Stability Domains, Substrate-induced Conformational Changes, and Hinge-bending Motions in a Psychrophilic Phosphoglycerate Kinase

The cold-active phosphoglycerate kinase from the Antarctic bacterium Pseudomonas sp. TACII18 exhibits two distinct stability domains in the free, open conformation. It is shown that these stability domains do not match the structural N- and C-domains as the heat-stable domain corresponds to about 80 residues of the C-domain, including the nucleotide binding site, whereas the remaining of the protein contributes to the main heat-labile domain. This was demonstrated by spectroscopic and microcalorimetric analyses of the native enzyme, of its mutants, and of the isolated recombinant structural domains. It is proposed that the heat-stable domain provides a compact structure improving the binding affinity of the nucleotide, therefore increasing the catalytic efficiency at low temperatures. Upon substrate binding, the enzyme adopts a uniformly more stable closed conformation. Substrate-induced stability changes suggest that the free energy of ligand binding is converted into an increased conformational stability used to drive the hinge-bending motions and domain closure.

Phosphoglycerate kinase (PGK 2 ; EC 2.7.2.3) catalyzes the reversible high energy phosphoryl transfer from 1,3-diphospho-D-glycerate to Mg 2ϩ -ADP to yield Mg 2ϩ -ATP and 3-phospho-D-glycerate (3-PGA) in the glycolytic pathway. Besides its key metabolic function in living organisms, PGK has been widely used as a model enzyme to study domains in proteins and their relative motions (1,2). Indeed, the structure of PGK consists of two globular domains of nearly equal size, referred to as the N-and C-terminal domains, linked by a hinge helix. In this structure, the C terminus of the protein crosses back the hinge region and folds into the N-domain, bringing the N-and C-extremities in close proximity where they generally interact via ion pairs (3,4). The triose substrate binds to a basic patch in the N-domain (5), whereas the nucleotide substrate is bound to the C-domain (6). Crystallographic studies have shown that the free or single substrate-liganded forms adopt an open conformation in which the distance between the substrates is too long to permit catalysis, probably to avoid the futile hydrolysis of 1,3-PGA or ATP (7). By contrast, the binding of both substrates induces a closed conformation resulting from hinge-bending motions of 32°at the level of the hinge helices and leading to domain closure into a catalytically competent conformation with well aligned substrates for phosphoryl transfer (7,8).
The stability and folding of yeast PGK has been extensively investigated and deconvolution of its single DSC heat absorption peak suggests the occurrence of two overlapping transitions that have been tentatively assigned to the N-and C-domains (9). Furthermore, folding studies using guanidinium chloride as denaturation agent have suggested that the C-domain is the most stable domain, although contradicting results have been reported (9 -13). In this context, the PGK from the Antarctic bacterium Pseudomonas sp. TACII18 (PsPGK) represents a unique model to study stability domains in PGK. The high level of amino acid sequence similarity of PsPGK with phosphoglycerate kinases from other species has allowed the building of a homology model based on the available crystal structures (14). Furthermore, crystals of PsPGK have been obtained (15), and its structure solved at 2.1 Å resolution indeed reveals a general fold identical to that of other monomeric PGKs. 3 However, the preliminary characterization of PsPGK by DSC (14) has shown that it possesses two well defined transitions in the unliganded conformation (see also Fig. 1). It follows that the enzyme contains two types of structures or regions differing by their intrinsic stability and responsible for the occurrence of a heat-labile domain (melting point (T m ) ϳ50°C) and a heat-stable domain (T m ϳ60°C) when thermal unfolding is recorded by DSC. Such stability domains within the protein may or may not correspond to the structural N-and C-domains. One cannot exclude for instance that each structural domain contains both heatlabile and heat-stable structures that unfold sequentially. We report here the unusual stability pattern of this enzyme and the complex effects of substrate binding inducing hinge-bending motions. The results have several implications in folding studies of PGK and of proteins in general as well as in the field of enzyme adaptations to extreme temperatures.

