Sex pheromone biosynthetic pathways are conserved between moths and the butterfly Bicyclus anynana

Although phylogenetically nested within the moths, butterflies have diverged extensively in a number of life history traits. Whereas moths rely greatly on chemical signals, visual advertisement is the hallmark of mate finding in butterflies. In the context of courtship, however, male chemical signals are widespread in both groups although they likely have multiple evolutionary origins. Here, we report that in males of the butterfly Bicyclus anynana, courtship scents are produced de novo via biosynthetic pathways shared with females of many moth species. We show that two of the pheromone components that play a major role in mate choice, namely the (Z)-9-tetradecenol and hexadecanal, are produced through the activity of a fatty acyl Δ11-desaturase and two specialized alcohol-forming fatty acyl reductases. Our study provides the first evidence of conservation and sharing of ancestral genetic modules for the production of FA-derived pheromones over a long evolutionary timeframe thereby reconciling mate communication in moths and butterflies.

T he order Lepidoptera comprises an estimated 160,000 species 1 and is thought to have arisen in concert with flowering plants about 140 Myr ago 2 . Traditionally, it has been divided into two major subgroups comprising the moths (suborder Heterocera) and butterflies (suborder Rhopalocera) 3 , the latter having diverged from their common moth ancestors in the Early Cretaceous about 110 to 100 Myr ago 2,4 , with subsequent lineage-specific bursts of radiations occurring about 70 Myr ago 2 . The ca 18,000 described species of extant butterflies (Papilionoidea and Hesperiidae) 4 have diverged extensively from their moth relatives in a number of life history traits including among others a diurnal lifestyle, bright appearances, body structures, antennal shapes and mate-finding behaviours [5][6][7] .
The switch from nocturnal to diurnal behaviour and the corresponding increased dependence on visual communication for food and mate finding have led to the widespread assumption that this group of insects has undergone a global decrease in their olfactory capabilities. Indeed, female butterflies seem to have lost the long-distance pheromones on which their moth relatives rely almost exclusively, and mate finding is typically performed by patrolling males in stereotyped and visually oriented search behaviours 8 . Nevertheless, recent studies on the genomic architecture of the olfactory protein and receptor repertoires 9 , the neuroanatomy 10,11 and the behavioural responses to plant and conspecific odours 12,13 have begun to reveal remarkable commonalities in the olfactory system of butterflies and their nocturnal cousins, the moths.
Also, chemical signalling in the form of scent bouquets disseminated by courting males at close range has long been known to complement visual patterns in many butterfly species 8 , which may play a decisive role in species recognition and female mate choice 5,14 . The extant butterfly diversity is paralleled by a great chemical diversity in volatile male pheromone components, which include alkaloid derivatives, terpenoids, aromatics or carboxylic acids that are in many cases suggested or proven to be plant derived 5,7,12,[14][15][16] . The great variety in male scent suggests, however, that the butterfly signalling traits and underlying biosynthetic machineries have evolved multiple times independently during the course of evolution 17 . Scentreleasing structures can be found on virtually any part of the body (for example, abdominal hair pencils, wing androconia and brushes), which argues further in favour of the multiple origins of the associated odours 6 .
The double bond in D9 position is a widespread structural feature in nature owing to the presence of integral metabolic membrane fatty acyl-CoA D9-desaturases 22 . In insects, D9-desaturases serve in biosynthetic pathways towards cuticular hydrocarbons in the fruit fly Drosophila melanogaster 23,24 , scentmarking pheromones in male bumblebees 25 and long-range sex pheromones in some female moths 26,27 . However, in the latter group, molecular reconstructions of biosynthetic pathways revealed that the majority of chemical structures derive from the evolution of novel desaturase multigene families absent from other insect orders 27,28 . Specifically, D11-desaturases form a key gene lineage [29][30][31] , of which orthologues encode diverse functional classes with particular regiospecificities (for example, D6 (ref. 32), D10 (ref. 33), D11 (refs 28,31), bifunctional D10-12 (refs 34,35) and D11-13 (ref. 36)). When acting in concert with chainshortening enzymes, desaturases take part in the production of highly diverse unsaturated FA structures, precursors of the large repertoire of moth pheromones, including D9-isomers 29,30 . Owing to the fact that the D11-desaturase lineage arose in Lepidoptera before the divergence of the butterflies and moths more than 120 Myr ago 28 , butterfly pheromone production likely evolved under a scenario involving either the evolutionary conservation and specialization of moth-like D11-desaturases to serve male reproductive functions or the loss of D11-desaturases in connection with the disappearance of female pheromones in butterflies followed by recruitment of a D9-desaturase for male pheromone production.
