The chitinolytic system of Lactococcus lactis ssp. lactis comprises a nonprocessive chitinase and a chitin-binding protein that promotes the degradation of α- and β-chitin

Authors


G. Vaaje-Kolstad, Department of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences, PO Box 5003, 1432 Ås, Norway
Fax: +47 64965901
Tel: +47 64965905
E-mail: gustav.vaaje-kolstad@umb.no

Abstract

It has recently been shown that the Gram-negative bacterium Serratia marcescens produces an accessory nonhydrolytic chitin-binding protein that acts in synergy with chitinases. This provided the first example of the production of dedicated helper proteins for the turnover of recalcitrant polysaccharides. Chitin-binding proteins belong to family 33 of the carbohydrate-binding modules, and genes putatively encoding these proteins occur in many microorganisms. To obtain an impression of the functional conservation of these proteins, we studied the chitinolytic system of the Gram-positive Lactococcus lactis ssp. lactis IL1403. The genome of this lactic acid bacterium harbours a simple chitinolytic machinery, consisting of one family 18 chitinase (named LlChi18A), one family 33 chitin-binding protein (named LlCBP33A) and one family 20 N-acetylhexosaminidase. We cloned, overexpressed and characterized LlChi18A and LlCBP33A. Sequence alignments and structural modelling indicated that LlChi18A has a shallow substrate-binding groove characteristic of nonprocessive endochitinases. Enzymology showed that LlChi18A was able to hydrolyse both chitin oligomers and artificial substrates, with no sign of processivity. Although the chitin-binding protein from S. marcescens only bound to β-chitin, LlCBP33A was found to bind to both α- and β-chitin. LlCBP33A increased the hydrolytic efficiency of LlChi18A to both α- and β-chitin. These results show the general importance of chitin-binding proteins in chitin turnover, and provide the first example of a family 33 chitin-binding protein that increases chitinase efficiency towards α-chitin.

Abbreviations
CBM

carbohydrate-binding module

CBP

chitin-binding protein

FnIII

Fibronectin-III

LAB

lactic acid bacterium

4MU-(GlcNAc)3

4-methylumbelliferyl-β-d-N,N′,N′′-diacetylchitobioside

TEV

tobacco etch virus

Chitin is a widespread biopolymer composed of β(1,4)-linked N-acetylglucosamine that provides structural and chemical resistance in the exoskeleton of crustaceans and arthropods, as well as in the cell wall of fungi. Chitin exists almost exclusively in an insoluble crystalline form that complexes with proteins and/or minerals to form a robust composite material. Three naturally occurring crystalline polymorphs have been described in the literature: the dominant polymorph α-chitin (antiparallel packing of the chitin chains); β-chitin (parallel packing of the chitin chains); and the minor polymorph γ-chitin (mixture of parallel and antiparallel chain packing) [1,2]. In nature, chitin is only exceeded in abundance by the structural biopolymers of plants (cellulose and hemicellulose) and is an important source of energy for a variety of organisms.

The primary degraders of chitin are microorganisms that secrete one or several chitin-degrading enzymes (chitinases). On the basis of sequence and structure, chitinases are classified into two distinct families (18 and 19) of glycoside hydrolases [3,4]. Recently, a complete survey of Trichoderma chitinases suggested a further classification of family 18 chitinases into subgroups A (bacterial/fungal), B (plant/fungal) and C (killer toxin-like chitinases) [5]. Family 18 chitinases are represented in most living organisms, whereas family 19 enzymes are mostly found in plants, where they contribute to defence against chitinous pathogens.

As a result of the recalcitrance of chitinous matrices, microorganisms have devised a variety of complementary strategies to gain access to and degrade individual polymer chains. First, the chains are degraded by both endochitinases, that attack the chitin chain randomly, and exochitinases, that attack the chitin chains from either the reducing or nonreducing end [6,7]. As endo-acting enzymes increase substrate availability for exo-acting enzymes, synergistic effects are observed [8–10]. Second, some chitinases act processively, that is, they remain associated with one and the same polymer chain whilst cleaving off consecutive dimers (also called ‘multiple attack’ mechanism [11]). Processivity is considered to be beneficial when degrading crystalline substrates, because it prevents detached individual polymer chains from re-associating with insoluble material [12,13]. Furthermore, the majority of chitinases targeting crystalline chitin are equipped with additional chitin-binding domains [also called modules or carbohydrate-binding modules (CBMs)] that are thought to increase the affinity of the enzyme for the insoluble substrate [14–16]. In addition to the enzyme machinery that decomposes the polymers, chitin-degrading microorganisms produce an N-acetylhexosaminidase (chitobiase) that converts chitobiose to N-acetylglucosamine.

Recently, an additional strategy for chitin degradation was identified, which involves the secretion of a nonhydrolytic chitin-binding protein (CBP) that acts synergistically with chitinases, presumably by increasing substrate accessibility [10,17]. These nonhydrolytic proteins are classified as family 33 CBMs [3,18], but, with one exception [19], they occur as individual proteins rather than as auxiliary domains in hydrolytic enzymes. Genome analyses indicate that secreted family 33 CBPs are produced by most chitin-degrading microorganisms [17], but only a few have been characterized biochemically. Binding studies of family 33 CBPs have been conducted for CBP21 from Serratia marcescens [17,20], ChbB [21] and Chb3 [22] from Streptomyces coelicolor, CHB1 from Streptomyces olivaceoviridis [23], CHB2 from Streptomyces reticuli [24], CbpD from Pseudomonas aeruginosa [25] and proteins E7 and E8 from Thermobifdia fusca [26], showing a large diversity of binding preferences. The function of family 33 CBPs was first demonstrated for CBP21 from S. marcescens [10], and a second example has been described recently in a study on carbohydrate-binding proteins and domains from T. fusca [26].