EXPERIMENTAL PROCEDURES
Mutagenesis of PsPGK-The pgk gene from the Antarctic bacterium Pseudomonas sp. TACII18 has been cloned in the expression vector pET22b as previously described (14). Genetic engineering was performed on the gene or on gene fragments inserted in the plasmid pSP73. Mutations were introduced by polymerase chain reaction as described (16). The mutation Trp 305 3 Phe was introduced by changing the codon TGG into TTC on the gene fragment generated by SalI-PstI digestion. The mutation Phe 135 3 Trp was introduced by changing the codon TTT into TGG on the gene fragment generated by XbaI-SalI digestion. Deletion of domain coding regions was carried out by inverse PCR using silent primers, 24 nucleotides in length, starting in 5Ј at codon for Thr 1 (reverse primer) and for Ala 172 (forward primer) for the * This work was supported by the Fonds National de la Recherche Scientifique, Belgium (Grants FRFC 2.4515.00 and 2.4536.04). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. deletion of the N-domain and starting in 5Ј at codon for Leu 170 (reverse primer) and at the stop codon (forward primer) for the deletion of the C-domain. The amplification products were purified (QIAquick PCR purification kit, Qiagen), phosphorylated by T4 polynucleotide kinase and circularized by ligation before transformation. The sequence of these constructions was checked by double-strand sequencing on an ALF DNA sequencer Express II (Amersham Biosciences) and transferred to the expression vector using appropriate restriction sites.
Protein Expression and Purification-The recombinant proteins were expressed in Eschericha coli BL21(DE3) grown at 18°C as described previously (14), except that Terrific broth was used for cultivation and induction with 0.5 mM isopropyl 1-thio-␤-galactopyranoside was performed at A 600 nm of the culture between 3 and 5. The recombinant proteins were purified essentially as described (14) following high pressure or freeze-thawing cell disruption, protamine sulfate precipitation of nucleic acids, ammonium sulfate fractionation, phenyl-Sepharose, and Q-Sepharose chromatography. The double mutant Phe 135 3 Trp/Trp 305 3 Phe was precipitated at 50% ammonium sulfate saturation, whereas the recombinant isolated domains were brought to 30% ammonium sulfate saturation before phenyl-Sepharose chromatography. For the domain proteins, the buffer was 50 mM Tris, 0.5 mM EDTA, 5 mM dithiothreitol, 0.1 mM benzamidine, pH 7.5, and a second Q-Sepharose chromatography was required. All the proteins were pure on the basis of a single band on SDS-PAGE, a single N-terminal amino acid sequence, and a single ion distribution in mass spectrometry.
Analytical Procedures-N-terminal amino acid sequences were determined using a pulsed liquid-phase protein sequencer (Procise 492, Applied Biosystems, PerkinElmer Division) fitted with an on-line phenylthiohydantoin analyzer. Electrospray mass spectrometry of the proteins (5-10 M) was performed on a Q-TOF Ultima mass spectrometer (Micromass). Absorbance spectra were recorded on an Uvikon XS spectrophotometer (Bio-Tek Instruments). Fluorescence spectra were recorded on an SLM-Aminco model 8100 spectrofluorimeter (Spectronic Instruments). Circular dichroism spectra were acquired under constant nitrogen flush using a J-810 spectropolarimeter (Jasco Corp.) in a 0.2-cm path length cell for the far-UV region. All spectrophotometric analyses were performed at 20°C.
Differential Scanning Calorimetry-Measurements were performed using a MicroCal MCS-DSC instrument as detailed (17). Samples (ϳ3 mg/ml) were dialyzed overnight against 30 mM Mops, pH 7.5. Thermograms of the isolated domains and for reversibility tests were recorded in the presence of 0.5 M 3-(1-pyridinio)-1-propanesulfonate to delay aggregation. Thermograms were analyzed according to a non-two-state model in which the T m , the calorimetric enthalpy (⌬H cal ), and the van't Hoff enthalpy (⌬H vH ) of individual transitions are fitted independently using the MicroCal Origin software (version 2.9). The magnitude and source of the errors in the T m and enthalpy values have been discussed elsewhere (18). Fitting standard errors on a series of three DSC measurements made under the same conditions in the present study were found to be Ϯ0.1 K on T m and Ϯ1% on both enthalpies. Fitting standard errors on four batches of purified PsPGK were found to be Ϯ 1 K on T m and Ϯ 10% on both enthalpies.

