Abstract

The endoplasmic reticulum malectin is a highly conserved protein in the animal kingdom that has no counterpart so far in lower organisms. We recently determined the structure of its conserved domain and found a highly selective binding to Glc2Man9GlcNAc2, an intermediate of N-glycosylation. In our quest for putative ligands during the initial characterization of the protein, we noticed that the malectin domain is highly specific for diglucosides but quite tolerant towards the linkage of the glucosidic bond. To understand the molecular requirements for the observed promiscuity of the malectin domain, here we analyze the binding to a range of diglucosides through comparison of the protein chemical shift perturbation patterns and the saturation transfer difference spectra of the ligands including two maltose-mimicking drugs. A comparison of the maltose-bound structure of the malectin domain with the complex of the native ligand nigerose reveals why malectin is able to tolerate such a diversity of ligands.

Introduction

One of the main functions of the endoplasmic reticulum (ER) is the biogenesis of glycoproteins. More than 50% of all proteins are predicted to become glycosylated with a diverse range of oligosaccharides (Apweiler et al. 1999) that are specific for species, tissue and cell type (for an overview about attached glycosides, see e.g., Lütteke et al. 2005). Most of the proteins use N-linked glycosylation to a NXS or NXT motif. Current estimations suggest that there may be as many as 1000 different mammalian N-glycans. The attached oligosaccharides confer increased stability and solubility to the proteins (Shental-Bechor and Levy 2008) and are used as sorting signals in intracellular trafficking or intercellular signaling. Correct folding of proteins destined for compartments in the secretory and endocytic pathways, the plasma membrane and for secretion is monitored by a quality control machinery that retains and ultimately disposes of misfolded proteins (Helenius 2001; Helenius and Aebi 2004; Anelli and Sitia 2008).

Glycoprotein synthesis starts in the ER with a series of identical steps conserved in almost all eukaryotes. A preassembled, lipid-linked oligosaccharide Glc3Man9GlcNAc2 is transferred, cotranslationally, by oligosaccharyltransferase from the lipid-linked sugar donor, Glc3Man9GlNAc2-diphosphate dolichol, to selected asparagine residues on nascent polypeptide chains. Thereafter, the terminal glucoses of the N-glycan are sequentially trimmed by glucosidases I and II (Helenius and Aebi 2004). Folding is assisted by the ER resident chaperones calnexin and calreticulin (Hebert et al. 1996) that recognize the monoglycosylated form of the high mannose N-glycan on the polypeptide chain. Cleavage of the terminal glucose residue by glucosidase II releases the glycoprotein from calnexin/calreticulin. Incomplete folding is sensed by UDP-Glc:glycoprotein glucosyltransferase that reglucosylates the glycoprotein and starts a new round of assisted folding (Sousa et al. 1992). After repeated rounds of folding, correctly folded glycoproteins exit to the Golgi apparatus where further processing takes place, whereas prolonged misfolded proteins are directed to the endoplasmic reticulum associated protein degradation pathway (Vashist and Ng 2004). In this system, the partially trimmed N-glycans serve as crucial signals that invoke the next step in glycoprotein maturation through interaction with the lectin domains of the involved proteins.

We recently discovered a novel but highly conserved carbohydrate-binding protein from the ER of—at that time—unknown function, which we named malectin (Schallus et al. 2008). Nuclear magnetic resonance (NMR) structure determination showed that its luminal part is a carbohydrate-binding domain. We could further show that it interacted with nigerose, the diglucoside Glcα1-3Glc. Together with the ER location of malectin, this pointed to an interaction with the early N-glycan intermediates, the only oligosaccharides of the cell to contain glucose moieties. Carbohydrate microarray analyses corroborated this and revealed an intense and highly selective binding to the Glc2Man9GlcNAc2 (Glc2-N-glycan) rather than the Glc3-N-glycan or the Glc1Man9GlcNAc2 (Glc1-N-glycan) N-glycans.

Despite any sequential homology, the malectin domain unexpectedly revealed a close structural similarity to carbohydrate-binding modules (CBMs) of bacterial glycosylhydrolases. CBMs in general are bacterial non-enzymatic carbohydrate-binding domains found in glycoside hydrolases (Wormald et al. 2002; Boraston et al. 2004). Their task is to direct the parent enzymes to the target polysaccharides of the plant cell wall. Observed binding affinities of CBMs to the oligosaccharides are often low in the range of 105 to 104 M1, but this can be compensated by either multivalent binding sites or by several CBMs acting in tandem. Currently, 39 families have been described on the basis of amino acid sequence similarity (Hashimoto 2006; Cantarel et al. 2009).

Another classification scheme of CBMs relies on the interaction with the substrate (Wormald et al. 2002; Boraston et al. 2004). Type A modules bind to the flat surfaces of crystalline polysaccharides and correspondingly show a flat interface composed of aromatic residues. Type B modules recognize single polysaccharide chains that fit into a groove on the protein surface. The depth of the cleft varies from shallow to being able to accommodate the entire pyranose ring and, as in type A modules, is lined with aromatic residues. Type C modules finally interact with mono-, di- or trisaccharides and in this respect resemble eukaryotic lectins.

With respect to both the total number of families and the number of entries in databases, the dominant fold among CBMs and also several animal lectins is the β-sandwich fold, further subdivided into β-jelly roll and immunoglobulin like (Wormald et al. 2002; Boraston et al. 2004; Hashimoto 2006). Prominent non-CBM members of the first group are galectins (Bourne et al. 1994) and calnexin (Schrag et al. 2001), and malectin also adopts this fold. However, malectin in contrast to many lectins and bacterial CBMs does not contain a calcium ion to stabilize the fold. Another distinct criterion of malectin is that carbohydrate binding in β-jelly roll CBMs and lectins is often mediated by clefts on the surface located on the concave β-sheet side, whereas the malectin binding cleft is formed by the loops connecting the β-sheets, a binding mode less commonly observed (Hashimoto 2006).

Initial ligand-screening studies showed that malectin not only binds to nigerose but also to maltose and related diglucosides, on the basis of which we chose the protein name (maltose binding lectin). Here, we compare the interactions of the protein to a range of diglucosides as well as to two maltose-mimicking drugs by NMR spectroscopy. 15N-relaxation dynamic studies and the comparison of the maltose-bound structure with the nigerose-bound form of the malectin domain reveal why malectin is able to tolerate such a diversity of ligands.

