By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
B. L. Møller, Plant Biochemistry Laboratory, Department of Plant Biology and Biotechnology, University of Copenhagen, Thorvaldsensvej 40, DK-1871 Frederiksberg, Denmark Fax: +45 35 333 333 Tel: +45 35 333 352 E-mail: email@example.com Website: http://www.plbio.life.ku.dk/English/Sections/plchem.aspx
Nitrile formation in plants involves the activity of cytochrome P450s. Hydroxynitrile glucosides are widespread among plants but generally do not occur in glucosinolate producing species. Alliaria petiolata (garlic mustard, Brassicaceae) is the only species known to produce glucosinolates as well as a γ-hydroxynitrile glucoside. Furthermore, A. petiolata has been described to release diffusible cyanide, which indicates the presence of unidentified cyanogenic glucoside(s). Our research on A. petiolata addresses the molecular evolution of P450s. By integrating current knowledge about glucosinolate and hydroxynitrile glucoside biosynthesis in other species and new visions on recurrent evolution of hydroxynitrile glucoside biosynthesis, we propose a pathway for biosynthesis of the γ-hydroxynitrile glucoside, alliarinoside. Homomethionine and the corresponding oxime are suggested as shared intermediates in the biosynthesis of alliarinoside and 2-propenyl glucosinolate. The first committed step in the alliarinoside pathway is envisioned to be catalysed by a P450, which has been recruited to metabolize the oxime. Furthermore, alliarinoside biosynthesis is suggested to involve enzyme activities common to secondary modification of glucosinolates. Thus, we argue that biosynthesis of alliarinoside may be the first known case of a hydroxynitrile glucoside pathway having evolved from the glucosinolate pathway. An intriguing question is whether the proposed hydroxynitrile intermediate may also be converted to novel homomethionine-derived cyanogenic glucoside(s), which could release cyanide. Elucidation of the pathway for biosynthesis of alliarinoside and other putative hydroxynitrile glucosides in A. petiolata is envisioned to offer significant new knowledge on the emerging picture of P450 functional dynamics as a basis for recurrent evolution of pathways for bioactive natural product biosynthesis.
Bioactive natural products containing a nitrile functional group (RC≡N) are widespread in biological systems and have been described in animals, plants, algae, sponges, fungi and bacteria [1,2]. Cyanogenic glucosides represent the largest group of nitrile glucosides occurring in more than 2650 higher plant species within pteridophytes, gymnosperms and angiosperms as well as in a few arthropod species [3–5]. Cyanogenic glucosides are β-glucosides of α-hydroxynitriles and in some plant species they co-occur with structurally related non-cyanogenic β- and γ-hydroxynitrile glucosides . Another type of natural products containing a nitrile functional group is the cyanolipids, a small group of fatty acid esters of α- or γ-hydroxynitriles. Their occurrence appears to be restricted to plant species within the Sapindaceae and Hippocastanaceae [6,7]. In the Brassicales order, nitriles may be detected as degradation products of glucosinolates and as intermediates in the biosynthesis of tryptophan-derived phytoalexins. Indolacetonitrile may also serve as an intermediate in the biosynthesis of the phytohormone indolyl-3-acetic acid [8–10]. Natural products harbouring the nitrile functional group are all thought to be derived from amino acids in plants, arthropods, bacteria and fungi [1,6,11]. Amino acids can be converted into oximes via N-hydroxylation, decarboxylation and dehydration reactions. The resulting oximes are subsequently converted into the corresponding nitriles by dehydration. In plants, these reactions are all catalysed by cytochrome P450 monooxygenases (P450s) belonging to the CYP71 clan [3,6,9,12,13].
Alliarinoside is a α-β unsaturated γ-hydroxynitrile glucoside found in Alliaria petiolata (M.Bieb.) Cavara and Grande (garlic mustard), a member of Brassicaceae . Recently, A. petiolata has developed into an aggressive invasive species in North America and research efforts to determine its chemical ecology and how the arsenal of natural products may contribute to invasiveness have been initiated. A. petiolata is the only species known to contain glucosinolates as well as a γ-hydroxynitrile glucoside. Furthermore, A. petiolata has been reported to release hydrogen cyanide , which indicates the presence of an unidentified cyanogenic glucoside. This review focuses on the co-occurrence of glucosinolates, alliarinoside and possibly also a cyanogenic glucoside in A. petiolata and addresses the knowledge about P450 evolution that may be gained from elucidation of the evolutionary relationship between biosynthesis of these three classes of bioactive natural products.
A. petiolata is a biennial herb belonging to the Brassicaceae. It is native to the Eurasian temperate zone and is an invasive species in North American deciduous forests, where it has spread to at least 34 US states and four Canadian provinces since its first reported occurrence in 1868 . The introduction of A. petiolata has presumably occurred from multiple European sources as indicated by analyses of genetic variation [17,18]. The successful invasion of North America may be attributed to the chemical arsenal carried by A. petiolata mediating resistance to herbivory by endemic species [14,19–21] as well as inhibiting the growth, survival and reproduction of surrounding plant species, a phenomenon known as allelopathy [22–28].
Alliarinoside was initially isolated from the foliage of A. petiolata and characterized as (Z)-4-(β-d-glucopyranosyloxy)but-2-enenitrile  (Fig. 1). It has not been described in any other species and has insect-deterring properties. Using bioassays with larvae of the butterfly Pieris napi oleracea, which uses various other Brassicaceae species as host plants, isolated alliarinoside was found to inhibit herbivory [14,21]. The larvae were observed to feed for 4 h before they became motionless and stopped further feeding attempts. Hence, the effect of alliarinoside on the larvae was post-ingestive . This was contrasted by a more immediate deterring effect observed following ingestion of isovitexin 6″-d-β-glucopyranoside, another bioactive natural product present in A. petiolata . This flavone diglucoside has also been found in the distantly related Gentiana arisanensis (Gentianaceae) . A. petiolata contains a range of other flavonoid glucosides [20,21]. Flavonoid biosynthesis includes a number of P450-catalysed reactions beyond the scope of this nitrile-focused review.
