Identification of sequences within the gamma-carboxylase that represent a novel contact site with vitamin K-dependent proteins and that are required for activity.

The vitamin K-dependent (VKD) carboxylase converts clusters of Glu residues to gamma-carboxylated Glu residues (Glas) in VKD proteins, which is required for their activity. VKD precursors are targeted to the carboxylase by their carboxylase recognition site, which in most cases is a propeptide. We have identified a second tethering site for carboxylase and VKD proteins that is required for carboxylase activity, called the vitamin K-dependent protein site of interaction (VKS). Several VKD proteins specifically bound an immobilized peptide comprising amino acids 343-355 of the human carboxylase (CVYKRSRGKSGQK) but not a scrambled peptide containing the same residues in a different order. Association with the 343-355 peptide was independent of propeptide binding, because the VKD proteins lacked the propeptide and because the 343-355 peptide did not disrupt association of a propeptide factor IX-carboxylase complex. Analysis with peptides that overlapped amino acids 343-355 indicated that the 343-345 CVY residues were necessary but not sufficient for prothrombin binding. Ionic interactions were also suggested because peptide-VKD protein binding could be disrupted by changes in ionic strength or pH. Mutagenesis of Cys(343) to Ser and Tyr(345) to Phe resulted in 7-11-fold decreases in vitamin K epoxidation and peptide (EEL) substrate and carboxylase carboxylation, and kinetic analysis showed 5-6-fold increases in K(m) values for the Glu substrate. These results suggest that Cys(343) and Tyr(345) are near the catalytic center and affect the active site conformation required for correct positioning of the Glu substrate. The 343-355 VKS peptide had a higher affinity for carboxylated prothrombin (K(d) = 5 microm) than uncarboxylated prothrombin (K(d) = 60 microm), and the basic VKS region may also facilitate exiting of the Gla product from the catalytic center by ionic attraction. Tethering of VKD proteins to the carboxylase via the propeptide-binding site and the VKS region has important implications for the mechanism of VKD protein carboxylation, and a model is proposed for how the carboxylase VKS region may be required for efficient and processive VKD protein carboxylation.