Results

Chemical shift assignment

Most of the resonances of the free malectin domain could be assigned through standard heteronuclear backbone, side-chain and NOESY experiments with the exception of four loops (L1–L4), the residues of which experienced severe exchange broadening. In the ligand-bound form, several new peaks became visible in the 1H15N-HSQC spectrum that had been barely detectable in the free protein (Figure 1A and supplemental Figure S1). A new set of heteronuclear backbone experiments recorded in the presence of 4 mM maltose allowed us to identify these resonances as those of the missing residues (with the exception of G62, Y67, Y117 and D186). In the following, several previously unidentified side chain systems of the 13C-edited NOESY and TOCSY spectra could be connected to the corresponding residues, resulting in almost complete assignment. The subsequent structure determination revealed that all four loops (G62-G68, T86-N90, E114-A118 and Y185-N187) were arranged on one side of the domain (Schallus et al. 2008).

Fig. 1

(A) Superposition of 1H15N-HSQC spectra of the malectin domain in free (black), maltose-bound (red) and nigerose-bound (green) forms recorded at 600 MHz. Cross peaks are labeled with the amino acid number and type. The two expanded views show regions of the spectrum where several of the residues crucial for the interaction resonate. (B) Chemical shift changes of maltose induced by binding to the malectin domain. Superposition of the 1H13C-HSQC spectra of 13C-maltose in free (black) and protein-bound (red) forms. The spectra show a mixture between α- and β-maltose because both forms interconvert in solution. The resonances of the reducing residue in the β-enantiomer that could be fully assigned as well as the anomeric carbons are labeled. Peaks labeled with an asterisk originate from natural abundance protein resonances.

Fig. 1

(A) Superposition of 1H15N-HSQC spectra of the malectin domain in free (black), maltose-bound (red) and nigerose-bound (green) forms recorded at 600 MHz. Cross peaks are labeled with the amino acid number and type. The two expanded views show regions of the spectrum where several of the residues crucial for the interaction resonate. (B) Chemical shift changes of maltose induced by binding to the malectin domain. Superposition of the 1H13C-HSQC spectra of 13C-maltose in free (black) and protein-bound (red) forms. The spectra show a mixture between α- and β-maltose because both forms interconvert in solution. The resonances of the reducing residue in the β-enantiomer that could be fully assigned as well as the anomeric carbons are labeled. Peaks labeled with an asterisk originate from natural abundance protein resonances.

Four aromatic residues are the key residues for the interaction

In the presence of maltose or nigerose, four aromatic ring systems also appeared in the 2D 1H-NOESY spectra, which had no counterpart in the free domain but showed NOEs to the carbohydrate moiety (supplemental Figure S2 A). This matched with the fact that five ring systems had remained undetected in the free domain (Y67, Y89, Y116, F117 and Y185). Due to inherent line broadening observed for the five residues in the backbone experiments, it was impossible to unambiguously connect the new aromatic systems to their corresponding backbone resonances. Therefore, we chose a mutational strategy: each of the unassigned tyrosine residues was exchanged to phenylalanine and the phenylalanine F117 to tyrosine, respectively, and 1H-NOESY spectra were recorded in free and maltose-bound forms. These conservative mutations were chosen to retain binding.

As expected, each of the unassigned aromatic ring systems disappeared in turn for four of the five mutations, while resonances of a new aromatic system reappeared (supplemental Figure S2 B–E), enabling us to assign the corresponding aromatic residues. The NOEs between the protein and the carbohydrate showed that Y67 and Y89 contacted the terminal, non-reducing glucose of maltose, F117 contacted the reducing glucose, and Y116 contacted both. Y185 showed no NOEs to the sugar moiety. With the help of these NOEs, we were able to calculate the structures of the malectin domain in complex with nigerose (2K46; Schallus et al. 2008) and maltose (vide infra).

To investigate the role of the other residues within the four loops experiencing exchange-broadening, we created several further mutants and tested them for maltose interaction. Hydrophobic residues were exchanged against alanine; hydrophilic ones against serine. In this way, we mutated R63A64S65D66Y67 of loop 1 to SASSA, E87R88Y89 of loop 2 to SSA, E114V115Y116 of loop 3 to SVS and Y184Y185D186N187 of loop 4 to SASS. Single mutations comprised R88S and E114S.

Whereas the R63A64S65D66Y67-SASSA, Y184Y185D186N187-SASS and E114S mutant domains still showed chemical shift changes, indicating a weakened interaction with maltose (supplemental Figure S3 A–C), the E114V115Y116-SVS mutation completely abrogated the interaction of malectin with maltose (supplemental Figure S3 D). Two mutations in the second loop, R88S and E87R88Y89-SSA, caused unfolding of the domain as demonstrated by 1H15N-HSQC spectra (supplemental Figure S3 E and F). Since the Hη protons of R88 may form salt bridges to E42, D55 and the Hε to E58 according to the structure, R88 is apparently needed to stabilize the fold.

Our results show that, although the resonances of all four loops are affected by the interaction with maltose, only few like Y116 are truly important for the binding event.

Carbohydrate specificity

During our search for putative carbohydrate ligands, we observed that exclusively diglucose residues bind (Schallus et al. 2008). Although the affinity was highly specific, the malectin domain tolerated several isomer diglucoside ligands. We assayed a range of possible diglucosides and several analogs by chemical shift perturbation in 1H15N-HSQC spectra. The 1H15N-HSQC spectrum of a protein is regarded as its fingerprint within a specific environment. Chemical shifts are sensitive to alterations in the electronic and/or conformational environments as induced by binding of a ligand molecule. Changes in 15N and 1HN chemical shifts are generally employed to identify the residues involved in binding and to discriminate between different binding modes.

Figure 2 shows a comparison of the chemical shift changes of the malectin domain induced by the addition of either the native ligand nigerose (Glcα1-3Glc), or maltose (Glcα1-4Glc), kojibiose (Glcα1-2Glc) or cellobiose (Glcβ1-4Glc). All four diglucosides cause changes in the same areas, mainly the loops 61–69, 85–91, 110–117 and 182–191 and residues of the adjacent loop 137–148 (Figure 2A–D). Minor differences could be found for individual residues. One should, however, bear in mind that a large shift change cannot be interpreted as a direct interaction with the ligand, as e.g., F117 in hydrophobic contact with the reducing end glucose (as demonstrated by the presence of intermolecular NOEs) had a very low chemical shift change, whereas E114 displayed the highest observed shift change without any direct contact.