Different glucosinolates have been described in A. petiolata (Fig. 1). 2-Propenyl glucosinolate has been reported most frequently and appears to be the most abundant glucosinolate in aerial tissues [28,30–33]. However, in one study 2-propenyl glucosinolate was detected in leaves and stems at an early stage of development in autumn but not after the onset of flowering in spring. In roots, 2-propenyl glucosinolate was present at both analysed stages . This indicates that A. petiolata exhibits developmental- and tissue-specific patterns in the accumulation of glucosinolates, a phenomenon also observed in other glucosinolate producing species [34,35]. In support of this notion, it was concluded in a review of glucosinolate content of various species that there is quantitative and frequently also qualitative variation in the glucosinolate content between A. petiolata tissues, during the growth cycle, and between seeds of different origin . This may explain why other glucosinolates have not been reported consistently in analyses of A. petiolata glucosinolates. Benzyl glucosinolate has been found in roots and aerial tissues [28,31,32,36]. There is one report of phenylethyl glucosinolate from roots , while another study could not detect it in foliage . The presence of 3-butenyl glucosinolate in amounts comparable with those of 2-propenyl glucosinolate has been deduced from measurements of the degradation product but-3-enylisothiocyanate . Indol-3-ylmethyl glucosinolate and 4-hydroxy-indol-3-ylmethyl glucosinolate have been detected in all plant parts [30,33]. In addition, Huang et al. found indications of an unidentified glucosinolate, which was not 2-propenyl glucosinolate, benzyl glucosinolate or phenylethyl glucosinolate, but was reported as being aromatic . In summary, A. petiolata has been reported to contain glucosinolates derived from homomethionine, dihomomethionine, phenylalanine, homophenylalanine and tryptophan. These glucosinolates may be differentially expressed in different ecotypes or populations of A. petiolata as observed in Arabidopsis . In a study comparing the concentration of defence compounds in A. petiolata collected at different sites, the observed population differences in unspecified glucosinolate content disappeared when field-collected seedlings were transplanted to pots and grown in a greenhouse under controlled conditions. Hence, the site-based variation in total glucosinolate content was apparently due to environmental rather than genetic differences . This illustrates that glucosinolate profiles as well as those of other classes of natural products are regulated and highly variable. The content of alliarinoside has not been found to vary significantly between different populations grown under controlled conditions [39,40]. Changes in the concentration of alliarinoside have been observed in reaction to jasmonic acid treatment  supporting the notion of alliarinoside being a herbivore defence compound. It remains to be clarified whether the alliarinoside content is subjected to regulation in response to other environmental stimuli.
To enable genetic fingerprinting and the investigation of breeding patterns of A. petiolata, microsatellite loci have been identified . A genome sequencing project has not yet been initiated. Reports of chromosome numbers of 2n = 14 or 42 have led to the assumption that n =7 and hence that A. petiolata exists as diploid and hexaploid variants. The diploid variants occur in Western Asia with an accession from Lesser Caucasus being the westernmost diploid A. petiolata described . Hexaploids occur in Central and Western Europe and North America, although a few accessions from the Netherlands and Sweden have also been reported as having 2n = 36 [42,43]. Investigations of A. petiolata from Southeast Europe and Turkey are needed for determining the borders of cytotype distribution . Most investigations of the natural product composition of A. petiolata have been carried out using plants obtained from North America and Western Europe and are therefore assumed to involve hexaploid plants [14,19–21,24,28,30,33,39,40]. Elucidation of the biosynthesis of natural products such as alliarinoside in A. petiolata would be simpler using diploid specimens. It remains to be established whether hexaploid A. petiolata is autopolyploid or allopolyploid, meaning that the polyploidy originates from conspecific parents or from hybridization of two or more species. In the case of allopolyploidy, the natural product profile found in diploids will most probably differ both qualitatively and quantitatively from the profile of polyploids. It is noticeable that while alliarinoside has been detected consistently in all analysed populations from Western Europe and North America [14,21,24,39,40], no trace of alliarinoside was detected in A. petiolata collected in the sub Mediterranean region of South Croatia , a geographic site approaching the diploid distribution area. However, in the latter study volatile aglucones obtained following β-glucosidase treatment of O-glucosides were detected using gas chromatography and gas chromatography–mass spectrometry and 20% of the aglucones in A. petiolata remained unidentified and may have included the aglucone of alliarinoside. Hence, possible differences in metabolite profiles between diploid and hexaploid A. petiolata remain to be resolved.
Cyanogenic glucosides are present in many plants and have traditionally been viewed as phytoanticipins defending the plant against attacking herbivores and pathogens. When the plant tissue is disrupted, the compartmentalized cyanogenic glucoside is brought into contact with hydrolytic enzymes. A β-glucosidase cleaves the glucosidic linkage yielding the corresponding α-hydroxynitrile (cyanohydrin). The α-hydroxynitrile dissociates into a ketone or an aldehyde and hydrogen cyanide (HCN) either catalysed by an α-hydroxynitrile lyase or spontaneously due to the labile nature of the α-hydroxynitrile at pH values above 6. Due to the release of toxic volatile HCN, the process is denoted cyanogenesis. In addition to functioning as defence compounds, cyanogenic glucosides are emerging as storage and transport forms of reduced nitrogen and glucose with turnover into primary metabolism [6,44].
