Application of machine learning models to animal health pharmacovigilance: A proof-of-concept study.

Machine learning (ML) models were applied to pharmacovigilance (PV) data in a two-component proof-of-concept study. PV data were partitioned into Training, Validation, and Holdout datasets for model training and selection. During the first component ML models were challenged to identify factors in individual case safety reports (ICSRs) involving spinosad and neurological and ocular clinical signs. The target feature for the models were these clinical signs that were disproportionately reported for spinosad. The endpoints were normalized coefficient values representing the relationship between the target feature and ICSR free text fields. The deployed model accurately identified the risk factors "demodectic," "demodicosis," and "ivomec." In the second component, the ML models were trained to identify high quality and complete ICSRs free of confounders. The deployed model was presented with an external Test dataset of six ICSRs, one that was complete, of high quality, and devoid of confounders, and five that were not. The endpoints were model-generated probabilities for the ICSRs. The deployed ML model accurately identified the ICSR of interest with a greater than 10-fold higher probability score. Although narrow in scope, the study supports further investigation and potential application of ML models to animal health PV data.

Regioselectivity mechanism of the Thunbergia alata Δ6-16:0-acyl carrier protein desaturase.

Plant plastidial acyl-acyl carrier protein (ACP) desaturases are a soluble class of diiron-containing enzymes that are distinct from the diiron-containing integral membrane desaturases found in plants and other organisms. The archetype of this class is the stearoyl-ACP desaturase which converts stearoyl-ACP into oleoyl (18:1Δ9cis)-ACP. Several variants expressing distinct regioselectivity have been described including a Δ6-16:0-ACP desaturase from black-eyed Susan vine (Thunbergia alata). We solved a crystal structure of the T. alata desaturase at 2.05 Å resolution. Using molecular dynamics (MD) simulations, we identified a low-energy complex between 16:0-ACP and the desaturase that would position C6 and C7 of the acyl chain adjacent to the diiron active site. The model complex was used to identify mutant variants that could convert the T. alata Δ6 desaturase to Δ9 regioselectivity. Additional modeling between ACP and the mutant variants confirmed the predicted regioselectivity. To validate the in-silico predictions, we synthesized two variants of the T. alata desaturase and analyzed their reaction products using gas chromatography-coupled mass spectrometry. Assay results confirmed that mutants designed to convert T. alata Δ6 to Δ9 selectivity exhibited the predicted changes. In complementary experiments, variants of the castor desaturase designed to convert Δ9 to Δ6 selectivity lost some of their Δ9 desaturation ability and gained the ability to desaturate at the Δ6 position. The computational workflow for revealing the mechanistic understanding of regioselectivity presented herein lays a foundation for designing acyl-ACP desaturases with novel selectivities to increase the diversity of monoenes available for bioproduct applications.

Castor Stearoyl-ACP Desaturase Can Synthesize a Vicinal Diol by Dioxygenase Chemistry.

In previous work, we identified a triple mutant of the castor (Ricinus communis) stearoyl-Acyl Carrier Protein desaturase (T117R/G188L/D280K) that, in addition to introducing a double bond into stearate to produce oleate, performed an additional round of oxidation to convert oleate to a trans allylic alcohol acid. To determine the contributions of each mutation, in this work we generated individual castor desaturase mutants carrying residue changes corresponding to those in the triple mutant and investigated their catalytic activities. We observed that T117R, and to a lesser extent D280K, accumulated a novel product, namely erythro-9,10-dihydroxystearate, that we identified via its methyl ester through gas chromatography-mass spectrometry and comparison with authentic standards. The use of 18O2 labeling showed that the oxygens of both hydroxyl moieties originate from molecular oxygen rather than water. Incubation with an equimolar mixture of 18O2 and 16O2 demonstrated that both hydroxyl oxygens originate from a single molecule of O2, proving the product is the result of dioxygenase catalysis. Using prolonged incubation, we discovered that wild-type castor desaturase is also capable of forming erythro-9,10-dihydroxystearate, which presents a likely explanation for its accumulation to ∼0.7% in castor oil, the biosynthetic origin of which had remained enigmatic for decades. In summary, the findings presented here expand the documented constellation of di-iron enzyme catalysis to include a dioxygenase reactivity in which an unactivated alkene is converted to a vicinal diol.

Amino Acid Change in an Orchid Desaturase Enables Mimicry of the Pollinator's Sex Pheromone.

