The structure of FCGBP is formed as a disulfide‐mediated homodimer between its C‐terminal domains

Abstract

Mucus in the colon is crucial for intestinal homeostasis by forming a barrier that separates microbes from the epithelium. This is achieved by the structural arrangement of the major mucus proteins, such as MUC2 and FCGBP, both of which are comprised of several von Willebrand D domains (vWD) and assemblies. Numerous disulfide bonds stabilise these domains, and intermolecular bonds generate multimers of MUC2. The oligomeric nature of FCGBP is not known. Human hFCGBP contains 13 vWD domains whereas mouse mFCGBP consists of only 7. We found unpaired cysteines in the vWD1 (human and mouse) and vWD5 (mouse)/vWD11 (human) assemblies which were not involved in disulfide bonds. However, the most C‐terminal vWD domains, vWD7 (mouse)/vWD13 (human), formed disulfide‐linked dimers. The intermolecular bond between C₅₂₈₄ and C₅₄₀₃ of human hFCGBP was observed by using mass spectrometry to generate the dimer. Cryo‐EM structure analysis of recombinant mouse mFCGBP revealed a compact dimer with two symmetric intermolecular disulfide bonds between C₂₄₆₂ and C₂₅₈₁, corresponding to the dimerising cysteines in the human hFCGBP. This compact conformation involves interactions between the vWD assemblies, but although the domains involved at the interface are the same, the nature of the interactions differ. Mouse mFCGBP was also found to exist in a semi‐extended conformation. These different interactions offer insights into the dynamic nature of the FCGBP homodimer.

Intestinal mucus is produced by goblet cells and forms a protective layer on the epithelium. Mucus is composed of several different proteins, of which FCGBP is one of the more abundantly present. We now reveal the core structure of FCGBP showing a disulphide‐stabilised dimer involving its most C‐terminal von Willebrand D domain. This dimer is one of the building blocks in the organisation of the mucus layer.

Abbreviations

cryo‐EM

cryo‐electron microscopy

SAXS

small‐angle X‐ray scattering

TIL

trypsin inhibitor like

vWD

von Willebrand domain

vWF

von Willebrand factor

Introduction

The outer luminal barrier of the gastrointestinal tract is constituted of a mucus layer that under healthy conditions protects the epithelial surface form various hazards and is of particular importance in colon where it keeps the microbiota at a distance from the epithelium [1]. Defects in this protection are associated with bacterial penetration and colitis [1, 2, 3]. Proteomic analyses have revealed a set of core mucus proteins, such as MUC2, chloride channel accessory‐1 (CLCA1), zymogen granule protein‐16 (ZG16), anterior gradient protein‐2 (AGR2), trefoil factor‐3 (TFF3) and the IgG Fc‐binding protein (FCGBP) [4, 5, 6]. The MUC2 mucin builds the structural core of mucus with its highly O‐glycosylated long central domains contributing to the gel‐like properties of mucus. The N‐ and C‐terminal parts of MUC2 are cysteine‐rich with three von Willebrand D (vWD) assemblies in the N‐terminal part and one in its C‐terminal part. MUC2 is stabilised by disulfide bonds and forms oligomers by intermolecular disulfide bonds during biosynthesis. MUC2 forms covalent dimers between the C‐terminal domains followed by multimer formation with bonds between the N‐terminal domains [7, 8, 9, 10]. After its secretion, MUC2 is further stabilised by transglutaminase‐catalysed intermolecular isopeptide bonds, which is important for protection against proteolytic cleavage and the prevention of colitis [11]. FCGBP is one of the most abundant proteins in secreted mucus and has been suggested to be an IgG‐binding protein in human colonic tissue [12], but such function has been contradicted by more recent studies using purified proteins under native conditions, suggesting its name to be misleading [13]. It is also suggested to bind the mucus protein TFF3 [14] and to crosslink the main intestinal mucin MUC2 [4], but a quantitative approach revealed that the vast majority of secreted murine FCGBP is soluble in chaotropic buffers and does not form a covalent complex with MUC2 [13]. The structural organisation of FCGBP is characterised by repeated vWD domains. This structural arrangement is remarkable as other vWD domain‐containing proteins, such as the secreted mucins, zonadhesin, SCO‐spondin and otogelins combine their vWDs with other types of domains. Despite its secretion into mucus together with mucin, the vWD domains in FCGBP show the highest sequence homology to SCO‐spondin, zonadhesin and otogelin and not with the mucins or the von Willebrand factor (vWF) [15]. SCO‐spondin is a prime component of Reisner's fibre, and deletion of this protein in zebrafish abolished fibre formation and which is crucial for correct development of the body axis during embryogenesis [16]. This indicates that the structural arrangements of FCGBP in mucus could be involved in protein–protein interactions, which might be of importance for its protective function.

Human FCGBP (hFCGBP) consists of thirteen vWD domains, with the first twelve being complete assemblies containing consecutive vWD, C8, TIL and E domains (Fig. 1A). The murine orthologue has only seven vWD domains, where the first six are full vWD assemblies, the vWD3‐5 assemblies of the mouse FCGBP (mFCGBP) are repeated three times in the human sequence (vWD3‐5, vWD6‐8, vWD9‐11). Eleven of the vWD domains in hFCGBP are autocatalytically cleaved via GDPH (Gly‐Asp‐Pro‐His) motifs, whereas mFCGBP contains five cleaved vWD domains. Therefore, all vWD domains except the two most C‐terminal ones, are cleaved in both species. The vWD assemblies in FCGBP seem to differ compared to vWD assemblies in other proteins and the GDPH cleavage infers stability to the domains [17]. The cleaved vWD domains remain linked together by a single disulfide bond that tethers each consecutive fragment [13]. These cross‐links could potentially be used for controlled dissociation through changes in redox potential or physical forces through sheer stress. hFCGBP has an extra N‐terminal domain (FCGBP‐N1) which can be found in bacteria which utilise helical gliding motility, such as Myxococcus xhantus [18], but its role in FCGBP is still not clear. This domain is however not found in mFCGBP.

Fig. 1.

Domain organisation of FCGBP and SDS/PAGE migration suggesting a size larger than a monomer in cell lysates, mucus and when recombinantly expressed. (A) Domain organisation of the human and mouse FCGBP with the domains colour coded, GDPH cleavages marked and repeated domains explained. (B) Western blots using the antisera α‐FCGBPD13 showing nonreduced mouse (mus) and human (hum) FCGBP in cell lysates and mucus revealing high‐molecular mass complexes with sizes over 460 kDa for human hFCGBP and almost 460 kDa for mouse mFCGBP. Representative of n = 3 experiments. (C) Purified recombinant mouse mFCGBP separated by SDS/PAGE in nonreduced and reduced conditions stained with Coomassie. Nonreduced bands are just above 460 kDa while the reduced sample showed fragments as cleaved at the GD/PH sequence with sizes between 31 and 55 kDa, and the C terminus of around 117 kDa. Representative of n = 4 experiments. Gaps between lanes in B and C indicate gel images spliced together.

