Molecular cloning of acetone cyanohydrin lyase from flax (Linum usitatissimum). Definition of a novel class of hydroxynitrile lyases.

Acetone cyanohydrin lyase from Linum usitatissimum is a hydroxynitrile lyase (HNL) which is involved in the catabolism of cyanogenic glycosides in young seedlings of flax. We have isolated a full-length cDNA clone encoding L. usitatissimum HNL (LuHNL) from a cDNA expression library by immunoscreening. LuHNL cDNA was expressed in Escherichia coli and isolated from the respective soluble fraction in an active form which was biochemically indistinguishable from the natural enzyme. An open reading frame of 1266 base pairs encodes for a protein of 45,780 kDa. The derived amino acid sequence shows no overall homologies to the to date cloned HNLs, but has significant similarities to members of the alcohol dehydrogenase (ADH) family of enzymes. In particular, the cysteine and histidine residues responsible for coordination of an active site Zn2+ and a second structurally important Zn2+ in alcohol dehydrogenases are conserved. Nevertheless, we found neither alcohol dehydrogenase activity in LuHNL nor HNL activity in ADH. Moreover, well known inhibitors of ADHs, which interfere with the coordination of the active site Zn2+, fail to affect HNL activity of LuHNL, suggesting principally different mechanisms of cyanohydrin cleavage and alcohol oxidation. Interestingly, LuHNL like ADH and Prunus serotina (PsHNL) possesses an ADP-binding βαβ unit motif, pointing to the possibility that the non-flavoprotein PsHNL and the flavoprotein LuHNL have developed from two independent lines of evolution of a common ancestor with an ADP-binding βαβ unit.

Hydroxynitrile lyases (HNLs) 1 catalyze the decomposition of cyanohydrins (␣-hydroxynitriles) into the corresponding aldehyde or ketone and cyanide (1). All HNLs described so far are found in cyanogenic plants. In these plants, HNLs are involved in the catabolization of cyanogenic glycosides during cyanogenesis or in the metabolization of these compounds during seedling development (1)(2)(3)(4). In the presence of high concentrations of HCN and aldehydes or ketones, HNLs can be used as biocatalysts for the stereoselective synthesis of a wide array of cyanohydrins, important building blocks in the pharmaceutical and fine chemical industries (5).
In recent years the HNLs from Prunus serotina (PsHNL) (6), Sorghum bicolor (SbHNL) (7), Manihot esculenta (MeHNL) (8), and Hevea brasiliensis (HbHNL) (9) have been molecularly cloned. Analysis of the cDNA-derived amino acid sequences revealed that these enzymes belong to three different classes of HNL. While MeHNL and HbHNL share 74% identity (9), no sequence homologies can be found to PsHNL or SbHNL. The lack of sequence homologies between the cloned HNLs correlates with the fundamental differences between these enzymes with regard to molecular weight, subunit composition, glycosylation, FAD content, and substrate specificity. 2 The flavoprotein PsHNL has moderate homologies to various other flavoproteins, especially to various types of dehydrogenases and oxidases (6). In particular, a stretch of 27 amino acid residues near the N terminus of PsHNL fulfill Wierenga's rule for an ADP-binding ␤␣␤ unit (6), while MeHNL and HbHNL show homologies to two proteins of unknown function from rice (9). However, the most intriguing homologies are found for SbHNL, namely, that this HNL possesses up to 50% homology over the whole sequence to serine carboxypeptidases, which belong to the structurally well investigated group of ␣/␤ hydrolase fold enzymes (5,7). In particular, sites critical for function and structural integrity of serine carboxypeptidases are conserved, suggesting that SbHNL is also a ␣/␤ hydrolase fold enzyme. All ␣/␤ hydrolase fold enzymes have a "nucleophile-histidine-acid" catalytic triad found in common with the subtilisin and chymotrypsin class of serine proteases (10). In all these enzymes, the nucleophile is part of the consensus motif Gly-X-Ser/Cys-X-Gly/ Ala-Gly/Ala (10). There is functional evidence by site-directed mutagenesis for the use of a catalytic triad by MeHNL and HbHNL, as well (9,11). Moreover, the order of the catalytic triad residues in primary sequence suggests that these HNLs also belong to the ␣/␤ hydrolase fold group of enzymes despite having no sequence homologies to SbHNL (11).
