The X-ray Crystal Structures of Human α-Phosphomannomutase 1 Reveal the Structural Basis of Congenital Disorder of Glycosylation Type 1a*

Congential disorder of glycosylation type 1a (CDG-1a) is a congenital disease characterized by severe defects in nervous system development. It is caused by mutations in α-phosphomannomutase (of which there are two isozymes, α-PMM1 and α-PPM2). Here we report the x-ray crystal structures of human α-PMM1 in the open conformation, with and without the bound substrate, α-d-mannose 1-phosphate. α-PMM1, like most haloalkanoic acid dehalogenase superfamily (HADSF) members, consists of two domains, the cap and core, which open to bind substrate and then close to provide a solvent-exclusive environment for catalysis. The substrate phosphate group is observed at a positively charged site of the cap domain, rather than at the core domain phosphoryl-transfer site defined by the Asp19 nucleophile and Mg2+ cofactor. This suggests that substrate binds first to the cap and then is swept into the active site upon cap closure. The orientation of the acid/base residue Asp21 suggests that α-phosphomannomutase (α-PMM) uses a different method of protecting the aspartylphosphate from hydrolysis than the HADSF member β-phosphoglucomutase. It is hypothesized that the electrostatic repulsion of positive charges at the interface of the cap and core domains stabilizes α-PMM1 in the open conformation and that the negatively charged substrate binds to the cap, thereby facilitating its closure over the core domain. The two isozymes, α-PMM1 and α-PMM2, are shown to have a conserved active-site structure and to display similar kinetic properties. Analysis of the known mutation sites in the context of the structures reveals the genotype-phenotype relationship underlying CDG-1a.

Glycoproteins play fundamental roles in humans (1,2) where they are active in cell recognition and adhesion, development, and homeostasis (3)(4)(5)(6)(7). Defects in protein glycosylation lead to a variety of disease states known collectively as congenital disorders of glycosylation (CDG). 2 The most common CDG, congenital disorder of glycosylation type 1a (CDG-1a) (8), manifests in infancy with physical and developmental abnormalities. Numerous clinical genotypes have been identified, with the range of phenotypes extending from mild to severe (9,10). CDG-1a is caused by mutations in the gene encoding ␣-phosphomannomutase (␣-PMM) (11)(12)(13). The most common mutant, R141H (40% of disease alleles of CDG-1a patients), has a high carrier frequency of 1/70 in the general population. Individuals homozygous for this mutation have not been identified, suggesting that this genotype is lethal (12).
Human ␣-PMM, which catalyzes the conversion of D-mannose 6-phosphate (Man-6-P) to ␣-D-mannose 1-phosphate (␣-Man-1-P) (see Fig. 1) is required for GDP-mannose and dolichol-phosphate-mannose biosynthesis. Two isoforms are found; PMM2 is expressed in all tissues, whereas expression of PPM1 (65% sequence identity with PMM2) is restricted to the brain and lungs (14). ␣-PMM belongs to the haloalkanoic acid dehalogenase superfamily (HADSF), distinguishing it both structurally and mechanistically from the phosphomanno/phosphoglucomutase enzymes of the phosphohexomutase superfamily that are integral to glycolysis in prokaryotes and eukaryotes (15) and to alginate biosynthesis in Gram-negative bacteria (16).
Within the HADSF, ␣-PMM is distinguished from the bacterial ␤-phosphoglucomutase (␤-PGM) in its C(1) anomer specificity as well as in the position and fold of the cap domain which serves as a lid over the conserved catalytic core domain. In HADSF members, wherein phosphatases predominate, the conserved aspartate nucleophile of the core domain mediates the transfer of the phosphoryl group. Remarkably, the ␣-PMM and ␤-PGM evolved independently to arrive at an adaptation of the phosphatase scaffold, which allows the aspartylphosphate of the core domain to discriminate between solvent and the sugarphosphate substrate as the phosphoryl group acceptor.