The apparent molecular mass of the protein observed by different methods under nonreducing conditions indicates that mFCGBP is larger than a monomer [13]. In mucins, a CTCK domain mediates covalent dimerisation through disulfide bonds [19, 20, 21] at the C terminus, but mFCGBP lacks this domain. N‐terminal multimerisation mediates elongation of the mucin and the vWF [8, 9, 22], but such interactions are not known for FCGBP. Furthermore, FCGBP obtained from mucosal tissue can be fragmented to its cleaved domains by reduction suggesting that it is not cross‐linked by any nonreducible bond.

Since interactions and structural arrangements are important for understanding the role of FCGBP in forming a protective mucus barrier, the aim of the study was to identify cysteines involved in the formation of intermolecular disulfide bonds. This was studied using recombinantly expressed domains of mFCGBP and by solving the core structure of recombinant mFCGBP to identify the core building blocks of FCGBP in mucus.

Results

Human and mouse FCGBP are larger than a predicted monomer in both tissue and mucus

The full‐length FCGBP, in both human and mouse, is fully cleaved at all its GDPH motifs, but its integrity is retained through disulfide bonds (Fig. 1A) [13]. The predicted molecular mass of FCGBP is 569 kDa for the human and 272 kDa for the mouse protein, excluding the signal peptide and N‐glycans. O‐glycosylation has not been studied for FCGBP, but as no clear PTS (Pro‐Thr‐Ser) regions can be found in the FCGBP sequence a high degree of O‐glycosylation, as found in the gel‐forming mucins, is unlikely. The migration of nonreduced human and murine FCGBP in a western blot of colonic cell lysates and mucus shows an interspecies size difference as expected with a larger protein present in humans (Fig. 1B). The molecular mass of the band for hFCGBP in both cell lysate and mucus is outside the range of the molecular mass marker (>460 kDa) and can thus not be estimated. A band between 460 kDa and 268 kDa is observed for mFCGBP in both cell lysate and mucus. mFCGBP has nine potential N‐glycosylation sites, which could result in an increased molecular mass up to approximately 290 kDa. However, the recombinantly expressed and purified mFCGBP migrated as a band around 460 kDa under nonreducing conditions, indicating a molecular mass larger than that predicted for the monomer (Fig. 1C). The shape of the protein influences the apparent molecular mass where cysteine bonds typically preserve a compact structure in proteins, causing them to migrate faster in SDS/PAGE when compared to its reduced state, making size estimates uncertain. Reduction of FCGBP results in its fragmentation due to the cleaved GDPH motifs, making it impossible to prove disulfide multimerisation simply by using SDS/PAGE and different redox conditions. The analysis of reduced fragments of recombinant mFCGBP resulted in products in the size range of 31–117 kDa, where the smaller fragments correspond to the polypeptides between two GDPH cleavages in two adjacent vWD assemblies (Fig. 1C). The two most C‐terminal vWD domains lacking GDPH cleavage sites correspond to the larger band with an estimated size of 117 kDa [13].

Several theoretically unpaired cysteines do not form intermolecular bonds

Based on SDS/PAGE analyses, the FCGBP protein appears to form larger molecules through disulfide‐bond mediated interactions. However, the specific cysteine residues forming these intermolecular bonds have not been identified. To identify potentially unpaired cysteine residues, we aligned the human and mouse FCGBP protein sequences. This revealed several differences including some in the position of cysteines. The first part of the human N terminus has long N1 domain, absent in mouse mFCGBP, which contains five such cysteines (Fig. 2A). Whether any of these cysteines were involved in N‐terminal dimerisation was investigated by analysing the recombinant hFCGBP N1‐N2 protein by SDS/PAGE under reducing and nonreducing conditions. A band of almost the same size, around 50 kDa, was observed under both conditions (Fig. 2B). This corresponds to the size of a monomer (51 kDa), which suggests that N‐terminal cysteines are not involved in the formation of disulfide‐bonded dimers.

Fig. 2.

Several cysteines are not involved in disulfide bonds. (A) The N‐terminal amino acid sequence of human hFCGBP (top, NP_003881.2) and mouse mFCGBP (bottom, NP_001116075.1) with the signal sequence (brown), the N1 region (black) (absent in mouse) and the N2 region (purple). The N1 region has five cysteines (yellow), whereas N2 has four cysteines. (B) Purified human hFCGBP‐N1N2 separated by SDS/PAGE under reducing and nonreducing conditions and stained by Coomassie revealed monomers under both conditions. Representative of n = 3 experiments. (C) Alignment of part of the human (NP_003881.2) and murine (NP_001116075.1) vWD1‐assembly by Clustal Omega (EMBL‐EBI) revealed an extra cysteine (red) in the TIL1 domain located after the C8 domain (pink) in mouse and an extra cysteine in the vWD1 (blue) in human. (D) Purified recombinant mouse mFCGBP N2‐vWD2 (left, representative of n = 2 experiments.) and human hFCGBP N1‐vWD2 (right, representative of n = 4 experiments) was analysed by SDS/PAGE and Coomassie stained under nonreducing and reducing conditions (indicated by DTT − or +). (E) Alignment of human (NP_003881.2) and mouse (NP_001116075.1) FCGBP vWD assemblies (end of the assembly sequence shown) by Clustal Omega (EMBL‐EBI) reveal elongation at the end of human assemblies 3, 6 and 9 and mouse assembly 3. An additional cysteine in the human vWD11 assembly and the corresponding mouse vWD5 assembly are marked (red). (F) SDS/PAGE of purified mouse mFCGBP vWD5‐D6 under reducing and nonreducing conditions reveal bands of similar sizes (around 75 kDa) corresponding to monomers. The corresponding human hFCGBP D11‐D12 was expressed in CHO‐K1 cells and analysed by western blot (α‐Flag) revealing a band in the size of the monomer (indicated by the arrow) with mock transfection as control. Both are representative of n = 2 experiments. Gaps between lanes in D and F indicate gel images spliced together.

All vWD domains are in assemblies that contain C8, TIL and E domains except for the last vWD domain, which is alone without the other associated domains. Aligning the FCGBP vWD domains revealed an extra cysteine in the TIL1 domain of the mFCGBP vWD1 assembly when compared to the human sequence (Fig. 2C). The human vWD1 assembly instead contains a cysteine in the end of the vWD1 domain that is not present in mouse (Fig. 2C). Intermolecular homodimerisation using these cysteines was investigated by SDS/PAGE of expressed recombinant truncated proteins. The mFCGBP N2‐D2 was observed as a 75 kDa band under nonreducing conditions, corresponding to a monomer with a predicted size of 88 kDa (Fig. 2D). The nonreduced hFCGBP N1‐D2 protein show a band at 100 kDa, most likely corresponding to a monomer as its predicted size is 132 kDa, the observed faster migration during SDS/PAGE is likely due to a smaller Stokes radius (Fig. 2D). The reduced proteins are dissociated into the GDPH cleavage products. The vWD1 and vWD2 as well as the N‐terminal domain of hFCGBP were observed as 37–50 kDa bands (Fig. 2D). The mouse N2 domain on the other hand, released by reduction and GDPH cleavage in WD1, was not observed due to its small molecular mass (6 kDa).