Alcohol-producing fatty acyl-CoA reductases (FARs) catalyse the reduction of FAs to fatty alcohols and are also found widely in nature from plants 37,38 , invertebrates [39][40][41] including moths [42][43][44][45][46] , to birds 47 and mammals 48 . There is a considerable diversity of FAR genes underlying similar functions and activity profiles in spite of phylogenetic unrelatedness. This is notably illustrated by FARs involved in wax ester biosynthetic pathways in a copepod 39 , several hymenopteran 40,41 and bird 47 species and FARs involved in pheromone pathways in moth species 43,46 ; all these FARs use the ubiquitous palmitic acid (16:acid) to produce hexadecanol (16:OH), without being necessarily encoded by orthologous genes. By analogy to female moth biosynthesis, the presence of alcohol and aldehyde (the latter presumably being derivative of 16:OH) as functional groups in B. anynana male compounds suggests that reduction must occur; yet the involved molecular players could have distinct origins compared with the genes involved in female moth pheromone production.
Due to its unusual pheromone composition Bicyclus allows exploration of the evolutionary origins of the genes involved in shaping the biosynthetic modules in butterfly pheromone pathways. Specifically, we dissect the genetic and functional organization of the two major biosynthetic steps-the desaturation and the reduction steps-and using a multidisciplinary approach, we show that Bicyclus wing pheromone signalling shares its genetic basis and functional mechanism with those of moths. In addition, we explore other butterfly genomes and provide evidence for the conservation of biosynthetic gene clades in important butterfly systems such as Danaus and Heliconius. Altogether, our findings bring the first critical evidence to reconcile the evolutionary history of genes that are involved in the production of volatile chemicals in Bicyclus, with appropriate tools to start investigating the molecular underpinnings of FA secretions 49 , utilized within the butterflies.

Results
The Bicyclus pheromone pathway follows moth biosynthesis. Our first aim was to experimentally test putative biosynthetic routes that could lead to the production of adult male pheromone components in B. anynana, the Z9-14:OH and 16:Ald, two compounds likely produced de novo as part of the insect FA metabolism. The third presumptive pheromone component, trimethylpentadecan-2-ol, was not part of the study as it is a phytol-derived compound 50 expected to originate from the diet and not related to FA biochemistry.
First, we prepared wing extracts from 4-day-old male B. anynana individuals. The first male-specific component, Z9-14:OH is equally distributed in the forewing and the hindwing 18 (Fig. 1a). The second male component, 16:Ald is found almost exclusively in the hindwing, more specifically in the androconia ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms4957 (Fig. 1a). We confirmed that its immediate precursor, 16:OH, is undetectable, which is most likely the consequence of its rapid conversion into the final aldehyde form.
To validate the alternative biosynthetic routes, we carried out topical application of deuterium-labelled C16 and C14 precursors on the respective fore-and hindwing androconial areas. We applied D 3 -16:acid, D 3 -14:acid, D 9 -Z11-16:acid and D 9 -Z9-14:acid as single-compound solutions onto the androconial region of each wing. Our analyses of wing extracts revealed that deuterium atoms from D 3 -16:acid, D 9 -Z11-16:acid and D 9 -Z9-14:acid were incorporated into Z9-14:OH in both fore-and hindwings (Fig. 1c,d, Supplementary Fig. 1). In contrast, there was no label incorporation from D 3 -14:acid into Z9-14:OH (Fig. 1c,d). Deuterium atoms from D 3 -16:acid were incorporated into the 16:Ald in hindwings (Fig. 1e), whereas diagnostic ions indicative of label incorporation were absent in DMSO controls in all analyses (Fig. 1c-e).    Whereas Drosophila melanogaster harbours seven FAD loci of which DesatF, Desat1 and Desat2 encode D9-desaturases functioning in female and male long-chain cuticular hydrocarbon pheromone synthesis 23,24,52 , the domesticated silkmoth Bombyx mori genome contains at least 16 FAD paralogues (Fig. 3). Of these, four have orthologues in the fruit fly (Bmo-1 is orthologous to Desat1 and Desat2, whereas Bmo-5, Bmo-6 and Bmo-7, respectively, are orthologous to Dmel/CG15531, Dmel/CG9743 and Dmel/CG9747). Of the twelve predicted B. mori FAD genes absent in flies, five paralogues derive from an expansion of the Lepidoptera-specific D11-like gene subfamily (Fig. 3). With 10 and 11 predicted paralogues, the butterfly desaturase subfamilies thus appear well conserved with the B. mori genome (Fig. 3) and each pair of butterfly desaturase orthologues displays a high level of identity at the amino-acid level and a conserved gene structure ( Supplementary Fig. 2, Supplementary Table 1). Also, the evolution of the distinct FAD subfamilies in butterflies was likely accompanied by a successive loss of introns (Supplementary  (Fig. 4) 43,44 . Interestingly, whereas B. mori and D. plexippus harbour three gene copies of the moth-specific pheromone production subfamily (pgFAR) 44 , that is, the pheromone production enzyme Bmo-pgFAR plus two other genes named Bmo-swdb1 and Bmo-swdb2 for the silkmoth 43 , and Dpl-ctg2, Dpl-ctg3 and Dpl-ctg8 for the Monarch, respectively, H. melpomene seems to have undergone a burst of lineage-specific duplications. It harbours eight pgFAR-like orthologues, of which only one gene Hmel-ctg8 may be nonfunctional as a consequence of a frame shift in exon 1. All other predicted pgFAR-like genes (Hmel-ctg1 to Hmel-ctg7) are contained in two distinct scaffolds presumably as a result from tandem duplications and possess the characteristics of functional FARs.
