X-ray diffraction structure of a plant glycosyl hydrolase family 32 protein: fructan 1-exohydrolase IIa of Cichorium intybus


  • Maureen Verhaest,

    1. Laboratorium voor Analytische Chemie en Medicinale Fysicochemie, Faculteit Farmaceutische Wetenschappen, K.U. Leuven, E. Van Evenstraat 4, B-3000 Leuven, Belgium
    Search for more papers by this author
  • Wim Van den Ende,

    1. Laboratorium voor Moleculaire Plantenfysiologie, Departement Biologie, Faculteit Wetenschappen, K.U.Leuven, Kasteelpark Arenberg 31, B-3001 Heverlee, Belgium
    Search for more papers by this author
  • Katrien Le Roy,

    1. Laboratorium voor Moleculaire Plantenfysiologie, Departement Biologie, Faculteit Wetenschappen, K.U.Leuven, Kasteelpark Arenberg 31, B-3001 Heverlee, Belgium
    Search for more papers by this author
  • Camiel J. De Ranter,

    1. Laboratorium voor Analytische Chemie en Medicinale Fysicochemie, Faculteit Farmaceutische Wetenschappen, K.U. Leuven, E. Van Evenstraat 4, B-3000 Leuven, Belgium
    Search for more papers by this author
  • André Van Laere,

    1. Laboratorium voor Moleculaire Plantenfysiologie, Departement Biologie, Faculteit Wetenschappen, K.U.Leuven, Kasteelpark Arenberg 31, B-3001 Heverlee, Belgium
    Search for more papers by this author
  • Anja Rabijns

    Corresponding author
    1. Laboratorium voor Analytische Chemie en Medicinale Fysicochemie, Faculteit Farmaceutische Wetenschappen, K.U. Leuven, E. Van Evenstraat 4, B-3000 Leuven, Belgium
      (fax +32 16 32 34 69; e-mail anja.rabijns@pharm.kuleuven.ac.be).
    Search for more papers by this author

(fax +32 16 32 34 69; e-mail anja.rabijns@pharm.kuleuven.ac.be).


Fructan 1-exohydrolase, an enzyme involved in fructan degradation, belongs to the glycosyl hydrolase family 32. The structure of isoenzyme 1-FEH IIa from Cichorium intybus is described at a resolution of 2.35 Å. The structure consists of an N-terminal fivefold β-propeller domain connected to two C-terminal β-sheets. The putative active site is located entirely in the β-propeller domain and is formed by amino acids which are highly conserved within glycosyl hydrolase family 32. The fructan-binding site is thought to be in the cleft formed between the two domains. The 1-FEH IIa structure is compared with the structures of two homologous but functionally different enzymes: a levansucrase from Bacillus subtilis (glycosyl hydrolase family 68) and an invertase from Thermotoga maritima (glycosyl hydrolase family 32).


Fructans occur in 15% of flowering plants as well as in many bacterial species and some fungi (Hendry, 1993). They can be considered to be a cellular extension of vacuolar sucrose reserves in plants (Wiemken et al., 1995). Fructans temporarily withhold hexoses from mainstream metabolism and form a carbohydrate reserve, different from starch in their chemical nature (fructose versus glucose), cellular localization (vacuole versus plastid) and physical state (dissolved versus crystalline-precipitated) (Van Laere and Van den Ende, 2002). Moreover, fructans can stabilize cellular membranes (Hincha et al., 2000; Vereyken et al., 2003) and might enhance plant tolerance to cold and drought (Parvanova et al., 2004).

In food, fructans are considered as prebiotics as they selectively promote growth of beneficial intestinal bacteria such as bifidobacteria and lactobacilli. Moreover, they appear to promote Ca2+ uptake in the colon and may be anti-carcinogenic (Roberfroid and Delzenne, 1998). Fructans are therefore added to an increasing number of so-called functional foods and feeds. So far, mainly inulin, extracted from the roots of chicory (Cichorium intybus L.), is used for this purpose as there is a vast agricultural experience with this crop and has a satisfactory yield (Wilson et al., 2004).