The alignment of all vWD assemblies from human and mouse FCGBP revealed an that the human vWD3, vWD6 and vWD9 and mouse vWD3 contained an additional sequence (approximately 25 amino acids, Fig. 2E). In addition, there is an extra cysteine in the human vWD11 and mouse vWD5 (Fig. 2E). These cysteines do also not form intermolecular bonds as the murine mFCGBP vWD5‐D6 protein as well as the human hFCGBP D11‐D12 migrated as predicted for monomers (around 80 kDa) on SDS/PAGE under nonreducing conditions (Fig. 2F). The reduced form migrates slightly slower matching the theoretical monomeric mass as a short sequence before the GDPH cleavage is lost.

Cysteines in vWD7 are involved in dimer formation of recombinant human and mouse FCGBP

The localisation of the disulfide bonds is known from the structure of the vWD1‐3 in the human vWF. These vWD domains were aligned with C‐terminal FCGBP vWD domains of the human (vWD13) and mouse (vWD7). This revealed that C₅₂₈₄ and C₅₄₀₃ of vWD13 (human) or C₂₄₆₂ and C₂₅₈₁ vWD7 (mouse) were lacking their expected interacting cysteines (Fig. 3A).

Fig. 3.

Cysteines in the C‐terminal vWD domain of FCGBP mediate intermolecular bonds. (A) Alignments of human vWD13 (NP_003881.2, aa. 5236‐5405) and mouse vWD7 (NP_001116075.1, aa. 2414‐2583) (FCGBP in red) with vWD1‐4 of the vWF (P04275.4, vWD1‐3 aa. 23‐1137) by Clustal Omega (EMBL‐EBI). The proposed disulfide‐bond configuration is indicated by yellow, orange, magenta and green lines (cysteines are underlined). Cysteines lacking conserved partners are highlighted by arrows. (B–D) Recombinant purified mouse mFCGBP vWD7 (B), human hFCGBP vWD13 (C) and human hFCGBP D12‐D13 (D) were analysed by SDS/PAGE and stained with Coomassie under reducing and nonreducing conditions. All these nonreduced recombinant proteins migrated as sizes corresponding to dimers. All images are representative of n = 3 experiments. Gaps between lanes in B indicate gel images spliced together. (E–H) SAXS results of purified recombinant mouse mFCGBP in (E) as scattering plot, in (F) as Guinier plot, in (G) as Kratky plot, and in (H) as its Pr function. The results show a well‐folded globular protein with a Dmax of 42 nm. The molecular mass estimation using the Bayesian inference approach in primus, suggest a Mr of 679–829 kDa, suggesting that the mouse mFCGBP is a dimer. (I) The SAXS envelope of the mouse mFCGBP suggests a dimer. Inserted is the structure revealed by Cryo‐EM. Scale bar is 5 nm.

To investigate dimer formation between the C‐terminal vWD domains, recombinant human vWD13 and mouse vWD7 were produced. Both domains migrated with an estimated size of almost 37 kDa under nonreducing conditions, that upon reduction were reduced to smaller than 20 kDa (Fig. 3B,C). This suggests that the vWD13 and vWD7 domains in human and mouse, respectively, form disulfide‐bonded dimers. This model is supported by analysis of recombinant hFCGBP‐vWD12‐vWD13 (predicted mass of around 65 kDa) showing a dimer >100 kDa under nonreducing conditions and a band smaller than 75 kDa after reduction (Fig. 3D). These results show that both human and mouse FCGBP can form dimers in their most C‐terminal vWD domain. This agrees with the observed sizes of the full‐length FCGBP molecules (Fig. 1).

To obtain a better understanding of the disulfide‐stabilised dimer, we performed small‐angle X‐ray scattering (SAXS) of recombinantly expressed full‐length mFCGBP. The SAXS results (Fig. 3E–H, Table 1) revealed a well‐folded structure and confirm that mFCGBP is a homodimer in solution, according to the Bayesian molecular weight estimation of approximately 715 kDa. The Kratky plot (Fig. 3G) shows a bi‐modal distribution, characteristic of multi‐domain proteins such as FCGBP. The pairwise distance distribution (p(r)) function (Fig. 3H), which describes the paired set of distance between all the electrons in the macromolecular structure, shows two maxima and a rather extended conformation with a Dmax of 42 nm. This type of shape suggests that mFCGBP is an elongated multi‐domain protein. The SAXS ab initio model indicated a dimer with a central core and outward‐facing extensions (Fig. 3I).

Table 1.

SAXS data collection, processing and modelling statistics of full‐length mouse Fcgbp.

Data collection parameters
Beamline	ESRF BM29, Grenoble, France
Beam geometry (μm2)	200 × 200
Wavelength (nm)	0.08266
Detector	Pilatus 2 M in vacuum
Detector distance (m)	2.867
q‐range (nm−1)	0.05–6.2
Sample environment	Quartz glass capillary, 1 mm diameter
Exposure time (s) per frame	1 s
Temperature (°C)	20
Concentration range measured (mg·mL−1)	0.12–1
Concentration used (mg·mL−1)	Merged dataset (0.56 mg·mL−1)
Structural parameters
Guinier analysis
I(0) (cm−1)	1934.49 ± 12.71
Rg (nm)	10.02 ± 0.23
q Rg‐range	0.49–1.3
P(r) analysis
I(0) (cm−1)	1990
Rg (nm)	10.77
q‐range (nm−1)	0.05–0.8
D max (nm)	42
Total quality estimate (GNOM)	0.75
Porod volume estimate, Vp (nm3)	1461.60
Molecular weight determination
Calculated monomeric MW from sequence, kDa	276
Volume of correlation (Vc)	2947
MW from Bayesian assessment (kDa) [23] [credibility interval], probability	714.8 [679.10–829.45], 90.41%
Software employed
Primary data reduction	BsxCuBE
Data processing	Primusqt [24] – ATSAS package version 3.0.0 [31]
Ab initio model	DAMMIN [25]
Model validation, averaging and final refinement	DAMAVER [26]
3D graphics representations	UCSF ChimeraX [27, 28]
SASBDB ID	SASDTV8

Mouse FCGBP forms covalent dimers with a compact core structure and a semi‐extended conformation

The mouse FCGBP was analysed by Cryo‐EM. This confirmed the presence of a dimeric interface in the vWD7 domain including two intermolecular disulfide bonds. The globular core formed a convoluted “eight shaped” dense structure composed of vWD4 to vWD7 (Fig. 4A). The three most N‐terminal vWD assemblies were not possible to observe, most likely due to flexibility in this part of the protein. This could be explained by the long linker region between the vWD3 and vWD4 assemblies. The main particles identified revealed a compact structure composed of two disulphide linked monomers of vWD4 to vWD7, with a global resolution of 3.6 Å displaying a C2 symmetry (Figs 4A, 5A–E, Table 2). The dimer is connected between the C‐terminal vWD7 domains by covalent intermolecular disulfide bonds between Cys₂₄₆₂‐Cys₂₅₈₁. The interface of this interaction spans an area of approximately 1350 Ų and contains eight hydrogen bonds (Table 3). The compact structure is further stabilised by vWD4‐vWD7′ interactions (Fig. 4A). This interface is formed between the C8 domain of the vWD4 assembly and amino acids of the D domain of the vWD7′ from the other vWD4′ to vWD7′ assembly, specifically those proximal to its calcium‐binding site (Fig. 4D). This interface is almost 290 Ų and includes one hydrogen bond (Table 3). The GDPH sites in both vWD4 and vWD5 are cleaved as previously noted [13, 17].

