Ecology Letters, (2013) 16: 1157–1167 LETTER €rgens,1* Suk-Ling Wee,2 Andreas Ju Adam Shuttleworth1 and Steven D. Johnson1 doi: 10.1111/ele.12152 Chemical mimicry of insect oviposition sites: a global analysis of convergence in angiosperms Abstract Floral mimicry of decaying plant or animal material has evolved in many plant lineages and exploits, for the purpose of pollination, insects seeking oviposition sites. Existing studies suggest that volatile signals play a particularly important role in these mimicry systems. Here, we present the first large-scale phylogenetically informed study of patterns of evolution in the volatile emissions of plants that mimic insect oviposition sites. Multivariate analyses showed strong convergent evolution, represented by distinct clusters in chemical phenotype space of plants that mimic animal carrion, decaying plant material, herbivore dung and omnivore/carnivore faeces respectively. These plants deploy universal infochemicals that serve as indicators for the main nutrients utilised by saprophagous, coprophagous and necrophagous insects. The emission of oligosulphide-dominated volatile blends very similar to those emitted by carrion has evolved independently in at least five plant families (Annonaceae, Apocynaceae, Araceae, Orchidaceae and Rafflesiaceae) and characterises plants associated mainly with pollination by necrophagous flies and beetles. Keywords Amorphophallus, chemical ecology, pollination syndrome, floral evolution, flower scent, Rafflesia, Stapelia. Ecology Letters (2013) 16: 1157–1167 INTRODUCTION Flowers that mimic dung or carrion have evolved many times during the angiosperm radiation (Vereecken & McNeil 2010; Urru et al. 2011). To lure insects (typically flies or beetles) to their flowers, these plants deploy the scent and visual (and sometimes even temperature) signals that these insects use to locate oviposition sites on animal dung or carrion (Stensmyr et al. 2002; B€anziger & Pape 2004; Vereecken & McNeil 2010). There is increasing evidence that scent plays the central role in oviposition site floral mimicry systems. Volatile organic compounds (VOCs) emitted by these flowers are often identical to those recorded from samples of carrion or dung (Stensmyr et al. 2002; Johnson & J€urgens 2010). These compounds include oligosulphides (e.g. Dimethyl disulphide, DMDS; Dimethyl trisulphide, DMTS) associated with carrion, aliphatic acids generally associated with decaying materials, and p-cresol associated with dung (e.g. Kite et al. 1998; J€urgens et al. 2006; Johnson & J€urgens 2010; van der Niet et al. 2011). The function of these compounds as pollinator attractants has been demonstrated in both physiological experiments and field and laboratory bioassays (Kite et al. 1998; Stensmyr et al. 2002; Shuttleworth & Johnson 2010; van der Niet et al. 2011). Although botanists have tended to use collective terms, such as ‘saprophilous’ or ‘carrion flowers’, to refer to plants that mimic insect oviposition sites, there is increasing evidence that plants use different chemical strategies to attract saprophagous, coprophagous, and necrophagous insects respectively (J€urgens et al. 2006; Johnson & J€urgens 2010). Floral exploitation of each of these groups of insects depends on emission of particular blends of volatiles that they innately associate with their oviposition sites. These combinations of volatiles have been proposed to constitute ‘adaptive peaks’ 1 School of Life Sciences, University of KwaZulu-Natal, P. Bag X01 Scottsville, Pietermaritzburg, 3209, South Africa in ‘floral phenotype space’ as suggested by Ollerton & Raguso (2006) (see also Raguso 2003). It can be predicted that in this phenotype space, oviposition site mimics will show patterns of advergent evolution (adaptive phenotypic resemblance of one species to another, but not vice versa) relative to their models, and convergent evolution among each other when the same model is involved. Recent studies that have explored the volatile phenotype space of flowers have supported the idea that there is chemical mimicry of a range of different oviposition sites (J€urgens et al. 2006; Ollerton & Raguso 2006; Johnson & J€urgens 2010). Most of these studies have limited power as they used either presence/absence data to describe patterns in the scent chemistry of oviposition site mimics, restricted their analysis to one plant lineage only, or based their analysis on a relative small data set (e.g. Ollerton & Raguso 2006; Johnson & J€urgens 2010; Urru et al. 2011; Schiestl & D€ otterl 2012). In addition, very few of these studies have formally tested for mimicry, by comparing the volatile profiles of flowers to those of oviposition site materials, such as decaying plant material, animal dung and carrion, or for convergent evolution, by comparing volatile profiles of oviposition site mimics with those of a control group of related plants with other pollination systems. To address these shortfalls, we conducted the first large-scale study of the evolution of scent signals in plants that mimic insect oviposition sites. MATERIALS AND METHODS Data preparation for the meta-analysis To analyse patterns in VOCs among oviposition site mimics and their relatives with other pollination systems, we collated floral volatile data for 61 species that were considered or suspected to be ovi2 School of Environmental & Natural Resource Sciences, Faculty of Science & Technology, University Kebangsaan Malaysia, Bangi, Selangor, 43600, Malaysia *Correspondence: E-mail: juergensa@ukzn.ac.za © 2013 John Wiley & Sons Ltd/CNRS €rgens et al. 1158 A. Ju position site mimics and 111 species with other pollination systems. To avoid taxonomic bias, species with other pollination systems were selected to be closely related to species with oviposition site mimicry. The composite phylogeny is therefore not a random selection of species but includes species with