Crystal structure of human Dickkopf 4 cysteine-rich domain 1 and evaluation of conformational rigidity

This article describes the structural characterization of cysteine-rich domain 1 of Dickkopf 4. Comparison with the solution NMR structure indicates that regions displaying increased conformational variability may be relevant for protein–protein interactions.

Abstract

Dickkopf (DKK) family proteins (DKK1–DKK4), which function as extracellular modulators of Wnt signaling, contain two cysteine-rich domains: CRD1 and CRD2. In DKK1, CRD1 modulates interaction with its receptor low-density lipoprotein receptor-related protein (LRP) 5/6, whereas CRD2 directly binds to LRP5/6. The crystal structure of human DKK4-CRD1 was determined at 1.83 Å resolution. Crystals were obtained from refolded protein expressed as inclusion bodies and belonged to space group P2₁, with two molecules in the asymmetric unit. Initial molecular-replacement attempts using the solution NMR structure were unsuccessful, whereas an AlphaFold2-predicted model provided a clear solution. The refined structure reveals a compact fold comprising N- and C-subdomains connected by a linker region and stabilized by five conserved disulfide bonds. The crystal structure closely resembles the AlphaFold2 model, but shows larger deviations from the NMR ensemble. ANSURR analysis and hydrogen-bond comparisons indicate that the NMR models underestimate structural rigidity, particularly in β-sheet regions, owing to fewer stabilizing hydrogen bonds. Notably, enhanced conformational variability is observed in the N-subdomain, suggesting a potential role for structural plasticity in ligand recognition.

1. Introduction

The Dickkopf (DKK) family of secreted glycoproteins represents a pivotal class of extracellular modulators of Wnt signaling, a pathway essential for embryonic development, stem-cell regulation and tissue homeostasis. In mammals, the DKK family comprises DKK1, DKK2, DKK3, DKK4 and the related DKKL1 (Soggy). DKK1–DKK4 contain two conserved cysteine-rich domains (CRD1 and CRD2) and an N-terminal signal peptide (Niehrs, 2006 ▸). In contrast, DKKL1 contains a domain similar to the N-terminal region of DKK3 but lacks both CRD1 and CRD2. Despite these shared structural features, the family members exhibit distinct receptor affinities, expression profiles and biological functions (Giralt et al., 2021 ▸; Kikuchi et al., 2022 ▸). Comparative sequence analyses reveal moderate overall amino-acid identity (approximately 30–50%) among DKK1–DKK4, with the C-terminal CRD2 region being the most conserved. CRD2 contains ten invariant cysteine residues that stabilize protein folding through disulfide bonding (Brott & Sokol, 2002 ▸; Cheng et al., 2011 ▸; Ahn et al., 2011 ▸; Zebisch et al., 2016 ▸). DKK1, DKK2 and DKK4 function as canonical Wnt/β-catenin pathway inhibitors by binding to LRP5/6 and Kremen1/2 co-receptors, thereby promoting their internalization and suppressing Wnt ligand-induced signaling (Bafico et al., 2001 ▸; Mao et al., 2002 ▸; Zebisch et al., 2016 ▸).

CRD1 can influence the specificity and strength of DKK–LRP interactions indirectly by maintaining the proper structural conformation of the protein (Brott & Sokol, 2002 ▸; Niehrs, 2006 ▸; Ahn et al., 2011 ▸; Giralt et al., 2021 ▸). CRD2 is responsible for LRP6 binding and Wnt inhibition, whereas CRD1 modulates this interaction (Brott & Sokol, 2002 ▸). However, this functional role of CRD1 appears to be clearly established primarily for DKK1.

Although CRD2 is recognized as the principal LRP5/6-binding module, the structural and mechanistic basis by which CRD1 influences DKK function remains poorly understood. It remains unclear how conformational features of CRD1 contribute to receptor specificity or modulation of DKK–LRP interactions across family members. Compared with DKK1, structural and functional information on DKK4 is limited, despite its reported involvement in tissue-specific Wnt regulation and disease-associated processes. Previous solution NMR studies of DKK4-CRD1 provided an initial structural framework for the isolated domain and revealed regions exhibiting increased conformational variability (Patel et al., 2018 ▸). A high-resolution crystal structure of DKK4-CRD1 would therefore provide a reference for evaluating domain rigidity, conformational variability and the reliability of existing solution models. Such structural information would also facilitate comparison with computational predictions and support future studies aimed at identifying interaction partners of DKK4-CRD1.