Fig. 4.

The Cryo‐EM structure of mouse mFCGBP reveals a compact and a semi‐extended dimer. (A) Analysis of results from Cryo‐EM of the recombinant mFCGBP shown as a 3D structure reveal an anti‐parallel dimer forming the dense ‘eight shaped’ core containing the vWDs 4‐vWD7′, at a resolution of 3.6 Å in a twofold symmetry. The main domain interfaces are between vWD7‐vWD7′ and vWD4‐vWD7′ of the anti‐parallel complex. The dimer is shown in two projections with a 90° angle and the vWD domains are indicated by numbers. The rest of the molecule (N2‐vWD3) are likely flexible and remain unresolved (indicated by black arrows). The structures are presented using UCSF Chimera. (B) Cryo‐EM of mFCGBP revealed a second semi‐extended conformation with an asymmetric dimer containing one monomer in a conformation similar to the compact form while the other only include the vWD7. The remaining part was not resolved in the Cryo‐EM maps likely due to flexibility (indicated by a blue arrowhead) as well as the N‐terminal domains (indicated by a black arrow). This structure with a global resolution of 3.9 Å is shown in two projections with a 90° angle and the vWD domains are indicated by numbers. The structures are presented using UCSF Chimera. (C) The vWD7‐vWD7 dimer is stabilised by two disulfide bonds between C₂₄₆₂ and C₂₅₈₁ of each monomer with Lys₂₄₇₁ involved in hydrogen bonds with different amino acids in the compact and semi‐extended conformations. The structures are presented using Pymol. (D) Comparison of the vWD4‐vWD7′ interaction between the compact and semi‐extended conformations reveal differences in the loop formed by Ser₂₅₆₄‐Ile₂₅₇₀. The structures are presented using UCSF Chimera.

Fig. 5.

Cryo‐EM density map of the compact mouse mFCGBP C‐terminal. (A) Map at 3.60 Å resolution coloured by local resolution. The structures are presented using UCSF Chimera. (B) Fourier correlations (FSC) for the final density map. (C) Per‐particle distribution over azimuth and elevation angles for the final density map. (D) Representative image of a particle. (E) Representative 2D class averages. Details of the experiment are presented in Table 2.

Table 2.

Cryo‐electron microscopy parameters for the analysis of mouse C‐terminal Fcgbp.

Structure	Fcgbp	Fcgbp
	Compact	Semi‐extended
Data accession
PDB	8RDE	8R0T
EMDB	EMD‐19070	EMD‐18803
Data collection
Microscope	TFS KRIOS	TFS KRIOS
Voltage (kV)	300	300
Detector	GATAN K3 detector	GATAN K3 detector
Pixel size (Å)	0.86	0.86
Electron exposure (e−/Å2)	40	40
Defocus range (μm)	−1.0 to −2.0	−1.0 to −2.0
Micrographs collected	6355	6355
Reconstruction
Software	cryosparc (v3.2)	cryosparc (v3.2)
Micrographs used	3055	3055
Particles used in refinement	184 175	167 045
Symmetry imposed	C2	C1
Overall resolution (Å) FSC = 0.143 (masked)	3.60	3.89
Map sharpening B‐factor (Å2)	81.6	102.4
Local resolution range (Å)	2.2–7.0	2.2–7.0
Model refinement
Software	phenix (v1.19.2‐4158)	phenix (v1.19.2‐4158)
Nonhydrogen atoms	20 050	11 461
Protein residues	2684	1534
Ligands	20	11
Average B factors (Å2)
Protein	40	40
Ligands	40	40
R.M.S. deviations
Bond length (Å)	0.006	0.006
Bond angle (°)	1.168	1.292
Ramachandran statistics (%)
Outliers	0.07	0.00
Allowed	6.29	7.6
Favoured	93.64	92.4
MolProbity score	2.58	3.03
Model vs. Map FSC
FSC = 0.5 (masked, Å)	4.4	4.5

Table 3.

Details of the interactions between vWD domains in the Fcgbp dimer.

Structure	Interface	Interface area (Å2)	Solvation ΔiG (kcal·mol−1)	Hydrogen bonds a	Salt bonds	Disulphide bonds	Peptidic bond
Compact	vWD7‐vWD7′	1348.2	−13.7	8	0	2	0
Compact	vWD4‐vWD7′	288.6	−3.1	1	0	0	0
Semi‐extended	vWD7‐vWD7′	1158.3	−16.4	6	0	2	0
Semi‐extended	vWD4‐vWD7′	539.6	−5.9	1	1	0	0
Compact	vWD4‐vWD5	741	−9.1	0	0	0	1
Compact	vWD5‐vWD6	918	0.1	3	1	0	1
Compact	vWD6‐vWD7	378	−4.8	1	1	0	1

^a Under 3 Å distance.

During data processing, another abundant particle type was observed and studied. This map was partly identical to the compact dimer with one compact mFCGBP monomer, while most of the other mFCGBP was not visible in the Cryo‐EM images. The refined map could fit the structure of the vWD4‐vWD7 of one monomer, but only the vWD7′ of the second mFCGBP with a global resolution of 3.9 Å (Figs 4B, 6A–E, Table 2). This indicates that the vWD1‐vWD6 of the second mFCGBP monomer adopts a different, likely more flexible conformation compared to the compact structure. This semi‐extended conformation of the dimer involves interacting interfaces between vWD7 and vWD7′, as well as between vWD4 and vWD7′, forming an asymmetric dimer (Fig. 4B).

Fig. 6.

Cryo‐EM density map of the semi‐extended mouse mFCGBP C‐terminal. (A) Map at 3.89 Å resolution coloured by local resolution. The structures are presented using UCSF Chimera. (B) Fourier correlations (FSC) for the final density map. (C) Per‐particle distribution over azimuth and elevation angles for the final density map. (D) Representative image of a particle. (E) Representative 2D class averages. Details of the experiment are presented in Table 2.

A comparison between the interacting interfaces of the compact and the semi‐extended conformation indicates that the two covalently interacting vWD7 domains to be fairly similar (Fig. 4C). The area of the interacting surface is similar, but the calculated solvation‐free energy change (ΔⁱG), and the number of hydrogen bonds differ (Table 3). The main differences between the density maps of the vWD7 domains of the two conformations are found in the interactions between the loop containing amino acids 2576 to 2581 forming multiple hydrogen bonds with Lys₂₄₇₁ of the other vWD7 domain in the compact conformation. However, the most pronounced differences between the two conformations are observed in the vWD4‐vWD7′ interface. This interface exhibits one hydrogen bond between Thr₁₄₆₅ and Asp′₂₅₅₆, but the semi‐extended vWD4‐vWD7′ interface has a larger interface area, lower solvation‐free energy change (ΔⁱG), and includes an additional salt bridge between Glu₁₄₅₃ and Lys′₂₅₆₇ compared to the compact conformation (Fig. 4D, Table 3). In the compact conformation of mFCGBP, the vWD7′ loop, which contains amino acids Ser₂₅₆₄ to Ile₂₅₇₀, is structured, while in the semi‐extended it is less structured and protrudes towards the C8 domain of the vWD4. This suggests that upon the initial extension of one monomer of the mFCGBP dimer the remaining vWD4‐vWD7′ interface acquires a more stable interaction.