Genes encoding family 33 CBPs occur even in bacteria containing otherwise seemingly simple chitinolytic machineries, such as in the lactic acid bacterium (LAB) Lactococcus lactis ssp lactis IL1403. LABs are Gram-positive, facultatively, anaerobic, fermentative bacteria that are of major importance in the food industry for the generation of fermented products. In general, there is not much known about the ability of LABs to degrade chitin, but one study has shown that L. lactis is able to grow on a minimal medium containing N-acetylglucosamine oligomers as the sole carbon source [27]. According to the CAZy database [3], only a few of the sequenced LAB genomes contain genes that together encode a complete chitinolytic machinery. The genome sequence of L. lactis [28] shows three genes potentially involved in chitin turnover, coding for the following: a secreted family 18 chitinase (gene name chiA; protein referred to as LlChi18A); a secreted family 33 CBP (yucG; protein referred to as LlCBP33A); and a family 20 N-acetylhexosaminidase (LnbA). The chiA and yucG genes are separated by 19 bp in an operon starting with a putative transcriptional regulator positioned 166 bp upstream from the chitinase start codon. In this study, we have followed a biochemical approach to the question of whether L. lactis contains a functional chitinolytic machinery. The genes encoding LlChi18A and LlCBP33A were cloned and the gene products were characterized. In addition to yielding insight into the chitinolytic potential of L. lactis, the present results provide only the third example of the role of family 33 CBPs in the degradation of recalcitrant polysaccharides. Furthermore, the results provide the first example of a family 33 CBP that promotes the degradation of α-chitin, the most abundant chitin form in nature.

Results and Discussion

Preliminary assessment of the production of chitinases by L. lactis

Apart from one study showing that L. lactis can grow on chito-oligosaccharides [27], nothing is known about the ability of LABs to metabolize chitin. We attempted to culture L. lactis IL1403 on minimal medium containing various chitin forms as the sole carbon source. The chitin-containing media (sterilized by autoclaving) were inoculated with cells from an overnight culture that had been washed in sterile 0.9% saline buffer in order to remove traces of glucose. Under these conditions, the bacterium did not grow, and we could not detect chitinolytic activity in the culture supernatants even after several days of incubation.

Most microorganisms secrete a variety of hydrolytic enzymes when starved, in order to access new sources of carbon. In order to further analyse whether L. lactis would look for chitin as an alternative source of carbon, the bacterium was grown in a medium containing only 0.1% (w/v) glucose. During the growth period and starvation period, culture samples were taken and assayed for chitinolytic activity. Chitinolytic activity was detected, peaking 7 h after inoculation (Fig. 1). After 7 h, chitinolytic activity declined, but still remained significant. We could not detect chitinolytic activity in uninoculated culture medium or in cultures grown with normal glucose concentrations.

Figure 1.

 Chitinolytic activity produced by cultured L. lactis. Bar chart of chitinolytic activity measured in the culture supernatant of a starved L. lactis culture at specific time points. The bar labelled as ‘LM17’ indicates the chitinolytic activity present in fresh culture medium. Activity was recorded by measuring the hydrolysis of the fluorogenic substrate 4MU-(GlcNAc)3. All experiments were run in triplicate.

Cloning and purification of LlChi18A and LlCBP33A

The gene fragments coding for the mature proteins of LlChi18A and LlCBP33A were successfully cloned into the pETM11 and pET30 Xa/LIC expression vectors, respectively.

When expressed in Escherichia coli BL21 DE3, both proteins were produced in large amounts, although partly (LlChi18A) or almost exclusively (LlCBP33A) in an insoluble form (inclusion bodies). The culture conditions (temperature, isopropyl thio-β-d-galactoside concentrations and duration of culture) were varied in an attempt to obtain soluble protein. For LlChi18A, this resulted in the production of sufficient amounts of soluble protein. Soluble LlCBP33A was obtained through refolding of protein obtained from the inclusion bodies. After testing several denaturation and refolding protocols, we adopted a protocol based on denaturation in 8 m urea, pH 8.0 for 3 h and refolding through dialysis of concentrated denatured protein in a large volume of 20 mm Tris/HCl, pH 8.0 (see Materials and methods for more details). At most, the purification scheme resulted in 10 mg of purified LlChi18A and 7.1 mg of purified LlCBP33A per litre of culture. After purification, His-tags were removed from LlChi18A and LlCBP33A with tobacco etch virus (TEV) protease and factor Xa, respectively, with no significant loss of cleaved protein. The purity of the recombinant proteins after His-tag removal was assessed by SDS-PAGE to be better than 95%.

Sequence analysis and modelling of LlChi18A and LlCBP33A

The closest homologue of LlChi18A (when performing a standard blast search with the LlChi18A sequence) is ChiC1 from S. marcescens (49% sequence identity when aligning full-length sequences, 78.5% when aligning catalytic domains only). Like ChiC1 from S. marcescens, LlChi18A is predicted to be a three-domain protein consisting of a catalytic domain belonging to glycoside hydrolase family 18 subgroup A, according to the classification of family 18 chitinases suggested by Seidl et al. [5], followed by a Fibronectin-III (FnIII) module and a family 5 CBM [3,18], respectively. ChiC1 has the same domain structure, but the FnIII domain is followed by a family 12 CBM, which is distantly related to the family 5 CBM found in LlChi18A. Sequence analysis also shows that the catalytic module lacks an α + β-fold insertion between β-sheets 7 and 8 of the TIM-barrel fold (Fig. 2A), which is responsible for deepening the substrate-binding groove in many family 18 chitinases [29]. A deep substrate-binding groove is considered to be characteristic of enzymes that act in an exo-fashion and/or that tend to stick tightly to the substrate whilst degrading it in a processive manner [30,31]. Enzymes lacking the α + β-fold insertion have a shallow catalytic cleft, as illustrated by the crystal structure of the plant family 18 subgroup B chitinase hevamine [32]. Such shallow catalytic clefts are typically seen amongst endo-acting, nonprocessive carbohydrate-degrading enzymes. Detailed studies using chitosan as substrate have shown that ChiC1 from S. marcescens is indeed a nonprocessive endo-acting enzyme [30,33]. A model of LlChi18A automatically generated by 3d-jigsaw [34] using the structure of hevamine (Protein Data Bank code: 2HVM) as template suggested that the two proteins indeed have similar shallow and open substrate-binding clefts (results not shown).

Figure 2.