In this study, the crystal structure of human DKK4-CRD1 was determined and compared with previously reported NMR models and an AlphaFold2 prediction to assess domain architecture and conformational properties.

2. Materials and methods

2.1. Macromolecule production

2.1.1. Cloning

The gene for human DKK4 was synthesized by Thermo Fisher Scientific as a DNA fragment with codon optimization for Escherichia coli expression (Table 1 ▸). The pET-47b vector (Novagen) was linearized and amplified by the inverse PCR method (Triglia et al., 1988 ▸; Ochman et al., 1988 ▸). The CRD1 region (Gly38–Thr97), containing an N-terminal 6×His-tag, of the gene, including a 17 bp overlap homologous to the ends of the linearized vector, was amplified by PCR. The insert and the linearized vector were assembled using In-Fusion Snap Assembly (Takara Bio) and transformed into Escherichia coli DynaCompetent Cells JetGiga DH5α (BioDynamics Laboratory). The cell suspension was plated on Luria–Bertani (LB) medium containing 100 µg ml⁻¹ ampicillin.

Table 1. Macromolecule-production information.

Source organism	Homo sapiens
DNA source	Synthetic gene (GeneArt Strings DNA Fragments, Thermo Fisher Scientific)
DNA sequence	ATGGCACATCACCACCACCATCACGGTAGCCAGTGTCTGAGCGATACCGATTGTAATACCCGTAAATTTTGTCTGCAGCCTCGTGATGAAAAACCGTTTTGTGCAACCTGTCGTGGTCTGCGTCGTCGTTGTCAGCGTGATGCAATGTGTTGTCCGGGTACACTGTGTGTTAATGATGTTTGTACCACCATGGAAGATGCAACCTAA
Forward primer (insert)†	5′-CATCACCACCACCATCACGGTAGCCAGTGTCTGAGCGATAC-3′
Reverse primer (insert)†	5′-TGTTTAGCAGCCTAGGTTTAGGTTGCATCTTCCATGGTGGTACAAACATCATTAACAC-3′
Forward primer (vector)†	5′-ACCTAGGCTGCTAAACAAAGC-3′
Reverse primer (vector)†	5′-GTGATGGTGGTGGTGATGTGCC-3′
Expression vector	pET-47b
Expression host	E. coli BL21-Gold (DE3)
Complete amino-acid sequence of the construct produced	MAHHHHHHGSQCLSDTDCNTRKFCLQPRDEKPFCATCRGLRRRCQRDAMCCPGTLCVNDVCTTMEDAT

^†The overlapped sequence to assemble the DNA fragments is underlined.

2.1.2. Protein expression

Protein expression for crystallization was performed by cultivation of E. coli BL21-Gold (DE3) cells (Agilent Technologies) carrying the expression plasmid in LB medium at 37°C. Induction of protein expression was initiated when the OD₆₀₀ reached 0.5 by adding 1 mM isopropyl β-d-1-thio­galactopyranoside. After 4 h of induction, the cells were harvested by centrifugation at 10 000g for 5 min and then resuspended in phosphate-buffered saline (PBS). The cells were pelleted again by centrifugation and frozen at −80°C until purification.