The Cryo‐EM model was placed in the SAXS ab initio model showing that the model occupied the central core, and the N‐terminal parts not observed at Cryo‐EM extend out in two directions (Fig. 3I).

Mouse mFCGBP vWD assemblies in the compact interactions are different

The structural conformation of the compact mFCGBP is based on vWD assembly interactions, vWD4 to vWD5, vWD5 to vWD6 and vWD6 to vWD7 within the same monomer. The interaction interface between vWD4 and vWD5 assemblies aligns with that of vWD5 and vWD6 (Fig. 7A–C). This interface is characterised by the interaction of the vWD and C8 domains of the foremost assembly with the TIL domain of the next assembly and occasionally its E domain. Additional interactions can be found between the E domain and the following vWD domain, given their contiguous nature in the protein sequence (Fig. 7A,B). Despite the identical structural organisation of vWD4‐vWD5 and vWD5‐vWD6, it is noteworthy that the interfaces exhibit markedly different amino acid compositions and are driven by distinct interactions. The vWD4‐vWD5 interface is primarily driven by hydrophobic interactions, indicated by the low solvation‐free energy (ΔⁱG, Fig. 7A, Table 3). In contrast, the vWD5‐vWD6 interface is stabilised by three hydrogen bonds and one salt bridge (Fig. 7B, Table 3). For the vWD6‐vWD7 interaction, the pattern is disrupted by the truncated vWD7 (Fig. 7D). The vWD domain in vWD7 thus interacts with the E domain of the vWD6 assembly in a unique fashion, alongside its dimerising interactions with the adjacent monomer domain vWD7′ and with vWD4′.

Fig. 7.

The interacting interfaces between the vWD assemblies of the compact mFCGBP structure. (A) The vWD4‐vWD5 assembly interface of the compact conformation involves the D and C8 domains of the vWD4 assembly and TIL and E domains of the vWD5 assembly. The interface is mainly formed by interactions of hydrophobic residues. Calcium is green. (B) The vWD5‐vWD6 assembly interface of the compact conformation involves the D and C8 domains of the vWD5 assembly and TIL and E domains of the vWD6 assembly. The interface is mainly mediated by hydrogen bonds. (C) Superposition of the structures of the interacting vWD4‐vWD5 assemblies and vWD5‐VWD6 assemblies reveal a very similar structural organisation although the interface interactions differ. (D) The vWD6 assembly‐vWD7 domain interface of the compact conformation is different as the last domain is not a full assembly. This interface mainly involves the E domain of the vWD6 assembly and the D domain of the vWD7 and is based on a combination of hydrophobic interactions and a hydrogen bond and a salt bridge. The structures are presented using Pymol.

Structural modelling suggests dimerisation involving the predicted unpaired C‐terminal cysteines in human hFCGBP

To elucidate the cysteines of human hFCGFP involved in dimerisation, structural models of the human vWD13 domain were generated using AlphaFold2 and AlphaFold‐Multimer. In the generated models, the vWD13 was composed of a vWD domain followed by a C‐terminal disordered tail. The structure revealed that C₅₂₈₄ and C₅₄₀₃ are exposed and located too far away from each other to allow for the formation of an intramolecular disulfide bond. The dimer model instead predicts a symmetric intermolecular bond formed between the C₅₂₈₄ present in the β5‐β6 loop of the D13 domain from one monomer and the C₅₄₀₃ in the C‐terminal tail of the other (Fig. 8A). The interaction surface area spans approximately 1185 Ų, representing 12% of the total surface of a monomer. The interaction of the D13 domains through the same side of the β‐sandwiches creates a hydrophobic core that accommodates both C‐terminal tails. The interaction is mainly hydrophobic and includes symmetric interactions between 32 residues, 8 hydrogen bonds, one salt bridge and the described disulfide bond. FCGBP is thus likely to form disulfide‐linked dimers through its C‐terminal vWD domain. The predicted disulfide bridges connecting C₅₂₈₄ and C₅₄₀₃ of the AlphaFold 2 model were confirmed by mass spectrometry identifying the two connected peptides (Fig. 8B).

Fig. 8.

Cysteines in the C‐terminal vWD domain of human hFCGBP mediate intermolecular bonds. (A) A model of the potential FCGBPvWD13 dimer based on intermolecular bonds between C₅₂₈₄ and C₅₄₀₃ was generated by AlphaFold2 and AlphaFold‐Multimer. The structure is presented using Pymol. All cysteines are numbered in the model. The monomers are coloured brown and green, and the disulfide bonds are shown in yellow. (B) MS2 fragment spectrum of the parent ion [M + 3H]³⁺ at m/z 902.76. The results confirm that a disulfide bridge connects C₅₂₈₄‐C₅₄₀₃. The b ions are labelled in red, and y ions are labelled in blue.

Discussion

Intestinal mucus contributes to intestinal protection in many ways. To achieve this, several molecules with different functional properties contribute to the mucus organisation. How these molecules achieve this is not known today, but the emerging detailed structural information of its components will be very important for understanding their function and the organisation of secreted mucus. The most abundant mucus proteins, identified by proteomics, are MUC2, FCGBP and CLCA1 [4, 5, 6]. Both MUC2 and CLCA1 are suggested to form multimeric structures but nothing is known about the oligomeric nature of FCGBP.

The multimeric nature of the high‐molecular mass complexes of the full‐length human and murine FCGBP as previously observed by SDS/PAGE [13] are not possible to estimate by comparing reduced and nonreduced samples as FCGBP is fragmented upon reduction due to the GDPH cleavages [13]. In this study, we have compared the human and murine FCGBP sequence to each other in relation to vWF, an approach that revealed several differences in the number and positions of involved cysteine residues. Both FCGBP species have an unpaired cysteine in the vWD1 assembly but at different positions, namely the vWD1 domain in hFCGBP and in the TIL1 domain of mFCGBP. For the human variant, this could maybe be coupled to the uneven number of cysteines found in the N1 domain, not present in mFCGBP. However, dimers are not observed for the recombinant hFCGBP N1‐D2 or the murine mFCGBP N2‐D2. This suggests that the uneven number of cysteines in the human N1 domain as well as the unpaired cysteine in the vWD1 assembly from both species are not involved in the formation of intermolecular bonds. An extra cysteine was also observed in vWD5 and vWD11 assemblies of mouse and human FCGBP, respectively. However, once again these residues did not form any intermolecular bonds when these domains were analysed as recombinant proteins. The Cryo‐EM structure of mFCGBP confirmed that this vWD5‐free cysteine is not involved in forming a disulfide bond. As the human sequence contains a triplet of the central three vWD domains, it is interesting to note that this extra cysteine is only present in vWD11, close to the C‐terminal two vWD domains which lack the GDPH cleavage motif. This organisation could indicate a functional role for this amino acid, but this is not yet understood.