Fructans are biosynthesized by transferring β (2,1) and/or β (2,6)-linked fructofuranosyl units to one (or more) of the three primary hydroxyl groups of sucrose. Most common are linear (2,1)-type fructans (inulins; principally in dicot plants) and linear or branched (2,6)-type fructans (levans or graminans; mostly in monocots and bacteria) (Ritsema and Smeekens, 2003).

Inulin synthesis requires two distinct enzymes in plants (Edelman and Jefford, 1968; Koops and Jonker, 1996; Lüscher et al., 1996; Van den Ende and Van Laere, 1996). First, the trisaccharide 1-kestose and a free glucose is formed from two sucrose molecules by sucrose:sucrose 1-fructosyltransferase (1-SST). Next, fructan:fructan 1-fructosyltransferase (1-FFT) catalyzes chain elongation by transferring a fructosyl residue from one inulin molecule to another. The wide variation in fructan structure and degree of polymerization, apparent throughout plant species as well as tissues can be attributed mainly to differences in properties of their fructosyl transferases (Hellwege et al., 2000; Vergauwen et al., 2003). In contrast, bacterial levans or inulins are biosynthesized by one single enzyme, levansucrase or inulosucrase (Van Hijum et al., 2003). These bacterial enzymes can also hydrolyze fructans and sucrose depending on the conditions.

Fructan breakdown is catalyzed by fructan hydrolases (FH) or fructanases. Depending on the linkage type attacked, 1-FH (inulinase) and 6-FH (levanase) type enzymes can be distinguished. Unlike bacteria and fungi, where both exo- and endo-type fructan hydrolases occur (Van Damme and Derycke, 1983), plants apparently contain only fructan exohydrolases (FEH) releasing terminal fructose units. The most studied plant FEHs come from chicory (C. intybus) where three different forms have been described: 1-FEH I, IIa, and IIb (Van den Ende et al., 2001). In contrast to microbial FEHs (β-fructosidases), all plant FEHs so far purified are unable to degrade sucrose (Van Laere and Van den Ende, 2002). Plant FEHs are known to fulfill various roles in fructan plants by adjusting endogenous fructan concentrations and degree of polymerization. Unexpectedly, FEHs also occur in a wide array of non-fructan plants (Van den Ende et al., 2003) further emphasizing the importance of these enzymes in plants. In these plants FEHs might have a defense-related role acting on microbial (exogenous) fructans.

Close relationships at the biochemical and molecular levels are found between fructan biosynthesizing and degrading enzymes on the one hand and invertases on the other. This strongly supports the idea that all fructan-metabolizing enzymes evolved from invertases by relatively few mutational changes (Vijn and Smeekens, 1999). Most of these enzymes do show some invertase activity at low sucrose concentrations. However, plant FEHs seem to have completely lost the ability to degrade sucrose. Probably, an ancient β-fructosidase gene that is also found in microorganisms has been duplicated in plants: one branch evolved into specific FEHs and the other into invertases, allowing independent control of sucrose and fructan concentrations.

All plant fructan enzymes essentially transfer a fructose moiety from a donor to an acceptor substrate, but each enzyme has a preferred or even a unique substrate as donor (fructan, sucrose) and/or acceptor (fructan, sucrose, or water) as indicated in Table 1. All these enzymes have very homologous amino acid sequences and are believed to be glycoproteins. They belong to glycosyl hydrolase (GH) family 32 in the carbohydrate-active enzyme database (http://afmb.cnrs-mrs.fr/CAZY), a classification based on overall amino acid sequence similarities (Henrissat, 1991; Henrissat and Davies, 1997). All enzymes of one family probably have a common three-dimensional structure. Families that are grouped in clans also show structural similarities as well. GH family 32 is combined with GH family 68 in clan GH-J. GH family 68 groups the bacterial levansucrases and invertases. The members of this GH-J clan all probably share three critical acidic residues in their active site. Most likely, they work via a ping-pong mechanism (Chambert et al., 1974; Vergauwen et al., 2003) with retention of the anomeric configuration (Alberto et al., 2004).