Herein we report the x-ray crystallographic structures of ␣-PMM1 in the Mg 2ϩ -bound form and in complex with the substrate of the reaction, ␣-Man-1-P. These structures are analyzed in the context of the coordinates for uncomplexed ␣-PMM2 deposited in the Protein Data Bank by the Center for Eukaryotic Structural Genomics (accession code 2AMY) to identify active-site residues with roles in binding and catalysis as well as residues involved in subunit dimerization. The structure and kinetic analyses of ␣-PMM1 and ␣-PMM2 provide for the first time insight into how quaternary structure, stability, and catalytic efficiency are altered in clinically important mutants.

EXPERIMENTAL PROCEDURES
Materials-Except where indicated, all chemicals were obtained from Sigma. Primers, T4 DNA ligase, and restriction enzymes were from Invitrogen. The Geneclean spin kit and the Qiaprep spin miniprep kit were from Qiagen. Host cells were purchased from Novagen. The Biolmol green phosphate assay kit was purchased from Biolmol Research Laboratories Inc. The ␣-mannose 1,6-bisphosphate used in the kinetic assays was prepared according to published procedure (13).
␣-PMM-1 was purified from E. coli Rosetta TM (DE3) cells containing the pET-28b(ϩ)-pmm1 plasmid. Transformed cells (4L) were grown at 37°C in Luria broth (or minimal medium containing selenomethionine) containing 50 g/ml ampicillin and induced for 4 h with 0.4 mM isopropyl ␤-D-thiogalactopyranoside. Cells were harvested by centrifugation and the pellet (11 g) was suspended in 100 ml of ice-cold lysis buffer (50 mM Hepes, pH 7.5, 5 mM DTT). Cells were lysed by sonication (6 ϫ 6 min at 50% power and 30% duty cycle) with a Branson sonifier 250 (VWR Scientific Inc.). The cell lysate was centrifuged at 4°C for 60 min at 18,000 rpm (38,384 ϫ g). The protein precipitated with 40 -50% (NH 4 ) 2 SO 4 was harvested by centrifugation at 15,000 rpm (26,895 ϫ g) for 15 min. Following dissolution in 25 ml of lysis buffer and dialysis against buffer A (50 mM Hepes, pH 7.5, 0.5 mM DTT) overnight, the protein was loaded onto a 100-ml DEAE-Sepharose column (Amersham Biosciences) pre-equilibrated with buffer A. The column was washed with 100 ml of buffer A, and eluted with a 0.7-liter linear gradient of 0 -0.5 M NaCl in buffer A. Fractions containing ␣-PMM1 were combined and (NH 4 ) 2 SO 4 was added to 1 M. The solution was centrifuged at 15,000 rpm (26,895 ϫ g) for 5 min to remove precipitated protein. The supernatant was loaded onto a 25-ml butyl-Sepharose column (Amersham Biosciences) pre-equilibrated with 1 M (NH 4 ) 2 SO 4 in buffer A. The column was washed with 50 ml 1 M (NH 4 ) 2 SO 4 in buffer A and then eluted with a 200-ml linear gradient of 1 to 0 M (NH 4 ) 2 SO 4 in buffer A. ␣-PMM1 was concentrated and stored at Ϫ80°C. The protein, shown to be Ͼ95% pure by SDS-PAGE, was obtained in a yield of 2 mg of protein/g of cell paste.
Selenomethionine-substituted ␣-PMM1 was expressed in E. coli BL21-CodonPlus cells. The cells were grown in minimal medium pre-pared by mixing 200 ml of 20ϫ M9 (10 g of NH 4 Cl, 30 g of KH 2 PO 4 , 60 g of Na 2 HPO 4 , 500 ml of H 2 O, pH 7.5), 20 ml of amino acid mix I (the 20 standard amino acids excluding Met, Tyr, Trp, and Phe, each at 4 mg/ml), 40 ml of amino acid mix II (Tyr, Trp, and Phe, each at 2 mg/ml in distilled/deionized H 2 O, pH 9.0), 1 ml of vitamin mixture (riboflavin, niacinamide, pyridoxine monohydrochloride, and thiamine, each at 1 mg/ml), 4 ml of 1 M MgSO 4 , 4 ml of 12.5 mg/ml FeSO 4 , 20 ml of 40% (w/w) glucose, and 4 ml of 10 mg/ml selenomethionine in a final volume of 2 liters. Chloramphenicol (16.5 mg/liter) and kanamycin (50 mg/liter) were added to the culture. The culture was grown at 30°C with shaking at 200 rpm until the A 600 reached 0.8 -1.1 (24 h), at which point ␣-PMM1 expression was induced by adding IPTG to a final concentration of 0.4 mM. The cells were grown for an additional 10 h at 25°C and harvested by centrifugation. The protein was purified using the procedure described above, except that the DTT concentration was increased to 10 mM, and 1 mM EDTA was added throughout.