Characterization of a male B. anynana biosynthetic D11-FAD. All FADs including our in silico gene predictions of butterfly orthologues share conserved protein motifs ( Supplementary Fig. 2). These structural features allowed us to screen the male Bicyclus wing transcriptome using oligonucleotide primers targeting histidine-rich regions (HIS1 and HIS3) of desaturase genes in a PCR-based approach 26 . We characterized two full-length cDNAs, one of which shared high sequence conservation with moth D9-desaturases (C184C16) (Fig. 3) and corresponded to a previously identified D9-like EST clone (see methods) 55 . In light of our in vivo data, this gene is unlikely involved in pheromone biosynthesis. The second cDNA shared high sequence conservation with the predicted D11-like genes from Heliconius (71 and 67% to Hmel-ctg0 and Hmel-ctg1, respectively) and Danaus (68% to Dpl-ctg1) and other previously characterized moth D11-desaturases (Fig. 3, Supplementary  Fig. 2). Our maximum-likelihood phylogenetic reconstruction ( Fig. 3) based on representative members of the different lepidopteran desaturase subfamilies further supports the notion that the B. anynana candidate FAD, hereafter named Ban-D11, is orthologous to members of the pheromone-producing D11-desaturase lineage (Fig. 3).
Characterization of male B. anynana pgFAR-like cDNAs. We used the predicted candidate biosynthetic butterfly FAR gene sequences to probe existing EST libraries from B. anynana 55 in search of orthologues of the pheromone FAR family alongside a RACE-PCR cDNA amplification approach followed by Sanger sequencing from male hindwing tissue. We were able to characterize two full-length FAR cDNA candidates, hereafter named B. anynana wing FARs 1 and 2 (Ban-wFAR1 and Ban-wFAR2). These share 76% identity between their coding regions. Both deduced protein sequences display high amino-acid identity with predicted FAR genes from Danaus and Heliconius (77 and 75% with Dpl-ctg3 and 76 and 64% with Hmel-ctg1, respectively) and belong to the moth pheromone gland biosynthetic FAR subfamily (pgFARs, Fig. 4).
Heterologous expression of a butterfly D11-FAD ortholog.
To test for a biochemical role in male signal production, that is, the ability of Ban-D11 to catalyse the desaturation of palmitic acid, we placed its ORF under the control of a copper-dependent CUP1 promoter then transformed the resulting pYEX-CHT-Ban-D11 plasmid in InvSc1 Saccharomyces cerevisiae yeast strain and induced protein expression through activation of the CUP1 promoter. Both D9-desaturation and elongation occur in the InvSc1 yeast, leading to the presence of minor amounts of Z11-16:acid (1,948±95 ng) that results from elongation of Z9-14:acid as seen in control yeast extracts (Fig. 5a). After addition of Cu 2 þ however, the relative production of Z11-16:acid consistently increased threefold (5,881 ± 355 ng) in yeast expressing the pYEX-CHT-Ban-D11 construct (independent samples t-test; Po0.001) compared with control cultures (Fig. 5b,c). In addition, the relative proportions of endogenous FAs remained constant with or without the addition of copper (15:Me, P ¼ 0.729; Z9-15:Me, P ¼ 0.242; relative ratio between 15:Me and Z9-15;Me, P ¼ 0.348) (Fig. 5b), indicating that Ban-D11 conferred on the yeast the ability to catalyse the desaturation of palmitic acid and produce higher amounts of Z11-16:acid. No additional monounsaturated compounds were identified, indicating that Ban-D11 encodes a functional D11-desaturase specifically using palmitic acid as a substrate. This result supports the proposed pheromone biochemical pathway towards the production of the intermediate precursor Z11-16:acid.