Table 1.  Fructosyl transferases transfer a fructose from a donor to an acceptor, each enzyme having a preferred or unique substrate for donor and acceptor
Fructosyl transferaseFructosyl donorFructosyl acceptor
1-FEH (fructan 1-exohydrolase)InulinWater
6-FEH (fructan 6-exohydrolase)LevanWater
1-SST (sucrose: sucrose 1-fructosyl transferase)SucroseSucrose
1-FFT (fructan: fructan 1-fructosyl transferase)InulinInulin/sucrose

The 3-D structures of two non-glycosylated prokaryotic enzymes from this clan have recently been resolved: a levansucrase (GH family 68) from Bacillus subtilis (Meng and Fütterer, 2003) and an invertase (GH family 32) from Thermotoga maritima (Alberto et al., 2004), which is in fact a β-fructosidase (Liebl et al., 1998). In both cases, a five-bladed propeller fold was found. Here, we describe the 3-D structure of a GH 32 enzyme from a higher plant, namely a highly specific fructan 1-exohydrolase (1-FEH IIa), a glycoprotein from C. intybus. In contrast to the multi-functional prokaryotic levansucrases and β-fructosidases, this enzyme completely lacks the ability to break down sucrose (the primary function of the ancestral enzyme). The structure provides a template for all plant members of the family GH 32 and may prove to be an important tool for understanding which molecular determinants are important for the different substrate specificities within the group.


Overall fold

The crystal structure of C. intybus 1-FEH IIa was determined by the single anomalous dispersion method at a resolution of 2.35 Å and consists of an N-terminal fivefold β-propeller domain followed by a C-terminal domain formed by two β-sheets (Figure 1a). Only two α-helices and four 310 helices are present.

Figure 1.

Three-dimensional structure of 1-FEH IIa.
(a) The overall three-dimensional structure of 1-FEH IIa with the two glycosylation sites indicated by red balls.
(b) Schematic diagram of the topology of 1-FEH IIa. β-strands are depicted by arrows, α-helices by cylinders. Asterisks represent the glycosylation sites.
(c) Superposition of 1-FEH IIa (green) with levansucrase (purple) and invertase (blue).
(d, e) Distribution of the electrostatic potential on the molecular surfaces of 1-FEH IIa (d) and invertase (e). Blue corresponds to positive potential and red to negative potential. The active site is situated in the red colored depression on the left site of each molecule. In invertase, the groove is occluded near the active site (see arrow).

The propeller domain is based on a fivefold repeat of blades (numbered I–V), each composed of four antiparallel β-strands, around a central axis (Figure 1b). Strands are labeled A, B, C, and D from the inside of the propeller outwards. A loop is inserted in two blades, creating a disruption in strand A of blade III from residue 147 to 151 and in strand D of blade IV between 254 and 265. A long loop between strands B and C of blade V is interrupted by an α-helix (from residues 298 to 306), followed by a short β-strand that is hydrogen bonded to blade I.

So far, the number of fivefold β-propeller structures reported is limited to four: (i) an invertase/β-fructosidase from T. maritima (Alberto et al., 2004; r.m.s. deviation of 1.8 Å for 380 amino acids); (ii) a levansucrase from B. subtilis (Meng and Fütterer, 2003; r.m.s. deviation of 2.0 Å for 211 amino acids); (iii) an α-l-arabinase 43A from Cellvibrio japonicus (Nurizzo et al., 2002; r.m.s. deviation of 2.1 Å for 207 amino acids); (iv) a tachylectin-2 from Tachypleus tridentatus (Beisel et al., 1999; r.m.s. deviation of 2.5 Å for 50 amino acids). The first three proteins were found in the above order in structure homology searches using DALI (Holm and Sander, 1993). Tachylectin-2 was found only at the 40th position, preceded by different neuraminidases and sialidases, confirming the predicted similarity with these sixfold β-propeller structures (glycosyl hydrolase families 33 and 34) (Pons et al., 2000). The structures of invertase/β-fructosidase and levansucrase are presented in Figure 1(c).

