Structural insight into bicarbonate-mediated carboxylation by human vitamin K-dependent carboxylase

Wu, Ke; Wang, Zheng; Yao, Deqiang; Li, Shaobai; Wang, Xiaozhu; Zhang, Yuanyuan; Cao, Mi; Shen, Yafeng; Xing, Shunpeng; Wu, Jian; Lei, Ming; Lan, Pengfei

doi:10.1038/s41467-025-65488-3

Download PDF

Article
Open access
Published: 25 November 2025

Structural insight into bicarbonate-mediated carboxylation by human vitamin K-dependent carboxylase

Nature Communications volume 16, Article number: 10480 (2025) Cite this article

4030 Accesses
1 Citations
1 Altmetric
Metrics details

Subjects

Abstract

Vitamin K-dependent (VKD) carboxylation, mediated by γ-glutamyl carboxylase (GGCX), is essential for the maturation of VKD proteins involved in critical physiological processes such as blood clotting, vascular calcification and bone metabolism. Here, we present cryo-electron microscopic structures of human GGCX alone and in complex with VKD proteins, vitamin K, and inhibitor anisindione. GGCX specifically recognizes diverse VKD substrates through high-affinity propeptide binding, while substrates like osteocalcin utilize a secondary exosite to enhance interaction. GGCX employs a conserved dipeptide anchoring mechanism that ensures processive carboxylation of glutamate residues. GGCX undergoes allosteric conformational changes that enable coordinated binding of vitamin K and glutamate substrates, facilitating the catalytic process. Additionally, we reveal a bicarbonate-mediated CO₂ capture mechanism that is conserved across bacterial and eukaryotic species, suggesting that this strategy for CO₂ utilization is both ancient and universal. Our findings lay the foundation for developing targeted anticoagulant drugs and innovative enzymatic CO₂ fixation strategies.

Structure and mechanism of vitamin-K-dependent γ-glutamyl carboxylase

Article 29 January 2025

Structural insights into the vitamin K-dependent γ-carboxylation of osteocalcin

Article 02 September 2025

Structural basis for the carboxylation and epoxidation of human gamma-glutamyl carboxylase

Article Open access 27 March 2026

Introduction

Vitamin K-dependent (VKD) carboxylation is an essential post-translational modification that converts specific glutamate (Glu) residues in VKD proteins into gamma-carboxyglutamate (Gla)^1,2. This modification is crucial for the biological functions of VKD proteins, which play key roles in various physiological processes such as blood clotting, vascular calcification and bone metabolism^1,3,4. γ-glutamyl carboxylase (GGCX) catalyzes this reaction, utilizing reduced vitamin K (KH₂), carbon dioxide (CO₂), and oxygen (O₂) as co-factors^5,6,7(Fig. 1a). Each carboxylation event is coupled to the oxidation of KH₂ into vitamin K 2,3-epoxide (VKO), which is then recycled back to KH₂ through the vitamin K cycle^8,9(Fig. 1a). Mutations in GGCX impair VKD protein carboxylation, leading to bleeding disorders and skin abnormalities^{10,11,12,13,14}. Mice lacking GGCX die perinatally from hemorrhage¹⁵. Additionally, targeting the vitamin K cycle has proven effective in anticoagulant therapy, with warfarin acting as a specific inhibitor of vitamin K epoxide reductase complex 1 (VKORC1), while anisindione—classically viewed as a VKORC1 antagonist—may also compete for the vitamin K–binding pocket of GGCX^16,17,18,19.

**Fig. 1: Overall structure of the GGCX-FIX complex.**

Despite its biological significance, key features of GGCX remain poorly understood. GGCX shares no known homology with other enzyme families and selectively processes diverse VKD proteins while excluding non-VKD substrates^{1,20,21,22,23}. Among these VKD proteins are GLA-domain containing coagulation factors, bone Gla protein (BGP), and matrix Gla protein (MGP), which share no sequence homology and display distinct structural characteristics³. These VKD proteins are guided to GGCX through propeptide sequences that bind a specialized exosite of conserved residues critical for GGCX interaction^{21,24,25,26,27}. Additionally, GGCX displays cooperativity in binding vitamin K and VKD proteins^28,29,30, yet the molecular details underlying this cooperativity remain elusive. Unlike other carboxylases, GGCX lacks structural similarity to known CO₂-binding enzymes¹, leaving its CO₂ uptake and trapping mechanisms unknown. GGCX displays processivity and allosteric regulation, modifying multiple Glu residues within a single substrate-binding event^31,32(Fig. 1a), yet the underlying mechanism still remains poorly understood.

At the catalytic core, GGCX performs two coupled half-reactions, KH₂ oxygenation and Glu carboxylation. Early chemical modeling studies proposed a “base strength amplification” model, positing that a weak base within the active site initially deprotonates KH₂, enabling its reaction with O₂ to generate a much stronger base^33,34,35,36. Previous studies have identified residue Lysine 218 of GGCX as the general base, responsible for the initial KH₂ deprotonation^37,38. CO₂ is essential for coupling KH₂ oxygenation and Glu carboxylation^1,39. Mutation of Histidine 160 impairs both CO₂ binding and Glu deprotonation, indicating a concerted mechanism where CO₂ incorporation and proton abstraction occur simultaneously³⁹. Despite these extensive studies, the configuration of the active sites and the catalytic mechanism remain largely elusive. Recently, three independent cryo-EM studies have reported structures of GGCX, revealing how the propeptide, Glu residues and vitamin K are recognized and offering mechanistic insight into its catalysis^40,41,42. However, several key questions remain unresolved, including how CO₂ is captured and positioned in the active site, how processive carboxylation of multiple glutamate residues is achieved, how substrate binding drives the conformational changes required for catalysis, and how Lys218 is maintained in its unprotonated state and then recycled.

Here, we present cryo-EM structures of human GGCX alone and in complex with various VKD proteins and vitamin K or its potential inhibitor. These structures define how GGCX recognizes and distinguishes VKD substrates, elucidate its allosteric regulation and processivity, and reveal the active sites governing KH₂ oxygenation and Glu carboxylation. Collectively, our findings provide an integrated framework for understanding how GGCX coordinates multiple substrates and cofactors, and offer a comprehensive mechanistic insight for this essential enzyme.

Results

Overall structure of human GGCX bound with FIX

To unveil the mechanism underlying the recognition and maturation of coagulant factors mediated by GGCX, we determined the structure of GGCX bound to coagulation factor IX (FIX), a representative VKD protein. GGCX contains three distinct domains, the transmembrane domain (TMD), the lumenal domain, and the C-terminal domain (CTD) (Fig. 1b, c). FIX is composed of a propeptide region, a GLA domain and a C-terminal functional domain. The GGCX-FIX complex was expressed in Expi293F cells, purified in detergent, and subjected to structural and enzymatic activity analyses (Supplementary Fig. 1a–k and Supplementary Table 1). In the presence of excess KH₂, the GGCX-FIX complex exhibits a robust carboxylase activity (Supplementary Fig. 1b). Cryo-EM analysis yielded a high-quality density map at a resolution of 2.62 Å, which allowed us to build an atomic model of GGCX, except for three disordered loops (residues 1–31, 627–652, and 727–758) (Fig. 1c, d and Supplementary Fig. 1g). Of the FIX substrate, only residues −18 to −3 of the propeptide region and residues 5 to 9 of the GLA domain were clearly resolved (Fig. 1c, d and Supplementary Fig. 1i).

The TMD of GGCX is made up of nine transmembrane (TM) helices and tightly packed with one cholesterol and additional phospholipids (Fig. 1b, d and Supplementary Fig. 1j), which likely contribute to the stability of the TM helices. The lumenal domain is composed of two structural “clamps” that sit atop the TMD, creating a cleft for propeptide binding (Fig. 1b, d). Clamp A is characterized by a central three-stranded β-sheet, while clamp B displays a prominent jelly-roll fold consisting of two stacked layers of four-stranded antiparallel β-sheets (Fig. 1d). The CTD of GGCX connects the lumenal and TMD domains via two long α-helical segments (α1, α2) (Fig. 1b, d), while the two amphipathic helices (α3, α4) extend into the lumenal leaflet of the ER membrane, running parallel to the membrane plane and further stabilizing the TM helices (Fig. 1b, d). Multiple lumenal loops, including the highly conserved lumenal loop LL9⁴³, further connect the lumenal domain and TMD (Fig. 1b, d and Supplementary Fig. 1k). This unique architecture of GGCX positions lipid-soluble and water-soluble substrates, as well as key catalytic residues into close vicinity, at the lumenal ER membrane interface (Fig. 1b, d), presumably enabling efficient catalysis.

Vitamin K binding

In the cryo-EM density map of GGCX-FIX, we observed a density in the lumenal-side membrane leaflet, which is surrounded by the four-helix bundle TM5 to TM8 and covered by the lumenal “cap” with three short helical motifs (Supplementary Fig. 1h). This density is clearly visible and can be unambiguously assigned with a vitamin K molecule (Supplementary Fig. 1h). Given that no vitamin K was added during purification, the cofactor most likely derives from endogenous sources during expression.

The vitamin K is deeply buried into a long-extended and highly hydrophobic pocket, formed by TM5 to TM8 and covered by the connecting cap domain and lumenal loop LL9 (Fig. 2a–d). The naphthoquinone ring of vitamin K lies perpendicular to the membrane surface, while its isoprenyl tail makes a sharp turn and meander through the monotopic TM6 and TM7, running parallel with the membrane surface (Fig. 2a).

