Development of a calreticulin-targeting glycan ligand based on a hybrid binding concept

Abstract

Calreticulin (CRT), a chaperone that possesses both lectin and chaperone domains, is localized in the endoplasmic reticulum (ER). CRT has diverse functions and localizations; thus, CRT is a multifunctional protein. Particularly in the ER, CRT mainly aids in the proper folding of nascent glycoproteins as lectin chaperones. Approximately one-third of cellular proteins, including disease-related proteins, are synthesized in the ER. The lectin chaperones CRT and calnexin facilitate the correct folding of these glycoproteins; hence, these chaperones are essential for cells. Various CRT ligands have been reported, mainly composed of Glc₁Man₉GlcNAc₂-type glycan. However, it remains problematic for the complicated synthesis and preparation, and it interacts with glycoprotein folding-related proteins in the ER other than CRT. This suggests that the development of CRT ligands still can be improved. In this study, we developed a hybrid binding concept, which encompasses concurrent binding of ligands to CRT lectin and chaperone domains. We synthesized a CRT-targeting glycan ligand with a glycan and hydrophobic aglycone for CRT lectin and chaperone domain binding, respectively. The thermal shift assay with the CRT-targeting glycan demonstrated that binding was enhanced by simultaneous glycan and hydrophobic aglycone binding. The affinity of the CRT-targeting ligand showed isothermal titration calorimetry approximately 50-fold stronger than that of the glycan alone, thereby supporting the hybrid binding concept. In addition, the CRT-targeting ligand inhibited chaperone function. Overall, these results indicate that the hybrid binding concept may be useful as a novel strategy for the development of CRT ligands and inhibitors.

calreticulin, chaperone inhibitor, glycan ligand, hybrid binding

Introduction

Calreticulin (CRT) is a soluble endoplasmic reticulum (ER)-resident lectin chaperone (Michalak et al. 1992; Coppolino and Dedhar 1998; Krause et al. 1997; Michalak et al. 1999) that possesses a globular N-domain (monoglucosylated glycan-binding lectin domain), a flexible P-domain (chaperone domain) involved in client protein binding and chaperone function, and a C-domain (Ca²⁺-binding domain). The chaperones main function is to facilitate glycoprotein folding and to maintain Ca²⁺ homeostasis in the ER (Michalak et al. 2009). Specifically, CRT and calnexin (CNX), a membrane-bound lectin chaperone (Ware et al. 1995), aid the folding of nascent glycoproteins, which serves as a central glycoprotein folding mechanism in the ER. Considering that approximately one-third of all proteins are synthesized in the ER (Juszkiewicz and Hegde 2018; Wiseman et al. 2022), CRT-facilitated glycoprotein folding is an important process. In addition, CRT localizes to areas other than the ER, such as the cell surface and extracellular matrix, suggesting multifunctional roles in cells (Johnson et al. 2001). The accumulation of misfolded (glyco)proteins induces ER stress and has been implicated in the pathogenesis of neurodegenerative disorders and diabetes mellitus (Wang and Kaufman 2016). Given the importance of CRT in cells, CRT ligands and inhibitors have been reported (Kapoor et al. 2003; Ito et al. 2004; Arai et al. 2005; Gopalakrishnapai et al. 2006), and monoglucosylated glycan (Glc₁Man₉GlcNAc₂)-type ligands are the most potent thus far. Considering that the glycoform is a known ligand and/or substrate for other glycoprotein folding-related proteins such as CNX, glucosidase II and mannosidase in the ER (Caramelo and Parodi 2015), the development of a CRT-targeting ligand other than the Glc₁Man₉GlcNAc₂-type is needed. In this study, we developed a novel CRT-targeting glycan ligand. We designed and synthesized the ligand with a glycan and a hydrophobic aglycone to bind to the CRT lectin and chaperone domains, respectively. We called this the hybrid binding concept. Using the CRT-targeting glycan ligand, we analyzed its binding ability and potential for the inhibition of chaperone function.

