SEARCH

SEARCH BY CITATION

Keywords:

  • bacterial hydroxynitrile lyase;
  • cupin structure;
  • enzyme catalysis;
  • metalloenzyme;
  • site-directed mutagenesis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Hydroxynitrile lyases (HNLs), which catalyse the decomposition of cyanohydrins, are found mainly in plants. In vitro, they are able to catalyse the synthesis of enantiopure cyanohydrins, which are versatile building blocks in the chemical industry. Recently, HNLs have also been discovered in bacteria. Here, we report on the detailed biochemical and structural characterization of a hydroxynitrile lyase from Granulicella tundricola (GtHNL), which was successfully heterologously expressed in Escherichia coli. The crystal structure was solved at a crystallographic resolution of 2.5 Å and exhibits a cupin fold. As GtHNL does not show any sequence or structural similarity to any other HNL and does not contain conserved motifs typical of HNLs, cupins represent a new class of HNLs. GtHNL is metal-dependent, as confirmed by inductively coupled plasma/optical emission spectroscopy, and in the crystal structure, manganese is bound to three histidine and one glutamine residue. GtHNL displayed a specific activity of 1.74 U·mg−1 at pH 6 with (R)-mandelonitrile, and synthesized (R)-mandelonitrile with 90% enantiomeric excess at 80% conversion using 0.5 m benzaldehyde in a biphasic reaction system with methyl tertiary butyl ether.


Abbreviations
ee

enantiomeric excess

HNL

hydroxynitrile lyase

ICP-OES

inductively coupled plasma/optical emission spectroscopy

MTBE

methyl tertiary butyl ether

RMSD

root-mean-square deviation

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Hydroxynitrile lyases (HNLs) are enzymes that catalyse the reversible cleavage of cyanohydrins, yielding HCN and aldehydes or ketones. In the last two decades, the enzymes have been investigated extensively, especially regarding their potential use as catalysts in organic chemistry [1-3]. In chemical synthesis, reversal of the natural cleavage reaction is employed. Cyanohydrins constitute important building blocks for fine chemicals, as their synthesis creates a new stereocentre and adds a versatile functional group (the nitrile group) at the same time [4-6]. HNLs comprise a diverse group of enzymes that vary in terms of their substrate specificity, enantioselectivity and the need for a co-factor. This may account mainly for the fact that they belong to different protein families with no significant sequence similarity. The structures solved so far show similarity to α/β-hydrolases, oxidoreductases, carboxypeptidases or Zn2+-dependent alcohol dehydrogenases [2, 7]. Interestingly, members of the α/β-hydrolase fold show both (R)- and (S)-selectivity [8, 9]. Most HNLs have been discovered in plants, and several have been cloned successfully and expressed in Escherichia coli or yeasts such as Pichia pastoris [10, 11]. Very recently, the discovery of hydroxynitrile lyases in endophytic bacteria has been described by our group [12]. Interestingly, they are not related to any of the HNLs characterized so far, but show sequence similarity to proteins of the cupin superfamily, which comprises a large number of small β-barrel-fold proteins with diverse functionalities [13-15]. As HNL activity has not previously been described for proteins of this superfamily, cupins represent a novel class of these industrially useful enzymes.

Although cupins are structurally conserved and usually contain two conserved motifs, G-(X)5-H-X-H-(X)3,4-E-(X)6-G (motif 1) and G-(X)5-P-X-G-(X)2-H-(X)3-N (motif 2), the overall sequence identity is low among members of this superfamily [13]. The two motifs also include the residues for metal binding. Most cupins are metal-binding proteins that bind divalent cations such as iron, zinc, manganese, copper, nickel or cadmium [16]. The metal is usually involved in the enzymatic reaction either directly in the reaction mechanism or at least via an interaction with the substrate.

In this paper, we describe the biochemical and structural characterization of the manganese-dependent hydroxynitrile lyase from the acidobacterium Granulicella tundricola (GtHNL), which shows high sequence similarity to two previously described cupins with HNL activity [12] and was heterologously expressed in E. coli.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Based on sequence similarities to recently discovered bacterial hydroxynitrile lyases with a cupin fold from Pseudomonas mephitica (PsmHNL) and Burkholderia phytofirmans (BpHNL) [12], several highly similar proteins were discovered using a BLASTp search on the UniProt (Universal Protein Resource) server. Most of them originate from bacterial genome or metagenome sequencing projects and are not annotated, thus their natural function is unknown. Some of them were obtained as synthetic genes, cloned into expression vectors, successfully expressed as soluble proteins in E. coli, and tested for HNL activity. One of them, a protein originating from Granulicella tundricola with 79.1% sequence similarity to BpHNL, showed higher cyanolysis activity in preliminary screenings and was chosen for detailed biochemical and structural analyses.

Mandelonitrile cyanolysis activity of GtHNL is dependent on manganese

GtHNL was expressed at a very high yield as soluble protein in E. coli (data not shown), with yields of > 50% of total soluble protein. In initial experiments, the cyanolysis activity of GtHNL was assessed using mandelonitrile as substrate. The enzyme cleaved (R)-mandelonitrile to benzaldehyde and HCN, as indicated by a sensitive colony-based filter assay for HCN (data not shown), as well as by the increase in absorption at 280 nm attributed to released benzaldehyde measured in a spectrophotometric assay using cleared lysate (Fig. 1). The enzymatic activity is clearly distinguishable from the background reaction in the same buffer.

image

Figure 1. Release of benzaldehyde from 18 mm (R)-mandelonitrile at pH 5.5 (80 mm MES, 80 mm sodium oxalate buffer), catalysed by 50 μg manganese-reconstituted GtHNL apoprotein (black squares, solid line) at 25–27 °C, compared to the buffer background (grey crosses, dashed line).