**Fig. 2: Vitamin K binding and inhibition by anisindione.**

In the bottom of the pocket, the bulky naphthoquinone ring is tightly enclosed by a panel of aromatic and hydrophobic sidechains. In particular, it is sandwiched between sidechains of Met401 at the top while Phe286 at the bottom (Fig. 2b, d). In addition, the vitamin K mediates two hydrogen bonds with the sidechains of Lys218 and Asp263 (Fig. 2b, d), providing further specificity for the naphthoquinone ring binding. The isoprenyl tail of vitamin K snugly lies in the long hydrophobic tunnels formed by TM7 and the cap domain of GGCX (Fig. 2c, d).

Potential inhibition of GGCX by anisindione

Anisindione, an FDA-approved vitamin K antagonist, is used clinically to prevent thrombosis^44,45, prompted us to investigate whether it could interact with GGCX. To this end, we determined the cryo-EM structure of GGCX in complex with anisindione at a resolution of 2.8 Å (Supplementary Fig. 2a–i). The structure of anisindione-bound GGCX is almost identical to GGCX-FIX complex (Supplementary Fig. 2j).

The density for anisindione was of high quality, allowing unambiguous assignment of this drug molecule in the structure (Supplementary Fig. 2f, i). Occupying the same binding site as vitamin K, anisindione almost completely fills the quinone ring-binding region while leaving the isoprenyl chain tunnel vacant (Fig. 2a, e, f, g). The bicyclic 1,3-indandione core of anisindione matches well with the naphthoquinone ring of vitamin K, and contacts with the same group of residues that coordinate the binding of vitamin K ring (Fig. 2g). Due to its anisole extension, anisindione resides even deeper in the pocket, explaining of how this drug prevents vitamin K from binding to GGCX (Fig. 2g). In aggregate, the high-resolution structure of GGCX-anisindione complex reported here implies that anisindione may serve as a competitive antagonist of GGCX by occupying the vitamin K–binding pocket and preventing cofactor engagement.

Recognition of coagulation factor FIX

In the GGCX-FIX complex structure, FIX adopts an extended conformation that engages two distinct sites on GGCX (Figs. 1d, and 3a). The propeptide is recognized by a unique exosite of GGCX, formed by the lumenal bipartite clamp that creates a hydrophobic cleft ~25 Å away from the active site (Fig. 3a, b). The propeptide is anchored in this cleft primarily through the conserved hydrophobic residues, including (−17) Val, (−16) Phe, (−7) Ile, and (−6) Leu of FIX, which occupy a hydrophobic groove in GGCX (Fig. 3b and Supplementary Fig. 3a). At the N-terminus, (−17) Val^FIX is surrounded by Val430^GGCX, Phe431^GGCX, and Leu426^GGCX, while (−16) Phe^FIX is stabilized by Leu426 and Tyr415 of GGCX (Supplementary Fig. 3b). At the C-terminus, (−6) Leu^FIX is deeply buried in a hydrophobic pocket formed by the clamp B of GGCX (Fig. 3b and Supplementary Fig. 3c). Notably, the central region of the propeptide of FIX adopts a helical conformation, and makes a sharp turn at (−10) Ala^FIX, whose methyl group closely contacts with the aromatic ring of Tyr425^GGCX (Supplementary Fig. 3d). Additional extensive hydrogen bonds between FIX propeptide backbone and the GGCX clamp domain, further reinforce the recognition of FIX (Supplementary Fig. 3c, d).

Despite varying affinities among VKD propeptides²⁶, these core hydrophobic and hydrogen-bonding interactions are conserved, indicating a common recognition mechanism across VKD substrates.

Consistent with the structural information, hydrophobic residues, (−16) Phe, (−10) Ala, and (−6) Leu, are highly conserved across VKD proteins (Supplementary Fig. 3a), suggestive of their key roles in mediating the propeptide binding with GGCX. Indeed, previous biochemical analyses have shown that mutations such as F(−16)A and A(−10)T could impair the carboxylation of FIX^24,27. Furthermore, patients carrying mutations A(−10)T or A(−10)V in FIX exhibited hypersensitivity to warfarin inhibition, underscoring the functional importance of these alterations in human diseases^27,46.

The intervening region of FIX, which connects the propeptide to the carboxylation residue in the GLA domain, is poorly defined in the density map likely due to its intrinsic flexibility (Supplementary Fig. 1i). We propose that this flexibility allows intramolecular rearrangements within the GLA domain, enabling the sequential repositioning of Glu residues for processive carboxylation. This flexible linker consistently contains ~7–8 amino acids with varying sequences across all VKD proteins (Supplementary Fig. 3e). It is noteworthy that the complex structure reveals that the length of 7–8 amino acids is just enough for the positioning of the first carboxylation residue in the active site (Fig. 3a). We speculate that the dynamic nature and optimal length of the propeptide linker ensure efficient modification of multiple Glu sites in a processive reaction.

Dipeptide anchoring mechanism in the catalytic pocket

Once the propeptide is anchored, the downstream Glu residues are positioned into the catalytic site, forming a second key interface between FIX and GGCX (Figs. 1d, and 3a, c). Strikingly, this interface is dominated by a simple FIX dipeptide, Leu6-Glu7, which is snugly fitted into a pocket formed by ER lumenal loops LL3, LL7, and LL9 of GGCX (Fig. 3c). The Glu carboxyl group is oriented to mediate three electrostatic interactions with the sidechains of Asn159^GGCX, His160^GGCX and Tyr395^GGCX (Fig. 3c). In accordance with the structure, mutations of His160 greatly impaired the carboxylation reaction³⁹. Another salient feature of this dipeptide binding is the stacking between Leu6^FIX and Tyr395^GGCX, which guides the Glu sidechain in the correct orientation (Fig. 3c). In addition, a hydrogen bond between R436^GGCX and the backbone of Leu6^FIX further stabilizes the dipeptide (Fig. 3c).

Next, to examine whether this surprisingly simple dipeptide anchoring mechanism is conserved for all carboxylation sites in VKD proteins, we surveyed the immediately upstream residues (hereafter refer to as “Eup” residues) of each carboxylated Glu in human VKD proteins. This analysis clearly revealed a strong preference for hydrophobic amino acids or those with extended aliphatic side chains at this position (Fig. 3d). All these “Eup” residues presumably could engage in a stacking interaction with Tyr395^GGCX, ensuring the precise Glu positioning in the same dipeptide manner, thereby enabling processive carboxylation. Previous mutational studies affirmed the importance of this interaction, as alterations in Tyr395^GGCX could severely impair Glu binding and reduce carboxylation efficiency⁴³.

While this dipeptide pattern is broadly conserved, it is often does not apply to the final one or two potential carboxylation sites in the VKD proteins (Fig. 3d). These sites—often not essential for VKD function—may reduce substrate binding affinity and could serve as a natural “stop signal” for processive carboxylation to facilitate product release¹. Another apparent exception to the dipeptide pattern is the GLA-domain-containing VKDs, such as coagulation factors, where multiple two-consecutive-Glu sites cluster at the N-termini of the proteins (Fig. 3d). Once an upstream Glu is carboxylated to Gla, its newly acquired negative charge could, in principle, interfere with the Tyr395-stacking, raising the question how the subsequent Glu is carboxylated.

To address this issue, we synthesized a FIX peptide bearing two consecutive Gla-Glu residues (FIX_GlaE), with the upstream Glu pre-carboxylated to Gla. Cryo-EM analysis of the GGCX-FIX_GlaE complex resulted in an EM-density map at 2.4 Å resolution (Supplementary Fig. 4a–e), clearly resolving vitamin K, the propeptide, and residues 5–10 of FIX_GlaE (Supplementary Fig. 4f–j). In particular, the density for the Gla residue is well-resolved and easily distinguishable from others (Supplementary Fig. 4i). Structural comparison of GGCX complexed with FIX versus with FIX_GlaE reveals a subtle yet significant conformational change in the GGCX-FIX_GlaE complex, resulting in electrostatic interactions between carboxylated sidechain of Gla7^FIX with His440^GGCX and Arg406^GGCX (Fig. 3e and Supplementary Fig. 4k, l). This neutralization ensures the formation of the dipeptide anchor, allowing Gla7^FIX to maintain its stacking with Tyr395^GGCX (Fig. 3e). In contrast, the sidechains of His440^GGCX and Arg406^GGCX are shorter in distance to mediate direct interactions even when Leu6 is replaced with a carboxylated Gla in GGCX-FIX complex (Supplementary Fig. 4l).

Taken together, these observations suggest a conserved dipeptide anchoring mechanism, wherein the “Eup” residue, regardless of being hydrophobic or carboxylated, is uniquely recognized by the GGCX pocket (Fig. 3c, e). While hydrophobic residues stack with Tyr395 (Fig. 3c), carboxylated Gla is neutralized by Arg406 and His440 to facilitate their aliphatic chain stacking with Tyr395 (Fig. 3e). This ensures optimal substrate orientation and efficient, processive Glu carboxylation.

Allosteric regulation in GGCX

GGCX activity is governed by an intricate mechanism that synchronizes substrate-VK binding and catalysis^31,32,47,48. To understand this mechanism, we determined cryo-EM structure of GGCX in its apo state (Supplementary Fig. 5a–g). Comparison of this apo state with the FIX-bound GGCX structure reveals a series of cooperative conformational changes involving both clamp domains, the vitamin-K binding cap domain, and the lumenal loop LL9 upon FIX and VK binding (Fig. 4a and Supplementary Movie 1). In the apo structure, clamp domains A and B adopt a wedge angle of ~65°, resulting in a rather open conformation whose landscape is unsuitable for propeptide binding (Fig. 4a, b). In contrast, upon FIX association the clamps of GGCX undergo coordinated rigid-body rotations, reducing the wedge angle to ~45° and transitioning to a more closed conformation (Fig. 4a, b). We propose that this propeptide-induced clamp closure plays an important role in stably entrapping the propeptide in the clamp wedge during the processive carboxylation process.