Structural Basis of CDG1a
Structure Determination and Model Refinement-Because structure determination by molecular replacement was unsuccessful, the structure was phased by the multiple wavelength anomalous diffraction method, using three 2.1 Å SeMet data sets. The selenium substructure was solved using BnP (18), which found three of the four selenium sites in the protein. The sites were confirmed by comparison with the anomalous difference Patterson map. The N-terminal methionine, along with the next 10 residues, was disordered. An initial model was built using RESOLVE (19) and improved by manual rebuilding in the graphics suite COOT (20). This partial model was refined in the REFMAC5 (21) module of the CCP4 suite (22) using the Hendrickson-Lattman coefficients from BnP. Electron density maps calculated from the combined phases were of very high quality and permitted rapid rebuilding of the model. Iterative cycles of restrained refinement in REFMAC5 and rebuilding in COOT ultimately yielded a model containing 244 of 262 residues, 171 water molecules, and 2 Mg 2ϩ ions. The quality of the model at this point was assessed by MOLPROBITY (23), and 235 residues (96.3%) were in the favored regions, 9 residues were in the additionally allowed regions, and no residues were in the disallowed regions.
The ligand-bound structure was solved by Fourier difference methods using phases from the 2.1 Å SeMet native model. After rigid body refinement in REFMAC5 the electron density for Man-1-P was clearly defined. The Man-1-P was not modeled until the model reached an R work of 22.3% and an R free of 27.4%. After several more rounds of model building and refinement, the final ␣-PMM1⅐Man-1-P model was comprised of residues 13-258 of ␣-PMM1, 235 water molecules, two Mg 2ϩ ions, and one molecule of Man-1-P. Data collection and refinement statistics for both models are reported in Table 1.
Steady-state Kinetics-All 1-ml kinetic assays were carried out at 25°C in 50 mM K ϩ Hepes, pH 6.5, containing 10 mM MgCl 2 . The initial velocity of ␣-PMM catalyzed Glc-6-P formation from ␣-Glc-1-P was measured in the presence of 50 M ␣-Glc-1,6-bisP, 0.2 mM NADP, and 5 units/ml Glc-6-P dehydrogenase by monitoring the increase in absorbance at 340 nm (⌬⑀ ϭ 6.2 mM Ϫ1 cm Ϫ1 ) resulting from the Glc-6-P dehydrogenase catalyzed reduction of NADP. The initial velocity of ␣-PMM-catalyzed Man-6-P formation from ␣-Man-1-P was measured in the presence of 50 M ␣-Man-1,6-bisP (13), 5 units/ml glucose-6phosphate dehydrogenase, 5 units/ml phosphomannose isomerase, 5 units/ml phosphoglucose isomerase by monitoring the increase in absorbance at 340 nm (⌬⑀ ϭ 6.2 mM Ϫ1 cm Ϫ1 ) resulting from the Glc-6-P dehydrogenase-catalyzed reduction of NADP. The k cat and K m values were determined from the initial velocity data using the equation is the ␣-Glc-1-P or ␣-Man-1-P concentration, V 0 is the initial velocity, V m is the maximum velocity, and K m is the Michaelis constant. The k cat was calculated from the ratio of V max and the ␣-PMM concentration.