Fig. 2

Chemical shift changes of the malectin domain in (A) nigerose, (B) maltose, (C) kojibiose, (D) cellobiose, (E) α-nitrophenylmaltoside, (F) β-nitrophenylmaltoside, (G) α-methylglucose, (H) β-methylglucose, (I) acarbose© and (J) miglitol©. Residues that experience line broadening in the free form are marked with an asterisk. The protein concentration was 0.5 mM; the ligands were added in 10-fold excess.

Fig. 2

Chemical shift changes of the malectin domain in (A) nigerose, (B) maltose, (C) kojibiose, (D) cellobiose, (E) α-nitrophenylmaltoside, (F) β-nitrophenylmaltoside, (G) α-methylglucose, (H) β-methylglucose, (I) acarbose© and (J) miglitol©. Residues that experience line broadening in the free form are marked with an asterisk. The protein concentration was 0.5 mM; the ligands were added in 10-fold excess.

Since carbohydrate microarray analysis had revealed that malectin binds specifically to a Glc2-N-glycan, where the diglucoside is followed by mannose attached to the α-diglucoside, we also analyzed binding to α- and β-nitrophenylmaltoside (Figure 2E and F). The reducing end of a carbohydrate exists in two slowly interconverting anomers, depending on whether the hydroxyl group on the C1 atom is in axial or equatorial position. Appending the aromatic residue to the reducing end blocks mutarotation, and therefore only one conformation exists. Although the α-nitrophenylmaltoside would correspond to the glucosidic bond found in native ligand, malectin exhibited no preference for either of the two enantiomers. Likewise, the attached aromatic systems did not cause changes in the protein chemical shifts. This suggests that this part of the reducing sugar sticks out of the binding groove.

As this showed that two glucose residues are sufficient, we assayed whether we could further shorten the carbohydrate moiety and tested α- and β-methylglucoside (Figure 2G and H). Despite the fact that free glucose did not bind to the malectin domain, the attachment of a methyl group attached to the anomeric center of glucose was already sufficient to enable binding (Figure 2E and F). In this case, there was a clear preference for the α-enantiomer, whereas the β-enantiomer showed only few affected shifts.

Dissociation constants for all compounds were determined by isothermal titration calorimetry to compare the chemical shift patterns with the affinity of the interaction. The order of KD values was 26.3 ± 0.7 µM for nigerose (Schallus et al. 2008), 50 ± 0.5 µM for maltose (Schallus et al. 2008), 37 ± 1 µM for α-nitrophenylmaltoside, 51 ± 4 µM for β-nitrophenylmaltoside (Feher et al. 2008), 210 ± 6 µM for kojibiose (Schallus et al. 2008) and 500 ± 26 µM for α-methylglucoside. The KD values of β-methylglucoside and cellobiose could not be determined due to low affinity.

Titration of unlabeled protein to 13C-labeled maltose, performed as complementary experiment (Figure 1B), revealed that the presence of the malectin domain caused large effects on the sugar. The assignment of the resonances, however, was precluded by line broadening affecting the resonances of the non-reducing residue as well as those of the reducing residue of the α-enantiomer (apart from the anomeric proton), which disappeared completely. Only the resonances of the reducing residue of the β-enantiomer could be assigned by HSQC-TOCSY (data not shown) and are labeled in Figure 1B.

All compounds were further analyzed by saturation transfer difference (STD), a technique which monitors the transfer of magnetization from protons located in the binding pocket of the protein to protons of the ligand in the proximity of the binding site. Thus, only protons directly contacting malectin give rise to a resonance signal. The STD spectra with the assignment of the corresponding protons are shown in Figure 3. The STD spectrum of β-methylglucoside was empty, corroborating that this compound did not interact with malectin. Likewise, cellobiose (Glcβ1-4Glc) also revealed only weak signals. Resonances of the non-reducing residues of nigerose, maltose, kojibiose and nitrophenyl-maltosides gave 2- to 3-fold higher intensities than those of the second ring, which implied that they were in close contact to the protein.

Fig. 3

One-dimensional STD spectra for glucose and glucose disaccharides. (A) nigerose, (B) maltose, (C) kojibiose, (D) cellobiose, (E) α-nitrophenylmaltoside, (F) β-nitrophenylmaltoside, (G) α-methylglucose, (H) β-methylglucose, (I) acarbose© and (J) miglitol©. α and β denote the stereoisomers at the anomeric center of the glucose ring at the reducing end. The ring protons of the non-reducing residues are labeled 1-6; those of the reducing residues, 1′-6′. The protein concentration was 20 µM; the ligands were added in hundred fold excess.

Fig. 3

One-dimensional STD spectra for glucose and glucose disaccharides. (A) nigerose, (B) maltose, (C) kojibiose, (D) cellobiose, (E) α-nitrophenylmaltoside, (F) β-nitrophenylmaltoside, (G) α-methylglucose, (H) β-methylglucose, (I) acarbose© and (J) miglitol©. α and β denote the stereoisomers at the anomeric center of the glucose ring at the reducing end. The ring protons of the non-reducing residues are labeled 1-6; those of the reducing residues, 1′-6′. The protein concentration was 20 µM; the ligands were added in hundred fold excess.

Addition of the specific carbohydrate ligand stabilizes the binding loops

Line broadening as observed for the ligand-binding loops of the malectin domain is often caused by motion on the micro- or millisecond time scale. The fact that the aromatic ring resonances of Y67, Y89, Y116 and F117 were also exchange-broadened corroborates the idea that the entire loops are affected by conformational exchange in the unbound state.