**Fig. 4: Allosteric regulation of GGCX.**

Another major conformational change of GGCX upon FIX binding is from the VK-binding cap domain. Strikingly, the cap domain is almost completely disordered in the apo-state structure, suggestive of a highly flexible conformation in the absence of substrate (Supplementary Fig. 5h–j). The propeptide-induced clamp closure coincides with a large translational movement of the clamp domain B for ~10 Å (Fig. 4a). This movement allows the clamp to form contacts with the stabilized VK-binding cap domain via a network of direct interactions, which otherwise does not exist in the apo structure (Fig. 4c). In particular, Arg513 from the bridge helix in clamp B moves downward to stack against Tyr227 from the VK-binding cap domain, while Arg476 and Asn474 from the bridge loop in clamp A mediate polar contacts with Asp222 in the cap (Fig. 4c). It is noteworthy that mutations R513K and R476C that could disrupt the clamp-cap interaction have been identified in VKD coagulation factor deficiency type 1 (VKCFD1) patients, underscoring the functional significance of the clamp induced cap stabilization^12,49(Supplementary Fig. 6).

Notably, it is the cap domain surface opposite to the clamp-binding site that forms the VK-tail binding tunnel, highlighting a coupled structural rearrangement between propeptide and VK binding (Fig. 4a, d). This structural rearrangement can be further extended to the LL9 of GGCX, which also undergoes a substantial conformational change and flips away from the catalytic center such that sidechain of Met402 no longer occupies the position of Leu6^FIX in the dipeptide binding pocket (Fig. 4e). This coordinated structural rearrangement is consistent with previous hydrogen exchange mass spectrometry studies, which identified LL9 and the bridge helix in clamp B in GGCX as the regions undergoing substantial conformational changes upon substrate binding⁵⁰.

CO₂ trapping mechanism

Close examination of the high quality cryo-EM density map of GGCX-FIX revealed a clear linear density within the catalytic pocket, strongly indicative of a bound CO₂ molecule proximal to the Glu γ-carboxylation site (Supplementary Fig. 7a). This CO₂ molecule is precisely positioned and fully encapsulated within a hydrophobic pocket formed by a group of hydrophobic and aromatic residues of GGCX, in conjunction with the Glu substrate and vitamin K (Fig. 5a). This hydrophobic pocket very likely provides a highly selective environment for CO₂ binding, ensuring its stabilization and exclusion of competing molecules such as water. Notably, CO₂ interacts with the highly conserved Phe299^GGCX via a face-on acromatic π stacking interaction, driven by CO₂’s quadrupole alignment with the phenylalanine’s π-electron cloud (Fig. 5a and Supplementary Fig. 7a). Comparison of the apo and FIX-bound structures clearly reveals that the simultaneous presence of the Glu substrate and VK induces conformational changes of Met401^GGCX and Met402^GGCX to facilitate CO₂ trapping (Figs. 4e, and 5a), suggesting that CO₂ capture by GGCX is a substrate-assisted process.

**Fig. 5: Active site and catalytic mechanism.**

To corroborate our structural analyses, we performed two 1-μs all-atom molecular dynamics (MD) simulations of GGCX in a membrane-mimetic environment—one in the apo state and the other with FIX bound. Consistent with the cryo-EM observations, the apo enzyme exhibits globally elevated Root Mean Square Fluctuation (RMSF) values, with peaks at residues 220–255 (Cap) and 603–758 (CTD), which are unresolved in the apo map and therefore intrinsically disordered. FIX binding promotes clamp closure and Cap stabilization, attenuating fluctuations across Clamp A/B, LL9, and the Cap and yielding much lower RMSF than apo (Fig. 4a, b and Supplementary Fig. 7b–e). We then assessed CO₂ trapping in the GGCX–FIX simulation. The CO₂ molecule was stably bound within the hydrophobic pocket throughout the entire trajectory (Fig. 5a and Supplementary Fig. 7f and Supplementary Movie 2). Clustering analysis of CO₂ positions confirmed that it remained tightly localized within the pocket, highlighting the stability of its binding (Supplementary Fig. 7f). Furthermore, the radial distribution function (RDF) showed a complete absence of density of water molecules within a 5 Å radius of the CO₂ pocket (Supplementary Fig. 7g). This effective exclusion of water prevents undesired side reactions such as protonation¹. In contrast, simulations of GGCX mutants, F299A and H160A, revealed significant disruptions to CO₂ trapping (Supplementary Fig. 7f and Supplementary Movies 3, 4). Clustering analysis showed that, unlike the wild-type, CO₂ dispersed across multiple regions away from the hydrophobic pocket, illustrating the loss of binding stability in the mutants (Supplementary Fig. 7f and Supplementary Movies 3, 4). Correspondingly, both RDFs exhibited increased water density starting around 3 Å, indicating partial water penetration into the pocket originally occupied by CO₂ (Supplementary Fig. 7g).

The critical role of Phe299 in CO₂ trapping and GGCX function is highlighted by the identification of the F299S mutation as a pathogenic variant in VKD coagulation factor deficiency type 1 (VKCFD1) patients⁵¹ (Supplementary Fig. 6). This mutation completely abolished the GGCX enzymatic activity⁵², emphasizing its indispensable role in catalysis. Patients with VKCFD1, harboring different GGCX mutations, are often treated with high doses of vitamin K, but the therapeutic response varies among mutants⁵³. The F299S mutation, in particular, showed no response to vitamin K treatment⁵³. Our structural analysis provides a clear explanation for this phenomenon, as the F299S mutation disrupts CO₂ binding within the catalytic pocket, rendering GGCX completely inactive⁵².

Active site and catalytic mechanism

The GGCX-FIX structure reveals two spatially separated but functionally related active sites for KH₂ deprotonation and Glu carboxylation, respectively (Fig. 5b–d and Supplementary Fig. 8a, b). At the center of the KH₂ deprotonation site, Lys218—previously identified as the key general base^37,38—is properly oriented and stabilized by a panel of interactions (Fig. 5b, c and Supplementary Fig. 8a, b). Specifically, Lys218 is fixed in position through a stacking interaction with the conserved Trp223^GGCX, which restricts the flexibility of Lys218 aliphatic sidechain (Fig. 5b, c). The ε-amino group of Lys218 forms a hydrogen bond with the main-chain carbonyl of Met401 (Fig. 5c). Strikingly, the EM density map clearly reveals that the ε-amino group of Lys218 points away from the carboxyl group of Asp263 despite their close proximity (~4 Å) (Supplementary Fig. 8c). This suggests that there likely exists a strong electrostatic attraction that pulls Lys218 to the opposite direction of Asp263. Notably, this differs from the conformation reported in a recent study⁴⁰, where Lys218 and Asp263 were proposed to interact more directly.

The active site for Glu carboxylation is characterized by the optimal spatial arrangement of CO₂ and the glutamate substrate, consistent with a tightly coupled reaction mechanism⁵⁴. CO₂ is positioned ~4 Å from the glutamate γ-carbon and is stabilized by stacking with Phe299^GGCX (Fig. 5b). The carboxyl group of the Glu^FIX residue is stabilized by hydrogen-bonding interactions with Asn159^GGCX, His160^GGCX and Tyr395^GGCX, which could increase the electrophilicity of the γ-carbon of Glu^FIX and enhance its susceptibility to deprotonation (Fig. 5d). Mutational studies reveal that substitutions such as N159A, H160A, W223A, F299A and Y395A significantly impair carboxylase activity (Fig. 5e), highlighting the indispensable roles of these residues.

In the GGCX-FIX complex structure, it is noteworthy that the 1-hydroxyl of VK, which is proposed to be oxygenated to form a dialkoxide and function as the strong base for Glu deprotonation, is positioned ~6 Å away from the γ-carbon of Glu, while in sharp contrast the 3-hydroxyl of VK is only ~3 Å from Lys218 (Fig. 5b). Given that the initial deprotonation of KH₂ is the rate-limiting step⁵⁵, this distance difference suggests that the GGCX-FIX complex EM structure likely captures a pre-catalytic state before KH₂ deprotonation. To trap this state, all purification and grid-preparation steps were performed at 4 °C without any addition of KH₂, with warming to 25 °C serving as the trigger for carboxylation. This is supported by both biochemical and structural evidence. First, in vitro activity assays using purified GGCX-FIX (residues 1–100), GGCX-FIX and GGCX-BGP complexes without exogenous KH₂ displayed a detectable activity at 25 °C, strongly arguing that the reduced form of vitamin K (KH₂) is bound in the structure (Supplementary Fig. 9a). Second, the reactions exhibited a strong temperature dependence, with catalytic activity clearly observed at 25 °C but nearly undetectable at 4 °C (Supplementary Fig. 9a), suggesting that thermal activation is likely required to initiate the carboxylation reaction, even when KH₂ is already bound. Third, the structure of GGCX-FIX bound to VKO is almost identical to the KH₂-bound structure, suggesting that each structure represents a catalytically inactive state and pre-catalytic state, respectively (Supplementary Fig. 9b–f).

Although Lys218 has been identified as the general base responsible for KH₂ deprotonation^37,38, it is still not clear how its ε-amino group returns back to the deprotonated state after Lys218 abstracts a proton from KH₂ for continuous catalysis. Another unsolved puzzle is how the CO₂ molecule is captured by GGCX. The high-resolution GGCX-FIX complex structure bound with KH₂ provides us a unique opportunity to address these issues.

