Coexpression Analysis Identifies Two OxidoreductasesInvolved in the Biosynthesis of the Monoterpene AcidMoiety of Natural Pyrethrin Insecticides inTanacetum cinerariifolium1[OPEN]

Haiyang Xu,a Gaurav D. Moghe,b,2 Krystle Wiegert-Rininger,c Anthony L. Schilmiller,d Cornelius S. Barry,c

Robert L. Last,b,e and Eran Picherskya,3

aDepartment of Molecular, Cellular, and Developmental Biology, University of Michigan, Ann Arbor,Michigan 48109bDepartment of Biochemistry, Michigan State University, East Lansing, Michigan 48824cDepartment of Horticulture, Michigan State University, East Lansing, Michigan 48824dMass Spectrometry and Metabolomics Core Facility, Michigan State University, East Lansing, Michigan 48824eDepartment of Plant Biology, Michigan State University, East Lansing, Michigan 48824

ORCID IDs: 0000-0002-6954-0000 (H.X.); 0000-0002-8761-064X (K.W.-R.); 0000-0001-7069-7120 (A.L.S); 0000-0003-4685-0273 (C.S.B.);0000-0001-6974-9587 (R.L.L.); 0000-0002-4343-1535 (E.P.).

Flowers of Tanacetum cinerariifolium produce a set of compounds known collectively as pyrethrins, which are commerciallyimportant pesticides that are strongly toxic to flying insects but not to most vertebrates. A pyrethrin molecule is an esterconsisting of either trans-chrysanthemic acid or its modified form, pyrethric acid, and one of three alcohols, jasmolone,pyrethrolone, and cinerolone, that appear to be derived from jasmonic acid. Chrysanthemyl diphosphate synthase (CDS), thefirst enzyme involved in the synthesis of trans-chrysanthemic acid, was characterized previously and its gene isolated. TcCDSproduces free trans-chrysanthemol in addition to trans-chrysanthemyl diphosphate, but the enzymes responsible for the conversionof trans-chrysanthemol to the corresponding aldehyde and then to the acid have not been reported. We used an RNA sequencing-based approach and coexpression correlation analysis to identify several candidate genes encoding putative trans-chrysanthemoland trans-chrysanthemal dehydrogenases. We functionally characterized the proteins encoded by these genes using a combinationof in vitro biochemical assays and heterologous expression in planta to demonstrate that TcADH2 encodes an enzyme that oxidizestrans-chrysanthemol to trans-chrysanthemal, while TcALDH1 encodes an enzyme that oxidizes trans-chrysanthemal into trans-chrysanthemic acid. Transient coexpression of TcADH2 and TcALDH1 together with TcCDS in Nicotiana benthamiana leaves resultsin the production of trans-chrysanthemic acid as well as several other side products. The majority (58%) of trans-chrysanthemic acidwas glycosylated or otherwise modified. Overall, these data identify key steps in the biosynthesis of pyrethrins and demonstratethe feasibility of metabolic engineering to produce components of these defense compounds in a heterologous host.

A small group of plants in the Asteraceae family,including Tanacetum cinerariifolium (formerly Chrysan-themum cinerariifolium), make insecticides knowncollectively as pyrethrins, with the highest levels ofproduction observed in flowers (Casida, 1973; Casida

and Quistad, 1995; Crombie, 1995). Due to their effec-tive toxicity against a wide range of insect species, lowtoxicity to warm-blooded animals, and propensity fordegradation by sunlight and oxidation, pyrethrins havebeen used for pest control since medieval times (Casidaand Quistad, 1995; Katsuda, 1999). The commercialproduction of natural pyrethrins involves drying andgrinding the flowers and dissolving the powder in anorganic solvent, which can be used directly for spray-ing. However, the concentration of pyrethrins in thepowder is low at about 2% (Casida, 1973). Syntheticpyrethrins, called pyrethroids, which are chemicallysimilar but not identical to natural pyrethrins, arecheaper and used at much higher quantities world-wide. However, these synthetic products are generallyless biodegradable and photolabile, persisting in theenvironment longer than natural pyrethrins and, thus,give rise to the emergence of resistance among insects.In addition, some pyrethroids are toxic to mammalsand fish (Katsuda, 2012).

1 This work was supported by the National Science Foundationcollaborative research grants 1565355 to E.P. and 1565232 to R.L.L.

2 Current address: Section of Plant Biology, School of IntegrativePlant Sciences, Cornell University, Ithaca, NY 14853.

3 Address correspondence to lelx@umich.edu.The author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policy de-scribed in the Instructions for Authors (www.plantphysiol.org) is:Eran Pichersky (lelx@umich.edu).

H.X., R.L.L., and E.P. designed the experiments; H.X., G.D.M., A.L.S.,and K.W.-R. conducted the experiments; H.X., C.S.B., R.L.L., and E.P.wrote the article; all authors edited the article.

[OPEN] Articles can be viewed without a subscription.www.plantphysiol.org/cgi/doi/10.1104/pp.17.01330

524 Plant Physiology�, January 2018, Vol. 176, pp. 524–537, www.plantphysiol.org � 2018 American Society of Plant Biologists. All Rights Reserved. www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from

Copyright © 2018 American Society of Plant Biologists. All rights reserved.

Natural pyrethrins comprise a group of six com-pounds. The type I pyrethrins (pyrethrin I, jasmolin I,and cinerin I) are esters of the monoterpenoid trans-chrysanthemic acid with one of the three respective fattyacid-derived alcohols pyrethrolone, jasmolone, and cin-erolone. The type II pyrethrins (pyrethrin II, jasmolin II,and cinerin II) are esters of pyrethric acid with one ofthese respective three alcohols. Pyrethric acid is identicalto trans-chrysanthemic acid, except that its C8 positionhas a methylated carboxyl group (Fig. 1A).The synthesis of both the acid and alcohol moieties

of pyrethrins is not fully understood. With respect tothe acid moiety, Rivera et al. (2001) demonstrated thatT. cinerariifolium flowers contain the enzyme trans-chrysanthemyl diphosphate synthase (CDS; EC 2.5.1.67),which condenses twodimethyl allyl diphosphate (DMAPP)molecules viawhat is knownas an irregularC19-2-3 linkageto form mostly trans-chrysanthemyl diphosphate (CDP;Fig. 1B) as well as a small amount of lavandulyl diphos-phate. The identities of the enzymes responsible for thesubsequent hypothetical steps are not clear. The conversionof CDP to trans-chrysanthemol could be catalyzed by amember of the terpene synthase family similar to thoseterpene synthases catalyzing the conversion of geranyl di-phosphate to the alcohol monoterpenes geraniol and lin-alool in many species (Chen et al., 2011), or CDP could behydrolyzed by phosphatases to give trans-chrysanthemol,similar to the conversion of geranyl diphosphate to geraniolin rose (Rosa spp.)flowers (Magnard et al., 2015). However,no such enzymatic activities were reported thus far in T.cinerariifolium. A recent study by Yang et al. (2014) reportedthat prolonged incubation of Escherichia coli-produced re-combinantCDSproteinwith low concentrations ofDMAPPled to the detection of trans-chrysanthemol in addition toCDP, suggesting thatCDScanproduce trans-chrysanthemolin vivo. Furthermore, the enzymes responsible for the sub-sequent oxidation reactions of trans-chrysanthemol to trans-chrysanthemal and then to trans-chrysanthemic acid (Fig. 1B)had not yet been reported.To facilitate the characterization of pyrethrin bio-

synthesis, a transcriptome assembly of T. cinerariifolium

was generated from RNA sequencing (RNAseq) anal-ysis of leaf and flower tissues harvested at differentstages of development. Candidate genes for trans-chrysanthemic acid biosynthesis were identified basedon coexpression analysis with two previously func-tionally identified genes in pyrethrin biosynthesis,TcCDS and TcGLIP, the latter being the gene encoding aGDSL-family lipase that combines the acid and alcoholmoieties into an ester (Kikuta et al., 2012). This analysisled to the identification of the genes TcADH2 andTcALDH1, encoding two flower-expressed oxidore-ductases that respectively catalyze the two sequentialoxidation steps from trans-chrysanthemol to trans-chrysanthemic acid. The discovery of these enzymesfacilitated the reconstruction of trans-chrysanthemicacid biosynthesis from the precursor DMAPP, bothin vitro and in vivo, the latter by transient expression ofmultiple genes in Nicotiana benthamiana.