Results

Design and synthesis of CRT-targeting glycan ligand

To design a CRT-targeting glycan ligand, we focused on the binding of CRT to client proteins. CRT binds to both non-glycosylated and glycosylated proteins (Svaerke and Houen 1998; Wijeyesakere et al. 2013), suggesting that the molecule can bind to its client proteins in both a glycan-dependent and -independent manner. These results imply that glycans provide selectivity for CRT, because other chaperones in the ER, excluding CNX, do not have a lectin domain. During the early and intermediate folding steps of glycoprotein biosynthesis, the hydrophobic peptide sequences of glycoproteins are protected by CRT and CNX. Totani et al. reported a folding sensor enzyme; UDP-glucose:glycoprotein glucosyltransferase, which senses glycoprotein folding states, showed high glucose transfer activity for chemically synthesized Man₉GlcNAc₂-type glycan with Fmoc (Totani et al. 2009). This suggests that the hydrophobic surface of glycoprotein folding steps can be mimicked by Fmoc. Therefore, we selected the Fmoc group of a hydrophobic aglycone. The Fmoc group is also used as a general label for oligomannose-type glycans derived from natural sources (Makimura et al. 2012; Wang et al. 2015; Hirose et al. 2024). In addition, synthetic monoglucosylated oligomannose-type glycans (Glc₁Man₉GlcNAc₂) with several types of hydrophobic aglycones strongly bind to CRT in vitro (Hirano et al. 2015). Therefore, we hypothesized that the hybrid binding concept, that is simultaneous glycan and hydrophobic aglycone binding, is a specific feature of CRT. This indicates that Glc₁Man₉GlcNAc₂ with a hydrophobic aglycone is a suitable backbone for the development of CRT-targeting glycan ligands. However, the synthesis of Glc₁Man₉GlcNAc₂ is difficult (Matsuo et al. 2006) and several glycoprotein-related proteins in the ER bind to and/or act on the glycan structure (Caramelo and Parodi 2015). Considering these limitations, we focused on determining the minimal glycan structure required for CRT. Although CRT binds to disaccharides (Glc-α1,3-Man) (Spiro et al. 1996; Kapoor et al. 2003), Williams et al. demonstrated greater binding for trisaccharides than that for disaccharides (Vassilakos et al. 1998). Based on these findings, we selected the trisaccharide Glc-α1,3-Man-α1,2-Man (GMM) as the glycan moiety. The glycan and hydrophobic aglycone (Fmoc) were connected to an octa ethylene glycol (EG8) linker, which was estimated to have almost the same overall length as the effective CRT ligand Glc₁Man₉GlcNAc₂-Gly-Fmoc. The synthetic pathway and structure of the target glycan ligand (GMM-EG8-Fmoc) are shown in Scheme 1.

Scheme 1

Synthetic pathway and structure of calreticulin (CRT)-targeting glycan ligand. a) NaOMe, tetrahydrofuran (THF)/MeOH = 1:1, v/v. b) Tetra-n-butylammonium fluoride (TBAF), THF, 40 °C. c) Cerium ammonium nitrate (CAN), MeCN/H₂O = 3:1, v/v. d) Pd(OH)₂/C, H₂, dimethylformamide (DMF). GMM, Glc-α1,3-Man-α1,2-Man; EG8, octa ethylene glycol; Fmoc, hydrophobic aglycone. PFP = pentafluoropropionyl.