Download figure to PowerPoint

Metal dependence

Most cupins are metal-binding proteins that bind Mn2+, Fe2+, Ni2+, Cu2+ or Zn2+ in the active site [16]. The characteristic cupin domain comprises two conserved motifs, G-(X)5-H-X-H-(X)3,4-E-(X)6-G (motif 1) and G-(X)5-P-X-G-(X)2-H-(X)3-N (motif 2), which are both involved in metal ion binding [15]. The two His residues and the Glu/Gln residue in motif 1 as well as the His residue in motif 2 act as ligands for binding of the metal ion in the active site. These conserved motifs, including the putative metal-binding amino acids, are conserved in the sequence of GtHNL, indicating that GtHNL is also metal-dependent. Preliminary experiments using a colony-based filter assay on colonies grown on LB agar supplemented with various divalent cations (Fe2+, Zn2+, Mn2+, Ni2+, Co2+ or Cu2+) showed that HNL activity increased in the presence of Mn2+ in the medium, compared to colonies without supplementation (data not shown). The results of a detailed metal analysis are described below.

pH dependence

The stability and activity of HNLs at acidic pH is especially important as the non-specific background reaction is reduced at low pH. Purified protein, which was expressed in the presence of 100 μm MnCl2, catalysed the cyanolysis of (R)-mandelonitrile in aqueous sodium oxalate buffer (which was identified as the most appropriate buffer) at pH 4.5, with a Km of 10.3 ± 2 mm and a kcat of 0.03 s−1, corresponding to an enzymatic activity of 0.12 ± 0.009 U·mg−1 at Vmax (Fig. S1). Unsurprisingly, the enzyme is more active at less acidic pH, and the activity of the protein increased to 1.74 U·mg−1 at pH 6. Protein reconstituted with manganese in vitro showed an even higher activity (> 4 U·mg−1 at pH 6.0), which may be explained by the higher manganese loading achieved in vitro. Overall, the specific activity increased ~ 40-fold from pH 4.5 to pH 6.0 (Fig. 2). The activity was not measured at higher pH, as the enzyme catalysis becomes indistinguishable from the base-catalysed chemical background reaction at pH values above 7. The pH stability of GtHNL at pH 4.0 was confirmed by circular dichroism spectroscopy (Fig. S2A). No unfolding was observed within 1 h, indicating that the overall fold is stable at low pH.

image

Figure 2. Specific activity (U·mg−1) of GtHNL (50 μg Mn2+-reconstituted apoprotein) at various pH values (80 mm MES, 80 mm sodium oxalate buffer, 18 mm (R)-mandelonitrile, 25–27 °C); the chemical background reaction at the various pH values was subtracted. *Sodium oxalate at pH 4.5 without MES.

Download figure to PowerPoint

Metal analysis

As described above, addition of manganese to the expression medium led to an increase in HNL activity. To prevent interference of the His-tag with the metal content and analysis, the protein was purified without the tag. Surprisingly, the first purification step using anion exchange chromatography resulted in almost pure protein (Fig. S3), and the purity was further improved by size-exclusion chromatography. The yield of purified protein was up to 50–60 mg per litre of culture, corresponding to ~ 20 mg of purified protein per g wet cell paste. The metal contents of purified protein samples, which were either grown in LB medium supplemented with various metals or deprived of metal using the chelating agent 2,6-pyridinedicarboxylic acid and reconstituted with various metals, were analysed by inductively coupled plasma/optical emission spectroscopy (ICP-OES). As GtHNL is a monocupin with one proposed metal-binding site, one metal ion is expected per subunit. This was also confirmed by X-ray crystallography (see below).

When the protein was expressed in LB medium without addition of metals, only traces of zinc and iron were detected during analysis (3.8 mol% Zn2+ and 5.3 mol% Fe2+ per metal-binding site). These two metals are generally present in LB medium, which was confirmed by ICP-OES analysis of the medium. As Zn2+ and Fe2+ were incorporated into the protein in only very low amounts, and the expression in minimal medium without any metal was significantly lower than in LB medium, further experiments were performed in complex medium. Interestingly, when 100 μm MnCl2 (final concentration) was added to the medium during induction with isopropyl thio-β-d-galactoside, the resulting enzyme showed a significantly increased manganese loading, corresponding to 58.9 mol% manganese per monomer. Addition of 100 μm ZnSO4 or FeSO4 to the expression medium did not lead to an increase in incorporation of these metals into the protein in vivo. In another set of experiments, apoprotein was produced by dialysing purified protein against a buffer containing the strong chelating agent 2,6-pyridinedicarboxylic acid. Circular dichroism spectroscopy confirmed that the apoprotein was still folded correctly, even after incubation for 1 h at pH 4.0 (Fig. S2B). According to ICP-OES measurements, the apoprotein contained residual metal (mainly iron) amounting to < 10% of all metal-binding sites. Subsequently, the apoprotein was incubated with MnCl2, ZnSO4 and FeSO4. In addition, purified protein expressed without addition of metal ions to the medium was also loaded with the same metals in vitro, as it contains almost no metal (data not shown). In both cases, the protein efficiently took up manganese, zinc or iron (Fig. 3A). Interestingly, only the protein reconstituted with manganese regained clearly measurable activity with respect to mandelonitrile cyanolysis (Fig. 3B). The activity of apoprotein reconstituted with Zn2+ or Fe2+ was slightly higher than that of the metal-free apoprotein, but was not clearly distinguishable from the high background reaction.

image

Figure 3. (A) Metal contents (determined by ICP-OES) of apoprotein after treatment with chelating agent, and subsequent incubation with MnCl2, ZnSO4 or FeSO4: green bars, manganese; purple bars, zinc; blue bars, iron. (B) Cyanolysis assay (67 mm sodium oxalate buffer pH 4.5, 18 mm (R)-mandelonitrile, 25 °C) with 50 μg of purified apoprotein (grey diamonds, dashed line) or apoprotein reconstituted with either manganese (green squares, solid line), zinc (purple circles, dashed line) or iron (blue triangles, solid line). Black crosses represent the buffer control.