RESULTS

Identification of Pyrethrins and Their TerpenoidPrecursors in T. cinerariifolium Flowers

The flower heads of T. cinerariifolium consist of acollection of ray florets on the inside and disc floretson the outside, with both types set on a receptacle(Ramirez et al., 2013). The pyrethrin and terpenoidprecursor contents in T. cinerariifolium leaves andflowers of different developmental stages (Fig. 2A)were determined by analysis of methyl tert-butylether (MTBE)-extracted macerated tissues by gaschromatography-mass spectrometry (GC-MS). Asobserved previously (Kikuta et al., 2012; Ramirezet al., 2012), pyrethrin content in flowers increased asthey matured (Fig. 2B), and leaves contained negli-gible amounts of pyrethrins compared with theamounts observed in flowers. Also as described previ-ously (Casida, 1973), pyrethrin I was the most abundantpyrethrin (Fig. 2C). In addition to pyrethrins, we alsoobserved trans-chrysanthemol, trans-chrysanthemal,

Figure 1. A, Structures of pyrethrins. B,Proposed pathway for the biosynthesis oftrans-chrysanthemic acid.

Plant Physiol. Vol. 176, 2018 525

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

and trans-chrysanthemic acid in floral extracts, andquantitation of trans-chrysanthemic acid indicated thatits concentration also increased as the flower matured(Fig. 2, D and E).

Identification of Candidate Genes Involved in Trans-Chrysanthemic Acid Biosynthesis

To identify the genes encoding the enzymes re-sponsible for the conversion of trans-chrysanthemolto trans-chrysanthemal and trans-chrysanthemal to

trans-chrysanthemic acid, transcriptome assemblieswere constructed from RNAseq libraries constructedfrom eight different T. cinerariifolium tissue samples:leaves, flowers at stage 1, flowers at stage 2, flowers atstage 3, ray florets at stage 4, disk florets at stage 4, rayflorets at stage 5, and disk florets at stage 5 (Fig. 3A).Interrogating our database set (http://sativa.mcdb.lsa.umich.edu/blast/) for oxidoreductases with plant alco-hol dehydrogenase (ADH) and aldehyde dehydrogen-ase (ALDH) sequences (see “Materials and Methods”)identified 12 transcripts encoding putative alcohol de-hydrogenases (named TcADH1 to TcADH12) and three

Figure 2. GC-MS analysis of pyrethrins and terpenoids from T. cinerariifolium leaves and flowers at different stages of devel-opment. A, Flowers of different stages of development and a leaf of T. cinerariifolium. B, Changes in relative concentrations ofpyrethrin I during floral development. Pyrethrin I is the most abundant pyrethrin in the flower, and changes in the concentrationsof other pyrethrins follow the same pattern as those of pyrethrin I. C, GC-MS chromatogram of the total ionmode ofMTBE extractsfrom leaves and flowers harvested at stage 4. In each flower/leaf comparison, samples are shown with the same relative y axisscale, but the 7.2- to 15.2-min section is shown at a smaller scale tomagnify the peaks. Peaks identified as terpenoids and internalstandard (tetradecane) are labeled. D, GC-MS chromatogram (total ion mode) of MTBE extracts from leaves and flowers ofdifferent stages of development, showing the trans-chrysanthemic acid levels in each sample. E, Concentrations of trans-chrysanthemic acid in the leaf and in different stages of flowers. Quantification was achieved by normalization of the peaksin D to the tetradecane internal standard and comparison with a standard curve of authentic trans-chrysanthemic acid (n = 3;means 6 SD). FW, Fresh weight.

526 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

transcripts for putative aldehyde dehydrogenases (namedTcALDH1 to TcALDH3). Four of the putative alcohol de-hydrogenase genes, TcADH1, TcADH2, TcADH4, andTcADH6, and one putative aldehyde dehydrogenase gene,TcALDH1, showed the highest degree of coexpressionwiththe known genes of pyrethrin biosynthesis, TcCDS orTcGLIP (Fig. 3B; Supplemental Table S1). However, themuch lower relative transcript abundance of TcADH4 andTcADH6 compared with that of TcCDS and TcGLIP in

flowers of T. cinerariifolium argued against their involve-ment in trans-chrysanthemic acid biosynthesis.

The expression patterns of TcADH1, TcADH2, andTcALDH1 deduced from the RNAseq read frequencieswere confirmed by quantitative reverse transcription(qRT)-PCR together with the transcript levels of TcCDSand TcGLIP (Fig. 3C). The transcripts of all five geneswere significantly more abundant in floral tissues com-pared with leaves, roots, or stems. Generally, transcripts

Figure 3. Identification of candidate ADH and ALDH genes for trans-chrysanthemic acid biosynthesis. A, Images of T. cine-rariifolium flowers of different stages and of leaves from which RNA samples were obtained for RNAseq analysis. B, Average-linkage hierarchical clustering of relative transcript abundance of putative ADHs and ALDHs with TcCDS and TcGLIP based onthe number of reads of each transcript in each RNAseq library. The tree and heat map were generated by Cluster 3.0 software (see“Materials and Methods”). C, Verification of levels of expression of TcCDS, TcGLIP, TcADH1, TcADH2, and TcALDH1 by qRT-PCR. Transcript levels are expressed relative to that of GAPDH (glyceraldehyde-3-phosphate dehydrogenase) in each sample(n = 4; means6 SD). **, P, 0.01 and *, P, 0.05. The differences between leaf, stem, and root data points and any flower datapoints are all significant at P , 0.001.

Plant Physiol. Vol. 176, 2018 527

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

increased initially from floral stage 1 to floral stage 2 andthen remained at similar levels or decreased somewhat,particularly in ray florets, a pattern consistent with thepreviously reported localization of pyrethrin accumula-tion (Kikuta et al., 2012; Ramirez et al., 2012).

Phylogenetic analysis that included T. cinerariifoliumADHs and other ADHs with assigned functions revealedthat TcADH1 and TcADH2 are most closely related toterpene-modifyingADHs, includingADH1 fromArtemesiaannua, annotated in the National Center for BiotechnologyInformation (NCBI) as artemisinic alcohol dehydrogenase(accession no. AEI16475) and 8-hydroxygeraniol oxidore-ductase fromCatharanthus roseus (Miettinen et al., 2014; Fig.4A). Similarly, TcALDH1 ismost closely related (Fig. 4B) toALDH1 from A. annua, an enzyme that converts artemi-sinic aldehyde and dihydroartemisinic aldehyde to arte-misinic acid and dihydroartemisinic acid, respectively(Teoh et al., 2009). No prediction for organelle targetingwas obtained for the inferred TcADH1, TcADH2, andTcALDH1 protein sequences using the TargetP 1.1 Server(http://www.cbs.dtu.dk/services/TargetP/) and WoLFPSORT (https://wolfpsort.hgc.jp/) programs (Nakai andHorton, 1999; Emanuelsson et al., 2007).

TcADH2 Is a Trans-Chrysanthemol Dehydrogenase

Based on results of the coexpression and phylogeneticanalyses that identified TcADH1 and TcADH2 as candi-dates for the conversion of trans-chrysanthemol to trans-chrysanthemal,we tested their activities in vitro. TcADH1and TcADH2 proteins with a fused N-terminal His6 tagwere expressed in E. coli and purified. To obtain the trans-chrysanthemol substrate, which is not commerciallyavailable, for the enzymatic assays, we used recombinantTcCDS to convert DMAPP to CDP, which was hydro-lyzed subsequently by the addition of commercial alka-line phosphatase to give trans-chrysanthemol (Fig. 5A).As reported previously (Rivera et al., 2001; Yang et al.,2014), the CDS-catalyzed reaction also produces smallamounts of lavandulyl diphosphate, causing the trans-chrysanthemol preparation to contain a small amount oflavandulol (Fig. 5A). Under our assay conditions, we didnot detect trans-chrysanthemol production byCDS for upto 24 h (Fig. 5A).