Two of the cysteines known to be involved in disulfide‐bond formation in the vWF were not present in the last vWD domains of either human or mouse FCGBP. Instead, two other cysteines lacking their interacting cysteine partners were now shown to stabilise a C‐terminal dimer by two intermolecular disulfide bonds. This dimerisation was shown by analysing truncated recombinant mouse FCGBP vWD7 and human FCGBP vWD13, both of which are held together as reducible dimers. The hFCGBP vWD12D13 forms a dimeric complex, whereas hFCGBP vWD11D12 exists only as a monomer. The C‐terminal dimer is thus dependent on the small vWD13 domain corresponding to vWD7 in mouse. The resolved mouse mFCGBP structure confirmed the dimer and intermolecular disulfide bonds between C₂₄₆₂ and C₂₅₈₁ of the vWD7 domains. Modelling of the vWD13 in hFCGBP indicated C₅₂₈₄ and C₅₄₀₃ that correspond to C₂₄₆₂ and C₂₅₈₁ of mouse to be the most likely candidates to form such intermolecular cross‐links. This assumption was then confirmed by mass spectrometry. The predicted dimeric interface of both mouse and human FCGBP involves an extended hydrophobic region along the external side of the β‐sheet 2. The hydrophobic core of the β‐sandwich is stabilised by the presence of numerous specific polar contacts, which make dimerisation likely to occur independent of the free cysteines. The intermolecular disulfide bonds would therefore be important to stabilise and lock the dimeric state. The dimer formation, likely to take place already in the endoplasmic reticulum, is thus similar in both mouse and human FCGBP.

At the neutral pH conditions used for Cryo‐EM, mFCGBP forms a compact dimeric structure mediated by the four C‐terminal vWD assemblies. However, the first three vWD assemblies were not observed due to flexibility. Although the structure of the vWDs are similar, the interaction interfaces between the vWD4‐vWD5 and vWD5‐vWD6 are drastically different suggesting specificity in different vWD assembly interactions. This indicates that the seemingly similar vWD assemblies have high specificity in their interactions that might be the basis for a predetermined organisation of interactions between not only FCGBP, but also other molecules with vWD assemblies such as the gel‐forming mucins.

Cryo‐EM analysis of mFCGBP also revealed a semi‐extended conformation where one monomer contained the compact shape with minor changes while the other monomer was not visible by Cryo‐EM likely due to increased flexibility. Although this conformation could be an artefact, the presence of these two distinct conformations may indicate that mFCGBP can transition between different conformations; at least between a compact and a semi‐extended form. This could suggest that FCGBP could act as a spring, potentially absorbing mechanical forces. The compact conformation would be energetically favourable as it contains all interactions between the two molecules. However, it is also possible that mFCGBP could adopt a fully extended conformation, but as this lacks both vWD4‐vWD7′ interactions it would eventually likely return to the more compact state where both monomers exhibit a stable vWD4‐vWD7′ interaction. The energy required to break the vWD4‐vWD7′ interface differs between the two conformations. In the compact conformation, this interaction is weaker compared to the semi‐extended conformation.

SAXS analysis of recombinantly expressed mouse mFCGBP gave a molecular weight between 679 and 829 kDa with a 90% confidence interval, indicating mFCGBP to form at least a dimer, as a monomer of mouse mFCGBP is approximately 276 kDa. mFCGBP has nine predicted N‐glycosylation sites per monomer adding around 3.8 kDa per site of the CHO cell‐expressed recombinant protein in addition to potential O‐glycans [29]. The addition of glycans considerably increases the molecular mass of the protein, in part explaining the higher molecular mass observed in the SAXS analysis. The SAXS envelope shape of the dimer suggest the localisation of the Cryo‐EM solved vWD4‐vWD7 dimer in the centre and that each N‐terminal extension comprising N2vWD1‐vWD3 point out in opposite directions (Fig. 3I). While we have only been able to address the molecular interactions within the central core and the C‐terminal part of FCGBP, it is important to consider potential interactions of the extending N‐terminal regions. This part may facilitate the binding and interaction of mFCGBP with other components of the mucus, including other mFCGBP dimers, contributing to the overall functionality of the molecule. The presence of an extra N1 domain in the hFCGBP with unknown function, not present in mouse, further highlights the importance of the N‐terminal part of FCGBP.

Despite that FCGBP is one of the main proteins in mucus, few studies have addressed its function. One reason for this is its name and assumption that it was binding immunoglobulins [30] and later MUC2 or TFF3 [4, 14]. Binding to immunoglobulins has not been possible to reproduce [13]. The idea of binding other proteins was driven by the observation of GDPH sequences that upon cleavage form reactive internal anhydrides. Structural studies have now shown that the potentially active Asp is buried inside the molecule and that the cleavage is important for stabilisation of the vWD assemblies [17]. FCGBP is stored together with the MUC2 mucin in the secretory granules of the goblet cell‐like LS174T cells [31, 32]. MUC2 and FCGBP were also suggested to interact noncovalently [32], something that is not unlikely considering the interaction possibilities by their common vWD assemblies. The strong association of mucin and FCGBP expression is best illustrated by the dramatic induction of FCGBP seen in the transition from healthy lungs to the induction of enhanced mucus production associated with lung diseases such as COPD [33]. The current structural understanding argues for FCGBP to have a role in the organisation and function of mucus and mucus layers in several organs.

In conclusion, our study has identified free cysteines with different positions in human and mouse FCGBP and shown that while several cysteine residues are present in FCGBP, only those in the last vWD domain are involved in covalent dimer formation. The vWD assemblies of mFCGBP can interact in different ways using the same interface which might infer functional specificity. The structure further indicates a dynamic protein which can accommodate different conformations. These results add to the previously known structures of mucus proteins, such as MUC2 [9, 10], involved in building the intestinal mucus, which are likely to influence its properties. The structural arrangement and rearrangements of mucus protein complexes depending on the environment could mediate the different properties of mucus. Our observations suggest FCGBP to be a structural mucus component involved in maintaining gut homeostasis, but its function is still largely unknown.

Materials and methods

Human samples

Sigmoid colon biopsies were obtained from patients referred for colonoscopy at Sahlgrenska University Hospital, Gothenburg, Sweden, in compliance with the Declaration of Helsinki and approved by the human research ethical committee in Gothenburg, Sweden (040‐08, 136‐12 and 2020‐03196) as previously described [13]. All included patients were adults (above age 18) and gave written informed consent. Only patients with normal intestinal macroscopy were included in this study.

Animals

C57BL/6N (WT) mice were bred and kept at the animal core facility at the University of Gothenburg with an environment of 21–22 °C, 12 h day/night cycles, and food and water ad libitum. Male mice, 10–12 weeks of age, were euthanized through sedation with isoflurane, followed by cervical dislocation and dissected to collect the samples. Animal experiments were approved by the Swedish Laboratory Animal Ethical Committee in Gothenburg (number: 73‐2015 and 2285‐2019).

