Structural insights into the substrate uptake and inhibition of the human creatine transporter (hCRT)

Significance

Creatine is well known as an energy booster for muscles, but it also plays a key role in brain function and has recently been recognized as a neurotransmitter. Its transport into cells relies on a protein called human creatine transporter (hCRT). Malfunction of hCRT causes cerebral creatine deficiency syndrome 1 (CCDS1), and some cancers exploit this transporter to meet their energy demands. To address these issues, we resolved the high-resolution structures of hCRT. These atomic-level views reveal how hCRT recognizes and transports creatine, how a cancer drug candidate can inhibit its function, and how CCDS1 mutations impair its activity. This structural insight lays the groundwork for therapies targeting CCDS1 and cancers reliant on creatine metabolism.

Abstract

Creatine plays a vital role in cellular energy production and adenosine triphosphate (ATP) homeostasis and has also been identified as a neurotransmitter in the mammalian brain. Creatine is transported into cells by the human creatine transporter (hCRT) (SLC6A8), an Na⁺/Cl⁻-dependent symporter encoded on the X chromosome. Mutations in hCRT cause cerebral creatine deficiency syndrome 1, a neurological disorder marked by intellectual disability, speech delay, and seizures. Beyond its role in the brain and muscle, hCRT is highly expressed in metabolically active tumors. Many cancer cells, including colorectal cancer and glioblastoma, upregulate hCRT to sustain intracellular creatine levels and buffer ATP under energy stress. Pharmacological blockade of hCRT by RGX202 has been shown to impair tumor growth by disrupting energy homeostasis. Here, we report the high-resolution cryo-Electron Microscopy (cryo-EM) structures of human hCRT in three states: apo, creatine-bound, and RGX202-bound. hCRT adopts a canonical LeuT-fold with 12 transmembrane helices and two pseudosymmetric inverted repeats. Creatine is coordinated in the central substrate-binding site through interactions with transmembrane helices TM1, TM3, TM6, and TM8, while the inhibitor RGX202 occupies the same binding pocket, engaging in overlapping contacts that competitively block creatine access. Our structural and mechanistic findings clarify substrate recognition and inhibitory binding of hCRT, providing a molecular rationale for targeting hCRT in both inherited metabolic diseases and cancer therapy.

Creatine is an important molecule in regulating cellular energy metabolism, playing a crucial role in ATP homeostasis by participating in the creatine-phosphocreatine cycle, particularly in neurons and muscle cells (1–7). In this system, creatine is phosphorylated by creatine kinase to generate phosphocreatine, which acts as a temporal energy buffer during periods of high metabolic demand (3, 6, 8, 9). Accordingly, creatine supplementation is commonly used to enhance physical performance and energy buffering capacity (3, 5, 10, 11). Notably, a recent study has identified creatine as a neurotransmitter in the mammalian brain (12). It meets several criteria for neurotransmitter classification: It is synthesized in neurons, stored in synaptic vesicles (SVs), released upon stimulation, and interacts with postsynaptic receptors (12).

The human creatine transporter (hCRT, encoded by the SLC6A8 gene on the X chromosome) mediates cellular uptake of creatine in an Na⁺/Cl⁻-dependent manner (13–15). hCRT is highly expressed in tissues with elevated energy requirements, including brain astrocytes (16), skeletal muscle, heart, kidney, and small intestine (8), where it plays a central role in maintaining ATP levels and supporting bioenergetic homeostasis. hCRT belongs to the Neurotransmitter Sodium Symporters (NSSs) (8, 17) and facilitates the cotransport of neurotransmitters and other small molecules. Structural studies of proteins in NSSs reveal a LeuT-fold architecture (18–22), consisting of 12 transmembrane helices (18, 22, 23). Despite structural conservation, NSSs exhibit divergent substrate specificities: For example, hCRT selectively transports the zwitterionic compound creatine, whereas other NSS members such as dopamine transporter (DAT) (24–26) and serotonin transporter (SERT) (23, 27, 28) transport monoamines including dopamine and serotonin. High-resolution structures have been determined for several NSS proteins, including GABA transporter 1 (GAT1) (29–31), DAT (24–26), SERT (23, 27, 28), norepinephrine transporter (NET) (32–35), and taurine transporter (TauT) (36), offering valuable mechanistic insights into substrate recognition and transport. However, until recently, structural understanding of hCRT remained limited.

Mutations in hCRT cause cerebral creatine deficiency syndrome 1 (CCDS1), an X-linked neurological disorder characterized by intellectual disability, speech delay, seizures, and impaired energy metabolism (37–39). Male patients typically exhibit more severe symptoms, including cognitive impairment, behavioral abnormalities, and muscle weakness (13, 39, 40). In neurons, hCRT is indispensable for sustaining ATP production, synaptic transmission, and neural signaling (13). Disruption of creatine transport has also been implicated in mood disorders and cognitive decline (9, 13, 41). Emerging evidence suggests that creatine is also a critical element of the tumor microenvironment (42–45). Rapidly proliferating cancer cells, including those in colorectal cancer (44) and glioblastoma (46), upregulate hCRT expression to enhance creatine uptake, thereby sustaining intracellular energy reserves under metabolic stress (11, 46). Pharmacological blockade of hCRT using competitive inhibitors such as RGX202 (also known as ompenaclid) has been shown to impair tumor growth in preclinical models by restricting creatine-dependent ATP production (12). In addition to its antitumor effects, hCRT inhibition sensitizes cancer cells to chemotherapeutic agents and modulates immune responses within the tumor microenvironment (44, 46). These findings highlight hCRT as a promising therapeutic target for metabolically active malignancies (46, 47).