Recombinant TcADH1 and TcADH2 were initiallytested with a variety of alcohol substrates at 0.3 mM con-centration and 1mMNAD+orNADP+ (Table I).With bothTcADH1 and TcADH2, NAD+ was a more effective co-factor; with NADP+, product yield with each substratewas less than 5% compared with NAD+. TcADH1, how-ever, had no activity with trans-chrysanthemol, insteadshowing its highest level of activity with nerol and re-duced activity with geraniol (48%; Table I), both mono-terpene compounds. TcADH2 had the highest activitywith trans-chrysanthemol, but it also had some activity(70% or less compared with its activity with trans-chrysanthemol) with several other monoterpene alco-hols (Table I). GC-MS analysis of the reaction productsshowed that TcADH2 converted trans-chrysanthemol to

trans-chrysanthemal but did not oxidize lavandulol (Fig.5B). Kinetic analysis revealed that TcADH2 had a Kmvalue of 236 6 5.8 mM and a turnover rate of 0.75 60.0032 s21 for trans-chrysanthemol, while theKm value forNAD+ was 192.6 6 8.7 mM.

TcALDH1 Converts Trans-Chrysanthemal to Trans-Chrysanthemic Acid

Based on results of the coexpression and phylogeneticanalyses that identified TcALDH1 as a candidate for theconversion of trans-chrysanthemal to trans-chrysanthemicacid,we tested its activity in vitrowith trans-chrysanthemaland several other substrates. Trans-chrysanthemal wasproduced in a coupled enzymatic method employingTcCDS, alkaline phosphatase, and TcADH2 (Fig. 5A; see“Materials andMethods”).N-terminally taggedTcALDH1purified from E. coli showed highest activity with trans-chrysanthemal substrate but also displayed substantialactivity with several additional aliphatic and aromaticsubstrates (Table I). The enzymehadaKmvalue of 4.66 1.8mM for trans-chrysanthemal when NAD+ was providedand a Km value of 4.46 2.2 mM for trans-chrysanthemal inthe presence of NADP+ (Table II). BothNAD+ andNADP+

served as cofactors for the enzyme, with Km of 20.4 mM forNAD+ and 68.6 mM for NADP+ (Table II). Incubation oftrans-chrysanthemol with both TcADH2 and TcALDH1led to the production of trans-chrysanthemic acid (Fig. 5C).

Transient Coexpression of TcADH2 and TcALDH1 withCDS in N. benthamiana results in Trans-ChrysanthemicAcid Production

To test the activities of TcADH2 and TcALDH1 inplanta, N. benthamiana leaves were infiltrated with Agro-bacterium tumefaciens strains harboring plasmids contain-ing TcCDS, TcADH2, and TcALDH1. For controls, weinfiltrated N. benthamiana leaves with TcCDS alone,TcCDS and TcADH2, or EGFP (Enhanced Green FluorescentProtein; see “Materials and Methods”) as a control. Inthese experiments, the complete open reading frame ofTcCDS was used, including the transit peptide that wasshown to direct the protein to the plastids (Yang et al.,2014). Transformed leaves were harvested 10 d after in-filtration, and the products were extracted and analyzedby GC-MS. As expected, transient expression of TcCDSalone resulted in the production of trans-chrysanthemolat 18.3 nmol g21 fresh weight and trace amounts of trans-chrysanthemic acid (Fig. 6). Notably, coexpression ofTcCDSwith TcADH2 inN. benthamiana leaves resulted inthe production of trans-chrysanthemic acid at 328 nmolg21 fresh weight (Fig. 6), a 48-fold increase over the ex-pression of TcCDS alone. Finally, when all three genes(TcCDS, TcADH2, and TcALDH1) were coexpressedin N. benthamiana leaves, the concentration of trans-chrysanthemic acid was 818.4 nmol g21 fresh weight, a122-fold increase over the expression of TcCDS alone.

In addition to trans-chrysanthemic acid and itsprecursors, additional volatile derivatives of these

528 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

Figure 4. Phylogenetic analysis of candidate T. cinerariifolium dehydrogenases for trans-chrysanthemic acid biosynthesis basedon protein sequences. A, Phylogenetic tree for TcADH1 and TcADH2. B, Phylogenetic tree for TcALDH1. The protein sequencesfrom other species are of functionally characterized enzymes whose sequences were identified by BLAST search to be mostclosely related to the T. cinerariifolium sequences. Phylogenetic analyses were conducted in MEGA7 (Kumar et al., 2016) withthe following parameters: multiple sequence alignment with ClustalW, phylogenetic construction with the maximum likelihoodmethod, and bootstrap tests of 1,000 replicates.

Plant Physiol. Vol. 176, 2018 529

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

metabolites were identified. A peak with a mass frag-mentation spectrum similar to trans-chrysanthemol(Supplemental Fig. S1C) was found in N. benthamianaleaves expressing TcCDS; it eluted with a retention time

of 30.16 min, compared with a retention time of 10.08 minfor trans-chrysanthemol, and was labeled Unknown1 in Figure 6. Despite the lack of an authentic standardfor this compound, which precluded the determinationof its actual concentration, we could measure changesin its relative concentrations. When TcCDS-expressingN. benthamiana leaf extract was treated with glycosidase,Unknown 1 was no longer observed (Supplemental Fig.S2A). When the same extract was treated with NaOH,there were no changes in Unknown 1 and trans-chrysanthemol (Supplemental Fig. S2B).

In leaves expressing both TcCDS and TcADH2, Un-known 1 concentration was reduced by 30-fold. Inaddition, a new compound judged to be related to trans-chrysanthemic acid by its mass fragmentation spectrumand designatedUnknown 2was detectedwith a retentiontime of 15.49 min (Fig. 6; Supplemental Fig. S1D). Inleaves expressing TcCDS, TcADH2, and TcALDH1, thelevels of Unknown 2 increased 7-fold over those observedin leaves expressing TcCDS and TcADH2 withoutTcALDH1. In addition, leaves expressing all three geneshad a new trans-chrysanthemic acid-related compound,designatedUnknown 3,with a retention time of 31.98min(Fig. 6; Supplemental Fig. S1E). Unknown 2 and Un-known 3 were eliminated when the extract was treatedwith either glycosidase or NaOH (Supplemental Fig. S3,A and B), indicating that they are likely to be esters oftrans-chrysanthemic acid. It is notable that the concen-tration of trans-chrysanthemal was always below detec-tion levels in allN. benthamiana infiltrated leaves, whetherexpressing TcCDS by itself, TcCDS and TcADH2, orTcCDS, TcADH2, and TcALDH1.

To search for nonvolatile trans-chrysanthemol andtrans-chrysanthemic acid conjugates, aliquots of theleaf MTBE extracts from these experiments were dried,dissolved in 70:30 acetonitrile:water (v/v) solution, andanalyzed by liquid chromatography (LC)-quadrupoletime of flight-mass spectrometry (MS). The samplefrom leaves expressing TcCDS alone contained a peakwith mass-to-charge ratio (m/z) 803.3742 (Fig. 7B). Themass spectrum-mass spectrum (MS/MS) of this me-tabolite (Fig. 7E) was identical to that of a compoundproduced in tobacco (Nicotiana tabacum) expressingTcCDS (Yang et al., 2014) that these workers putativelyidentified as trans-chrysanthemyl malonylglucoside. Insamples expressing TcCDS + TcADH2, a peak with m/z831.3328 was detected with an MS/MS that closelymatched the spectrum of another monoterpene acidglucoside, geranyl-6-O-malonyl-b-D-glucopyranoside(Yang et al., 2011). Based on this similarity, the m/z831.3328 compound is putatively identified as the di-mer ion (2M-H) of a trans-chrysanthemic acid malo-nylglucoside conjugate (Fig. 7, C and F). When TcCDS,TcADH2, and TcALDH1 were coexpressed in N. ben-thamiana, the level of a product with identical elutiontime and accurate mass increased ;150-fold based onpeak area compared with TcCDS and TcADH2 ex-pression without TcALDH1 (Fig. 7, C and D). To de-termine the portion of trans-chrysanthemic acid presentas malonylated glucoside in plants simultaneously

Figure 5. Gas chromatography analyses of products obtained in in vitrobiochemical assays of TcADH2 and TcALDH1. For all assays analyzed,reaction products were extracted with 100 mL of MTBE and run on anRxi-5Sil column. Tetradecane was used as an internal standard. A, Syn-thesis of trans-chrysanthemol substrate. Top trace, 1 mM of a commerciallyavailable standard of a trans- and cis-chrysanthemol mixture; middletrace, reaction products obtained by incubating 30 mg of recombinantTcCDS with 2.5 mM DMAPP in a 50-mL reaction for 24 h at 30°C;bottom trace, reaction products obtained by incubating the products ofthe TcCDS-catalyzed condensation of DMAPP with 5 units of alkalinephosphatase (ALP) for 1 h at 37°C. B, In vitro production of trans-chrysanthemol. Reaction products obtained by incubating 0.64 mM

trans-chrysanthemol and 1.5 mM NAD+ with 5 mL of eluted protein fromempty vector (top trace) or 1.25mg of purified TcADH2 (bottom trace) in a60-mL reaction volume for 5 min. C, Production of trans-chrysanthemicacid from trans-chrysanthemol in a coupled assay containing 0.64 mM

trans-chrysanthemol and 1.5mMNAD+with 1.25mg of purified TcADH2and 6 mg of purified TcALDH1 in a 60-mL reaction volume for 5, 10, 15,25, and 45 min. A control reaction was performed using 5 mL of elutedprotein from the empty vector. The bottom trace shows 0.3 mM of acommercial trans-chrysanthemic acid.