oviposition site mimicry and closely related taxa with other pollination systems. We also included data for 19 bat-pollinated species as some flowers pollinated by bats have been reported to emit sulphur-containing compounds (e.g. Bestmann et al. 1997), and thus potentially share some emission profiles with oviposition site mimics. Data were obtained from published studies as well as unpublished studies by Shuttleworth & J€urgens (in prep.), Wee et al. (in prep.), and van der Niet et al. (in prep.). The detailed references and the data set used for the analysis are listed in the Data S1. We surveyed the literature on oviposition site mimics using an ISI Web of Sciences search (on 10 March 2012) for the keywords related to topic (A) saprophilous OR carrion OR carcass OR carrion flower* OR sapromyiophilous OR sapromyophilous OR dung beetle* OR carrion beetle* OR dung OR oviposition site AND topic (B) GC-MS OR scent OR volatile* OR odour* OR odour* OR headspace OR fragrance* OR VOC* AND topic (C) flower* OR floral OR plant* OR pollination OR pollinator* OR flower visitor* and also integrated literature from our own literature collection. To produce a consistent and comparable data set, we converted all reported values for volatile emissions into percentage amounts; in some cases, averages were calculated from several samples reported for a given plant species (see Data S1). We included mainly studies that used headspace collection techniques so that the data reflect similar techniques used for collecting scent. However, the sensitivity of methods (detection limits) has improved since 1985, the earliest study in our data set. In several studies, the precise isomers present were not determined. In these cases, compounds were grouped together under the name with the least assumption on the specific configuration of the compound. For example, lilac aldehyde or lilac alcohol may include one or more of the four possible isomers A, B, C and D. Methodological considerations for comparing data from different studies Statistical analyses of floral volatiles All identified compounds were included in the analysis, and their percentage amount was the basis for all statistical analyses. If the compound was given as ‘trace amount’ without any further information on the exact relative amount we used 0.1% for the statistical analysis. In total 184 samples (172 species, 95 genera, 36 families, 20 orders) comprising 580 compounds were included in the analysis (see Table 1, Data S1). Non-metric multidimensional scaling (NMDS) based on Bray– Curtis similarities of square root transformed values, as implemented in Primer 6 (see Clarke & Gorley 2006), was used to visualise similarities among samples. An ANOSIM test (Clarke & Gorley 2006) was performed using Primer 6 to test for differences in scent among floral oviposition site mimics and those with other pollination systems including bat-pollinated flowers. To separate the effect of plant ‘phylogeny’ from the effect of ‘pollinator type’, ANOSIM was calculated in a 2-way crossed layout (factors: group of pollinator; plant family; Clarke & Gorley 2006). ANOSIM calculates the test statistic R, a relative measure of separation between a-priori defined © 2013 John Wiley & Sons Ltd/CNRS Letter groups, based on differences of mean ranks between and within groups. R ranges between 0 and 1 ( 1) with an R-value of zero indicating completely random grouping, whereas a value of one indicating that samples within groups (e.g. pollination systems) are more similar to each other than to any sample from a different group (Clarke & Gorley 2006). Statistical significance of R was assessed by 10 000 random permutations of the grouping vector to obtain an empirical distribution of R under the null model. We graphically analysed the floral scent composition of floral oviposition site mimics (see Fig. 3) using three critera: (1) the number of species having a certain compound (x-axis), (2) the maximal relative amount reached in any species (y-axis) and (3) the average relative amount across all species (size of bubble). A high number of species emitting a given compound may indicate that this compound might play a functional role, particularly if this compound is always present in high relative amounts. However, alternative chemical strategies for pollinator attraction may be detected if single uncommon compounds reach a high relative amount in only one or a few species. We used principal components analysis (PCA) in PAST (Hammer et al. 2001) as a multivariate tool to visualise the different chemical strategies found within the species that mimic floral oviposition sites. The advantage of PCA is that the multivariate dataset can be visualised as a set of coordinates that provides a lower dimensional picture of the data. It can thus be used as a tool for revealing the inner structure of the VOC data. Identification of chemical groupings was based on the loadings of compounds on the PCA axes and examination of the ordination. Because compositional floral scent data are bounded (see Ranganathan & Borges 2011) and therefore violate a distribution assumption of PCA, we followed the recommendation of Aitchison & Egozcue (2005) to perform centred log-ratio transformation (clr) after replacing zero values using the rounded zero replacement algorithm in the software program CODAPACK v 2.01.8 (available at http://imae. udg.edu/codapack/; Department of Computer Sciences and Applied Mathematics at the University of Girona, Spain; see also Thiό-Henestrosa & Comas 2011). The same method was used to transform compound class data in the phylogenetically informed generalised estimating equations (GEE) models (see below). Phylogenetically informed analyses Our analyses were based on a phylogeny constructed using the online software PHYLOMATIC (Webb & Donoghue (2005); http:// phylodiversity.net/phylomatic/html/pm1.html), and resolved further using