2.1.3. Protein purification

The frozen cells were thawed, resuspended in phosphate-buffered saline (PBS) supplemented with 0.1% (w/v) Triton X-100 and disrupted by sonication. The lysate was centrifuged at 30 240g for 30 min at 4°C to collect the inclusion bodies containing DKK4-CRD1. The precipitate was solubilized in buffer A (10 ml per gram of cell pellet) consisting of 50 mM potassium phosphate pH 8.0, 20 mM imidazole pH 8.0, 6 M guanidinium hydrochloride pH 8.0, 5 mM dithiothreitol (DTT) and incubated at room temperature for 2 h. The insoluble fraction was pelleted by centrifugation at 48 000g for 20 min at 20°C. The supernatant containing DKK4-CRD1 was loaded onto an Ni–NTA agarose column (1.5 × 3 cm) equilibrated with buffer A and eluted by increasing the concentration of imidazole to 250 mM. The DKK4-CRD1-containing fractions were mixed and diluted to 2 mg ml⁻¹ by adding buffer A. After DTT powder had been added to the sample to a final concentration of 40 mM, the sample was incubated at room temperature for 1 h. The purified inclusion bodies were refolded by dropwise dilution into a 70-fold volume of 50 mM potassium phosphate pH 8.0, 0.5 Ml-arginine pH 8.0, 2 mMdl-cystine at room temperature with continuous stirring. After stirring for 1 h at room temperature, the refolded DKK4-CRD1 was collected by adding 5 ml cOmplete His-tag purification resin (Roche). The refolded sample was concentrated and further purified using a HiLoad 26/600 Superdex 200 pg size-exclusion column (Cytiva) equilibrated with 25 mM HEPES pH 7.5, 200 mM NaCl. Fractions containing DKK4-CRD1 were pooled and concentrated using a Vivaspin 15R (5 kDa cutoff) centrifugal concentrator (Sartorius).

2.2. Apparent molecular-weight estimation

Protein standards (conalbumin, ovalbumin, carbonic anhydrase, ribonuclease and aprotinin) and Blue Dextran from the Gel Filtration LMW Calibration Kit (Cytiva) were dissolved in 20 mM HEPES–NaOH pH 7.5, 200 mM NaCl at a final concentration of 1.0 mg ml⁻¹ each. A 500 µl aliquot of each of the five protein standards, Blue Dextran 2000 (1 mg ml⁻¹) and DKK4-CRD1 was separately injected onto a Superdex 200 Increase 10/300 GL (Cytiva) column connected to an ÄKTApure 25 chromatography system (Cytiva) and eluted in 20 mM HEPES–NaOH pH 7.5, 200 mM NaCl. The partition coefficient (K_av) was calculated using the equation K_av = (V_e − V₀)/(V_c − V₀), where V_e is the elution volume, V₀ is the column void volume and V_c is the geometric column volume. Apparent molecular weight (M_r) was estimated using a calibration curve obtained by plotting K_av versus log(M_r) of the protein standards.

2.3. Crystallization

The purified DKK4-CRD1 was concentrated to 2.0 mg ml⁻¹ and then set up for crystallization in an MRC 96-Well 2-Drop Crystallization plate using the commercial crystallization screens Morpheus, BCS Screen Eco Screen, JCSG-plus Eco Screen, Proplex Eco Screen, Stura FootPrint Combination Eco Screen and SG1 Eco Screen (Molecular Dimensions). Crystals only appeared in SG1 Eco Screen tube 2-18 [0.2 M ammonium nitrate, 20%(w/v) PEG 3350] (Table 2 ▸). Three crystals were picked up and cryoprotected in mother liquor with an additional 10% glycerol and were flash-cooled in liquid nitrogen.

Table 2. Crystallization.

Method	Vapor diffusion, sitting drop
Plate type	96-well plate, MRC
Temperature (K)	277
Protein concentration (mg ml−1)	2.0
Buffer composition of protein solution	20 mM HEPES–NaOH pH 7.5, 0.2 M NaCl
Composition of reservoir solution	0.2 M ammonium nitrate, 20%(w/v) PEG 3350
Volume and ratio of drop	1.6 µl, 0.8:0.8 (protein:reservoir)
Volume of reservoir (µl)	30

2.4. Data collection and processing

Diffraction data were collected on the BL32XU beamline at SPring-8, Japan. Data collection and processing were performed automatically using the ZOO system (Hirata et al., 2019 ▸). The data sets from each crystal were indexed and integrated using XDS (Kabsch, 2010 ▸). The final data set was obtained by merging and scaling of multiple small wedges of data from the three crystals using KAMO (Yamashita et al., 2018 ▸). X-ray data-collection and processing statistics are summarized in Table 3 ▸.