530 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

expressing the three enzymes, we used GC-MS to com-pare the amount of free trans-chrysanthemic acid in ex-tract hydrolyzedwith 0.4 N NaOH at 80°C (which left nodetectable trans-chrysanthemic acid malonylglucosideas determined by LC-quadrupole time of flight-MS;Supplemental Fig. S4) with that in nonhydrolyzed ex-tract and found that 58% of the total trans-chrysanthemicacid in this extract was esterified, including with malo-nylated glucoside (Fig. 7G). Since the concentration offree trans-chrysanthemic acid in these leaves was de-termined to be 818.4 nmol g21 fresh weight (Fig. 6C), thetotal amount of trans-chrysanthemic acid produced inthese leaves can be calculated to be 1,946.6 nmol g21

fresh weight.

DISCUSSION

Coexpression Analysis and Biochemical Assays IndicateThat TcADH2 and TcALDH1 Catalyze Reactions in theTrans-Chrysanthemic Acid Biosynthetic Pathway

T. cinerariifolium is an important commercial sourceof the biodegradable natural pyrethrin insecticides,which are very efficient against flying insects and safefor humans and other vertebrates. It has been shown

that pyrethrin biosynthesis begins at early stages inthe developing achenes (dry fruits) in the inflorescenceand reaches peak accumulation in the mature achene(Ramirez et al., 2012). Our observations on the patternof accumulation of pyrethrins in the inflorescence (Fig.2B) were consistent with these previous observations.Based on this information, we proceeded to performtranscriptomic profiling on RNA samples collectedfrom five different stages of developing inflorescencesas well as from leaf material. The last two stages offloral development, stages 4 and 5, afforded enoughmaterial to do separate analyses on ray florets (flowerson the outside perimeter, which have large petals) anddisk florets (flowers on the inside, with small petals).

The transcriptomic profiling data (assembled tran-scripts and the number of reads from each transcript)were used to perform coexpression cluster analysis(Eisen et al., 1998) to identify candidate ADHandALDHgenes whose expression patterns most resembled that ofTcCDS, the gene encoding the key enzyme in the path-way for trans-chrysanthemic acid (Rivera et al., 2001;Yang et al., 2014), as well as that of TcGLIP, encoding theenzyme that forms the final pyrethrin product (Kikutaet al., 2012). Both genes were shown to have peak tran-script levels at the earliest developmental stage (stage 2)

Table II. Kinetic properties of recombinant TcADH2 and TcALDH1

Data are presented as means 6 SD from triplicate independent assays. All assays were performed by GC-MS.

Enzyme Substrate Km Kcat Kcat/Km

mM s21M21 s21

TcADH2 Trans-chrysanthemola 236.0 6 5.8 0.75 6 0.0032 3,186.2NAD+b 192.6 6 8.7 0.64 6 0.0034 3,345.5

TcALDH1 Trans-chrysanthemalc 4.4 6 2.2 0.11 6 0.0049 25,122.4NADP+d 20.4 6 7.1 0.090 6 0.0013 4,391.8Trans-chrysanthemale 4. 6 6 1.8 0.096 6 0.0032 20,992.3NAD+f 68.6 6 17.1 0.086 6 0.0012 1,256.07

aKinetic parameters were determined with 1 mM NAD+. bKinetic parameters were determined with 0.64 mM trans-chrysanthemol. cKineticparameters were determined with 1 mM NADP+. dKinetic parameters were determined with 0.04 mM trans-chrysanthemal. eKinetic param-eters were determined with 1 mM NAD+. fKinetic parameters were determined with 0.04 mM trans-chrysanthemal.

Table I. Relative activities of recombinant TcADH1, TcADH2, and TcALDH1 with selected substrates

The activities of TcADH1 and TcADH2 were measured with 0.3 mM alcohols and 1 mM NAD+. The activities of TcALDH1 were measured with0.04 mM aldehydes and 1 mM NADP+. Data are expressed as relative mean percentages from triplicate independent assays.

Alcohols TcADH1 TcADH2 Aldehydes TcALDH1

Cis-3-hexenol N.D.a N.D. Trans-2-heptenal 47Perillyl alcohol N.D. 35 Octanal 90Nerol 100b 42 Dodecanal 47Geraniol 48 38 Citral 20(S)-b-Citronellol 8 22 Perillaldehyde 40Trans,trans-farnesol N.D. N.D. Farnesal 54Trans-chrysanthemol N.D. 100c Trans-chrysanthemal 100d

8-Hydroxygeraniol N.D. 70 (S)-(2)-Citronellal 84Benzyl alcohol N.D. N.D. Benzaldehyde 90Cinnamyl alcohol N.D. N.D. Trans-cinnamaldehyde 8

aN.D., Not detectable (less than 5% of highest activity). b100% relative activity, corresponds to 0.44 mmol min21 mg21 citral. c100%relative activity, corresponds to 0.5 mmol min21 mg21 trans-chrysanthemal. d100% relative activity, corresponds to 0.11 mmol min21 mg21 trans-chrysanthemic acid.

Plant Physiol. Vol. 176, 2018 531

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

that could be examined (Ramirez et al., 2012). This anal-ysis identified TcADH1, TcADH2, and TcALDH1 as thestrongest candidate enzymes for the synthesis of the ter-pene moiety of pyrethrins. The results of the qRT-PCRanalysis of the levels of transcripts of these three genes,as well as those of TcCDS and TcGLIP (Fig. 3C), weregenerally consistent with the read frequencies obtained inthe transcriptomic profiling experiments (SupplementalTable S1). However, the transcript levels wemeasured forall five genes do not show a steep decline in later stages offloral development, aswas reported previously forTcCDsand TcGLIP (Ramirez et al., 2012), although the discrep-ancy may be due to differences in the delineation of de-velopmental stages and consequent differences in theactual ages of the materials examined.

Based on the identification of TcADH1, TcADH2, andTcALDH1as themost likely candidates to be involved in theconversion of trans-chrysanthemol to trans-chrysanthemicacid, we proceeded to produce recombinant proteins fromthese three genes and test the catalytic activities of theseproteins in in vitro assays. TcADH1 exhibited no activitywith trans-chrysanthemol, but TcADH2 was able tocatalyze the conversion of trans-chrysanthemol totrans-chrysanthemal, with a Km value of 236 mM for trans-chrysanthemol (Table II). Similar in vitro assays demon-strated that TcALDH1 catalyzes the conversion oftrans-chrysanthemal to trans-chrysanthemic acid,with aKmvalue of 4.4 mM for trans-chrysanthemal. Transient coex-pression of TcADH2 and TcALDH1 together with TcCDS inN. benthamiana leaves resulted in appreciable amounts of

Figure 6. Production of trans-chrysanthemol, trans-chrysanthemic acid, and related compounds in N. benthamiana leavestransiently expressing TcCDS, TcADH2, and TcALDH1 proteins. A, GC-MS chromatograms of MTBE extracts ofN. benthamianaleaves expressing EGFP (control), TcCDS, TcCDS and TcADH2, and TcCDS with TcADH2 and TcALDH1. For terpenes, m/z =123 was monitored, and for the internal control tetradecane, m/z = 198 was monitored. Peaks related to trans-chrysanthemol,trans-chrysanthemic acid, and internal standard are labeled. B and C, Concentrations of free trans-chrysanthemol (B) and freetrans-chrysanthemic acid (C) inN. benthamiana leaves expressing the indicated constructs were determined by comparison withan authentic standard. FW, Fresh weight. D to F, Relative levels of Unknown 1 (D), Unknown 2 (E), and Unknown 3 (F) in N.benthamiana leaves expressing the indicated constructs. For each compound, the plant material expressing a specific constructthat showed the highest levels (average of three biological replicates) was set at 100%. The data in B to F represent means 6 SD

from triplicate biological replicates. N.D., Not detected.