Most β-propeller structures are stabilized by a closure of the ring of blades by a combination of β-strands from the N and C termini in the last blade, called a ‘molecular velcro’ (Fülöp and Jones, 1999). Although different propellers, such as propyl oligopeptidase (Fülöp et al., 1998), tricorn protease (Brandstetter et al., 2001), dipeptidyl peptidase (Engel et al., 2003), phytase (Ha et al., 2000), α-l-arabinase Arb43A (Nurizzo et al., 2002), and levansucrase (Meng and Fütterer, 2003), have been reported to have no or only atypical molecular velcros, the closure of the 1-FEH IIa propeller has yet another arrangement. Here a β-strand coming out of blade V (amino acids 306–308, strand C) forms three hydrogen bonds with blade I (amino acids 18–20, strand A) (see Table 2). Similar hydrogen bonds exist in invertase (Alberto et al., 2004). A second closure is made by a β-strand coming from the second domain (amino acids 532–536) and forming four hydrogen bonds with the D β-strand of blade I at the N terminus (amino acids 63–65). Consequently, blade I has two small extra β-strands (one at the beginning and one at the end) forming a kind of double closure consisting of six β-strands in a 1 + 4 + 1 arrangement (Figure 1b). This double closure is absent in invertase, where a 1 + 4 arrangement is observed. Furthermore, like invertase (Alberto et al., 2004), 1-FEH IIa shows an extra hydrogen bond between the loop at the N-terminus and the third strand of blade V (amino acids 13 and 311), again forming an extra closing contact.

Table 2.  Hydrogen bonds in the molecular velcro
  Distance (Å)

In most β propeller proteins, the β-blades are similar in terms of sequence repetition, for example, the Asp box ([ST]xDx[GY]xx[WFY]), which is found in the structurally similar sialidases and neuraminidases (Copley et al., 2001). However, 1-FEH IIa lacks a clear and conserved motif. By comparing the 1-FEH IIa sequence with this Asp box, we see that there is a repetition of an aspartate in the loops between the C and D β-strand of each blade of 1-FEH IIa. A serine is present in the C β-strand in three of the five blades, and the D β-strands contain a tryptophan (except in blade IV which has a phenylalanine). However, when we superimpose the five blades (Figure 2), only the aspartate and serine of blades I and III, and the tryptophan in the first three blades could really be matched. So, although the amino acid sequences show some similarity, these residues are not really structurally conserved.

Figure 2.

Superposition of the five blades of the propeller.
(a) Superposition of the backbone of each of the five blades of the propeller, showing structural conservation of the different blades. Blade I in yellow, blade II in light green, blade III in dark green, blade IV in gray, and blade V in blue.
(b) The side chains of the conserved sequences are superimposed in a backbone representation. Only a few amino acids, like Trp63, Trp125, Trp184, Ser57, Ser178, Asp59 and Asp180, are structurally conserved.

The second domain starts near Leu337 and consists of two six-stranded β-sheets. The two β-sheets are composed of antiparallel β-strands and form a sandwich-like fold. Structural homology searches for this β-sheet domain by DALI found correspondence with lectins, which are sugar-binding proteins (Holm and Sander, 1993). This domain displays one disulfide bridge between Cys393 and Cys440. In contrast to the invertase results (Alberto et al., 2004), sequence analysis of this C-terminal domain revealed sequence similarity with the C-terminus in other GH 32 proteins. This domain could therefore function as a model for other GH 32 proteins. Almost all β-strands of this domain are longer in 1-FEH IIa than in invertase. The loop in 1-FEH IIa from amino acids Leu373 to Pro406, containing a small α-helix, is very small in invertase.

There is a cleft at the interface between the two domains near the pocket-shaped active site of the propeller domain (Figure 1d), the function of which is discussed below.