We performed 15N-relaxation experiments to study the backbone dynamics of the malectin domain in free and maltose-bound forms (Figure 4). We obtained data for 119 residues at 500 MHz or 131 at 600 MHz out of 190 residues. The remaining residues were not analyzed due to either peak overlap or because their resonances were weak or missing. Addition of maltose allowed us to analyze 18 (500 MHz) or 19 (600 MHz) further residues located in the binding loops. Apart from the N- and C-termini, the domain shows fairly uniform values with little variations in the R1, R2 or heteronuclear NOE values. Residues of the ligand-binding loops that could only be analyzed in the maltose-bound form did not differ from the bulk of the protein, showing they are not subject to fast picosecond time scale motion. Addition of the ligand caused an overall reduction in the R2 values, which dropped from an average of 19.2 to 16.68 s1 at 500 MHz and from an average of 19.1 to 17.9 s1 at 600 MHz, while the R1 values were marginally affected. From the average R2/R1 values for rigid residues, we fitted a correlation time of 13.2 ± 0.09 ns for the free and 12.6 ± 0.09 ns for the ligand-bound protein and order parameters S2 with an average value of 0.85 (500 MHz) and 0.88 (600 MHz) in the free form and 0.91 (500 MHz) and 0.90 (600 MHz) in the bound form. The decrease in R2 and correlation time τC upon binding of maltose could be caused by a small contribution of dimerization or oligomerization of the protein in the free form; however, dynamic light scattering showed a sharp mono-disperse particle fraction with an average diameter of 2.2 nm, excluding the possibility that a noteworthy dimer fraction exists. The concomitant increase in the hetNOE indicated that the protein becomes more rigid upon binding of the ligand and that this is not confined to the loops.

Fig. 4

15N relaxation data for malectin at (A) 500 MHz and (B) 600 MHz. Below, the order parameters S2 derived from Lipari-Szabo model free analysis are illustrated. The values of free malectin are shown as bars, and the values of the maltose-bound protein are superimposed as black circles. Bars in the top panel denote the four binding loops.

Fig. 4

15N relaxation data for malectin at (A) 500 MHz and (B) 600 MHz. Below, the order parameters S2 derived from Lipari-Szabo model free analysis are illustrated. The values of free malectin are shown as bars, and the values of the maltose-bound protein are superimposed as black circles. Bars in the top panel denote the four binding loops.

Structural comparison of maltose- and nigerose-bound malectin domain (2KR2, 2K46)

Next to nigerose, maltose had the highest binding affinity to the malectin domain among the diglucosides tested (Schallus et al. 2008). Analogous to the nigerose complex, the structure determination of the maltose complex was based on 30 intermolecular NOEs between malectin and maltose in 13C-filtered NOESY spectra. Very little NOE information was obtained for the ligand-contacting side chains apart from the intermolecular NOEs both in the maltose and the nigerose complexes. As a consequence, in both complexes the ligand-binding amino acid side chains and the carbohydrate adopt a range of possible orientations. In Figure 5, a representative structure is shown for each complex. The mode of interaction in the maltose complex (2KR2) is comparable to that of the malectin nigerose interaction (2K46; Schallus et al. 2008). Maltose like nigerose is located in the binding cleft lined by the aromatic residues Y67, Y89, Y116 and F117 (Figure 5). The observed NOEs of the non-reducing glucose residue to Y67, Y89 and Y93 were essentially the same for maltose and nigerose. Accordingly, the non-reducing residue is placed in a similar position for both complexes. At the bottom of the cleft, the carboxyl group of E87 can form hydrogen bonds to the hydroxyl groups of C3 and/or C4 and the amine group of K189 to the hydroxyl groups of C3 or C2 of the non-reducing glucose residue in both structures. It is unclear whether N90 also participates in the interaction. In some structures, its amide side-chain group forms a hydrogen bond to the C2 hydroxyl group of the non-reducing residue.

Fig. 5

Comparison of the interaction of (A) maltose (2KR2) and (B) nigerose (2K46; Schallus et al. 2008) with the malectin domain. Stereo images of the carbohydrate-binding pocket are shown. The residues interacting with the diglucoside are displayed and labeled with their amino acid number. GLCr and GLCnr denote the reducing and non-reducing glucose residues of maltose and nigerose. The figures were produced with MOLMOL (Koradi et al. 1996).

Fig. 5

Comparison of the interaction of (A) maltose (2KR2) and (B) nigerose (2K46; Schallus et al. 2008) with the malectin domain. Stereo images of the carbohydrate-binding pocket are shown. The residues interacting with the diglucoside are displayed and labeled with their amino acid number. GLCr and GLCnr denote the reducing and non-reducing glucose residues of maltose and nigerose. The figures were produced with MOLMOL (Koradi et al. 1996).

The reducing glucose moiety is differently oriented in the two structures. The α1-4 glucosidic bond of maltose leads to an inversion of the A and B face of the reducing residue with respect to the α1-3 bond in nigerose. The intermolecular NOE pattern reflects this: for maltose, the Y116 Hδ and Hε protons are in close contact to H1′ and H3′, whereas in nigerose they contact H2′ and H4′. The inversion of the reducing glucose residue places the methyl C6′ group of maltose outside the binding pocket, whereas in the nigerose complex this group points into the cleft and can potentially form a hydrogen bond to the carboxyl group of D186 (Schallus et al. 2008). Alternatively, the latter is also in hydrogen bond distance to the hydroxyl group of C4′of nigerose.

Interaction with maltose-mimicking drugs

In diabetes treatment, maltose-mimicking drugs are in use that have been developed from α-glycosidase inhibitors. We tested two of these for interaction with malectin to see whether they could be of potential use in functional cellular studies.

Acarbose© bound to malectin and induced similar shift changes as maltose (Figures 2I and 3I), indicating that it binds to the interaction cleft (KD determined by isothermal titration calorimetry 340 ± 25 µM). This observation was not too unexpected as the third and fourth sugar rings of acarbose© correspond to maltose in composition and stereochemistry. Miglitol©, a compound derived from fungal norijimicin, is chemically more diverse, but it also interacted with malectin (Figures 2J and 3J). However, the binding seemed to be much weaker as only minor shifts could be observed. A KD of ~2 mM was determined for the miglitol© malectin interaction by STD competition experiments with maltose (Feher et al. 2008). This also confirmed that miglitol© interacts with malectin via the same binding site. In this respect, acarbose© and miglitol© are perfect candidates as lead compounds for the development of a malectin inhibitor.