532 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

trans-chrysanthemic acid produced, 1,946.6 nmol g21 freshweight, further demonstrating the ability of TcADH2 andTcALDH1 to catalyze the sequential oxidation of trans-chrysanthemol to trans-chrysanthemic acid.

Chrysanthemic Acid Can Be Made Efficiently by in PlantaHeterologous Expression of Only TcCDS, TcADH2,and TcALDH1

Yang et al. (2014) reported that transgenic tobacco plantsexpressing only TcCDS produce trans-chrysanthemol andtrans-chrysanthemyl malonylglucoside. Our analysis ofN.benthamiana leaves transiently expressing TcCDS identifiedthe production of these two metabolites as well as a com-pound by GC-MS that appears to be a modification oftrans-chrysanthemol (Unknown 1). When extracts ofN. benthamiana leaves transiently expressing TcCDS wastreated with glycosidase, both trans-chrysanthemyl malo-nylglucoside and Unknown 1 disappeared, and theamount of trans-chrysanthemol increased by 35-fold, in-dicating that most of the trans-chrysanthemol in theseleaves is in a glycone form. These results indicate (as alsonoted by Yang et al. [2014]) that trans-chrysanthemol isproduced directly by TcCDS and/or that phosphatasespresent in the host plant can hydrolyze the two phos-phate groups from trans-chrysanthemyl diphosphate

to give trans-chrysanthemol. But in addition to trans-chrysanthemol and its derivatives, we also detected traceamounts of trans-chrysanthemic acid in theN. benthamianaleaves transiently expressing TcCDS (Fig. 6), showing thatendogenous enzymes found in plant hosts are capable offurther modifying the heterologously produced trans-chrysanthemol by additional oxidation reactions to givethe corresponding acid. This observation is similar to theobserved conversion of the monoterpene alcohol geraniolto geranial and geranic acid in transgenic tomato (Solanumlycopersicum) fruits (Davidovich-Rikanati et al., 2007).

The coexpression of TcCDS with TcADH2, and particu-larly the coexpression of these two genes with TcALDH1,greatly enhanced the production of trans-chrysanthemicacid in N. benthamiana leaves (Fig. 6). While trans-chrysanthemyl malonylglucoside and Unknown 1 (amodified trans-chrysanthemol) were no longer detected inplants expressing all three genes, the malonylglucosideester of trans-chrysanthemic acid and two other esters ofthis acid, Unknown 2 and Unknown 3, were present at acombined concentration exceeding the levels of free trans-chrysanthemic acid by a factor of 1.5. These results indicatethat, in addition to dehydrogenases that can act on trans-chrysanthemic acid precursors, N. benthamiana leaves alsocontain enzymes that canuse trans-chrysanthemic acid as asubstrate and modify it further. The observation that en-dogenous enzymes in a plant cell engineered to make

Figure 7. LC-MS analysis of N. benthamiana leaves simultaneously expressing the three enzymes TcCDS, TcADH2, andTcALDH1. A to D, Extracted ion chromatograms of m/z 803.37 and 831.33 are shown for EGFP single expression control (A),TcCDS (B), TcCDS + TcADH2 (C), and TcCDS + TcADH2 + TcALDH1 (D). Chromatograms are all scaled the same, as indicatedby the ion current (1.01e7) in the top right corner of each chromatogram. The sample for Dwas diluted 10-fold comparedwith theother three samples due to the high concentration ofm/z 831.33 in the sample. E and F, MS/MS for ionsm/z 803.37 (E) andm/z831.33 (F) along with the proposed compound structures based on exact mass, fragmentation pattern, and similarity to previouslypublished data. G, Relative levels of trans-chrysanthemic acid in N. benthamiana leaves expressing all TcCDS, TcADH2, andTcALDH1 with or without sodium hydroxide treatment.

Plant Physiol. Vol. 176, 2018 533

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

compounds not previously present in the cell can reactwith such new compounds has been made before(Lewinsohn and Gijzen, 2009). For a successful geneticengineering attempt to reconstruct the pyrethrin biosyn-thetic pathway in a heterologous system, it will be nec-essary to counteract such nonproductive side reactions.

Biosynthesis of Pyrethric Acid

In three of the six pyrethrins that T. cinerariifolium syn-thesizes (pyrethrin II, cenerin II, and jasmolin II), pyrethricacid rather than trans-chrysanthemic acid constitutesthe acid moiety. Pyrethric acid is identical to trans-chrysanthemic acid except that the C8 position (some-times referred to as C10) has a methylated carboxyl groupand, therefore, is likely derived from trans-chrysanthemicacid by a series of enzymatic reactions involving first thehydroxylation of C8, then two successive oxidations of C8to give an aldehyde and then a carboxyl group, and finallya carboxylmethylation. The hydroxylation of C8 is likely tobe catalyzed by an enzyme of the cytochrome P450 oxi-doreductase family. The next two oxidation reactions,however, are equivalent to the reactions catalyzed byTcADH2 and TcALDH1, respectively. Our coexpressionanalysis identified only TcADH1, TcADH2, and TcALDH1as likely candidate genes for involvement in the syn-thesis of the terpene moiety of pyrethrins. TcADH4and TcADH6 were also identified as potential candi-dates, but the low levels of their transcripts in pyrethrin-producing tissue made them less likely to be involved intrans-chrysanthemic acid biosynthesis. Therefore, it ispossible that the TcADH4 or TcADH6 proteins are in-volved in trans-chrysanthemic acid biosynthesis. There-fore, it is possible that the TcADH6 protein is involved inthe biosynthesis of pyrethric acid by catalyzing the con-version of the C8 alcohol to the C8 aldehyde, as pyrethrinscontaining the pyrethric acid moiety are much less abun-dant that those containing trans-chrysanthemic acid. It isalso possible that TcADH1, which had no activity withtrans-chrysanthemol, catalyzes this reaction, and perhapseven TcADH2 possesses this activity. It is notable thatall three of these ADHs are evolutionarily closely relatedto 8-hydroxygeraniol oxidoreductase from Catharanthusroseus, an enzyme that catalyzes the same C8 alcoholoxidation reaction on anothermonoterpenoid (Miettinenet al., 2014). Finally, it is possible that TcALDH1 catalyzesthe conversion of the C8 aldehyde to give a carboxylgroup. The lack of suitable substrates currently preventsthe testing of these hypotheses, but the availability oftransgenic plants producing some of the precursors in thepathway to pyrethric acid may ameliorate this problem.

MATERIALS AND METHODS

Plant Materials and Chemicals

Various tissues, including flowers, leaves, stems, and roots, were collectedfrom Tanacetum cinerariifolium grown in a greenhouse with a 16/8-h day/nightphotoperiod. Harvested tissues were flash frozen in liquid nitrogen and stored

at 280°C until use. All commercial chemicals were purchased from Sigma-Aldrich with the exception of 8-hydroxygeraniol, which was obtained fromSanta Cruz Biotechnology, and trans-chrysanthemol and trans-chrysanthemal.

Trans-chrysanthemolwasproducedviaanenzymaticmethodbyTcCDSplusalkaline phosphatase from DMAPP, and its concentration was calculatedaccording on GC-MS based on the standard curve of a commercial mixture oftrans- and cis-chrysanthemol. Trans-chrysanthemal was produced by TcADH2and NAD+ from trans-chrysanthemol, extracted by MTBE from the reactionvolume, dried, and dissolved in water.