Table 3. Data collection and processing.

Values in parentheses are for the outer shell.

Diffraction source	SPring-8 beamline BL32XU
Wavelength (Å)	1.0000
Temperature (K)	100
Detector	Dectris EIGER X 9M
Crystal-to-detector distance (mm)	210.6
Rotation range per image (°)	0.1
Exposure time per image (s)	0.020
Space group	P21
a, b, c (Å)	23.17, 67.55, 32.79
α, β, γ (°)	90, 91.65, 90
Resolution range (Å)	33.78–1.83 (1.94–1.83)
Total No. of reflections	40860 (5098)
No. of unique reflections	8880 (1421)
Completeness (%)	99.1 (98.6)
Multiplicity	4.60 (3.59)
〈I/σ(I)〉	5.77 (1.57)†
R meage	0.180 (0.583)
R meas	0.203 (0.682)
CC1/2	0.981 (0.710)†
Overall B factor from Wilson plot (Å2)	18.79

^†The I/σ(I) in the outer shell is 1.57, where CC_1/2 reaches 0.710. It falls below 2 in the second outermost shell at 1.94 Å resolution. Analyses of merged CC_1/2 correlations between intensity estimates from half-data sets (Karplus & Diederichs, 2015 ▸) were used to influence the high-resolution cutoff for data processing.

2.5. Structure solution and refinement

The first trial of molecular replacement (MR) using the solution NMR structure of DKK4-CRD1 (PDB entry 5o57; Patel et al., 2018 ▸) and Phaser (McCoy et al., 2007 ▸) was unsuccessful. The initial phases were obtained from the MR solution using the core CRD1 region (residues 39–94) of the AlphaFold2 (AF2) model (AF-Q9UBT3-F1) of DKK4 as a template. The MR solution was refined by iterative rounds of refinement in phenix.refine (Liebschner et al., 2019 ▸) and model building in Coot (Emsley & Cowtan, 2004 ▸). The resulting structure-refinement data are provided in Table 4 ▸. The atomic coordinates and structure factors were deposited in the Protein Data Bank with accession code 9x64. Root-mean-square deviation (r.m.s.d.) calculation and superimposition of the crystal, solution NMR and AF2 structures were carried out using SUPERPOSE (Krissinel & Henrick, 2004 ▸) in CCP4 (Agirre et al., 2023 ▸). Hydrogen-bond analysis was performed using USCF Chimera (Pettersen et al., 2004 ▸).

Table 4. Structure solution and refinement.

Values in parentheses are for the outer shell.

Resolution range (Å)	33.77–1.83 (1.94–1.83)
Completeness (%)	98.9 (97.3)
No. of reflections, working set	7957 (1303)
No. of reflections, test set	881 (140)
Final Rcryst	0.1851 (0.2456)
Final Rfree	0.2457 (0.2988)
No. of non-H atoms
Protein	886
Ion	9
Water	79
Total	974
R.m.s. deviations
Bond lengths (Å)	0.007
Angles (°)	0.773
Average B factors (Å2)
Protein	24.36
Ion	24.13
Water	27.17
Ramachandran plot
Most favored (%)	100.0
Allowed (%)	0.0
Outliers (%)	0.0

3. Results and discussion

3.1. Crystallization and structure determination

Crystals of DKK4-CRD1 were obtained using protein refolded from guanidinium-solubilized inclusion bodies. The crystals belonged to space group P2₁, with unit-cell parameters a = 23.17, b = 67.55, c = 32.79 Å, α = 90.00, β = 91.65, γ = 90.00°. Molecular replacement using the solution NMR structure (PDB entry 5o57) was unsuccessful despite 100% sequence identity. In contrast, the core CRD1 region (residues 39–94) of the AF2-predicted structure (AF-Q9UBT3-F1) yielded a clear solution with two molecules in the asymmetric unit. The structure was refined to a final resolution of 1.83 Å, with R_work and R_free values of 0.1851 and 0.2457, respectively (Table 4 ▸).