published phylogenies (see Data S1). Branch lengths were scaled to be approximately equal to time, with divergence time estimates obtained from Bell et al. (2010) using the ‘BLADJ’ function in PHYLOCOM version 3.41 (Webb et al. 2006). Phylogenetically informed analyses were performed with the APE library of the R package (R version 2.15.0) (Paradis et al. 2004). To test for putative correlations between the percentage of various compound classes (sulphur compounds, benzenoids, monoterpenoids and aliphatic acids) and compounds (DMDS, p-cresol, indole, phenol) in the total scent emission of each species and the two categories of pollination systems considered (oviposition site floral mimicry systems, other pollination systems) we used GEE models that incorporate a phylogenetic distance matrix that accounts for species relatedness (Paradis & Claude 2002; Paradis et al. 2004). The applicability of GEE analyses for comparative studies was demon- Letter Evolution of floral scent in plants 1159 Table 1 Main floral scent compounds of 61 plants with oviposition site floral mimicry and the response of insects (+ = attractive; = repellent; 0 = indifferent; nr = no EAG response) to these compounds as reported by listed authors (for references see supplementary material S1). For complete list of compounds, species and references see Data S1 Relative amounts Occurrence Fly response Compound Compound class S G F Max Mean Physiological Behavioural Dimethyl disulphide (DMDS) Dimethyl trisulphide (DMTS) p-Cresol Octanal Benzaldehyde Hexanoic acid Linalool 4-Methylpentanoic acid Phenol Acetoin Butanoic acid 3-Methyl-1-butanol Benzoic acid (E)-Ocimene Indole Trimethylamine (TMA) Butyl acetate Ethyl acetate Acetophenone oct-1-en-3-ol b-Citronellene a-Pinene Benzyl alcohol 3-Isopentyl 2,5-dimethyl-pyrazine Heptanal Geranyl acetone Phenylethyl alcohol 3-Methyl-2-pentanone Decanoic acid Limonene (E)-Oct-1-en-5-ol Nonanal Dimethyl tetrasulphide Dimethyl sulphone Ethanol (E)-Oct-2-en-1-ol Heptan-2-one 3,7-Dimethyl-1-octene 3-Methylbutanoic acid Heptan-2-ol b-Caryophyllene Acetic acid Decanal Nonanoic acid SCC SCC BC AC, aldehyde BC, aldehyde AC, acid MT, acyclic AC, acid BC AC, ketone AC, acid AC, alcohol BC, acid MT, acyclic NC NC AL, ester AC, ester BC AC, alcohol MT MT, cyclic BC, alcohol M AC, aldehyde IT BC, alcohol AC, ketone AC, acid MT AC, alcohol AC, aldehyde SCC SCC AC, alcohol AC, alcohol AC, ketone AC, alkene AC, acid AC, alcohol Sesquiterpene AC, acid AC, aldehyde AC, acid 34 34 14 16 31 10 14 2 13 6 10 5 11 8 18 2 1 6 9 6 4 14 24 1 13 13 12 14 13 18 1 8 10 8 5 1 11 2 8 7 15 7 6 12 16 15 10 12 21 9 11 1 10 2 8 1 9 5 15 2 1 3 8 6 2 10 21 1 11 9 9 10 11 13 1 7 4 7 2 1 8 7 7 5 10 5 6 11 7 7 5 2 9 3 8 1 3 2 3 1 2 5 7 2 1 2 2 4 1 4 8 1 2 1 5 2 1 5 1 6 3 1 1 1 6 1 3 2 5 4 5 2 93.8 96.0 72.7 29.6 31.9 95.0 41.5 100.0 48.3 39.9 56.5 67.4 22.3 67.5 28.4 85.0 83.2 60.0 39.0 47.5 46.0 30.9 18.0 60.3 16.7 10.1 27.1 15.6 9.0 14.3 39.6 15.8 15.0 34.3 18.7 38.0 7.9 26.4 27.9 30.2 6.8 18.7 20.5 8.9 17.3 14.1 3.3 2.9 2.4 2.1 2.0 1.9 1.8 1.7 1.6 1.6 1.5 1.5 1.4 1.4 1.3 1.2 1.2 1.2 1.2 1.1 1.1 1.0 1.0 0.7 0.7 0.7 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.5 0.5 0.5 0.5 0.5 +MF5,11 +MF5,12,11 +CP14, nrMF5 +CF1, +MF1,-MF12 Models in which VOCs were reported nrMF5 C5, D6, E6,7,10 A6, C5, E6,7,10,F12 A15,B8,15,C5 D9,E10 A6,B8,D6,9,E7 C5,E7,10 +MF5 A6,B8,C5,D6,9,E10 +SC3 +MF5 +MF4 5 4 +MF +MF2 +MF +OF13,+MF13 nrMF5 C5,D6,E7,10 A15,B15 A6,B8,C5,D6,9,E7 E7,10 E10 A15,E7 D6,E7,10 A15,B15 A15,B15,C5 A6,B8,D6 D9,E7,10 +MF5 C5 nrMF5 C5 A6,B8,15,C5, E10 SC3 A15,B8,D6,9,E7,10 C5 B8 E7,10 +MF5 nrMF5 C5,D6 A15,B8,15 C5,D6,E7,10 8 B,E7,10 E10 Occurrence: S = Number of species, G = genera, or F = families were a compound was detected. Compound Class: AL = aliphatic compound, BC = benzenoids, NC = nitrogen compounds, SCC = sulphur compounds; MT = monoterpenoids, ST = sesquiterpenoids, M = Miscellaneous. Response: CF = calliphroids; MF = muscids; SC = sarcophagids; OF = other flies; NB = necrophagus beetles; CP = coprophagus beetles. Models: A = horse faeces; B = cattle faeces; C = pig manure; D = dog faeces; E = dead vertebrate; F = fungi. strated by Paradis & Claude (2002). GEEs incorporate a phylogenetic distance matrix into the framework of a general linear model (Paradis & Claude 2002) and they are more general than most comparative methods such as Phylogenetic Generalised Least Squares (PGLS) and Phylogenetically Independent Contrasts (PICs). They are especially suitable in analyses of data that include categorical, as predictor, and continuous, as response, variables (Paradis 2006). The sensitivity of our results to branch length and phylogenetic uncertainty was tested by the following: (1) performing all phylogenetically informed analyses across a set of 50 trees where polytomies were randomly resolved using the MESQUITE software package for phylogenetic computing (MESQUITE 2.74; Maddison & Maddison 2010) and (2) comparing results for scaled branch lengths to results when setting all branch lengths to 1. We also tested whether the occurrence of sulphur compounds was significantly different © 2013 John Wiley & Sons Ltd/CNRS €rgens et al. 1160 A. Ju between species with oviposition site floral mimicry systems compared to all other species using Pagel’s (1994) correlation test that is part of the correl package in the MESQUITE software program. Pagel’s test uses likelihood to test whether the evolution of two binary characters, in our case pollination system (oviposition site floral mimicry systems vs. other) and presence of sulphur compounds, is independent. RESULTS VOCs reported from floral oviposition site mimics The 61 plant species known or suspected to mimic insect oviposition sites represent 11 plant families distributed across the core Eudicots, Magnoliids and Monocots (Fig. 1). In total, 370 VOCs have been identified in the floral headspace of these species. Many