Mimicry illustrates the power of selection to produce phenotypic convergence in biology [1]. A striking example is the imitation of female insects by plants that are pollinated by sexual deception of males of the same insect species [2-4]. This involves mimicry of visual, tactile, and chemical signals of females [2-7], especially their sex pheromones [8-11]. The Mediterranean orchid Ophrys exaltata employs chemical mimicry of cuticular hydrocarbons, particularly the 7-alkenes, in an insect sex pheromone to attract and elicit mating behavior in its pollinators, males of the cellophane bee Colletes cunicularius [11-13]. A difference in alkene double-bond positions is responsible for reproductive isolation between O. exaltata and closely related species, such as O. sphegodes [13-16]. We show that these 7-alkenes are likely determined by the action of the stearoyl-acyl-carrier-protein desaturase (SAD) homolog SAD5. After gene duplication, changes in subcellular localization relative to the ancestral housekeeping desaturase may have allowed proto-SAD5's reaction products to undergo further biosynthesis to both 7- and 9-alkenes. Such ancestral coproduction of two alkene classes may have led to pollinator-mediated deleterious pleiotropy. Despite possible evolutionary intermediates with reduced activity, amino acid changes at the bottom of the substrate-binding cavity have conferred enzyme specificity for 7-alkene biosynthesis by preventing the binding of longer-chained fatty acid (FA) precursors by the enzyme. This change in desaturase function enabled the orchid to perfect its chemical mimicry of pollinator sex pheromones by escape from deleterious pleiotropy, supporting a role of pleiotropy in determining the possible trajectories of adaptive evolution.

Remote control of regioselectivity in acyl-acyl carrier protein-desaturases.

Regiospecific desaturation of long-chain saturated fatty acids has been described as approaching the limits of the discriminatory power of enzymes because the substrate entirely lacks distinguishing features close to the site of dehydrogenation. To identify the elusive mechanism underlying regioselectivity, we have determined two crystal structures of the archetypal Δ9 desaturase from castor in complex with acyl carrier protein (ACP), which show the bound ACP ideally situated to position C9 and C10 of the acyl chain adjacent to the diiron active site for Δ9 desaturation. Analysis of the structures and modeling of the complex between the highly homologous ivy Δ4 desaturase and ACP, identified a residue located at the entrance to the binding cavity, Asp280 in the castor desaturase (Lys275 in the ivy desaturase), which is strictly conserved within Δ9 and Δ4 enzymes but differs between them. We hypothesized that interaction between Lys275 and the phosphate of the pantetheine, seen in the ivy model, is key to positioning C4 and C5 adjacent to the diiron center for Δ4 desaturation. Mutating castor Asp280 to Lys resulted in a major shift from Δ9 to Δ4 desaturation. Thus, interaction between desaturase side-chain 280 and phospho-serine 38 of ACP, approximately 27 Å from the site of double-bond formation, predisposes ACP binding that favors either Δ9 or Δ4 desaturation via repulsion (acidic side chain) or attraction (positively charged side chain), respectively. Understanding the mechanism underlying remote control of regioselectivity provides the foundation for reengineering desaturase enzymes to create designer chemical feedstocks that would provide alternatives to those currently obtained from petrochemicals.

Stearoyl-acyl carrier protein desaturases are associated with floral isolation in sexually deceptive orchids.

The orchids Ophrys sphegodes and O. exaltata are reproductively isolated from each other by the attraction of two different, highly specific pollinator species. For pollinator attraction, flowers chemically mimic the pollinators' sex pheromones, the key components of which are alkenes with different double-bond positions. This study identifies genes likely involved in alkene biosynthesis, encoding stearoyl-acyl carrier protein (ACP) desaturase (SAD) homologs. The expression of two isoforms, SAD1 and SAD2, is flower-specific and broadly parallels alkene production during flower development. SAD2 shows a significant association with alkene production, and in vitro assays show that O. sphegodes SAD2 has activity both as an 18:0-ACP Δ(9) and a 16:0-ACP Δ(4) desaturase. Downstream metabolism of the SAD2 reaction products would give rise to alkenes with double-bonds at position 9 or position 12, matching double-bond positions observed in alkenes in the odor bouquet of O. sphegodes. SAD1 and SAD2 show evidence of purifying selection before, and positive or relaxed purifying selection after gene duplication. By contributing to the production of species-specific alkene bouquets, SAD2 is suggested to contribute to differential pollinator attraction and reproductive isolation among these species. Taken together, these data are consistent with the hypothesis that SAD2 is a florally expressed barrier gene of large phenotypic effect and, possibly, a genic target of pollinator-mediated selection.

Metabolic engineering of seeds can achieve levels of omega-7 fatty acids comparable with the highest levels found in natural plant sources.