Heterologous expression of the butterfly FAR orthologues. To determine the putative biological role of the two characterized B. anynana biosynthetic FAR cDNA candidates, their respective ORFs were cloned under the control of a GAL1 promoter in the pYES2.1 vector and assayed in InvSc1 yeast cells. We established earlier that no long-chain fatty alcohols occur naturally in InvSc1 yeast carrying the control pYES2.1 expression vector, ensuring that the production of alcohol in yeast extracts is conferred by the heterologous insect gene 43 . Yeast cells bearing either gene construct were grown in the presence of galactose and 0.5 mM of individually supplemented Z9-14:Me or 16:Me as biosynthetic precursor. Yeast cells expressing either Ban-wFAR1 or Ban-wFAR2 catalysed the formation of primary fatty alcohols indicating that both genes are functional FARs (Fig. 6), but with distinct substrate preferences: Ban-wFAR1 specifically reduced the 16:Acyl biosynthetic precursor (Fig. 6a,c)    ARTICLE whereas Ban-wFAR2 specifically reduced the Z9-14:Acyl towards Z9-14:OH (Fig. 6b,d). Ban-wFAR1 produces eight times more 16:OH than Ban-wFAR2 produces Z9-14:OH ( Supplementary  Fig. 3), indicative not only of substrate selectivity but also distinct relative affinity for their respective substrate. Ban-wFAR1 is capable of reducing not only exogenous but endogenous 16:acid naturally occurring in yeast (Fig. 6c), and in a fraction of samples, we also detected minor amounts of Z9-16:OH (o2% of the total enzyme alcohol production, Fig. 6c, Supplementary Fig. 3), resulting from conversion of Z9-16:Me naturally occurring in yeast. Finally, both enzymes display a similar minor activity on myristic acid (Fig. 6, Supplementary Fig. 3), which also naturally occurs in yeast. However, in the insect wings myristic acid and Z9-16:acid represent less than 3.5% of the C14-C16 wing lipid content and are not biosynthetic precursors (Figs 1 and 2). In addition, semi-quantitative RT-PCR analyses indicated that in contrast to female moths, whose D11 and pgFARs implicated in pheromone production typically exhibit pheromone glandspecific expression patterns 43,54 , the functional butterfly FAD and FAR transcripts show a broader expression pattern (Fig. 7,  Supplementary Fig. 4).

Discussion
We first investigated the biochemical basis of production of the FA derivatives Z9-14:OH and 16:Ald, the two key pheromone components in Bicyclus butterfly males, and demonstrate that they are produced de novo along pheromone biosynthetic routes similar to female moths. Our labelling data support that the production of Z9-14:OH in male wings proceeds in a three-step pathway from palmitic acid. Palmitic acid, which is also known as an essential precursor for female moth pheromone biosynthesis 29 , undergoes a desaturation towards Z11-16:acid, followed by one cycle of chain shortening by b-oxidation towards Z9-14:acid and a final reduction to the corresponding alcohol (Fig. 2a). The alternative hypothesis involving a D9-desaturase (Fig. 2a) can readily be excluded based on the absence of incorporation from labelled myristic acid. Second, the 16:Ald is synthesized from palmitic acid, via a two-step pathway including reduction to hexadecanol (16:OH), presumably followed by oxidation to the final aldehyde (Fig. 2b). The biochemical pathways identified here clearly suggest that the Bicyclus male pheromone components are produced via de novo moth-like biosynthetic routes, involving both a FAD and at least one fatty acyl-CoA reductase (FAR).  H m e l-c t g 4 H m e l-c t g 1 H m e l-c tg 2 H m e l-c tg 7 D p l-c t g 3 B a n w F A R 1 B a n w F A R 2 Dpl-ctg2 Dpl -ctg 8 Hmel-ct g3 Dpl-ctg10 Dpl-ct g12 In order to find genomic evidence of conserved pheromone production pathways and associated biosynthetic genes in butterflies, we mined the available genomes of two Nymphalidae, the monarch and postman butterflies. We show that both of the butterfly genomes harbour diversified FAD and FAR gene families, among which are found orthologues to previously identified lepidopteran pheromone production genes. The butterfly desaturase families appear well conserved with other lepidopteran genomes and in agreement with the evolutionary history of dipteran and lepidopteran ancestors (Fig. 3), and the high diversification of FAR subfamilies in butterflies is concordant with the 22 predicted paralogues in B. mori (Fig. 4). There is thus a compelling similarity between the repertoires of pheromone biosynthetic genes of butterflies and moths. This is remarkable considering (i) the loss of long-range female-produced sex pheromones in butterflies, (ii) the chemical unrelatedness of many plant-derived male butterfly pheromonal odours and (iii) the evolutionary timeframe since divergence from a common ancestor.
Functional evidence for shared biosynthetic ancestry of pheromone production between moths and butterflies is first revealed through characterization and heterologous expression of the B. anynana butterfly D11-desaturase ortholog (Figs 3 and 5).