of these VOCs have been shown to elicit electrophysiological and behavioural responses of insects (particularly flies and beetles), and to be present in the typical emissions from decaying plant material, fungi, animal dung and carrion (Table 1, Data S1). The oligosulphides DMDS and DMTS are by far the most common VOCs emitted from flowers that mimic insect oviposition sites (Figs 1 and 2). Multivariate analysis of VOC data The multivariate analysis showed that oviposition site mimics form a distinct group based on their VOC composition (2-way ANOSIM with pollination system and family as factors) R = 0.451; P < 0.01 (Fig. 3). Species with other pollination systems, mainly visited by nectarivorous insects, show high relative amounts of monoterpenoids and aromatic aldehydes, alcohols and esters (Fig. 2), while those visited by insects that oviposit on dead or decaying organic material are characterised by emission of volatiles indicative of microbial degradation/fermentation of plant and animal material, such as sulphur compounds (e.g. DMDS, DMTS), nitrogen containing compounds (e.g. indole, skatole), aliphatic acids accompanied by aliphatic alcohols and esters, and much lower amounts of monoterpenoids and aromatic compounds (Figs 2 and 3). Furthermore, within the general syndrome of oviposition site mimicry, separation between plants that mimic decaying plant material, fermenting carbohydrate sources and decaying animal material (degrading protein) is evident (Fig. 3). This division is supported by a principal components analysis, based on 20 compounds that make up the highest relative amounts in the data, which showed that three different chemical strategies are evident (Fig. 4). In the first group, we find species dominated by DMDS and DMTS, compounds that are normally products of microbial degradation of the amino acids methionine and cysteine (Fig. 4a). The second group consists of species dominated by aromatic compounds and hydrocarbon aldehydes, compounds that are typically found as products of microbial degradation of lipids and amino acids (Fig. 4b). In the third group, we Letter find species dominated by hydrocarbon acids, alcohols and esters, compounds often found as fermentation products of carbohydrates and amino acids [Fig. 4c; see Dekeirsschieter et al. (2009); Paczkowski & Sch€utz (2011) and additional references in both papers]. Flowers with the highest similarity to the scent of carrion form a core group due to their relatively high emission of DMDS and DMTS and low relative amounts of aromatics and monoterpenoids (Figs 3 and 4). Phylogenetically informed analysis of VOC data Using phylogenetic GEE, we found significant differences between pollination systems in the relative amounts of sulphur compounds emitted (Table 2), both for models using phylogenies with branch lengths calculated from dated trees and for those with branch lengths set arbitrarily to 1. Higher relative amounts of monoterpenoids and benzenoids in species not associated with oviposition site mimicry were only significantly higher when branch lengths were set arbitrarily to 1. No differences were found in the relative amounts of aliphatic acids (Table 2). For some of the single compounds (DMDS, p-cresol, indole and phenol), we found also significant differences between pollination systems tested (Table 2). Oviposition site mimicry systems emitted significantly higher amounts of DMDS and indole. However, for indole these differences were only found using phylogenies with branch lengths of 1. No differences between pollination systems were found for phenol and p-cresol (Table 2). Similarly, analyses using Pagel’s (1994) correlation test with data for 172 species confirmed that the occurrence of sulphur compounds (likelihood difference = 25.26, P < 0.001) or aliphatic acids (likelihood difference = 14.69, P < 0.01) is associated with oviposition site mimicry, whereas the occurrence of monoterpenoids is associated with species not belonging to the oviposition site mimicry (likelihood difference = 14.38, P < 0.01). No differences were found in the occurrence of benzenoids (likelihood difference = 4.13, P = 0.39); for details see also Data S2. DISCUSSION VOCs emitted from floral oviposition site mimics and their role as pollinator attractants Our analysis shows that floral oviposition site mimicry makes up a distinctive scent syndrome within the general odour space of angiosperms. The VOCs of floral oviposition site mimics are universal infochemicals widely occurring in nature and representing a wide range of different degradation/fermentation products of plant and animal material (Table 1). Within the scent syndrome of floral oviposition site mimicry, three main groups are distinguishable: (A) species emitting compounds indicative of (microbial) degradation of protein (including true carrion flowers; see Paczkowski & Sch€utz Figure 1 Composite phylogeny of 172 angiosperm species used to analyse floral scent characteristics of species with oviposition site floral mimicry systems (n = 61). Species names indicate pollination system; red (pollination by saprophilous insects suspected or confirmed, see Data S1), green (bat pollination), black (other pollination systems). Red branch colour indicates occurrence of sulphur compounds. Phylogenetically informed analyses confirmed that the occurrence and amount of sulphur compounds was correlated with oviposition site floral mimicry systems. Numbers at nodes indicate reference points according to Bell et al. (2010) used for calculating branch lengths (see Data S1). Representative flowers: A = Rafflesia cantleyi, B = Orbea variegata, C = Orbea verrucosa, D = Huernia hystrix, E = Stapelia leendertziae, F = Aristolochia cymbifera, G = Eucomis bicolor, H = Ferraria crispa, I = Satyrium pumilum. See Data S3 for a larger version of Figure 1. © 2013 John Wiley & Sons Ltd/CNRS Letter Evolution of floral scent in plants 