The vitamin K-dependent (VKD) 1 or ␥-carboxylase converts Glus to ␥-carboxylated Glus (Glas) in VKD proteins as they transit through the endoplasmic reticulum (1,2). Most of the VKD proteins are secreted out of the cell, and carboxylation of their Gla domain confers the ability to bind phospholipid bilayers, where these proteins exert their effects. Carboxylation is thus required for the biological activity of VKD proteins, which function in hemostasis, calcium homeostasis, and growth control. In addition, a novel subset of mammalian VKD proteins with potential functions in signal transduction has recently been identified by sequence homology (3,4). Unlike the other VKD proteins, these proteins apparently have a single-pass transmembrane domain with the extracellular domain containing the predicted carboxylated region. Inhibition of VKD protein activities forms the basis of anticoagulant therapies with warfarin and coumadin, in which the carboxylation of hemostatic VKD proteins, as well as the other VKD proteins, is reduced by limiting the supply of vitamin K cofactor to the carboxylase.
Although the carboxylase was first identified in mammals, carboxylase homologs and activity have been found in fish, the fish-hunting cone snails of the genus Conus, and the fruit fly Drosophila (5)(6)(7)(8)(9). All chordates appear to contain the hemostatic VKD proteins (10). The known VKD proteins of Conus, however, have a distinct function where the VKD proteins are neurotoxic venom peptides (11)(12)(13)(14)(15). VKD proteins have not yet been isolated in Drosophila, and so the function of carboxylation in the fruit fly is not currently known.
The carboxylase modifies VKD proteins by using O 2 and vitamin K hydroquinone (KH 2 ) to abstract the ␥-hydrogen of glutamyl residues to form a carbanion intermediate, which then incorporates CO 2 via nucleophilic attack to form the Gla (1,2). During each Glu to Gla conversion, one molecule of KH 2 is oxidized to vitamin K epoxide, and the carboxylase is also an epoxidase. Insights into the molecular mechanism for this reaction have only recently been revealed. Early studies with thiol-specific inhibitors implicated Cys residues as part of the carboxylase active site (16). Chemical modeling based on those studies led to a proposed base strength amplification mechanism where a weak base (thiolate) initiates KH 2 oxygenation to generate a strong base that can abstract the ␥-hydrogen of Glu to form the carbanion intermediate (17). We determined that two Cys residues in the carboxylase, Cys 99 and Cys 450 , were modified by N-ethylmaleimide, a thiol-specific inhibitor that inactivated both carboxylation and epoxidation (18). Individual mutation of each Cys residue to a Ser greatly reduced both carboxylation and epoxidation, and kinetic analysis suggested that the Glu substrate coordinates at least one Cys residue and KH 2 in a complex required to initiate oxygenation of KH 2 (18).
Multiple Glu residues are carboxylated in VKD proteins, with as many as 12 Glus converted to Glas per molecule (in human factor IX (fIX)) (1,2). Carboxylation occurs through a processive mechanism in which the multiple Glu to Gla conversions all result from a single binding event between VKD protein and carboxylase (19,20). One of those studies (19) also indicated that there is a driving force for VKD protein carboxylation, which is likely based on the ability of the carboxylase to distinguish Glus from Glas. However, how such a distinction is accomplished is not known.
Processivity is mediated by the binding of VKD proteins to the carboxylase through a carboxylase recognition sequence (CRS) in VKD proteins that tethers them to the carboxylase throughout the reaction. This sequence is immediately adjacent to the Gla domain where modification occurs. In most instances, VKD proteins are synthesized as precursor proteins where the CRS is a propeptide that is cleaved in the Golgi subsequent to carboxylation, and the CRS is frequently referred to as the propeptide (1,2). This sequence confers high affinity binding; covalent attachment to small peptides derived from the Gla domain reduces the K m values of such peptides by 10 3 -10 4 -fold, from the mM to the M range (1). Three nearly invariant residues in the propeptide have been shown to be important for carboxylase recognition. Other residues clearly affect binding, however, because propeptides with these three residues exhibit up to a 100-fold difference in affinities (21). To date, the propeptide-binding site is the only known tethering point between the carboxylase and its substrate protein.
The propeptide-binding site, like most functional regions of the carboxylase, has not been identified. Biochemical mapping of the carboxylase has been difficult because the carboxylase is a large (95 kDa) integral membrane protein that is purified in a micelle. Consequently, most structural information has been low in resolution. In two conflicting reports, cross-linking studies mapped the propeptide-binding site to either amino acids 50 -225 (22) or ϳ349 -500 (23). Because of the difficulties in biochemical mapping, alternative approaches for analyzing carboxylase structure function have been pursued. Two human genetic variants with phenotypic consequences have been identified (24,25), and the low number detected is not surprising given the broad range of VKD protein-mediated functions affected by the carboxylase. Mutagenesis studies have also identified potentially important carboxylase residues whose alterations affect activity (26,27); however, to date there has been no independent confirmation by biochemical mapping to support their functional significance. Thus, high resolution biochemical information has been limiting but is essential for elucidating the structural basis for the carboxylase mechanism.
We discovered a region of the human carboxylase, amino acids 343-355, that binds VKD proteins with high affinity. This region binds VKD proteins that lack the CRS/propeptide, showing that it defines a new tethering site on the carboxylase distinct from the CRS/propeptide-binding site. Biochemical identification of 8 amino acids within this site that are required for VKD protein binding and subsequent mutational analysis showed that this region is important for carboxylation and epoxidation. A new site of interaction likely affects the efficiency of protein carboxylation. Results showing that this region has a higher affinity for carboxylated than uncarboxylated VKD protein and the previous processivity studies showing that the VKD protein remains tethered to the carboxylase throughout the reaction to become multiply and fully carboxylated (19) suggest that this region facilitates the driving force that accomplishes processive, comprehensive carboxylation.

Carboxylase-derived Peptides
Peptide synthesis, HPLC purification, and structure verification by mass spectrometry were performed by the Lerner Research Institute Molecular Biotechnology Core as described at www.lerner.ccf. org/services/molecbiotech/peptide.html.