RESULTS
General Properties-The thermogram of PsPGK (Fig. 1a) shows that its heat-induced unfolding is characterized by two distinct heat absorption peaks separated by a temperature interval of about 10°C (TABLE  ONE) and henceforward referred to as the heat-labile (T m ϭ 49.8°C) and the heat-stable (T m ϭ 60.6°C) transitions (with T m , the melting point at the top of the transition). In our experimental conditions, no aggregation occurred in the heated sample and a stable post-transition baseline was recorded allowing the calculation of a progress baseline accounting for both transitions (Fig. 1a). On this basis, an estimation of the heat capacity difference between the unfolded and the native states can be made for the whole molecule (⌬C p ϭ 7.47 kcal mol Ϫ1 K Ϫ1 ), for the heat-labile (⌬C p ϭ 5,93 kcal mol Ϫ1 K Ϫ1 ) and the heat-stable (⌬C p ϭ 1,54 kcal mol Ϫ1 K Ϫ1 ) transitions. A ⌬C p value of 7.5 kcal mol Ϫ1 K Ϫ1 has been reported for yeast PGK (11), and the specific value of 19.3 cal/mol residue for PsPGK (386 amino acid residues) is in perfect agreement with data reported for large proteins (17). As far as heat absorption during unfolding is concerned, the normalized surface (⌬H cal , corre-  sponding to all enthalpy-driven interactions disrupted during unfolding) below the heat-labile transition is nearly twice that of the heatstable transition ( Fig. 1b; TABLE ONE), indicating that a large fraction of the native PsPGK conformation has a low stability. As shown in Fig. 2, thermal unfolding of the heat-labile transition is fully reversible. In addition, the ⌬H cal /⌬H vH ratio for this transition (where ⌬H vH is the van't Hoff enthalpy calculated from the slope of the transition) is ϳ1.5, indicating that the heat-labile transition may be composed by more than one cooperative unit, as confirmed below. By contrast, thermal unfolding of the heat-stable transition is not reversible and, as a result, the ⌬H cal /⌬H vH ratio decreases to ϳ0. 35. It should be noted that unfolding of the latter transition drastically reduces the reversibility of the heat-labile transition (Fig. 2). This suggests that the protein regions involved in both transitions are structurally connected. Addition of 5 mM dithiothreitol does not improve the reversibility, indicating that oxidation of the three free cysteines (Cys 56 , Cys 132 , and Cys 207 ) is not a major determinant of the irreversible unfolding. Mutagenesis Study of the Structural Domains-PsPGK contains one single Trp residue at position 305 in the C-domain. As shown by heatinduced unfolding monitored by fluorescence (excitation wavelength: 295 nm), this Trp is quenched in its microenvironment. Indeed, unfolding results in fluorescence increase (Fig. 1c) and the appearance of a hyperfluorescent intermediate. The temperature of maximal dequenching closely corresponds to the melting point of the heat-labile transition observed by DSC. The mutation Trp 305 3 Phe totally abolishes the dequenching transition (data not shown) therefore ascertaining the assignment of the fluorescence signal to Trp 305 . Interestingly, the DSC thermogram of this mutant displays an altered heat-labile transition (Fig. 3), as indicated by the decrease of T m (Ϫ5°C) and of ⌬H cal (Ϫ30%), whereas the heat-stable transition remains almost unaffected. One can conclude that the C-domain mainly contributes to the heat-labile transition. On the other hand, the mutation Phe 135 3 Trp has been introduced in the N-domain and the DSC trace of this mutant (Fig. 3) also exhibits a drastic destabilization of the heat-labile transition (T m Ϫ8°C, ⌬H cal Ϫ50%). These results indicate that both N and C structural domains of PsPGK contribute to the heat-labile DSC transition. Accordingly, each DSC transition of PsPGK cannot be assigned to the unfolding of a specific structural domain. One can also note a slight increase of the T m for the heat-stable transition in both mutants. This can be tentatively related to the relaxation of weak interactions between interacting domains in the protein when one of these domains is destabilized (9,11).