Discussion

Only few other animal lectins are known that have a similar specificity for glucan oligosaccharides like malectin. The carbohydrate recognition domains of the calnexin family involved in the first steps of the N-linked glycosylation in the ER (calnexin family) recognize the terminal glucose of the Glc1-N-glycan. Insect β-1,3-glucan recognition protein (β-GRP) binds to β-1,3-glucan, a component of fungal cell walls (Takahasi et al. 2009). Surfactant protein D (SP-D) was shown to interact with maltose (Shrive et al. 2003), although its physiological ligand seems to be pustulan (β-1,6-linked glucose homopolymer; Allen et al. 2001). Finally, dectin-1 exclusively binds to 1,3-linked glucose oligomers with a minimum length of 10 residues (Brown and Gordon 2001; Palma et al. 2006). Apart from calnexin, these proteins are cell-surface carbohydrate receptors involved in innate immunity to pathogenic microorganisms, but they are not homologous to each other (Lawson and Reid 2000; Brown et al. 2003; Herre et al. 2004). β-GRP belongs to the Ig-like fold and thus is a member of the β-sandwich fold like calnexin and malectin. It binds the oligosaccharide mainly through hydrogen bonds between residues on the β-sheet side and the OH-groups of the carbohydrate moiety comparable to the calnexin carbohydrate interaction (Schrag et al. 2001). The other two proteins are members of the C-type lectin family, which have a different fold and interaction mode.

The malectin binding cleft is composed of four loops located on the apical side of the domain. This is untypical for β-sandwich animal lectins. In β-jelly roll CBMs, binding via the β-sheet connecting loops is only found in few members (belonging to CBM6 and CBM35; Tunnicliffe et al. 2005; Hashimoto 2006), whereas the binding cleft of the majority is found on the concave side of the β-sheet. Among the closest structural homologs, Clostridium thermocellum Xyn10B CBM22-2 (Xie et al. 2001) and the sugar binding domain of Fbs1 (SBD-Fbs1), an F-box lectin (Mizushima et al. 2004), only SBD-Fbs1 also binds with the loops connecting the β-sheets, whereas CBM22-2 binds its substrate via the β-sheet face. Thus, a structural homology in carbohydrate recognition domains does not automatically imply identical binding sites.

Apart from its high selectivity towards glucose residues, malectin is fairly promiscuous towards the glucosidic bond. At first glance, this was astonishing because sterically different glucosidic linkages place the second residue and especially its polar OH groups in different spatial positions with respect to the first glucose residue. This should be especially detrimental for protein–carbohydrate hydrogen bond interactions that depend on defined, well-oriented positions between donor and acceptor. Also, aromatic ring stacking between the apolar side of the carbohydrates and the protein recognizes a well-defined interaction surface. Most prokaryotic carbohydrate-binding modules that use these two ligation types in general select their substrates with high specificity (CBMs; families 4, 6, 9 and 22; Boraston et al. 2002, 2004; Pires et al. 2004).

A possible explanation for malectin’s promiscuity is that the interactions with the terminal non-reducing glucose, which is inserted more deeply into the cleft, provide the driving energy for binding (Figure 5). The position of this residue does not differ much in the maltose-bound and nigerose-bound complexes, nor probably in all the other disaccharides tested. The requirements for the second residue, which is only partially buried, apparently are lower since substrates as diverse as α-methylglucose or miglitol can bind. Essentially, this residue needs to provide some hydrophobic interaction site and should not have hydroxyl groups in axial position that would interfere with the surrounding aromatic side chains as, e.g., mannose or galactose (Schallus et al. 2008). The steric rearrangements of the side chains necessary to adapt to the diverse positions of the hydroxyl groups and the bulky methyl-hydroxy group arising from different glucosidic bonds are probably achieved through the intrinsic flexibility of the binding loops. The entire binding site and the surrounding residues on the four loops show a high mobility in the free form that freezes upon ligand binding. It is conceivable that, among the many possible interconverting loop conformations, each ligand selects for the one that best accommodates the respective carbohydrate.

The actual binding cleft may extend towards the two aromatic residues Y184 and Y185 adjacent to the binding interface. In the current structural models, they do not interact with either nigerose or maltose (Figure 5) or with the aromatic ring of α- and β-nitrophenylmaltoside (Figure 2E and F). This may be different for longer carbohydrate chains. In Glc2-N-glycan, the mannose adjacent to the glucose is attached via an α-1-3-bond which places it straight outside the binding cleft. The following mannoses, however, are linked via α-1-2-bonds which cause a bend in the oligosaccharide chain. Based on considerations on the published structures of Glc3Man (Petrescu et al. 1997; Mackeen et al. 2009) and the high Mannose Man9-glycan (Woods et al. 1998), it may be possible that either the next or second next mannose could curve back to the binding cleft and interact with these tyrosines. Several charged or polar residues are lying on the opposite loop and could complement the interaction.

Binding to more than one specific oligosaccharide is not a specific feature of malectin but has also been observed for galectins (Solomon et al. 1991; Schwarz et al. 1998; Ford et al. 2003) and some CBMs, e.g., CtCBM11 (Carvalho et al. 2004). To our current knowledge, malectin is exposed to only very few diverse oligosaccharides in the cell, which are derivatives of the Glc3-N-glycan precursor and composed of glucose, mannose and N-acetyl-glucosamine. In the protected environment of the ER, it mainly needs to discriminate between one, two or three terminal glucose residues but not between the many variations of carbohydrate structure that occur, e.g., on cell surfaces. Malectin selects with a very high specificity for the Glc2-N-glycan, albeit Glc3-N-glycan and Glc1-N-glycan also interact to some extent (Schallus et al. 2008). This promiscuity in binding could be potentially used to develop inhibitors of the protein useful for in vivo studies of the protein to elucidate its cellular function. Our findings that miglitol© also interacts with malectin denotes this compound as a putative lead compound.

Materials and methods

Cloning, expression and purification of the malectin constructs

Cloning, expression and purification of the malectin domain have been previously described (Schallus et al. 2008). Mutations were generated using the Quick Change Site-Directed Mutagenesis Kit (Stratagene). All sequences were verified by DNA sequencing.

NMR experiments

The carbohydrates were purchased from Sigma-Aldrich with the exception of nigerose, which was purchased from COSMO BIO Co., Ltd, miglitol© (purchased from Carbomer Inc.), and acarbose© (purchased from Hangzhou VIWA Co., Ltd, China).