Binding Assay for VKD Protein-Carboxylase Peptide Interaction
Column Chromatography Assay-Peptides derived from the carboxylase were coupled to CNBr-activated Sepharose (Amersham Pharmacia Biotech) at a final concentration of 500 M. Mixtures of prothrombin (PT) (Enzyme Research Laboratories) and BSA (Pierce), both at 1 M in a volume of 10 ml of buffer A (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5% Nonidet P-40) were passed over 2-ml peptide columns. The columns were rinsed with 20 ml of buffer A, then rinsed with 20 ml buffer A lacking Nonidet P-40, and then eluted with 2 ml of 100 mM sodium citrate, pH 2.5, followed by neutralization of the eluant with 1.5 M Tris-HCl, pH 8.8. Aliquots corresponding to 1% of the starting sample, flow through, and eluant were analyzed by SDS-PAGE and Coomassie staining. A similar assay was also performed with fIX (Enzyme Research Laboratories) and ovalbumin (Sigma), except that loading concentrations of 0.2 M were used.
Human or rabbit serum was also chromatographed on carboxylasederived peptide columns as above. The eluants were analyzed in a Western blot using affinity-purified polyclonal Abs against fIX (28), protein C (29), factor X (American Diagnostica), or PT. The anti-PT Ab was rabbit Ab against human PT and was affinity-purified using a PT-Sepharose column (5 ml, 2 mg/ml PT). All Abs were used at 0.2 g/ml, and detection was by ECL (Amersham Pharmacia Biotech).
Determination of the K d -Human PT (Enzyme Research Laboratories) and a carboxylase-derived peptide (343-355 of the human carboxylase) that had a biotin group at the C terminus were incubated for 1 h at 21°C in 120 l of phosphate-buffered saline containing 0.05% Tween 20. The concentrations of PT and VKS peptide varied from 2 to 20 M, and duplicate samples were analyzed at each concentration. The samples (100 l) were then chromatographed on 2-ml G-75 Sephadex columns, and 100-l fractions were collected to separate free from PTbound peptide. Preliminary experiments indicated that the PT-peptide complex did not dissociate during the time frame of the chromatography (ϳ2 min). Initial experiments also verified that the void volume, which contained PT and the PT-peptide complex, was well separated from free peptide.
The amounts of PT and 343-355 peptide in the void volume were quantitated by PT and carboxylase peptide (343-355) ELISAs. For the PT ELISA, the samples were diluted in phosphate-buffered saline, 0.05% Tween 20, 1% BSA adjusted to 0.5 M NaCl to disrupt PT-peptide binding. A monoclonal anti-PT Ab (Biodesign International) was used for antigen capture, and detection was with affinity-purified polyclonal anti-PT Ab. Pure human PT was used for the standard curve (0.03-0.9 nM). The carboxylase peptide (343-355) ELISA used Neutravidin plates (Pierce; capacity, 25 pmol biotin/well) to capture the biotinylated peptide and polyclonal anti-peptide Ab for detection. The anti-peptide Ab was against the human 343-355 peptide coupled to keyhole limpet hemacyanin and was generated in rabbits. The antiserum was chromatographed on a 5-ml 343-355 peptide column (2 mg/ml, coupled to Sepharose) and then on a protein A-Sepharose column (5 ml; Sigma) to generate a homogeneous preparation of Ab as determined by SDS-PAGE and Coomassie staining. A standard curve of biotinylated peptide (5-40 nM) was used for quantitation.

Assay for Propeptide fIX-Carboxylase Complex Dissociation
The propeptide fIX-carboxylase complex was isolated from a BHK cell line (cultured without vitamin K) expressing recombinant human fIX and recombinant human carboxylase (30) and tested for stability in the presence of various peptides. Solubilized microsomes from the recombinant human fIX, recombinant human carboxylase BHK cell line (1 ml, 2 mg/ml protein and 5 ϫ 10 6 cpm/hr of carboxylase peptide activity) were incubated with anti-fIX Ab resin (300 l, 5 g/l of affinity-purified polyclonal anti-fIX Ab) overnight at 4°C. The resin was washed six times at 4°C with 1 ml of buffer B (50 mM Tris-HCl, pH 7.4, 500 mM NaCl, 0.25% CHAPS, 0.25% phosphatidyl choline, and 5 mM dithiothreitol) by gently rocking the resin in buffer B and then centrifuging for 1 min at 1000 ϫ g to remove the supernatant. After the last wash, the resin was resuspended in 1 ml of buffer B, and the slurry was divided into aliquots (100 l). Equal volumes (100 l) of various peptides, which are described under "Results," were then added (each at a final concentration of 100 M). The samples were rocked overnight at 4°C and then centrifuged at 1000 ϫ g for 1 min at 4°C, and the supernatant was assayed for peptide carboxylation activity.

Construction of Carboxylase Mutants
The C343S, Y345F, and K346A/R347A carboxylase mutants were each generated using 10 overlapping oligonucleotides that replaced the 145-base pair MscI-MscI fragment of pGC2 (a pUC118 plasmid bearing the 3Ј EcoRI-BamHI fragment of the carboxylase cDNA (18)). The C343S mutation changed the TGT codon to TCC, the Y345F mutation changed the TAT codon to TTC, and the K346A/R347A mutation changed the AAG⅐AGG codons to GCT⅐GCG. A silent mutation, G1043T, was also introduced into each mutant to create a new HindIII site that facilitated screening for positive clones. All of the mutants were confirmed by double-strand DNA sequencing, and the EcoRI-BstBI fragment (bases 954 -2196 of the sequence in GenBank TM accession number M81592) bearing these mutations was used to replace the corresponding wild type sequence in a plasmid containing full-length wild type carboxylase in a pBakPAK1vector (CLONTECH) (18).

Construction of Carboxylase-expressing Baculovirus
Plasmids bearing carboxylase cDNAs in pBacPAK1 (500 ng) were cotransfected with 100 ng of Bsu36I-digested viral DNA (BacPAK6, CLONTECH) using Lipofectin on SF21 cells. Individual plaques (6 -10 for each mutant) were isolated and analyzed by preparing SF-21-infected cell lysates that were screened for carboxylase peptide activity, as previously described (31). Representative viruses for each mutant were then scaled up, and microsomes were prepared, as described (18).