1161 (a) (f) (b) (g) (c) (d) (h) (e) (i) © 2013 John Wiley & Sons Ltd/CNRS €rgens et al. 1162 A. Ju Letter (a) (b) Figure 2 (a) Floral volatile organic compounds (VOCs) in 61 species (370 compounds) with oviposition site floral mimicry systems; (b) Floral VOCs in 92 species (328 compounds) with other pollination systems (e.g. bees, flies, butterflies, moths, hawkmoths, beetles). Number of species with a given compound (x-axis), maximum relative amount (%) reached in any species (y-axis), and average relative amount (size of bubble; see lower right corner for examples of average relative amounts of 1, 2 and 5%). © 2013 John Wiley & Sons Ltd/CNRS Letter Evolution of floral scent in plants 1163 (a) (b) Figure 3 Non-metric multidimensional scaling of floral volatile organic compounds of 172 species of flowering plants with different pollination systems. (a) Species labelled according to pollination system. (b) Species labelled according to plant family. 2-D Stress value = 0.27. © 2013 John Wiley & Sons Ltd/CNRS €rgens et al. 1164 A. Ju Letter (b) (c) (a) Figure 4 Principal component analysis of floral volatile organic compounds in 61 species with oviposition site floral mimicry systems. Analysis was based on the 20 compounds with the highest relative amounts (see Table 1). The principal components analysis (PCA) revealed that species can be divided into three chemical groups. (A) Species dominated by DMDS and DMTS; compounds that are normally products of microbial degradation of the amino acids methionine and cysteine. (B) Species dominated by aromatic compounds and hydrocarbon aldehydes; these compounds are typically found as products of microbial degradation of lipids and amino acids. (C) Species dominated by hydrocarbon acids, alcohols and esters; these compounds are often found as fermentation products of carbohydrates and amino acids [see Dekeirsschieter et al. (2009); Paczkowski & Sch€utz (2011) and additional references in both papers]. 2011 and references therein), (B) species emitting chemicals indicative of microbial degradation (and fermentation) of protein plus carbohydrates (see Schulz & Dickschat 2007 and references therein), (C) species emitting compounds indicative of (microbial) degradation of protein plus fat (Fig. 4; see Schulz & Dickschat 2007 and references therein). Most of the dominant VOCs in the scent of floral oviposition site mimics are known to elicit responses at the antenna and/or receptor level and/or be attractive to insects that oviposit on dead or decaying plant material (Table 1). The most characteristic compound class in plants pollinated by saprophilous insects are the sulphur compounds (Fig. 1). Some other compounds, like 4methyl pentanoic acid or 3-methyl-1-butanol, occurred only in one or very few species but in high relative amounts. They might represent specific pollinator attractants, but their role in this regard needs to be tested (see Fig. 2). 3-Methyl-1-butanol has been reported as the main compound emitted during fermentation of a sucrose solution with baker’s yeast (Goodrich et al. 2006) and may be an important infochemical indicative of fermentation. The aromatic compound benzaldehyde was the most common compound in oviposition site floral mimicry systems at the genus and family level (Table 1, Fig. 2) and this compound is one of the most widespread compounds emitted from angiosperm flowers (Knudsen et al. 2006). It would thus seem unlikely that this compound could act as a specific signal. However, benzaldehyde has been © 2013 John Wiley & Sons Ltd/CNRS reported from VOC samples of carrion (Paczkowski & Sch€utz 2011) and Sarcophagidae have been shown to be attracted to benzaldehyde (James 2005). Convergent evolution of sulphur compounds in oviposition site mimics Oviposition site mimics characterised by high relative amounts of sulphur compounds and low relative amounts of monoterpenes and benzenoids have evolved independently in five angiosperm families: Annonaceae, Apocynaceae, Araceae, Rafflesiaceae and Orchidaceae (Fig. 1) and include angiosperm species like Duguetia cadaverica, Rafflesia cantleyi, Stapelia gigantea, Helicodiceros muscivorus (Dead-horse arum) and Amorphophallus titanium (Titan arum). The differences in the results of our phylogenetic GEEs when using chronograms (branch length proportional to divergence time) compared to phylogenetic trees with branch lengths set arbitrary to 1 suggests that the model of evolution is an important aspect to consider when analysing scent data. Several studies have shown that branch lengths are an integral component of the statistical null model because they indicate how likely a given trait changes from one node to another along a phylogenetic tree (see Garland et al. 2005 and references therein). One possible scenario is that the rate of evolution for floral scent compounds is not constant and that phenotypes respond rapidly to changes in the environment by switching biosynthetic pathways on Letter Evolution of floral scent in plants 1165 Table 2 Results for phylogenetic generalised estimating equations (GEE) models of composition VOC data testing for difference in the total relative amounts of (A) compound classes and (B) single compounds among 172 angiosperm species that were classified as associated with (i) saprophilous insects (n = 61) and (ii) other animal pollinators (n = 111) Trait Branch length (A) Compound classes Monoterpenoids Chronogram 1 Sulphur compounds Chronogram 1 Benzenoids Chronogram 1 Aliphatic acids Chronogram 1 (B) Single compounds Dimethyl disulphide Chronogram 1 p-Cresol Chronogram 1 Phenol Chronogram 1 Indole Chronogram 1 d.f. Median t CI of t Median P 58.2 30.6 58.2 30.6 58.2 30.6 58.2 30.6 1.49 2.24 2.03 3.37 1.81 2.62 0.93 0.95 0.33 0.35 0.39 0.36 0.27 0.30 0.31 0.29 0.14 0.03* 0.047* 0.002* 0.08 0.014* 0.35 0.32 58.2 30.6 58.2 30.6 58.2 30.6 58.2 30.6 6.12 5.79 0.93 0.19 0.29 0.35 4.82 1.70 0.18 0.16 0.08 