All spectra were acquired at 22°C in 20 mM potassium phosphate buffer pH 6.8, 150 mM KCl and 2 mM 1,4 dithiothreitol. Details on the experiments used for chemical shift assignment of the protein and carbohydrates have been previously published (Schallus et al. 2008). 1H-NOESY and 13C-half-filtered NOESY spectra of 0.8 mM uniformly 13C/15N labeled malectin and 4 mM maltose were recorded at 600 MHz on a Bruker DRX600 equipped with a cryoprobe with mixing times of 80 ms (1H-NOESY) and 150 ms (13C-half-filtered NOESY). For ligand interaction studies, shifting peaks were traced in 1H15N-HSQC spectra. An 1H15N-HSQC spectrum of 500 µM uniformly 15N-labeled protein in free form was recorded and compared to a spectrum in the presence of a 10-fold excess of ligand, which was added to the protein solution from a concentrated stock solution. The difference in chemical shift Δδ was calculated with the formula Δδ = √ (δ2HN + (δN/10)2).

A sample of 0.3 mM 13C-labeled maltose (Omicron Biochemicals) with 1 mM unlabeled malectin domain was used to determine the resonances of the bound form of maltose. Data were processed with NMRPIPE (Delaglio et al. 1995) and analyzed using NMRVIEW (Johnson 2004).

Saturation transfer difference experiments

Protein samples of 20 μM were used with a 100-fold excess of ligand. The spectra were measured using a pulse sequence in which the difference between the on- and off-resonance experiments was created by phase cycling (Mayer and Meyer 2001). Presaturation of the protein NMR signals was performed using a train of selective Gaussian pulse of a duration of 49 ms and a field strength of 75 Hz each that were separated by a short delay of 1 ms. The on-resonance frequency used for presaturation of the protein signals was set to 0.8 ppm, while the off-resonance irradiation was applied at −40 ppm outside the area of NMR resonances. For the suppression of background protein signals, a 30-ms T spinlock was used as relaxation filter with a field strength of 5 kHz. For the epitope mapping analysis, STD signals were assigned using the assignment of the free ligands (Schallus et al. 2008). In addition, STD-TOCSY and STD-HSQC spectra were acquired when necessary. Quantitative evaluation of the STD signal was prevented by the strong overlap in the ring proton region; nevertheless, saturation time buildup curves were recorded to assess the relative signal intensities qualitatively.

15N relaxation measurements

Longitudinal and transverse relaxation times as well as heteronuclear NOE were measured at 500 (Bruker DRX500) and 600 MHz at 22°C at a protein concentration of 0.7 mM and a maltose concentration of 3.5 mM. Relaxation delays for T1 were varied between 10.8 and 2074 ms for 500 MHz and between 10.8 and 1642 ms for 600 MHz (Farrow et al. 1994). T2 of backbone amide 15N was obtained from T measurements (Massi et al. 2004) where a spinlock field strength of 2000 Hz was applied. The T decay was sampled at relaxation delays between 10 and 350 ms (500 and 600 MHz) with one duplicate point to test for instabilities. The heteronuclear NOE was determined as the signal intensity ratio of the HN/N cross peaks with and without 1H saturation. All experiments were recorded in an interleaved manner (Kay et al. 1989). The water signal was suppressed with a combination of the water-flip-back and WATERGATE scheme in all cases. The model-free analysis was carried out with the program Tensor2 (Dosset et al. 2000).

Structure calculation

Structure calculation of the maltose–malectin complex was achieved with ARIA1.2/CNS as previously described (Linge et al. 2001; Schallus et al. 2008). α-Maltose was attached to malectin in the course of the calculation through 30 NOEs that were determined in 13C-edited half-filtered experiments with a mixing time of 150 ms. The resulting structural bundle was further refined with COSMOS to analyze the hydrogen bond pattern between the protein and the carbohydrate (Sternberg and Möllhoff 2001). The COSMOS-NMR force field allows structure refinements with quantum chemically calculated chemical shifts and their derivatives with respect to atomic coordinates using the Bond Polarization Theory (BPT; Sternberg 1988; Prieß and Sternberg 2001). In contrast to traditional force fields, the one in COSMOS-NMR is not parameterized using fixed atomic charges, but the charges are calculated by BPT routines. NMR data like distances from NOE spectra can be incorporated in the refinement as well.

Structural statistics can be found in the supporting information (Table S1).

Supplementary data

Supplementary data for this article is available online at http://glycob.oxfordjournals.org/.

Funding

This work was supported by the German Research Foundation [Mu-1606/1-5 to C.M.-G.].

The authors would like to thank Dr. M. Biskup (KIT, Karlsruhe) for valuable discussions and for critical reading of the manuscript. We thank Prof. M. Sattler (EMBL) for access to the NMR spectrometers.