In the EM density map, we observed a clear, distinct density in close proximity to Lys218, appearing to coordinate an electrostatic interaction with the ε-amino group of Lys218 (Supplementary Fig. 8a–c). On one hand, this density is close to the solvent surface of GGCX, and on the other hand it is only ~7–8 Å from the CO₂ molecule. Based on these observations, we propose that the identity of this unique density is likely a bicarbonate molecule (HCO₃⁻). Indeed, an HCO₃⁻ model matches with the density and fits well with the local stereo-chemical environment (Supplementary Fig. 8a–c). In our 2.6-Å density map, the ε-amino nitrogen of Lys218^GGCX and the basic oxygen of HCO₃⁻ are connected by clear electron density with a distance of ~2 Å, indicative of a low-barrier hydrogen bond⁵⁶ that would lower the pKa of its ε-amino group and enhance its basicity (Fig. 5c and Supplementary Fig. 8a–c). Meanwhile, the carbonyl and hydroxyl groups of HCO₃⁻ form hydrogen bonds with mainchain amino group of Met401^GGCX and mainchain carbonyl of Ala214^GGCX, respectively (Fig. 5c and Supplementary Fig. 8a). This extensive interaction network could stabilize the bound HCO₃⁻ molecule, explaining its observed clear electron density.

Asp263^GGCX, located immediately adjacent to Lys218^GGCX, has been implicated in an acid–base relay that aligns Lys218 toward its deprotonated state⁴⁰. Although this interaction likely facilitates Lys218 for KH₂ deprotonation, protonation of Asp263–COO⁻ would not be effectively reversible under physiological conditions and thus likely could not serve as the ultimate proton acceptor. We therefore propose that, once Lys218–NH₂ abstracts a proton from KH₂, the resulting Lys218–NH₃⁺ transfers this proton directly to HCO₃⁻, forming transient carbonic acid (H₂CO₃) that rapidly decomposes into CO₂ and H₂O (Fig. 5f). This mechanism regenerates Lys218 in its deprotonated form without necessitating protonation–deprotonation cycling of Asp263.

The identification of bicarbonate in the active site near Lys218 allows us to propose a catalytic mechanism for GGCX, in which HCO₃⁻ both serves as the proton acceptor and upon decomposition, supplies the CO₂ for substrate carboxylation. In this model, CO₂ and HCO₃⁻ are in equilibrium in the aqueous environment. An HCO₃⁻ molecule enters GGCX and binds to the KH₂ deprotonation active site in close proximity to Lys218 via a network of hydrogen bonds (Fig. 5c, f). We propose that this delicate interaction geometry enables HCO₃⁻ to immediately accept a proton from Lys218 after it abstracts one from KH₂ (Fig. 5f). This coupled VK→Lys218→HCO₃⁻ proton transfer process maintains Lys218 in a constantly unprotonated state, ensuring its role as a general base in continuous KH₂ deprotonation. Upon accepting the proton from Lys218, HCO₃⁻ would spontaneously undergo hydrolysis to produce water (H₂O) and carbon dioxide (CO₂) molecules. This CO₂ molecule is then delivered into the CO₂ trapping pocket (Fig. 5a, f), where it is poised to react with the γ-carbanion of the Glu substrate in the next carboxylation cycle. Notably, no CO₂ density is observed in either the VKO- or anisindione-bound maps, and HCO₃⁻ remains locked in place without hydrolysis owing to these inhibitors’ binding (Supplementary Fig. 9g, h). This suggests that the CO₂ seen in our KH₂-bound structures likely originates specifically from HCO₃⁻ hydrolysis in the previous catalytic cycle rather than diffusion from solution. The deprotonation of KH₂, catalyzed by Lys218, enables O₂ to react with KH₂ to generate a highly reactive strong base, which abstracts the γ-proton from Glu (Fig. 5f). This base then forms a stabilized carbanion intermediate, which subsequently reacts with CO₂ to produce γ-carboxyglutamate (Fig. 5f).

In summary, this coordinated mechanism integrates coupled VK→Lys218→HCO₃⁻ proton transfer, bicarbonate hydrolysis, CO₂ delivering and trapping, and substrate carboxylation into a single catalytic cycle, with each cycle consuming one molecule of KH₂, one molecule of HCO₃⁻, and one molecule of CO₂.

Recognition of osteocalcin (BGP) and matrix Gla protein (MGP)

BGP and MGP exhibit structural distinctions compared to GLA-domain-containing VKD proteins, particularly in the distribution of Gla residues and propeptide affinity to GGCX²⁷ (Supplementary Fig. 10a). To elucidate how GGCX processes such diverse substrates, we determined the structures of GGCX bound to BGP and MGP at resolutions of 2.6 Å and 3.1 Å, respectively (Supplementary Fig. 10b–k). While these structures share conserved features of propeptide binding and Glu substrate recognition observed in the GGCX-FIX complex (Supplementary Fig. 10l–q), notable differences exist.

In the GGCX-BGP complex, BGP interacts with GGCX through two distinct binding motifs, the canonical propeptide region and a second exosite binding domain (EBD2)^52,57. Structural analysis of the propeptide interface shows that high-affinity substrates such as FIX and MGP have a conserved alanine at the –10 position that packs against Tyr425^GGCX, whereas BGP carries a glycine at this site, preventing that hydrophobic contact and accounting for its lower affinity (Supplementary Figs. 3d, and 10l–n). EBD2 fits into a concaved hydrophobic pocket formed by the CTD domain and lumenal loop LL1 of GGCX (Supplementary Fig. 11a, b). We propose that EBD2 compensates for the low affinity of the BGP propeptide, facilitating effective substrate recognition. The disease-associated R704X mutation⁵⁸, which truncates the GGCX CTD, could compromise the interactions with EBD2 (Supplementary Fig. 6), selectively impairing BGP carboxylation while sparing other VKD proteins⁵².

The structures also explain the unique Gla residue distribution patterns in BGP and MGP (Supplementary Fig. 11c, d). In BGP, Glu31 remains unmodified due to its proximity to the stable EBD2-binding interface, which blocks access to the active site (Supplementary Figs. 10j, and 11c). In MGP, the centrally positioned propeptide most likely uses two flexible linkers to deliver Gla residues at the N- and C-termini to the active site (Supplementary Figs. 10i, k, and 11d). However, the distances between residues such as Glu5^MGP, Glu8^MGP, and Glu11^MGP in linker 1 and the beginning of propeptide (Leu12^MGP: 30 Å from the active center) are too short for these residues to be positioned in the active site for carboxylation (Supplementary Figs. 10i, k, and 11d). In contrast, Glu2^MGP that is close enough to the active site can undergo modification⁵⁹(Supplementary Figs. 10i, k, and 11d).

Collectively, our findings reveal how GGCX processes structurally diverse VKD substrates. The discovery of EBD2 highlights a BGP-specific adaptation that enhances substrate recognition, while MGP’s two flexible linkers illustrate GGCX’s ability to process dispersed Gla residues at both sides of the propeptide.

Discussion

In this study, we elucidate the molecular mechanism of VKD carboxylation catalyzed by GGCX. By integrating cryo-EM structures, MD simulations, and biochemical assays, we propose an integrated model that explains how GGCX orchestrates processive carboxylation for diverse VKD substrates. Furthermore, we also reveal a distinct CO₂ fixation and carboxylation mechanism, emphasizing the dual role of bicarbonate (HCO₃⁻) in facilitating both CO₂ capture and catalytic activity for vitamin K-dependent carboxylase.

In this model, GGCX recruits VKD substrates through distinct mechanisms, high-affinity propeptide substrate binds to the canonical propeptide-binding cleft (Fig. 6a, steps 1–2), while low-affinity propeptide substrates like BGP employs a second high-affinity site (EBD2) to compensate for weak propeptide binding (Supplementary Fig. 11a). Propeptide binding triggers a large conformational change (Fig. 6a, steps 1–2), serving as the primary allosteric event to prime the enzyme for cofactor and substrate engagement, although the precise sequence of these binding events remains to be determined (Fig. 6a, steps 1–2). Glu residues are anchored in a dipeptide pattern by Tyr395^GGCX, which stacks with the “Eup” residue (Fig. 6a, step 3). Although the precise order of dipeptide modification awaits further study, we hypothesize that the Leu–Glu dipeptide closest to the propeptide anchor is preferentially presented to the active site due to minimal spatial repositioning, initiating the processive cycle. Each reaction consumes KH₂, producing VKO and converting Glu into Gla (Fig. 6a, step 4). The newly formed Gla, encapsulated in a hydrophobic environment, would destabilize its position and exit the active site after reaction, allowing the next Glu to enter (Fig. 6a, step 4). For the subsequent Glu residue, carboxylated Gla can itself function as an anchor, with its γ-carboxyl group stabilized by Arg406^GGCX and His440^GGCX, maintaining the stacking interaction with Tyr395 (Figs. 3e and 6a, step 5). This dipeptide anchoring would enable efficient processive carboxylation while the propeptide remains stably bound (Fig. 6a, steps 6–8). As carboxylation progresses, the final one or two Glu residues often deviate from the conserved dipeptide pattern. This deviation likely acts as a “stop” signal, promoting the release of the fully carboxylated product from GGCX (Fig. 6a, steps 8–1).