In this study, we determined the cryo-EM structures of hCRT in three states: the apo state, the creatine-bound state, and its competitive inhibitor RGX202-bound state at the atomic level. These structures elucidate the molecular basis of substrate recognition and competitive inhibition. Furthermore, our integrative analysis of CCDS1-associated mutations reveals how these variants perturb substrate coordination, ion coupling, or conformational gating, offering a structural basis for rational correction of hCRT dysfunction.

Results

Overall Architecture of hCRT.

To gain a comprehensive understanding of the molecular mechanisms underlying substrate uptake and inhibition of hCRT, we designed a C-terminally Strep-tagged SLC6A8 and utilized the HEK293F system to express recombinant hCRT for cryo-EM analysis (SI Appendix, Extended Data Fig. S1). This approach generated structures embedded in Lauryl Maltose Neopentyl Glycol (LMNG) micelles in three distinct states: apo (hCRTapo), creatine-bound (hCRTC), and RGX202-bound (hCRTR) (Fig. 1). Three-dimensional (3D) reconstruction and de novo calculations using cryoSPARC resulted in high-resolution structures, with resolutions of 3.28 Å, 2.85 Å, and 2.75 Å, respectively (SI Appendix, Extended Data Figs. S2–S4) (Table 1).

Fig. 1.

Overall architecture of hCRT. (A) Cryo-EM maps of the hCRT structure in three states: hCRTapo (cyan) at 3.28 Å, hCRTC (slate) at 2.85 Å, and hCRTR (salmon) at 2.75 Å resolution. The detergent micelles are shown as gray density. (B) Topological schema of hCRT, displaying transmembrane helices (TM), extracellular loops (EL), and intracellular loops (IL). Substrate sites are marked with *, and disulfide bonds with S–S. (C) Ribbon models of hCRTapo (cyan), hCRTC (slate), and hCRTR (salmon) with bound Na⁺ and Cl⁻ ions (purple and green spheres, respectively).

Table 1.

Cryo-EM data collection, refinement and validation statistics

	hCRT–apo	hCRT–Creatine	hCRT–RGX202
Data collection and processing
Magnification	105,000	105,000	105,000
Voltage (kV)	300	300	300
Electron exposure (e–/Å2)	50	50	50
Defocus range (μm)	−1.5 ~ −2.5	−1.5 ~ −2.5	−1.5 ~ −2.5
Pixel size (Å)	0.856	0.8433	0.8433
Symmetry imposed	C1	C1	C1
Initial particle images (no.)	1,800 K	1,085 K	2,073 K
Final particle images (no.)	132 K	79 K	85 K
Map resolution (Å) FSC threshold	3.28 0.143	2.85 0.143	2.75 0.143
Map resolution range (Å)	3.0–4.6	2.7–4.6	2.6–4.6
Refinement
Initial model used (PDB code)	de novo, AlphaFold	de novo, AlphaFold	de novo, AlphaFold
Model resolution (Å) FSC threshold	3.2	3.1	3.0
Map sharpening B factor (Å2)	187.2	130.5	128.2
Model composition
Non-hydrogen atoms	3,766	4,038	4,039
Protein residues	476	504	505
Ligands	1	4	4
B factors (Å2)
Protein	51.29	35.32	33.69
Ligand	28.58	24.96	30.85
rmsd
Bond lengths (Å)	0.004	0.004	0.003
Bond angles (°)	0.809	0.625	0.648
Validation
MolProbity score	2.14	2.16	1.33
Clashscore	13.30	20.18	5.25
Poor rotamers (%)	0.00	0.00	0.00
Ramachandran plot
Favored (%)	91.42	94.94	97.78
Allowed (%)	8.58	4.86	2.22
Disallowed (%)	0.00	0.2	0.00

All three structures adopt the canonical LeuT-fold architecture characteristic of the NSS family (18–22), with 12 transmembrane (TM) helices arranged in two pseudosymmetrical repeats (TM1-5 and TM6-10), measuring 60 Å × 55 Å (Fig. 1 A and B). Both the N and C termini extend to the cytoplasmic side, while due to highly disordered nature, models for these regions were difficult to build. Notably, the well-defined peripheral loop contains a glycosylation site at N548, a crucial modification that is functionally linked to protein activity and transport efficacy. Additionally, a conserved disulfide bond (C172-C181) in the EL2 region of the substrate-bound hCRT structure likely plays a role in maintaining conformational stability (Fig. 1C).

Creatine-Bound Structure and Substrate-Binding Pocket of hCRT.

We determined the creatine-bound structure of hCRT (hCRTC) at 2.85 Å resolution, allowing precise identification of substrates, ions, and water molecules (Figs. 1 and 2 and SI Appendix, Extended Data Fig. S3). Surface Plasmon Resonance (SPR) analysis yielded an affinity of 3.2 µM for creatine (Fig. 2C). hCRTC structure adopts an occluded state. Creatine binds between TM1 and TM6 unwound region and is oriented parallel to the membrane, preventing access to either the extracellular or intracellular sides (Figs. 1A and 2H). TM1a bends over the cytoplasmic cavity and occludes the intracellular pathway and a hydrophobic cap formed by the side chains of Y148/TM3 and F315/TM6 blocks the substrate-binding site. The substrate-binding pocket is sealed by interaction networks of polar and π–π interactions (Fig. 2I).

Fig. 2.