Abbreviations

    Abbreviations
  • β-GRP

    β-1,3-glucan recognition protein

  • BPT

    Bond Polarization Theory

  • CBMs

    carbohydrate-binding modules

  • ER

    endoplasmic reticulum

  • NMR

    nuclear magnetic resonance

  • NOESY

    nuclear Overhauser effect spectroscopy

  • STD

    saturation transfer difference

  • TOCSY

    total correlation spectroscopy

References

Allen
MJ
Laederach
A
Reilly
PJ
Mason
RJ
Polysaccharide recognition by surfactant protein D: novel interactions of a C-type lectin with nonterminal glucosyl residues
Biochemistry
 , 
2001
, vol. 
40
 (pg. 
7789
-
7798
)
Anelli
T
Sitia
R
Protein quality control in the early secretory pathway
EMBO J
 , 
2008
, vol. 
27
 (pg. 
315
-
327
)
Apweiler
R
Hermjakob
H
Sharon
N
On the frequency of protein glycosylation, as deduced from analysis of the SWISS-PROT database
Biochim Biophys Acta
 , 
1999
, vol. 
1473
 (pg. 
4
-
8
)
Boraston
AB
Bolam
DN
Gilbert
HJ
Davies
GJ
Carbohydrate-binding modules: fine-tuning polysaccharide recognition
Biochem J
 , 
2004
, vol. 
382
 (pg. 
769
-
781
)
Boraston
AB
Ghaffari
M
Warren
RA
Kilburn
DG
Identification and glucan-binding properties of a new carbohydrate-binding module family
Biochem J
 , 
2002
, vol. 
361
 (pg. 
35
-
40
)
Bourne
Y
Bolgiano
B
Liao
DI
Strecker
G
Cantau
P
Herzberg
O
Feizi
T
Cambillau
C
Crosslinking of mammalian lectin (galectin-1) by complex biantennary saccharides
Nat Struct Biol
 , 
1994
, vol. 
1
 (pg. 
863
-
870
)
Brown
GD
Gordon
S
Immune recognition. A new receptor for beta-glucans
Nature
 , 
2001
, vol. 
413
 (pg. 
36
-
37
)
Brown
GD
Herre
J
Williams
DL
Willment
JA
Marshall
AS
Gordon
S
Dectin-1 mediates the biological effects of beta-glucans
J Exp Med
 , 
2003
, vol. 
197
 (pg. 
1119
-
1124
)
Cantarel
BL
Coutinho
PM
Rancurel
C
Bernard
T
Lombard
V
Henrissat
B
The Carbohydrate-Active EnZymes database (CAZy): an expert resource for glycogenomics
Nucleic Acids Res
 , 
2009
, vol. 
37
 (pg. 
D233
-
238
)
Carvalho
AL
Goyal
A
Prates
JA
Bolam
DN
Gilbert
HJ
Pires
VM
Ferreira
LM
Planas
A
Romao
MJ
Fontes
CM
The family 11 carbohydrate-binding module of Clostridium thermocellum Lic26A-Cel5E accommodates beta-1, 4- and beta-1, 3–1, 4-mixed linked glucans at a single binding site
J Biol Chem
 , 
2004
, vol. 
279
 (pg. 
34785
-
34793
)
Delaglio
F
Grzesiek
S
Vuister
GW
Zhu
G
Pfeifer
J
Bax
A
NMRPipe: a multidimensional spectral processing system based on UNIX pipes
J Biomol NMR
 , 
1995
, vol. 
6
 (pg. 
277
-
293
)
Dosset
P
Hus
JC
Blackledge
M
Marion
D
Efficient analysis of macromolecular rotational diffusion from heteronuclear relaxation data
J Biomol NMR
 , 
2000
, vol. 
16
 (pg. 
23
-
28
)
Farrow
NA
Muhandiram
R
Singer
AU
Pascal
SM
Kay
CM
Gish
G
Shoelson
SE
Pawson
T
Forman-Kay
JD
Kay
LE
Backbone dynamics of a free and phosphopeptide-complexed Src homology 2 domain studied by 15N NMR relaxation
Biochemistry
 , 
1994
, vol. 
33
 (pg. 
5984
-
6003
)
Feher
K
Groves
P
Batta
G
Jimenez-Barbero
J
Muhle-Goll
C
Kover
KE
Competition saturation transfer difference experiments improved with isotope editing and filtering schemes in NMR-based screening
J Am Chem Soc
 , 
2008
, vol. 
130
 (pg. 
17148
-
17153
)
Ford
MG
Weimar
T
Kohli
T
Woods
RJ
Molecular dynamics simulations of galectin-1-oligosaccharide complexes reveal the molecular basis for ligand diversity
Proteins
 , 
2003
, vol. 
53
 (pg. 
229
-
240
)
Hashimoto
H
Recent structural studies of carbohydrate-binding modules
Cell Mol Life Sci
 , 
2006
, vol. 
63
 (pg. 
2954
-
2967
)
Hebert
DN
Foellmer
B
Helenius
A
Calnexin and calreticulin promote folding, delay oligomerization and suppress degradation of influenza hemagglutinin in microsomes
EMBO J
 , 
1996
, vol. 
15
 (pg. 
2961
-
2968
)
Helenius
A
Quality control in the secretory assembly line
Philos Trans R Soc Lond B Biol Sci
 , 
2001
, vol. 
356
 (pg. 
147
-
150
)
Helenius
A
Aebi
M
Roles of N-linked glycans in the endoplasmic reticulum
Annu Rev Biochem
 , 
2004
, vol. 
73
 (pg. 
1019
-
1049
)
Herre
J
Willment
JA
Gordon
S
Brown
GD
The role of Dectin-1 in antifungal immunity
Crit Rev Immunol
 , 
2004
, vol. 
24
 (pg. 
193
-
203
)
Johnson
BA
Using NMRView to visualize and analyze the NMR spectra of macromolecules
Methods Mol Biol
 , 
2004
, vol. 
278
 (pg. 
313
-
352
)
Kay
LE
Torchia
DA
Bax
A
Backbone dynamics of proteins as studied by 15N inverse detected heteronuclear NMR spectroscopy: application to staphylococcal nuclease
Biochemistry
 , 
1989
, vol. 
28
 (pg. 
8972
-
8979
)
Koradi
R
Billeter
M
Wuthrich
K
MOLMOL: a program for display and analysis of macromolecular structures
J Mol Graph
 , 
1996
, vol. 
14
 