**Fig. 6: The processive carboxylation mechanism of VKD substrates by GGCX and its evolutionary conservation.**

Our study also uncovers a previously unrecognized CO₂ fixation strategy mediated by GGCX, revealing a unique mechanism distinct from all other known carboxylases and enzymes^1,60, which stabilizes CO₂ through metal ions or reactive intermediates^60,61. We propose HCO₃⁻ plays a dual role in the catalytic cycle of GGCX. First, it stabilizes Lys218^GGCX, enabling it to deprotonate KH₂ as a general base (Fig. 5f). Second, HCO₃⁻ accepts a proton from Lys218^GGCX, which results in its hydrolysis into water (H₂O) and carbon dioxide (CO₂) (Fig. 5f). The released CO₂ is then delivered to the trapping pocket for the subsequent reaction with the γ-carbanion of the Glu substrate (Fig. 5f). This interplay among proton transfer, bicarbonate hydrolysis, and CO₂ capture highlights the intricate coordination during GGCX catalysis. The bicarbonate-mediated proton transfer closely couples the redox reactions of vitamin K, which in turn connects the oxidation of vitamin K to the carboxylation of glutamate. This interplay creates a tightly integrated cycle of oxidation and reduction of VK and HCO₃^-/CO₂, enabling efficient carbon fixation within the GGCX catalytic cycle.

Evolutionary analyses demonstrate that this CO₂ fixation strategy is not restricted to eukaryotes but is broadly conserved across the bacterial kingdom (Fig. 6b). Critical residues involved in CO₂ stabilization and KH₂ deprotonation are highly conserved among orthologues (Fig. 6b). Indeed, the bacterial orthologue MloH from Streptomyces paromomycinus catalyzes aspartyl carboxylation during malonomycin biosynthesis⁶², underscoring the adaptability of this mechanism across diverse biological systems. The broad conservation of GGCX-like enzymes across bacteria and eukaryotes suggests that this strategy for CO₂ fixation may represent an ancient and widespread solution to a fundamental biochemical challenge.

This work not only advances our understanding of natural carboxylation mechanisms mediated by vitamin K-dependent carboxylase but also paves the way for exploring this underappreciated CO₂ fixation strategy in nature.

Methods

Construction of plasmids and cells

The GGCX coding sequence (Uniprot ID: P38435, a.a.1-758), containing an N-terminal Flag tag and a C-terminal Strep tag, was cloned into the lentiviral pLVX plasmid and stably transfected into Expi293F cells for protein purification. For in vivo activity assays, an untagged version of GGCX was cloned into the pcDNA3.1 plasmid. To obtain the GGCX-FIX, GGCX-BGP, and GGCX-MGP complexes, the GGCX coding sequence with a C-terminal Strep tag and the cDNAs of FIX (Uniprot ID: P00740, a.a. 1-461), BGP (Uniprot ID: P02818, a.a. 1–100), and MGP (Uniprot ID: P08493, a.a. 1-103) with C-terminal Flag tag were cloned into the pLVX vector. The GGCX-Strep construct was co-transfected with either FIX-Flag, BGP-Flag, or MGP-Flag into Expi293F cells to enable complex formation. Stable Expi293F cell lines expressing these complexes were first grown as adherent monolayers in DMEM+10% FBS, then transitioned to suspension culture in Expi293F medium (Cellwise, CW001). HEK293T cells were obtained from ATCC (#CRL-3216), Expi293F cells were obtained from Gibco (#A14527), GGCX knockout HEK293 cells were obtained from Ubigene (#YKO-XN26522) and further confirmed by western blot and sequencing.

Expression and purification of recombinant proteins

All proteins were stably expressed in Expi293F cells cultured at 37 °C with 5% CO₂. Once the cell density reached 4 × 10⁶ cells/mL, cells were harvested by centrifugation at 1000 × g for 20 min. For the purification of GGCX-Apo, GGCX-FIX, GGCX-BGP, and GGCX-MGP complexes, the cell pellet was resuspended in low-salt buffer (10 mM NaCl, 10 mM HEPES-NaOH, pH 7.5) and homogenized using a Dounce homogenizer (Sigma-Aldrich). Cell membranes were collected by ultracentrifugation at 100,000 × g for 50 min, followed by solubilization in buffer containing 150 mM NaCl, 1% LMNG, 0.1% CHS, and 25 mM HEPES-NaOH (pH 7.5) for 2 h. Insoluble material was removed by centrifugation at 30,000 × g for 50 min, and the solubilized protein was purified by incubation with Strep-Tactin XT 4Flow resin (IBA Lifesciences) for 2 h. The resin was washed with 20 column volumes (CV) of buffer containing 150 mM NaCl, 0.01% LMNG, 0.002% GDN, 0.001% CHS, and 25 mM HEPES-NaOH (pH 7.5). Bound protein was eluted with the same buffer supplemented with 50 mM biotin (Sigma-Aldrich). The eluted protein was further purified by incubation with Anti-Flag (DYKDDDDK) G1 Affinity Resin (GenScript) overnight, followed by elution using wash buffer containing 200 µg/mL 3× Flag peptide. Finally, the protein was subjected to size exclusion chromatography (SEC) on a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with SEC buffer (150 mM NaCl, 0.005% LMNG, 0.001% GDN, 0.0005% CHS, and 25 mM HEPES-NaOH, pH 7.5). Peak fractions were concentrated to 12 mg/mL using a 100-kDa molecular weight cutoff concentrator for cryo-EM sample preparation. No vitamin K was supplemented during purification or during preparation of cryo-EM specimens. For the purification of GGCX-Anisiondone, 10 µM anisindione was added during cell culture, and 200 µM anisindione was maintained in all buffers throughout the purification process. For GGCX-FIX_GlaE complex, FIX_GlaE peptides were added during the Flag resin binding step. To purify the GGCX-VKO (Vitamin K1 2,3-epoxide) complex, 100 μM warfarin and 10 μM VKO were added during cell culture. Additionally, all buffers used in the purification process were supplemented with 100 μM VKO.

In vitro carboxylation activity assay

Vitamin K (VK) solution (10 mg/mL) was mixed with a threefold volume of reaction buffer containing 20 mM Tris-HCl (pH 8.5), 200 mM DTT, and 500 mM NaCl to reduce VK to KH₂. The reaction was carried out at 37 °C overnight in darkness. The carboxylation reaction of FIX was performed in a total volume of 125 µL. The reaction mixture contained 25 mM HEPES-NaOH (pH 7.5), 150 mM NaCl, 20 µg/mL GGCX-FIX complex, 0.005% LMNG, 0.001% GDN, 0.0005% CHS and 5 mM NaHCO₃. The reaction was initiated by adding 0.5 mM of the reduced KH₂ solution and incubated at 25 °C. Samples were collected at 0, 0.5, 1, 2, 4, and 6 h, and the reaction was terminated by adding Sodium Dodecyl Sulfate (SDS). The collected samples were analyzed by sodium dodecyl sulfate - polyacrylamide gel electrophoresis (SDS-PAGE) and subsequently subjected to Western blotting using an anti-Gla-specific antibody (Invitech Limited, 3570), with anti-Strep (Genscript, A01732) and anti-Flag (Cell Signaling, D6W5B) antibodies used as controls.

In vivo carboxylation activity assay

GGCX mutants were generated by bridging PCR using the wild-type sequence as a template and cloned into the pcDNA3.1 vector. A GGCX knockout HEK293 cell line (Ubigene) was validated by sequencing and Western blotting. Full-length FIX cDNA with a C-terminal Flag tag was cloned into the pLVX vector and stably transfected into the knockout cells. Equal numbers of cells were seeded in 6-well plates, and after attachment, 2 µg of pcDNA3.1 plasmids carrying GGCX wild-type, mutants, or empty vector were transiently transfected. Cells were incubated for 18 h in serum-free medium containing 50 µM vitamin K, and supernatants were collected. For detection, 96-well plates were coated with 1 µg/mL rabbit anti-Flag antibody overnight at 4 °C, washed, and blocked for 1 h at 37 °C. Supernatants were added alongside serially diluted FIX-Flag protein standards and incubated for 1 h at 37 °C. After washing, 1 µg/mL mouse anti-Gla antibody and HRP-conjugated goat anti-mouse secondary antibody were sequentially applied for 1 h each at 37 °C. The reaction was developed using TMB substrate for 15 min, terminated with 1 M sulfuric acid, and absorbance was measured at 405 nm. All samples were tested in triplicate.

Cryo-EM sample preparation and data acquisition

Protein samples at 10–15 mg/mL were applied to glow-discharged Quantifoil Cu R1.2/1.3 grids, blotted for 1 s at 8 °C and 100% humidity, and vitrified in liquid ethane using an FEI Vitrobot Mark IV. Cryo-EM data were collected on a 300 kV FEI Titan Krios G3i microscope equipped with a K3 direct electron detector and an energy filter (20 eV slit width). For GGCX-Apo, GGCX-FIX, GGCX-Anisindione, GGCX-MGP, and GGCX-BGP, data were acquired at 81,000× magnification (pixel size: 1.1 Å) using EPU software for automated data acquisition. Each image was dose-fractionated into 32 frames over a 2.7-s exposure, with a total electron dose of ~50 e⁻/Å² per stack and a defocus range of −0.9 to −1.7 µm. The datasets included 10,105, 13,187, 9591, 1533, and 7375 movies, respectively. For GGCX-FIX_GlaE, 7,500 movies were collected at 105,000× magnification with a pixel size of 0.824 Å and a defocus range of −0.7 to −2.5 µm. For GGCX-VKO, 8745 movies were collected at 105,000× magnification with a pixel size of 0.86 Å and a defocus range of −0.9 to −1.7 µm.