Carboxylase Activity Assays
Carboxylase peptide activity in solubilized microsomes from baculovirus-infected insect cells (10 -50 l, 2 mg/ml) was measured in a 150-l reaction mixture, as described (30). The reaction was initiated by the addition of microsomes, then incubated for 15 min at 21°C, and terminated by the addition of trichloroacetic acid (1 ml, 10%), followed by boiling and counting 14 C cpm as previously described (31). The epoxidase assay was carried out in the same manner except that 2.5 mM cold NaHCO 3 was used, and the reaction was terminated by the addition of isopropyl alcohol:hexane (3:2 v/v). Samples were processed for HPLC analysis as described (18), using a vitamin K epoxide standard (provided by Dr. James Sadowski and quantitated by absorbance). In experiments that compared the carboxylase and epoxidase activities of carboxylase mutants, the assays were performed on the same day using aliquots of the same microsomal sample.
To measure the effect of varying cofactor and substrate concentrations on enzyme activity, solubilized microsomes (50 l) were incubated in a 200-l reaction mixture (30) in a 15-min assay initiated by the addition of microsomes. The K m for CO 2  Carboxylase carboxylation was assayed using the same reaction mixture as in the peptide assay (30), except that EEL was omitted. The reaction was initiated by the addition of solubilized microsomes (25 l) to reaction mixture (50 l), and aliquots at 30 s and 1 min were taken and quenched by the addition of SDS-PAGE loading dye. The samples were electrophoresed on an 8% acrylamide SDS-PAGE gel along with a [ 14 C]BSA standard curve. The [ 14 C]BSA used was first chromatographed on a G-75 Sephadex column to remove any unincorporated radioactivity. After electrophoresis, the gel was rinsed in destain (5% acetic acid, 40% methanol), then dried, and exposed for 1 month using a 14 C-sensitive PhosphorImager screen (Molecular Dynamics). The rate of carboxylase carboxylation was calculated by measuring the amount of 14 CO 2 incorporation into the carboxylase, and the activities were normalized to the relative amounts of carboxylase protein as determined by a quantitative Western blot.

Quantitative Western Blot Analysis
Carboxylase levels in solubilized microsomes were quantitated using fluorescence-based Western blot analysis, as described (18). Briefly, several aliquots corresponding to ϳ10 -100 ng of carboxylase were coelectrophoresed with the same amount of pure human carboxylase on SDS-PAGE, followed by Western blot analysis using affinitypurified, anti-C-terminal human carboxylase anti-peptide Ab (1 g/ ml) (18) and fluorescein-conjugated goat anti-rabbit Ab (ECF; Amersham Pharmacia Biotech) and quantitation on a Storm Imager (Molecular Dynamics).

Preparation of Uncarboxylated PT
A human PT cDNA, provided by Dr. Sandra Friezner-Degen (32), was subcloned into the ZEM229 vector (33), and BHK cell lines stably expressing PT were generated by transfection and selection for resistance to dihydrofolate reductase. To prepare uncarboxylated PT, a clonal isolate was cultured in the absence of vitamin K. Secreted protein was purified using anti-human PT Ab, affinity-purified and coupled to CNBr-activated Sepharose. The preparation was shown to be homogeneous by SDS-PAGE and Coomassie staining, and the secreted PT did not contain detectable propeptide, as determined by N-terminal sequence analysis (data not shown).
Uncarboxylated PT was also generated by heat decarboxylation (34) of pure plasma PT. PT (3.5 mg, Enzyme Research Laboratories) was dialyzed against 0.1 M NH 4 HCO 3 pH 8.0, and then lyophilized to dryness. The sample was heated at 110°C for 5 h under nitrogen, then resuspended in 1 ml of 50 mM Tris-HCl, pH 7.4, 100 mM NaCl, and dialyzed against the same solution. The sample was quantitated alongside the initial carboxylated PT in a protein assay (BCA, Pierce), and SDS-PAGE and Coomassie staining showed that the heat-decarboxylated PT was fully intact. Gla quantitation (29) was used to verify decarboxylation and showed that the number of Glas/molecule was reduced from 10 (starting PT) to 2 (heat-decarboxylated PT).

Identification of a Site in the Carboxylase That Interacts with
Vitamin K-dependent Proteins-A region of the carboxylase that binds to VKD proteins was discovered while generating anti-carboxylase antibodies. Antisera from rabbits immunized with a peptide containing amino acids 343-355 of the human carboxylase were fractionated on an immobilized peptide column, revealing multiple proteins that specifically bound to the peptide column (Fig. 1A). The presence of nonimmunoglobulin bands, which was not observed in several other anti-peptide Ab purifications, was due to serum proteins binding to either the peptide or the anti-peptide Ab. To distinguish these two possibilities, the peptide column eluant was fractionated further using protein A-Sepharose. The Ab fractionated away from the other bands in this experiment (Fig. 1A), suggesting that the original binding was to the 343-355 peptide.
Several of the proteins in the 343-355 peptide column eluant had molecular weights similar to known serum VKD proteins. Because the peptide was derived from the carboxylase that modifies these proteins, we tested whether the peptide specifically bound serum VKD proteins in the absence of anti-peptide Ab. Human serum was fractionated on columns containing either the 343-355 peptide or a scrambled peptide with the same amino acid composition but in a different order, and the eluants were analyzed in a Western blot. Both PT and fIX were bound and eluted from the 343-355 column but not from the scrambled peptide column (Fig. 1B). Similar results were obtained for two other VKD proteins, factor X and protein C (data not shown). Therefore, the 343-355 carboxylase-derived sequence specifically bound multiple VKD proteins from human serum.
To test whether the VKD proteins were binding directly to the 343-355 peptide, a mixture of pure PT and BSA was fractionated on the peptide column (Fig. 2). PT was specifically and quantitatively bound, showing that the interaction of PT and the 343-355 peptide was direct and not mediated by an intermediate serum protein. A similar result was obtained using a mixture of fIX and ovalbumin (data not shown). Because the combined chromatography data indicated that the peptide binds VKD proteins, we refer to the 343-355 region of the carboxylase as the vitamin K-dependent protein site of interaction (VKS).