Download figure to PowerPoint

Enantioselective synthesis of mandelonitrile

The ability of the enzyme to catalyse the synthesis of mandelonitrile from benzaldehyde and HCN was examined using a two-phase system consisting of benzaldehyde and HCN dissolved in methyl tertiary butyl ether (MTBE) as the organic phase, and an aqueous phase comprising concentrated cleared lysate, acidified to a final pH of 4 using sodium acetate buffer. The reaction was tested at 15 °C and 5 °C. As the reactions at 5 °C yielded a significantly higher enantiomeric excess (data not shown), all further reactions were performed at 5 °C. The enzyme expressed in the presence of manganese is highly enantioselective, yielding (R)-mandelonitrile with 90% enantiomeric excess (ee) at 80% conversion of the initial 0.5 m benzaldehyde after 6 h. As expected, when cleared lysate of cells grown without addition of manganese to the expression medium was used, the yield and enantiomeric excess were markedly lower: < 50% conversion and ~ 80% ee (Fig. 4).

image

Figure 4. Mandelonitrile synthesis in a two-phase system consisting of 500 μL of cleared cell lysates expressed with and without Mn2+ addition (50 mg·mL−1 total protein in 100 mm sodium acetate, pH 4.0) as the aqueous phase and 1 mL MTBE containing 0.5 m benzaldehyde and 2 m HCN as the organic phase, shaken at 1000 rpm and 5 °C. The curves show the percentage enantiomeric excess (ee) of (R) for the enzyme expressed with Mn2+ addition (black circles, dashed line) and without Mn2+ addition (grey diamonds, dashed line), and conversions for the enzyme expressed with Mn2+ addition (black squares, solid line) and without Mn2+ addition (grey triangles, solid line) and the empty vector control expressed with Mn2+ addition (grey crosses, dotted/dashed line).

Download figure to PowerPoint

As it is known that a chemical background reaction proceeds readily in aqueous medium [5], a buffer control, as well as cleared lysate of E. coli without GtHNL (empty vector control), were used as controls. The lysate of the empty vector control showed 15% conversion after 6 h, compared with 80% for the lysate expressing GtHNL, and the control reaction showed no enantioselectivity. Purified GtHNL had lower apparent activity (data not shown) and consequently yielded mandelonitrile with lower enantiomeric excess even when added at approximately the same amount as present in the cleared lysate. A possible reason for this may be the stabilizing effect of E. coli proteins in the cleared lysate, especially considering the harsh reaction conditions of pH 4 in the aqueous phase, the contact to MTBE, and the high concentrations of benzaldehyde and HCN at the interphase. In another cyanohydrin synthesis study, it was shown that addition of albumin stabilized the hydroxynitrile lyase from Hevea brasiliensis in the presence of high concentrations of HCN in buffer-saturated solvent [17]. On the other hand, it has been reported that an HNL from Arabidopsis thaliana, which was expressed in E. coli and purified, retained considerable activity even though it precipitated in a buffer-saturated organic system [18].

Additionally, apoprotein and reconstituted proteins were used in mandelonitrile synthesis reactions (Table 1). The apoprotein was clearly less active and less enantioselective compared to the metal-containing proteins (22% conversion, 40% ee). The zinc-reconstituted protein showed only slightly higher conversion than the apoprotein (32%), but the product had a significantly higher ee of 74%. The iron-reconstituted protein (81% conversion, 87% ee) showed comparable activity and enantioselectivity to the manganese-reconstituted aliquot (76% conversion, 89% ee), although the difference in their cyanolysis activities was more pronounced (Fig. 3B). GtHNL did not accept aromatic ketones, such as acetophenone and propiophenone, as substrates.

Table 1. Conversions and enantiomeric excess of mandelonitrile synthesis reactions with purified GtHNL either as apoprotein or reconstituted with Mn2+, Zn2+ or Fe2+. The two-phase system consisted of 500 μL of aqueous phase (20 mg·mL−1 purified protein in 100 mm sodium acetate, pH 4.0) and 1 mL MTBE containing 0.5 m benzaldehyde and 2 m HCN as the organic phase, shaken at 1000 rpm, 5 °C, with 6 and 24 h reaction times.
SamplePercentage conversion% ee (R)
 6 h24 h6 h24 h
GtHNL Apo11 ± 1.522 ± 1.840 ± 1.540 ± 0.1
Apo + Mn35 ± 1.776 ± 1.490 ± 0.189 ± 0.1
Apo + Zn11 ± 1.032 ± 1.071 ± 0.074 ± 0.1
Apo + Fe39 ± 8.981 ± 8.488 ± 0.387 ± 0.4

Overall three-dimensional structure of GtHNL

The His-tagged protein was purified efficiently using affinity chromatography and subsequent size-exclusion chromatography as described previously [19]. The protein crystals diffracted to ~ 2.5 Å, and contained two homotetramers in the asymmetric unit (Fig. 5A). The electron density of the polypeptide chain backbones was well-ordered in all subunits containing all 131 amino acid residues (see Table 2).

Table 2. Data collection and refinement statistics.
Data processing statistics
Beam lineESRF/ID23-1
Wavelength [Å]1.0329
Resolution range (outer shell) [Å]37.72–2.46 (2.52–2.46)
Space groupC 2 2 21
Unit cell

a = 126.31

b = 254.78

c = 82.47

α = β = γ = 90.0

R merge 0.121 (0.58)
R meas 0.131 (0.227)
II11.3 (3.6)
No. of reflections352 885 (26 703)
No. of unique reflections48 423 (3563)
Completeness [%]99.3 (99.8)
Multiplicity7.3 (7.5)
Refinement statistics
Resolution range [Å]2.46–37.72
Protein residues1048
No. atoms8768
Non-solvent8061
Solvent707
Ligands8
Rfactor0.1767 (0.2128)
Rfree0.2329 (0.2867)
CC (1/2)99.6 (92.8)
Deviation from ideal values
Bonds0.015
Angles1.216
Chirality0.080
Planarity0.006
Dihedral16.183
Min. nonbonded distance1.856
Molprobity statistics
All-atom clashscore6.55
Ramachandran plot
Favoured (%)97.58
Allowed (%)1.94
Outliers (%)0.48
Rotamers outliers (%)6.24
C-beta deviations1.00
Wilson B-factors36.67
image

Figure 5. (A) Cartoon depiction of the overall structure of the two homotetramers of GtHNL. The subunits are coloured in blue, magenta, light and dark green. (B) Stick representation of the amino acids in the metal-binding site. Magenta, metal-binding residues; purple, metal ion; light green, amino acids close to the metal-binding site. The figures were prepared using the PyMOL Molecular Graphics System (version 1.5.0.4, Schrödinger LLC), Portland, OR, USA.