GC-MS Analysis of Pyrethrins and Terpenoids from T.cinerariifolium Leaves and Flowers of Different Stagesof Development

Leaf andflower tissues of different stages of developmentwere cut into smallpieces, 100 mg of which was transferred into a tube containing 200 mL of MTBEwith 0.01 ng mL21 tetradecane as an internal standard. The tube was vortexedfor 3 min at maximum speed, incubated at room temperature for 2 h, and theMTBE phase was then collected. Sequential extractions with MTBE on testsamples indicated that, in the initial extraction, more than 94% of the chrys-anthemic acid partitioned into the MTBE phase. The MTBE extracts wereanalyzed by GC-MS. The measurement of trans-chrysanthemic acid wasperformed based on the corresponding standard curve.

RNAseq Analysis

RNAseq analysiswas done essentially according toMoghe et al. (2017). TotalRNA was extracted from leaf and flower parts at different stages of develop-ment using the Total RNA Isolation Kit from Omega. The quantity and qualityof extracted RNA for sequencing were determined with Qubit (Thermo FisherScientific) and Bioanalyzer (Agilent Technologies). Libraries for all eight sam-ples (flowers at stage 1, flowers at stage 2, flowers at stage 3, ray florets at stage4, disk florets at stage 4, ray florets at stage 5, disk florets at stage 5, and leaf)were made using the KAPA stranded RNAseq library preparation kit and se-quenced at theMichigan State University Genomics Core on the Illumina HiSeq2500 with HiSeq SBS reagents in 2- 3 150-bp format. Initial reads have beendeposited in the NCBI (bioproject accession no. PRJNA399494).

To assemble the reads, paired endRNAseq readswere trimmed byquality scoresusingTrimmomaticversion 0.32 (Bolger et al., 2014)with the followingparameters(PE -threads 4 -phred33 ILLUMINACLIP:all_adapters_combined.fasta:2:30:10LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 HEADCROP:10MINLEN:50). Prior to generating a full-scale RNAseq assembly, we tested theimpact of changing the k-mer value for de novo assembly using a set of normalizedreads obtained using the insilico_read_normalization.pl function provided in Trinityversion 2.2.0 (Haas et al., 2013) with the following options (–seqType fq–JM 100G–max_cov 50–pairs_together–SS_lib_type RF–CPU 10–PARALLEL_STATS–KMER_SIZE 25–max_pct_stdev 200). The assemblies were analyzed forcompleteness by screening theMetazoan and Eukaryote comparison databases ofBUSCO version 1.22 (Simão et al., 2015) using default parameters. This analysisrevealed k = 31 to be the most optimal value of k for assembling complete tran-scripts. This result also was supported by analysis of the length distribution oftranscripts using the abyss_fac function inAbySS 1.9.0 (Simpson et al., 2009), withk = 31 producing the longest transcripts (N50 = 1,810 bp). Thus, we performed anassembly of all RNAseq reads with Trinity version 2.2.0 using the followingparameters (–seqType fq–KMER_SIZE 31–max_memory 120G–SS_lib_typeRF–CPU 16–min_kmer_cov 2). This assembly was used to identify ADH-likeand ALDH-like transcripts using BLAST.

The expression levels of all transcripts were estimated using thealign_and_estimate_abundance.pl function of the Trinity software using the followingparameters (–transcripts Trinity.fasta–prep_reference–left PYR*_R1_*.fastq–rightPYR*_R2_*.fastq–est_method RSEM–aln_method bowtie–SS_lib_typeRF–thread_count 10–max_ins_size 1000–trinity_mode–seqType fq), whichestimated transcript abundance using the RSEM function (Li and Dewey,2011). These expression values were used to identify transcripts correlatedwith CDSases and GLIP transcripts across all tissues.

Identifying ADH and ALDH Sequences in the T.cinerariifolium Transcriptome

To find candidate ADH and ALDH genes in our T. cinerariifolium tran-scriptome database, we queried it on the publicly accessible site (http://sativa.

534 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

mcdb.lsa.umich.edu/blast/) with various plant ADH and ALDH sequences(e.g. cinnamyl alcohol dehydrogenase from Populous trichocarpa, accession no.ACC63874; geraniol dehydrogenase from Ocimum basilicum, Q2KNL6; benz-aldehyde dehydrogenase from Antirrhinum majus, ACM89738; aldehydedehydrogenase1 from Artemisia annua, ACR61719.1) using the TBLASTNfunction. We also screened the entire annotated database for the designationsalcohol dehydrogenase and aldehyde dehydrogenase and added transcriptsidentified in this way to our list of candidates.

Coexpression Analysis

The number of reads from each transcript in the various RNAseq analysesof RNA samples from different stages of development and parts of the plantwere used to perform cluster analysis using the freely available Cluster 3.0software program (http://bonsai.hgc.jp/;mdehoon/software/cluster/software.htm#ctv). This program is based on an algorithm developed byEisen et al. (1998), and the analysis results in a tree and heat map that matcheseach gene with another gene whose expression pattern matches best with thatof the first one.

qRT-PCR Analysis of TcCDS, TcGLIP, TcADH1, TcADH2,and TcALDH1 Transcript Levels

For real-time RT-PCR analysis of transcripts in different tissues, RNA wasisolated using the Total RNA Isolation Kit from Omega with a DNA digestionstep. cDNAswerepreparedusing theHighCapacity cDNAReverseTranscriptionKit from Thermo Fisher Scientific following the manufacturer’s instructions. Primerdesign and real-time PCR were performed following the manufacturer’s instruc-tions. Assays were performed using four independent biological replicates. Therelative amounts of transcripts for different genes were normalized to GAPDH tran-script levels using LinRegPCR software (http://www.hartfaalcentrum.nl/index.php?main=files&fileName=LinRegPCR.zip&description=LinRegPCR:%20qPCR%20data%20analysis&sub=LinRegPCR). The CT mean values for GAPDH areshown in Supplemental Figure S5. The statistical assay (unpaired Student’s t test,two-tailed option) was performed via GraphPad Prism software (https://www.graphpad.com/scientific-software/prism/).

Generation, Expression, and Purification of RecombinantTcCDS, TcADH1, TcADH2, and TcALDH1

The open reading frames ofTcADH1,TcADH2, andTcALDH1were obtainedby RT-PCR from prepared cDNA of flowers at stage 1 of T. cinerariifolium. Thefull-length cDNAs were introduced into the expression vector pET28a+ orpHIS8, in each case generating a fusion gene that encoded a tag of His6 residuesat the N terminus for expression in Escherichia coli. To obtain soluble proteins forexpression in E. coli, a truncated open reading frame of TcCDS, missing the first50 codons, was obtained by RT-PCR from prepared cDNA of flowers at stage1 of T. cinerariifolium. The truncated TcCDS open reading frame was insertedinto the pEXP5-CT/TOPO vector following the manufacturer’s instructions,generating a fusion gene that encodes a tag ofHis6 residues at the C terminus forexpression in E. coli. The purification of recombinant proteins was performed asdescribed previously (Xu et al., 2013). All constructs were transformed intoE. coli BL21(+) cells. Transformed E. coli cells were grown in Luria-Bertanimedium containing appropriate antibiotics until optical density of the cultureat 600 nm reached 0.6, and then recombinant gene expression was inducedwiththe addition of isopropylthio-b-galactoside to a final concentration of 0.15 mM

and cells were grown overnight at 16°C. The resulting His-tagged fusion pro-teins were purified using Ni-NTA affinity columns.

Enzymatic Assays of Recombinant TcCDS

The enzymatic assay for TcCDS broadly followed the protocol described byRivera et al. (2001) as follows. The reaction was initiated by adding 30 mg ofaffinity-purified His-tagged enzyme in a final volume of 50 mL of assay buffer(pH 7.5) containing 50 mM Tris-HCl, 2 mM DTT, 5 mM MgCl2, and 2.5 mM

DMAPP. The assaywas incubated at 30°C for 24 h. To analyze the production ofCDP in this assay by GC-MS, hydrolysis of CDP to trans-chrysanthemol by5 units of Roche rAPid alkaline phosphatase (Sigma-Aldrich) was preformedfollowing the manufacturer’s instructions at 37°C for 2 h. Reaction productswere extracted with 100 mL of MTBE, and the MTBE extract was injected andanalyzed by GC-MS.