Substrate recognition, ion coordination, and creatine-induced conformational changes in hCRT. (A) Heatmap of surface electrostatics around hCRTC’s central binding pocket, with creatine shown as sticks. (B) Cryo-EM density map of the substrate-binding pocket in hCRTC, showing creatine (orange sticks), Na⁺ ions (purple spheres), and Cl⁻ ion (green sphere) at 5σ contour level. (C) SPR analysis of creatine binding to hCRT, with a dissociation constant (KD) of 3.2 µM. (D) Enlarged view of the creatine binding site. (E–G) Coordination networks for Na⁺1, Na⁺2, and Cl⁻ in the hCRTC structure. (H) Structural alignment of hCRTapo (cyan) and hCRTC (slate), highlighting conformational changes upon substrate binding. The boxed region marks the positional shift of TM1a and TM6b. (I) Magnified view of the boxed region in (H), illustrating the inward tilt of TM1a (~21.3°) and lateral displacement of TM6b (~1.7 Å) upon creatine binding. (J) Superposition of TM1a, TM5, and TM6b helices between hCRTapo and hCRTC. Structural rearrangements of residues such as F68, M63, and L327 support gating transitions. (K) Comparison of ion-binding sites between hCRTapo and hCRTC. Coordinating residues are shown as slate sticks, with creatine in orange line and interactions indicated by dashed lines.

The local resolution of the hCRTC substrate-binding pocket reaches 2.7 Å (SI Appendix, Extended Data Fig. S3D). Within this pocket, creatine is stabilized by potential hydrogen bonds with the side chain of Y148, the carbonyl oxygens of F68 and F315, the nitrogen atom of backbone Gly71, and water molecule. Notably, three aromatic residues (F68, Y148, and F315) surround the guanidinium group of creatine to form strong π–π interactions (Fig. 2 D and I). Substitution of aromatic residues by alanine results in a 20-fold decrease in affinity, suggesting that the aromatic rings are critical in substrate binding (SI Appendix, Extended Data Fig. S6 A–F). Meanwhile, a thiol-mediated hydrogen bond with C144 and creatine further anchors the substrate. C144, a unique feature of hCRT and corresponds to a more hydrophobic residue in other NSSs (SI Appendix, Extended Data Fig. S7), is essential for stabilizing the substrate-binding site through its interaction with the guanidinium group (Fig. 2D). This interaction enhances the affinity of creatine to the transporter and ensures efficient substrate recognition and translocation, and it also significantly influences the conformation of the binding pocket to promote optimal substrate arrangement and improve transport efficiency. Studies have shown that deprotonation of Cys144 enables the formation of a salt bridge with the guanidine moiety of creatine, and mutations at this position not only impair creatine transport but also drastically increase GABA uptake (48). Substitution of cysteine with alanine reduces the affinity for creatine by over 30-fold, suggesting that the C144 is critical in substrate binding (SI Appendix, Extended Data Fig. S6C).

Ion Coordination in hCRTC.

hCRT uptake activity requires the participation of Na⁺ and Cl^– ions (49). In the occluded state, two Na⁺ ions and one Cl^– ion are observed, which aligns with the conserved ion-binding patterns observed in other NSS family members, located between TM1 and TM8 and between TM1 and TM6/TM7, respectively (Figs. 1C and 2 B and E–G). Na⁺1, located between TM1 and TM6, forms a five-coordinate geometry with the carboxyl group of creatine, the hydroxyl group of S316, the amide groups of N74/N348, and the main chain carboxyl group of A69, with coordination distances of approximately ~2.5 Å (Fig. 2E). Na⁺2, positioned between TM1 and TM8, coordinates with the carbonyl oxygens of the main chains of G67, V70, and L413, as well as the hydroxyl group of S417, with a coordination distance of average approximately 2.5 Å (Fig. 2F). The Cl^– ion, located between TM1, TM6, and TM7, interacts with the side chains of N74, Y94, S316, and S352 (Fig. 2G), forming a putative quadrilateral geometry. These ion coordination networks are conserved among NSS family members, but notable differences in local residues distinguish hCRT’s specificity.

Structural Basis of Creatine-Induced Conformational Changes.

To detect the structural basis of substrate-induced conformational changes in hCRT, we compared the hCRTapo and hCRTC and examined the rearrangements at the substrate-binding core. Despite an overall good alignment between the two structures (Cα rmsd = 0.845 Å over 438 residues), a pronounced conformational shift is observed in the gating helix TM1a, which undergoes a tilt of approximately 21.3° during the transition from the occluded to the inward-open state to facilitate creatine release (Fig. 2 H and I). Structural comparison between the hCRTapo and hCRTC states revealed that the movement of TM1a initiates from the substrate-binding site and extends toward the N terminus. As TM1a swings, the side chain of F68 shifts to occupy the position previously held by L327. To accommodate this change, L327 itself undergoes a positional adjustment. Concomitantly, residues on TM5 also respond to the gate opening. In particular, M63 extends further toward the intracellular side, leading to coordinated displacements along TM5 (Fig. 2J). Interestingly, a lateral displacement of approximately 1.7 Å was detected in the intracellular portion of TM6b, which movement may arise from steric clashes and relieving spatial compression and allowing the hydrophobic environment surrounding creatine to stabilize (Fig. 2 H and I). Notably, this shift occurs independently of TM 1b and TM 6a, both of which retain relatively stable conformations during the transition, likely to maintain closure of the extracellular gate and insulate the binding site from extracellular influences (Fig. 2H). The transition from the occluded to inward-open state reveals a gating mechanism in which TM1a swings away from the substrate-binding site to open the cytoplasmic gate, facilitating the release of creatine into the intracellular space.

In the hCRTapo structure, only the conserved Cl⁻ ion is clearly resolved, whereas the densities for the two Na⁺ ions are absent. To investigate the structural consequences of ion loss, we compared the ion-coordinating residues between the two states. In the hCRTapo structure, TM1a undergoes a pronounced displacement, disrupting the canonical Na⁺ coordination geometry, which likely contributes to the absence of Na⁺ density. Notably, S417, a coordinating residue for the Na2 site, undergoes a marked backbone rotation exceeding 60°, forming a new 3.4 Å hydrogen bond with an adjacent backbone carbonyl of L413. This rearrangement is associated with a slight outward tilt of TM8 (~2°), moving away from the Na2 site (Fig. 2K). The transition from the occluded to inward-open state induces subtle displacements in surrounding helices, which in turn influence Cl⁻ coordination. Residues such as N74 and S352 shift slightly, resulting in an ~1 Å positional displacement of the Cl⁻ ion in hCRTapo compared to hCRTC (Fig. 2K).