Here we describe the molecular cloning of LuHNL, which, like MeHNL and HbHNL, has acetone cyanohydrin as its natural substrate. However, in contrast to these presumed ␣/␤ hydrolase fold enzymes, LuHNL was found to be structurally related to the alcohol dehydrogenase class of enzymes. In particular, amino acid residues of ADHs important for structural integrity or coordinating Zn 2ϩ are conserved in LuHNL. However, despite having all the conserved residues responsible for Zn 2ϩ binding, LuHNL neither exerts ADH activity nor is inhibited by reagents interfering with Zn 2ϩ coordination in liver ADH.

EXPERIMENTAL PROCEDURES
Plant Material-Seeds of Linum usitatissimum L. were obtained from Frank AG (Herrenberg, Germany). Seeds were germinated for * This work was supported by Deutsche Forschungsgemeinschaft Grant WA 1025/1-1 and by the Bundesministerium fü r Forschung und Technologie, Germany, Grant A03U-ZSP Stuttgart. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) Y09084.
Chemicals-Except where noted, chemicals were purchased from Sigma (Deisenhofen, Germany). Chromatography resins, AutoRead sequencing kit, and cDNA synthesis kit were obtained from Pharmacia LKB Biotechnology Inc. (Freiburg, Germany). Lambda ZAP DNA was from Stratagene, and the bicinchoninic acid protein assay kit was from Pierce.
Enzyme Assays-The activity of LuHNL was measured as described by Selmar et al. (12). The amount of HNL which decomposes 1 mol of acetone cyanohydrin in 1 min under the conditions described in Selmar et al. (12) is defined as 1 unit. Total protein for calculation of specific activities was determined with the bicinchoninic acid protein assay reagent from Pierce according to the manufacturer's recommendations. Enzyme assays for liver alcohol dehydrogenase (Sigma) were essentially performed as described by Dunn and Bernhard (13). For inhibition studies, enzymes were incubated in the appropriate reaction buffer containing the respective reagent for 30 min at room temperature. The remaining activity was then determined as described above.
Purification of LuHNL and Immunization-Purification of LuHNL from flax seedlings and the immunization procedure were performed as described previously (14). Purification of recombinant LuHNL was performed as follows. E. coli cells expressing LuHNL were harvested by centrifugation, resuspended in binding buffer (50 mM sodium phos-phate, pH 8.0, 300 mM NaCl), and disrupted by sonication on ice. The supernatant was cleared by centrifugation and applied on a 1-ml Ninitrilotriacetic acid Superflow column (Quiagen, Minden, Germany). After washing with 10 ml of binding buffer, bound proteins were eluted with a 30-ml gradient of 0 -0.5 M imidazole in binding buffer.
Generation and Screening of a Lambda ZAP II cDNA Library-Total RNA was isolated from equal amounts of 5-, 7-, and 10-day-old seedlings of L. usitatissimum according to the guanidine chloride procedure described by Logeman et al. (15). Poly(A) ϩ -enriched RNA was obtained using Oligotex(dT) particles (Quiagen, Minden, Germany). Oligo(dT)primed cDNA was obtained with a cDNA synthesis kit (Pharmacia) according to the manufacturer's recommendations and cloned in lambda ZAP II (Stratagene). A total of 1 ϫ 10 5 recombinants were screened with polyclonal antisera (1:1000) against LuHNL as described by Young and Davis (16). Positive clones were visualized by detection of the formed complex of anti-LuHNL antibodies and anti-mouse (Ig) antibodies conjugated with alkaline phosphatase (0.4 g/ml; Dianova, Hamburg, Germany). Detection was performed with a combination of 0.01% 5-bromo-4-chloro-3-indolyl phosphate and 0.02% nitro blue tetrazolium complemented with 4 mM MgCl 2 as a substrate. After in vivo DNA Sequencing and Analysis-Sequencing of double-stranded DNA templates was achieved by a modified chain-termination method (17) using T7 DNA polymerase and the ALF express DNA analysis system (Pharmacia). For priming, T3 and T7 primer or specific oligonucleotides based on the preceding sequences were used. Obtained DNA and amino acid sequences were compiled and analyzed using the HUSAR (Heidelberg UNIX Sequence Resources) software.
SDS-PAGE and Immunoblotting-Proteins were separated on 15% (w/v) polyacrylamide gels according to Laemmli (18) and either silver stained according to Blum et al. (19) or transferred to nitrocellulose as described by Towbin et al. (20). The immunoblots were further handled as described elsewhere (14).