When both mutations are combined in the double mutant Phe 135 3 Trp/Trp 305 3 Phe, it is expressed and purified as an inactive and partly folded protein, as demonstrated by the weak DSC heat  absorption peak (Fig. 3, inset). It was found that after addition of 20 mM sulfate ions, the double mutant recovered 20% of the wild-type PsPGK activity. In the presence of sulfate, the DSC thermogram of the mutant exhibits two slight transitions (Fig. 3, inset). This indicates that the sulfate ion behaves like a folding effector of PsPGK.
The Isolated N and C Structural Domains-To further characterize the contribution of the PsPGK structural domains in both DSC transitions, the isolated N-and C-domains were expressed in E. coli from the pgk gene truncated in the middle of the first hinge helix. The plasmid construct coding for the N-domain contained the nucleotide sequence from Met-Thr 1 to Leu 170 (mature protein numbering; 18,107.9 Da for 170 aa), whereas the sequence for Met-Thr 1 -Ala 172 to Ala 386 was used for the C-domain (21,996.3 Da for 216 aa). It should be recalled that because the C-terminal end of PGK folds into the N-domain, truncation in the middle of the gene results in the expression of an N-domain devoid of ϳ15 residues that remain attached to the recombinant C-domain. N-terminal amino acid sequencing and mass spectrometry analyses (N-domain: 18,107 Da; C-domain: 21,999 Da) indicated that the initiating formylmethionine has been removed and that no other posttranslational modification occurred during the expression and purification processes.
The UV absorbance spectra well reflect the aromatic amino acid content of the analyzed proteins. The N-domain (six Phe, three Tyr) lacking Trp has a low absorption coefficient (⑀ 280 nm ϭ 3,960 M Ϫ1 cm Ϫ1 ) and its spectrum is dominated by Tyr residues, with three slight shoulders from Phe between 260 and 270 nm (Fig. 4a). Having an additional Trp residue, the absorption coefficient of the C-domain (nine Phe, three Tyr, one Trp) is much higher (⑀ 280 nm ϭ 9,530 M Ϫ1 cm Ϫ1 ). An equimolar mix of both N-and C-domains restores the absorbance value of the native PsPGK but the imperfect superimposition of the spectra indicates slight differences in the aromatic microenvironment within the recombinant domains. The fluorescence emission spectrum of PsPGK (Fig. 4b) is poorly defined as a result of the above-mentioned quenching of Trp305. Similarly, the N-domain fluorescence spectrum is dominated by Tyr residues. However, the pronounced red shift of the C-domain fluorescence spectrum shows that Trp 305 is dequenched and therefore, this residue is not in its native microenvironment in the recombinant domain. It follows that the equimolar mix of both domains restores the maximal fluorescence intensity of the native PsPGK but at a higher emission wavelength. The far-UV circular dichroism spectrum of the native PsPGK (Fig. 4c) displays two strong ellipticity minima at 208 and 222 nm indicating a high content of secondary structures. The spectra of both isolated N-and C-domains also present these minima but at a weaker ellipticity, especially for the N-domain. As a result, the equimolar mix of both domains reveals a lower secondary structure content as compared with the native enzyme. Finally, the near-UV CD spectra indicate an asymmetric environment of aromatic residues (and possibly of cysteines) typical of a folded conformation in the three proteins and the equimolar mix (Fig. 4d). From these spectrophotometric analyses, it can be concluded that the isolated N-and C-domains possess well organized secondary structures and a defined tertiary structure which however differ from the native state of PsPGK.