Download figure to PowerPoint

The α-carbons of the subunits superimpose very well, with an RMSD (using chain A as the reference) of 0.23 Å. Each monomer comprises eleven β-strands and does not contain any helices. It forms a jellyroll β-sandwich with a topology that is characteristic of the cupin-barrel fold. The name cupin is derived from the Latin term ‘cupa’, which means small barrel and refers to the shape of the structure. The tetramer is formed by two dimers (mean interface area for A to B, C to D, E to F, G to H = 1554 Å2), whereby the N-terminus of chain A interacts with a β-sheet (G83–F88) from chain B and vice versa. The same holds for chain C and D. The two dimers are then twisted by 90° relative to each other (with chain A mainly interacting with chain C and chain B mainly interacting with chain D; mean total interface area of dimer AB to CD = 619.2 Å2). Each monomer has a metal-binding site at the base of the β-barrel (Fig. 5B). The metal ion interacts with three His residues and one Gln, a metal-binding site that is rare but has previously been reported in other cupins [16, 20]. The metal-binding residues are His53, His55, Gln59 and His94, with distances to the metal of 2.2, 2.2, 2.0 and 2.4 Å, respectively (calculated after alignment of all chains to the reference chain A as the distance between the metal ion and the mean position of metal-coordinating atoms). They align well with metal-binding residues in other cupins in terms of both sequence and structure. Interestingly, another histidine residue, His96, is located within binding distance to the metal (mean 2.4 Å). However, a metal-binding site comprising five binding residues has not been described in cupins previously, and His96 is located on the opposite side of the residues that are usually involved. The remaining coordination sites of the metal ion may interact with the substrate. The refinement with manganese as the metal ion supports the biochemical data and produces the best refinement statistics. Unfortunately, no substrate could be co-crystallized or soaked into the active site so far. Docking experiments with mandelonitrile resulted in several possible binding modes, most likely due to the large cavity.

A structural similarity search for GtHNL using the DALI server [21] showed that the closest structure is 2F4P, an uncharacterized monocupin protein from Thermotoga maritima, which was also used as a template for molecular replacement (Z-score 21.9, RMSD 1.7 over 125 amino acids). However, it has only 40% sequence identity to GtHNL. The other structures showed even lower similarity [1VJ2: Z-score 14.5, RMSD 1.9, 20% sequence identity; 1O4T: Z-score 14.5, RMSD 2.0, sequence identity 21% (both uncharacterized monocupins from Thermotoga maritima); 3HT2: Z-score 14.5, RMSD 2.4, 14% sequence identity (a zinc-containing polyketide cyclase RemF from Streptomyces resistomycificus [22])]. The gene corresponding to 2F4P was obtained as a synthetic gene and re-cloned, and the protein was expressed as described for GtHNL, but showed only very weak hydroxynitrile lyase activity, which was not investigated any further (data not shown).

Structure-guided mutagenesis

In addition to the amino acids of the metal-binding site, amino acids in and around the main cavity, which harbours the active site in all cupins characterized so far, were identified by structural analysis and mutated by site-directed mutagenesis (Fig. 6).

image

Figure 6. Stick representation of several amino acids in or close to the active site. Purple, metal-binding residues and metal ion; light green, amino acids in or close to the active site. The figure was prepared using the PyMOL Molecular Graphics System (version 1.5.0.4, Schrödinger LLC).

Download figure to PowerPoint

The effect of mutations on the cyanolysis reaction was qualitatively analysed using a colony-based assay (see Fig. 7A for metal-binding sites). Several of the mutations had no effect on the activity (F44A, L61A, Q110A) and were not investigated any further. The most interesting variants were also examined in the synthesis direction, analogous to the wild-type protein (Table 3). As it was shown that manganese is important for catalysis, the presumed metal-binding residues His53, His55, Gln59 and His94 were all mutated to alanine. In addition, Gln59 was mutated to glutamate, as Glu is more abundant in the metal-binding sites of cupins than Gln. However, no difference in expression or activity was observed (data not shown). All four alanine variants were expressed at wild-type level (Fig. 7A, upper row). Moreover, CD spectroscopy confirmed that the overall fold of the variants did not change (Fig. S4 and Table S3). Nevertheless, three of these alanine variants lost cyanolysis and cyanohydrin synthesis activity completely (Fig. 7A and Table 3). Interestingly, although mutant H94A was still active with respect to cyanolysis (approximately the same activity, 0.13 U·mg−1, as the wild-type enzyme in the photometric assay at pH 4.5), it showed a dramatic loss in synthesis activity and enantioselectivity with respect to mandelonitrile (Table 3). Metal analysis by ICP-OES showed that this variant was able to bind manganese efficiently, in contrast to mutants of the other three residues (Fig. 7B). To rule out stability problems for variant H94A due to contact with organic solvents in the synthesis reaction in contrast to the aqueous cyanolysis conditions, acidified cleared lysate (pH 4.0) containing the enzyme (mutant H94A and wild-type as control) was shaken with MTBE for 1 h. The subsequent filter assay showed no apparent loss of cyanolysis activity, indicating that the two-phase reaction system was not responsible for the lack of activity.

Table 3. Conversions and ee of mandelonitrile in synthesis reactions with cleared lysates of GtHNL variants. The two-phase system consisted of 500 μL of aqueous phase (50 mg·mL−1 of total protein in 100 mm sodium acetate, pH 4.0) and 1 mL MTBE containing 0.5 m benzaldehyde and 2 m HCN as the organic phase, shaken at 1000 rpm, 5 °C, with a 6 h reaction time. ND, not determined due to the low conversion.
VariantPercentage conversion% ee (R)
Wild-type6486
T50A5791
H96A 6ND
H96D 5ND
H106A 5ND
H106D2654
H55A 2ND
H94A1121
image

Figure 7. (A) Colony-based filter assay for the mutants H96A, H53A, H55A, Q59A and H94A, wild-type and vector control for mandelonitrile cyanolysis. Blue colour indicates the release of HCN. The incubation time in minutes is shown on the right. The top row shows the band at ~ 15 kDa obtained upon SDS/PAGE of the soluble and insoluble protein fractions of wild-type and each variant, respectively. (B) Metal contents (analysed by ICP-OES) for purified variant proteins expressed in LB medium with addition of 100 μm MnCl2; white bars, manganese; grey bars, zinc; black bars, iron.