3.2. Overall structure

As expected, the crystal structure of DKK4-CRD1 closely resembles the AF2 model, which was used for the MR template, with r.m.s.d.s of 0.572 Å for chain A and 0.882 Å for chain B (Fig. 1 ▸). These values are same or smaller than the value (0.882 Å) between chain A and chain B of the crystal structure. The solution NMR structures also display a similar overall fold, but they show significantly higher r.m.s.d. values against the crystal structures [1.402–2.301 Å (average 1.734 Å) for chain A and 1.374–2.703 Å (average 1.896 Å) for chain B]. The crystal structure of DKK4-CRD1 comprises the N-subdomain (Gln40–Thr65) and C-subdomain (Arg71–Thr91) and a linker (Cys66–Arg70; N–C linker) connecting these subdomains. The N- and C-subdomains share a similar architecture containing a 3₁₀-helix [N, Ser43–Cys47 (3₁₀-N); C, Arg75–Cys79 (3₁₀-C)], an antiparallel β-sheet [N, Lys51–Leu54 and Phe62–Thr65 (β-N); C, Leu84–Val86 and Val89–Thr91 (β-C)] and two β-turns [N, Asn48–Lys51 (T1-N) and Pro56–Glu59 (T2-N); C, Cys80–Thr83 (T1-C) and Val86–Val89 (T2-C)], which is similar to the solution NMR structure except that the 3₁₀-C region displays a loop structure (Figs. 1 ▸ and 2 ▸). DKK4-CRD1 has five disulfide bonds, as previously found in the solution NMR structure (Patel et al., 2018 ▸; Fig. 1 ▸a). The first disulfide bond (Cys41–Cys53) tethers the N-terminal loop composed of Ser39–Leu42 to the first β-strand (Lys51–Leu54) of β-N. Cys47 forms the second disulfide bond with Cys63 of the second β-strand (Phe62–Thr65) of β-N. The third disulfide bond (Cys66–Cys80) tethers the N–C linker to the C-subdomain. The fourth (Cys73–Cys85) and fifth (Cys79–Cys90) disulfide bonds in the C-subdomain correspond to the first and second disulfide bonds in the N-subdomain, respectively.

Figure 1.

Overall X-ray structure of DKK4-CRD1. (a) Cartoon diagram (transparent sky blue) and C^α model (gray) with disulfide bonds (yellow ball-and-stick representation) of chain A. (b) Comparison of the C^α models of the AF2 model (light green ball-and-stick representation) with X-ray chain A (gray sticks) and chain B (hot pink sticks). (c) Comparison of the C^α models of representative solution NMR structures [#15 (best ANSURR score, gold ball-and-stick representation) and #26 (minimum r.m.s.d. against X-ray chain A, turquoise ball-and-stick representation)] with the AF2 model (light green sticks) and X-ray chain A (gray sticks) and chain B (hot pink sticks).

Figure 2.

Main-chain hydrogen bonds of DKK4-CRD1. (a) X-ray structure and (b) solution NMR structure #15 with the best ANSURR score. The hydrogen bonds conserved in the X-ray, AF2 and NMR structures are shown as broken lines colored depending on conservation level in all NMR models. Dotted lines indicate the hydrogen bonds conserved only in X-ray and AF2 [black in (a)] and NMR [dark blue in (b)]. Hydrogen bonds with a conservation level of ≤10% in the NMR ensembles are ignored in (b).