First, we synthesized a set of acceptors and donors for the target glycan moiety. The detailed synthesis pathway for each monosaccharide unit is shown in Scheme S1. Monosaccharide mannosyl acceptors 1 and 2 were synthesized from common starting material S1. Monosaccharide glucosyl donor 3 was prepared from commercially available per-acetylated β-D-glucopyranoside S9. The glycosylation of mannosyl acceptor 1 with donor 2 produced disaccharide 4 (Scheme 1, upper pathway). Owing to the pentafluoropropionyl group (Takatani et al. 2003) at the C-3 position of 2, the glycosylation followed by quenching with Et₃N produced the disaccharide acceptor 5 in a single pot, achieving a 61% isolated yield with α-selectivity over two steps. The disaccharide acceptor 5 was glycosylated with glucosyl donor 3 to afford 47% product with perfect α-selective product 6. The 1,2-cis-α-glucosylation is a difficult glycosylation reaction to perform to construct α-selective product (Nigudkar and Demchenko 2015). Moreover, construction of a 1,2-cis-α-glucosyl bond was required for our trisaccharide synthesis. Therefore, to obtain the target glycan backbone, we applied our previous developed glycosylation method, which is a silyl-assisted 1,2-cis-α-glucosylation reaction (Totani et al. 2015; Kawanobe et al. 2022). The method enabled us to produce perfect α-selective trisaccharides. After the four-step deprotection reaction, we obtained the deprotected target glycan 10 as GMM. Next, we derivatized GMM to the target glycan ligand. First, amination of the reducing end in aqueous saturated NH₄HCO₃ was performed, followed by a reaction with Fmoc-Gly-Cl. This resulted in 53% yield of glycan-glycine-Fmoc 12 and 47% recovery of GMM (Scheme 1, middle pathway). Fmoc deprotection was conducted with piperidine to obtain 13 and condensation with COOH-EG8-Fmoc resulted in 14 as GMM-EG8-Fmoc at 72% yield (Scheme 1, lower pathway), thereby achieving the synthesis of the target ligand GMM-EG8-Fmoc.

Binding assay of CRT-targeting glycan ligand

To confirm the binding of GMM-EG8-Fmoc to CRT, we performed a thermal shift assay (Pantoliano et al. 2001) with recombinant human CRT (hCRT) according to our previously reported method (Hirano et al. 2015). We expressed hCRT and confirmed its purity using Coomassie brilliant blue staining (Fig. S1). We examined GMM, maltotriose, COOH-EG8-Fmoc, or GMM-EG8-Fmoc as ligands for hCRT (Fig. 1). Maltotriose was used as a negative control. The melting peaks indicated the denaturing temperature (Td) of hCRT in the absence or presence of ligands (Fig. 1a). In the absence of ligand, the Td of the hCRT was 42.5 °C (Fig. 1, green line). In the presence of GMM or COOH-EG8-Fmoc, no statistically significant shift of the Td was observed (43.0 °C for GMM, 42.8 °C for maltotriose, 42.9 °C for COOH-EG8-Fmoc) (Fig. 1, cyan, orange, and gray lines, respectively). By contrast, in the presence of GMM-EG8-Fmoc, a significant increase of the Td (44.1 °C) was observed (Fig. 1, purple line). This clearly showed that hCRT with GMM-EG8-Fmoc was more heat-stable than that with GMM and COOH-EG8-Fmoc, suggesting that hCRT was bound to GMM-EG8-Fmoc relative strongly. Strong binding of the glycan ligands to hCRT would be achieved using ligands with both glycan and hydrophobic aglycone moieties.

Fig. 1

Thermal shift assay of human calreticulin (hCRT) with ligands. a) Melt peaks of hCRT in the absence or presence of ligands. A typical peak from three independent experiments was shown. b) Denaturing temperature (td) values of hCRT with or without ligands. Data with error bars indicate mean values with standard deviation of three independent experiments. Statistical significance was calculated using one-way ANOVA with post-hoc Tukey test, ^**P < 0.01. GMM, Glc-α1,3-man-α1,2-man; PEG, polyethylene glycol; Fmoc, hydrophobic aglycone.