Most of the over 60 known cyanogenic glucosides are derived from the five amino acids valine, leucine, isoleucine, phenylalanine and tyrosine. In addition, Passifloraceae contains cyanogenic glucosides derived from the rare non-protein amino acid 2-cyclopentenylglycine, and nicotinic acid seems to be the precursor in Acalypha indica (Euphorbiaceae) [45,46]. The enzymes catalysing cyanogenic glucoside biosynthesis and the corresponding structural genes were initially identified from Sorghum bicolor (L.) Moench (Poaceae), which synthesizes the tyrosine-derived cyanogenic glucoside dhurrin. Two membrane-anchored multifunctional P450s and a soluble UDP-glucosyl transferase (UGT) catalyse dhurrin biosynthesis in a highly channelled manner [47–55]. The first P450, CYP79A1, converts tyrosine into the corresponding oxime through multiple reaction steps involving two molecular oxygen- and NADPH-dependent N-hydroxylations (Fig. 2). The oxime is further metabolized into the corresponding α-hydroxynitrile through dehydration and an NADPH-dependent C-hydroxylation catalysed by another multifunctional P450, CYP71E1. Finally, UGT85B1 catalyses glucosylation of the α-hydroxynitrile to form the cyanogenic glucoside dhurrin.
The genes encoding the enzymes catalysing the entire biosynthetic pathway have been identified in S. bicolor, Manihot esculenta and Lotus japonicus [12,48,52,54,56–59] (Table 1). M. esculenta and L. japonicus both synthesize two cyanogenic glucosides, lotaustralin and linamarin, derived from isoleucine and valine, respectively. In other species, including Linum usitatissimum, Hordeum vulgare, Prunus dulcis, Trifolium repens, Triglochin maritima and Ribes uva-crispa, the biosynthetic pathway has been partially or fully biochemically elucidated via enzyme activity studies using radiolabelled precursors and intermediates, but most or all of the responsible genes remain to be identified [60–66]. In all studied cases, the classes of intermediates involved in cyanogenic glucoside biosynthesis from different amino acids have been the same as initially found in the S. bicolor system. All presently identified enzymes catalysing the amino acid to oxime conversion belong to the CYP79 family and exhibit high substrate specificity for the amino acid corresponding to the cyanogenic glucosides found in the plant in which the enzyme was identified [3,6,49,56,57] (Table 1). Opposed to this, the second P450 and the UGT catalysing the subsequent biosynthetic steps are promiscuous enzymes capable of metabolizing oximes and hydroxynitriles, respectively, which are derived from various amino acids [6,49,54,55].
Table 1. Overview of P450s involved in the biosynthesis of hydroxynitrile glucosides, glucosinolates and camalexin. In the column stating the metabolic activity of the P450s the substrate and product of the reaction are indicated. The main references are highlighted in bold. In addition, the P450s involved in the biosynthesis of glucosinolates and the tryptophan-derived phytoalexin camalexin have been reviewed in references 8,13,80.
Cyanogenic glucosides are widely distributed in the plant kingdom. Aromatic cyanogenic glucosides are present in pteridophytes, gymnosperms and angiosperms, while aliphatic cyanogenic glucosides are found in the monocot order Poales as well as in eudicots. Accordingly, the ability to synthesize cyanogenic glucosides has been regarded as an ancient trait and it has been assumed that the pathways for aliphatic cyanogenic glucosides evolved from the aromatic pathways . As more genome sequences have become available and biosynthetic genes have been identified from more species, the data obtained suggest recurrent evolution of cyanogenic glucoside biosynthesis. It was recently discovered that in L. japonicus the oxime to nitrile conversion is catalysed by CYP736A2, which is not a CYP71 family member like the other identified nitrile-forming enzymes in cyanogenic glucoside synthesis but belongs to another family in the diverse multifamily CYP71 clan comprising half of all known higher plant P450s [12,67]. Furthermore, the amino acid to oxime catalysing enzymes in L. japonicus, CYP79D3 and CYP79D4, are related but not orthologous to the oxime-metabolizing CYP79s present in S. bicolor and M. esculenta . This strongly indicates that the ability to synthesize cyanogenic glucosides has evolved independently in plants at least twice. The families of CYP79s and CYP71s are absent from gymnosperms and pteridophytes and appear to have arisen in angiosperms . This suggests that biosynthesis of cyanogenic glucosides in the two former taxa is catalysed by P450s from other families than the presently elucidated angiosperm pathways and could thus represent other example(s) of independent recruitment of P450s to catalyse amino acid to oxime conversion and subsequent hydroxynitrile formation. The primary approach for identifying genes involved in biosynthesis of cyanogenic glucoside has been to search for genes showing high sequence homology to the biochemically identified S. bicolor genes. This strategy may be futile in some species, if independent evolution of cyanogenic glucosides is more widespread than previously expected . If the finding of genomic clustering of the cyanogenic glucoside biosynthetic genes in L. japonicus, S. bicolor and M. esculenta is a general feature, this may guide the identification of genes involved in cyanogenic defence pathways in a more diverse range of plant species .