The active site

The structure of 1-FEH IIa shows a common fold (Figure 1c) with invertase, another member of the GH family 32 proteins (Alberto et al., 2004) and with the GH family 68 levansucrase protein (Meng and Fütterer, 2003). Through a superposition of these structures, we observe that the active sites of both GH families are located in the β-propeller domain and show extensive overlap with one another therein (Figure 3a) (Meng and Fütterer, 2003). Homologous residues of 1-FEH IIa, invertase, and levansucrase are shown in Figures 3(a) and 4. A few additional homologous amino acids near the active site are worth mentioning. The homolog of Trp19 (1-FEH IIa) and Trp14 (invertase) is an Asp83 in levansucrase. Phe46 (1-FEH IIa) differs from Trp41 in invertase and misses an equivalent in levansucrase. Ser101 in 1-FEH IIa varies from Tyr92 in invertase and Phe182 in levansucrase. 1-FEH IIa has a Gln107 similar to levansucrase (Gln190), while invertase has a Glu103. Finally, Trp292 and Trp256 are equal in 1-FEH IIa and invertase, respectively, but a Tyr429 is present in levansucrase.

Figure 3.

Stereo view of the three-dimensional structure of the active site.
(a) Superposition of the active site of 1-FEH IIa (green) with invertase (purple) and levansucrase (blue) clearly showing the similarities and variations in the catalytic region.
(b) Active site of 1-FEH IIa. The key residues of the active site and one glycerol are displayed. Bonding interactions are shown as dashed lines, while the spheres represent water molecules. The corresponding distances are given in Å.

Figure 4.

Sequence alignment of 1-FEH IIa (Cichorium intybus), invertase (Thermotoga maritima), and levansucrase (Bacillus subtilis) based on their three-dimensional structures.
Marked sequences are conserved sequence motifs within the GH 32 family. Arrows indicate the three crucial catalytic acids as well as a conserved Y which might be involved in pKa modulation of the acid–base catalyst.

The active site of 1-FEH IIa is composed of amino acids that are conserved within the GH family 32: Asp22, Asp147, and Glu201. These three acidic side chains are spaced 5.0–6.9 Å from each other (Figure 3b). Asp22 forms hydrogen bonds with NH of Asn21 (2.71 Å), OH of Ser83 (2.79 Å) and a glycerol lying in the middle of the active center (2.76, 3.10, and 3.17 Å, respectively). Glu201 forms hydrogen bonds with the hydroxyl group of Tyr274 (2.49 Å), with the NH2 of Arg146 (2.81 Å) and with a water molecule (2.87 Å). The hydroxyl group of Tyr274 also binds a water molecule (2.75 Å). Asp147 forms bonds with the glycerol in the active center (3.03 and 3.02 Å) and with the N of Cys202 (2.95 Å).


The general glycosyl hydrolase reaction mechanism involves the protonation of the glycosidic oxygen followed by a nucleophilic attack on the anomeric carbon of the sugar substrate by a carboxylate group (Koshland and Stein, 1954). The highly conserved amino acids Asp22, Asp147, and Glu201 in 1-FEH IIa are ‘members’ of the conserved regions NDPNG, FRDP, and WECPD within the GH family 32 (Figure 4). Mutation experiments at these three positions in proteins homologous to 1-FEH IIa indicate the crucial role of these residues in the catalytic mechanism for the hydrolysis of the glycosidic bond (Batista et al., 1999; Reddy and Maley, 1990, 1996). The fact that these amino acids cluster in a putative active site is consistent with these findings. By analogy with a study of Reddy and Maley (1996) on yeast invertase, the homologous amino acid Asp22 of 1-FEH IIa was identified as the nucleophile and Glu201 as the acid/base catalyst. Hence, the reaction scheme for 1-FEH IIa can be summarized as follows: an inulin-type fructan binds to the active site where its glycosidic oxygen is protonated by Glu201. Subsequently, a nucleophilic attack is performed by the carboxylate of Asp22 forming a covalent fructose-enzyme intermediate. Finally, this intermediate is hydrolyzed releasing fructose and the free enzyme.