Thermodynamic and kinetic analysis of CRT-targeting glycan ligand

The limitations of the thermal shift assay include its inability to determine parameters such as stoichiometry and binding affinity. Therefore, kinetic and thermodynamic analyses of each ligand were performed using isothermal titration calorimetry (ITC). We analyzed binding constant (K_b) and thermodynamic parameters (stoichiometry, ∆H, and ∆S) with a series of ligands. The hCRT bound to GMM and GMM-EG8-Fmoc but not to COOH-EG8-Fmoc and maltotriose (Fig. S2). The detailed properties are summarized in Table 1. Similar to the results of the thermal shift assay (Fig. 1), GMM-EG8-Fmoc was strongly bound to hCRT and the binding constant (K_b = 5.0 × 10⁶ M⁻¹) was approximately 50-fold higher than that of GMM (K_b = 1.4 × 10⁵ M⁻¹) (Table 1). The affinity for the hCRT was comparable with that of the natural-type glycan Glc₁Man₉GlcNAc₂-OC₃H₇ (K_b = ∼5.2 × 10⁶ M⁻¹) (Ito et al. 2005). Furthermore, to the best of our knowledge, the affinity of GMM-EG8-Fmoc is stronger than that of previously reported CRT ligands (Kapoor et al. 2003; Ito et al. 2004; Arai et al. 2005; Gopalakrishnapai et al. 2006). The ITC data suggests that the glycan moiety is essential for binding to hCRT and that EG8-Fmoc additionally enhances binding. The stoichiometry (n) of GMM-EG8-Fmoc for hCRT was 1:1, which supported the hybrid binding concept. Thermodynamic parameters (∆G, ∆H, and − T∆S) showed that the binding was an enthalpy-driven spontaneous reaction. This suggests that binding is driven by specific interactions with the ligand that likely include hydrogen-bonding and Van der Waals interactions, and not by desolvation of the lipophilic Fmoc group (Table 1). These analyses revealed that GMM-EG8-Fmoc increases the affinity compared to GMM, and the enthalpy-favored reaction. Taken together, the ITC results suggest that the validity of the hybrid binding concept in terms of increasing affinity.

Table 1

Thermodynamic parameters of hCRT ligands.

Ligand	n (sites)	Kb (M−1)	∆Gb (kcal/mol)	∆H (kcal/mol)	−T∆S (kcal/mol)
GMM	1.16	1.4 × 105	−6.78	−13.3	6.52
GMM-EG8-Fmoc	1.03	5.0× 106	−8.82	−15.4	6.58
COOH-EG8-Fmoc	−	n.d.a	−	−	−
Maltotriose	−	n.d.a	−	−	−

^aNot determined. b. ∆G = ∆H − T∆S

Inhibition of chaperone activity using CRT-targeting glycan ligand

Finally, we investigated whether GMM-EG8-Fmoc inhibited the chaperone activity of hCRT. To examine the chaperone activity of hCRT, we performed a classical aggregation assay using citrate synthase (CS) (Buchner et al. 1998). Briefly, the heat-induced aggregation of CS was determined by monitoring the increase in turbidity changing in absorbance at 360 nm. Under hCRT conditions, CS aggregation is suppressed and CRT chaperone activity can be analyzed. Therefore, we used the chaperone activity assay to examine the inhibitory effects of glycan ligands. The CS aggregation was reduced by approximately one third with the addition of hCRT (Fig. 3, comparison between CS only and CS + hCRT). In the presence of GMM-EG8-Fmoc, CS aggregated in a dose-dependent manner (1 to 100 μM) compared to CS alone, and almost all the same absorbance was observed at 100 μM, whereas 100 μM GMM or COOH-EG8-Fmoc did not promote CS aggregation (Fig. 2). Although a previous report demonstrated that glycan alone can inhibit the chaperone activity of CRT, the G1M3 structure is required to inhibit chaperone activity (Saito et al. 1999). This may explain why G1M2 alone did not inhibit chaperone activity. Taken together, these results indicate that hybrid binding of hCRT inhibits its chaperone activity.