Cryo-EM data processing

Movies were dose-weighted, aligned, and binned using MOTIONCOR2⁶³, and the contrast transfer function (CTF) of each micrograph was estimated with CtfFind 4.1⁶⁴. Particle picking was performed using Gautomatch (http://www.mrc-lmb.cam.ac.uk/kzhang/Gautomatch/) or Topaz v0.3.0⁶⁵. For the GGCX-Apo dataset, 12,112,131 particles were picked and extracted at a pixel size of 2.2 Å. After two rounds of 2D classification in CryoSPARC v4.4.1⁶⁶, 10,161,759 particles were subjected to heterogeneous refinement with ab-initio generated reference maps, reducing the dataset to 1,051,989 particles. These particles were re-extracted in RELION 5.0^67,68 at 1.1 Å pixel size for global 3D classifications, yielding two dominant classes; one represents the apo state (82,480 particles), and the other reveals copurified endogenous substrates and vitamin K (412,263 particles). Non-uniform refinement produced density maps at 3.24 Å and 3.63 Å, respectively. Repicking and further heterogeneous refinement yielded improved maps at 2.92 Å for the endogenous substrate-bound form and 3.46 Å for GGCX-Apo (347,146 particles), while focused 3D classification without alignment further improved the GGCX-Apo map to 3.4 Å (166,899 particles; Supplementary Table 1). A similar strategy was applied to the GGCX-MGP, GGCX-BGP, GGCX-FIX, GGCX-Anisindione, GGCX-FIX_GlaE and GGCX-VKO datasets, extracting 2,247,455, 15,006,064, 16,040,397, 11,207,547, 8,120,760, 7,107,588 particles, respectively, and yielding density maps at resolutions of 3.1 Å, 2.6 Å, 2.6 Å, 2.8 Å, 2.4 Å and 3 Å (Supplementary Table 1).

Model building and refinement

The structure was refined starting from an AlphaFold-predicted model (AF-P38435-F1-v4)⁶⁹, followed by iterative cycles of manual model building in COOT 0.9.4.1⁷⁰ and real-space refinement using Phenix 1.20.1⁷¹. Several ordered water molecules were modeled into the electron density maps with resolutions exceeding 2.8 Å. Validation statistics were obtained through MolProbity, integrated within the Phenix validation tools⁷². Structural figures were prepared using ChimeraX 1.6.1⁷³, PyMOL 3.1.3⁷⁴, and UCSF Chimera 1.16⁷⁵.

Molecular dynamics simulation

All-atom MD simulations were performed on four systems—apo GGCX, GGCX–FIX, and two active-site mutants (H160A, F299A)—using our cryo-EM structures as starting points. The apo system was built from the apo structure of GGCX with bound cholesterol, whereas the GGCX-FIX and the corresponding mutant systems consisted of GGCX, FIX peptide (residues −18 to 9), bound vitamin K, CO₂ and cholesterol. The active-site mutants (H160A and F299A) were created using PyMOL 3.1.3⁷⁴. Each model was embedded in an ER-like bilayer generated with Packmol-Memgen⁷⁶, with the lower leaflet composed of CHL1:DPPA:SOPA:SOPC:SLPC:POPE:SOPE:SLPE:SLPG:POPS:SOPS:SLPS:PSM:LSM by the ratio of 20:2:2:20:34:9:10:18:10:2:8:2:25:10, and the upper leaflet has the same composition with a different ratio of 20:2:2:6:10:14:16:28:10:7:29:6:13:8. The AMBER ff14SB⁷⁷ and Lipid21⁷⁸ force fields were applied for proteins and lipids, respectively. CO₂ and VK were parameterized using the GAFF2 force field and assigned AM1-BCC charges⁷⁹. The system was solvated with TIP3P water⁷⁸, and K⁺ and Cl⁻ ions were added to neutralize the charge. Energy minimization was first performed for 1000 cycles on CPU, followed by 10,000 cycles on GPU to remove steric clashes. The systems were heated to 100 K under constant volume, and subsequently to 300 K. Pre-equilibration was carried out under isothermal–isobaric (NPT) conditions with positional restraints progressively released. Then, an additional 100-ns NPT simulation was performed without restraints. Finally, production simulations of 1 μs were performed for each system, maintaining a temperature of 300 K with Langevin dynamics (collision frequency: 2.0 ps⁻¹) and regulating pressure using a Monte Carlo barostat. Electrostatic interactions were calculated using the particle mesh Ewald method, with real-space electrostatics and van der Waals interactions truncated at 10 Å. SHAKE constraints were applied to hydrogen atoms to enable a 2 fs time step⁸⁰, while hydrogen mass repartitioning⁸¹ was employed during final production simulations to extend the time step to 4 fs. All simulations were run on GPUs using the pmemd.cuda module in AMBER24⁷⁹ (all the simulation input files and output coordinates are provided in Supplementary Data 1 and simulation setup is shown in Supplementary Tables 2 and 3). Scatter plots of carbon atom positions in the CO₂ molecule and RDFs of water density versus radial distance from the CO₂ carbon atom were analyzed using cpptraj in AmberTools24.

Evolutionary analysis

To identify homologous sequences of GGCX protein, a BLASTP v2.15.0+ search⁸² was conducted using the human GGCX sequence as a query against the UniProtKB database. A total of 131 homologous protein sequences from species across Eukaryota and Bacteria were manually selected for evolutionary analysis. Protein sequences were aligned using MAFFT v7.48⁸³, and a phylogenetic tree was constructed with the maximum likelihood (ML) method implemented in IQ-TREE v2.3.6⁸⁴. The resulting tree was visualized using ggtree v3.10.1⁸⁵.

Statistical analyses

The statistical details and methods are indicated in the figure legends. All error bars represent the standard deviation of the mean. All tests for statistical significance were carried out using two-tailed unpaired t-test in GraphPad Prism. Experiments shown are representative of at least three biological replicates.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Atomic models have been deposited in the PDB, cryoEM maps deposited in the EMDB, and cryoEM maps deposited in the EMPIAR with accession codes: GGCX-Apo (PDB 9L1Y, EMD-62758, EMPIAR-12781), GGCX-FIX (PDB 9L21, EMD-62765, EMPIAR-12802), GGCX-FIX_GlaE (PDB 9L25, EMD-62769, EMPIAR-12807), GGCX-BGP (PDB 9L23, EMD-62767, EMPIAR-12805), GGCX-MGP (PDB 9L24, EMD-62768, EMPIAR-12806), GGCX-Anisindione (PDB 9L20, EMD-62764, EMPIAR-12803), GGCX-VKO (PDB 9L54, EMD-62822, EMPIAR-12804). Source data are provided with this paper. Please contact the corresponding author for any other materials. Source data are provided with this paper.