 Sequence alignments for LlChi18A and LlCBP33A. (A) Catalytic domains of LlChi18A (chitinase of L. lactis ssp. lactis), ChiC1 (chitinase C from S. marcescens BJL200), Heva (hevamine from Hevea brasiliensis), ChiA (chitinase A from S. marcescens BJL200) and ChiB (chitinase B from S. marcescens BJL200). The ChiC1 and LlChi18A sequences are aligned with a previously generated structural alignment of ChiA, ChiB and hevamine (see [49]). Conserved residues are shaded black. The stretches of residues constituting the α + β domain present in ChiA and ChiB, but lacking in LlChi18A, ChiC1 and hevamine, are shaded grey. Asterisks mark residues that are identical in LlChi18A and ChiC1. Small insertions in the hevamine sequence have been replaced by the letter ‘X’. Diagnostic sequence motifs containing residues that are crucial for catalysis (SXGG and DXXDXDXE) are shown below the alignment. Arrows indicate Ala126, replacing S in the SXGG motif, as well as two other residues, Tyr48 and Asn230, that presumably play a major role in catalysis (see text). (B) Full-length sequences of LlCBP33A (family 33 CBP of L. lactis ssp. lactis), ChbB (family 33 CBP from B. amyloliquefaciens) and CBP21 (family 33 CBP from S. marcescens). Fully conserved residues are shaded in black. Asterisks indicate residues that are thought to be located in the binding surface for chitin (as derived from the crystal structure of CBP21, as well as mutagenesis studies [10,17]). Residues involved in the chitin-binding and functional properties of CBP21 [10,17], but not conserved in LlCBP33A or ChbB, are shaded grey. The arrow indicates the terminal amino acid of the N-terminal signal sequence for all three proteins. The putatively surface-exposed aromatic amino acids in the first LlCBP33A insert are indicated by (; Trp51) and (; Phe55).

As shown in Fig. 2A, LlChi18A contains all residues known to be important for catalysis in family 18 chitinases, except for the serine in the diagnostic SXGG sequence motif, which is replaced by alanine (residue 126 in LlChi18A). The role of this serine in the catalytic mechanism of family 18 glycosyl hydrolases is to help in the stabilization of a temporary surplus of negative charge that develops on the first aspartate of the catalytic sequence motif DXDXE during catalysis [7,35]. For ChiB from S. marcescens, it was shown that this charge stabilization is in fact achieved by two residues: serine in the SXGG motif and a tyrosine residue. Although LlChi18A lacks serine, it does contain this tyrosine residue (Tyr48, corresponding to Tyr10 in S. marcescens ChiB). A multiple sequence alignment of the 50 family 18 catalytic modules that are most similar to the LlChi18A catalytic module (not shown) showed that about one-half of the proteins had a substitution at either the conserved serine or tyrosine, whereas none had substitutions at both positions. Thus, it appears that family 18 glycosyl hydrolases are tolerant to substitutions of either of the discussed amino acids, as long as both are not substituted.

Another conspicuous sequence characteristic of LlChi18A is the presence of an asparagine residue at position 230. The presence of an asparagine at this position is characteristic for family 18 chitinases with acidic pH optima for activity, whereas enzymes with more neutral pH optima have an aspartic acid at this position. For the latter type of enzyme, it has been shown that mutation of aspartic acid to asparagine leads to a drastic acidic shift of the pH optimum [35]. Indeed, LlChi18A was found to have an acidic pH optimum for activity (see below).

LlCBP33A is a family 33 CBP. The only available three-dimensional structure of a family 33 CBP is that of CBP21 from S. marcescens, which binds exclusively to β-chitin [17,20]. The combination of sequence and structural information with the results of site-directed mutagenesis studies showed that the surface of family 33 CBPs contains a patch of highly conserved, mostly polar residues that are important for binding to chitin and for the positive effect on chitinase efficiency [10,17] (Figs 2 and 3). Comparison of the LlCBP33A and CBP21 sequences shows two substitutions in the conserved surface patch, both concerning residues that are known to be important for CBP21 functionality [10]: (a) Ser63 occurs at a position at which CBP21 has a tyrosine (Tyr54) and where several other family 33 CBPs have another aromatic residue, tryptophan (e.g. Trp57 in CHB1 from St. olivaceoviridis, which has been shown to be important for the ability of CHB1 to bind α-chitin [36]); (b) Asn64 occurs instead of a glutamate residue (Glu55 in CBP21). Interestingly, the closest homologue of LlCBP33A from species other than L. lactis is ChbB from Bacillus amyloliquefaciens (66% sequence identity), which binds both α- and β-chitin [21]. As shown in Fig. 2B, ChbB differs from CBP21 in the same two positions as LlCBP33A: Tyr54 is replaced by Asp62 and Glu55 is replaced by Asn63. In addition to these sequence differences, LlCBP33A and ChbB differ from CBP21 in that they have two short inserts (Figs 2B and 3). Although it is not possible to model the structural position of these inserts accurately, it is clear that they are located close to the binding surface and may thus affect functionality (Fig. 3B). The possible implications of the observed differences within family 33 CBPs are discussed further in the context of the experimental results (see below).

Figure 3.

 Structural comparison of CBP21 and LlCBP33A. Illustrations of the CBP21 structure (A) and a structural model of LlCBP33A (B) shown in a surface representation. The surface thought to be involved in chitin binding is coloured blue. The side-chains of residues marked with an asterisk in the sequence alignment of Fig. 2B are shown as blue sticks. Residues important for chitin binding and the function of CBP21 [10,17], but not conserved in LlCBP33A, are shown as blue sticks and labelled. For illustration purposes only, the figure also shows the small inserts in LlCBP33A (orange) as they were rendered by the structure prediction program. Note that, as no template structure residues are available for modelling the inserts, the structural prediction of these inserts is highly inaccurate. Phe55 is coloured magenta and its side-chain is shown. Trp (Trp51) in the LlCBP33A insert is hidden from view. The model of LlCBP33A was generated by SwissModel (http://swissmodel.expasy.org//SWISS-MODEL.html; [50]), using CBP21 (Protein Data Bank code: 2BEM) as structural template. The model of LlCBP33A is deposited in the PMDB database (PMDB code: PM0075054).

Enzyme pH optimum, stability and kinetics

Activity measurements with the artificial substrate 4-methylumbelliferyl-β-d-N,N′,N′-diacetylchitobioside [4MU-(GlcNAc)3] showed that LlChi18A has a narrow pH activity profile with an optimum at pH 3.8 (Fig. 4A). Studies on pH stability showed that the enzyme was unstable at pH 3.8 and below, whereas enzyme activity remained stable for more than a week at bench temperature when dissolved in buffers with a pH higher or equal to pH 5 (results not shown). At shorter incubation times (e.g. up to the 20 min used in the enzyme assays), LlChi18A was stable at pH values as low as pH 3.4. Thus, kinetic parameters could be determined with confidence at the pH optimum.

Figure 4.