β- and γ-hydroxynitrile glucosides
Several plant species containing aliphatic cyanogenic glucosides also produce structurally related β- and γ-hydroxynitrile glucosides derived from the same amino acid as the cyanogenic α-hydroxynitrile glucosides present [6,45,61]. β- and γ-hydroxynitrile glucosides are not as widespread in the plant kingdom as cyanogenic glucosides but have been described in species within Poaceae, Fabaceae, Rosaceae, Crassulaceae, Grossulariaceae and possibly also Euphorbiaceae [6,45,61], demonstrating that they occur in monocots as well as eudicots. Alliarinoside in A. petiolata is the single known case of γ-hydroxynitrile glucoside occurrence in Brassicaceae, where no β-hydroxynitrile glucosides have been described. The findings of co-occurrence and a common amino acid precursor of cyanogenic glucosides and β- and γ-hydroxynitrile glucosides in a number of species strongly imply that β- and γ-hydroxynitrile glucosides are not only structurally but also biosynthetically related to α-hydroxynitriles . This is supported by evidence that the same isoleucine-derived oxime is a shared intermediate in the biosynthesis of the α-, β- and γ-hydroxynitrile glucosides present in L. japonicus [12,57,68]. Based on current knowledge, different evolutionary scenarios may be proposed for the biosynthesis of β- and γ-hydroxynitrile glucosides. It has been suggested to reflect evolutionary diversification of the second P450 in the cyanogenic glucoside pathway . Based on the suite of leucine-derived hydroxynitrile glucosides present in H. vulgare, it has been proposed that a single P450 is capable of hydroxylating all individual carbon atoms in the nitrile intermediate . Subsequent glucosylation, either directly or following dehydration reactions and additional C-hydroxylation(s) of the nitrile intermediate, could suffice to yield the entire range of saturated and unsaturated α-, β- and γ-hydroxynitrile glucosides in H. vulgare. This biosynthetic scheme can also be adopted to explain the isoleucine-derived hydroxynitrile glucosides in species of Ribes and Rhodiola . However, it is also possible that parallel oxime-metabolizing P450s could be operating in the respective pathways for cyanogenic glucoside and β- and γ-hydroxynitrile glucoside formation . This notion is supported by the recent discovery that, in L. japonicus, the oxime-metabolizing P450 in the β- and γ-hydroxynitrile glucoside pathway differs from that catalysing the cyanogenic glucoside pathway (Søren Bak, personal communication). The existence of parallel oxime-metabolizing P450s could originate from duplication of the gene encoding the oxime-metabolizing P450 in cyanogenic glucoside biosynthesis and subsequent evolution of one of the paralogues to catalyse β- and γ-hydroxynitrile formation . However, it is also possible that a P450 was recruited from outside the cyanogenic glucoside pathway to catalyse β- and γ-hydroxynitrile glucoside formation. Such P450s would be expected to belong to the CYP71 clan but not necessarily to be a member of the CYP71 family.
In summary, co-occurrence of α-, β- and γ-hydroxynitrile glucosides could reflect evolutionary expansion of the substrate specificity of a single oxime-metabolizing P450, paralogous oxime-metabolizing P450s, or recruitment of a new P450 to catalyse β- and γ-hydroxynitrile glucoside biosynthesis. In light of the evidence of recurrent evolution of cyanogenic glucosides , different evolutionary events may turn out to be the underlying cause in different species.
Hydrolysis of the β-glucosidic bond in β- and γ-hydroxynitrile glucosides results in formation of stable hydroxynitriles and accordingly no cyanogenesis. L. japonicus contains two β-glucosidases capable of hydrolysing the β- and γ-hydroxynitrile glucosides in this species, but only one of these enzymes also hydrolyses the two present α-hydroxynitrile glucosides in vivo. Hence, catabolism of cyanogenic and acyanogenic hydroxynitrile glucosides is potentially uncoupled in L. japonicus. Accordingly, the hydroxynitrile glucosides may play separate biological roles . Not much is known about the function of β- and γ-hydroxynitrile glucosides. They may serve as nitrogen storage compounds in parallel with cyanogenic glucosides [6,44]. γ-Hydroxynitrile glucosides are dominant compared with α- and β-hydroxynitrile glucosides in H. vulgare and in Ribes and Rhodiola species. Biosynthesis of the γ-hydroxynitrile glucosides is energetically more costly than cyanogenic glucoside biosynthesis, as γ-hydroxynitrile glucoside formation requires one or two additional hydroxylations. Thus, their strong dominance implies a biological function . As mentioned above, H. vulgare contains a suite of leucine-derived β- and γ-hydroxynitrile glucosides and a single cyanogenic glucoside in the leaves. The absence of an endogenous β-glucosidase, except in the endosperm of the germinating seed, indicates that their biological role in H. vulgare does not rely on hydrolysis in planta . However, as some herbivores contain β-glucosidases capable of cleaving hydroxynitrile glucosides , a role in plant defence involving hydrolysis of the β-glucosidic linkage is most likely. Investigations of H. vulgare during infection with the powdery mildew-causing fungus Blumeria graminis have indicated a protective role of β- and γ-hydroxynitrile glucosides in comparison with the α-hydroxynitrile glucoside . Hence, in the light of powdery mildew infections it has been suggested advantageous for H. vulgare to produce β- and γ- hydroxynitrile glucosides instead of an α-hydroxynitrile glucoside to prevent attack by the fungus . This may explain why β- and γ-hydroxynitrile glucosides constitute by far the major part of the hydroxynitrile glucoside content in commercial H. vulgare cultivars [65,71]. It has been suggested that the obligate biotrophic fungus uses the hydroxynitrile glucosides in host recognition and as a nutrient source of reduced nitrogen and glucose [44,71]. However, the mechanisms in powdery mildew–H. vulgare hydroxynitrile glucoside interactions remain to be further explored.
In recent investigations of hydroxynitrile glucosides in L. japonicus, it was found that following β-glucosidase cleavage the aglucone of the γ-hydroxynitrile glucoside rhodiocyanoside A may cyclize into a toxic lactone (Søren Bak, personal communication). A mechanism for lactone formation from γ-hydroxynitrile glucosides has previously been suggested . This would involve conversion of the nitrile group of the aglucone into a carboxyl group by the action of a nitrilase. Subsequent cyclization into the lactone may be spontaneous or enzymatically catalysed and was shown in L. japonicus to be dependent on the presence of the double bond and the (Z)-configuration of rhodiocyanoside A (Søren Bak, personal communication). Considering the structure of alliarinoside in A. petiolata, lactone formation may also occur upon hydrolysis of this unsaturated γ-hydroxynitrile glucoside. Furanones (e.g. lactones) from plants, algae and fungi are known to have antimicrobial and antifungal properties [72,73]. In agreement with this, growth of mycorrhizal fungi is inhibited in A. petiolata inhabited soil, thereby diminishing the emergence, growth and survival of mycorrhizal plant species . The antifungal effect has also been reported in soils treated with extracts of A. petiolata enriched in alliarinoside and unspecified flavonoids but devoid of glucosinolates. Hence, the authors attributed the mycorrhizal inhibition to flavonoids . However, it is likely that alliarinoside contributed to the observed allelopathic effect of A. petiolata, possibly through anti-mycorrhizal lactone formation.