Nevertheless, the DSC thermograms of these isolated domains reveal their individual contribution to the biphasic profile of the native PsPGK (Fig. 5). The recombinant C-domain unfolds at a temperature closely corresponding to the heat-stable transition and accounts for about 80% of heat absorption (⌬H cal ϭ 48 kcal mol Ϫ1 ). By contrast, the N-domain is strongly destabilized and only accounts for less than 10% (⌬H cal ϭ 11 kcal mol Ϫ1 ) of the heat-labile transition. It seems therefore that only the C-domain contributes to the heat-stable DSC transition, whereas the structures of both domains responsible for the heat-labile DSC transition have not a compact conformation in the recombinant fragments. The spectroscopic and calorimetric data therefore indicate that the heat-labile structures in the recombinant domains tend to have molten globule-like properties, i.e. a defined but non-compact conformation with a low intrinsic conformational stability (Figs. 4 and 5). This contrasts with the well folded isolated domains obtained from more stable PGKs (10,19,20).
Interestingly, the isolated C-domain still binds the substrate Mg 2ϩ -ADP (Fig. 5, inset, and TABLE ONE), as revealed by the large (⌬H cal ϩ50 kcal mol Ϫ1 ) and sharp transition at a higher T m (ϩ2°C). On the other hand, both 3-PGA and sulfate seem to bind to the isolated N-domain, but these effects cannot be quantified as a result of the weakness of the heat absorption peak.
Substrate-induced Stability Changes-Crystallographic studies have shown that the PGK from Trypanosoma brucei can be trapped in the closed, active conformation (7) upon incubation with one substrate (Mg 2ϩ -ADP) and one product (3-PGA). In these conditions, the DSC thermogram of PsPGK (Fig. 6) reveals that substrate binding and the resulting hinge-bending motions drive the free (open) PsPGK into a closed, compact and uniformly more stable conformation, as indicated by the single sharp transition at a higher T m . The ⌬H cal /⌬H vH ratio of ϳ0.9 also indicates that unfolding of the closed conformation is cooperative and approaches a two-state mechanism which however is poorly reversible (ϳ6%; data not shown). The nonhydrolyzable nucleotide analog AMP-pnp has been used to study hinge bending but the crystal structures of the porcine 3-PGA-AMP-pnp (21) and the yeast 3-PGA-AMP-pnp (22) complexes do not adopt a closed conformation. As a matter of fact, the thermogram of PsPGK incubated with Mg 2ϩ AMPpnp and 3-PGA (Fig. 6) is indicative of a less compact and stable conformation (estimated from the sharpness and Tm of the transition, respectively) than that recorded for the closed conformation. It should also be noted that in the absence of Mg 2ϩ , the conformational changes promoted by ADP and 3-PGA are less pronounced (Fig. 6), underlining that the true physiological substrate is the complex Mg 2ϩ -ADP.
Crystallographic studies have shown that the individual substrates do not induce the closed conformation (5,6). However, the DSC thermograms of PsPGK incubated with one substrate (Fig. 6) display drastic alterations of both the heat-labile and heat-stable transitions, revealing significant modifications of the structure energetics. Finally, the DSC traces of PsPGK in the presence of Mg 2ϩ -ADP and of Mg 2ϩ -ATP are markedly different suggesting that upon phosphoryl transfer to ATP, the enzyme undergoes a noticeable conformational change. It should be  noted that all these substrate-induced stability changes only affect the T m values and the unfolding cooperativity of the transitions. Indeed, as shown in TABLE ONE, there is no significant variation of the surface of the transitions (⌬H cal ) for the binary and ternary PsPGK complexes, whereas the sharpness (⌬H vH ) and T m of the transitions are modified.