Download figure to PowerPoint

Among the residues in or close to the active site, Thr50 and two histidines, His96 and His106, were considered as candidate residues involved in catalysis. In the HNLs from Hevea brasiliensis (HbHNL) and Prunus amygdalus (PaHNL), a histidine is proposed to be involved in catalysis [23, 24]. In PaHNL, the histidine may act directly as a general base, abstracting the proton from the mandelonitrile hydroxyl group, whereas in HbHNL, the histidine is part of a catalytic triad formed by Asp, His and Ser, in which the serine deprotonates the cyanohydrin [24]. Thr50, which is located in the second shell of the active site, may form a hydrogen bond with the vicinal His96 in the active site. His106 is located further away from Thr50 in the solved structure. However, the mutant T50A did not lose activity in any of the reactions. It showed a comparable conversion of ~ 58% and an even higher ee value (91.5%) than the wild-type protein (64% conversion, 86.6% ee), indicating that this residue is not essential for mandelonitrile synthesis (Table 3). His96, which is located within binding distance of the metal ion in the structure of GtHNL, was mutated to alanine. This variant was efficient at binding manganese (Fig. 7B) and folded correctly as determined by CD spectroscopy (Fig. S4), but it nevertheless lost most of its cyanolysis activity, based on the colony assay (Fig. 7A). Moreover, purified H96A had no detectable cyanolysis activity even at pH 6.0 with oxalate buffer in the standard cuvette assay (data not shown). In addition, His96 was mutated to lysine or arginine to introduce a positive charge or aspartate to introduce a negative charge. The H96R mutation resulted in insoluble protein, most likely due to clashes with the metal-binding site, while a lysine or aspartate at this position led to a soluble but inactive enzyme. All variants at position His96 were inactive in terms of mandelonitrile synthesis (Table 3). Variants of the other histidine residues in the active site, H106A and H106K, were expressed as soluble proteins in amounts comparable to the wild-type, but were inactive in the photometric mandelonitrile assay and the synthesis reaction (Table 3). Some weak activity was detectable in the more sensitive colony-based assay for cyanolysis. Interestingly, mutant H106D showed almost wild-type activity in the colony-based filter assay, and converted 26% of benzaldehyde to mandelonitrile (clearly above the background reaction), but with a low ee (Table 3).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We have cloned, crystallized and characterized a novel cupin hydroxynitrile lyase biochemically as well as by site-directed mutagenesis. To the best of our knowledge, the data shown here represent the first crystal structure, as well as the first detailed biochemical characterization of a cupin hydroxynitrile lyase. The enzyme has measurable, but relatively low, cyanolysis activity compared to previously described plant hydroxynitrile lyases [10]. Specific activity reached more than 4 U·mg−1 at pH 6.0 in the cyanolysis direction (Fig. 2). However, HNLs are distinguished by their ability to catalyse cyanohydrin synthesis from carbonyl and cyanide enantioselectively, rather than by their activity in the cyanolysis direction [1, 6]. The enzyme catalyses mandelonitrile formation from benzaldehyde and HCN with high stereoselectivity (Fig. 4). The enzyme thus represents a novel (R)-selective hydroxynitrile lyase. The activity was shown to be manganese-dependent in the cyanolysis direction. The manganese content and the activity increased significantly when the protein was expressed in medium supplemented with manganese. Zinc and iron did not significantly bind to the protein in vivo during expression in E. coli, although the two metal ions are abundant in the medium, but availability in the cytoplasm of E. coli is tightly regulated [25, 26]. Interestingly, zinc and iron bind to the protein in vitro after incubation of purified metal-free protein with either metal. Additionally, if manganese was removed in vitro from purified protein using chelating agents, the activity was completely lost. The same result was obtained by mutations in the metal-binding site. Activity was restored by loading the wild-type apoprotein with manganese. Under the significantly different conditions of the synthesis reaction, with high substrate loading and much longer reaction times, a metal dependence without clear preference for manganese was detected. CD spectra showed that all proteins and variants were correctly folded, indicating that the metal ion is not necessary for overall folding of the protein. However, small changes in the metal-binding site cannot be assessed by this method, and detailed X-ray crystallography and NMR studies are ongoing. Previously, it was reported that the manganese-dependent oxalate decarboxylase from Bacillus subtilis, which is a cupin, requires addition of manganese salts to the expression medium when over-expressed in E. coli [27]. However, in the case of the Bacillus enzyme, the manganese salts were required for proper folding. While the data strongly suggest that manganese is the preferred metal for GtHNL, this may only be taken as an indication, as the natural intracellular concentrations of the relevant metals are not known for Granulicella tundricola. In nature, cells tightly regulate the concentrations of various metals in their cytoplasm in order to survive, and utilize elaborate mechanisms in vivo to ensure that the various metalloproteins incorporate the correct metal [28, 29]. It has been shown previously that a marine cyanobacterium controls incorporation of manganese versus copper in two cupins with identical metal-binding amino acid residues by folding the protein in either the cytoplasm or the periplasm [30]. The organism thus achieves selectivity even though the metal-binding site of each enzyme is capable of accepting both metals in vitro.