References

Rishavy, M. A. & Berkner, K. L. Vitamin K oxygenation, glutamate carboxylation, and processivity: defining the three critical facets of catalysis by the vitamin K–dependent carboxylase. Adv. Nutr. 3, 135–148 (2012).
Article CAS PubMed PubMed Central Google Scholar
Presnell, S. & Stafford, D. The vitamin K-dependent carboxylase. Thromb. Haemost. 87, 937–946 (2017).
Article Google Scholar
Berkner, K. L. & Runge, K. W. Vitamin K-dependent protein activation: normal gamma-glutamyl carboxylation and disruption in disease. Int. J. Mol. Sci. 23. https://doi.org/10.3390/ijms23105759 (2022).
Alonso, N., Meinitzer, A., Fritz-Petrin, E., Enko, D. & Herrmann, M. Role of vitamin K in bone and muscle metabolism. Calcif. Tissue Int. 112, 178–196 (2023).
Article CAS PubMed Google Scholar
Suttie, J. W. Vitamin K-dependent carboxylase. Annu. Rev. Biochem. 54, 459–477 (1985).
Article CAS PubMed Google Scholar
Presnell, S. R. & Stafford, D. W. The vitamin K-dependent carboxylase. Thromb. Haemost. 87, 937–946 (2002).
Article CAS PubMed Google Scholar
Berkner, K. L. The vitamin K-dependent carboxylase. Annu. Rev. Nutr. 25, 127–149 (2005).
Article CAS PubMed Google Scholar
Oldenburg, J., Marinova, M., Muller-Reible, C. & Watzka, M. The vitamin K cycle. Vitam. Horm. 78, 35–62 (2008).
Article CAS PubMed Google Scholar
Stafford, D. W. The vitamin K cycle. J. Thromb. Haemost. 3, 1873–1878 (2005).
Article CAS PubMed Google Scholar
Spronk, H. M., Farah, R. A., Buchanan, G. R., Vermeer, C. & Soute, B. A. Novel mutation in the gamma-glutamyl carboxylase gene resulting in congenital combined deficiency of all vitamin K-dependent blood coagulation factors. Blood 96, 3650–3652 (2000).
Article CAS PubMed Google Scholar
Li, Q. et al. Mutations in the GGCX and ABCC6 genes in a family with pseudoxanthoma elasticum-like phenotypes. J. Invest. Dermatol. 129, 553–563 (2009).
Article CAS PubMed Google Scholar
Watzka, M. et al. Bleeding and non-bleeding phenotypes in patients with GGCX gene mutations. Thromb. Res. 134, 856–865 (2014).
Article CAS PubMed Google Scholar
De Vilder, E. Y., Debacker, J. & Vanakker, O. M. GGCX-associated phenotypes: an overview in search of genotype-phenotype correlations. Int. J. Mol. Sci. 18 https://doi.org/10.3390/ijms18020240 (2017)
Tie, J.-K. et al. Characterization of vitamin K–dependent carboxylase mutations that cause bleeding and nonbleeding disorders. Blood 127, 1847–1855 (2016).
Article CAS PubMed PubMed Central Google Scholar
Zhu, A. et al. Fatal hemorrhage in mice lacking gamma-glutamyl carboxylase. Blood 109, 5270–5275 (2007).
Article CAS PubMed PubMed Central Google Scholar
Grosset, A. B., Allen, J. E. & Rodgers, G. M. Anticoagulation with anisindione in patients who are intolerant of warfarin. Am. J. Hematol. 46, 138–140 (1994).
Article CAS PubMed Google Scholar
Zhou, Y. et al. TTD: therapeutic target database describing target druggability information. Nucleic Acids Res. 52, D1465–D1477 (2024).
Article PubMed Google Scholar
Liu, S. et al. Structural basis of antagonizing the vitamin K catalytic cycle for anticoagulation. Science 371 https://doi.org/10.1126/science.abc5667 (2021)
Chen, X., Jin, D. Y., Stafford, D. W. & Tie, J. K. Evaluation of oral anticoagulants with vitamin K epoxide reductase in its native milieu. Blood 132, 1974–1984 (2018).
Article CAS PubMed PubMed Central Google Scholar
Benton, M. E., Price, P. A. & Suttie, J. W. Multi-site-specificity of the vitamin K-dependent carboxylase: in vitro carboxylation of des-gamma-carboxylated bone Gla protein and Des-gamma-carboxylated pro bone Gla protein. Biochemistry 34, 9541–9551 (1995).
Article CAS PubMed Google Scholar
Price, P. A., Fraser, J. D. & Metz-Virca, G. Molecular cloning of matrix Gla protein: implications for substrate recognition by the vitamin K-dependent gamma-carboxylase. Proc. Natl Acad. Sci. USA 84, 8335–8339 (1987).
Article CAS PubMed PubMed Central ADS Google Scholar
Engelke, J. A., Hale, J. E., Suttie, J. W. & Price, P. A. Vitamin K-dependent carboxylase: utilization of decarboxylated bone Gla protein and matrix Gla protein as substrates. Biochim. Biophys. Acta 1078, 31–34 (1991).
Article CAS PubMed Google Scholar
Viegas, C. S. et al. Gla-rich protein (GRP), a new vitamin K-dependent protein identified from sturgeon cartilage and highly conserved in vertebrates. J. Biol. Chem. 283, 36655–36664 (2008).
Article CAS PubMed PubMed Central Google Scholar
Jorgensen, M. J. et al. Recognition site directing vitamin K-dependent γ-carboxylation resides on the propeptide of factor IX. Cell 48, 185–191 (1987).
Article CAS PubMed Google Scholar
Huber, P. et al. Identification of amino acids in the gamma-carboxylation recognition site on the propeptide of prothrombin. J. Biol. Chem. 265, 12467–12473 (1990).
Article CAS PubMed Google Scholar
Stanley, T. B., Jin, D. Y., Lin, P. J. & Stafford, D. W. The propeptides of the vitamin K-dependent proteins possess different affinities for the vitamin K-dependent carboxylase. J. Biol. Chem. 274, 16940–16944 (1999).
Article CAS PubMed Google Scholar
Hao, Z., Jin, D. Y., Stafford, D. W. & Tie, J. K. Vitamin K-dependent carboxylation of coagulation factors: insights from a cell-based functional study. Haematologica 105, 2164–2173 (2020).
Article CAS PubMed PubMed Central Google Scholar
Higgins-Gruber, S. L. et al. Effect of vitamin K-dependent protein precursor propeptide, vitamin K hydroquinone, and glutamate substrate binding on the structure and function of gamma-glutamyl carboxylase. J. Biol. Chem. 285, 31502–31508 (2010).
Article CAS PubMed PubMed Central Google Scholar
Knobloch, J. E. & Suttie, J. W. Vitamin K-dependent carboxylase. Control of enzyme activity by the “propeptide” region of factor X. J. Biol. Chem. 262, 15334–15337 (1987).
Article CAS PubMed Google Scholar
Soute, B. A. et al. Congenital deficiency of all vitamin K-dependent blood coagulation factors due to a defective vitamin K-dependent carboxylase in Devon Rex cats. Thromb. Haemost. 68, 521–525 (1992).
Article CAS PubMed Google Scholar
Stenina, O., Pudota, B. N., McNally, B. A., Hommema, E. L. & Berkner, K. L. Tethered processivity of the vitamin K-dependent carboxylase: factor IX is efficiently modified in a mechanism which distinguishes Gla’s from Glu’s and which accounts for comprehensive carboxylation in vivo. Biochemistry 40, 10301–10309 (2001).
Article CAS PubMed Google Scholar
Lin, P. J., Straight, D. L. & Stafford, D. W. Binding of the factor IX gamma-carboxyglutamic acid domain to the vitamin K-dependent gamma-glutamyl carboxylase active site induces an allosteric effect that may ensure processive carboxylation and regulate the release of carboxylated product. J. Biol. Chem. 279, 6560–6566 (2004).
Article CAS PubMed Google Scholar
Dowd, P., Ham, S. W. & Hershline, R. Role of oxygen in the vitamin K-dependent carboxylation reaction: incorporation of a second atom of oxygen 18 from molecular oxygen-18O2 into vitamin K oxide during carboxylase activity. J. Am. Chem. Soc. 114, 7613–7617 (1992).
Article CAS ADS Google Scholar
Kuliopulos, A. et al. Dioxygen transfer during vitamin K dependent carboxylase catalysis. Biochemistry 31, 7722–7728 (1992).
Article CAS PubMed Google Scholar
Dowd, P., Hershline, R., Ham, S. W. & Naganathan, S. Vitamin K and energy transduction: a base strength amplification mechanism. Science 269, 1684–1691 (1995).
Article CAS PubMed ADS Google Scholar
Dowd, P., Ham, S.-W., Naganathan, S. & Hershline, R. The mechanism of action of vitamin K. Annu. Rev. Nutr. 15, 419–440 (1995).
Article CAS PubMed Google Scholar
Rishavy, M. A. et al. Brønsted analysis reveals Lys218 as the carboxylase active site base that deprotonates vitamin K hydroquinone to initiate vitamin K-dependent protein carboxylation. Biochemistry 45, 13239–13248 (2006).
Article CAS PubMed Google Scholar
Rishavy, M. A. et al. A new model for vitamin K-dependent carboxylation: the catalytic base that deprotonates vitamin K hydroquinone is not Cys but an activated amine. Proc. Natl Acad. Sci. USA 101, 13732–13737 (2004).
Article CAS PubMed PubMed Central ADS Google Scholar
Rishavy, M. A. & Berkner, K. L. Insight into the coupling mechanism of the vitamin K-dependent carboxylase: mutation of histidine 160 disrupts glutamic acid carbanion formation and efficient coupling of vitamin K epoxidation to glutamic acid carboxylation. Biochemistry 47, 9836–9846 (2008).
Article CAS PubMed Google Scholar
Cao, Q. et al. Molecular basis of vitamin-K-driven γ-carboxylation at the membrane interface. Nature 639, 816–824 (2025).
Article CAS PubMed ADS Google Scholar
Wang, R. et al. Structure and mechanism of vitamin-K-dependent γ-glutamyl carboxylase. Nature 639, 808–815 (2025).
Article CAS PubMed ADS Google Scholar
Zhong, Q. et al. Molecular basis of vitamin K-dependent protein γ-glutamyl carboxylation. Cell Res. https://doi.org/10.1101/2025.02.09.637231 (2025).
Article PubMed PubMed Central Google Scholar
Mutucumarana, V. P., Acher, F., Straight, D. L., Jin, D. Y. & Stafford, D. W. A conserved region of human vitamin K-dependent carboxylase between residues 393 and 404 is important for its interaction with the glutamate substrate. J. Biol. Chem. 278, 46488–46493 (2003).
Article CAS PubMed Google Scholar
Spyropoulos, A. C., Hayth, K. A. & Jenkins, P. Anticoagulation with anisindione in a patient with a warfarin-induced skin eruption. Pharmacother. J. Hum. Pharmacol. Drug Ther. 23, 533–536 (2003).
Article Google Scholar
Matschinske, J. et al. Individuating possibly repurposable drugs and drug targets for COVID-19 treatment through hypothesis-driven systems medicine using CoVex. ASSAY Drug Dev. Technol. 18, 348–355 (2020).
Article CAS PubMed PubMed Central Google Scholar
Chu, K., Wu, S. M., Stanley, T., Stafford, D. W. & High, K. A. A mutation in the propeptide of Factor IX leads to warfarin sensitivity by a novel mechanism. J. Clin. Invest. 98, 1619–1625 (1996).
Article CAS PubMed PubMed Central Google Scholar
Sugiura, I., Furie, B., Walsh, C. T. & Furie, B. C. Propeptide and glutamate-containing substrates bound to the vitamin K-dependent carboxylase convert its vitamin K epoxidase function from an inactive to an active state. Proc. Natl Acad. Sci. 94, 9069–9074 (1997).
Article CAS PubMed PubMed Central ADS Google Scholar
Li, S., Furie, B. C., Furie, B. & Walsh, C. T. The propeptide of the vitamin K-dependent carboxylase substrate accelerates formation of the gamma-glutamyl carbanion intermediate. Biochemistry 36, 6384–6390 (1997).
Article CAS PubMed Google Scholar
Vanakker, O. M. et al. Pseudoxanthoma elasticum-like phenotype with cutis laxa and multiple coagulation factor deficiency represents a separate genetic entity. J. Investig. Dermatol. 127, 581–587 (2007).
Article CAS PubMed Google Scholar
Parker, C. H. et al. A conformational investigation of propeptide binding to the integral membrane protein gamma-glutamyl carboxylase using nanodisc hydrogen exchange mass spectrometry. Biochemistry 53, 1511–1520 (2014).
Article CAS PubMed Google Scholar
Dordoni, C. et al. Characterization of a Pseudoxanthoma elasticum-like patient with coagulation deficiency, cutaneous calcinosis and GGCX compound heterozygosity. J. Dermatol. Sci. 89, 201–204 (2018).
Article PubMed Google Scholar
Hao, Z. et al. gamma-Glutamyl carboxylase mutations differentially affect the biological function of vitamin K-dependent proteins. Blood 137, 533–543 (2021).
Article CAS PubMed PubMed Central Google Scholar
Ghosh, S. et al. GGCX mutations show different responses to vitamin K thereby determining the severity of the hemorrhagic phenotype in VKCFD1 patients. J. Thromb. Haemost. 19, 1412–1424 (2021).
Article CAS PubMed Google Scholar
Rishavy, M. A., Hallgren, K. W. & Berkner, K. L. The vitamin K-dependent carboxylase generates gamma-carboxylated glutamates by using CO2 to facilitate glutamate deprotonation in a concerted mechanism that drives catalysis. J. Biol. Chem. 286, 44821–44832 (2011).
Article CAS PubMed PubMed Central Google Scholar
Davis, C. H., Deerfield, D., Wymore, T., Stafford, D. W. & Pedersen, L. G. A quantum chemical study of the mechanism of action of Vitamin K carboxylase (VKC): III. Intermediates and transition states. J. Mol. Graph. Model. 26, 409–414 (2007).
Article CAS PubMed Google Scholar
Cleland, W. W., Frey, P. A. & Gerlt, J. A. The low barrier hydrogen bond in enzymatic catalysis *. J. Biol. Chem. 273, 25529–25532 (1998).
Article CAS PubMed Google Scholar
Houben, R. J. et al. Osteocalcin binds tightly to the gamma-glutamylcarboxylase at a site distinct from that of the other known vitamin K-dependent proteins. Biochem J. 341, 265–269 (1999).
Article CAS PubMed PubMed Central Google Scholar
Darghouth, D. et al. Compound heterozygosity of a W493C substitution and R704/Premature stop codon within the γ-glutamyl carboxylase in combined vitamin K-dependent coagulation factor deficiency in a French family. Blood 114, 1302–1302 (2009).
Article Google Scholar
Price, P. A. & Williamson, M. K. Primary structure of bovine matrix Gla protein, a new vitamin K-dependent bone protein. J. Biol. Chem. 260, 14971–14975 (1985).
Article CAS PubMed Google Scholar
Bierbaumer, S. et al. Enzymatic conversion of CO(2): from natural to artificial utilization. Chem. Rev. 123, 5702–5754 (2023).
Article CAS PubMed PubMed Central Google Scholar
Prywes, N., Phillips, N. R., Tuck, O. T., Valentin-Alvarado, L. E. & Savage, D. F. Rubisco function, evolution, and engineering. Annu. Rev. Biochem. 92, 385–410 (2023).
Article CAS PubMed Google Scholar
Law, B. J. C. et al. A vitamin K-dependent carboxylase orthologue is involved in antibiotic biosynthesis. Nat. Catal. 1, 977–984 (2018).
Article CAS Google Scholar
Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).
Article CAS PubMed PubMed Central Google Scholar
Rohou, A. & Grigorieff, N. CTFFIND4: Fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015).
Article PubMed PubMed Central Google Scholar
Bepler, T. et al. Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs. Nat. Methods 16, 1153–1160 (2019).
Article CAS PubMed PubMed Central Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS PubMed Google Scholar
Kimanius, D., Forsberg, B. O., Scheres, S. H. & Lindahl, E. Accelerated cryo-EM structure determination with parallelisation using GPUs in RELION-2. eLife 5, e18722 (2016).
Article PubMed PubMed Central Google Scholar
Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519–530 (2012).
Article CAS PubMed PubMed Central Google Scholar
Abramson, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024).
Article CAS PubMed PubMed Central ADS Google Scholar
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr D. Biol. Crystallogr. 60, 2126–2132 (2004).
Article PubMed ADS Google Scholar
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D. Biol. Crystallogr. 66, 213–221 (2010).
Article CAS PubMed PubMed Central ADS Google Scholar
Davis, I. W. et al. MolProbity: all-atom contacts and structure validation for proteins and nucleic acids. Nucleic Acids Res. 35, W375–W383 (2007).
Article PubMed PubMed Central ADS Google Scholar
Pettersen, E. F. et al. UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82 (2021).
Article CAS PubMed Google Scholar
DeLano, W. The PyMOL Molecular Graphics System. http://www.pymol.org (2002).
Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).
Article CAS PubMed ADS Google Scholar
Schott-Verdugo, S. & Gohlke, H. PACKMOL-Memgen: a simple-to-use, generalized workflow for membrane-protein-lipid-bilayer system building. J. Chem. Inf. Model. 59, 2522–2528 (2019).
Article CAS PubMed Google Scholar
Maier, J. A. et al. ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB. J. Chem. Theory Comput. 11, 3696–3713 (2015).
Article CAS PubMed PubMed Central Google Scholar
Dickson, C. J., Walker, R. C. & Gould, I. R. Lipid21: complex lipid membrane simulations with AMBER. J. Chem. Theory Comput 18, 1726–1736 (2022).
Article CAS PubMed PubMed Central Google Scholar
Case, D. A. et al. AmberTools. J. Chem. Inf. Model. 63, 6183–6191 (2023).
Article CAS PubMed PubMed Central Google Scholar
Ryckaert, J.-P., Ciccotti, G. & Berendsen, H. J. C. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J. Comput. Phys. 23, 327–341 (1977).
Article CAS ADS Google Scholar
Sahil, M., Sarkar, S. & Mondal, J. Long-time-step molecular dynamics can retard simulation of protein-ligand recognition process. Biophys. J. 122, 802–816 (2023).
Article CAS PubMed PubMed Central ADS Google Scholar
Camacho, C. et al. BLAST+: architecture and applications. BMC Bioinform. 10, 421 (2009).
Article Google Scholar
Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol. Biol. Evol. 30, 772–780 (2013).
Article CAS PubMed PubMed Central Google Scholar
Nguyen, L. T., Schmidt, H. A., von Haeseler, A. & Minh, B. Q. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol. Biol. Evol. 32, 268–274 (2015).
Article CAS PubMed Google Scholar
Yu, G., Smith, D., Zhu, H., Guan, Y. & Lam, T. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol. Evol. 8 https://doi.org/10.1111/2041-210X.12628 (2016).
Laskowski, R. A. & Swindells, M. B. LigPlot+: multiple ligand-protein interaction diagrams for drug discovery. J. Chem. Inf. Model. 51, 2778–2786 (2011).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