Distinct from cuticular hydrocarbon synthesis in the Diptera 23 , and similar to pheromone biosynthesis in many moth species, we demonstrate that B. anynana pheromone production does not proceed through D9-desaturation despite the conservation of metabolic 27 D9-FAD gene members in butterflies, and that D11-desaturation is of key importance in the genus Bicyclus. D11-desaturation has previously been recognized as an important innovation of the Lepidoptera that is widely used in sex pheromone biosynthetic pathways in extant ditrysian female moths 27,29,56,57 . This finding thus provides the very first demonstration of the sharing of conserved pheromoneproducing genes and biosynthetic modules between female moths and male butterflies across the Lepidoptera. We next characterize two duplicate genes of the Lepidopteraspecific pgFAR lineage in B. anynana (Fig. 4) and demonstrate that they encode specialized pheromone biosynthetic reductases (Fig. 6). The discovery of substrate-specific FAR gene functions may contrast with previously identified broad-range alcoholproducing FARs from the silkmoth 42 , small ermine moths 43 , the majority of corn borer species 44,45 and heliothine moth species 46 , adding to the degree of specialization and range of FAR biosynthetic activities found among lepidopteran species and 16  Neither the functional Ban-D11-desaturase nor the Ban-wFAR transcripts appear to be male or wing specific (Fig. 7) but their expression is nevertheless consistent with FA precursor distribution as the reaction product of Ban-D11, the Z11-16:acid and its shortened derivative the Z9-14:acid occur throughout both types of male butterfly wings (Fig. 1). We, however, did not focus specifically on investigating whether these precursors can be found in other adult tissues. In light of the broad messenger RNA distribution pattern of Ban-wFAR1, which encodes the enzyme producing the 16:OH, one could speculate that the aldehydeproducing oxidase accounting for the final biosynthetic step is likely to confer the tissue specificity of the aldehyde pheromone component. Finally, Ban-wFAR2 encodes a transcript highly expressed in both male forewings and hindwings but not restricted to the androconia (Fig. 7), again in agreement with the presence of Z9-14:OH throughout both types of wings (Fig. 1).
We show that FA desaturases and reductases orthologous to moth-biosynthetic genes are active in a butterfly, specifically B. anynana; yet, male abdominal glands of several Heliconius butterflies have also been shown to contain a large variety of FA-derived compounds including saturated and unsaturated long-chain alcohol and acetate chemicals 49,58 . This suggests that these pathways and enzymes are likely conserved in other butterfly systems. Although no detailed studies have investigated as yet the molecular underpinnings of these secretions, it is realistic to suggest that at least some of the predicted orthologous FAD and FAR genes have remained functional, however different the ethological role that FA derivatives might fulfil in butterflies in general 49 . Altogether, this supports the idea that moth-like biochemical pathways and the underlying genetic networks have survived over long evolutionary times across butterfly lineages to act in the de novo biosynthesis of diverse butterfly chemical secretions.
The Bicyclus male pheromone has been shown to play a determinant role in female mate choice by guiding their acceptance or rejection of courting males 18,19 . Individual males have also been shown to differ in their absolute pheromone titre and ratio at different ages, making the male pheromone composition both a reliable predictor of age and individuality 13 and a honest signal of an individual male's characteristics and associated fitness to prospective mates 13,18 . Nieberding et al. 13 further demonstrated that the overall increase in the titre of 16:Ald is critical for females to discriminate between young and old males whereas an overall increase in total amounts of Z9-14:OH seems to participate in how females can differentiate young (3 days old) from middle-aged males (14-day to 21 day-old individuals). The two pheromone biosynthetic FARs involved in the pathways towards the production of these two key pheromone components provide an opportunity for uncoupling the male pheromone signals and may contribute to pheromone variation in Bicyclus. Age-dependent modulation of male Bicyclus signals could take place through changes in FAR enzyme activities and provides the exciting opportunity to further investigate the mechanistic link between male condition, pheromone composition and female choice.
We show that the forewing and hindwing FARs are capable of producing 14:OH as a byproduct of the reduction step and arose following a recent gene duplication event (Fig. 4), suggesting that the ability to reduce 14:Acyl was acquired independently or more

Ban-Δ11
Ban-wFAR1 Ban-wFAR2 16s RNA  parsimoniously that the ancestral copy was less specialized. Hence, FARs characterized from moths so far are often capable of simultaneously reducing C14 and C16 substrates (except in Ostrinia 45 ), which could imply that the ancestral state in Lepidoptera was in many instances a broad-range FAR that underwent further specialization including this butterfly system. To test whether the ability to produce 16:OH and Z9-14:OH hence follows this subfunctionalization scenario or arose subsequently by neofunctionalization in one or both duplicates will, however, require further comparative molecular and functional analyses. Genetic reshuffling, that is, gene duplication, can provide a mechanism for new pheromones to arise and pheromone variation has been suggested to be positively correlated with the number of genetic elements involved in biosynthetic pathways 59 . Regardless of whether this pattern of specialization at the reduction step is unique to Bicyclus or widespread among other butterfly species, gene duplication in this species likely provided the necessary raw material that allowed the dissociation of biosynthetic gene functions. Decreasing functional constraints on gene regulatory elements 60,61 can promote independent variation of gene expression whereas relaxed constraints on the involved gene products could lead to the rapid evolution of new enzymatic functions, respectively, in turn facilitating intraspecific variation in the male signal 62,63 .
In conclusion, our central finding, that moth and butterfly lepidopteran lineages share a conserved pheromone production genetic network provides important insights into the evolutionary origin of FA-derived butterfly pheromones, and opens the door to address whether the evolution of dissociated biosynthetic pathways may have contributed to facilitate modulation in the released courtship signals.