This manganese dependence of a hydroxynitrile lyase represents a new catalytic role for manganese, which is present in a wide range of enzymes, such as hydrolases, dismutases, ligases, oxidoreductases and isomerases [31]. From a chemical perspective, manganese is known to act as a Lewis acid, and has been applied in organometallic catalysis for activation of benzaldehyde [32]. However, zinc is also known to act as a Lewis acid in enzymatic reactions on carbonyls, e.g. in alcohol dehydrogenases [33]. Moreover, a Zn2+-dependent HNL from Linum usitatissimum (LuHNL) with high sequence similarity to alcohol dehydrogenases was discovered many years ago [34]. However, no exact reaction mechanism has yet been published. LuHNL and the cupin HNLs do not share any conserved motifs. Even the metal-binding site is different, as the Zn2+ ion binds to two Cys and one His residue in LuHNL. The exact reason for the metal preference in GtHNL and the apparent difference in the synthesis and cyanolysis direction remains to be elucidated.

In addition to manganese, two histidine residues appear to be necessary for efficient cyanolysis, both of which are found near the metal-binding site. It has been postulated that HNLs require a positive charge and a general base for activity [23, 24] – functions that may both be fulfilled by histidines under different protonation states.

Conclusion

The cupin-fold HNLs appear to constitute a mechanistically distinct class of HNLs with no obvious similarity to previously described enzymes. Further research is necessary to elucidate the exact mechanism of this unusual, manganese-dependent, novel type of hydroxynitrile lyase, as well as to identify the full substrate range of this enzyme.

Experimental procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA) unless specified otherwise.

Cloning, mutagenesis and expression

The sequence encoding AciX9_0562 (gene ID 322434201) was obtained as a synthetic gene (GeneArt/Life Technologies, Carlsbad, CA, USA) that was codon-optimized for E. coli. The coding region was directly flanked by NdeI and HindIII restriction sites (ThermoScientificBio, Waltham, MA, USA), which were used to clone the gene into the expression vector pET26b(+) (Novagen/Merck, Darmstadt, Germany). Mutations were introduced by overlap-extension PCR, and the PCR products were ligated into vector pET28a(+) using the NcoI and HindIII restriction sites (see Table S1 for primer sequences). The constructs were confirmed by sequencing (LGC Genomics, Berlin, Germany). E. coli BL21 -Gold(DE3) (Stratagene, La Jolla, CA, USA) was used as the expression host. All cells were grown in LB (lysogeny broth, Lennox) medium (Carl Roth GmbH, Karlsruhe, Germany) supplemented with kanamycin sulfate (40 mg·L−1 final concentration). The protein was routinely expressed from pre-cultures diluted to an attenuance at 600 nm of ~ 0.1, and grown in baffled flasks at 37 °C and 120 rpm until an attenuance at 600 nm of ~ 0.6–0.8 was reached, after which the cultures were cooled to 25 °C, and expression was induced by addition of 0.1 mm isopropyl thio-β-d-galactoside (Carl Roth GmbH). When indicated, 100 μm of MnCl2, FeSO4 or ZnSO4 was added concomitantly with the induction. The induced cultures were harvested after 20–21 h at 25 °C. Protein expression and localization in cell extract fractions were monitored by SDS/PAGE. The percentage of expressed protein in cleared lysates was determined by quantifying the respective band on a Coomassie-stained SDS/PA gel using a G-box HR16 device (Syngene, Synoptics, Cambridge, UK), the software gene snap version 7.05 and Gene Tools version 4.00, using rolling disc baseline correction set at 30.

Cells were disrupted by sonification using a Branson sonifier S-250 (80% duty cycle, output control 7), twice for 3 min each, and cooled on ice. The crude lysates were cleared by centrifugation for 1 h at 48 250 g and 4 °C.

Colony-based filter assay

A colony-based filter assay suitable for high-throughput screening was performed as described previously [35]. Four identical spots of each clone were stamped adjacent to each other on a nylon membrane (Biodyne A, 0.2 μm, PALL Life Sciences, Port Washington, NY). The membranes were placed on LB agar (Lennox) plates supplemented with 40 mg·L−1 kanamycin, and the cells were grown overnight at 37 °C. The membranes with the colonies were then transferred onto LB agar plates containing 40 mg·L−1 kanamycin, 0.1 mm isopropyl thio-β-d-galactoside and 100 μm MnCl2, and the protein was expressed overnight at 20 °C. The colonies were pre-incubated with 100 mm citrate buffer, pH 3.5, after which buffer with 12 mm (R)-mandelonitrile (97%) was added. A piece of plastic mosquito net was placed on top of the membranes to prevent wetting, and overlaid by a filter paper (Whatman No. 1, GE Healthcare, Uppsala, Sweden) soaked with copper-(II) ethyl acetoacetate and 4,4′-methylenebis(N,N-dimethyl-aniline). Development of a blue colour indicates release of HCN. In the case of cleared lysates, the assay was adapted by placing a detection filter paper on top of a 96-well microtitre plate containing reaction mixtures comprising 50 μL cleared lysate and 100 μL 18 mm (R)-mandelonitrile in sodium oxalate buffer, pH 4.5.

Protein purification

Crude lysates of cells expressing GtHNL (calculated pI = 5.74, according to Protparam [36]) were produced by lysing cells by sonication (see above) in buffer A (50 mm Bis-Tris/HCl, pH 6.8, with 50 mm NaCl), and cleared by centrifugation (1 h, 48 250 g, 4 °C). Wild-type GtHNL and variants were purified in a first step on a Q-Sepharose anion-exchange column (HiTrap™ Q FF, 5 mL, GE Healthcare, Uppsala, Sweden). The column was loaded with ~ 40–50 mg total protein per ml matrix volume. GtHNL was eluted in a single step using 10% buffer B (50 mm Bis-Tris/HCl, pH 6.8, containing 1 m NaCl). The fractions were analysed by SDS/PAGE. Positive fractions were pooled, concentrated using Vivaspin 20 centrifugal filter units (10 000 Da molecular mass cut-off; Sartorius, Göttingen, Germany), and subsequently loaded onto a Superdex 200 16/60 size-exclusion column (GE Healthcare) pre-equilibrated with 20 mm Tris/HCl, pH 7.5, containing 200 mm NaCl.

Protein concentration determination

Protein concentrations of cleared lysates were routinely determined using the Bradford assay (Bio-Rad, Hercules, CA, USA). Concentrations of purified proteins were determined using a Nanodrop spectrophotometer (model 2000c, Peqlab, Erlangen, Germany), using an absorbance of 24.3 for a 1% solution (10 g·L−1) at 280 nm, as calculated based on the amino acid sequence using Protparam [36].