Plant oils containing ω-7 fatty acids (FAs; palmitoleic 16:1Δ(9) and cis-vaccenic 18:1Δ(11)) have potential as sustainable feedstocks for producing industrially important octene via metathesis chemistry. Engineering plants to produce seeds that accumulate high levels of any unusual FA has been an elusive goal. We achieved high levels of ω-7 FA accumulation by systematic metabolic engineering of Arabidopsis (Arabidopsis thaliana). A plastidial 16:0-ACP desaturase has been engineered to convert 16:0 to 16:1Δ(9) with specificity >100-fold than that of naturally occurring paralogs, such as that from cat's claw vine (Doxantha unguis-cati). Expressing this engineered enzyme (Com25) in seeds increased ω-7 FA accumulation from <2% to 14%. Reducing competition for 16:0-ACP by down-regulating the ß-ketoacyl-ACP synthase II 16:0 elongase further increased accumulation of ω-7 FA to 56%. The level of 16:0 exiting the plastid without desaturation also increased to 21%. Coexpression of a pair of fungal 16:0 desaturases in the cytosol reduced the 16:0 level to 11% and increased ω-7 FA to as much as 71%, equivalent to levels found in Doxantha seeds.

Revealing the catalytic potential of an acyl-ACP desaturase: tandem selective oxidation of saturated fatty acids.

It is estimated that plants contain thousands of fatty acid structures, many of which arise by the action of membrane-bound desaturases and desaturase-like enzymes. The details of "unusual" e.g., hydroxyl or conjugated, fatty acid formation remain elusive, because these enzymes await structural characterization. However, soluble plant acyl-ACP (acyl carrier protein) desaturases have been studied in far greater detail but typically only catalyze desaturation (dehydrogenation) reactions. We describe a mutant of the castor acyl-ACP desaturase (T117R/G188L/D280K) that converts stearoyl-ACP into the allylic alcohol trans-isomer (E)-10-18:1-9-OH via a cis isomer (Z)-9-18:1 intermediate. The use of regiospecifically deuterated substrates shows that the conversion of (Z)-9-18:1 substrate to (E)-10-18:1-9-OH product proceeds via hydrogen abstraction at C-11 and highly regioselective hydroxylation (>97%) at C-9. (18)O-labeling studies show that the hydroxyl oxygen in the reaction product is exclusively derived from molecular oxygen. The mutant enzyme converts (E)-9-18:1-ACP into two major products, (Z)-10-18:1-9-OH and the conjugated linolenic acid isomer, (E)-9-(Z)-11-18:2. The observed product profiles can be rationalized by differences in substrate binding as dictated by the curvature of substrate channel at the active site. That three amino acid substitutions, remote from the diiron active site, expand the range of reaction outcomes to mimic some of those associated with the membrane-bound desaturase family underscores the latent potential of O(2)-dependent nonheme diiron enzymes to mediate a diversity of functionalization chemistry. In summary, this study contributes detailed mechanistic insights into factors that govern the highly selective production of unusual fatty acids.

Stereochemistry of Delta4 dehydrogenation catalyzed by an ivy (Hedera helix) Delta9 desaturase homolog.

The stereochemistry of palmitoyl-ACP Delta(4) desaturase-mediated dehydrogenation has been examined by tracking the fate of deuterium atoms located on stereospecifically monodeuterated substrates-(4S)- and (4R)-[4-(2)H(1)]-palmitoyl-ACP and (5S)- and (5R)-[5-(2)H(1)]-palmitoyl-ACP. It was found that the introduction of the (Z)-double bond between C-4 and C-5 of a palmitoyl substrate occurs with pro-R enantioselectivity-a result which matches that obtained for a closely related homolog-castor stearoyl-ACP Delta(9) desaturase. These data show that despite the difference in regioselectivity between the two enzymes, the stereochemistry of hydrogen removal is conserved.

The crystal structure of the ivy Delta4-16:0-ACP desaturase reveals structural details of the oxidized active site and potential determinants of regioselectivity.