RESULTS AND DISCUSSION
Overall Structure-Recombinant human ␣-PMM1 was purified by conventional chromatographic techniques and crystallized from a solution of polyethylene glycol 3350, MgCl 2 , ␤-mercaptoethanol, and DLmalic acid at pH 7.0 for x-ray structure determination. Both the native and SeMet ␣-PMM1 crystallized in space group P4 3 2 1 2 (unit cell dimensions a ϭ b ϭ 50.9 Å, c ϭ 213.0 Å) with one molecule in the asymmetric unit. Molecular replacement using the structure of ␣-PMM2 (65% identical, Protein Data Bank accession code 2AMY, unpublished submission by the Center for Eukaryotic Structural Genomics) as the search model did yield a viable solution; however, the resulting model did not refine well, and several surface loops appeared to be disordered. To obtain experimental phases, the structure was determined using multiple wavelength anomalous diffraction phasing of the selenomethionine-substituted protein. The final model contained 245 of 262 residues. The first 11 residues at the N terminus and the last 6 residues at the C terminus were disordered. The refined model of the Mg 2ϩ -bound form of ␣-PMM1 was used to phase the structure of ␣-PMM1 in complex with ␣-D-mannose 1-phosphate (␣-PMM1⅐Man-1-P) obtained by soaking the Mg 2ϩ -bound crystals with ␣-Man-1-P. ␣-PMM1, like other HADSF phosphotransferases, is comprised of two domains, a core domain and a smaller cap domain (Fig. 2). The overall dimensions of the monomer are ϳ63 ϫ 31 ϫ 42 Å. The core domain (residues 1-90 and 198 -262) consists of a six-stranded parallel ␤-sheet (␤1-␤5 and ␤10-␤11) flanked by five ␣-helices (␣1-␣2 and ␣7-␣9) (Fig. 2B) with typical ␤-␣-␤ connections. The cap domain (residues 95-194) is inserted between ␤5 and ␣7 of the core domain. The ␣␤␤(␣)␣␤␤ topology of the cap domain, as well as its position in the primary sequence between motifs 2 and 3 (see Ref. 24 for review), place this enzyme in subclass IIb of the HADSF. The secondary structure elements are arranged to form a four-stranded antiparallel ␤-sheet (␤6-␤7 and ␤8-␤9) topped by three ␣-helices (␣4 -␣6). Two flexible "hinge" regions comprising residues 91-94 and 195-197 connect the two domains. The hinge regions were identified using visual inspection of the structure together with the program FlexOracle (25). In both structures reported herein, the cap domain is in an open conformation, with the active site exposed to solvent.
The 29.7-kDa monomer forms a homodimer in solution as demonstrated by size-filtration chromatography. This dimer is observed in the crystal structure of ␣-PMM1, with the interface consisting exclusively of interactions between the cap domains (Fig. 3A). Two major interactions dominate the dimer interface: a helix-helix interaction between the first ␣-helices (␣4) and an antiparallel ␤-sheet interaction between the first ␤-strands (␤6) of the two cap domains. The interaction between the strands creates an extended eight-stranded ␤-sheet across the two monomers, holding the two core domains away from one another. Residues Leu 105 , Ile 109 , Leu 113 , Phe 128 , Ile 129 , and Phe 131 form a hydrophobic core at the dimer interface. Several hydrogen bonds and a salt bridge anchor the interface at each end (Fig. 3B). The same oligomerization interface can be observed in the structure of human ␣-PMM2, demonstrating that this is a conserved feature of human ␣-PMMs. One of the more common deleterious mutations of ␣-PMM2, F119L (Phe 128 in ␣-PMM1), is located in the center of the dimer interface (Fig. 3B), highlighting the potential importance of the quaternary structure for function and/or stability of the protein.
Based on previous studies of catalytic cycling in other HADSF enzymes, it is known that the cap domain must dissociate from the core domain to allow substrate to bind, then associate to seal the active site from aqueous solvent (26,27). A portion of the cap domain that would contact the core domain in ␣-PMM carries a large number of positively charged residues (Figs. 4 and 5). These include the three conserved residues Arg 150 , Arg 143 , and Arg 132 in ␣-PMM1 and Arg 141 , Arg 134 , and Arg 123 in ␣-PMM2. One might expect that there would be complementary negatively charged residues on the surface of the core domain; however, the corresponding surface of the core retains an overall positive charge contributed by conserved residues Arg 28 and Lys 58 in ␣-PMM1 and Arg 21 and Lys 51 in ␣-PMM2. Thus, rather than acting as a "clasp" these two surfaces may act as an electrostatic wedge retaining the correct interdomain orientation. Indeed, the importance of these residues to protein function is highlighted by the fact that the mutation R141H (Arg 150 in ␣-PMM1) has been identified as one of the major deleterious polymorphic mutations of ␣-PMM2.