0.01 0.02 0.01 0.08 0.03 < 0.001* < 0.001* 0.34 0.85 0.774 0.73 0.019* 0.099 Significant results are indicated by an asterisk. Phylogenetically informed analyses were performed across a set of 50 trees where polytomies were randomly resolved using Mesquite software package for phylogenetic computing (MESQITE 2.74; Maddison & Maddison 2010). Results for scaled branch lengths (Chronogram) and setting all branch lengths to 1 are given. The reported median t and P values are based on the results from 50 trees. and off (Dudareva & Pichersky 2000). Another aspect that might complicate the situation is that different plants may synthesise the same VOCs in different ways. For example, sulphur-containing compounds may originate from methionine and cysteine. In Arabidopsis, methanethiol is produced by the cleavage of methionine, while in Allium cepa (onion) and A. sativum (garlic) a series of volatile sulphur compounds are generated by cleavage of S-alk(en)yl cysteine sulfoxide (Schwab et al. 2008 and references therein). The functional role of sulphur compounds as pollinator attractants in the carrion fly-pollinated Dead-horse arum has been demonstrated by Stensmyr et al. (2002). Shuttleworth & Johnson (2010) found DMDS and DMTS in two South African carrion fly-pollinated Eucomis species. Moreover, they showed that experimental addition of these compounds to the scent of closely related species induced a shift to fly pollination, suggesting that sulphur compounds may mediate pollinator shifts in the genus. Although many VOCs have been reported to be emitted from carrion, DMDS is reported to be produced at an early stage in the post-mortem decay process (e.g. Statheropoulous et al. 2005, 2007). It should be noted that the VOC composition emitted from carrion and dung (but also that of flowers) may change over time (see Paczkowski & Sch€utz 2011 and references therein). Thus, Fig. 3 represents only a snapshot of a more dynamic picture, where species and biological material change with time in odour space. The fact that DMDS (and other oligosulphides) are early indicators of decaying animal material might explain why this compound is such a strong attractant to necrophagous flies (Stensmyr et al. 2002; Shuttleworth & Johnson 2010; van der Niet et al. 2011). Many necrophagous flies oviposit or larviposit on carrion only during early decay stages to offer more resources for their offspring but also to reduce competition and risk of fungal infection (Archer & Elgar 2003). From this, it can be hypothesised that the strong selection pressure on necrophagous insects to evolve a fine tuned olfactory system for detecting early decaying stages of carrion has led to the convergent evolution of oligosulphide emissions in floral oviposition site mimics. Interestingly, the combined occurrence of DMDS and DMTS might have consequences for pollinator attraction and pollinator specificity in different ways. In the carrion beetle Nicrophorus vespilloides (L.) (Silphidae: Nicrophorinae), DMDS alone was not sufficient for precise orientation to the odour source and a combination of several sulphur-containing compounds was required (Kalinova et al. 2009). A synergetic effect of DMDS and DMTS on N. vespilloides was also reported in a study using pitfall traps by Podskalska et al. (2009). DMTS might, however, act as a repellent for other insect species with different trophic preferences since this compound has been reported as a fungal VOC emitted by Phoma and Rhizopus strains; and the presence of DMTS significantly reduced oviposition by Musca domestica (Lam et al. 2010). Flower visitors to many of the species that emit typical carrion compounds are, as expected, also attracted to actual carrion. However, in an unusual case, flowers of the tropical understory tree Duguetia cadaverica appear to mimic a stinkhorn fungus (Phallales) and thereby attract a specialised group of beetles that feed on decaying fruiting bodies of the fungus (Teichert et al. 2012). This example demonstrates that the emission of VOCs from flowers may reflect various successional stages of a multi-trophic system including the presence of bacteria and fungi. Since reduction of competition with fungi that colonise carrion and/or faeces plays an important role in host selection for coprophagous fly species (Lam et al. 2010), it is likely that VOCs typically emitted by fungi, e.g. several eight-carbon alcohols and ketones such as oct-1-en-3-ol, octan-3-ol, octan-1-ol and octan-3-one (e.g. Combet et al. 2006), may affect pollinator composition. The occurrence of VOCs, indicative of fungi in the floral bouquet might therefore be a filter mechanism making the flower unattractive to insects with strong preferences for early decomposition stages. However, in some systems where pollinators are normally attracted to fungi as breeding sites fungal VOCs are likely to play a key role for pollinator attraction (Kaiser 2006; Teichert et al. 2012). Visualisation of our data via NMDS shows that the floral scent of many New World bat-pollinated species, and some Old world species, is convergent with that of some floral oviposition site mimics, on account of the shared emission of high relative amounts of sulphur compounds (Figs 1 and 3). However, bat-pollinated species overall are more widely distributed in the odour space (Fig. 3) since the majority of Old World species do not emit sulphur compounds (Pettersson et al. 2004). Sulphur compounds have been shown to be potent attractants of flower-visiting bats of the genus Glossophaga (von Helversen et al. 2000). Although flowers of the New world species Lecythis poiteaui have been reported as breeding sites for stratiomyid flies that are thought to be generalist associates of decaying matter of plant and animal origin (Feinstein et al. 2008), these flies do not appear to contribute to pollination of these flowers. Since bat- and carrion fly-pollinated flowers partition