Glucosinolates and their biosynthesis
Glucosinolates are S-glucopyranosyl thiohydroxymates occurring in the Brassicales order and in the unrelated genus Drypetes (Euphorbiaceae, Malpighiales) [74,75]. Like cyanogenic glucosides, they play important roles as defence compounds following hydrolysis by myrosinases (thioglucoside glucohydrolases). The hydrolysis products are glucose and an unstable aglucone, which rearranges to form bioactive isothiocyanates, nitriles and other products . The type of breakdown products formed depends on the reaction conditions, the structure of the glucosinolate and its subsequent modifications, and the presence of cofactors and catalysing proteins [13,76,77].
Elucidation of the biosynthetic pathway has benefited greatly from the prior elucidation of the genes encoding the cyanogenic glucoside pathway and from the molecular and genetic tools tied to the model plant Arabidopsis thaliana (Brassicaceae), which produce about 40 different glucosinolates, mainly derived from methionine and tryptophan. A recent special issue of Phytochemistry Reviews focuses on many aspects of glucosinolate research, including their role in invasive plants such as A. petiolata [78,79].
Overall, glucosinolate biosynthesis may involve three major steps: (a) chain elongation of the amino acid precursor; (b) formation of the glucosinolate core structure; and (c) secondary modifications of the amino acid side chain [8,80]. The more than 120 known glucosinolates are derived from only eight different amino acids – alanine, leucine, isoleucine, valine, methionine, phenylalanine, tyrosine and tryptophan . The high structural diversity is due to secondary modifications such as hydroxylations, desaturations and glucosylations, and in some cases chain elongation of methionine and phenylalanine by insertion of methylene groups prior to glucosinolate core structure formation.
The pathway for the core structure involves intermediates common to all glucosinolates (Fig. 3). As in cyanogenic glucoside biosynthesis, the amino acid precursors are converted into oximes by CYP79 family enzymes, which are the only side-chain-specific enzymes in the core glucosinolate pathway [8,81,82]. Another P450 of the CYP83 family oxidizes the oxime, thereby forming a highly reactive product proposed to be either an aci-nitro compound or the dehydrated analogue, a nitrile oxide [83,84]. Based on the level of amino acid sequence homology, CYP83s ought to have been assigned to the CYP71 family. However, to avoid confusion the originally published nomenclature has been kept [3,67]. CYP79 homologues catalysing oxime formation in glucosinolate biosynthesis have been identified in A. thaliana and Sinapis alba [81,82,85–89]. The function of five of the seven CYP79 homologues in the A. thaliana genome has been characterized. They are all involved in glucosinolate biosynthesis and show narrow substrate specificity for particular amino acids or their chain-elongated amino acid derivatives  (Table 1).
The ability to produce cyanogenic glucosides has generally been found to be mutually exclusive to the ability to synthesize glucosinolates. CYP79-catalysed oxime formation from amino acids and the subsequent CYP71/83-catalysed conversion of the oximes are common features of the pathways of cyanogenic glucoside and glucosinolate biosynthesis. Accordingly, it has been hypothesized that genes have been recruited from the pathway of cyanogenic glucoside biosynthesis into the more recently evolved glucosinolate pathway [3,8]. It has been suggested that glucosinolates evolved as a means of detoxifying intermediates in a mutated cyanogenic glucoside pathway . A mutation in the oxime-metabolizing CYP71 could result in a toxic product such as the reactive aci-nitro compound or nitrile oxide, which the plant subsequently had to find means to process. Accordingly, the enzymes downstream of CYP83 in the core glucosinolate pathway may have been recruited from other pathways to promote the detoxification. In accordance with this hypothesis, glucosyltransferases as well as sulfotransferases are mediators of detoxification processes occurring widely in nature . However, in light of the recent evidence of recurrent evolution of cyanogenic glucosides, a likely scenario may be that the glucosinolates did not originate from the cyanogenic glucoside pathway . It is equally likely that the pathways for cyanogenic glucosides and glucosinolates evolved independently from the predisposition to produce oximes by CYP79 homologues and the subsequent need to convert the oximes into less labile and toxic compounds. CYP79 encoding genes are present in species producing neither glucosinolates nor cyanogenic glucosides, e.g. Glycine max and Populus trichocarpa, and some cyanogenic species harbour additional CYP79 genes unlikely to be involved in cyanogenic glucoside synthesis, e.g. L. japonicus . In members of Brassicaceae, some of these CYP79s are involved in oxime formation directed towards biosynthesis of tryptophan-derived phytoalexins  (Table 1). The function of the majority of these additional CYP79 genes remains to be resolved . Hence, CYP79s appear to be characterized by the ability to convert amino acids into oximes with high substrate specificity, but they are not confined to the pathways of cyanogenic glucosides or glucosinolates and recent phylogenetic analysis shows that CYP79s involved in glucosinolate and cyanogenic glucoside biosynthesis are not orthologous . The presence of glucosinolates in Brassicales as well as in Drypetes strongly suggests that glucosinolate biosynthesis evolved independently twice . This could represent different events of recruitment of P450s from the CYP79 family and CYP71 clan equivalent to the repeated evolution of cyanogenic glucosides. The enzymes downstream of CYP83 in the glucosinolate core pathway could have been recruited for detoxification purposes . Acquisition of more genomic sequences, expressed sequence tags and cDNA sequences from glucosinolate-containing species and subsequent characterization of the responsible gene products may help resolve the question of a relationship or independent evolution of the glucosinolate and cyanogenic glucoside pathways. In particular, it would be interesting to elucidate from where the genes for glucosinolate biosynthesis were recruited in Drypetes as well as in Brassicales species from other families than Brassicaceae. This may clarify whether the CYP79s and CYP83s in glucosinolate biosynthesis evolved from cyanogenic CYP79s and CYP71s, respectively, or if they are derived from other CYP79 and CYP71 homologues recruited into the pathway to catalyse amino acid and oxime metabolism. Functional characterization of CYP79s in species not producing glucosinolates and cyanogenic glucosides could shed further light on this issue.
