusion protein and solubilized with detergents (sarkosyl). The presence of the secondary struc- ture as assessed by CD spectra suggested that the regulatory domain could fold independently. However, DLS and PFG-NMR experiments clearly showed that the protein aggregated under a variety of conditions. The aggregation tendency coupled with the small size makes this domain particularly difficult to analyze with respect to ligand binding. Nonetheless, the DT values obtained from solu- tions of RBD were nearly identical whether determined by DLS or NMR. This establishes the usefulness of both methods in studying the hydrodynamic properties and quaternary structures of macromolecules, and demonstrat- ed that the regulatory domains alone form an even more complex quaternary structure. The new enzymes created by recombinant methods provided a step back in the evolutionary chain. Rather than stringing together multiple functional units we can dissect the contribution of individual domains towards the complex regulation and cooperativity observed within this enzyme system. The role of the tetrameric PGDH evolved to provide a means of regulating serine production within prokaryotes and lower plants. Although at the outset we predicted, based upon PGDH structural data and homol- ogous dimeric enzymes, an easily manipulated oligomeric structure, we were foiled by the complexities of heretofore unrevealed subunit:subunit contacts. Loss of one of the obvious tetrameric interfaces still results in a tetrameric enzyme. We continue our studies of this new subunit contact by looking at the native enzyme and why this interface may be beneficial. Fig. 7. Model of regulatory domain subunit:subunit interface proposed conformational changes. In this representation of PGDH only half of the tetramer is depicted. The domains are labeled NAD-BD, nucleo- tide binding domain; SBD, substrate binding domain; and RBD, regulatory binding domain. The arrows describe the positions of twofold rotation axes in the plane of the drawing. The third dyad associated with the 222 symmetrical tetramer is indicated by the black ellipse located at the intersection of the dyad arrows. In the inhibited state of PGDH (A), serine molecules are depicted as black stars, and the regulatory domains form an extended b sheet with the serine molecules bridging the two subunits. The crosses (substrate) located between the SBDs and NAD-BD domains indicate the location of the active sites. In this schematic model, the uninhibited state of PGDH (B) differs by the reorientation of all three domains. The new confor- mational state now contains a more closed conformation at the active site. The NSDs in (C) lack the RBDs. In this form, new subunit interfaces form across the dyad perpendicular to the plane of the drawingandatetramerresults. Ó FEBS 2002 D -3-Phosphoglycerate DH: an active, truncated form (Eur. J. Biochem. 269) 4183 ACKNOWLEDGEMENTS This work was funded by National Science Foundation grants MCB9318699 to L. J. B. and MCB9986278 to J. E. B. and a grant from the National Institutes of Health (GM56676) to G. A. G. The authors are grateful to both M. Lees and J. Bratt of the Banaszak laboratory for assistance in preparation of DNA constructs and protein purification. The authors would also like to thank Shou Lin Chang of the Mayo laboratory at the University of Minnesota for conducting the PFG-NMR experiments and K. Mayo for use of the Jasco 710 CD spectrophotometer. We gratefully acknowledge the help of J. Barycki in the preparation of this report. REFERENCES 1. Pizer, L. (1963) The pathway and control of serine biosynthesis in Escherichia coli. J. Biol. Chem. 238, 3934–3944. 2. Slaughter, J.C. & Davies, D.D. (1968) Inhibition of 3-phos- phoglycerate dehydrotenase by l-serine. Biochem. J. 109, 749–755. 3. Willis, J.E. & Sallach, H.J. (1964) The occurrence of D-3-phos- phoglycerate dehydrogenase in animal tissues. Biochim. Biophys. Acta 81, 39–54. 