Enzyme Active Site-The active site of ␣-PMM1 is located at the interface of the cap and core domains (Fig. 4). By analogy to other HADSF phosphatases, the key active-site residues are the catalytic nucleophile Asp 19 , the putative acid/base residue Asp 21 (contributed by motif 1), Ser 54 , which helps position the phosphoryl group (contributed by motif 2), the salt bridge residue Lys 198 (contributed by motif 3), and the metal-binding residue Asp 218 (contributed by motif 4). In both the dimer. The molecule has been rotated 90°on the horizontal axis relative to Fig. 2. The cap and core domains of monomer A are magenta and cyan, respectively, while the cap and core of monomer B are red and blue, respectively. The image was rendered using MOLSCRIPT (38) and POVRAY. B, schematic representation of the interactions between monomers at the dimer interface. For clarity, the distances are shown for only the right half of the dimer interface. Green circles denote residues involved in hydrogen bonds, blue circles denote residues involved in salt bridges, and yellow circles denote residues forming the hydrophobic core of the interface. Pro 122 , which imparts a kinked turn necessary for positioning Arg 125 , is shown as a white circle.
native and ␣-PMM1⅐Man-1-P structures, the acid/base residue Asp 21 , stationed two residues upstream from the nucleophile (thus denoted the Asp-plus-2 residue), is held in position for proton transfer by a hydrogen bond to residue Gln 62 contributed by ␣2, as has been noted previously for the corresponding residue in HADSF type IIA (28), type IIB (26), and type III (29) phosphatases. Because Asp 21 is positioned to activate a water molecule as well as the C(6)OH of ␣-Man-1-P for attack at the aspartylphosphate, it is clear that a different mechanism for ensuring that phosphomutase activity predominates over phosphatase activity is operative in ␣-PMM than the substrate induced-fit mechanism, which is operative in ␤-PGM. In the case of ␤-PGM, the Asp general acid/base residue is pinned to the linker in the cap-open conformation and to the active site in the cap-closed conformation. The bound substrate and not a water molecule functions to stabilize the cap-closed conformation for catalytic turnover.
Substrate Binding-The structure of ␣-PMM1 with ␣-D-mannose 1-phosphate bound identifies the active-site residues contributing to substrate specificity. Since this structure was obtained by soaking preformed crystals of ␣-PMM1 with Man-1-P, the domain motions required to close the active site could not occur, because the enzyme conformation was held fast by the crystal lattice. The Man-1-P-bound structure was surprising because it was anticipated from observations of  (38) and POVRAY. C, schematic representation of the interactions between ␣-PMM1 and Man-1-P (black). The residues of the cap and core domains and the primary and secondary substrate specificity loops are colored as described for B. The distances of the interactions are given in Angstroms.
phosphate (and phosphate analog) binding to other HADSF members that the electropositive environment created by the active-site Ser 54 , Lys 198 , and Mg 2ϩ would constitute a more competent phosphate-binding site than that of the cap domain (26,27,30). Thus, ␣-PMM1 is unique in that the phosphorylated substrate is seen bound primarily by interactions with the cap domain ( Figs. 2A and 4). The electron density corresponding to the ligand was observed at 3.0 in a ͉F o ͉ Ϫ ͉F c ͉ electron density map (Fig. 4A). The ␣-PMM1⅐Man-1-P structure can be said to represent a collision complex, the first encounter of enzyme and substrate, and not the Michaelis complex immediately preceding the first chemical step. Indeed, if the intermediate ␣-mannose 1,6-(bis)phosphate is overlaid with the Man-1-P substrate in the ␣-PMM1⅐Man-1-P complex, the C(6)-phosphate group is 10 Å from the Asp 19 nucleophile (that is, there is no phosphate group in the phosphoryl-transfer site). The implication of this observation is that substrate initially binds to the cap and not the core domain portion of the active site. Once the substrate is bound, there is presumably a change in the conformation of the hinge region that pushes the substrate into the active site as the cap binds to the core domain.