Enzymatic Assays of Recombinant TcADH1 and TcADH2

Dehydrogenase activity assaysweredone essentially according to Iijimaet al.(2006). To assay the substrate specificity of TcADH1 and TcADH2, reactionswere initiated by adding 1.25 mg of affinity-purified His-tagged enzyme in afinal volume of 50 mL of assay buffer (pH 8) containing 50 mM Tris-HCl, 2 mM

DTT, 1 mM NAD+, and 0.3 mM selected alcohols. The assays were incubated at30°C for 10 min, after which reaction products were extracted with 100 mL ofMTBE. The MTBE extract was analyzed by GC-MS for product and remainingsubstrate.

To determine the kinetic parameters of TcADH2, a similar protocol wasfollowed. The Km value for NAD+ was determined by using 0.64 mM trans-chrysanthemol, whereas the Km value for trans-chrysanthemol was deter-mined with 1 mM NAD+. Km and kcat values were calculated from initial ratedata by using the hyperbolic regression analysis method in Hyper32 software(version 1.0.0; http://hyper32.software.informer.com/).

Enzymatic Assays of Recombinant TcALDH1

To test the substrate specificity of TcALDH1, reactions were initiated byadding 2.5 mg of affinity-purified His-tagged enzyme in a final volume of 50 mLof assay buffer (pH 8.5) containing 50 mM Tris-HCl, 2 mM DTT, 1 mM NADP+,and 40 mM selected aldehydes. The assays were incubated at 30°C for 10 min,after which reaction products were extracted with 100 mL of MTBE. The MTBEextract was analyzed by GC-MS. The reaction rate was calculated according tothe decrease of substrate based on the corresponding standard curve.

To determine the kinetic parameters of TcALDH1, a similar protocol wasfollowed. The Km value for NAD+ and NADP+ was determined by using 40 mM

trans-chrysanthemal, whereas the Km value for trans-chrysanthemal was deter-minedwith 1mMNAD+ orNADP+. The reaction rate was calculated according tothe production of trans-chrysanthemic acid based on the corresponding standardcurve. Km and kcat values were calculated as described above.

Plasmid Construction for Transient Expression in N.benthamiana Leaves

The complete open reading frame of EGFP was amplified from pSAT6-EYFP-N1 vector and used as a control gene in this study. The EGFP, TcCDS,TcADH2, and TcALDH1 genes were spliced into pEAQ-HT binary vectorbetween AgeI and XhoI restriction sites using the NEBuilder HiFi DNA AssemblyCloningKit (https://www.neb.com/products/e5520-nebuilder-hifi-dna-assembly-cloning-kit; NEB) according to the manufacturer’s instructions to express themunder the control of the 35S promoter.

Transient Expression in Leaves of N. benthamiana

Agrobacterium tumefaciens strain GV3101 infiltration (agroinfiltration) wasperformed as described previously (Sainsbury et al., 2009). Briefly, A. tumefa-ciens strain GV3101 was grown at 28°C at 200 rpm for 24 h in Luria-Bertanimedium containing kanamycin (50 mg L21), rifampicin (50 mg L21), and gen-tamycin (25 mg L21). Cells were collected by centrifugation for 15 min at 3,000gat 20°C and then resuspended in 10 mM MES buffer containing 10 mM MgCl2and 100 mM acetosyringone (49-hydroxy-39,59-dimethoxyacetophenone) to a fi-nal OD600 of 0.4, followed by incubation at 20°C for 3 h. For coinfiltration, equalnumbers of cells from each of the cultures of strains harboring different binaryplasmids were mixed together, collected by centrifugation, and resuspended asabove to a final OD600 of 0.4 per strain. N. benthamiana plants were grown fromseeds on soil in a greenhouse with a 16/8-h day/night photoperiod at 25°C.Leaves of 4-week-old N. benthamiana plants were infiltrated using a 2-mL sy-ringe without a needle. N. benthamiana plants transformed with the binaryvector harboring EGFP alone were used as a control. The infiltrated leaves werecollected 10 d after infiltration.

To analyze the compounds produced in the leaves, the harvested plantmaterials were flash frozen and ground into a fine power in liquid nitrogen.Three grams of powder was extracted with 4 mL of MTBE. The extracts werebriefly vortexed for 3 min at maximum speed and then incubated at roomtemperaturewith shakingat 50 rpm for3h, followedbycentrifugation for 15minat 8,000g. The MTBE layer was transferred to a fresh vial, dehydrated usinganhydrous Na2SO4, and concentrated by evaporating the solvent to a finalvolume of about 0.3 mL. Analysis of the samples was performed with anRxi-5Sil column on a Shimadzu QP-2010 GC-MS system.

Plant Physiol. Vol. 176, 2018 535

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

GC-MS and LC-MS Analyses

Analytes from 1-mL samples were separated by the Shimadzu QP-2010GC-MS system equipped with the Rxi-5Sil column (30 m3 0.25 mm3 0.25 mmfilm thickness; Restek) using helium as the carrier gas at a flow rate of 1.4 mLmin21. The injector was used in split mode at a ratio of 1:2 with the inlet tem-perature set to 240°C. The initial oven temperature of 50°C was increased afterholding for 3 min to 110°C at a rate of 10°C min21, then increased to 150°C at arate of 5°C min21, held for 3 min at 150°C, increased to 300°C at a rate of 10°Cmin21, and finally held for 3 min at 300°C. Compounds were identified bycomparison of mass spectra and retention times with those of the authenticstandards, when available, or with known retention indices and mass frag-mentation from the literature and NIST library.

EachMTBE extract (4 mL)was evaporatedwith the BUCHI Rotavapor anddissolved in 0.5 mL of 70% acetonitrile and 30% water for analysis using aWaters Xevo G2-XS quadrupole time of flight-mass spectrometer interfacedwith a Waters Acquity binary solvent manager and 2777c autosampler.Samples (5 mL each) were injected onto an Acquity BEH C18 ultra-performance liquid chromatography column (2.1 3 100 mm, 1.7 mm parti-cle size; Waters) at 40°C. Initial conditions were 0.3 mL min21 99% solvent A(water + 0.1% formic acid) and 1% solvent B (acetonitrile). Following injec-tion, solvent Bwas increased in a linear gradient over 16 min to 99%, followedby a hold at 99% B for 2 min, then return to 99% A at 18.01 min and equili-bration for 2 min before starting the next sample. Ions were generated byelectrospray ionization in negative-ion mode. Capillary voltage was 2 kV,sample cone voltage was 40 V, and source temperature was 100°C. Des-olvation temperature was 350°C, and desolvation gas flow was 600 L h21.Data were acquired using a data-independent method providing both non-fragmenting and fragmenting conditions for each run, and lock mass cor-rection was performed using a Leu enkephalin standard. MS/MS analysiswas performed for selected ions in separate runs.

Hydrolysis of Modified Trans-Chrysanthemol and Trans-Chrysanthemic Acid

Acylglucosides such as chrysanthemic acid glucoside can typically be hy-drolyzed by base, but chrysanthemyl glucosides cannot. Some nonspecificglycosidases can hydrolyze both classes of compounds. Therefore,we used bothand glycosidase treatments to hydrolyze volatile and nonvolatile derivatives oftrans-chrysanthemol and trans-chrysanthemic acid produced inN. benthamianaleaves. Samples were prepared by flash freezing leaves and grinding them intoa fine power in liquid nitrogen. For base hydrolysis, homogenates (0.5 g persample) were mixed completely with 50 mL of 4 N NaOH and incubated at 80°Cfor 20 min, followed by neutralization with 50 mL of 4 N HCl, and then adding1 mL of MTBE containing 0.01 ng mL21 tetradecane as an internal standard andvortexing at maximum speed for 4 min. MTBE extracts were transferred to afresh vial, dehydrated, and concentrated to a final volume of about 0.2 mL forGC-MS analysis or dried and dissolved in 0.3 mL of 70% acetonitrile and 30%water for LC-MS analysis. For glycosidase treatment, homogenates were trea-ted with 4 mg of b-glucosidase (EC 3.2.1.21; purchased from Sigma-Aldrich) at37°C for 1.5 h and then extracted and analyzed as described above. GC-MSanalysis was used to measure volatile compounds as described above. LC-MSanalysis was performed to check for disappearance of the nonvolatile trans-chrysanthemic acid conjugates also as described above.

Accession Numbers

The sequence data used in this study can be obtained from the NCBI withthe following GenBank accession numbers: TcADH1, MF497443; TcADH2,MF497444; TcALDH1, MF497445. The bioproject accession number for theRNAseq data is PRJNA399494.