4. Grant, G. (1989) A new family of 2-hydroxyacid dehydrogenases. Biochem. Biophys. Res. Comm. 165, 1371–1374. 5. Lamzin,V.S.,Dauter,Z.,Popov,V.O.,Harutyunyan,E.H.& Wilson, K.S. (1994) High resolution structures of holo and apo formate dehydrogenase. J. Mol. Biol. 236, 759–785. 6. Goldberg, J.D., Yoshida, T. & Brick, P. (1994) Crystal structure of a NAD-dependent D -glycerate dehydrogenase at 2.4 A ˚ resolution. J. Mol. Biol. 236, 1123–1140. 7. Schuller, D.J., Grant, G.A. & Banaszak, L.J. (1995) The allosteric ligand site in the V max -type cooperative enzyme phosphoglycerate dehydrogenase. Nat. Struct. Biol. 2, 69–75. 8. Grant, G.A., Schuller, D.J. & Banaszak, L.J. (1996) A model for the regulation of D -3-phosphoglycerate dehydrogenase, a V-max- type allosteric enzyme. Prot. Sci. 5, 34–41. 9.Hurley,J.H.,Faber,H.R.,Worthylake,D.,Meadow,N.D., Roseman, S., Pettigrew, D.W. & Remington, S.J. (1993) Structure of the regulatory complex of Escherichia coli IIIglc with glycerol kinase. Science 259, 673–677. 10. Stoll, V.S., Kimber, M.S. & Pai, E.F. (1996) Insights into substrate binding by D -2-ketoacid dehydrogenases from the structure of Lactobacillus pentosus D -lactate dehydrogenase. Structure 4,437– 447. 11. Schuller, D.J., Getter, C.H., Banaszak, L.J. & Grant, G.A. (1989) Enhanced expression of the Escherichia coli serA gene in a plasmid vector. J. Biol. Chem. 264, 2645–2648. 12. Tillett, M.L., Lian, L.Y. & Norwood, T.J. (1998) Practical aspects of the measurement of the diffusion of proteins in aqueous solu- tion. J. Magn. Reson. 133, 379–384. 13. Karger, J., Pfeifer, H. & Heink, W. (1988) Adv. Magn. Reson. 12, 1–89. 14. Dubrow, R. & Pizer, L. (1977) Transient kinetic and deuterium isotope effect studies on the catalytic mechanism of phosphogly- cerate dehydrogenase. J. Biol. Chem. 25, 1539–1551. 15. Sugimoto, E. & Pizer, L. (1968) The mechanism of end product inhibition of serine biosynthesis I. Purification and kinetics of phosphoglycerate dehydrogenase. J. Biol. Chem. 243, 2081–2089. 16. Zhao, G. & Winkler, M.E. (1996) A novel alpha-ketoglutarate reductase activity of the serA encoded 3-phosphoglycerate dehydrogenase of Escherichia coli K-12 and its possible implica- tions for human 2-hydroxyglutaric aciduria. J. Bacteriol. 178,232– 239. 17. Sugimoto, E. & Pizer, L. (1968) The mechanism of end product inhibition of serine biosynthesis II. Optical studies of phos- phoglycerate dehydrogenase. J. Biol. Chem. 243, 2090–2098. 18. Achouri, Y., Rider, M.H., van Schaftingen, E. & Robbi, M. (1997) Cloning, sequencing and expression of rat liver 3-phosphoglyce- rate dehydrogenase. Biochem. J. 323, 365–370. 19. Kochhar, S., Hottinger, H., Chuard, N., Taylor, P.G., Atkinson, T., Scawen, M.D. & Nicholls, D.J. (1992) Cloning and over- expression of Lactobacillus helvetic D -lactate dehydrogenase gene in Escherichia coli. Eur. J. Biochem. 208, 799–805. 20. Klomp, L.W., de Konig, T.J., Malingre, H.E., van Beurden, E.A.,Brink,M.,Opdam,F.L.,Duran,M.,Jaeken,J.,Pineda, M., van Maldergem, L., Poll-The, B.T., van den Berg, I.E. & Berger, R. (2000) Molecular characterization of 3-phosphoglyce- rate dehydrogenase deficiency – a neurometabolic disorder asso- ciated with reduced L -serine biosynthesis. Am.J.Hum.Genet.67, 1389–1399. 21. Bell, J.K., Grant, G.A. & Banaszak, L.J. (2001) Disrupting sub- unit communication: Point mutations at the subunit interfaces of D -3-phosphoglycerate dehydrogenase. FASEB J. 15, A535. 4184 J. K. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002 . regulatory domain of PGDH is at the top followed by the sub- strate binding domain and finally the NAD binding domain at the bottom of the figure. The overlay of. regulatory domain showed the presence of higher oligomers instead of the predicted dimer. We have concluded that the removal of the regulatory domain is sufficient