Methods
Insect rearing. A butterfly population of B. anynana (Satyrinae) was established in Lund from hundreds of eggs originating from a lab colony stock originally derived from 80 gravid females collected in Malawi in 1988, and maintained at Leiden and at Yale University since then at a size of about 200-300 breeding individuals each generation.
Larvae were raised in a climate chamber with controlled environmental conditions including a 12L-12D light-dark cycle with 70% relative humidity and 27±1°C degree Celsius. They fed on regular supplies of fresh young maize plants until pupation. For experiments, newly emerged adults were separated on the day of eclosion (day 0) and held in single-sex cohorts with fresh banana slices as food source.
Biochemical labelling and biosynthetic pathway. In order to determine the relative wing composition of Z9-14:OH and 16:Ald in the fore-and hindwings, we excised individual forewings or hindwings of 4-day-old individuals and extracted them for 30 min in 1.5 ml glass vials containing n-hexane followed by gas chromatography (GC) analysis as described under the GC-mass spectrometry (GC-MS) analyses section below.
The aldehyde and alcohol biosynthetic pathways were probed using topical wing application of deuterium-labelled FAs. The D 3 -16:acid, D 3 -14:acid, D 9 -Z11-16:acid and D 9 -Z9-14:acid were dissolved individually in dimethylsulphoxide (DMSO) at a concentration of 20 mg ml À 1 . Natural butterfly movements can cause cross-wing contamination during incubation with the labelled chemicals and therefore, we precisely removed either the fore-or hindwings before topical application on the remaining pair of wings. Insects were anesthetized with carbon dioxide and 1 ml of each single-compound solution was applied to the androconial region, which for forewings constitutes a spot of differentiated scales located on the anal vein on the ventral side 13 .
For hindwing labelling, the labelled compound was applied onto the dorsal side of each hindwing in the region comprised between the subcostal and radial veins 13 .
The labelling was carried out 4 h before the onset of scotophase and males were thereafter kept individually for a 24-h incubation period. Labelled androconia from the two hindwings or the two forewings, respectively, were excised and extracted in 200 ml hexane for 30 min. The extracts were analysed by GC-MS using selected ion monitoring (SIM) along with wing extracts from control individuals treated with DMSO only.
To analyse the biosynthetic fatty acyl precursors, wing tissues recovered after the hexane extraction for volatile content analysis were subsequently extracted with 100 ml chloroform:methanol (2:1 v:v) for 24 h at room temperature. The extract was dried under a stream of nitrogen, and the residues were subjected to base methanolysis to convert fatty acyl moieties to the corresponding methyl esters 64 and subsequently analysed by GC-MS and SIM as described in the next paragraph.
GC-MS analysis of male wing extracts. Fore-and hindwing hexane extracts and the corresponding FA methyl esters (FAME) were analysed on a Hewlett Packard HP 6890 GC system (Agilent, Palo Alto CA, USA) coupled to a mass selective (MS) detector (HP-5975) equipped with a HP-5MS capillary column (30 m Â 0.25 mm, Agilent technologies) with helium as carrier gas and an average velocity of 30 cm s À 1 .
The oven temperature was set at 80°C, held for 1 min, then increased to 210°C at a rate of 10°C min À 1 , held for 12 min and finally increased to 250°C at a rate of 10°C min À 1 , held for 5 min. SIM was used to detect the native pheromone components and the corresponding deuterium-labelled compounds. Z9-14:OH was detected using the characteristic ions at m/z 31, 194 and 212; ions at m/z 197 and 203 were used to monitor the corresponding D 3 -Z9-14:OH and D 9 -Z9-14:OH, respectively. 16:Ald was monitored with the characteristic ions at m/z 196 and 222 and the incorporation of D 3 -16:Ald was monitored with ion at m/z 199. The identity of all components was confirmed by comparing retention times and mass spectra with those of reference standards.
Butterfly FAD and FAR gene annotation and phylogeny. The Danaus and Heliconius butterfly genomes were searched against tBLASTx databases at NCBI 65 using the deduced protein sequences of active pheromone biosynthetic moth genes as queries. All scaffolds predicted to harbour desaturase and FAR-like genes were retrieved and annotated in Geneious Pro 5.6.4 (ref. 66); intron-exon boundaries were predicted using tBLASTx, Softberry FGENESH and ExPasy prediction tools, manually curated and verified. A number of automated predictions for Danaus and Heliconius FAD proteins and for Danaus FAR proteins were made available through NCBI in the course of this study. For Danaus FADs, some automated gene predictions contained verifiable inaccuracies mainly in exon 1, for which we provide the corrected gene structures and associated nt and aa sequences in Supplementary Table 1 and Supplementary Data 1, respectively. For Danaus FARs, only a small number of genes were predicted compared with the number retrieved in our extensive in silico analysis. Scaffold and structural information for Heliconius and Danaus FAR genes are listed in Supplementary Tables 3,4, alongside accession numbers corresponding to automated predictions when available, and the associated nt and aa sequences are provided in Supplementary Data 2 and 3, respectively.