Structural Determinants of Substrate Specificity.

To gain deeper insights into the molecular basis of substrate specificity among NSS family members, we performed a comparative structural analysis of the substrate-binding pocket of the hCRT alongside representative transporters, including SLC6A1 (GAT1), SLC6A2 (NET), SLC6A3 (DAT), SLC6A4 (SERT), SLC6A6 (TauT), and SLC6A9 (GlyT1) (Fig. 3). Structural alignment revealed that hCRT possesses a highly specialized substrate-binding site (Fig. 3 C–E). This specificity is primarily shaped by a unique combination of nonconserved residues. Most notably, C144 on TM3, F68, and G71 on TM1 together establish a precise spatial and chemical microenvironment tailored for recognizing and coordinating the guanidinium moiety of creatine (Fig. 3 B–G).

Fig. 3.

Comparative analysis of the substrate-binding pockets across hCRT and other SLC6 family transporters. (A) Structural alignment of the creatine-bound hCRT (hCRTC, gray) with representative SLC6 family transporters, including SLC6A1 (GAT1, light pink, PDB: 7Y7W), SLC6A2 (NET, dusty rose, PDB: 8ZPB), SLC6A3 (DAT, olive green, PDB: 8Y2D), SLC6A4 (SERT, olive yellow, PDB: 7LIA), SLC6A6 (TauT, teal blue, PDB: 9JD5), and SLC6A9 (GlyT1, brownish maroon, PDB: 8WFI). All structures are shown in cartoon representation, with bound substrates depicted as lines. (B–G) Detailed views of substrate-binding pockets from panel (A), highlighting key differences in substrate recognition. Each panel corresponds to the aligned transporter and retains the same color scheme; creatine from the hCRTC structure is overlaid in each comparison are shown as slate lines. Conserved and nonconserved residues within the binding pockets are shown as sticks.

Compared to monoamine transporters (MATs) such as NET, DAT, and SERT, hCRT displays notable divergence in the architecture of the unwound regions of TM1 and TM6. In hCRT, G71 and L321 create a pocket that facilitates the anchoring of creatine, whereas the equivalent positions in MATs are occupied by bulkier acidic or aromatic residues (e.g., aspartate and phenylalanine), resulting in a more compact pocket and repositioning the substrate closer to TM3. In contrast, the smaller side chain of G71 in hCRT permits creatine to be centrally positioned within the substrate-binding cavity. At this interface, hCRT uniquely utilizes C144 to anchor the guanidinium group of creatine via a thiol-mediated salt bridge, a mode of interaction not observed in MATs (Fig. 3 C–E).

Comparison with GlyT1, a member of the amino acid transporter clade within the SLC6 family, further underscores hCRT’s structural divergence. In GlyT1, a bulky tryptophan residue (W322) narrows the binding pocket and positions glycine closer to TM1, contrasting with hCRT, where L321 occupies the analogous position and supports a broader, more open pocket accommodating the larger creatine molecule (Fig. 3G). Although hCRT, GAT1, and TauT are all members of the GABA transporter subclade of the SLC6 family and exhibit comparable structural frameworks, hCRT displays a different mode of substrate positioning and stabilization. In hCRT, C144 positions creatine nearly parallel to the membrane plane, whereas in GAT1, Y60 anchors GABA at a tilted angle (Fig. 3A). Furthermore, structural comparison with TauT reveals that both creatine and taurine align similarly within the binding cavity and are coordinated by conserved tyrosine residues (Y148 in hCRT, Y138 in LeuT). However, hCRT utilizes C144 to interact with the guanidinium group, whereas in TauT, the distal sulfonate group of taurine is stabilized by E406—a bulky acidic residue replaced by glycine (G421) in hCRT (Fig. 3F).

In addition to these positional differences, hCRT employs a set of aromatic and polar interactions critical for substrate stabilization. Notably, F68, Y148, and F315 form a tight network of π–π stacking and hydrogen bonding interactions that anchor the creatine molecule within the binding pocket (Fig. 2 D and I). Sequence alignment further shows that these positions are frequently occupied by aromatic residues such as phenylalanine and tyrosine across the NSS family, suggesting a conserved role for π–π stacking in substrate recognition (SI Appendix, Extended Data Fig. S7). Moreover, comparative alignment of creatine transporters orthologs from different species reveals high sequence similarity, particularly at residues involved in substrate and ion coordination (SI Appendix, Extended Data Fig. S8). The combination of these structural features underlies the exceptional substrate selectivity and functional specialization of hCRT.

hCRT Inhibition by RGX202.

RGX202 is a competitive inhibitor of hCRT currently undergoing clinical evaluation for colorectal cancer therapy (44). As a substituted guanidine derivative, RGX202 contains a canonical guanidinium group and adopts a linear configuration, with the guanidinium moiety connected via a flexible alkyl linker to a terminal carboxylic acid. This arrangement closely mimics the electrostatic profile and spatial orientation of creatine. This resemblance allows RGX202 to occupy the hCRT substrate-binding pocket in a conformation similar to that of creatine, effectively mimicking its coordination mode (Fig. 4I). SPR assays revealed that RGX202 binds hCRT with an affinity of 3.39 μM, comparable to that of creatine (Fig. 4C).

Fig. 4.