 Enzymatic properties of LlChi18A. (A) Relative specific activities of LlChi18A measured at pH values of 3.4, 3.8, 4.0, 4.2, 4.6, 5.0, 6.0, 7.0 and 8.0 using 4MU-(GlcNAc)3 as substrate at 37 °C. (B) Kinetics of LlChi18A towards 4MU-(GlcNAc)3 at pH 3.8 and 37 °C. The data were fitted to the Michaelis–Menten equation by nonlinear regression (represented by the curve drawn). The kinetic parameters kcat and Km derived from the data are shown in the figure. (C) Time course of the degradation of (GlcNAc)3 (◆) and (GlcNAc)4 (bsl00001) by LlChi18A, illustrated by the production of (GlcNAc)2 during the initial linear phase of the degradation reaction. Note that the enzyme concentrations used in the two reactions differed by a factor of 10 (see Materials and methods). (D) Chromatogram of (GlcNAc)6 degradation products generated by LlChi18 after 2 min of incubation with 1 nm of enzyme. The double peaks represent the α- and β-anomers of the oligomers. Using standard curves, the total concentrations of dimer, trimer and tetramer were calculated to be 25, 10 and 24 μm, respectively. The peak marked ‘X’ represents a nonhydrolysable background oligosaccharide that is also seen (with equal peak area) in control samples without enzyme. GlcNAc was not observed before all (GlcNAc)6 was degraded. Although the experiments in (D) were not conducted to preserve anomeric ratios generated by the enzyme, one important trend is still visible: the combination of a relative predominance of β-anomers for the (GlcNAc)2 product and the approximately equilibrium anomeric ratio for the tetrameric product suggests that the conversion of (GlcNAc)6 to (GlcNAc)2 and (GlcNAc)4 primarily results from binding of the nonreducing end of the substrate in subsite −2.

Both artificial substrates [4-methylumbelliferyl N-diacetyl-β-d-glucosaminide (4MU-(GlcNAc)2) and 4MU-(GlcNAc)3] were used to determine the enzyme kinetics of LlChi18A. Degradation of 4MU-(GlcNAc)2 gave sigmoidal kinetics that proved difficult to interpret (results not shown). 4MU-(GlcNAc)3, however, gave a regular hyperbolic curve that could be fitted to the Michaelis–Menten equation using nonlinear regression (Fig. 4B). The curve fitting showed LlChi18A to have a turnover rate (kcat) of 2.8 ± 0.2 s−1 and a Km value of 94 ± 10 μm. These are typical values for family 18 chitinases with shallow substrate-binding clefts [37–39]. Processive chitinases with their characteristic deep substrate-binding grooves usually have about 10-fold higher kcat and 10-fold lower Km values for oligomeric substrates [39].

Initial rate measurements with (GlcNAc)3 and (GlcNAc)4 as substrates yielded specific activities of 0.64 and 11.6 s−1, respectively (Fig. 4C), within the range of other results reported in the literature (e.g. ChiC1 from S. marcescens [30]). The products observed for (GlcNAc)3 degradation were GlcNAc and (GlcNAc)2. (GlcNAc)4 degradation resulted in the exclusive formation of (GlcNAc)2, indicating preference for binding subsites −2 to +2. Analysis of the initial degradation products formed from (GlcNAc)6 showed a 1 : 1 ratio of (GlcNAc)2 to (GlcNAc)4, which indicates a nonprocessive mode of action (Fig. 4D). Processive chitinases tend to convert (GlcNAc)6 processively into three (GlcNAc)2 moieties, leading to a characteristic initial (GlcNAc)2/(GlcNAc)4 product ratio that is considerably larger than unity (see, for example [40]). The product profile obtained with (GlcNAc)6 further shows that approximately 30% of (GlcNAc)6 is converted into two (GlcNAc)3 molecules. The data suggest that conversion of (GlcNAc)6 to (GlcNAc)2 and (GlcNAc)4 predominantly results from binding of the substrate with its nonreducing end in subsite −2 (see legend to Fig. 4), meaning that the longer part of the substrate interacts with + subsites. In a detailed analysis of product profiles [39], a similar conclusion was drawn for ChiC1 from S. marcescens. The fact that the longer part of the substrate extends towards the + side of the catalytic centre is compatible with the notion that the C-terminal substrate-binding domains are likely to be located on this side, which again suggests that this side of the enzyme is optimized for interacting with the longer (polymeric) part of the substrate. In conclusion, these experimental data and the inferences made from the sequence and structural comparisons above indicate that LlChi18A is a nonprocessive endo-acting chitinase, with overall properties that are quite similar to those of, for example, the nonprocessive endochitinase ChiC1 from S. marcescens.

Binding preferences for LlCBP33A

Some family 33 CBPs bind to a broad selection of insoluble carbohydrates (e.g. ChbB, which binds both α- and β-chitin [21], and Chb3 from St. coelicolor, which binds α-chitin, β-chitin, colloidal chitin and chitosan [22]), whereas others bind only to a specific substrate variant (e.g. CBP21 from S. marcescens which strictly binds to β-chitin [20] and CHB1 from St. olivaceoviridis [23] and CHB2 from St. reticuli [24] which strictly bind to α-chitin). A common property is that binding is influenced by pH (e.g. CBP21 from S. marcescens does not bind at pH < 4.5 [20]).

The binding preferences of LlCBP33A were investigated by incubating the protein with various types of chitin and other insoluble polymeric substrates. As noncrystalline/amorphous chitin variants, chitin beads (re-acetylated chitosan beads) and colloidal chitin (chitin processed with strong acid to disrupt the ordered crystalline properties of native chitin to render it amorphous) were used. Preliminary experiments showed that binding of LlCBP33A to chitin was relatively slow and that approximately 24 h of incubation at room temperature were needed to reach binding equilibrium. The extent and specificity of LlCBP33A binding was analysed by SDS-PAGE (Fig. 5A,B). The amount of LlCBP33A bound was also analysed by determining the protein concentrations in the supernatants of the reaction mixtures after 24 h of incubation. The results (Fig. 5C) show that LlCBP33A binds equally well to α- and β-chitin (∼ 40% of the protein in solution was bound at equilibrium), whereas binding to chitin beads (noncrystalline chitin, chitin beads; no binding detected) and colloidal chitin (amorphous chitin; ∼ 10% bound) was lower. As no or low binding was observed for the amorphous/noncrystalline chitin variants, it seems that LlCBP33A has a preference for binding the ordered, crystalline chitin forms rather than individual chitin chains. Interestingly, LlCBP33A also showed some binding to Avicel (microcrystalline cellulose, ∼ 20% bound), as has also been observed for other family 33 CBMs [21,41].