We thank all the staff members of the Electron Microscopy System, Mass Spectrometry System and The Imaging System at Shanghai Institute of Precision Medicine for providing technical support and assistance in data collection. This work was supported by grants from the National Key Research and Development Program of China (2020YFA0509800 to P.L.), the National Natural Science Foundation of China (32471344 to P.L.,31570235 and 31770259 to L.M.).

Author information

These authors contributed equally: Ke Wu, Zheng Wang, Deqiang Yao.

Authors and Affiliations

Institute of Aging and Tissue Regeneration, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Ke Wu, Zheng Wang, Deqiang Yao, Ming Lei & Pengfei Lan
Shanghai Institute of Precision Medicine, Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Shaobai Li, Xiaozhu Wang, Yuanyuan Zhang, Mi Cao, Yafeng Shen, Jian Wu, Ming Lei & Pengfei Lan
Department of Critical Care Medicine, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Shunpeng Xing
State Key laboratory of Oncogenes and Related Genes, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Ming Lei & Pengfei Lan

Authors

Ke Wu
View author publications
Search author on:PubMed Google Scholar
Zheng Wang
View author publications
Search author on:PubMed Google Scholar
Deqiang Yao
View author publications
Search author on:PubMed Google Scholar
Shaobai Li
View author publications
Search author on:PubMed Google Scholar
Xiaozhu Wang
View author publications
Search author on:PubMed Google Scholar
Yuanyuan Zhang
View author publications
Search author on:PubMed Google Scholar
Mi Cao
View author publications
Search author on:PubMed Google Scholar
Yafeng Shen
View author publications
Search author on:PubMed Google Scholar
Shunpeng Xing
View author publications
Search author on:PubMed Google Scholar
Jian Wu
View author publications
Search author on:PubMed Google Scholar
Ming Lei
View author publications
Search author on:PubMed Google Scholar
Pengfei Lan
View author publications
Search author on:PubMed Google Scholar

Contributions

K.W. and Z.W. purified the complex. K.W., S.L., M.C., P.L., and Y.S. prepared cryo EM specimens, collected data sets and determined the structure. D.Y. and J.W. carried out model building and refinement. K.W., Z.W., and S.X. carried out biochemical analysis. S.L. carried out the MD simulation analyses. X.W. and Y.Z. performed the evolutionary analyses. All authors contributed to data interpretation and the writing of the manuscript. P.L., M.L., and K.W. wrote the manuscript with input from all authors. All authors edited and approved the final manuscript. P.L., M.L., and J.W. initiated and orchestrated the project.

Corresponding authors

Correspondence to Ming Lei or Pengfei Lan.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Takahiro Kosugi, Jian-Ke Tie, and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information (download PDF )

Peer Review File (download PDF )

Description of Additional Supplementary Files (download PDF )

Supplementary Data 1 (download ZIP )

Supplementary Movie 1 (download AVI )

Supplementary Movie 2 (download MP4 )

Supplementary Movie 3 (download MP4 )

Supplementary Movie 4 (download MP4 )

Reporting Summary (download PDF )

Source data

Source Data (download ZIP )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Wu, K., Wang, Z., Yao, D. et al. Structural insight into bicarbonate-mediated carboxylation by human vitamin K-dependent carboxylase. Nat Commun 16, 10480 (2025). https://doi.org/10.1038/s41467-025-65488-3

Download citation

Received: 26 March 2025
Accepted: 14 October 2025
Published: 25 November 2025
Version of record: 25 November 2025
DOI: https://doi.org/10.1038/s41467-025-65488-3