Structural basis of hCRT inhibition by RGX202. (A) Electrostatic surface potential of the hCRTR substrate-binding pocket. (B) Cryo-EM density map of hCRTR, with RGX202 shown as teal sticks, Na⁺ ions as purple spheres, and Cl⁻ ion as a green sphere. Density for RGX202 is contoured at 5σ. (C) SPR analysis of RGX202 binding to hCRT, with a KD of 3.39 µM. (D) Coordination of RGX202 in the hCRTR binding pocket. (E–G) Detailed views of ion coordination in hCRTR: (E) Na⁺1 site, (F) Na⁺2 site, and (G) Cl⁻ site. (H) Overall structural comparison between hCRTC and hCRTR, showing minimal conformational changes upon inhibitor binding (Cα rmsd = 0.224 Å over 451 residues). (I) Superimposition of substrate-binding pockets from hCRTC and hCRTR. Coordinating residues are shown as salmon sticks; RGX202 is shown as teal lines; hydrogen bonds and coordination interactions are shown as dashed lines.

To elucidate the structural basis of this interaction, we determined the cryo-EM structure of the RGX202-bound hCRT complex (hCRTR) at 2.75 Å resolution in the occluded conformation (SI Appendix, Extended Data Fig. S4). Within the binding pocket, RGX202 is oriented parallel to the membrane and surrounded by key residues F68, C144, Y148, G71, F315, and L321 (Fig. 4D). Mutagenesis of these residues to alanine or phenylalanine reduced RGX202 binding affinity by ~30-fold, underscoring their critical roles in ligand recognition and stabilization (SI Appendix, Extended Data Fig. S6 G–L). The conserved Na⁺ and Cl⁻ coordination networks are preserved in the hCRTR structure (Fig. 4 E–G). Compared to creatine, RGX202 contains a bulkier alkyl chain, which extends approximately 1 Å closer to the unwound region of TM1, enabling additional potential hydrogen bonds with the backbone amides of L72 and G73 (Fig. 4D). These interactions likely contribute to its enhanced binding stability. Notably, RGX202 induces a displacement of Na1 by ~1.1 Å, potentially optimizing ion coordination and further stabilizing the inhibitor within the binding pocket (Fig. 4I).

Structural Mapping of CCDS1 Mutations.

Mutations of the SLC6A8 gene are causally linked to cerebral CCDS1, a neurodevelopmental disorder characterized by impaired creatine uptake, intellectual disability, and seizures (15, 37, 50). While previous studies have attributed pathogenicity to loss of transporter expression or mislocalization, our western blot analysis demonstrates that multiple clinically observed hCRT variants are robustly expressed in cells at levels comparable to wild-type (WT) protein (Fig. 5B), indicating that impaired function arises from postexpression structural disruption rather than reduced abundance.

Fig. 5.

Mapping of the CCDS1-related mutations in hCRTR. (A) Mapped CCDS1 mutations in hCRTR. Yellow spheres indicate deletion mutations, and orange spheres indicate amino acid substitutions. (B) Western blot analysis of representative CCDS1-associated hCRT variants.

To elucidate the structural basis of dysfunction, we mapped CCDS1-associated mutations onto our high-resolution hCRT structures and analyzed their spatial distribution and potential mechanistic impacts (Fig. 5A). We categorized these mutations into four distinct groups based on their structural and functional impact: 1) Ion transport channel mutations: Disrupted ion transport channel mutations including Y80H (51), G381R (40, 52), G383C (53), N336del (54–56), and C337W (54), lie in or adjacent to the substrate and ion permeation pathway. For instance, C337W replaces a small cysteine with a bulky tryptophan near the Na⁺/Cl⁻-coordinating region, likely introducing steric hindrance that disrupts interhelical packing and disturbs the geometry required for efficient solute binding and translocation. 2) Protein folding mutations like G87R (52), C181R (54), and G132V (57) replace conserved glycine or cysteine residues with bulkier, often charged side chains. These positions are typically located within tight turns or disulfide-constrained loops, and such replacements are predicted to alter local conformational plasticity, disulfide formation, or helical anchoring. 3) Transmembrane helix interface mutations: Mutations in helical regions, such as Y80H, F107del (58), G132V, I347del (58), F408del (47, 59), A448D (53), C491W (57), and V539I (53), could weaken the stability of transmembrane helices and their interactions. Even minor changes in conserved residues could lead to alterations of local structures that impact protein function. (4) Transmembrane domain–solute interface mutations: Such as N336del, C337W, P390L (52), R391W (58), G499del (54), V552L (60), and P554L (52), likely interfere with the mechanical coupling of structural domains necessary for conformational cycling during transport, and affect the flexibility and stability of the hCRT structure, ultimately disrupting the solute transport process (Table 2).

Table 2.

Classification of CCDS1-associated mutations based on their structural and functional impacts on hCRTR

Mutations	Predicted structure and functional impact	Classification
Y80H (51), G381R (40, 52), G383C (53), N336del (54–56), C337W (54)	Impedes binding by ions and substrate	Ion transport channel mutations
G87R (52), C181R (54), G132V (57)	Disrupts local structural stability and disulfide-bond formation	Protein folding mutations
Y80H (51), F107del (58), G132V (57), I347del (58), F408del (47, 59), A448D (53), C491W (57), V539I (53)	Weakens stability of transmembrane helices	Transmembrane helix interface mutations
N336del (54–56), C337W (54), P390L (52), R391W (58), G499del (54), V552L (60), P554L (52)	Affect flexibility and stability of hCRT structure	Transmembrane domain-solute interface mutations

Discussion

As a substance and a molecule important for energy homeostasis, the recent description of creatine as a neurotransmitter (12) reignites the interest of its study. Given its pivotal role in energy homeostasis and its involvement in causing CCDS1 (15, 37, 52), unraveling the structure of hCRT is essential. Despite substantial structural insights available for members of the NSSs family, including GAT1 (29–31), DAT (24–26), SERT (23, 27, 28), NET (32–35), and TauT (36), high-resolution structural information for hCRT remained unavailable until this study.