their pollinators despite overlap in their volatile blends, we infer that features such as morphological traits (including height above ground), presence and amount of nectar, and acoustic cues play additional roles in the specialisation of these very different pollination systems. © 2013 John Wiley & Sons Ltd/CNRS €rgens et al. 1166 A. Ju Faecal mimicry It has been suggested that differences in VOCs from faeces of different animals can be related to their diet (B€anziger & Pape 2004). The higher emission of sulphur compounds from faeces of carnivores compared to herbivores is due to the higher content of the sulphur-containing amino acids methionine and cysteine in animal protein compared to plant protein (see Magee et al. 2000). This is supported by our analysis showing that the VOCs of lion faeces (carnivore) were more similar to carrion than dog faeces (omnivore), whereas horse faeces (herbivore) were clearly different to carrion and omnivore/carnivore faeces (Fig. 3). Since coprophagous beetles and flies show trophic preferences to faeces of different animals (Hanski 1987) the groupings found in scent space (Fig. 3) might therefore reflect adaptations to olfactory preferences of insects that have specific faecal preferences. There needs to be a better understanding of how the distribution of carnivores or herbivore herds has influenced the evolution of different chemical strategies found in oviposition site mimics. In conclusion, the convergent evolution of specific VOC patterns among angiosperms pollinated by saprophilous insects appears to reflect exploitation of the pre-existing olfactory adaptations and preferences of these insect for different types of decomposing plant and animal material. In particular, these patterns in the angiosperm floral scent space seems to reflect the importance of universal infochemicals that serve as indicators for the main nutrients utilised by saprophagous, coprophagous and necrophagous animals, namely products of protein degradation, protein plus fat degradation and fermented carbohydrates plus protein degradation. It thus seems likely that differences among dung and carrion insects in their preferences for certain nutrient compositions and successional stages of the nutrient sources has driven the evolution of specific signals in these floral mimics (see Schiestl & D€ otterl 2012), thus supporting Raguso’s (2003) concept of ‘…adaptive odour peaks in phenotype space’. ACKNOWLEDGEMENTS S.-L. Wee is supported by Malaysian Science Toray Foundation (STGL-011-2008) and Fundamental Research Grant Scheme (UKMST-06-FRGS015-2010). A. Shuttleworth was supported by a University of KwaZulu-Natal postdoctoral fellowship. S.D. Johnson and A. J€ urgens are supported by the National Research Foundation (South Africa). We are grateful to Nina Hobbhahn for valuable discussions on phylogeny based statistical methods. AUTHORSHIP AJ designed and performed the literature survey, data collection, statistical testing and wrote the manuscript. S-LW collected the floral scent of Rafflesia and wrote the manuscript. AS performed the floral scent analysis of several stapeliads and wrote the manuscript. SDJ assisted with the statistical analyses and wrote the manuscript. REFERENCES Aitchison, J. & Egozcue, J.J. (2005). Compositional data analysis: where are we and where should we be heading? Math. Geol., 37, 829–850. Archer, M.S. & Elgar, M.A. (2003). Effects of decomposition on carcass attendance in a guild of carrion-breeding flies. Med. Vet. Entomol., 17, 263–271. © 2013 John Wiley & Sons Ltd/CNRS Letter B€anziger, H. & Pape, T. (2004). Flowers, faeces and cadavers: natural feeding and laying habits of flesh flies in Thailand (Diptera: Sarcophagidae, Sarcophaga spp.). J. Nat. Hist., 38, 1677–1694. Bell, C.D., Soltis, D.E. & Soltis, P.S. (2010). The age and diversification of the angiosperms re-visited. Am. J. Bot., 97, 1296–1303. Bestmann, H.J., Winkler, L. & von Helversen, O. (1997). Headspace analysis of volatile flower scent constituents of bat-pollinated plants. Phytochemistry, 46, 1169–1172. Clarke, K.R. & Gorley, R.N. (2006). Primer v6: User Manual/Tutorial. Primer-E Ltd, Plymouth. Combet, E., Henderson, J., Eastwood, C.D. & Burton, K.S. (2006). Eight-carbon volatiles in mushrooms and fungi: properties analysis and biosynthesis. Mycoscience, 47, 317–326. Dekeirsschieter, J., Verheggen, F.J., Gohy, M., Hubrecht, F., Bourguignon, L., Lognay, G. et al. (2009). Cadaveric volatile organic compounds released by decaying pig carcasses (Sus domesticus L.) in different biotopes. Forensic Sci. Int., 189, 46–53. Dudareva, N. & Pichersky, E. (2000). Biochemical and molecular genetic aspects of floral scents. Plant Physiol., 122, 627–633. Feinstein, J., Purzycki, K.L., Mori, S., Hequet, V. & Berkov, A. (2008). Neotropical stratiomyids (Diptera) reared from fallen flowers of Lecythis poiteaui Berg (Lecythidaceae), Do bat-pollinated flowers attract soldier flies? J. Torrey Bot. Soc., 135, 200–207. Garland, T. Jr, Bennett, A.F. & Rezende, E.L. (2005). Phylogenetic approaches in comparative physiology. J. Exp. Biol., 208, 3015–3035. Goodrich, K.R., Zjhra, M.L., Ley, C.A. & Raguso, R.A. (2006). When flowers smell fermented: the chemistry and ontogeny of yeasty floral scent in pawpaw (Asimina triloba: Annonaceae). Int. J. Plant Sci., 167, 33–46. Hammer, Ø., Harper, D.A.T. & Ryan, P.D. (2001). PAST: Paleontological Statistics software package for education and data analysis. Palaeontol. Electronica, 4, 9. Hanski, I. (1987). Nutritional ecology of dung- and carrion-feeding insects. In: Nutritional Ecology of Insects, Mites, and Spiders (eds Slansky, F. Jr & Rodriguez, J.G.). John Wiley & Sons, Inc., New York, pp. 837–884. von Helversen, O., Winkler, L. & Bestman, H.J. (2000). Sulphur containing ‘perfumes’ attract flower-visiting bats. J. Comp. Physiol. A., 186, 143–153. James, D.G. (2005). Further field evaluation of synthetic