The multifunctional acyl-acyl carrier protein (ACP) desaturase from Hedera helix (English ivy) catalyzes the Delta(4) desaturation of 16:0-ACP and the Delta(9) desaturation of 18:0-ACP and further desaturates Delta(9)-16:1 or Delta(9)-18:1 to the corresponding Delta(4,9) dienes. The crystal structure of the enzyme has been solved to 1.95 A resolution, and both the iron-iron distance of approximately 3.2A and the presence of a mu-oxo bridge reveal this to be the only reported structure of a desaturase in the oxidized FeIII-FeIII form. Significant differences are seen between the oxidized active site and the reduced active site of the Ricinus communis (castor) desaturase; His(227) coordination to Fe2 is lost, and the side chain of Glu(224), which bridges the two iron ions in the reduced structure, does not interact with either iron. Although carboxylate shifts have been observed on oxidation of other diiron proteins, this is the first example of the residue moving beyond the coordination range of both iron ions. Comparison of the ivy and castor structures reveal surface amino acids close to the annulus of the substrate-binding cavity and others lining the lower portion of the cavity that are potential determinants of their distinct substrate specificities. We propose a hypothesis that differences in side chain packing explains the apparent paradox that several residues lining the lower portion of the cavity in the ivy desaturase are bulkier than their equivalents in the castor enzyme despite the necessity for the ivy enzyme to accommodate three more carbons beyond the diiron site.

The Arabidopsis stearoyl-acyl carrier protein-desaturase family and the contribution of leaf isoforms to oleic acid synthesis.

In plants, changes in the levels of oleic acid (18:1), a major monounsaturated fatty acid (FA), results in the alteration of salicylic acid (SA)- and jasmonic acid (JA)-mediated defense responses. This is evident in the Arabidopsis ssi2/fab2 mutant, which encodes a defective stearoyl-acyl carrier protein-desaturase (S-ACP-DES) and consequently accumulates high levels of stearic acid (18:0) and low levels of 18:1. In addition to SSI2, the Arabidopsis genome encodes six S-ACP-DES-like enzymes, the native expression levels of which are unable to compensate for a loss-of-function mutation in ssi2. The presence of low levels of 18:1 in the fab2 null mutant indicates that one or more S-ACP-DES isozymes contribute to the 18:1 pool. Biochemical assays show that in addition to SSI2, four other isozymes are capable of desaturating 18:0-ACP but with greatly reduced specific activities, which likely explains the inability of these SSI2 isozymes to substitute for a defective ssi2. Lines containing T-DNA insertions in S-ACP-DES1 and S-ACP-DES4 show that they are altered in their lipid profile but contain normal 18:1 levels. However, overexpression of the S-ACP-DES1 isoform in ssi2 plants results in restoration of 18:1 levels and thereby rescues all ssi2-associated phenotypes. Thus, high expression of a low specific activity S-ACP-DES is required to compensate for a mutation in ssi2. Transcript level of S-ACP-DES isoforms is reduced in high 18:1-containing plants. Enzyme activities of the desaturase isoforms in a 5-fold excess of 18:1-ACP show product inhibition of up to 73%. Together these data indicate that 18:1 levels are regulated at both transcriptional and post-translational levels.

A single mutation in the castor Delta9-18:0-desaturase changes reaction partitioning from desaturation to oxidase chemistry.

Sequence analysis of the diiron cluster-containing soluble desaturases suggests they are unrelated to other diiron enzymes; however, structural alignment of the core four-helix bundle of desaturases to other diiron enzymes reveals a conserved iron binding motif with similar spacing in all enzymes of this structural class, implying a common evolutionary ancestry. Detailed structural comparison of the castor desaturase with that of a peroxidase, rubrerythrin, shows remarkable conservation of both identity and geometry of residues surrounding the diiron center, with the exception of residue 199. Position 199 is occupied by a threonine in the castor desaturase, but the equivalent position in rubrerythrin contains a glutamic acid. We previously hypothesized that a carboxylate in this location facilitates oxidase chemistry in rubrerythrin by the close apposition of a residue capable of facilitating proton transfer to the activated oxygen (in a hydrophobic cavity adjacent to the diiron center based on the crystal structure of the oxygen-binding mimic azide). Here we report that desaturase mutant T199D binds substrate but its desaturase activity decreases by approximately 2 x 10(3)-fold. However, it shows a >31-fold increase in peroxide-dependent oxidase activity with respect to WT desaturase, as monitored by single-turnover stopped-flow spectrometry. A 2.65-A crystal structure of T199D reveals active-site geometry remarkably similar to that of rubrerythrin, consistent with its enhanced function as an oxidase enzyme. That a single amino acid substitution can switch reactivity from desaturation to oxidation provides experimental support for the hypothesis that the desaturase evolved from an ancestral oxidase enzyme.

A multifunctional acyl-acyl carrier protein desaturase from Hedera helix L. (English ivy) can synthesize 16- and 18-carbon monoene and diene products.