In the current study, we utilized cryoelectron microscopy to have resolved three high-resolution structures of hCRT in both inward-open (hCRTapo) and occluded states (hCRTC and hCRTR), derived from the full-length hCRT protein without auxiliary stabilization methods such as Fab fragments or nanobodies, allowing us to observe authentic conformational dynamics and ligand-induced structural rearrangements. Our results indicate that the helical region between TM1 and TM6 is crucial for recognizing substrates and facilitating conformational changes in response to drugs. Specifically, several key residues including F68 in TM1, Y148 in TM3, and F315 in TM6 with the interaction network formed orchestrate the gating and stabilization processes essential for substrate translocation.

Comparison of apo state and substrate-bound state revealed substantial conformational shifts, notably in TM1a and TM6b, which shifted by 21.3° and 1.7 Å, respectively. This gating mechanism is consistent with those observed in other NSSs transporters. Given the lack of the cyro-EM resolved outward-open state of hCRT, we employed AlphaFold2 to predict this missing state (Fig. 6A). The resulting model, with a mean Predicted Local Distance Difference Test (pLDDT) score of 83.57, suggests a high degree of structural reliability at the local residue level (Fig. 6B). Structurally, the predicted model displays hallmark features of the outward-open state commonly observed in NSS family transporters: TM1a and TM6b adopt a closed conformation, effectively sealing the intracellular gate, while TM1b and TM6a swing outward to form an open extracellular vestibule that permits substrate entry (SI Appendix, Extended Data Fig. S9A). Based on this conformation, we propose a stepwise mechanism for creatine transport: hCRT initially captures extracellular creatine or substrate analogs in the outward-open state; conformational rearrangement of TM1b and TM6a then transitions the transporter to an occluded state, locking the substrate inside and preventing reverse transport; subsequently, the intracellular gate opens via outward movement of TM1a and TM6b, enabling substrate release into the cytosol (Fig. 6C).

Fig. 6.

Proposed model of the hCRT transport cycle. (A) AlphaFold2-predicted structure of full-length hCRT (hCRT_AF, orange) in an outward-open conformation. (B) Predicted per-residue confidence score (pLDDT) plotted against residue number. The average pLDDT score is 83.57, with helical regions exhibiting high confidence (>90). (C) Schematic model of the hCRT transport cycle based on structural predictions. Key gating helices TM1a, TM1b, TM6a, and TM6b are shown in orange, cyan, and slate, and other helices in gray. Substrates and inhibitors are shown as orange spheres or sticks. Solid boxes denote experimentally determined cryo-EM structures obtained in this study; the dashed box indicates the AlphaFold2-predicted model.

Additionally, our cryo-EM structure of the hCRTR complex reveals that RGX202 occupies the same substrate-binding pocket as creatine, thereby acting as a competitive inhibitor that directly blocks substrate access. This competitive binding is mediated by key residues within a conserved interaction network crucial for ligand recognition and stabilization. High-resolution structural alignment with hCRTC confirms that RGX202 mimics the pose of creatine in the occluded conformation but introduces additional steric and polar interactions, including a bulkier alkyl chain and hydrogen bonding with L72 and G73, enhancing its binding stability. Functionally, this inhibition disrupts the creatine-phosphocreatine cycle, leading to reduced intracellular ATP availability—a mechanism particularly effective against metabolically active tumors such as colorectal cancer (44) and glioblastoma (46), which are highly dependent on creatine uptake for energy homeostasis. Understanding the structural basis of hCRT inhibition thus provides critical insight for the rational design of next-generation anticancer agents targeting the substrate-binding site. These findings open a therapeutic avenue by enabling the development of small-molecule inhibitors that exploit the creatine transport machinery to suppress tumor bioenergetics, offering a targeted strategy for cancers reliant on creatine-mediated ATP production.

Furthermore, we analyzed a panel of hCRT mutations (51–60) that are clinically associated with CCDS1. Despite prior hypotheses attributing pathogenicity to impaired protein biosynthesis or membrane trafficking, our western blot analyses revealed that these mutant variants are expressed at levels comparable to WT hCRT, suggesting that the primary cause of dysfunction lies in impaired transporter activity rather than protein abundance. Structural mapping of these mutations onto our high-resolution cryo-EM models allowed us to classify them into four mechanistic categories based on their spatial localization and functional disruption. These structural insights offer a detailed molecular basis for how individual missense or deletion mutations can compromise creatine transport efficiency and trigger disease phenotypes.

In conclusion, our study provides molecular understandings for substrate binding as well as the binding of Na⁺ and Cl⁻ and the conformational changes of hCRT during the transport cycle. Furthermore, we define the binding mode of RGX202 as a competitive inhibitor occupying the creatine-binding site, providing structural rationale for its therapeutic use in metabolically active tumors. Together, these findings establish a robust framework for structure-guided drug discovery targeting hCRT. Future research should extend the investigation into the interactions between hCRT and other cellular signaling pathways and would combine emerging technologies for real-time monitoring of the dynamic changes of hCRT to promote new progress toward the treatment of diseases including CCDS1, cancer, infections, and autoimmune disorders.

Materials and Methods

Materials.