Free VKS Peptide Does Not Alter Carboxylase-VKD Protein Association and Binds Carboxylated Protein with a Higher
Affinity than Uncarboxylated Protein-The VKD proteins in human serum (Fig. 1B) and those tested for direct binding to the VKS peptide (e.g. PT in Fig. 2) lacked the propeptide. Therefore, the VKS region of the carboxylase appeared to be distinct from the propeptide-binding site. We tested whether the VKS peptide affected VKD propeptide binding by assessing its ability to disrupt VKD protein-carboxylase association. In the absence of vitamin K, tight complexes between carboxylase and propeptide-VKD precursor accumulate in vivo (28,(35)(36)(37). We isolated such a preformed propeptide fIX-carboxylase complex by immobilization on anti-fIX Ab resin, then incubated the complex with various peptides, and monitored for carboxylase release from the propeptide fIX-anti fIX Ab resin (Table I). As expected from previous studies (28,35), the propeptide fIXcarboxylase complex was disrupted by incubation with a propeptide sequence. In contrast, when the VKS peptide or an extended peptide containing additional N-terminal sequences was incubated with the propeptide fIX-carboxylase complex, the amount of complex disruption was indistinguishable from background levels obtained in the absence of any peptide (Table I). These data support the conclusion from the chromatography results (Figs. 1 and 2) that the VKS region is distinct from the carboxylase propeptide-binding site.
The VKD proteins that bound the VKS peptide (Figs. 1 and 2) were carboxylated, and so uncarboxylated protein was also tested for its ability to bind the 343-355 peptide. Uncarboxylated PT, lacking the propeptide, was generated by production of recombinant protein in cells cultured without vitamin K and also by heat decarboxylation of purified plasma protein ("Experimental Procedures"), and the two preparations gave similar results. Carboxylated and uncarboxylated PT were tested in the chromatography assay as in Fig. 2, except that a NaCl gradient was used for elution. Uncarboxylated PT, like carboxylated PT, was quantitatively bound to the 343-355 VKS pep-tide but eluted at a lower ionic strength (0.3 M NaCl) than carboxylated PT (0.4 M NaCl; data not shown). Binding constants for the two forms were determined using a quantitative assay; VKS peptide and PT were coincubated and then passed over a size exclusion column to isolate and quantitate the PT-VKS peptide complex ("Experimental Procedures"). These