The surfaces of both domains at the Man-1-P-binding site carry an overall positive charge (Fig. 5). The residues constituting the Man-1-Pbinding site are positioned on two loops of the cap domain, termed the primary (residues 140 -162) and secondary (residues 181-190) substrate specificity loops. The corresponding loops have been described for BT4131, a homologous subclass IIb HADSF enzyme (26). ␣-PMM1 residues Arg 143 and Arg 150 , together with Ser 188 and the main chain amide of Met 186 , form the distal phosphate-binding site (Fig. 4, B and C). Other residues involved in Man-1-P binding are Asp 190 and Arg 132 , which hold the C-2 hydroxyl, and Arg 28 , the sole core domain contact, which interacts with the C-4 hydroxyl. Although Arg 132 is not part of either substrate specificity loop, it does directly contact the substrate. It is located on strand ␤6, which forms a major part of the dimer interface (see above) in both ␣-PMM1 and ␣-PMM2. The importance of this residue to substrate binding and its proximity to the dimer interface may explain, at least in part, the loss of activity observed when dimerization is disrupted by genetic mutation.
A Proposal for the Mechanism of Selective Phosphoryl Transfer-␣-PMM, like ␤-phosphoglucomutase, catalyzes the interconversion of the hexose 1-phosphate and hexose 6-phosphate via a hexose-1,6-(bis)phosphate intermediate generated by reaction with the phosphoenzyme (Fig. 1). For a HADSF enzyme to act as a phosphotransferase, as opposed to a phosphatase, the active site of the phosphorylated enzyme must be capable of discriminating between the substrate and water as the phosphoryl group acceptor. In both subfamily I and subfamily II HADSF enzymes, the cap domain closes on the active site, sealing it from solvent (31,32). In unliganded ␤-phosphoglucomutase, the Asp 10 acid/base is involved in hydrogen bond interactions with the main chain amides of two residues in the flexible hinge region. In this orientation, the Asp 10 acid/base is pointing away from the catalytic center, which prevents general base catalysis of phosphoryl transfer to water. Binding of substrate shifts the equilibrium between the open and closed forms of the enzyme, stabilizing the closed form. Upon cap closure, the ␤-phosphoglucomutase Asp 10 acid/base rotates into the active site to form a hydrogen bond with the C(6)OH of the ␤-glucose 1-phosphate. In this way, catalysis is linked to cap closure.
The unliganded and the MIP-bound structures of ␣-PMM1 suggest a unique mechanism of discrimination between substrate and solvent. The Asp 21 acid/base of ␣-PMM1 is engaged in a hydrogen bond with Gln 62 (2.9 Å) located on helix ␣2 of the core domain. In the cap-open conformation, Asp 21 is in position to function as an acid/base catalyst in the transfer of the phosphoryl group to and from the Asp 19 nucleophile. Thus, solvent discrimination is not accomplished by coupling the positioning of the Asp 21 acid/base catalyst with substrate induced cap closure. Rather, it must rely on transition-state stabilization by the unique active-site environment created in the cap-closed conformation. How, then, is substrate binding tied to cap closure? The ␣-PMM structures suggest that the positively charged surfaces at the interface of the cap and core domains (Fig. 5) observed in the unliganded enzyme may act as an electrostatic wedge. Binding of the negatively charged substrate would mitigate the electrostatic repulsion of the cap and core domains, favoring cap closure.
The Structural Effects of Clinically Observed Mutations-The correlation between mutation and ␣-PMM activity has been the object of only one study reported in 1999 by Pirard et al. (33). In this study, seven PPM2 mutants (D65Y, F119L, V129M, R141H, R162W, D188G, and V231M) were purified and their in vitro activity and thermal stability briefly surveyed. The R141H and D188G mutants proved to be the most catalytically impaired (0.4 and 2% of wild-type activity), correlating with the severity of the heterozygous R141H/D188G phenotype, while V231M proved to be the least stable at 40°C. Although not biochemi-  -and-stick) bound. A dot surface shows the overall charge of the cap and core domain surfaces at the Man-1-P-binding site. The image was rendered using CCP4MG (39).