Figure 5.

 Substrate preferences for LlCBP33A at pH 6.0. (A, B) Binding of LlCBP33A visualized by SDS-PAGE. (A) LlCBP33A present in the supernatant after 24 h of incubation with α-chitin (lane 2), β-chitin (lane 3), Avicel (lane 4), chitin beads (lane 5) and colloidal chitin (lane 6). Lane 1 shows the control incubation (0.4 mg·mL−1LlCBP33A incubated for 24 h in 50 mm citrate–phosphate buffer, pH 6.0). (B) LlCBP33A bound to α-chitin (lane 2), β-chitin (lane 3), Avicel (lane 4), chitin beads (lane 5) and colloidal chitin (lane 6). Lane 1 shows controls (LlCBP33A bound to the sample tube wall). The proteins were removed from the solid substrates by boiling in SDS-PAGE sample buffer after the substrates had been washed to remove nonspecifically bound protein. Note that the samples in (B) are approximately sixfold concentrated compared with the corresponding samples in (A) (A shows 20 μL of a 300 μL supernatant; B shows 20 μL samples of bound protein resolubilized in 50 μL of SDS-PAGE sample buffer). (C) Bar chart quantifying the binding of LlCBP33A to a variety of insoluble substrates. Bound protein was determined indirectly by measuring the concentration of free protein in the supernatants after 24 h of incubation.

In terms of binding to the various chitin forms, the characteristics of LlCBP33A are similar to those of ChbB from B. amyloliquefaciens, in that both proteins bind well to both α- and β-chitin. As noted above, ChbB is the closest homologue of LlCBP33A and the two proteins share sequence characteristics that separate them from the ‘one-substrate binders’ such as CBP21 [17,20] and CHB1 [23]. It is conceivable that the above-mentioned two mutations in the binding surface and the two insertions that are putatively close to this surface (Fig. 3) endorse LlCBP33A and ChbB with the ability to bind a wider variety of substrates than do CBP21 and CHB1.

Degradation of α- and β-chitin

The degradation rates of α- and β-chitin were assayed with LlChi18A in the presence or absence of LlCBP33A. As both chitin variants, and especially α-chitin, are highly resistant to enzymatic hydrolysis, the time span of the assay was 2 weeks using a relatively high concentration of LlChi18A and LlCBP33A (1.0 and 3.0 μm, respectively). The production of (GlcNAc)2 (the major end-product of enzymatic hydrolysis by family 18 glycosyl hydrolases) was recorded at regular time intervals.

The degradation of α-chitin by LlChi18A started with a rapid phase, regardless of the presence of LlCBP33A. In the presence of LlCBP33A, the fast initial phase was maintained longer than in the absence of LlCBP33A, indicating that LlCBP33A acts synergistically with LlChi18A. However, the effect of LlCBP33A was small and ceased after approximately 48 h (Fig. 6A). This indicates that LlCBP33A only acts on a specific minor subfraction of α-chitin. Thus, LlCBP33A has an effect on the degradation of α-chitin, but the effect is smaller than the effects of CBP21 [10] or LlCBP33A (below) on β-chitin.

Figure 6.

 Chitin degradation by LlChi18A in the absence and presence of LlCBP33A at pH 6.0, 37 °C. (A) Full lines show the degradation of 0.5 mg·mL−1α-chitin by LlChi18A (bsl00001) and LlChi18A in the presence of LlCBP33A (•) with nonstatic incubation. (B) Full lines show the degradation of 0.1 mg·mL−1β-chitin by LlChi18A (bsl00001) and LlChi18A in the presence of LlCBP33A (•) with static incubation. For comparison, the production of the minor end-product GlcNAc is also shown (dotted lines through squares for LlChi18A; dotted lines through circles for LlChi18A in the presence of LlCBP33A). The production of GlcNAc in the reaction with α-chitin could not be quantified accurately, but was of the same order of magnitude.

The degradation of β-chitin by LlChi18A was much more rapid than the degradation of α-chitin. Moreover, although about 85% of α-chitin was left after 2 weeks of incubation, all of the β-chitin was completely solubilized by LlChi18A, in both the absence and presence of LlCBP33A. In the absence of LlCBP33A, the end-point of the reaction (i.e. solubilization of all chitin) was reached after approximately 2 weeks. When LlCBP33A was present in the reaction, the degradation rate was substantially higher, the end-point being reached after approximately 48 h (Fig. 6B). Thus, LlCBP33A clearly acts synergistically with LlChi18A in the degradation of β-chitin. The increase in LlChi18A efficiency on addition of LlCBP33A is comparable with the increase observed when adding CBP21 during the degradation of β-chitin with ChiC1 from S. marcescens [10].

Although the occurrence of family 33 CBPs has been known for some time [23], the present results provide only the third demonstration of the accessory function of these proteins. The effect of LlCBP33A on β-chitin degradation is of the same order of magnitude as the effect of CBP21. The effect on α-chitin degradation is unique for LlCBP33A, but is rather modest (Fig. 6A). It should be noted that, in nature, chitin is often found as a composite where layers/sheets of chitin are interwoven with proteins and/or minerals in a recalcitrant heteropolymer. The crystalline chitin used in most experiments in the chitin/chitinase field has been treated by strong acids and bases in order to remove the protein and/or the mineral fraction. It is conceivable that the real natural substrates of the CBP proteins differ from the substrates used here and in other studies. There may exist composite natural chitinous substrates that are more susceptible to the action of CBPs.

Structure–function studies of CBP21 have shown that this protein may act by disrupting the crystalline substrate through interactions that involve polar residues in a conserved surface patch [10,17]. The lack of aromatic residues in the binding surface of CBP21 (there is only one, Tyr54) was somewhat surprising, because aromatic residues are generally considered to play important roles in enzyme–carbohydrate interactions [18]. As CBP21 and LlCBP33A have different binding properties, a structural comparison of the two proteins could provide more insight into the mechanism and specificity of CBP action. Unfortunately, despite extensive attempts, we have so far been unable to obtain crystals of LlCBP33A. The most obvious structural difference between the two proteins is formed by the two inserts in LlCBP33A that seem positioned close to the conserved surface patch and that could extend the binding surface (Fig. 3B). Interestingly, the largest insert contains two aromatic amino acids (Trp51 and Phe55), which could interact with the surface of a chitin crystal. Another interesting observation is that a disulfide bridge on the surface, close to the important Tyr54 in CBP21 (Cys41–Cys49 in CBP21), is missing in LlCBP33A, which has the 50–57 insert in this area (Fig. 2B). This could affect the binding properties of the protein, as it may introduce flexibility and/or structural changes in this crucial region.

