Isoprene (2-methyl-1,3-butadiene) is emitted from many plants and it appears to have an adaptive role in protecting leaves from abiotic stress. However, only some species emit isoprene. Isoprene emission has appeared and been lost many times independently during the evolution of plants. As an example, our phylogenetic analysis shows that isoprene emission is likely ancestral within the family Fabaceae (= Leguminosae), but that it has been lost at least 16 times and secondarily gained at least 10 times through independent evolutionary events. Within the division Pteridophyta (ferns), we conservatively estimate that isoprene emissions have been gained five times and lost two times through independent evolutionary events. Within the genus Quercus (oaks), isoprene emissions have been lost from one clade, but replaced by a novel type of light-dependent monoterpene emissions that uses the same metabolic pathways and substrates as isoprene emissions. This novel type of monoterpene emissions has appeared at least twice independently within Quercus, and has been lost from 9% of the individuals within a single population of Quercus suber. Gain and loss of gene function for isoprene synthase is possible through relatively few mutations. Thus, this trait appears frequently in lineages; but, once it appears, the time available for evolutionary radiation into environments that select for the trait is short relative to the time required for mutations capable of producing a non-functional isoprene synthase gene. The high frequency of gains and losses of the trait and its heterogeneous taxonomic distribution in plants may be explained by the relatively few mutations necessary to produce or lose the isoprene synthase gene combined with the assumption that isoprene emission is advantageous in a narrow range of environments and phenotypes.
While progress has been made in explaining why plants emit isoprene, there has been little progress in explaining why only some plants emit isoprene. It has been assumed that the capacity for enzyme-catalysed isoprene emission has evolved independently within distinct lineages of plants, and may have been lost from some lineages (Loreto et al. 1998; Harley et al. 1999; Sharkey et al. 2005); but whether these transitions occur frequently or rarely, and the distribution of such transitions across broad taxonomic groups, has not been discussed. In one of the few examples of independent origins that have been analysed, the isoprene synthase (ISPS) genes from aspen and kudzu share only 65% similarity in coding sequence (Sharkey et al. 2008); this result led the authors to suggest independent origins, albeit through informal inference. A cursory examination of the distribution of isoprene emission among different plant lineages also supports independent origins (Harley et al. 2004). For example, taxonomic groups as distinct as mosses and oak trees emit isoprene, but in groups as closely allied as mosses and hornworts or oak and maple trees we can find both emitters (mosses and oaks) and non-emitters (hornworts and maples) (Hanson et al. 1999; Harley et al. 1999; Lerdau & Gray 2003). In some groups with high taxonomic diversity, such as the Fabaceae (= Leguminosae), numerous isoprene-emitting genera exist and the trait is distributed among traditionally defined subfamilies. In other groups with high taxonomic diversity, such as the Poaceae, isoprene emission is relatively rare. Within the genus Quercus, it has been observed that trees of most species emit isoprene, but one lineage of trees does not emit isoprene, and in another lineage trees emit light-dependent monoterpenes (Loreto et al. 1998, 2002). Thus, there are complex patterns in the distribution of this trait; in some cases isoprene emission seems to be phylogenetically conserved, whereas in others, the trait seems randomly distributed.
The fact that isoprene emission is dispersed so heterogeneously among plant taxa must, in and of itself, have meaning as to the adaptive significance of the trait. That is, answers to the questions –‘why do plants emit isoprene’ and ‘why do only some plants emit isoprene’– are likely to emerge from common explanations, all dealing with the costs and benefits by which the trait influences fitness, and including accommodation for stochasticity. In this paper, we have conducted a more comprehensive phylogenetic analysis than has been attempted in the past with the aim of gaining a deeper perspective of the lability of isoprene emission. We assembled our emission trait data from a database on isoprene emissions that included approximately 1700 species of plants and were collated from numerous past ‘survey-type’ studies. To explore the evolutionary history of isoprene emission across plant diversity, we mapped isoprene emission on plant phylogenies that we constructed using DNA sequence data. Our data set was limited to plants that had both DNA sequence data and had been tested for isoprene emissions. Nevertheless, we found enough quantitative data to provide at least some broad insight into phylogenetic dispersion of the trait and the extent to which it has been gained or lost independently within well-defined lineages. After describing patterns of phylogenetic distribution of isoprene emission in plant lineages at several taxonomic scales, we present the hypothesis that this trait has been lost and gained frequently and stochastically due to a requirement for relatively few mutations to gain or lose a functional form of the enzyme isoprene synthase, combined with adaptive advantage in a narrow range of environments and phenotypes.