Fig. 2

Human calreticulin (hCRT) chaperone activity inhibition assay. Data with error bars show mean values with standard deviation (three independent experiments). Significance was calculated using one-way ANOVA with post-hoc Tukey test. P < 0.05 was considered statistically significant. ^*P < 0.05. GMM, Glc-α1,3-man-α1,2-man; EG8, octa ethylene glycol; Fmoc, hydrophobic aglycone.

Discussion

The Fmoc-binding site in CRT is an attractive topic. Thus far, we have identified two possibilities for the Fmoc binding site based on previous studies. The first is a polypeptide-binding site located in the lectin domain (Lum et al. 2016). Another possibility is that the hydrophobic amino acids contain peptide sequences in their flexible P-domain (Mohrlüder et al. 2007; Thielmann et al. 2009). To compare these two possibilities, we calculated the hydrophobicity of hCRT using PyMOL software (Eisenberg et al. 1984) (Fig. S3a). Using complex structure of hCRT with a heptasaccharide revealed by the cryo-electron microscopy (Domnick et al. 2022), we compared the two possible sites located in hCRT (Fig. S3b and S3c). A polypeptide-binding site in the lectin domain was located on the opposite side of the glycan-binding site (Fig. S3b), whereas a hydrophobic site in the flexible P-domain was located on the same side of the glycan-binding site. This was calculated as hydrophobic and contained aromatic amino acids, such as phenylalanine and tryptophan (Fig. S3c). From the ITC results (Table 1), we suggested that GMM-EG8-Fmoc exhibited enthalpy-driven binding to hCRT despite the addition of hydrophobicity by Fmoc. This unusual thermodynamic change might be explained by enthalpy-driven nonclassical hydrophobic interactions, which have been reported in several protein-ligand interactions (Meyer et al. 2003; Setny et al. 2010; Gómez-Velasco et al. 2020). Furthermore, this phenomenon might indicate that nonpolar interactions such as London dispersion forces and induced dipolar forces contribute to the binding affinity of GMM-EG8-Fmoc. Among these weak forces, π-π interactions of the aromatic ligand with the tryptophan of the protein have been indirectly shown to be an enthalpy-driven interaction (Shao et al. 2022). Additionally, other hydrophobic sites containing aromatic amino acids near the glycan-binding site and specific regions of the P-domain (Fig. S3a) were identified. Although further experimental evidence is needed, we predicted that the hydrophobic site containing aromatic amino acids near the glycan-binding site or in the P-domain may be a reliable Fmoc-binding site (Fig. S3d).

In this study, we designed and developed a CRT-targeting ligand, GMM-EG8-Fmoc. GMM-EG8-Fmoc bound to hCRT with micromolar affinity, and both GMM and Fmoc were necessary to maximize affinity, thereby supporting our proposed hybrid binding concept. Furthermore, inhibition of chaperone activity indicated that GMM-EG8-Fmoc inhibited hCRT in vitro. The glycoprotein folding steps contribute not only to physiological glycoprotein synthesis but also to the pathological process of some diseases (Ferris et al. 2014; Wang and Kaufman 2016). In addition, the multifunctionality of CRT (Johnson et al. 2001; Raghavan et al. 2013; Fucikova et al. 2021) may indicate the availability of GMM-EG8-Fmoc. Therefore, inhibition of CRT may be useful for expanding our understanding of CRT’s physiological and pathological functions. The hybrid-binding concept proposed in this study may provide a strategy for the design and development of ligands and CRT inhibitors.

Materials and methods

The procedure for chemical synthesis is included in the supplementary data.

General methods

All the purchased reagents were used without further purification. SYPRO^Ⓡ Orange dye was purchased from Invitrogen. A thermal shift assay was conducted using real-time PCR equipment (C1000™Thermal Cycler; Bio-Rad). Isothermal titration calorimetry was performed using an iTC 200 calorimeter (Malvern Instruments). Absorbance was measured using a micro plate reader (FlexStation^Ⓡ3; Molecular Devices).