The hydrolysis reaction, in general, results in two possible stereochemical outcomes, those associated with either the inversion or the retention of the anomeric configuration (Henrissat and Davies, 1997). Both mechanisms involve an oxocarbenium ion-like transition state. Hydrolases like 1-FEH IIa, which retain the anomeric configuration, use an enzyme-covalent intermediate via a double displacement mechanism (see Figure 3 in Reddy and Maley, 1996), whereas inverting enzymes release the products in a single step, using a nucleophile that activates a water molecule. The average distance between the acid/base catalyst and the nucleophile reflects the presence or absence of a water molecule, that being 5.5 Å for retaining enzymes and 10 Å for the inverting enzymes (Davies and Henrissat, 1995; McCarther and Withers, 1994). Here, however, we observe a distance of 6.9 Å between the proton donor and the nucleophile, a value comparable to that in levansucrase (6.82 Å for the free enzyme; Meng and Fütterer, 2003) and invertase (6.70 Å for the free enzyme; Alberto et al., 2004). The same aberrant value is seen in the inverting enzyme α-l-arabinanase of GH family 43 (5.90 Å for the free and 6.69 Å for the ligand-bound enzyme; Nurizzo et al., 2002). Hence, there is no difference in the distances between the catalytic residues in some of the retaining GH families 32 and 68 on the one hand and the inverting GH families 43 and 62 on the other.Alberto et al. (2004) propose that it is the different binding positions of the sugars in the −1 subsite, rather than the distances between the catalytic residues, are crucial in defining the catalytic mechanisms of the various enzymes (subsite nomenclature of Davies et al. 1997). As for the GH 43 enzyme α-l-arabinanase the distance between the nucleophile and the acid/base catalyst upon ligand binding did not change drastically from the ‘expected values’ for an inverting enzyme, it is to be expected that the same holds for 1-FEH IIa and the other members of the GH families 32 and 68.

By definition, as with all β-glycosyl hydrolases (White and Rose, 1997), the pKa of the acid catalyst Glu201 of 1-FEH IIa would need to change approximately two to three pH units before and after the formation of a covalent intermediate. Hence, this requires a modulation of the ionization state of the side chain of Glu201. Although it is not trivial for the carboxyl of Glu201 to act as an acid at the optimum pH of 5, an abnormally high pKa for Glu is seen in several proteins. In xylanases, the high pKa of Glu is due to electrostatic interactions with other carboxylates within the enzyme structure (Davoodi et al., 1995). The abnormally high pKa of Glu in β-glucosidase is believed to result from the hydrophobic environment of this residue (Keresztessy et al., 1994). We propose that the pKa of Glu201 in 1-FEH IIa is modulated by the proximity of Tyr274 (2.5 Å) which is strongly conserved among many GH 32 enzymes. Although mutation experiments of the homolog of Asp147 (i.e. Asp309 in levansucrase from Acetobacter diazotrophicus) did support the theory that this amino acid plays a role in catalysis (Batista et al., 1999), structural analysis does not suggest a direct function for the chemical reaction mechanism. The homologous Asp247 in the B. subtilis levansucrase–sucrose complex structure (Meng and Fütterer, 2003) can form hydrogen bonds with the fructosyl unit (C3′ and C4′ hydroxyls), but it is too far from the C2′ hydroxyl or the glycosidic oxygen to act as a catalytic residue. It is possible that this Asp stabilizes the oxocarbenium-like transition state of the anomeric carbon.