herbivore-induced plant volatiles as attractants for beneficial insects. J. Chem. Ecol., 31, 481–495. Johnson, S.D. & J€urgens, A. (2010). Convergent evolution of carrion and faecal scent mimicry in a stinkhorn fungus and fly-pollinated angiosperm flowers. S. Afr. J. Bot., 76, 796–807. J€urgens, A., D€ otterl, S. & Meve, U. (2006). The chemical nature of fetid floral odours in stapeliads (Apocynaceae-Asclepiadoideae-Ceropegieae). New Phytol., 172, 452–468. Kaiser, R. (2006). Flowers and fungi use scents to mimic each other. Science, 311, 806–807. Kalinova, B., Podskalska, H., R uzicka, J. & Hoskovec, M. (2009). Irresistible bouquet of death—how are burying beetles (Coleoptera: Silphidae: Nicrophorus) attracted by carcasses. Naturwissenschaften, 96, 889–899. Kite, G.C., Hetterscheid, W.L.A., Lewis, M.J., Boyce, P.C., Ollerton, J., Cocklin, E. et al. (1998). Inflorescence odours and pollinators of Arum and Amorphophallus (Araceae). In: Reproductive Biology (eds Owens, S.J. & Rudall, P.J.). Royal Botanical, Gardens, pp. 295–315. Knudsen, J.T., Eriksson, R., Gershenzon, J. & Stahl, B. (2006). Diversity and distribution of floral scent. Bot. Rev., 72, 1–120. Lam, K., Tsang, M., Labrie, A., Gries, R. & Gries, G. (2010). Semiochemicalmediated oviposition avoidance by female house flies, Musca domestica, on animal feces colonized with harmful fungi. J. Chem. Ecol., 36, 141–147. Maddison, W.P. & Maddison, D.R. (2010). Mesquite: A Modular System for Evolutionary Analysis. Version 2.73. Available at: http://mesquiteproject.org. Last accessed 20 April 2013. Magee, E.A., Richardson, C.J., Hughes, R. & Cummings, J.H. (2000). Contribution of dietary protein to sulphide production in the large intestine: an in vitro and a controlled feeding study in humans. Am. J. Clin. Nutr., 72, 1488–1494. van der Niet, T., Hansen, D.M. & Johnson, S.D. (2011). Carrion mimicry in a South African orchid: flowers attract a narrow subset of the fly assemblage on animal carcasses. Ann. Bot., 107, 981–992. Letter Ollerton, J. & Raguso, R.A. (2006). The sweet stench of decay. New Phytol., 172, 382–385. Paczkowski, S. & Sch€utz, S. (2011). Post-mortem volatiles of vertebrate tissue. Appl. Microbiol. Biotechnol., 91, 917–935. Pagel, M. (1994). Detecting correlated evolution on phylogenies: a general method for the comparative analysis of discrete characters. Proc. R. Soc. Lond. B, 255, 37–45. Paradis, E. (2006). Analysis of Phylogenetics and Evolution with R. Springer, New York, UK. Paradis, E. & Claude, J. (2002). Analysis of comparative data using generalized estimating equations. J. Theor. Biol., 218, 175–185. Paradis, E., Claude, J. & Strimmer, K. (2004). APE: analyses of phylogenetics and evolution in R language. Bioinformatics, 20, 289–290. Pettersson, S., Ervik, F. & Knudsen, J.T. (2004). Floral scent of bat-pollinated species: West Africa vs. The New World. Biol. J. Linn. Soc., 82, 161–168. Podskalska, H., R uzicka, J., Hoskovec, M. & Salek, M. (2009). Use of infochemicals to attract carrion beetles into pitfall traps. Entomol. Exp. Appl., 132, 59–64. Raguso, R.A. (2003). Olfactory landscapes and deceptive pollination: signal, noise and convergent evolution in floral scent. In: Insect Pheromone Biochemistry and Molecular Biology (eds Blomquist, G.J. & Vogt, R.G.). Academic Press, New York, NY, pp. 631–650. Ranganathan, Y. & Borges, R.M. (2011). To transform or not to transform: that is the dilemma in the statistical analysis of plant volatiles. Plant Signal. Behav., 6, 113–116. Schiestl, F. & D€ otterl, S. (2012). The evolution of floral scent and olfactory preferences in pollinators: coevolution or pre-existing bias? Evolution, 66, 2042–2055. Schulz, S. & Dickschat, J.S. (2007). Bacterial volatiles: the smell of small organisms Nat. Prod. Rep., 24, 814–842. Schwab, W., Davidovich-Rikanati, R. & Lewinsohn, E. (2008). Biosynthesis of plant-derived flavor compounds. Plant J., 54, 712–732. Shuttleworth, A. & Johnson, S.D. (2010). The missing stink: sulphur compounds can mediate a shift between fly and wasp pollination systems. Proc. Biol. Sci., 277, 2811–2819. Statheropoulous, M., Spiliopoulou, C. & Apapiou, A. (2005). A study of volatile organic compounds evolved from the decaying human body. Forensic Sci. Int., 153, 147–155. Evolution of floral scent in plants 1167 Statheropoulous, M., Apapiou, A., Spiliopoulou, C., Pallis, G.C. & Sianos, E. (2007). Environmental aspects of VOCs evolved in early stages of human decomposition. Sci. Total Environ., 385, 221–227. Stensmyr, M.C., Urru, I., Collu, I., Celander, M., Hansson, B.S. & Angioy, A. (2002). Rotting smell of dead-horse arum florets. Nature, 420, 625–626. Teichert, H., D€ otterl, S., Frame, D., Kirejtshuk, A. & Gottsberger, G. (2012). A novel pollination mode, saprocantharophily, in Duguetia cadaverica (Annonaceae): a stinkhorn (Phallales) flower mimic. Flora, 207, 522–529. Thiό-Henestrosa, S. & Comas, M. (2011). CoDaPack v2 USER’s GUIDE. http://ima.udg.edu/~thio/#Compositional%20Data%20Package. Last accessed 25 April 2013. Urru, I., Stensmyr, M.C. & Hansson, B.S. (2011). Pollination by brood-site deception. Phytochemistry, 72, 1655–1666. Vereecken, N.J. & McNeil, J.N. (2010). Cheaters and liars: chemical mimicry at its finest. Can. J. Zool., 88, 725–752. Webb, C.O. & Donoghue, M.J. (2005). Phylomatic: tree assembly for applied phylogenetics. Mol. Ecol. Notes, 5, 181–183. Webb, C.O., Ackerly, D.D. & Kembel, S.W. (2006). Phylocom: Software for the Analysis of Community Phylogenetic Structure and Trait Evolution. Version 3.40. Available at: www.phylodiversity.net/phylocom/. Last accessed 25 April 2013. SUPPORTING INFORMATION Additional Supporting Information may be downloaded via the online version of this article at Wiley Online Library (www.ecologyletters.com). Editor, Jessica Gurevitch Manuscript received 29 April 2013 First decision made 26 May 2013 Manuscript accepted 14 June 2013 © 2013 John Wiley & Sons Ltd/CNRS