51–55
(pg. 
29
-
32
)
Lawson
PR
Reid
KB
The roles of surfactant proteins A and D in innate immunity
Immunol Rev
 , 
2000
, vol. 
173
 (pg. 
66
-
78
)
Linge
JP
O'Donoghue
SI
Nilges
M
Automated assignment of ambiguous nuclear overhauser effects with ARIA
Methods Enzymol
 , 
2001
, vol. 
339
 (pg. 
71
-
90
)
Lütteke
T
Frank
M
von der Lieth
CW
Carbohydrate Structure Suite (CSS): analysis of carbohydrate 3D structures derived from the PDB
Nucleic Acids Res
 , 
2005
, vol. 
33
 (pg. 
D242
-
246
)
Mackeen
MM
Almond
A
Deschamps
M
Cumpstey
I
Fairbanks
AJ
Tsang
C
Rudd
PM
Butters
TD
Dwek
RA
Wormald
MR
The conformational properties of the Glc3Man unit suggest conformational biasing within the chaperone-assisted glycoprotein folding pathway
J Mol Biol
 , 
2009
, vol. 
387
 (pg. 
335
-
347
)
Massi
F
Johnson
E
Wang
C
Rance
M
Palmer
AG
3rd
NMR R1 rho rotating-frame relaxation with weak radio frequency fields
J Am Chem Soc
 , 
2004
, vol. 
126
 (pg. 
2247
-
2256
)
Mayer
M
Meyer
B
Group epitope mapping by saturation transfer difference NMR to identify segments of a ligand in direct contact with a protein receptor
J Am Chem Soc
 , 
2001
, vol. 
123
 (pg. 
6108
-
6117
)
Mizushima
T
Hirao
T
Yoshida
Y
Lee
SJ
Chiba
T
Iwai
K
Yamaguchi
Y
Kato
K
Tsukihara
T
Tanaka
K
Structural basis of sugar-recognizing ubiquitin ligase
Nat Struct Mol Biol
 , 
2004
, vol. 
11
 (pg. 
365
-
370
)
Palma
AS
Feizi
T
Zhang
Y
Stoll
MS
Lawson
AM
Diaz-Rodriguez
E
Campanero-Rhodes
MA
Costa
J
Gordon
S
Brown
GD
, et al.  . 
Ligands for the beta-glucan receptor, Dectin-1, assigned using “designer” microarrays of oligosaccharide probes (neoglycolipids) generated from glucan polysaccharides
J Biol Chem
 , 
2006
, vol. 
281
 (pg. 
5771
-
5779
)
Petrescu
AJ
Butters
TD
Reinkensmeier
G
Petrescu
S
Platt
FM
Dwek
RA
Wormald
MR
The solution NMR structure of glucosylated N-glycans involved in the early stages of glycoprotein biosynthesis and folding
EMBO J
 , 
1997
, vol. 
16
 (pg. 
4302
-
4310
)
Pires
VM
Henshaw
JL
Prates
JA
Bolam
DN
Ferreira
LM
Fontes
CM
Henrissat
B
Planas
A
Gilbert
HJ
Czjzek
M
The crystal structure of the family 6 carbohydrate binding module from Cellvibrio mixtus endoglucanase 5a in complex with oligosaccharides reveals two distinct binding sites with different ligand specificities
J Biol Chem
 , 
2004
, vol. 
279
 (pg. 
21560
-
21568
)
Prieß
W
Sternberg
U
Fast calculation of 13C NMR chemical shift tensors using the bond polarization model
J Mol Struct: THEOCHEM
 , 
2001
, vol. 
544
 (pg. 
181
-
190
)
Schallus
T
Jaeckh
C
Feher
K
Palma
AS
Liu
Y
Simpson
JC
Mackeen
M
Stier
G
Gibson
TJ
Feizi
T
, et al.  . 
Malectin: a novel carbohydrate-binding protein of the endoplasmic reticulum and a candidate player in the early steps of protein N-glycosylation
Mol Biol Cell
 , 
2008
, vol. 
19
 (pg. 
3404
-
3414
)
Schrag
JD
Bergeron
JJ
Li
Y
Borisova
S
Hahn
M
Thomas
DY
Cygler
M
The structure of calnexin, an ER chaperone involved in quality control of protein folding
Mol Cell
 , 
2001
, vol. 
8
 (pg. 
633
-
644
)
Schwarz
FP
Ahmed
H
Bianchet
MA
Amzel
LM
Vasta
GR
Thermodynamics of bovine spleen galectin-1 binding to disaccharides: correlation with structure and its effect on oligomerization at the denaturation temperature
Biochemistry
 , 
1998
, vol. 
37
 (pg. 
5867
-
5877
)
Shental-Bechor
D
Levy
Y
Effect of glycosylation on protein folding: a close look at thermodynamic stabilization
Proc Natl Acad Sci USA
 , 
2008
, vol. 
105
 (pg. 
8256
-
8261
)
Shrive
AK
Tharia
HA
Strong
P
Kishore
U
Burns
I
Rizkallah
PJ
Reid
KB
Greenhough
TJ
High-resolution structural insights into ligand binding and immune cell recognition by human lung surfactant protein D
J Mol Biol
 , 
2003
, vol. 
331
 (pg. 
509
-
523
)
Solomon
JC
Stoll
MS
Penfold
P
Abbott
WM
Childs
RA
Hanfland
P
Feizi
T
Studies of the binding specificity of the soluble 14,000-dalton bovine heart muscle lectin using immobilised glycolipids and neoglycolipids
Carbohydr Res
 , 
1991
, vol. 
213
 (pg. 
293
-
307
)
Sousa
MC
Ferrero-Garcia
MA
Parodi
AJ
Recognition of the oligosaccharide and protein moieties of glycoproteins by the UDP-Glc:glycoprotein glucosyltransferase
Biochemistry
 , 
1992
, vol. 
31
 (pg. 
97
-
105
)
Sternberg
U
Theory of the influence of the second co-ordination sphere on the chemical shift
Mol Phys
 , 
1988
, vol. 
63
 (pg. 
249
-
267
)
Sternberg
U
Möllhoff
M
Molecular mechanics with fluctuating atomic charges—a new force field with a semi-empirical charge calculation
J Mol Model
 , 
2001
, vol. 
7
 (pg. 
90
-
102
)
Takahasi
K
Ochiai
M
Horiuchi
M
Kumeta
H
Ogura
K
Ashida
M
Inagaki
F
Solution structure of the silkworm betaGRP/GNBP3 N-terminal domain reveals the mechanism for beta-1, 3-glucan-specific recognition
Proc Natl Acad Sci USA
 , 
2009
, vol. 
106
 (pg. 
11679
-
11684
)
Tunnicliffe
RB
Bolam
DN
Pell
G
Gilbert
HJ
Williamson
MP
Structure of a mannan-specific family 35 carbohydrate-binding module: evidence for significant conformational changes upon ligand binding
J Mol Biol
 , 
2005
, vol. 
347
 (pg. 
287
-
296
)
Vashist
S
Ng
DT
Misfolded proteins are sorted by a sequential checkpoint mechanism of ER quality control
J Cell Biol
 , 
2004
, vol. 
165
 (pg. 
41
-
52
)
Woods
RJ
Pathiaseril
A
Wormald
MR
Edge
CJ
Dwek
RA
The high degree of internal flexibility observed for an oligomannose oligosaccharide does not alter the overall topology of the molecule
Eur J Biochem
 , 
1998
, vol. 
258
 (pg. 
372
-
386
)
Wormald
MR
Petrescu
AJ
Pao
YL
Glithero
A
Elliott
T
Dwek
RA
Conformational studies of oligosaccharides and glycopeptides: complementarity of NMR, X-ray crystallography, and molecular modelling
Chem Rev
 , 
2002
, vol. 
102
 (pg. 
371
-
386
)
Xie
H
Gilbert
HJ
Charnock
SJ
Davies
GJ
Williamson
MP
Simpson
PJ
Raghothama
S
Fontes
CM
Dias
FM
Ferreira
LM
, et al.  . 
Clostridium thermocellum Xyn10B carbohydrate-binding module 22-2: the role of conserved amino acids in ligand binding
Biochemistry
 , 
2001
, vol. 
40
 (pg. 
9167
-
9176
)

Supplementary data