3.3. Intermolecular interactions in crystal packing and evaluation of the oligomeric state in solution

In the crystal, chain A interacts with chain B primarily through two major interfaces (Interface 1 and Interface 2; Supplementary Fig. S1). In both interfaces, residues from the C-subdomain of chain A predominantly participate in the interactions, whereas residues from both the N- and C-subdomains of chain B are involved. The buried surface areas of Interfaces 1 and 2 are 567 and 457 Ų, respectively. These values are close to the lower limit typically observed for specific protein–protein interactions (∼500 Ų; Chen et al., 2013 ▸). However, because the two interfaces exhibit distinct interaction modes, they are unlikely to represent biologically relevant, specific interactions such as the head-to-tail oligomeric or polymeric assemblies reported for other systems (Kim & Bowie, 2003 ▸; Schwarz-Romond et al., 2007 ▸; Bienz, 2014 ▸; Afroz et al., 2017 ▸; Fiedler et al., 2022 ▸). To further assess the oligomeric state of DKK4-CRD1 in solution, size-exclusion chromatography (SEC) was performed. DKK4-CRD1 eluted as a single peak at 17.892 ml (Supplementary Fig. S2). Based on a calibration curve generated using molecular-weight standards, the apparent molecular weight was estimated to be 9494 Da. Although this value is approximately 22% higher than the calculated molecular weight of the DKK4-CRD1 monomer (7751 Da), it is substantially lower than that expected for a dimer (15 502 Da), supporting a monomeric state in solution.

3.4. Evaluation of the solution NMR structure of DKK4-CRD1 based on the ANSURR score

Fowler et al. (2020 ▸) revealed that AF2 structures tend to be more accurate than NMR ensembles using ANSURR. ANSURR compares the Random Coil Index (RCI) calculated from backbone secondary chemical shifts and the local rigidity analyzed by a computational method, Floppy Inclusions and Rigid Substructure Topography (FIRST; Jacobs et al., 2001 ▸), calculated from NMR structures. The program provides two scores, correlation and RMSD, which indicate how well the local rigidities match. The former basically reflects the accuracy of the secondary structure and the latter shows the validity of the overall structural rigidity. To understand the reason for the difficulty in MR structure analysis using the NMR structure of DKK4-CRD1, the precomputed ANSURR scores obtained from the ANSURR web server (https://ansurr.com/) were assessed.

The correlation scores show a wide range from 13.3 to 87, whereas the RMSD scores (1.3–7.6) are in a narrow range and far worse. This seems to be common for solution NMR structures (Fowler et al., 2020 ▸, 2023 ▸; Fowler & Williamson, 2022 ▸). The authors suggested that a low RMSD score generally means far more flexibility than the rigidity indicated by backbone chemical shifts due to poor hydrogen bonds, especially in β-sheet regions (Fowler et al., 2020 ▸, 2023 ▸; Fowler & Williamson, 2022 ▸). Fowler and coworkers defined the ANSURR score as the sum of the correlation and RMSD scores of a structure (Fowler & Williamson, 2022 ▸). Therefore, ‘good’ models generally have both high correlation and RMSD scores. Supplementary Table S1 shows the ANSURR scores of the NMR ensembles of DKK4-CRD1. The ANSURR scores strongly correlate (r = 0.997) with the correlation scores, whereas they correlate very weakly with the RMSD scores (r = 0.360). Therefore, the correlation scores are a major determinant for the rank of ANSURR analysis. The rank of ANSURR scores shows no clear correlation with that of the r.m.s.d. values of the NMR ensembles against the X-ray and AF2 structures; the best and second-best ANSURR models are ranked from 9th to 55th in the r.m.s.d. values (Supplementary Table S1). The ANSURR score ranks of the NMR ensembles with the minimum r.m.s.d.s against X-ray chain A, chain B and AF2 are 18, 23 and 37, respectively.

3.5. Hydrogen-bond networks and structural rigidity

Supplementary Table S2 shows main-chain hydrogen bonds of X-ray, AF2 and the best, worst and minimum r.m.s.d. models of the NMR ensembles. The X-ray and AF2 structures contain significantly more hydrogen bonds than the NMR models listed in Supplementary Table S2, which is consistent with NMR ensembles with low RMSD scores tending to have fewer hydrogen bonds than predicted. The five main-chain hydrogen bonds only found in the X-ray and AF2 structures seem to contribute to stabilizing the overall fold by strengthening helix–loop, β-strand–loop and loop–loop interactions (Figs. 1 ▸c and 2 ▸, Supplementary Table S2). In particular, the N-terminal region is placed close to the second β-strand (Phe62–Thr65) of β-N by forming the Cys63 N–Ser39 O hydrogen bond in the X-ray and AF2 structures, whereas it is away from the N-subdomain in most of the NMR ensembles (Figs. 1 ▸a and 2 ▸). The Arg67 N–Met78 O hydrogen bond bridges 3₁₀-C and the N–C linker, which seems to stabilize the structure of the 3₁₀-C region. The Cys80 N–Arg67 O and Arg70 N–Cys90 O hydrogen bonds also contribute to the stability of the C-subdomain and the N–C linker by linking them.