Accession numbers for moth FAD and FAR sequences, as well as ESTs, used for the phylogenetic reconstructions are listed in Supplementary Tables 2 and 5 or were given elsewhere 28,43 and are available upon request. For each set of aminoacid sequences, multiple sequence alignments were generated using MAFFT v7 with E-INS-i algorithm and the BLOSUM45 scoring matrix 67 . Maximumlikelihood inference was carried out using the standalone version of PhyML 68 and the WAG þ I þ G model as determined by performing model selection in Topali v2. 5 (ref. 69). Clade support was evaluated using 100 bootstrap replicates. The cladograms were visualized and prepared using the online tool EvolView 70 .
Molecular characterization of FAD and FAR cDNA candidates. First-strand cDNA was synthesized using a Stratascript reverse transcriptase (Stratagene) from 1 mg total RNA extracted from the portion of 6-day-old male hindwings containing the androconia and hair pencils. We opted for screening the transcriptome of this tissue based on the rationale that all candidate biosynthetic genes were to be transcribed in the adult hindwing since it produces both pheromone components.
For the desaturase cDNA screening, B. anynana EST libraries 55 contained only information for a predicted D9-candidate desaturase transcript (EST GenBank Acc. nr GE668257 (Ban-delta9-like, Fig. 3). Therefore, we performed a complete PCR-based screen for candidate desaturase genes using male hindwing androconial cDNA as template in PCR reactions with oligonucleotide primers designed against conserved desaturase motifs 31 . PCR thermal cycling conditions consisted of 95°C for 5 min, 35 cycles at 95°C for 30 s, 50°C for 45 s, 72°C for 90 s and 72°C for 10 min. An amplicon with expected size of 560-bp was gel purified (Promega), ligated into the pCR2.1-TOPO TA cloning vector system (Invitrogen) and amplified in DH5a Escherichia coli cells (New England Biolabs). Purified plasmids were sequenced using the Big Dye Terminator cycle sequencing kit v1.1 followed by analysis on a capillary ABI 3100 sequencer instrument (Applied Biosystems). Double-stranded DNA sequence information encompassing the central fragment of desaturase genes was curated using BioEdit followed by BLAST searches to verify the desaturase gene identity 65 . Clones were identified corresponding to the above-mentioned EST D9-desaturase and a D11-like desaturase for which 3 0 -and 5 0 -cDNA end termini were amplified using the SMART RACE Kit (Clontech) and gene-specific RACE primers (Supplementary Table 6).
For the reductase cDNA screening, BLASTP and TBLASTN searches were used to screen B. anynana EST databases using the Danaus and Heliconius orthologues of biosynthetic FAR candidate genes as queries. Candidate ESTs under Acc. numbers GE680761 and GE680762 (adult head EST library; DOE Joint Genome Institute project ID 16936) were assembled in a partial B. anynana contig. The assembled contig clustered with moth-biosynthetic pgFARs and was used as sequence template to design 5 0 -and 3 0 -RACE primers in NT1 Vector (Invitrogen) (Supplementary Table 6), which served to obtain the full-length cDNA sequence. The B. anynana FAR EST 3 0 -RACE primer led to the amplification of a 2 kb fragment matching perfectly with the partial EST 5 0 region and the corresponding full-length cDNA was named Ban-wFAR1. Twenty-one of the 25 DNA bases from the 3 0 -RACE primer were conserved enough to amplify a distinct 1 kb long 3 0 -RACE DNA amplicon, which we found to encode the 3 0 cDNA-end of a distinct pgFAR-like transcript as confirmed by phylogenetic analyses. A set of gene-specific 5 0 -RACE primers was designed to amplify the second clone 5 0 -cDNA-end, and the compiled full-length cDNA was named Ban-wFAR2. The sequence integrity and distinctness of the two pgFAR-like gene candidates, which share 76% identity at the nt level, was confirmed by amplifying each ORF with gene-specific primers followed by Sanger DNA sequencing.
Tissue distribution of biosynthetic transcripts. Adult butterfly tissues from fifteen CO 2 -anaesthesized 6-day-old male individuals (head, antenna, thorax, legs, abdomen, forewing androconia, hindwing androconia, fore-and hindwing (minus androconia)) were dissected using microscissors and collected in RNA later. Tissues were also collected from whole fore-and hindwing tissue from fifteen 6-day-old females. Total RNA was isolated using the RNeasy Isolation kit and a DNase purification step (Qiagen). RT-PCR reactions were carried out using the SuperScript III One Step RT-PCR System with Platinum Taq (Invitrogen) in a 25-ml reaction containing 30 ng RNA and 0.8 mM GSP (Supplementary Table 6) and cycling conditions as follows: 55°C for 30 min, 94°C for 2 min, 35 cycles of 94°C for 15 s, 55°C for 30 s, 68°C for 45 s and 68°C for 2 min. PCR products were analysed on a 2% agarose gel. Parallel RT-PCR reactions were Exo-Sap purified and sequenced to confirm gene-specific amplification.