Extraction of human and mouse colonic cell lysate and secreted mucus

Extraction of mucus was done as previously described [13]. For extraction of human colonic mucus, the tissue was flushed and stored in Krebs transport buffer before mounting in an explant chamber system [34]. The apical surface was submerged in 150 μL of Krebs‐mannitol buffer. After 1 h, the secreted mucus was collected by gentle scraping. For extraction of mouse mucus, the dissected and opened distal to mid‐colon was mounted with the epithelium facing upwards on a silicon‐coated plate. The tissue was overlaid with 150–200 μL of PBS and the mucus gently scraped off. All samples were stored at −20 °C.

To prepare lysates, human and mouse colonic biopsies were placed in 400 μL of lysis buffer (50 mm Tris/HCl pH 8, 150 mm NaCl, 1% Triton X‐100) with 1× complete EDTA‐free protease inhibitor cocktail (Roche, Merck, Darmstadt, Germany), and homogenised using an Ultra Turrax IKA (Werke, Staufen, Germany), followed by incubation at RT for 20 min. Samples were centrifuged at 9000  g for 15 min at 4 °C. All samples were stored at −20 °C.

hFCGBP/mFCGBP expression vectors

Expression vectors for different domains of human FCGBP were previously generated [13]. These include FCGBP N1‐D2 (NP_003881.2, amino acids 24‐1250) inserted into the pSecTag(+)‐Myc‐His C vector (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), the FCGBP N1‐N2 (NP_003881.2, amino acids 1‐470) into the pcDNA3.1(+)‐Myc‐His B vector (Invitrogen, Thermo Fisher Scientific) and the FCGBP vWD13 (NP_003881.2, amino acids 5225‐5405) into the pSecTag(+)‐Myc‐His A vector (Invitrogen, Thermo Fisher Scientific). In this work, the FCGBP vWD12‐D13 sequence (NP_003881.2, amino acids 4823‐5405) was inserted in pSec‐Tag(+)‐Myc‐His A vector, and the FCGBP vWD11‐D12 (8xHis‐DDK‐FCGBP, NP_003881.2 amino acids 4467‐5234) vector was generated using the pcDNA3.1 vector by GenScript (Piscataway, NJ, USA).

For the mouse, the vector pCMV‐Fcgbp (amino acids 1‐2583, NP_001116075.1)‐Myc‐DDK (NM_001122603, Origene, Rockville, MD, USA) was used to express full‐length mouse mFCGBP protein as previously described [13] and as a template for the domain constructs. Fcgbp N2‐D2 (8xHis‐DDK‐Fcgbp, NP_001116075.1 amino acids 27‐835), Fcgbp vWD5‐D6 (8His‐DDK‐Fcgbp, NP_001116075.1 amino acids 1645‐2412) and Fcgbp vWD7 (8xHis‐DDK‐Fcgbp, NP_001116075.1 amino acids 2407‐2583) expression vectors were all generated in the pcDNA3.1 vector by GenScript.

Recombinant FCGBP protein expression and purification

Plasmids were transformed into competent E. coli XL1‐Blue (Agilent, Santa Clara, CA, USA), and DNA was purified using Qiagen plasmid Mini or Maxi kits (Qiagen, Hilden, Germany) according to manufacturer's instructions.

The truncated proteins were transiently expressed by FectoPRO (Polyplus, Illkirch, France) or NovaCHOice (Merck Millipore, Darmstadt, Germany) transfections of in suspension‐growing CHO‐S cells and purified from cell medium using Hi‐trap Chelating FF or HisTrap Excel affinity columns (GE Healthcare, Chicago. IL, USA) as previously described [13]. Transient CHO‐K1 expression of hFCGBP vWD11‐D12 was performed by transfection using Lipofectamine 2000 (Thermo Fisher Scientific) in IMDM medium (Gibco). Cells were grown in IMDM (Gibco) with 10% fetal bovine serum (Gibco) with G418 (Gibco) to generate a stably expression culture as described previously [13].

Full‐length recombinant mouse mFCGBP was expressed transiently in CHO‐S or stably in CHO‐K1 and purified by affinity chromatography using an EZview Red ANTI‐FLAG M2 affinity gel (Sigma‐Aldrich, Merck, Darmstadt, Germany). In addition, a final size exclusion or ion exchange purification using a Superose 6 Increase 10/300 or a MONO Q 5/50 column (GE Healthcare) was performed. The eluate was concentrated, buffer exchanged to 25 mm HEPES pH 7.4 with 100–150 mm NaCl and stored at −80 °C [13].

Electrophoresis and western blot

Nonreduced mucus, lysate and recombinant proteins were run on Mini Protean 4–16 or 4–20% TGX PAGE (Bio‐Rad, Hercules, CA, USA) or 10% SDS/PAGE gels using Laemmli buffers at 50–120 V. Gels were stained overnight at room temperature with Imperial Protein stain (Thermo Fisher Scientific) and imaged using a Licor Odyssey CLX imager (Li‐Cor Biotechnology, Lincoln, NE, USA) or used for western blot. For western blot, Immobilon‐P PVDF or PVDF‐FL membranes (Merck Millipore, Darmstadt, Germany) were used, and semi‐dry blotting was performed with 5–10% methanol in 48 mm Tris/HCl, 21 mm glycine and 1.3 mm SDS, running at 2.5 mA·cm⁻² current for 60–80 min. Blots were blocked in PBS with 0.01% Tween 20 (PBS‐T) and 5% milk. αFCGBP‐D13 and αFCGBP‐N antisera [13] were diluted 1:2000–5000 and incubated overnight at 4 °C in block solution. Commercially available mouse α‐Flag 9291 (1:2000) antibodies were used to stain tagged recombinant proteins. Secondary antibodies, goat anti‐rabbit or anti‐mouse IgG alkaline phosphatase (Southern Biotech, Birmingham, AL, USA) at 1:1000–2000 dilution were incubated at RT for 1 h and developed by bromo‐4‐chloro‐3‐indolyl phosphate/nitroblue tetrazolium (Southern Biotech).

For fluorescently developed blots, the αFCGBP‐N and αFCGBP‐D13 was diluted 1:5000 in PBS with 5% bovine serum albumin (BSA), or PBS with 5% BSA and 0.02% SDS, respectively, and incubated overnight at 4 °C. The secondary antibody goat anti‐rabbit IgG alexa fluor 790, diluted 1:20 000 (Thermo Fisher Scientific) were incubated for 60 min at room temperature in PBS‐T with 5% BSA and 0.02% SDS and developed using a Licor Odyssey CLX imager (Li‐Cor Biotechnology).

Structure predictions

In silico structure modelling of human vWD13 (amino acids 5235‐5405, NP_003881.2) was performed using alphafold v2.3.0 [35] and AlphaFold‐Multimer [36].