Per-residue r.m.s.d.s between X-ray chains A and B are generally higher in the N-subdomain than in the C-subdomain (Supplementary Fig. S3). This tendency holds true for NMR ensembles, especially in the N-terminal, T1-N and T2-N regions. This is consistent with the fact that the T1-N region of X-ray chain A has double conformations (Figs. 1 ▸b and 2 ▸a) and that the T2-N region of X-ray chain B has the highest B factor (Supplementary Fig. S4). In addition, the T2-N (Pro56–Glu59) and T1-C (Cys80–Thr83) regions show slightly elevated r.m.s.d. values between the X-ray and AF2 structures. These regions form minor crystal contacts in addition to Interfaces 1 and 2 in the crystal lattice, and the resulting structural perturbations may account for the increased r.m.s.d. values.

The pLDDT (predicted local distance difference test) scores of the AF2 structure are very high (≥90) across most of DKK4-CRD1, but decrease below 90 in the N-terminal, T1-N (Asn48–Lys51), T2-N (Pro56–Glu59) and C-terminal regions (Supplementary Fig. S3). However, even in these regions the scores remain mostly above 85, with the lowest value observed for Met93 (81.87), indicating consistently high confidence across the domain. Notably, regions with reduced pLDDT scores correspond closely to peaks in the r.m.s.d. profile (Supplementary Fig. S3). This observation is consistent with previous reports that residues with lower pLDDT values often exhibit increased structural flexibility (Guo et al., 2022 ▸; Vander Meersche et al., 2025 ▸).

Conformational fluctuations are often crucial for protein–protein interaction, especially in signal transduction. The most favored conformation of the fluctuating protein is selected by the binding of a rigid binding partner. The basic idea of the conformational selection model is that protein structure of ligand-bound state pre-exists even in the absence of binding partners (Boehr et al., 2009 ▸; Csermely et al., 2010 ▸). Indeed, the residues (His204–Ser207) of DKK1-CRD2 corresponding to the T2-N region of DKK4-CRD1 bind to the third β-propeller domain of LRP6 (Ahn et al., 2011 ▸; Cheng et al., 2011 ▸; Zebisch et al., 2016 ▸; Matoba et al., 2017 ▸), although physiological receptors for DKK4-CRD1 have not been identified (Patel et al., 2018 ▸; Giralt et al., 2021 ▸; Lou et al., 2021 ▸; Kikuchi et al., 2022 ▸).

4. Conclusions

The crystal structure of human DKK4-CRD1 was determined at 1.83 Å resolution, providing a high-resolution view of its domain architecture and intramolecular stabilization. The structure reveals a compact fold composed of N- and C-subdomains connected by a linker region and reinforced by five conserved disulfide bonds. Comparison with solution NMR models and an AlphaFold2 prediction highlights pronounced differences in local rigidity, particularly within the N-subdomain, which displays increased conformational variability across structural methods. Analysis of hydrogen-bond networks and ANSURR scores suggests that the NMR ensembles underestimate structural rigidity in β-sheet regions relative to the crystallographic structure. These results provide a structural reference for DKK4-CRD1 and a framework for interpreting the conformational variability observed in solution. The present structure will facilitate future studies aimed at identifying interaction partners and elucidating the functional role of CRD1 in DKK-mediated Wnt signaling modulation.

Supplementary Material

PDB reference: DKK4 CRD1 domain, 9x64

Funding Statement

The following funding is acknowledged: Japan Society for the Promotion of Science (grant No. 21H02420; grant No. 24K09357) and Hyogo Science and Technology Association, Japan. This research was partially supported by Research Support Project for Life Science and Drug Discovery (Basis for Supporting Innovative Drug Discovery and Life Science Research; BINDS) from AMED under Grant No. JP25am0101070 (support No. S01898003).