RESULTS AND DISCUSSION
Cloning and Nucleotide Sequence of HNL from L. usitatissimum-A size selected (800 -2500 base pairs) cDNA expression library in lambda ZAP comprising 5 ϫ 10 5 plaque-forming units per packing reaction (85% recombinants) was used without further amplification for screening with anti-LuHNL antisera. About 100,000 plaques were screened as described under "Experimental Procedures." Three positive clones, containing inserts of 1.3, 1.4, and 1.5 kilobase pairs in length were identified, converted into plasmid and used for further analysis. Both strands of the cDNAs were sequenced, and a common open reading frame of 422 amino acid residues downstream of the lacZ part was identified. The open reading frame encodes for a protein with a predicted molecular mass of 45,780 which is in good accordance with the molecular mass of 42,000 estimated for purified LuHNL by SDS-PAGE (21). In particular, the open reading frame contained sequences near the start methionine corresponding to the N-terminal sequence of Lu-HNL determined by Albrecht et al. (22) by Edman degradation. Two potential polyadenylation signals (AATAAA) occur at 1429 and 1507. The nucleotide sequence and the predicted amino acid sequence are shown in Fig. 1.
Protein Sequence Analysis-Searching for homologies using HUSAR and the TFASTA algorithms revealed that LuHNL shares up to 40% homology over the whole molecule with members of the zinc-containing alcohol dehydrogenase family of enzymes (Fig. 2). Of great interest is the fact that there are no significant homologies of LuHNL to the other HNLs cloned so far. Remarkably, residues which are structurally or functionally important in alcohol dehydrogenases are conserved (Fig.  2). Like LuHNL, alcohol dehydrogenases consist of two identical subunits (23). For ADHs it has been shown by resolution of the three-dimensional structure that each subunit is divided into two domains separated by the deep active site cleft (23). In ADHs one of these domains binds the coenzyme (coenzyme binding domain). The other domain binds two zinc ions (cata-lytic domain), one of which is located in the active-site pocket, whereas the other one is located on a lobe outside the catalytic center (23). All residues responsible for Zn 2ϩ binding in ADHs are conserved in LuHNL ( Fig. 2A). Three glycines (LuHNL: Gly 84 , Gly 95 , and Gly 104 ) of ADHs which are invariant because of lack of space for a side chain in hydrophobic cores are also conserved ( Fig. 2A). What is more, analyzing the deduced Lu-HNL amino acid sequence according to Wierenga's rules (24) revealed a ADP-binding ␤␣␤-unit motif comprising residues 219 -248 ( Fig. 2A). Based on analysis of several FAD-and NAD-binding proteins, Wierenga et al. (24) have defined 11 residues within a 29 -31 amino acid motif, which are necessary to allow the sequence folding into an ADP-binding ␤␣␤ unit. The ADP-binding ␤␣␤ unit motif of LuHNL, compared in Fig.  2B with those of other ADP-binding proteins, matched exactly with the consensus sequence and is therefore very likely to prove to be folded as an ADP-binding ␤␣␤ unit (Fig. 2B). Nevertheless, LuHNL catalyzes no net oxidation or reduction, suggesting that this fold is rather of structural than of catalytic  importance. Taking into account the above mentioned conservation of structurally important residues between LuHNL and ADHs, we propose that the overall structure of LuHNL is quite similar to that of ADHs. Inhibition Studies-Given the complete conservation of cysteine and histidine residues required for coordination of the active Zn 2ϩ in alcohol dehydrogenases, we proposed two questions. First, has LuHNL a side dehydrogenase activity or have ADHs a side HNL activity and second, are the above mentioned conserved residues functionally involved in LuHNL-catalyzed acetone cyanohydrin cleavage? As shown in Table I, we found no ADH activity with LuHNL in a standard ADH assay; likewise, we found no indication for hydroxynitrile lyase activity in a commercially available liver alcohol dehydrogenase preparation. As mentioned above, one of the two coordinated Zn 2ϩ of each ADH subunit is located in the active site and functionally involved in catalysis. Therefore both o-phenanthroline, which forms complexes with Zn 2ϩ , and diethyl pyrocarbonate, which modifies histidine side chains, are potent inhibitors of ADH activity (23). Taking into account that the above mentioned residues responsible for Zn 2ϩ coordination are conserved between LuHNL and ADHs, we inquired into a putative involvement in LuHNL-catalyzed cyanohydrin cleavage of these residues. Surprisingly, we found no influence of these inhibitors on LuHNL activity even after extended preincubation and at concentrations 10 to 20 times higher than those used for significant inhibition of ADH activity (Table I). The competitive inhibitor o-phenanthroline forms a ADH-Zn 2ϩ -o-phenanthroline complexes with the active-site Zn 2ϩ of ADHs (25) leading to reversible inactivation of these enzyms. The lack of inhibition of HNL-activity by this agent therefore suggests that Zn 2ϩ is not directly involved in LuHNL-catalyzed cyanohydrin cleveage.