Secondary side chain modification of the glucosinolate core structure occurs in organ- and developmental-specific patterns and may also differ between ecotypes [34,35,38]. Biochemical and genetic information about the different secondary modifications is generally limited. However, the enzymes and genes responsible for a major set of modifications in methionine-derived glucosinolates have been identified. These modifications involve S-oxygenation followed by cleavage of the terminal methylsulfinyl group and its replacement either by a terminal hydroxyl group or a vinylic double bond on the side chain (Fig. 4). In A. thaliana, five flavin monooxygenases (FMOs) have been found to catalyse the conversion of methylthioalkyl glucosinolates into methylsulfinylalkyl glucosinolates [90,91]. Two 2-oxoacid-dependent dioxygenases, AOP2 and AOP3, have been shown to metabolize the S-oxygenated glucosinolates in A. thaliana. Their names reflect their function as alkenyl/hydroxy(OH)alkyl producing enzymes. AOP2 catalyses the conversion to alkenyl glucosinolates, whereas AOP3 forms hydroxyalkyl glucosinolates [92,93] (Fig. 4). As the names indicate, a third enzyme, AOP1, has also been identified, but its function remains to be determined. Homologous AOP2 and AOP1 genes are present in Brassica oleracea (Brassicaceae), which also produces methionine-derived glucosinolates .
Suggested evolution of the pathway for alliarinoside from glucosinolate biosynthesis
A. petiolata is the only plant species in which a γ-hydroxynitrile glucoside has been reported to co-occur with glucosinolates. All aspects of alliarinoside production remain to be experimentally elucidated. However, by integrating the current knowledge about biosynthesis of hydroxynitrile glucosides and glucosinolates with biochemical reasoning, pathways for alliarinoside biosynthesis can be envisioned (Fig. 5). The reported occurrence of 2-propenyl glucosinolate and 3-butenyl glucosinolate in A. petiolata strongly implies the presence of a methionine chain-elongation machinery homologous to the one identified in A. thaliana . 2-Propenyl glucosinolate is derived from homomethionine, whereas 3-butenyl glucosinolate is formed from dihomomethionine. In A. thaliana, both methionine derivatives are metabolized into the corresponding oxime by CYP79F1. CYP79F1 metabolizes all chain-elongated methionine derivatives, whereas CYP79F2 is specific for long-chain-elongated pentahomomethionine and hexahomomethionine [81,88,95]. No methionine-derived hydroxynitrile glucosides have previously been described and the ability to convert methionine derivatives into oximes appears to be a specific feature of Brassicales. Accordingly, it is highly likely that A. petiolata contains a CYP79 enzyme, probably related to members of the CYP79F subfamily, which converts homomethionine into 4-(methylthio)butanal oxime. The subsequent steps in glucosinolate formation are expected to proceed using intermediates in core structure formation identical to those described in other glucosinolate accumulating species (Fig. 5). A genome or transcriptome sequencing program in A. petiolata would facilitate identification of the genes involved.
The β- and γ-hydroxynitrile glucosides known from other species are all derived from leucine or isoleucine [45,61]. Considering the structure of alliarinoside with the double bond and lack of branching (Fig. 1), neither leucine nor isoleucine would appear likely precursors. We suggest that alliarinoside is derived from homomethionine and that 4-(methylthio)butanal oxime is a shared intermediate in 2-propenyl glucosinolate and alliarinoside biosynthesis. The first committed step in the hydroxynitrile pathway is envisioned to be catalysed by a P450, which has been recruited to catalyse the oxime to nitrile conversion. Although the CYP71 family contains most known P450s converting oximes to nitriles, it is possible that this P450 has evolved from a different P450 family. The identification of CYP736A2 as the oxime-metabolizing enzyme in L. japonicus shows that a non-CYP71 family member has evolved to catalyse this type of reaction . Regardless of the origin, the P450 is thought to be multifunctional and catalyse nitrile formation and subsequent C-hydroxylation as seen for CYP736A2 and the CYP71E subfamily members involved in hydroxynitrile glucoside biosynthesis (Table 1). It is also possible that different P450s catalyse the dehydration and hydroxylation reactions. Recent investigations of the biosynthesis of the γ-hydroxynitrile glucoside rhodiocyanoside D in L. japonicus indicate that C-hydroxylation of the nitrile intermediate in this pathway may be catalysed by a different enzyme than the P450 catalysing dehydration of the oxime (Søren Bak, personal communication). A. thaliana contains the phytoalexin camalexin derived from tryptophan. In the biosynthesis of camalexin, a tryptophan-derived oxime is converted into the corresponding nitrile by CYP71A13 . The subsequent nitrile-metabolizing enzyme in the camalexin pathway remains unidentified. Hence, there are indications that some oxime-metabolizing P450s are not multifunctional and catalyse dehydration of the oxime only. Accordingly, oxime to nitrile conversion and subsequent hydroxynitrile formation may be catalysed by different P450s in the proposed pathway for alliarinoside biosynthesis. The C-hydroxylation may take place at different carbon atoms as indicated by the bifurcation of the pathway (Fig. 5). Whether the two suggested hydroxynitriles coincide in planta is currently an open question. Both are possible intermediates in alliarinoside biosynthesis and not mutually exclusive. CYP71s capable of hydroxylating any of the carbon atoms in the nitrile intermediate have been hypothesized to be involved in hydroxynitrile glucoside biosynthesis in H. vulgare and species of Ribes and Rhodiola [6,65]. A similar promiscuity or broad specificity of the P450 in A. petiolata is possible. Alternatively, different P450s may be working in parallel in each of the suggested branches. Formation of the double bond is thought to result from dehydration of the hydroxynitrile(s) or at a later step (Fig. 5). The dehydrating enzyme may be a P450, possibly identical or related to the P450 catalysing dehydration of the oxime. The terminal methylsulfinyl group is envisioned to be S-oxygenated and replaced by a hydroxyl group through reactions similar to the secondary modifications of methionine-derived glucosinolates. Hence, FMO- and AOP3-like enzyme activities are suggested to be present. Little is known about FMOs in plants, but sequence homology based genomic search has