Supplemental Data

The following supplemental materials are available.

Supplemental Figure S1. Mass spectral analysis of Unknown 1, Unknown2, and Unknown 3 produced in N. benthamiana leaves expressing TcCDS,TcADH2, and TcALDH1.

Supplemental Figure S2. GC-MS analysis of hydrolysis assays of tissues ofN. benthamiana leaves expressing TcCDS.

Supplemental Figure S3. GC-MS analysis of hydrolysis assays of tissues ofN. benthamiana leaves expressing TcCDS, TcADH2, and TcALDH1.

Supplemental Figure S4. LC-MS analysis of hydrolysis assays of tissues ofN. benthamiana leaves expressing TcCDS, TcADH2, and TcALDH1.

Supplemental Figure S5. The CT mean values of TcGAPDH.

Supplemental Table S1. Transcript abundance for selected candidates inthe RNAseq database of leaves and flowers at different stages fromT. cinerariifolium.

ACKNOWLEDGMENTS

We thankDaniel Jones (Michigan State University) for helpwith the analysisof nonvolatile compounds in N. benthamiana leaves transiently expressing var-ious T. cinerariifolium genes.

Received September 15, 2017; accepted November 6, 2017; publishedNovember 9, 2017.

LITERATURE CITED

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer forIllumina sequence data. Bioinformatics 30: 2114–2120

Casida JE (1973) Pyrethrum, the Natural Insecticide. Academic Press, New YorkCasida JE, Quistad GB, editors (1995) Pyrethrum Flowers: Production,

Chemistry, Toxicology, and Uses. Oxford University Press, New YorkChen F, Tholl D, Bohlmann J, Pichersky E (2011) The family of terpene syn-

thases in plants: a mid-size family of genes for specialized metabolism that ishighly diversified throughout the kingdom. Plant J 66: 212–229

Crombie L (1995) Chemistry of the pyrethrins. In JE Casida, GB Quistad,eds, Pyrethrum Flowers: Production, Chemistry, Toxicology, and Uses.Oxford University Press, New York, pp 123–193

Davidovich-Rikanati R, Sitrit Y, Tadmor Y, Iijima Y, Bilenko N, Bar E,Carmona B, Fallik E, Dudai N, Simon JE, et al (2007) Enrichment oftomato flavor by diversion of the early plastidial terpenoid pathway.Nat Biotechnol 25: 899–901

Eisen MB, Spellman PT, Brown PO, Botstein D (1998) Cluster analysisand display of genome-wide expression patterns. Proc Natl Acad SciUSA 95: 14863–14868

Emanuelsson O, Brunak S, von Heijne G, Nielsen H (2007) Locating proteinsin the cell using TargetP, SignalP and related tools. Nat Protoc 2: 953–971

Haas BJ, Papanicolaou A, Yassour M, Grabherr M, Blood PD, Bowden J,Couger MB, Eccles D, Li B, Lieber M, et al (2013) De novo transcriptsequence reconstruction from RNA-seq using the Trinity platform forreference generation and analysis. Nat Protoc 8: 1494–1512

Iijima Y, Wang G, Fridman E, Pichersky E (2006) Analysis of the enzy-matic formation of citral in the glands of sweet basil. Arch BiochemBiophys 448: 141–149

Katsuda Y (1999) Development and future prospects for pyrethroidchemistry. Pestic Sci 55: 775–782

Katsuda Y (2012) Progress and future of pyrethroids. Top Curr Chem 314: 1–30Kikuta Y, Ueda H, Takahashi M, Mitsumori T, Yamada G, Sakamori K,

Takeda K, Furutani S, Nakayama K, Katsuda Y, et al (2012) Identifi-cation and characterization of a GDSL lipase-like protein that catalyzesthe ester-forming reaction for pyrethrin biosynthesis in Tanacetum cine-rariifolium: a new target for plant protection. Plant J 71: 183–193

Kumar S, Stecher G, Tamura K (2016) MEGA7: Molecular EvolutionaryGenetics Analysis version 7.0 for bigger datasets. Mol Biol Evol 33:1870–1874

Lewinsohn E, Gijzen M (2009) Phytochemical diversity: the sounds of si-lent metabolism. Plant Sci 176: 161–169

Li B, Dewey CN (2011) RSEM: accurate transcript quantification from RNA-Seqdata with or without a reference genome. BMC Bioinformatics 12: 323

Magnard JL, Roccia A, Caissard JC, Vergne P, Sun P, Hecquet R, DuboisA, Hibrand-Saint Oyant L, Jullien F, Nicolè F, et al (2015) Plant vola-tiles: biosynthesis of monoterpene scent compounds in roses. Science349: 81–83

Miettinen K, Dong L, Navrot N, Schneider T, Burlat V, Pollier J, WoittiezL, van der Krol S, Lugan R, Ilc T, et al (2014) The seco-iridoid pathwayfrom Catharanthus roseus. Nat Commun 5: 3606

536 Plant Physiol. Vol. 176, 2018

Xu et al.

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.

Moghe GD, Leong BJ, Hurney SM, Jones AD, Last RL (2017) Evolutionaryroutes to biochemical innovation revealed by integrative analysis of aplant-defense related specialized metabolic pathway. eLife 6: 6

Nakai K, Horton P (1999) PSORT: a program for detecting sorting signals inproteins and predicting their subcellular localization. Trends BiochemSci 24: 34–36

Ramirez AM, Saillard N, Yang T, Franssen MC, Bouwmeester HJ,Jongsma MA (2013) Biosynthesis of sesquiterpene lactones in pyre-thrum (Tanacetum cinerariifolium). PLoS ONE 8: e65030

Ramirez AM, Stoopen G, Menzel TR, Gols R, Bouwmeester HJ, Dicke M,Jongsma MA (2012) Bidirectional secretions from glandular trichomesof pyrethrum enable immunization of seedlings. Plant Cell 24: 4252–4265

Rivera SB, Swedlund BD, King GJ, Bell RN, Hussey CE Jr, Shattuck-EidensDM, Wrobel WM, Peiser GD, Poulter CD (2001) Chrysanthemyl diphos-phate synthase: isolation of the gene and characterization of the recombinantnon-head-to-tail monoterpene synthase from Chrysanthemum cinerariaefolium.Proc Natl Acad Sci USA 98: 4373–4378

Sainsbury F, Thuenemann EC, Lomonossoff GP (2009) pEAQ: versatileexpression vectors for easy and quick transient expression of heterolo-gous proteins in plants. Plant Biotechnol J 7: 682–693

Simão FA, Waterhouse RM, Ioannidis P, Kriventseva EV, Zdobnov EM(2015) BUSCO: assessing genome assembly and annotation complete-ness with single-copy orthologs. Bioinformatics 31: 3210–3212

Simpson JT, Wong K, Jackman SD, Schein JE, Jones SJ, Birol I (2009)ABySS: a parallel assembler for short read sequence data. Genome Res19: 1117–1123

Teoh KH, Polichuk DR, Reed DW, Covello PS (2009) Molecular cloning ofan aldehyde dehydrogenase implicated in artemisinin biosynthesis inArtemisia annua. Botany 87: 635–642

Xu H, Zhang F, Liu B, Huhman DV, Sumner LW, Dixon RA, Wang G(2013) Characterization of the formation of branched short-chain fattyacid:CoAs for bitter acid biosynthesis in hop glandular trichomes. MolPlant 6: 1301–1317

Yang T, Gao L, Hu H, Stoopen G, Wang C, Jongsma MA (2014) Chrys-anthemyl diphosphate synthase operates in planta as a bifunctionalenzyme with chrysanthemol synthase activity. J Biol Chem 289: 36325–36335

Yang T, Stoopen G, Yalpani N, Vervoort J, de Vos R, Voster A, VerstappenFW, Bouwmeester HJ, Jongsma MA (2011) Metabolic engineering of geranicacid in maize to achieve fungal resistance is compromised by novel glyco-sylation patterns. Metab Eng 13: 414–425

Plant Physiol. Vol. 176, 2018 537

Biosynthesis of Trans-Chrysanthemic Acid

www.plantphysiol.orgon April 14, 2020 - Published by Downloaded from Copyright © 2018 American Society of Plant Biologists. All rights reserved.