Besides sequence and structural similarities between 1-FEH IIa, invertase and levansucrase, functional differences are reflected in structural modifications, which may play a role in the stabilization of different substrates. One example is the difference between Glu234 in 1-FEH IIa and Arg360 in levansucrase. The latter could be important in substrate recognition. Mutation studies of the Arg360 in levansucrase in B. subtilis, as well as the homologous His296 in Zymomonas mobilis levansucrase, showed their importance in fructan polymerization (Chambert and Petit-Glatron, 1991; Yanase et al., 2002). Mutation to Lys, Ser, or Leu reduced the enzyme's capability of forming the trisaccharide kestose. Arg360 in levansucrase is oriented toward the active site, whereas Glu234 of 1-FEH IIa is twisted over 180° to the opposite direction. Arg360 forms the binding site for the (n-1) or (n-2) residue of an n-meric levan acceptor (Meng and Fütterer, 2003). Most likely, the different orientation (Glu234 in 1-FEH IIa) or the complete absence (invertase, see Figure 3a) of an Arg360 equivalent makes binding of an acceptor substrate impossible. These amino acids may be crucial for differentiation between polymerizing and degrading enzymes, as the latter have only a small water molecule as an acceptor. Another structural difference in the active site is the Cys202 in 1-FEH IIa (Cys191 in invertase) and the corresponding Arg343 in levansucrase. Mutations in yeast invertase of the homologous Cys to Ala show a 70% decrease in enzyme activity, whereas the Glu201Ala mutation, in contrast, produces a 3000-fold reduction in Kcat(Reddy and Maley, 1996). Thus Cys202 plays an ancillary role in the catalytic process, for example, by maintaining a suitable microenvironment in the active site or perhaps in substrate binding. 1-FEH IIa Thr199 lacks the acidic side chain of the invertase (Glu188) and levansucrase (Glu340), respectively. Moreover, the side chain of the levansucrase Glu340 is slightly oriented differently. Trp82 (1-FEH IIa), Trp163 (levansucrase), and Phe74 (invertase) are similarly located, although the plane of the side chain is orientated differently in 1-FEH IIa. The equivalent Phe74 in invertase has the same orientation as in levansucrase.

Although structural differences near the active site in the first domain probably explain different substrate-binding specificities for two monocot fructosyltransferases (Altenbach et al., 2004), we believe that the second domain, consisting of two β-sheets, plays a major role in the recognition of fructans of different length. More precisely, the cavity between the two domains forms a cleft, emerging from the active site, which we believe to be the inulin-binding site. This idea is supported by the presence of four glycerols in this cleft. The presence of glycerol is a good indicator for a sugar-binding site. A substrate-binding cleft formed by two different domains was also found in GH family 13 (e.g. isomaltulose synthase, Zhang et al., 2003). The molecular surface (Figure 1d,e) shows a long open cleft in 1-FEH IIa. This cleft is also present in invertase, but the groove is occluded near the active site (see arrow in Figure 1e). In addition, the larger β-sheet domain and the extra loop of 1-FEH IIa compared with invertase might influence substrate binding.

Two glycosylation sites were assigned to Asn116 in the propeller domain and to Asn513 in the β-sheet domain (see red dots in Figure 1a). They are both at the surface of the protein, far from the active site and the putative sugar-binding groove. Post-translational modifications such as glycosylations do not occur in bacterial proteins.

The elucidation of the structure of this plant 1-FEH IIa enzyme is an important step toward the understanding of the molecular mechanisms of different plant GH family 32 proteins. Moreover, structural information is essential to manipulate the related fructosyltransferases (SST and FFT) by site-directed mutagenesis. In doing so, one might create plants that produce designer fructans better suited for different food and non-food applications. Structural work on an inactive Glu201Gln mutant co-crystallized with its substrate is in progress to further elucidate the mechanism of action of these enzymes.

Experimental procedures

Purification, crystallization, and data collection

Cichorium intybus 1-FEH IIa (mature protein region Q39QIEQ...VKSAA581; cDNA-based numbering) was heterologously expressed in Pichia pastoris and purified from the yeast's supernatant. It was crystallized by the hanging drop vapor diffusion method as previously described (Verhaest et al., 2004). A 2.35-Å resolution X-ray diffraction data set of the native crystals was collected at the X11 beam line of the DESY synchrotron (Hamburg, Germany) with a MAR CCD 165 detector at 100 K. Crystals belong to space group P41212. This data set was processed using DENZO and SCALEPACK (Otwinowski and Minor, 1997). Data statistics are summarized in Table 3.

Table 3.  Data collection and reduction statistics
 1-FEH IIa (native)1-FEH IIa (Hg derivative)
  1. Values in parentheses indicate data in the highest resolution shell.