Conclusions

The present data show that the putative chitinase and CBP genes in L. lactis code for a functional chitinolytic machinery capable of converting chitin to GlcNAc and (GlcNAc)2. The primary product of this machinery is (GlcNAc)2, which can be converted to mono-sugars by the putative N-acetylglucosaminidase encoded by LnbA. We were able to show that L. lactis indeed produces chitinolytic activity under certain conditions. However, further work is needed to analyse the role and regulation of the chitinolytic system of this bacterium. LlChi18A was shown to be active and relatively stable at low pH, which agrees with the ability of L. lactis to grow and thrive in mildly acidic environments.

The finding of nonhydrolytic accessory proteins for chitinases has reinforced interest in the question as to whether such proteins may also exist for cellulose. The existence of substrate-disrupting accessory proteins and domains that act synergistically with cellulases has been a topic in cellulose research ever since the studies by Reese et al. around 1950 [42]. Cases as clear cut as the two cases from the chitinase field ([10], this paper) do not yet exist. However, more recent studies indicate that nonhydrolytic proteins that are either dedicated to cellulose degradation [43,44] or that can be exploited for this purpose (expansins [45]; see also [46]) do exist.

Materials and methods

Bacterial strains and plasmids and cultivation

Lactococcus lactis ssp. lactis IL1403 is a derivative of the strain IL594 that was isolated from a cheese starter culture [47]. For the isolation of genomic DNA and the creation of stock cultures, the bacterium was grown overnight at 30 °C without aeration in M17 medium (Oxoid, Basingstoke, Hampshire, UK) supplemented with 0.2% (w/v) glucose (GM17). The bacterium was maintained as frozen stocks at −80 °C in liquid medium containing 17% (v/v) glycerol.

To investigate whether L. lactis was able to produce chitinolytic activity, overnight cultures of L. lactis grown in GM17 were diluted to an attenuance D at 600 nm of approximately 0.1 in modified M17 medium (LM17), composed of Maritex Fish peptone (5.0 g·L−1) [48], bacto yeast (Difco Laboratories, Sparks, MD, USA) (5.0 g·L−1), ascorbic acid (Sigma, St Louis, MO, USA) (0.5 g·L−1), magnesium sulfate (0.25 g·L−1) (Sigma), disodium glycerolophosphate (19 g·L−1) (Sigma) and manganese sulfate (0.05 g·L−1) (Sigma). As a carbon source, β-chitin isolated from squid pen (France Chitin, Marseille, France), α-chitin isolated from shrimp (Hov-Bio, Tromsø, Norway), colloidal chitin and glucose were used (all chitin variants at a final concentration of 1% w/v and glucose at final concentrations of 0.1% or 0.4% w/v). The cultures were incubated at 30 °C and samples were taken at various time points (4, 7, 8 and 10.5 h) in order to assay for chitinolytic activity in the culture supernatants (see below for assay details).

Cloning of L. lactis chitinases and CBP

Genomic DNA from L. lactis was isolated from an overnight culture using a midi-prep genomic DNA isolation kit (Qiagen, Venlo, The Netherlands) and stored at −20 °C. A 3392 bp long region of the genome containing a putative transcription regulator (GenBank ID: AAK06047.1), chitinase gene (GenBank ID: AAK06048.1) and gene encoding a family 33 CBP (GenBank ID: AAK06049.1) was amplified by PCR using primers flanking 100 bp upstream of the first ORF and 100 bp downstream of the third ORF (forward primer, 5′-GGATGAGCTCTATACTCACATCTTGAGC-3′; reverse primer, 5′-TTGTGGGCCCAACCAATCTATGAAGAATT-3′). The PCR product was cloned using the zero blunt TOPO-cloning kit (Invitrogen, Carlsbad, CA, USA). The resulting plasmid was transformed into E. coli TOP10 competent cells and the insert was sequenced using a series of sequencing primers evenly distributed along the cloned DNA sequence. Different strategies were used for subsequent separate cloning of the chitinase and CBP, as the N-terminus of the latter protein should be free of non-native amino acids after removal of the affinity tag (because amino acid one of the mature protein is conserved and seems important; see [17]). Both primary gene products were predicted to contain N-terminal leader peptides directing sec-dependent secretion. The genes were cloned without these leader peptide-encoding parts. The start of the mature proteins was assigned using the SignalP server (http://www.cbs.dtu.dk/services/SignalP/).

Primers for cloning of the putative chitinase were designed to amplify a fragment encoding amino acids 32–492 of the predicted gene product. The forward primer (5′-GGTCTCCCATGGATGCAGCTAGTGAAATGGTCA-3′) was designed with an NdeI-compatible BsaI restriction site at the 5′ end, leading to a one-residue (methionine) N-terminal extension of the gene product. The reverse primer (5′-CTCGAGTTATAGCTTTTTCCATGGACCAAAATCTC-3′) contained a XhoI restriction site starting immediately downstream of the stop codon of the chitinase gene. The amplified chitinase fragment was ligated into vector pCR®4Blunt-TOPO®Zero Blunt TOPO (Invitrogen), excised from the TOPO vector using XhoI and BsaI, and ligated to NdeI–XhoI-digested pETM11 vector (Günter Stier, EMBL Heidelberg, Germany). The pETM11 vector contains a T7 promoter sequence for expression and an N-terminal His6 tag for immobilized metal affinity chromatography purification.

The putative CBP was cloned using the pET30 Xa/LIC kit (Merck Chemicals Ltd, Nottingham, UK), which provides a ligation-independent method for cloning a gene of interest. The expression vector (pET30 Xa/LIC) provides an N-terminal His6 tag that can be removed from the N-terminus of the purified protein using activated factor X leaving no non-native amino acids. Cloning primers were designed according to the suppliers’ instructions, containing ends compatible with the expression vector (forward primer, 5′-GGTATTGAGGGTCGCCATGGTTATGTTCAATCACCA-3′; reverse primer, 5′-AGAGGAGAGTTAGAGCCTTACAAGAAGGGTCCAAAGA-3′). The PCR product was purified, treated with T4 exonuclease to create vector-compatible overhangs and annealed to a prepared expression vector (pET30 Xa/LIC) provided by the supplier.