For molecular cloning, the following reagents were utilized: a seamless assembly cloning kit (C5891) from CloneSmarter and a QuikChange site-directed mutagenesis kit (200518) from Agilent. In the field of cell culture, the human embryonic kidney (HEK) 293F cell line (R79007) was obtained from Thermo Fisher Scientific, with the cell culture medium (SMM 293-TII) obtained from Sino Biological. The Dulbecco’s modified Eagle’s medium (C11995599BT) and fetal bovine serum (10091148) were sourced from Gibco, while the penicillin-streptomycin solution (SV30010) was acquired from Hyclone. Additionally, linear polyethyleneimine (PEI, MW 25,000; 23966) was purchased from Polysciences. For protein purification, the following reagents were acquired: HEPES (4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid; H3375) and NaCl (S9888) from Sigma-Aldrich, and an inhibitor cocktail tablet (04693116001) from Roche. The 10:1 LMNG/Cholesteryl Hemisuccinate Tris Salt (CHS) premixed solution (NG310-CH210) was obtained from Anatrace. Strep-Tactin resin (2-1208-500) and D-desthiobiotin (2-1000-005) were procured from IBA. RGX202 (CAS No.: 353-09-3) and creatine (CAS No.: 57-00-1) were all acquired from MedChemExpress. For immunoblotting, mouse anti-Flag monoclonal antibody (catalog no. BE7015) and mouse anti-actin monoclonal antibody (catalog no. BE0021) were ordered from Bioeasytech, and the Goat Anti-Mouse IgG, HRP Conjugated Antibody (catalog no. CW0102) was ordered from CWbiotech.

SLC6A8 Construct Design and Cell Culture.

The full-length human SLC6A8 sequence was retrieved from the UniProt database (UniProt ID: Q8N697). Subsequently, the synthesis company performed codon optimization and synthesis of the SLC6A8 complementary DNA to enhance expression in mammalian systems. A custom construct was designed for protein expression and cellular assays and subsequently subcloned into the pCAGGS vector, which includes a C-terminal tandem Strep tag. These plasmids were propagated and amplified using the Escherichia coli strain DH5α, cultured at 37 °C in Luria-Bertani medium (Sigma). For protein expression studies, HEK293F cells were cultured in medium supplemented with 1× penicillin/streptomycin (Hyclone) under conditions of 37 °C and 5% CO₂, shaking at 120 rpm in a Multitron-Pro shaker (Infors).

Transporter Expression and Purification.

The SLC6A8 protein was expressed in cell culture using 800 mL of Sino Biological medium, incubated at 37 °C, 5% CO₂, and shaking at 120 rpm in a humidified environment. When the cell density reached 2 × 10⁶ per milliliter, transfection was initiated using 25 kDa linear PEI at a DNA ratio of 1:1.2. Two milligrams of plasmid DNA were incubated with 2.5 mg of PEI in 50 mL of fresh medium for 20 min to prepare the transfection mixture, which was then added to the cell culture. After an additional incubation of 2 d, the cells were harvested for protein purification. Approximately 6 L of transfected cells were centrifuged at 4,000 rpm to collect the cell pellet, which was then resuspended and lysed using a high-pressure homogenizer in buffer containing 25 mM HEPES (pH 7.4), 150 mM NaCl, protease inhibitors, and 10 mM β-mercaptoethanol. The lysate was centrifuged at 8,000 rpm for 45 min to remove cell debris, followed by ultracentrifugation at 45,000 rpm for 1 h to isolate membrane fractions.

The membrane fractions were solubilized in lysis buffer supplemented with 1% (w/v) LMNG and 0.1% (w/v) CHS, gently rotating for 2 h. The sample was then ultracentrifuged at 45,000 rpm for 1 h, and the supernatant was filtered through a 0.45 μm filter and loaded onto a column containing Strep-Tactin Sepharose resin. The column was washed with 50 column volumes of wash buffer (25 mM HEPES pH 7.4, 150 mM NaCl, 10 mM β-mercaptoethanol, 0.006% LMNG, and 0.0006% CHS) before eluting the target protein using the same buffer supplemented with 10 mM desthiobiotin. The eluted fraction was concentrated to approximately 100 μL using a 50 kDa cutoff centrifugal filter (Millipore) and further purified by size exclusion chromatography (SEC) (Superose 6 5/150, GE Healthcare) in SEC buffer containing 25 mM HEPES (pH 7.4), 150 mM NaCl, 0.006% (w/v) LMNG, and 0.0006% (w/v) CHS. The protein was analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and the suitable fractions for cryo-EM were concentrated to 25 to 30 mg mL⁻¹. All purification steps were performed at 4 °C or on ice to ensure protein integrity.

SPR Assay.

In this study, the SPR assay was employed to investigate the binding behavior of creatine and RGX202 as ligands, using SLC6A8 (hCRT) as the target protein. The experiments were conducted using a Biacore 8 K plus (Cytiva) equipped with a CM5 sensor chip. The protein and ligands were prepared in a buffer consisting of 25 mM HEPES, 300 mM NaCl, and 0.003% LMNG (pH 7.4). The purified hCRT protein (0.5 mg mL⁻¹) was immobilized on the CM5 sensor chip using amine coupling chemistry. The chip surface was activated using a mixture of 0.2 M EDC and 0.05 M NHS, followed by the injection of 50 µL of the hCRT solution. Unreacted sites were blocked with 1 M ethanolamine (pH 8.5). The ligand solutions of creatine and RGX202 were prepared at 25 µM and injected over the immobilized protein at a flow rate of 30 µL/min. Real-time changes in refractive index were monitored to capture the association and dissociation phases. The sensorgrams were analyzed using Biacore 8K plus evaluation software. The association rate constant (k_on) and dissociation rate constant (k_off) were determined by fitting the data to a 1:1 Langmuir binding model. The equilibrium dissociation constant (KD) was calculated based on the ratio of k_off to k_on. SPR measurements for hCRT mutants at key substrate-binding residues were also performed using the same procedure. Both creatine and RGX202 exhibited high-affinity binding to hCRT, validating the utility of SPR in studying protein–ligand interactions.

Cryo-EM Sample Preparation and Data Collection.