Substrate and Ion Binding-The available crystal structures of binary and ternary PGK complexes and a multiple sequence alignment allowed to identify the highly conserved residues of PsPGK involved in substrate and ion binding. 3-PGA binds to the N-terminal domain (residues Asp 20 , Asn 22 , Arg 35 , His 58 , Arg 61 , Arg 112 , Gly 141 , Arg 145 , Gly 362 , Gly 363 , and Ala 364 ) in the so-called basic patch, whereas ADP binds to the C-terminal domain (residues Ser 191 , Lys 196 , Ile 282 , Leu 283 , Asn 306 , Glu 313 , Gly 340 , and Thr 342 ). Fig. 7 shows the effect of increasing concentrations of both substrates on the DSC thermograms of PsPGK. It can be seen that upon 3-PGA binding, the T m of the heat labile transition is displaced toward higher values whereas the heat-stable transition remains unaffected. This confirms that the structural N-domain bearing the 3-PGA binding site belongs to the heat-labile DSC transition. On the other hand, upon ADP binding the stability of the heat-labile transition is improved, but in this case, the surface (⌬H cal ) of the heat-stable transition is also increased. This again confirms that the structural C-domain bearing the ADP binding site contributes to both the heatlabile and heat-stable DSC transitions. Interestingly, a plot of the T m increase for the heat-labile transition as a function of the substrate concentration can be fitted to a binding isotherm (Fig. 7, insets) providing apparent K d values of 0.51 mM for 3-PGA and 0.38 mM for ADP. These values are in agreement with the K m values previously determined by standard kinetic methods (14).
The above-mentioned experiments have been performed in the absence of Mg 2ϩ . This ion was shown to interact with an acidic side chain of the ADP binding site (7), corresponding to Asp 341 in the C-domain of PsPGK. Fig. 8 shows that increasing concentrations of MgCl 2 only affect the heat-stable DSC transition by decreasing its T m . At MgCl 2 concentrations higher than 2 mM, the heat-stable transition tends to fuse with the heat-labile DSC transition that remains unaltered but sample precipitation precluded further analysis of this phenomenon. These results clearly demonstrate that the Mg 2ϩ -binding site belongs to the heat-stable DSC transition. Finally, the sulfate ion can be viewed as a test case for our assignment of the DSC transitions because it has been reported to bind both to the N-(Arg 61 ) and C-(Lys 192 ) domains (23). And indeed, increasing Na 2 SO 4 concentrations affect both the heat-labile and heat-stable DSC transitions (Fig. 8), which fuse in a conformation mimicking the substrate-induced closed conformation, however, at a lower T m .

DISCUSSION
Stability and Structural Domains-Several multidomain proteins display distinct stability domains resulting in multiphasic DSC thermograms (24). In some cases, a close correlation was found between structural and stability domains, i.e. each DSC transition can be assigned to the unfolding of a specific structural domain (25)(26)(27)(28)(29). However, PsPGK clearly demonstrates that such correlation is not always valid as, in this case, we have shown that the heat-stable DSC domain belongs to the structural C-domain, whereas both N and C structural domains contribute to the main heat-labile DSC domain. The size of the heat-stable domain in PsPGK can be estimated to 100 residues in a first approximation, assuming that all 386 residues contribute equally to ⌬H cal of the DSC transitions (TABLE ONE). A more precise estimation of 80 residues can be based on ⌬C p of the transitions (5.93 and 1.54 kcal mol Ϫ1 K Ϫ1 for the first and second transitions, respectively) because this parameter has been found relatively constant for proteins of similar size, whatever their intrinsic stability (17). The spectroscopic and mutagenesis studies as well as the effects of substrates and ions on the DSC transitions indicate that this heat-stable domain includes or is very close to the nucleotide binding site. If the stability pattern of the PsPGK structural C-domain (i.e. containing both heat-labile and heat-stable struc-tures) is extended to other homologous PGK, this can partly explain the contradicting results reported for the stability of the C-domain (9 -13).