Because of its limited tissue distribution, mutations to ␣-PMM1 have not been linked to pathology. However, the PMM1 and PMM2 isozymes are quite similar (65% identical; Fig. 6), and the variable residues that are scattered throughout the structure are found mainly at the protein surface. The overall structures of ␣-PMM1 and ␣-PMM2 are very similar with root mean square deviations of 0.65 Å (for the cap) and 1.0 Å (for the core) when the two domains are overlaid separately. The dimer interface, active-site motifs (Fig. 6), and the hinge regions are conserved.
The numerous ␣-PMM2 mutation sites known to be associated with the CDG Type 1a phenotype are not localized to one particular region of the structure but are instead distributed throughout (Fig. 7). Based on their locations in the structure, the clinically important mutations can be classified on the basis of whether substrate binding and catalysis, dimerization, or protein stability are targeted. The most severe mutations, R141H (Arg 150 in PMM1) and D188G (Asp 197 ) belong in the first category, as does R123G (Arg 132 ). As seen in the ␣-PMM1⅐Man-1-P structure, Arg 150 (Arg 141 in ␣-PMM2) is a crucial part of the distal phosphate-binding site, and loss of the positive charge at this location would impair substrate binding. This conclusion is in good agreement with the finding reported in Pirard et al. (14,36) that the K m for Man-1-P increased ϳ10-fold in the R141H mutant. The effect of the D188G (Asp 197 ) mutation is equally dramatic. This residue is adjacent to the motif 3 lysine, which is required for proper orientation of the nucleophilic aspartate. The R123G (Arg 132 ) mutation is also likely to effect substrate binding, because, as seen in the Man-1-P-bound structure, Arg 132 of ␣-PMM1 makes direct interactions with the C-2 hydroxyl of Man-1-P. Of the mutations that appear to disrupt the dimer interface (F119L (Phe 128 ) and P113L (Pro 122 )), the F119L mutation would be expected to have the greatest ipmact, given that this residue is a major component of the hydrophobic core that comprises the dimer interface (Fig. 3B). The P113L mutation might alter the turn that positions Arg 125 (Arg 116 in ␣-PMM2) for formation of a salt bridge at the dimer interface.
The last category, mutations effecting folding and/or stability of the polypeptide, includes V129M (Ile 138 in PMM1), V44A (Val 51 ), D65Y (Asp 74 ), and V231M (Val 240 ). These mutations occur in the interior of the cap or core domains (Fig. 7) and are therefore expected to be especially detrimental to native structure. Finally, one mutation, F183S (Phe 192 ) appears to be unique in that it may alter the interdomain motion required for catalysis. Phe 183 is positioned at the start of the turn forming the second linker region, and it is possible that changes to this  residue interfere with the conformational changes expected to accompany substrate binding.
Kinetic Analysis-The structures of ␣-PMM1 and ␣-PMM2 are nearly identical, and the results of the kinetic assays reinforce this observation ( Table 2). In addition, the kinetic analysis establishes the integrity of the recombinant PMM enzymes.
Both ␣-PMM1 and ␣-PMM2 are efficient catalysts with high specificity for glucose 1-phosphate and mannose 1-phosphate, as indicated by the magnitude of k cat /K m . In two instances the observed kinetic parameters differ significantly from those of previous studies (14,36). First, the K m of ␣-PMM1 for ␣-Man-1-P of 54 M is larger than the published value of 3.2 M. Nevertheless, this larger value is reasonable given that the average serum mannose concentration in humans is 55 M (range 35-70 M) (37). Also, Pirard et al. (14) reported that ␣-PMM2 is 10-fold more reactive with ␣-Man-1-P than with ␣-Glc-1-P, while our results indicate that there is no significant difference in reactivity, consistent with the conservation of active site residues.
Conclusion-The structural and kinetic analyses reported here in are important for several reasons. First, the key structural and functional sites in the two ␣-PMM isozymes are conserved. Thus, we anticipate that mutations will effect both enzymes in similar manner. The clinical manifestation, however, will depend on the isozyme tissue distribution. Second, location of the positions of the clinical mutations in the ␣-PMM structures provide a link between phenotype and genotype, which can be explored in detail through future structure/function analysis of the mutant enzymes. Finally, the divergence of mutase function in HASDF subclass II, which we contend is mechanistically distinct from that observed in the HASDF subclass I, provides a birds eye view of the diversity of nature at the molecular level.