The final constructs (pETM11-LlChi18A and pET30XaLIC-LlCBP33A) were transformed into E. coli BL21Star (DE3) (Invitrogen). DNA sequencing was performed using a BigDye® Terminator v3.1 Cycle Sequencing Kit (Perkin-Elmer/Applied Biosystems, Foster City, CA, USA) and an ABI PRISM® 3100 Genetic Analyser (Perkin-Elmer/Applied Biosystems).

Protein expression and purification

Overnight cultures grown from −80 °C stocks of BL21Star (DE3) cells containing pETM11-LlA or pET30XaLIC-LlA were used to inoculate 150 mL of Luria–Bertani medium containing 50 μg·mL−1 of kanamycin. The cultures were incubated at 37 °C and 250 r.p.m. When the cell density reached 0.6 (D600), isopropyl thio-β-d-galactoside was added to a final concentration of 0.05 mm, and the culture was further incubated for 4 h at 37 °C, followed by harvesting by centrifugation (11 325 g 10 min at 4 °C). The cell pellet was resuspended in citrate–phosphate buffer pH 6.0, and the cells were lysed by sonication at 20% amplitude with 30 × 5 s pulses (with 5 s delay between pulses) on ice, with a Vibra cell Ultrasonic Processor, converter model CV33, equipped with a 3 mm probe (Sonics, Newtown, CT, USA). The sonicated material was centrifuged at 11 325 g for 10 min at 4 °C in order to pellet the insoluble cell remains. At this stage, LlChi18A was found in the soluble fraction and LlCBP33A was found as inclusion bodies in the insoluble fraction. Thus, two separate protocols were followed for subsequent purification. For LlChi18A, the cleared lysate was applied to a 3 cm × 5 cm Ni-NTA column (Qiagen, Venlo, The Netherlands) equilibrated with running buffer (100 mm Tris/HCl, pH 8.0). LlChi18A was eluted by running four column volumes of elution buffer (100 mm Tris/HCl, pH 8.0 and 100 mm imidazole) through the column. The peak containing chitinase was collected and concentrated using a Centricon P-20 unit (Millipore, Billerica, MA, USA) and dialysed overnight in 20 mm Tris/HCl, pH 8.0.

For LlCBP33A, the pellet resulting from centrifugation of the sonicated cells was resuspended in denaturing buffer containing 8 m urea, 0.1 m NaH2PO4, 10 mm Tris/HCl, pH 8.0 and 25 mm dithiothreitol, and incubated at room temperature for 3 h with gentle shaking. Subsequently, the unfolded protein was purified on an Ni-NTA column under denaturing conditions, using 8 m urea, 0.1 m NaH2PO4 and 10 mm Tris/HCl, pH 8.0 as running buffer, and 8 m urea, 0.1 m NaH2PO4, 10 mm Tris/HCl, pH 8.0 and 100 mm imidazole as elution buffer. The peak containing the pure protein was concentrated using a Centricon P-20 unit (Millipore) and the protein was refolded by extensive dialysis in 20 mm Tris/HCl, pH 8.0 at 4 °C (two buffer changes in 24 h).

The removal of the N-terminal His6 tags was performed by the addition of recombinant TEV protease (1 U per 3 μg of target protein) or activated factor X (Merck Chemicals Ltd.; 1 U per 70 μg of target protein) to purified His-tagged LlChi18A and LlCBP33A, respectively. TEV protease cleavage reactions were conducted in 20 mm Tris/HCl, pH 8.0, 0.25 mm EDTA and 1 mm dithiothreitol, incubated at 37 °C for 4 h. Factor Xa cleavage reactions were conducted in 100 mm NaCl, 50 mm Tris/HCl, 5 mm CaCl2, pH 8.0, incubated at room temperature for 16 h. Cleavage reactions were followed by immobilized metal affinity chromatography purification (as above), in which the nonbound protein (cleaved LlChi18A or LlCBP33A) was collected and the bound protein (free His6 tag and His-tagged TEV protease if present) was discarded. Factor X was removed from the LlCBP33A cleavage reaction by running the sample through a mini spin column containing 50 μL of Xarrest Agarose (Merck Chemicals Ltd). Both proteins were dialysed overnight in 20 mm Tris/HCl, concentrated using Centricon P-20 units (Millipore) and stored at 4 °C.

Protein purity was analysed by SDS-PAGE. Protein concentrations were determined using the Bradford micro-assay (Bio-Rad, Hercules, CA, USA) according to the instructions provided by the supplier, employing purified bovine serum albumin (New England Biolabs, Beverly, MA, USA) as standard.

Chitin-binding assays

Binding studies were conducted using powdered α-chitin from shrimp shells (Hov-Bio), powdered β-chitin from squid pen (France Chitin), chitin beads (New England Biolabs), colloidal chitin and Avicel (microcrystalline cellulose; Sigma). All chitin variants were suspended in ddH2O to yield a 20 mg·mL−1 stock suspension. Binding was assayed in 1 mL reactions in Eppendorf tubes containing 5 mg·mL−1 chitin and 400 μg·mL−1LlCBP33A in 50 mm citrate–phosphate buffer, pH 6.0. Reactions were mixed by vertical rotation (60 r.p.m.) at room temperature for 24 h. Subsequently, the chitin (with the bound protein fraction) was pelleted by spinning the sample tubes for 5 min at 15 700 g in a microcentrifuge. The relative amount of free protein in the supernatant was determined by measuring the absorption at 280 nm (Eppendorf Biophotometer, Eppendorf, Hamburg, Germany). All assays were performed in triplicate and with blanks (buffer + 5 mg·mL−1 of the appropriate substrate) and controls to correct for aspecific binding of the protein to the reaction vessels (buffer + 400 μg·mL−1LlCBP33A; the values derived from this control sample were considered to represent 0% binding). For further verification of binding, protein bound to the substrate was analysed by SDS-PAGE after removal of nonspecifically bound protein by washing with 1.5 mL of 50 mm Tris/HCl, pH 8.0. The pellets were resuspended in 50 μL SDS-PAGE sample buffer and boiled for 5 min in order to strip bound protein of the substrate. Finally, 20 μL of sample was run on a pre-cast SDS-PAGE gel (Novex 12%; Invitrogen). Gels were run for 30 min at 200 V, stained in a solution containing 0.5% (w/v) Coomassie brilliant blue, 50% (v/v) methanol and 10% (v/v) acetic acid, and destained in a solution containing 10% (v/v) methanol and acetic acid.