Metal exchange

The apoprotein was obtained by dialysing ~ 70–100 mg GtHNL (in ~ 20–50 mL of 20 mm Tris/HCl, pH 7.5, 200 mm NaCl) twice overnight against 500 mL chelating buffer (100 mm sodium acetate, 10 mm 2,6-pyridinedicarboxylic acid, pH 5.5, 200 mm NaCl) and once overnight against 4 L of the original buffer at 4 °C. The protein was reconstituted with metal by incubating 0.15–0.3 mm of purified protein in 20 mm Tris/HCl, pH 7.5, 200 mm NaCl with a tenfold molar excess (1.5–3 mm) of either MnCl2, FeSO4 or ZnSO4 for at least 4 h at room temperature (22 °C). Afterwards, the protein was transferred into metal-free buffer using PD-10 desalting columns (GE Healthcare). The samples were concentrated using Vivaspin 20 centrifugal filter units.

Metal analysis

The metal content was analysed by ICP-OES (Spectro Ciros Vision EOP, Spectro Analytical Instruments, Kleve, Germany) at the Institute of Analytical Chemistry and Food Chemistry, Technische Universität Graz, Austria. Purified water (18 MΩ·cm−1, Barnstead Nanopure, Thermo Fisher Scientific) and high purity acids (HCl and HNO3, Suprapur, Merck) were used throughout. ICP-OES calibration solutions (0.04–10 mg·L−1) were prepared from a 100 mg·L−1 multi-element stock (28-element ICP standard solution, Carl Roth) in 3% v/v HNO3. Enzyme samples weighing ~ 0.5 g (corresponding to 500 μL of 10–20 mg·mL−1 purified protein solutions in 20 mm Tris/HCl, pH 7.5, 200 mm NaCl) were digested with 1 mL HCl, 2 mL HNO3 and 2 mL H2O by means of a pressurized microwave digestion system (Multiwave 3000, Anton Paar, Graz, Austria). The microwave power was increased to 1400 W over a period of 10 min. During this period, the maximum working pressure of the PFA digestion vessels (HF rotor, 4 MPa) was reached. After a further 20 min of continued heating, the content of the vessels was cooled and diluted to a final volume of 20 mL. Scandium was added to the solutions as an internal standard. The concentrations of Cr, Cu, Fe, Mn, Ni and Zn were determined under robust plasma conditions [37] by ICP-OES. The ICP was operated at 1350 W RF power, 12.5 L·min−1 cooling gas flow, 0.6 L·min−1 auxiliary gas flow and 0.83 L·min−1 nebulizer gas flow. A standard torch with a 2.5 mm ID injector and a cross-flow nebulizer with a Scott-type spray chamber were used. The following emission lines were used: Cr, 205.552 nm; Cu, 219.226 nm; Fe, 238.204 nm; Mn, 259.373 nm; Ni, 231.604 nm; Sc, 361.384 nm; Zn, 206.191 nm. Metal loading was calculated based on the molar protein concentrations and the measured molar concentrations of the metals. As controls, LB medium, buffer and the flow-through of the Vivaspin 20 centrifugal filter units were analysed.

Mandelonitrile cyanolysis assay

The mandelonitrile cyanolysis assay at pH 4.5 was performed in microtitre plates using various final concentrations (3–48 mm) of (R)-mandelonitrile as substrate, 50–100 μg of purified protein per well in 50 μL of 20 mm Tris/HCl, pH 7.5, 200 mm NaCl and 100 μL of 100 mm sodium oxalate, pH 4.5, giving a final volume of 150 μL per well. The kinetic data were recorded at 25 °C on a Synergy MX SMATBLD microplate reader (Biotek, Winooski, VT, USA), running gen 5 software version 1.11. The increase in the reaction product, benzaldehyde, was followed at 280 nm and used for the calculation of activities (ε280 nm = 1.376 L·mmol−1·cm−1) [18, 35]. Km and Vmax values were calculated using SigmaPlot 11.0 (systat Software, Chicago, IL, USA) with the module ‘Enzyme Kinetics 1.3′ based on the Michaelis–Menten model. For determination of the cyanolysis activity at higher pH values, a two-component buffer with 100 mm oxalic acid and 100 mm MES was used, which was adjusted to pH 5.0, 5.5 or 6.0 with sodium hydroxide. The photometric assays at pH 5–6 were performed using 50 μg of purified protein in 500 μL (final volume) in 1 mL quartz cuvettes (1 cm pathlength) using a Beckman DU800 spectrophotometer (Beckman Coulter, Pasadena, CA, USA) without temperature control (the temperature in the cuvettes in the photometer was measured using a standard thermometer at several time points and revealed a constant temperature between 25 and 27 °C). The linear range of the photometer was probed by a standard curve with benzaldehyde, showing that the absorbance is perfectly linear up to a value of 3 (R2 = 0.998).

The substrate stock (90 mm (R)-mandelonitrile) was prepared in 3 mm oxalic acid, and 100 μL thereof were added to the cuvettes to start the reactions. The cuvettes were sealed immediately to prevent release of HCN. For all buffers and pH values, the corresponding chemical cyanolysis background values were subtracted. As the assay is monitored at 280 nm, at which proteins also absorb, variants with low activity could not be measured, as the detection limit of the photometer was reached due to the high amount of protein required in the assay.

One mandelonitrile cyanolysis unit was defined as the amount of enzyme that catalyses the release of 1 μmol of benzaldehyde from mandelonitrile dissolved in aqueous buffer in 1 min at 25 °C, under specified buffer and pH conditions.