However, a structural role of Zn 2ϩ ions in LuHNL cannot be ruled out by these data. LuHNL was poorly inhibited by high concentrations (10 mM) of the serine-modifying reagent diisopropyl fluorophosphate (Table I), whereas HNLs having a catalytic triad were almost completely inhibited by this compound in the low millimolar range (7,11). Hence, involvement of a serine residue in LuHNL catalysis is rather unlikely.
Functional Expression of LuHNL-To investigate the biochemical properties of recombinant LuHN, we cloned LuHNL cDNA in the inducible procaryotic expression vector pQE10 (Quiagen) and expressed the protein in E. coli (Fig. 3). Recombinant LuHNL was purified by affinity chromatography on a Ni-nitrilo triacetic acid Superflow column with a linear gradient from 0 to 0.5 M imidazole in binding buffer. The purification process yielded a pure enzyme preparation with a specific activity of 40 units/mg which is in accordance with the specific activity (34 units/mg) described for the natural enzyme. The recombinant enzyme exhibited a Michaelis-Menten kinetic with a K m for acetone cyanohydrin of 1.9 mM and V max of 71 mol/min/mg (Fig. 3), which again matched well with the values published for the natural enzyme by Xu et al. (21) (K m ϭ 2.5 mM). In contrast to other HNLs, the substrate specificity of LuHNL has not yet been studied in detail. Nevertheless, preliminary data of Albrecht et al. (22) suggested that LuHNL acts preferentially on aliphatic (R)-cyanohydrins. The molecular cloning of LuHNL described here should now allow the overexpression of this enzyme. A more thorough investigation of the technological potential of this enzyme with regard to its possible use as a biocatalyst in stereoselective synthesis of cyanohydrins should therefore be possible.
Phylogeny of Hydroxynitrile Lyases of Higher Plants-Previously, HNLs were divided into two fundamental classes according to their FAD content. However, comparison of the amino acid sequence of various recently cloned HNLs (PsHNL, Sb-HNL, MeHNL, HbHNL) (6 -9) revealed sequence homologies to other proteins, suggesting the existence of at least three phylogenetically independent groups of HNLs. One group formed by HbHNL and MeHNL, sharing 74% identity, exerts significant homologies to two proteins of as yet unknown function from rice (9). The other two groups are defined by SbHNL and PsHNL, respectively. While sequence analysis of the SbHNL revealed extensive homologies to serine carboxypeptidases, which belong to the structurally well investigated group of ␣/␤ hydrolase fold enzymes (5,7), the flavoprotein PsHNL shows moderate homologies to various flavoproteins, especially to dehydrogenases and oxidases. Remarkably, we found no sequence homologies of LuHNL with the HNLs from cassava (MeHNL) and rubber tree (HbHNL), despite a common natural substrate (acetone cyanohydrin). This lack of sequence homology is consistent with the discrepancy in biochemical properties of these enzymes (Table II). Therefore, we propose here that LuHNL defines a fourth group of HNLs. LuHNL, like PsHNL, has an ADP-binding ␤␣␤ unit motif matching strikingly with the conserved sequence defined by Wierenga et al. (24) for this fold. Otherwise, there are no overall sequence similarities between these two HNLs, suggesting that they have evolved independently from two different lines of evolution from an ancestoral  aϭ no homologies to proteins with known function but homologies to two proteins from rice of unknown function.
ADP-binding ␤␣␤ unit protein (Fig. 4). There is no indication of a phylogenetic relationship of LuHNL or PsHNL to the ␣/␤ hydrolase fold enzymes. Taking into account the lack of similarity in biochemical properties of the, to date, noncloned HNLs from Ximenia americana (26) and Phlebodium aureum (5) and the above discussed HNLs, it is likely that additional, phylogenetically defined groups of HNLs exist (Fig. 4).