A desaturase with 83% sequence identity to the coriander delta(4)-16:0-ACP desaturase was isolated from developing seeds of Hedera helix (English ivy). Expression of the ivy desaturase in Arabidopsis resulted in the accumulation of 16:1delta(4) and its expected elongation product 18:1delta(6) (petroselinic acid). Expression in Escherichia coli resulted in the accumulation of soluble, active protein that was purified to apparent homogeneity. In vitro assays confirmed delta(4) desaturation with 16:0-ACP; however, with 18:0-acyl acyl carrier protein (ACP) desaturation occurred at the delta(9) position. The ivy desaturase also converted 16:1delta(9)-ACP and 18:1delta(9)-ACP to the corresponding delta(4,9) dienes. These data suggest at least two distinct substrate binding modes, one placing C4 at the diiron active site and the other placing C9 at the active site. In the latter case, 18:0 would likely bind in an extended conformation as described for the castor desaturase with 9-carbons accommodated in the cavity beyond the dirron site. However, delta(4) desaturation would require the accommodation of 12 carbons for C16 substrates or 14 carbons for C18 substrates. The amino acids lining the substrate binding cavity of ivy and castor desaturases are conserved except for T117R and P179I (castor/ivy). Paradoxically, both substitutions, when introduced into the castor desaturase, favored the binding of shorter acyl chains. Thus, it seems likely that delta(4) desaturation would require a non-extended, perhaps U-shaped, substrate conformation. A cis double bond may facilitate the initiation of such a non-extended conformation in the monounsaturated substrates. The multifunctional properties of the ivy desaturase make it well suited for further dissection of the determinants of regiospecificity.

Evidence linking the Pseudomonas oleovorans alkane omega-hydroxylase, an integral membrane diiron enzyme, and the fatty acid desaturase family.

Pseudomonas oleovorans alkane omega-hydroxylase (AlkB) is an integral membrane diiron enzyme that shares a requirement for iron and oxygen for activity in a manner similar to that of the non-heme integral membrane desaturases, epoxidases, acetylenases, conjugases, ketolases, decarbonylase and methyl oxidases. No overall sequence similarity is detected between AlkB and these desaturase-like enzymes by computer algorithms; however, they do contain a series of histidine residues in a similar relative positioning with respect to hydrophobic regions thought to be transmembrane domains. To test whether these conserved histidine residues are functionally equivalent to those of the desaturase-like enzymes we used scanning alanine mutagenesis to test if they are essential for activity of AlkB. These experiments show that alanine substitution of any of the eight conserved histidines results in complete inactivation, whereas replacement of three non-conserved histidines in close proximity to the conserved residues, results in only partial inactivation. These data provide the first experimental support for the hypotheses: (i) that the histidine motif in AlkB is equivalent to that in the desaturase-like enzymes and (ii) that the conserved histidine residues play a vital role such as coordinating the Fe ions comprising the diiron active site.

Desaturation and hydroxylation. Residues 148 and 324 of Arabidopsis FAD2, in addition to substrate chain length, exert a major influence in partitioning of catalytic specificity.

Exchanging the identity of amino acids at four key locations within the Arabidopsis thaliana oleate desaturase (FAD2) and the Lesquerella fendleri hydroxylase/desaturase (LFAH) was shown to influence partitioning between desaturation and hydroxylation (Broun, P., Shanklin, J., Whittle, E., and Somerville, C. (1998) Science 282, 1315-1317). We report that four analogous substitutions in the FAD2 sequence by their equivalents from the castor oleate hydroxylase result in hydroxy fatty acid accumulation in A. thaliana to the same levels as for the wild-type castor hydroxylase. We also describe the relative contribution of these substitutions, both individually and in combination, by analyzing the products resulting from their expression in A. thaliana and/or Saccharomyces cerevisiae. Yeast expression showed that M324V, a change reachable by a single point mutation, altered the product distribution approximately 49-fold, and that residue 148 is also a predominant determinant of reaction outcome. Comparison of residues at position 148 of FAD2, LFAH, and the Ricinus oleate hydroxylase prompted us to rationally engineer LFAH-N149I, a variant with approximately 1.9-fold increase in hydroxylation specificity compared with that of wild-type LFAH. Control experiments showed that the wild-type Arabidopsis thaliana FAD2 desaturase has inherent, low level, hydroxylation activity. Further, fatty acid desaturases from different kingdoms and with different regiospecificities exhibit similar intrinsic hydroxylase activity, underscoring fundamental mechanistic similarities between desaturation and hydroxylation. For LFAH mutants the hydroxylation:desaturation ratio is 5-9-fold higher for 18-carbon versus 16-carbon substrates, supporting our hypothesis that substrate positioning in the active site plays a key role in the partitioning of catalytic specificity.