Mandelonitrile synthesis reaction

All experiments involving the possible release of HCN were performed in a well-ventilated hood, in which an HCN detector (Dräger PACIII, Lübeck, Germany) was placed for continuous monitoring. Mandelonitrile synthesis was performed in a two-phase system with a 0.5 mL aqueous phase containing the enzyme (cleared lysate with 40–50 mg·mL−1 of total protein or alternatively ~ 20 mg·mL−1 of purified enzyme at pH 4.0 in 100 mm sodium acetate buffer) and a 1 mL organic phase comprising 0.5 m freshly distilled benzaldehyde (molar ratio enzyme to benzaldehyde 1 : 1000), 2% v/v triethylbenzene or triisopropylbenzene as internal standard, and 2 m HCN in MTBE. The HCN/MTBE solution was prepared via in situ extraction of HCN released from sodium cyanide with HCl, as described previously [18]. The reaction mixtures were cooled to 5 °C in a thermomixer and shaken at 1000 rpm. Samples comprising 10 μL of the organic phase were withdrawn at appropriate time points through rubber septa to prevent the release of HCN, and diluted 1 : 100 in HPLC eluent (n-heptane/isopropanol/trifluoroacetic acid, 95 : 5 : 0.1). Chiral analysis by HPLC using a Chiracel OB-H column (4.5 μm × 250 mm, Chiral Technologies Europe SAS, Cedex, France) was performed as described previously [38]. Alternatively, a Daicel OD-H column (Chiral Technologies Europe SAS) was used (OD-H 25 0 mm× 4.6 mm × 5 μm; n-hexane/isopropanol/trifluoroacetic acid, 96 : 4 : 0.1 ; 20 °C; 1 mL·min−1; 4 MPa; retention times 3.4 min for triisopropylbenzene, 6.7 min for benzaldehyde, 24.7 min for (S)-mandelonitrile and 26.2 min for (R)-mandelonitrile). The conversion was calculated based on the decrease in the benzaldehyde peak area compared to the peak area of the internal standard. A sample was taken from the organic phase before addition of the aqueous phase, and the peak areas were taken as time point zero. No mandelonitrile was detectable in the starting samples. The enantiomeric excess was calculated by comparing the peak areas corresponding to (R)- and (S)-mandelonitrile. The analyte content in the aqueous phase compared to the organic phase was assumed to be negligible, and this was confirmed by running a blank reaction with only buffer as the aqueous phase. All experiments were performed in triplicate. Acetophenone (97%) and propiophenone (99%) (both from ACROS Organics/Thermo Fisher, Waltham, MA, USA) were used at 0.2 m concentration (tenfold molar excess of HCN), and the reactions were monitored for conversion in analogy to benzaldehyde.

Crystallization and structure determination

GtHNL was purified and crystallized as described previously [19]. In brief, the initial crystallization condition H02 (10% w/v poly(ethylene glycol) 8000, 20% v/v ethylene glycol, 0.02 m each of the amino acids d,l-Ala, d,l-Lys, d,l-Ser, l-Glu and Gly, 0.1 m MES/imidazole buffer, pH 6.5, Morpheus crystallization screen, Molecular Dimensions, Newmarket, UK [39]) and the purified protein sample (~ 10 mg·mL−1) were dispensed using an Oryx8 protein crystallization robot (Douglas Instruments Ltd, Hungerford, UK) onto a SWISSCi two-well crystallization plate (Molecular Dimensions). A serial dilution grid of both components was generated using the XStep program (Douglas Instruments Ltd) in order to improve the crystal quality. The best diffracting crystals were harvested under conditions comprising ~ 30% original protein solution and ~ 35% original precipitant solution. For data collection, crystals were frozen without cryoprotection. Three native data sets were collected at the European Synchrotron Radiation Facility (Grenoble, France) at beam lines ID23-2 and ID23-1 (see Table S2). The data were processed using program xia2 [40] in semi-automatic mode using 3d mode (XDS, XSCALE, LABELIT and CCP4 backends [41, 42]), analysed and converted to mtz format using the program POINTLESS [43]. The processing statistics as well as data quality analysis were performed using the program Xtriage of the PHENIX suite [44]. The data processing results for the best dataset are presented in Table 2. Molecular replacement was performed using BALBES [45], an automated molecular replacement pipeline (version 1.1.5_DB_1.1.5). In total, eight molecules were placed in the asymmetric unit based on template 2F4P, which is deposited in the Protein Data Bank as a cupin-like protein from Thermotoga maritima (tm1010). Initial automated model building was performed in PHENIX suite (versions 1.8.1-1168 and 1.8.2-1309) using the AutoBuild wizard. The final model was built manually using the program Coot (version 0.7) [46] and refined using the programs REFMAC [47] (version 5.7.0032) and phenix.refine from the PHENIX suite. The atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB ID 4BIF).

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

This work was supported by the Austrian Bundesministerium für Wirtschaft, Familie und Jugend, the Austrian Bundesministerium für Verkehr, Innovation und Technologie, the SFG, the Standortagentur Tirol, and the Technology Agency of the City of Vienna through the COMET Funding Program managed by the Austrian Forschungsförderungsgesellschaft. A research stay by I.H. at the Technische Universiteit Delft was supported by the Doktoratskolleg Molecular Enzymology (FWF project W901-B12). We thank Helmar Wiltsche, Institute of Analytical Chemistry and Food Chemistry, TU Graz for the ICP-OES analysis, Sanjib Kumar Karmee, Mandana Gruber and Romana Wiedner for their help with cyanohydrins synthesis reactions, and Tea Pavkov for help with the CD measurements. We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities, and we thank their scientific and technical personnel for assistance in using beamlines ID23-1 and ID23-2.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
febs12501-sup-0001-FigS1-S4-TableS1-S3.zipapplication/PDF521K

Fig. S1. Michaelis–Menten kinetics of purified GtHNL expressed with addition of 100 μm MnCl2.

Fig. S2. CD spectra showing the pH stability of purified GtHNL protein at pH 4.0.

Fig. S3. SDS/PAGE of cleared lysate and purified protein wild-type GtHNL and the mutants H96A, H53A, H55A, Q59A and H94A.

Fig. S4. CD spectra of GtHNL variants superimposed on the CD spectrum of the wild-type protein.

Table S1. Sequences of primers used for cloning and mutagenesis in this study.

Table S2. Data collection statistics.

Table S3. Predicted secondary structure composition.

Appendix S1. Methods for circular dicroism (CD) spectroscopy.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.