SAXS of recombinant mouse mFCGBP

Small‐angle X‐ray scattering (SAXS) measurements were carried out at 20 °C with a momentum transfer range from 0.05 to 6.2 nm⁻¹, at BioSAXS beamline BM29 at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. Serial dilution of full‐length mouse mFCGBP were done using the gel filtration buffer (25 mm Hepes pH 7.4100 mm NaCl 10 mm CaCl₂), and the different concentrations were measured in batch mode according to the parameters described in Table 1. Ten frames were collected for each dilution and for the buffer before and after the sample with 1 s/frame exposure time using an X‐ray wavelength of 0.08 nm in flow mode. Data reduction and 1D scattering intensities of the samples and buffers were done at the beamline BM29. Frames averaged, buffer subtraction and extrapolation were performed using primusqt from atsas 3.0.0 package [37] Although no measurable radiation damage or intermolecular interaction/repulsion were observed in the different concentrations, the scattering curve used to estimate the structural parameters (Table 1) was generated by merging the higher concentration dataset (1 mg·mL⁻¹) at high angles and the lower concentration (0.12 mg·mL⁻¹) dataset at low angles (scattering curve). The radius of gyration (Rg), forward scattering (I(0)), maximum particle dimension (D _max) and the distance distribution function were determined with GNOM [38]. The molecular weight was obtained using the Bayesian inference approach at atsas package [39].

Ab initio models were obtained by dummy atom modelling using DAMMIN [25]. The generated models were aligned, compared, averaged and filtered using DAMAVER [26], and the final envelop was representing using UCSF ChimeraX [27, 28]. The SAXS results are deposited to SASBDB with accession number SASDTV8.

Cryo‐EM grid preparation and data acquisition of recombinantly expressed mouse mFCGBP

The Cryo‐EM grid preparation was largely performed as previously described [10]. The recombinant full‐length mouse mFCGBP protein at a concentration of 1 mg·mL⁻¹ was incubated 2 h at 4 °C in a buffer solution containing 150 mm NaCl 25 mm HEPES and 5 mm CaCl₂ at pH 7.4. A 2 μL solution of this sample was loaded onto a Quantifoil UltAuFoil R1.2/1.3, 300 mesh gold grid that was previously glow discharged with 15 mA during 40 s with negative charge. The grid was plunge frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot plunger (Thermo Fisher Scientific) at 100% humidity and 4 °C temperature. Cryo‐EM data were collected on a Titan Krios G2 transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV. Movies of 80 frames each were recorded on a Gatan K3 BioQuantum camera (Gatan, Pleasanton, CA, USA). Movies were recorded in counting mode at a nominal magnification of 105 000 × zoom, corresponding to a physical pixel size of 0.86 Å. The total dose rate was set to 80.15 e‐/Ų. Nominal defocus range was −1 to −2 μm. EPU software (Thermo Fisher Scientific) in Aberration‐free image shift (AFIS) mode was used for automated data collection, in which two images were collected from each hole.

Cryo‐EM data analysis

For model building, refinement and analysis, Cryo‐EM movies data processing and image fitting was done using cryoSPARC software v3.2 [40]. Patch motion correction and patch CTF estimation was done over the total of 6355 movies collected. An initial particle picking was done using the Blob Picker followed by particle extraction, 2D classification and 3D heterogeneous refinement. This allowed for creating two different initial volume structures, from which 50 equally spaced re‐projections were created to perform a second picking using the Template picking function for each conformation identified. A total of 1.702.602 particles were extracted, 2D classified and filtered by class. Selected 2D classes were used for a 3D classification and discarding process that included several cycles of ab initio reconstructions for 2 3D‐classes, followed by heterogeneous refinement. For each cycle, the particles that gave the best volume were selected for another cycle of 3D selection until the volumes did not improve for each conformation. The finally selected particles were used for a deep picking process through Topaz wrapper in cryoSPARC [41]. All the selected particles were subjected to the same particle selection process where 184.178 particles were used for the compact mFCGBP structure, and 167.045 particles were used for the semi‐extended mFCGBP conformation. The particles and the corresponding 3D volume were refined using the 3D refinement tools and CTF refining tools, applying a C2 symmetry, with a final local refinement of the compact mFCGBP structure.

By manually fitting a monomer of the MUC2 model D3 dimer (PDB ID 6RBF) in each corresponding volume of the convoluted map, an initial backbone structure for the mouse mFCGBP domain modules D4, D5 and D6 of the mFCGBP dimer was obtained using UCSF Chimera. For the vWD7 domain backbone module, a truncated version of the MUC2 monomer was used by deleting the C8, TIL and E domains. The initial model was finally manually built in Coot [42], aided by the autobuild tool of Phenix [43] using the E9Q9C6 UniProt sequence. The final model was refined by cycles of real‐space refinement using Phenix and manual rebuilding and refining in Coot. The final structure refinement validation for each conformation was assessed by Molprobity [44]. The structure analysis and figures were generated using Pymol and UCSF Chimera. Detailed information in Table 2.

Mass spectrometry

The Coomassie‐stained band of the hFCGBP vWD13 dimer was excised and washed in acetonitrile (ACN) with an overnight digestion with 10 ng·μL⁻¹ trypsin at 37 °C (Promega) without reduction and alkylation as disulfide‐linked peptides were targets. The peptides were extracted in 66% ACN with 0.2% formic acid (FA), and the organic solvents were removed using speedvac. Samples were acidified in 5% acetic acid before C18 stage‐tip purification. Eluted samples were dried and resuspended in 0.2% FA prior to analysis in nanoLC‐MS–MS using a Q Exactive HF mass spectrometer (Thermo Fisher Scientific) coupled to an EASY‐nLC system (Thermo Fisher Scientific) with a 150 × 0.0075 mm column (New Objective, Woburn, MA, USA) loaded with Reprosil‐Pur C18‐AQ 3 μm particles (Dr. Maisch, Ammerbuch, Germany). MS runs were performed as previously described [13]. The stavrox software (version 3.6.6.) [45] was used to search for disulfide‐crosslinked peptides in the vWD13 (α‐subunit and β‐subunit) using the following conditions (i) tolerance of the parent ion 2 ppm; (ii) tolerance of fragment ions 20 ppm; (iii) ‐2H as composition of the cross‐linker between the corresponding cysteine residues; and (iv) full specificity for trypsin with one missed cleavage.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

EE, PG, STM, MJGB and CR performed the experiments and analysed the data. EE and MJGB purified the protein and performed the SDS/PAGE experiments, PG performed the Cryo‐EM studies, STM performed the AlphaFold2.3 prediction, MJGB performed the SAXS study and CR performed the MS experiments. MEVJ and GCH supervised the work and provided funding. MEVJ wrote the paper. All authors contributed to revision of the manuscript and approved the submitted version.

Peer review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/febs.17383.

Data availability statement

The data presented in the article are shown as alignments of available sequences, as full lane images with full gel images available upon request. The structural models are based on AlphaFold predictions of available sequences with annotations reported, and the full spectra for the MS analysis are presented. The SAXS data are deposited to the Small Angle Scattering Biological Data Bank, SASBDB, with ID SASDTV8, (https://www.sasbdb.org/data/SASDTV8/qro5nkcf1m). The structural data from Cryo‐EM are deposited to the PDB database with entry 8RDE, EMD‐19070 for the compact form and 8R0T, EMD‐18803 for the semi‐extended mFCGBP.