Functional assay of Ban-D11, Ban-wFAR1 and Ban-wFAR2. Gene-specific primers (Supplementary Table 6) were designed to amplify the desaturase and FAR candidate ORFs using male hindwing androconia cDNA as template in combination with the Advantage2 PCR system (Clontech). The desaturase ORF was ligated in the pYEX-CHT expression vector at the BamH1 and EcoR1 restriction sites and each FAR ORF was cloned in the pYES2.1 TA cloning expression vector (Invitrogen) downstream the GAL1 promoter. All constructs were verified by sequencing. The empty pYEX-CHT and pYES2.1 control plasmids as well as the distinct constructs were transformed into the InvSc1 strain of the yeast S. cerevisiae (MATa his3D1 leu2 trp1-289 ura3-52) (Invitrogen) and propagated on SC-U plates containing 2% glucose.
For desaturase expression in pYEX-CHT, yeast prototrophs were selected and grown in 20 ml SC-U 2% glucose and constant agitation for 48 h at 30°C and 300 r.p.m. in a shaking incubator prior to dilution to an OD 600 of 0.4 in 250 ml flasks containing 10 ml fresh selective medium. At this stage, yeast cultures of single transformants were run in parallel with or without addition of 2 mM Cu 2 þ (resulting in CUP promoter induction), that is, the addition of 20 ml CuSO 4 1 M. Equal amounts of cells (as measured by their OD) from CUP-induced and CUP-non-induced yeast cultures were harvested by centrifugation and the supernatant was discarded. The total lipid fraction in the yeast pellet was extracted with 0.5 ml chloroform:methanol (2:1; v-v) spiked with 10 ng ul À 1 triheptadecenoin (Larodan, Sweden) as internal standard, to extract the total cell lipid content prior to base methanolysis 28 . Hexane samples were stored at À 20°C until GC-MS analyses (See GC-MS section described below).
The FAR functional assay was performed following a protocol explained in detail elsewhere 43 with alcohol-free Z9-14:Me and 16:Me (Larodan) used as biosynthetic substrates in the yeast assays. Briefly, individual prototroph colonies were inoculated in 5 ml SC-U medium and incubated for 48 h at 30°C and 300 r.p.m. (Innova 42, New Brunswick Scientific), diluted to an OD600 ¼ 0.4 to a final volume of 20 ml SC-U 2% galactose and 0.1% glucose in 250-ml flasks, and incubated for 24 h at 30°C and 200 r.p.m. Yeast cultures were diluted to 1:10 in 2 ml SC-U 2% galactose, 1% tergitol (Nonidet P-40, Sigma), and 0.5 mM alcoholfree precursors in the form of methyl esters. After incubation for 24 h at 30°C and 300 r.p.m., cells were collected by centrifugation at 2,000 r.p.m. (Labofuge 200, Heraeus Instruments) and washed in sterile water. Cell pellets were extracted with 1 ml n-hexane spiked with 150 ng Z11-13:OH as an internal standard followed by shaking at 200 cycles/min (Vibramax 100, Heidolph) for 60 min. Hexane samples were stored at À 20°C until GC-MS analysis.
GC-MS analysis of yeast extracts. Prior to analysis, hexane extracts from base-methanolysed InvSc1 yeast cultures expressing the pYEX control, and the pYEX-Ban-D11 constructs with or without Cu 2 þ induction were concentrated under a gentle flow of pure nitrogen to a final volume of 25 ml and transferred into 1.5-ml vials containing glass inserts. One microlitre was injected on a Hewlett Packard HP 6890 GC system coupled to an automatic injector (HP-7683) and a mass selective (MS) detector (HP-5975). The GC was equipped with a polar HP-INNOWax column (100% polyethylene glycol, 30 m Â 0.25 mm Â 0.25 mm; Agilent Technologies) with helium as carrier gas and an average velocity of 30 cm sec À 1 . The MS was operated in electron impact mode (70 eV), the GC oven temperature was set at 50°C for 2 min and then rose at a rate of 10°C min À 1 up to 220°C, held for 20 min.
For analysis of hexane extracts from InvSc1 yeast cultures expressing the pYES-only, the pYES-BanFAR1 or pYES-BanFAR2 construct, extracts were concentrated to 50 ml. Two microlitres were manually injected on a gas chromatograph (Hewlett Packard HP 5890II GC system) coupled to a mass selective detector (HP 5972) and equipped with a polar INNOWax column (100% polyethylene glycol, 30 m Â 0.25 mm Â 0.25 mm, Agilent Technologies). The GC-MS was operated in electron impact mode (70 eV) and the GC injector was configured in splitless mode at 220°C with helium used as carrier gas (average velocity: 20 cm s À 1 ). The oven temperature was set at 50°C for 2 min and then rose at a rate of 10°C min À 1 up to 220°C, with a final hold at 220°C for 20 min.