Substrate-induced Stability Changes and Hinge-bending Motions-Considering the weak stability of the major DSC transition (Fig. 1), the free PsPGK appears to possess a relatively loose open conformation. Upon binding of a single substrate molecule, either Mg 2ϩ ADP or 3-PGA, the heat-labile structures evolve toward a more compact conformation (Fig. 7). This can be related to an induced-fit mechanism by which binding of the substrate induces a structural rearrangement of the binding site. It should be noted that this drastic modification of PsPGK conformational stability mainly involves the structure energetics because only discrete modifications have been reported between the crystal structures of free and single substrate-bound PGKs (5,6). The behavior of the isolated C-domain provides useful insights into the energetics of hinge-bending motions in PsPGK. Indeed, Fig. 5 (inset) and TABLE ONE show a drastic increase of ⌬H cal and of T m for this isolated domain upon Mg 2ϩ -ADP binding. An estimation of the conformational stability in the free and substrate-bound states can be obtained using Equation 1.
The validity of this approximation has been discussed elsewhere (30 -32). The computed stability curves indicate an increase of the conformational stability (free energy of unfolding) of 5.5 kcal mol Ϫ1 upon substrate binding. On the other hand, the free energy of substrate binding can be estimated to Ϫ5 kcal mol Ϫ1 using the K m value for the nucleotide (0.21 mM) and the following relation (Equation 2).
Comparison of both free energy estimates strongly suggests that the nucleotide binding free energy is transferred into conformational stability. However, such behavior of the heat-stable transition is not observed in the native PsPGK upon ADP (Fig. 7) and Mg 2ϩ (Fig. 8) binding. One can therefore propose that in the native enzyme, the gain in conformational stability arising from substrate binding is now converted into the energy required for hinge-bending motions and domain closure. The lack of ͚⌬H cal difference between the free and closed forms of PsPGK (TABLE ONE) also supports this view. Finally, in a thoughtful study of PGK activation, the authors concluded that it cannot be decided whether anion activation is a consequence of domain closure or anion binding itself mediates the closure by connecting the two domains (23). From our results, it seems that the latter hypothesis is valid, as demonstrated by the apparent domain closure induced by the sulfate anion activator (Fig. 8).
Activity at Low Temperatures-The unusual stability pattern of PsPGK should also be considered in the context of the enzyme adaptations to catalysis at low temperatures. In order to catalyze the essential glycolytic step at temperatures close to 0°C found in the Antarctic environment, this enzyme possesses a high specific activity to compensate for the reduction of chemical reaction rates at such low temperatures. In addition, PsPGK displays a high affinity for the nucleotide, therefore improving the k cat /K m ratio which is the relevant parameter for intracellular enzymes (14). It has been argued that the high activity of cold-adapted enzymes is gained through an improved flexibility or mobility of the active site and of the structures bearing the active site residues in order to facilitate molecular motions in low thermal energy conditions (32)(33)(34). In the case of PsPGK, the active site is formed by the  DECEMBER 16, 2005 • VOLUME 280 • NUMBER 50 interface of the N-and C-domains that have to move to bring the substrates in close proximity. Accordingly, the predominant weak stability of this enzyme can enhance the substrate-induced conformational changes (Figs. 6 and 7) and the resulting hinge-bending motions at low temperatures. It has also been suggested that during the course of evolution, some psychrophilic enzymes have reached a state close to the lowest stability compatible with the native state (17,33,35). This is supported here by the drastic effects of the single conservative mutations introduced in both N-and C-domains (Fig. 3), by the almost unfolded state of the double mutant (Fig. 3, inset) and by the lack of compact conformation of the structures forming the heat-labile DSC transition in the isolated domains (Figs. 4 and 5). As far as substrate binding is concerned, most psychrophilic enzymes have a reduced affinity arising from the active site mobility (33,36), whereas PsPGK has a better affinity for the nucleotide. One can therefore propose that the heat-stable domain of PsPGK provides a more compact and structured nucleotide binding site which, consequently, has a high affinity. This seems to be an elegant molecular adaptation allowing the optimization of both k cat and K m at low temperatures, not previously observed in other psychrophilic enzymes.