TABLE I
The VKS peptide does not disrupt VKD protein/carboxylase association Solubilized microsomes from BHK cells containing propeptide fIX/ carboxylase complex were adsorbed to anti-fIX Ab resin, and the resin was washed and incubated with the indicated peptides, as detailed under "Experimental Procedures." The resins were subsequently centrifuged, and the supernatant was assayed to quantitate the amount of carboxylase released from the propeptide fIX/anti-fIX Ab resin. The 333-355 N-terminally extended peptide was also tested because studies (described below) indicated that the N terminus of the 343-355 peptide was functionally more important for activity. The factor X propeptide sequence is SLFIRREQANNILARVTR. The ability of the 343-355 peptide to distinguish carboxylated from uncarboxylated PT raised the question of whether the VKS region plays a role in Glu substrate recognition. Therefore, the ability of the VKS peptide to compete with enzyme in a carboxylation reaction on peptide substrate was tested. The concentrations of peptide substrate (EEL) and VKS peptide used were chosen so that competition would be observable; the EEL concentration (0.1 mM) was 4-fold below K m (18), and the VKS peptide (1 mM) was in 10-fold excess of EEL, 10 4 -fold excess of carboxylase, and 16-fold above the K d for uncarboxylated PT and this peptide. The VKS peptide, however, had no effect upon the carboxylation of the EEL substrate (Table II). Similar results were obtained when the reaction was performed in the presence of propeptide, which activates the carboxylase (38) and therefore accounts for the higher activity observed (Table II). Thus, although the VKS efficiently bound VKD proteins, it did not compete for carboxylase association with EEL. One explanation that reconciles these data is that the EEL substrate does not have all of the sequences present in VKD proteins that are required for binding to the VKS peptide.
The N-terminal Portion of the VKS Peptide Is Essential for PT Binding-The lack of effect of the VKS peptide when added in trans indicated that a more informative way to test for a function for the VKS region would be doing so within the context of the entire carboxylase molecule. Therefore, VKS peptide residues to target for mutagenesis were identified using overlapping carboxylase-derived peptides of similar lengths and the column assay for VKD protein binding.
Comparison of the 343-355 VKS peptide to an overlapping 346 -362 peptide showed that the N-terminal CVY residues of the VKS peptide were required for PT binding (Fig. 3A). These residues were not sufficient for binding, however, because a peptide with these sequences (333-346) did not bind PT (Fig.  3B). Extending this peptide by 4 amino acids (333-350; Fig. 3C) led to PT binding, implicating the importance of the KRSRG residues C-terminal to CVY. The KRSRG sequence includes 3 positively charged residues, some of which probably are functionally important because PT binding was disrupted by increased salt concentration or changes in pH. Charge alone, however, was not sufficient for binding. A scrambled peptide with the same amino acid composition and predicted pI (11.4) as that of the VKS 343-355 peptide did not bind PT (Figs. 3D and 1B), indicating that the specific order of amino acids was important. In addition, the 346 -362 peptide did not bind PT (Fig. 3A) yet was even more basic (pI ϭ 12.8) than the 343-355 VKS peptide.
The results therefore indicated that VKD protein binding to the VKS region involved both the N-terminal CVY sequences and charged residues. At the time the binding experiments were performed, only the human and bovine carboxylase se-quences were known, and the CVY sequence was CMY in the bovine carboxylase (39). The evolutionary conservation implicated the C and Y residues as most critical for function, and these two residues were therefore targeted for mutagenesis. To analyze the effect of charge on function, two of the three amino acids within the 346 -350 sequence that facilitated PT binding (Fig. 3, B and C) were also chosen for mutagenesis. The dibasic pair was targeted because mutagenesis studies frequently implicate pairs of charged residues as important for protein function and protein-protein interaction (40 -43).
Cys 343 and Tyr 345 Are Important for Peptide Substrate Carboxylation and Epoxidation and for Carboxylase Carboxylation-Individual carboxylase mutants were made that converted the Cys at position 343 to Ser (C343S), the Tyr at position 345 to Phe (Y345F) and the Lys-Arg dibasic pair at positions 346 and 347 to Ala-Ala (K346A/R347A). All of the mutated carboxylases were produced in baculovirus-infected insect cells because these cells have no endogenous carboxylase (44), which would interfere with mutant analysis. A K346A/ R347A mutant had previously been generated in mammalian cells (26); however, the mutant was also produced here in insect cells to unequivocally assess its function in the absence of background carboxylase activity. Microsomes were prepared from insect cells expressing mutant or wild type carboxylase (Fig. 4), and the assayed activities were normalized to the relative amounts of carboxylase protein as determined by a quantitative Western blot.
When assayed for activity, the C343S and Y345F mutant enzymes showed 9 -13% of the wild type enzyme level, whereas the K346A/R347A mutant showed wild type activity (Table III). Both peptide carboxylation and KH 2 epoxidation were reduced in the C343S and Y345F mutants, and all three mutants, like wild type, gave a ratio of carboxylase activity to epoxidase activity of close to 1. Kinetic analyses using different concentrations of substrate or cofactor (Table IV) showed an increased K m for the EEL substrate for the enzymes with reduced activity. The Y345F mutant exhibited a 6-fold increase in K m , with a 20-fold decrease in catalytic efficiency with respect to wild type enzyme, and the C343S mutant exhibited a 5-fold increase in K m and a 33-fold decrease in catalytic efficiency. Both mutants also showed a decrease in K m for KH 2 and no difference in K m for CO 2 (Table IV).
We previously showed that the carboxylase undergoes carboxylation both in vitro and in vivo, in a reaction dependent upon vitamin K (30). The effect of the VKS mutants on carboxylase carboxylation was therefore also tested. Microsomes containing the wild type or mutant enzymes were prepared from insect cells cultured in the absence of vitamin K to produce uncarboxylated enzyme. Solubilized microsomes were in vitro carboxylated, and the rate of 14 C incorporation into carboxylase was quantitated by SDS-PAGE and PhosphorImager analysis (Table V). C343S and Y345F showed 14 and 12% of the rate of carboxylase carboxylation compared with wild type enzyme, respectively, whereas K346A/R347A showed only a slight reduction in carboxylase carboxylation. The reduction in C343S and Y345F carboxylase carboxylation paralleled the reduction in peptide carboxylation and epoxidation (Tables III and V). The reduced EEL carboxylation was not a consequence of reduced carboxylase carboxylation, because we have shown that the rate of carboxylation of a fIX substrate is not affected by the extent of carboxylase carboxylation (data not shown). Thus, mutated residues in the VKS region altered the carboxylase reaction for both the peptide and carboxylase substrates, consistent with Cys 343 and Tyr 345 playing a role in the carboxylase active site. Carboxylase activity on an EEL peptide was measured as described (30), except that the EEL concentration was 0.1 mM and ammonium sulfate was replaced by 0.1 M NaCl. Activity was measured with or without VKS peptide (final concentration, 1 mM) and with or without propeptide (20 M), using solubilized microsomes from insect cells expressing wild type human carboxylase.