Thermal shift assay for recombinant human CRT with ligands

Recombinant human CRT (4 μM) was mixed with 8 μM of ligands (GMM, GMM-EG8-Fmoc, COOH-EG8-Fmoc, maltotriose) in the reaction buffer [10 mM MOPS (pH 7.4), 5 mM CaCl₂, 150 mM NaCl and 4% DMSO]. The mixture was incubated at 4 °C for 1 h. After incubation, 60 × SYPRO^Ⓡ Orange dye (original stock solution: 5000 ×) was added, and half of the mixture was used for denaturation temperature analysis using real-time PCR equipment (CFX96™Real-Time System, C1000™Thermal Cycler, Bio-Rad). Protein denaturation was plotted as melt curves based on the fluorescence intensity of SYPRO^Ⓡ Orange dye (λ_em = 470 nm, λ_ex = 570 nm), and converted melt peaks indicated the Td. The analytical conditions consisted of heating the mixtures from 20 to 70 °C at increments 0.2 °C for 10 sec.

Isothermal titration calorimetry of recombinant hCRT with ligands

Recombinant human CRT (22 μM) in 10 mM MOPS (pH 7.4), 5 mM CaCl₂, 150 mM NaCl and 2.5% DMSO was added to a sample cell of an isothermal titration calorimeter (iTC 200 calorimeter, Malvern Instruments). Each ligand (GMM, GMM-EG8-Fmoc, COOH-EG8-Fmoc, or maltotriose) at a concentration of 250 μM in the same buffer was titrated to the sample cell at 750 rpm and 291 K. The titration was set by an initial injection of 0.5 μL, followed by 16 2.6-μL injections over 4 s with 150-s spacing intervals. The corrected data were fitted by nonlinear regression with one set of site models using the ITC analysis software package Origin 7 (OriginLab).

Inhibition assay for recombinant hCRT chaperone activity

Citrate synthase (1 μM) and recombinant hCRT (3 μM) were added to the reaction buffer (10 mM HEPES pH 7.4, 150 mM NaCl, and 5 mM CaCl₂) with or without ligands (1, 10, or 100 μM GMM-EG8-Fmoc, 100 μM GMM, or 100 μM COOH-EG8-Fmoc). After heating of the mixtures at 45 °C for 15 min, the CS aggregation was analyzed using 360-nm absorbance as an indicator using a microplate reader (FlexStation^Ⓡ3, Molecular Devices).

Author contributions

Taiki Kuribara (Writing—original draft, Writing—review and editing, Data curation, Investigation, Formal analysis, Supervision), Mitsuaki Hirose (Writing—original draft, Writing—review and editing), Yoichi Takeda (Writing—original draft, Writing—review and editing, Data curation, Investigation, Formal analysis), Kiichiro Totani (Writing—original draft, Supervision); Taiga Kojima (Data curation, Investigation, Formal analysis), Yuka Kobayashi (Data curation, Investigation, Formal analysis), Keita Shibayama (Data curation, Investigation, Formal analysis). All authors reviewed and approved the final manuscript.

Taiki Kuribara (Data curation [lead], Formal analysis [lead], Investigation [equal], Supervision [equal], Writing—original draft [lead], Writing—review & editing [equal]), Taiga Kojima (Data curation [supporting], Formal analysis [supporting], Investigation [lead]), Yuka Kobayashi (Data curation [equal], Formal analysis [supporting], Investigation [equal]), Mitsuaki Hirose (Writing—original draft [lead], Writing—review & editing [lead]), Keita Shibayama (Data curation [equal], Formal analysis [lead], Investigation [equal]), Yoichi Takeda (Data curation [equal], Formal analysis [equal], Investigation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Kiichiro Totani (Supervision [lead], Writing—original draft [lead]).

Funding

This work was partially supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI [Grant Number JP22K05335 to K.T.], and the Fujimori Science and Technology Foundation [to M.H.].

Conflict of interest

None declared.