Space groupP41212P41212
Unit-cell parameters (Å)
Wavelength used (Å)0.811000.934
Resolution limit (Å)2.35 (2.39–2.35)3.29 (3.47–3.29)
Total observations362 168354 539 (48 233)
Unique observations75 024 (3698)28 280 (3876)
Redundancy4.8312.5 (12.4)
Completeness (%)99.1 (98.7)99.2
Mean I/σ12.62 (2.23)7.3 (4.0)
Rsym (%)7.5 (39.3)9.2 (18.5)

Structure determination

The structure was determined by the single anomalous dispersion (SAD) method. C7H5HgO3Na was used to prepare the derivative by adding 20 μl of crystallization solution containing 0.5 mm C7H5HgO3Na to the hanging drop. The crystals were soaked for 28 h. Heavy atom-derivative crystals were cryoprotected as before (Verhaest et al., 2004). A highly redundant SAD data set of this derivative was measured to 3.29 Å resolution at the ESRF synchrotron in Grenoble (France) at beam line ID14–1 at 100 K. mosflm and scala (Collaborative Computational Project No. 1994) were used to process this data set. The statistics are summarized in Table 3. It was possible to locate three mercury-binding sites using the program autosharp (de La Fortelle and Bricogne, 1997). Subsequent phasing made it possible to calculate an electron density map. The automatic tracer programs arp/warp (Perrakis et al., 1999) and maid (Levitt, 2001) were used to trace the initial 1-FEH IIa model. Structure refinement was made by the CNS program (version 1.1) (Brünger et al., 1998). The first refinement cycle was made with 269 of 543 amino acids. This gave an Rwork of 47.93% and an Rfree of 49.53%. After several refinement cycles with intermittent manual rebuilding in O (Jones et al., 1991), we obtained the final values of Rwork of 18.29% and Rfree of 20.02%. The refinement statistics are summarized in Table 4. The final 1-FEH IIa structure consists of 537 amino acids (the first and the last five residues were omitted because of a lack of electron density), 331 water molecules, four glycerols, four N-acetyl-glucosamines and one mannose.

Table 4.  Refinement statistics
Reflections (working/test)66 415/3576
Total number of non-hydrogen atoms4709
 Protein atoms4274
 Water molecules331
 Glycerol molecules4
 N-acetyl-glucosamine molecules4
 Mannose molecules1
Rwork (%)18.29
Rfree (%)20.02
R.m.s.d. bond lengths (Å)0.0058
R.m.s.d. bond angles (°)1.417
R.m.s.d. B-factors, bonded main chain (Å2)1.154
R.m.s.d. B-factors, bonded side chain (Å2)2.159
Average B-factors, protein atoms (Å2)31.45
Average B-factors, other entities (Å2)41.00

Ramachandran statistics (procheck) (Laskowski et al., 1993) for 1-FEH IIa showed that 85.1% of the residues reside in the most favored region, 14.3% in the additionally allowed regions, and 0.4% in the generously allowed regions. One residue (His478) was found in a disallowed region of the Ramachandran plot, but this unusual main chain conformation was confirmed from the electron density map. Two glycosylation sites, as assigned by Q-TOF mass analysis (Van den Ende et al., 2001), are confirmed on residues Asn116 and Asn513. The electron density map clearly shows two succeeding N-acetyl-glucosamines at these two positions. On residue Asn513, an extra mannose was detected.

Figures were prepared with molscript (Kraulis, 1991) and raster3d (Merritt and Bacon, 1997) (Figure 1a,c), with visual molecular dynamics (Humphrey et al., 1996) and povray (http://www.povray.org) (Figures 2 and 3), or with grasp (Nicholls et al., 1991) (Figure 1d,e).

Atomic coordinates and structure factor files have been deposited in the Protein Data Bank (accession code 1ST8).


AR and WVdE are Postdoctoral Research Fellows of the Fund for Scientific Research-Flanders (Belgium) (FWO-Vlaanderen). We thank the beam line scientists at EMBL/DESY for technical support and the European Community for their support through the Access to Research Infrastructure Action of the Improving Human Potential Programme to the EMBL Hamburg Outstation, contract number HPRI-CT-1999–00017. We acknowledge the European Synchrotron Radiation Facility and the EMBL Grenoble Outstation for providing support for measurements at the ESRF under the European Union ‘Improving Human Potential Programme’.