For other samples, purified hCRT at a concentration of 25 to 30 mg mL⁻¹ was incubated with 10 mM creatine and 10 mM RGX202 separately for 1 h to prepare cryo-EM grids. Using the Vitrobot Mark IV (FEI) under controlled conditions of 100% humidity and 8 °C, 3 to 4 μL of the 25 to 30 mg mL⁻¹ sample was deposited onto glow-discharged holey carbon grids (Quantifoil Au R1.2/1.3, 300 mesh) for cryo-EM grid preparation. After an 8-s equilibrating period, the grid was blotted for 2 to 5 s and then immediately plunged into liquid ethane for rapid vitrification. The grids were initially screened using a 200 kV Tecnai Arctica microscope (FEI) equipped with a Falcon 3 direct electron detector (FEI). Grids meeting quality standards were transferred to a Titan Krios microscope (FEI) operating at 300 kV, equipped with a 20 eV slit width energy filter (GIF Quantum LS, Gatan) for detailed data collection. Images were acquired in superresolution mode using a K3 direct electron detector (Gatan), with a nominal magnification of 105,000× and a calibrated pixel size of 0.42165 Å. Data collection was automatically performed in movie mode using AutoEMation2 (61), with a frame exposure time of 0.04 s, a total exposure time of 1.28 s per stack, generating 32 frames per stack, and a total dose of approximately 50 e⁻ Å⁻². Image processing included aligning and summing all 32 frames of each stack using MotionCor2 (62), followed by merging the resultant images to achieve a final pixel size of 0.8433 Å (hCRTC and hCRTR) and 0.856 Å (hCRTapo).

Image Processing and 3D Reconstruction.

For the cryo-EM dataset of the hCRTapo, hCRTC, and hCRTR complexes, a specific number of micrographs were collected for each state. Image processing was performed using cryoSPARC (63). Initially, 4,252, 5,124, and 4,299 micrographs from the hCRTapo, hCRTC, and hCRTR datasets were imported, and CTF correction was performed using Patch CTF estimation. After discarding low-quality micrographs from hCRTapo, 200 micrographs were selected for automatic picking using blob picker, and these particles were subjected to two-dimensional (2D) classification. The class averages representing projections of hCRTapo in various orientations were selected as templates for template picking from the entire dataset. From the selected 4,175, 5,020, and 4,137 micrographs, particles were automatically picked using blob picker for 2D classification, and class averages were selected as templates for further picking. A total of 7,153 K, 6,450 K, and 6,391 K particles were extracted from the 4,175, 5,020, and 4,137 micrographs, subjected to double binning and 2D classification. After three rounds of 2D classification, approximately 1,800 K, 2,001 K, and 1,226 K particles exhibiting clear 2D averages were re-extracted and subjected to binning, followed by de novo reconstruction and heterogeneous refinement. Class average analysis in Chimera (64) led to further refinement to exclude suboptimal particles, resulting in high-quality particles that underwent additional processing to enhance data resolution. 3D reconstruction was performed in cryoSPARC via nonuniform (NU) refinement (65), achieving 3.28 Å, 2.85 Å, and 2.75 Å resolution maps for each complex without applying symmetry, and visualized in ChimeraX (66). Detailed information regarding these procedures is documented in SI Appendix, Extended Data Figs. S2–S4.

Model Building and Refinement.

The conformational model of SLC6A8 predicted by AlphaFold (67, 68) was integrated into the cryoelectron microscopy density map for model building. Based on the highest resolution of hCRTR, de novo model building was initially performed using the hCRTR map. The predicted structural model was docked into the cryoelectron microscopy map at a resolution of 3.02 Å, followed by manual refinement in COOT (69) to establish the atomic model. The main chain of the atomic model was refined using the real_space_refine module of PHENIX (70), incorporating secondary structure and geometric constraints to mitigate overfitting. After preliminary manual adjustments in COOT, the model underwent iterative refinement using PHENIX to optimize the fit to the density map. Detailed information on the cryoelectron microscopy data collection and refinement statistics can be found in Table 1.

Immunoblotting.

To assess whether WT hCRT and CCDS1-associated hCRT mutants exhibit comparable expression levels, HEK293F cells (10 mL culture per condition) were transiently transfected with hCRT-FLAG WT or mutant constructs using a DNA:PEI ratio of 10 μg:25 μg. After 48 h, cells were harvested and lysed in 200 μL of lysis buffer per mL of culture. The lysis buffer consisted of 25 mM HEPES (pH 7.4), 150 mM NaCl, 1% (w/v) Triton X-100, and a protease inhibitor cocktail. Cell lysates were incubated at 4 °C for 2 h with gentle rotation, followed by centrifugation at 15,000× g for 10 min to remove debris. Supernatants were collected, and 20 μL of each was mixed with SDS loading buffer. Equal amounts (10 μL) of each sample were resolved by 4 to 20% SDS-PAGE (GenScript) and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore). PVDF membranes were blocked with 5% nonfat dry milk (Bio-Rad) in PBS containing 0.1% Tween-20 (PBS-T) for 1 h at room temperature (RT), then incubated with mouse anti-FLAG monoclonal antibody (catalog no. BE7015; dilution 1:3,000) and mouse anti-actin monoclonal antibody (catalog no. BE0021; dilution 1:3,000) for 1 h at RT. After three washes in PBS-T (10 min each), membranes were incubated with goat anti-mouse IgG-HRP–conjugated secondary antibody (CWbiotech, catalog no. CW0102; dilution 1:3,000) for 1 h at RT. Membranes were then washed three additional times in PBS-T, and protein bands were visualized using enhanced chemiluminescence substrate (Pierce) and imaged with the Amersham Imager 600 system (GE Healthcare).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The 3D cryoelectron microscopy density maps and associated atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB) with the following accession codes: apo state (hCRTapo): [EMD-64855 (71) and pdb 9V8Y (72)]; creatine-bound state (hCRTC): [EMD-64854 (73) and pdb 9V8X (74)]; and RGX202-bound state (hCRTR): [EMD-64853 (75) and pdb 9V8W (76)].