Enzymology

The kinetic properties of LlChi18A were determined using the artificial substrate 4MU-(GlcNAc)3 (Sigma). The maximum substrate concentration used had to be limited to about twice Km because of the occurrence of substrate inhibition (which is usual in this type of assay; for example, see [8]). Standard reaction mixtures contained 2.0 nm of LlChi18A, 0.1 mg·mL−1 bovine serum albumin and 0–200 μm of the substrate in 50 mm citrate–phosphate buffer, pH 3.8. The reaction mixtures were incubated at 37 °C and product formation was monitored by taking out 50 μL samples at different time points (0–20 min), in which the reaction was stopped by the addition of 1.95 mL of 0.2 m Na2CO3. The amount of 4MU released was determined by measuring the fluorescence emitted at 460 nm on excitation at 380 nm, using a DyNA 200 fluorimeter (Hoefer Pharmacia Biotech, San Francisco, CA, USA). The release of 4MU proved to be linear with time for all substrate concentrations, allowing the straightforward calculation of enzyme velocities by linear regression (all curves had correlation coefficients above 0.99). Kinetic parameters were calculated by directly fitting the data to the Michaelis–Menten equation by nonlinear regression using graphpad prism (GraphPad Software Inc., San Diego, CA, USA).

The specific activity of LlChi18A towards a natural substrate was determined by monitoring initial product release during the degradation of (GlcNAc)3 and (GlcNAc)4 (Seikagaku Co., Tokyo, Japan). Reactions were performed in Eppendorf tubes containing 200 μm of oligosaccharide and 0.1 mg·mL−1 of bovine serum albumin in 50 mm citrate–phosphate buffer, pH 3.8. The reaction was initiated by the addition of LlChi18A, giving an end concentration of 15 or 1.5 nm of enzyme [for the degradation of (GlcNAc)3 and (GlcNAc)4, respectively]. Samples were taken at 0, 2, 4, 6, 8 and 10 min and mixed immediately 1 : 1 with 70% (v/v) acetonitrile to stop hydrolysis [35% (v/v) acetonitrile abolishes enzyme activity]. Samples were then analysed by isocratic HPLC employing a 0.46 × 25 cm Amide-80 column (Tosoh Bioscience, Montgomeryville, PA, USA), coupled to a Gilson Unipoint HPLC system (Gilson, Middleton, WI, USA). The liquid phase was 70% (v/v) acetonitrile and the flow rate was 0.7 mL·min−1. Twenty microlitre samples were injected using a Gilson 123 autoinjector. Eluted oligosaccharides were monitored by recording the absorption at 210 nm. Chromatograms were collected and analysed using Gilson unipoint software (Gilson). A standard solution containing 100 μm of (GlcNAc)1–4 was analysed at the start, in the middle and at the end of each series of samples. The resulting average values of the standards (displaying standard deviations of < 5%) were used for calibration. All measurements were performed in triplicate. Background was corrected for by subtracting the value of samples taken at t = 0 min.

The determination of the initial products from (GlcNAc)6 degradation was performed by incubating 1 nm of LlChi18A with 200 μm of (GlcNAc)6 in 50 mm citrate–phosphate buffer, pH 6.0. Products formed after 2 min of incubation at 37 °C were analysed using isocratic HPLC as described above.

The presence of chitinolytic activity in culture supernatants of L. lactis was assayed using 4MU-(GlcNAc)2 as substrate; 50 μL of supernatant was mixed with 50 μL of 50 mm citrate–phosphate buffer, pH 3.8, containing 50 μm of substrate and 0.1 mg·mL−1 of bovine serum albumin, giving a final volume of 100 μL. The reaction mixture was incubated at room temperature overnight as the chitinase concentration in the supernatant was low. The release of 4MU was determined as described above.

For the determination of the pH optimum, solutions of 4MU-(GlcNAc)3 (50 μm) were prepared using 50 mm citrate–phosphate buffer (pH range 3.4–7.0) and Tris/HCl at pH 8.0, containing 0.1 mg·mL−1 of bovine serum albumin. LlChi18A was added to a final concentration of 20 nm, and samples were taken at 3, 6 and 9 min to record the release of 4MU, as described above. All measurements were performed in triplicate. Product release was linear over time in all cases.

Degradation of α- and β-chitin

Determination of the enzyme activity towards insoluble chitin was performed using β-chitin from a squid pen (France Chitin) and α-chitin isolated from shrimp shells (Hov-Bio) as substrates. Reaction mixtures contained 1 μm of LlChi18A and/or 3 μm of LlCBP33A, 0.1 mg·mL−1 of purified bovine serum albumin, 0.1 mg·mL−1 of β-chitin powder or 0.5 mg·mL−1 of α-chitin powder, in 50 mm citrate–phosphate buffer, pH 6.0. The reaction was buffered at a higher pH than in the kinetic experiments as the long-term stability (incubations exceeding 1 h) of LlChi18A (in the presence of bovine serum albumin) was better at a near-neutral pH (at pH 6.0, there was no detectable loss of activity under the conditions described below; A. C. Bunæs and G. Vaaje-Kolstad, unpublished observations). Reaction mixtures were incubated at 37 °C for up to 2 weeks with vigorous shaking (α-chitin; 1300 r.p.m. in an Eppendorf Thermomixer comfort; Eppendorf) or without shaking (β-chitin). Initial experiments showed that the degradation rate of α-chitin was slow when using static incubation (results not shown); thus, to increase the amount of product formed, and thereby the reliability of the assay, vigorous shaking was applied to promote the chitin–LlChi18A and/or chitin–LlCBP33A contact. At time points ranging from 2 to 340 h, 60 μL of the reaction was taken and mixed with an equivalent amount of 70% acetonitrile in an Eppendorf tube (the presence of acetonitrile arrests all enzyme activity). All reactions were run in triplicate and all samples were stored at −20 °C until product analysis by HPLC as described above.

Acknowledgements

We thank Svein J. Horn for helpful discussions. This work was funded by the Norwegian Research Council, grants 171991 (GVK), 164653 (GVK), 159058 (GM) and 140497 (ACB, VE).

Ancillary