DISCUSSION
The VKS Domain Defines a New Site of Carboxylase-VKD Protein Contact-We used independent biochemical and genetic approaches to show that a region of the carboxylase that binds VKD proteins, called VKS, represents a new site of interaction between VKD proteins and the carboxylase that is independent of propeptide binding and that is important for both epoxidation and carboxylation. The VKS domain is distinct from the CRS/propeptide-binding site because VKD proteins lacking the propeptide were bound to a VKS peptide (Figs. 1 and 2) and because VKS peptides did not dissociate a propeptide fIX-carboxylase complex (Table I). In addition, the VKS peptide did not inhibit propeptide stimulation of carboxylase activity in a carboxylation reaction (Table II).
The affinity of VKS peptide-VKD protein binding is less than the affinity of the carboxylase for propeptide, with K d values of 5 or 60 M for VKS peptide and carboxylated or uncarboxylated PT, respectively. Although the K d values for VKD propeptides and the carboxylase have not yet been reported, inhibition of carboxylation of a fIX peptide by VKD propeptides shows K i values in the submicromolar range (e.g. 0.3 M for PT) (21). The higher propeptide affinity is consistent with its requirement for VKD protein binding and carboxylation in vivo; mutations in the propeptide abolish the normal intracellular association of carboxylase with VKD protein (28) and impair carboxylation (45)(46)(47). Thus, the VKS domain must have a distinct role from targeting VKD protein precursors to the carboxylase. It should be noted that tethering through the propeptide increases the local concentration of VKD protein with respect to the carboxylase. For example, if the site in VKD proteins bound by VKS is just downstream of the Gla domain, the local concentration of this site would be in the low millimolar range. Thus, binding of the VKD proteins at the carboxylase VKS domain is likely to be saturated.

Functional Role(s) of the VKS Region in VKD Protein
Carboxylation-Cys 343 and Tyr 345 within the VKS region are both important for function. A VKS-derived peptide lacking these residues did not bind PT (Fig. 3A), and both C343S and Y345F mutant enzymes showed significantly reduced EEL and carboxylase carboxylation (Tables III and V). Cys 343 is of interest because previous studies implicated thiols as part of the active site (16). However, Cys 343 is not one of these residues; we used N-ethylmaleimide modification and liquid chromatography electrospray mass spectroscopy to identify Cys 99 and Cys 450 as the two active site thiols (18). Cys 343 was not modified by N-ethylmaleimide, so it does not play the same integral roles of Cys 99 and Cys 450 in catalysis. Nonetheless, the significant reduction in both epoxidation and carboxylation for the mutant C343S and for Y345F indicates that both Cys 343 and Tyr 345 residues affect the active site.
Both C343S and Y345F mutations caused an increase in K m for the EEL substrate and a decrease in K m for the KH 2 cofactor (Table IV) that paralleled the reciprocal K m changes previously observed with the C99S and C450S active site mutants (18). Those studies suggested that the Glu substrate, KH 2 , and active site thiols form a coordinated complex, with the mutations disrupting the complex and leading to poorer Glu substrate binding and tighter but nonproductive KH 2 binding that yields poor activity. The similar kinetic perturbations caused by the C343S and Y345F mutations suggest that these amino acid side chain alterations also distort the Glu substrate-KH 2 complex, indicating that the VKS region is near the active site. The results are consistent, then, with the VKS domain playing a role in establishing the precise conformation of the carboxylase active site required for correct positioning of the Glu substrate-KH 2 complex and consequent epoxidation and carboxylation.
As described in the results, VKS domain function depends upon at least one other interaction than that mediated by Cys 343 and Tyr 345 , which is most likely facilitated by charged residues. A functional requirement for charge is attractive in providing a mechanism for how the VKS domain distinguishes carboxylated from uncarboxylated VKD proteins. At present, the specific charged residues that are important are not known. A K346A/R347A mutant enzyme showed wild type activity on a peptide (EEL) substrate (Table III). However, the importance of residues 346 and 347 was originally suggested by binding studies with PT, not EEL, and we are therefore currently developing an assay to test for VKD protein carboxylation by carboxylase mutants. Other or additional charged VKS residues may also contribute to function, because PT binding was more efficient with a VKS peptide containing additional charged residues besides Lys 346 and Arg 347 (Figs. 3C and 2).