identified various FMO homologues in distantly related plant species such as Oryza satica, Populus trichocarpa and Thalaginella halophila [90,91,96]. This indicates that FMOs are widespread in vascular plants. The five FMOs, FMOGS-OX1–5, which have been shown to catalyse S-oxygenation in glucosinolate modification, have been found to cluster in a separate subclade within a clade of sequences from distantly related species . FMOGS-OX1 was shown to catalyse conversion of methylthiobutane glucosinolate and desulfo-methylthiobutane glucosinolate with equal efficiency, whereas activity was not observed towards any upstream intermediates . This led to the conclusion that FMOGS-OX1 requires an S-glucose group but may function regardless of the sulfate group. The specificity of the other identified FMOGS-OX enzymes with respect to S-glucose requirement has not been determined. In a study of γ-glutamyl peptidases involved in glucosinolate core structure biosynthesis, A. thaliana mutants were found to accumulate core structure intermediates with sulfinyl side chains . This indicates S-oxygenation of intermediates upstream of glucosylation, which could be catalysed by FMOGS-OX(s) with no S-glucose group requirement. FMOGS-OX1, FMOGS-OX2, FMOGS-OX3 and FMOGS-OX4 have been found to have broad substrate specificity and catalyse methylthioalkyl glucosinolate metabolism independently of chain length. FMOGS-OX5, on the other hand, seemed specific for long-chain methylthiobutane glucosinolate . Thus, the FMO proposed to be involved in alliarinoside biosynthesis may be an FMOGS-OX homologue which has no requirement for the presence of an S-glucose group in its substrate. Alternatively, this conversion may involve an FMO recruited from outside the glucosinolate pathway.
Alk(en)yl and hydroxyalkyl glucosinolates accumulate in many Brassicaceae species, including A. petiolata and the closely related Thlaspi arvense [74,98]. This indicates that AOP2 and AOP3 activities are common in Brassicaceae. The genes encoding AOP1, AOP2 and AOP3 in A. thaliana were generated by two different gene duplication events, which have been suggested to predate A. thaliana speciation . Thus, AOP homologues could be present in A. petiolata. In secondary modification of glucosinolates AOP3 catalyses conversion of glucosidic substrates. However, the requirement for an S-glucose residue in the substrate has not been investigated. There are some indications that AOP3 is specific for the homomethionine-derived 3-methylsulfinylalkyl glucosinolate and does not catalyse conversion of a dihomomethionine-derived substrate . It is suggested that in A. petiolata an AOP3 homologue metabolizes an aglucone derived from homomethionine. The FMO and AOP3 homologue could potentially convert different substrates in the suggested alliarinoside pathway, as indicated in Fig. 5. The resulting terminal hydroxyl group is envisioned to be glucosylated by a UGT, thereby forming alliarinoside. Other hydroxylated intermediates in the pathway could also be imagined to be glucosylated. This could yield 4-(β-d-glucopyranosyloxy)-3-hydroxybutanenitrile, 3-(β-d-glucopyranosyloxy)-4-hydroxybutanenitrile and 4-(methylthio)-2-(β-d-glucopyranosyloxy)butanenitrile, which are γ-, β- and α-hydroxynitrile glucosides, respectively. The methylthio group in 4-(methylthio)-2-(β-d-glucopyranosyloxy)butanenitrile might be subject to secondary modification by an FMO and homologues of AOP2 or AOP3. This would potentially result in two different α-hydroxynitrile glucosides, 2-(β-d-glucopyranosyloxy)-but-3-enenitrile and 2-(β-d-glucopyranosyloxy)-4-hydroxybutanenitrile. The possible existence of the suggested five additional hydroxynitrile glucosides needs verification by experimentation.
Unidentified source of cyanide in A. petiolata
The cyanide release from A. petiolata reported by Cipollini and Gruner  is a highly interesting observation. Cyanide release strongly indicates the presence of a cyanogenic glucoside or a cyanolipid. However, the occurrence of cyanolipids is thought to be restricted to Sapindaceae and Hippocastanaceae [6,7] and with very few known exceptions cyanogenic glucosides do not co-occur with glucosinolates [99–105]. Using colorimetric detection of cyanide, homogenized leaves of A. petiolata were found to release a mean of 44 ± 2 p.p.m. diffusible hydrogen cyanide. Less cyanide release was detected from roots. In comparison, young seedlings of the cyanogenic species Sorghum sudanese were found to release five times more hydrogen cyanide than young seedlings of A. petiolata, whereas four other Brassicaceae species released 25–150 times less . The source of hydrogen cyanide release in A. petiolata has not been determined. Considering the presence of the γ-hydroxynitrile glucoside alliarinoside , the possible co-occurrence of a cyanogenic α-hydroxynitrile glucoside is intriguing. Experimental demonstration of the presence of cyanogenic glucosides in A. petiolata may serve to further support recurrent evolution of cyanogenic glucoside biosynthesis and reveal additional information on the possible evolutionary relationship between glucosinolates and cyanogenic glucosides. In this review, we suggest that alliarinoside biosynthesis evolved from the predisposition to produce a homomethionine-derived oxime. It is proposed that a P450 was recruited outside the glucosinolate pathway to catalyse oxime to nitrile conversion. A cyanogenic glucoside could also be the result of P450 recruitment after glucosinolate biosynthesis was established. Putative intermediates in the suggested alliarinoside pathway could be metabolized into a homomethionine-derived α-hydroxynitrile glucoside as outlined in Fig. 5. Another possibility is the presence of a cyanogenic glucoside derived from another oxime produced by A. petiolata. Glucosinolates derived from phenylalanine, homophenylalanine, homomethionine, dihomomethionine and tryptophan have been detected in A. petiolata [28,30–33,36,37]. This implies that A. petiolata is able to form oximes from all of these amino acids. Two Brassicales members, Carica papaya and Carica quercifolia (Caricaceae), produce benzylglucosinolate and the cyanogenic glucoside prunasin, both of which are derived from phenylalanine [99,100,102–105]. However, biosynthetic intermediates and the involved enzymes have not been identified. Thus, it remains to be resolved how glucosinolate and cyanogenic glucoside biosynthesis are related in these species and whether the information gained may specifically apply to Caricaceae or be useful for elucidation of hydroxynitrile metabolism in Brassicaceae species like A. petiolata.
Alliarinoside is also a hypothetical source of hydrogen cyanide release. Generally, γ-hydroxynitrile glucosides do not release HCN upon hydrolysis, because the aglucones formed are not cyanohydrins. However, the structure of alliarinoside contains a double bond that might give rise to formation of the corresponding epoxide. Epoxides of hydroxynitrile glucosides are known from other species and one of them, sarmentosin epoxide, was shown to slowly release HCN in aqueous solution [45,106]. Presumably, hydrolysis of the oxirane group led to cyanohydrin formation and subsequent HCN release . However, the presence of alliarinoside epoxide in A. petiolata has not been reported.
A third possibility is that degradation products from glucosinolates are responsible for the positive colorimetric response observed in the HCN assay . During degradation of some glucosinolates, specifier proteins can promote rearrangement of the aglucone released by myrosinase into a number of different products [76,77]. Depending on the side chain of the aglucone and the present cofactors and proteins, these include simple nitriles, epithionitriles and thiocyanates. In addition, isothiocyanates and oxazolidine-2-thiones can be formed spontaneously. The specificity of the different methods used for cyanide detection has been debated [107–109]. The Lambert method applied to A. petiolata is a variant of the colorimetric method based on the König synthesis of pyridine dyes [15,110,111]. The other colorimetric methods used are variants relying on the same basic principle. Cyanide is oxidized by an appropriate oxidizing agent to form a cyanogen halide, which reacts with pyridine and a coupling agent. The resulting dye formation is quantified by spectrophotometry. The Lambert method was described as being specific for the cyanide ion and interferences virtually non-existing, but this was not addressed experimentally [107,111]. Compounds which may produce CN or SCN radicals have been assumed to be detectable in variants of the König reaction assay . Experimental evidence shows that thiocyanates are detected to different extents depending on the assay details [112–115]. However, by including a diffusion step in which volatile hydrogen cyanide is trapped in an alkaline solution prior to the König reaction, interference by non-volatile hydrogen thiocyanate can be eliminated . The method applied to A. petiolata included such a diffusion step . Accordingly, thiocyanates are assumed not to have been the source of the positive response detected in the assay. However, nitriles cannot be excluded as interferents. In a study of nitriles and isonitriles as interfering agents in cyanide assays, phenylacetonitrile and other nitriles were found to show positive interference . The interference was observed in direct spectrophotometric assays and to lesser extent when a prior distillation or diffusion step was introduced. The level of interference generally increased with the ratio of interfering compound to cyanide. The analytical procedure used for these determinations differed somewhat from the Lambert method with diffusion. The effect of the presence of nitriles thus needs to be addressed experimentally. In this context it is important to notice that phenylacetonitrile is a degradation product from benzylglucosinolate, both of which have been detected in A. petiolata [28,31,32,36]. Furthermore, it remains to be investigated whether other degradation products from the large amounts of glucosinolates found in A. petiolata may be the reason for the positive colour reaction in the cyanide assay.
By integrating the current knowledge about biosynthesis of hydroxynitrile glucosides and glucosinolates, possible pathways for alliarinoside biosynthesis have been outlined and discussed. We argue that biosynthesis of alliarinoside may be the first known example of a hydroxynitrile glucoside evolved from the glucosinolate pathway. The ability to produce homomethionine and the corresponding oxime is proposed as the basis for recruitment of a P450 catalysing oxime to nitrile formation in a novel hydroxynitrile glucoside pathway. Furthermore, homologues to enzymes involved in secondary modification of glucosinolates are proposed to have evolved to metabolize aglucone intermediates in the alliarinoside pathway. Experimental elucidation of the biosynthesis of alliarinoside and other putative hydroxynitrile glucosides present in A. petiolata will resolve whether the suggested oxime-metabolizing P450 is able to hydroxylate different carbon atoms in the backbone of the nitrile intermediate or whether different P450s are working in parallel to synthesize different hydroxynitriles from the same oxime.
At present, the reported hydrogen cyanide release in A. petiolata raises more questions than it answers. It is essential that this finding is verified or explained by interference resulting from the presence of organic nitriles. Identification of the source would serve to shed new light on the chemical ecology of A. petiolata and could also have important implications for our understanding of nitrile glucoside metabolism and diversification. The possible presence of cyanogenic glucosides in A. petiolata and elucidation of their structures and routes of biosynthesis may provide new insights on recurrent and independent evolution of the pathways for cyanogenic glucoside and glucosinolate synthesis. The new visions on recurrent evolution of hydroxynitrile glucoside biosynthesis recently published  have served as a great source of inspiration behind this review. Elucidation of the pathway for biosynthesis of alliarinoside and other putative hydroxynitrile glucosides in A. petiolata is envisioned to offer significant new knowledge on the emerging picture of P450 functional dynamics as the basis for recurrent evolution of pathways for synthesis of bioactive natural products.
We thank Drs Nanna Bjarnholt and Fred Rook for fruitful discussions and Associate Professor Mohammed Saddik Motawia for advice with respect to the chemical nomenclature. BLM acknowledges financial support from the Villum Foundation research centre ‘Pro-Active Plants’, from the UNIK Center for Synthetic Biology funded by the Danish Ministry for Science, Technology and Innovation, and from a research grant from the Danish Research Council for Independent Research/Technology and Production Sciences. TF acknowledges a PhD stipend granted by the Faculty of Life Sciences, University of Copenhagen.