FIG. 4. Carboxylase expression in insect cells.
Solubilized microsomes (100 g of total protein) from insect cells infected with wild type (wt) or mutant carboxylase-containing baculoviruses were analyzed by chemiluminescence in a Western blot using affinity-purified anti-Cterminal carboxylase Ab, as described (18). Mock-infected insect microsomes (mock) and pure carboxylase (carb) were included as controls, and all lanes are from the same gel. The lower bands migrate with the molecular mass expected for unglycosylated carboxylase.
Anchorage of VKD proteins to the carboxylase through two sites likely contributes to the efficiency of protein carboxylation. Such efficiency is of interest with regard to recent results showing that VKD protein carboxylation occurs through a mechanism of tethered processivity, where the Gla domain of a VKD protein, tethered to the carboxylase via its propeptide, is scanned for Glus, which all become converted to Glas as a consequence of a single binding event (19). Those studies showed that the fIX Gla domain remains tightly associated with the carboxylase throughout the reaction, that carboxylation of the Gla domain occurs as rapidly as peptide (EEL) carboxylation, and that there is a driving force for the reaction based upon the ability of the carboxylase to distinguish Glus from Glas. Two points of tethering would facilitate the tight Gla domain-carboxylase association. In addition, a Gla domain secured close to the carboxylase active site could be more rapidly scanned for Glu residues to be modified, accounting for the fast reaction rate. Finally, the ability of the VKS region to bind carboxylated PT with a 12-fold higher affinity than uncarboxylated PT may also explain how the carboxylase provides the driving force for the reaction. Glu/ Gla positioning within the active site is a dynamic process because the multiple catalytic events within the Gla domain require that the Gla product, once generated, must exit the catalytic site to allow access to new Glu substrates. The VKS region may facilitate product exit with charged residues attracting the Gla away from the catalytic site. We propose, then, that the VKS region makes two functional contributions to the efficiency of carboxylation: Cys 343 -and Tyr 345mediated positioning of the Glu substrate for catalysis and charged residue-facilitated removal of Gla product from the catalytic site to provide the driving force. Both functions would contribute to a high k cat . A high catalytic rate versus low dissociation rate results in a processive reaction, and so the VKS region may facilitate carboxylase processivity.
The existence of a second tethering point between VKD proteins and the carboxylase has important implications for the unusual properties of bone Gla protein (BGP). The BGP propeptide has a much poorer affinity for the carboxylase than that of other VKD proteins, whereas BGP lacking the propeptide has an unusually high carboxylase affinity (21,48,49). These observations have raised questions as to the role of the propeptide in BGP carboxylation and where the high affinity carboxylase site on BGP is located. One possibility is that the VKS domain provides the high affinity contact for BGP that, along with the weaker propeptide binding, allows for efficient BGP carboxylation.
The VKS Region Is a Vertebrate-specific Carboxylase Domain-While this work was in progress, the carboxylase cDNA sequences from several mammalian species, a bony fish, and Drosophila melanogaster were reported (8,9,50,51). Comparison of carboxylase sequences between the closely related mammalian species shows that the VKS peptide is highly conserved in all mammals (Fig. 5A). Comparison of a mammalian consensus carboxylase sequence with the toadfish carboxylase sequence revealed an 8-amino acid insertion in the VKS peptide (Fig. 5B). Strikingly, the insertion point was within 1 amino acid of the minimal functional VKS peptide that we independently determined could bind to PT (Fig. 3C, peptide 333-350). In addition, the identical residues between toadfish and mammals included the Cys 343 and Tyr 345 amino acids that we showed by mutant analysis play an important role in carboxylase activity.
In contrast to the carboxylases in vertebrates, the carboxylase from Drosophila lacks the VKS domain. VKD proteins have not yet been identified in Drosophila; however, it is clear that carboxylases from distantly related species (Drosophila, mammals, and Conus, whose sequence has not yet been reported) have distinct contacts with their respective VKD substrates. The Conus and mammalian VKD substrates do not share sequence homology within their propeptides, and the propeptides of these two species are not functionally interchangeable (5-7). In addition, the mammalian and Conus propeptides do not  343 and Tyr 345 are required for carboxylation and epoxidation Solubilized microsomes from baculovirus-infected insect cells expressing wild type or mutant carboxylase were assayed for carboxylase or epoxidase activities, as described under "Experimental Procedures." The amount of carboxylase protein in each microsomal preparation was determined using a quantitative Western blot with pure carboxylase as a standard. The conversion of cpm to pmol in the carboxylase assay was based on a specific activity of 60 cpm/pmol for 14   function with the Drosophila enzyme (8,9). Thus, the propeptide sequences are not conserved between the vertebrate, fly, and snail enzymes. It is therefore not surprising that the second point of tethering between the vertebrate carboxylases and their VKD proteins is not conserved in the Drosophila enzyme.
Given that vertebrate carboxylases effect such extensive carboxylation (up to 12 versus, for example, 1-5 in Conus), one interesting possibility is that the VKS domain represents an evolutionary adaptation that facilitates processive and comprehensive carboxylation of VKD proteins that require many Glu to Gla conversions. Future studies that fully characterize the VKS region and the domain on VKD proteins bound to VKS will provide ways to assess the importance of this second tethering point on the efficiency of carboxylation.