Constructing a broad taxonomic database of isoprene emission capacity
We derived data on isoprene emission capacity using several sources. Most of the data were obtained from a list of ∼1300 species that was constructed and has been maintained by researchers at Lancaster University (http://www.es.lancs.ac.uk/cnhgroup/iso-emissions.pdf), hereafter called the Lancaster List. In this list, species are not always classified with regard to rates of emission, and many are simply classified as ‘emitters’ or ‘non-emitters’. We only used data for which we had quantitative measurements of emission rate. For those species in which quantitative data were included, the units used for emission rate varied. We used species emission rates that were based on leaf area as they were reported, and converted those based on leaf mass to an area basis using an estimated specific leaf area (cm2 g−1) of 150 for angiosperms and 75 for gymnosperms (from Reich, Wright & Lusk 2007). The choice of these values, while supported as averages obtained from broad surveys, is subjective, and we acknowledge that it introduces uncertainty into the analysis. In recognition of these uncertainties, as well as those due to differences in measurement method and seasonality that exist among the multiple studies in the Lancaster List, we classified the isoprene emission rate for specific taxa according to emission ranges – what we call ‘emission bins’. We obtained data on emission rates from several additional studies that were not represented in the Lancaster List, including surveys conducted in Costa Rica by Geron et al. (2002), southern Africa by Harley et al. (2003), the Amazon Basin by Harley et al. (2004), the Biosphere 2 facility by Pegoraro et al. (2006), southern California by Benjamin et al. (1996), China by Loreto et al. (2002) and Geron et al. (2006), and South Africa by Zunckel et al. (2007). The five emission bins that we used are designated by progressively greater numbers of asterisks within the presented phylogenies: no asterisk (emissions < 0.2 nmol m−2 s−1), * (emissions ≥0.2 and <10 nmol m−2 s−1), ** (emissions ≥10 and <30 nmol m−2 s−1), *** (emissions ≥30 and <50 nmol m−2 s−1), **** (emissions ≥50 nmol m−2 s−1). We only used values that were referenced to a photosynthetic photon flux density (PPFD) of 1000 µmol m−2 s−1 and leaf or air temperature of 30 °C.
Construction of phylogenies
We acquired DNA sequence data from GenBank (http://www.ncbi.nlm.nih.gov/GenBank/) for four groups of plants: the legumes (Fabaceae), the oaks (Quercus), the Fagales (an order of plants that includes the beech, birch, oak and walnut families), and the ferns (Pteridophyta). Multiple genes were used for each plant group: rbcL, trnL and matK for the legumes (Supporting Information Table S1); 5.8S, CLAW, matK, rbcL, the intergenic space (IGS) between psbA and nuclear internal transcribed spacer (ITS), and the IGS between rbcL and atpB for the oaks (Supporting Information Table S2); ITS, trnL/trnF and matK for the Fagales (Supporting Information Table S3), and rbcL, atpA, matK, rps4, and IGS between trnL and trnF for the ferns (Supporting Information Table S4). Sequence data were aligned using the program, Muscle (http://www.drive5.com/muscle/), and edited by eye using Seaview v4.3.3 (http://pbil.univ-lyon1.fr/software/seaview.html). We used MrBayes v3.2 (http://mrbayes.sourceforge.net/) to generate phylogenies for each plant group. We used a general time reversible substitution model with gamma-distributed rate variation and a proportion of invariable sites. Data sets were partitioned such that model parameters for each gene were estimated separately. We ran two independent runs (each started with a randomly chosen tree) with one cold chain and three heated chains per run. Each of the four analyses was initially set to run for 2 000 000 generations. We used a 25% burn-in (i.e. the first 25% of the cold chains were discarded) and used the average standard deviation of split frequencies between the two runs to determine if the phylogenies between the runs had converged. If the average deviation of split frequencies was not below 0.01 after 2 000 000 generations, additional 1 000 000 generations were added to the run. For Figs 2–5, isoprene presence or absence was depicted on the consensus tree and node values indicate the posterior distribution of the consensus topology. The number of state changes (i.e. gains and losses of isoprene emission) was determined using parsimony. If there were equally parsimonious reconstructions, we chose losses of emission over gains of emission.
Emissions among all orders within division Angiospermae
We have mapped the capacity for isoprene emission observed in past studies onto the complete tree of angiosperm orders developed by Soltis et al. (2011) using 17 genes and 640 taxa (Fig. 1). The distribution of isoprene emissions at this scale is not particularly informative as to independent state changes. However, it does reveal phylogenetic domains that have been understudied. For example, in the Malvideae high emission rates have been observed in the Sapindales and Myrtales, but nothing is known of emission potentials in the two orders that separate the Sapindales and Myrtales; that is, the Geraniales or Crossosomatales. Studies within Crossosomatales would be particularly interesting as this order includes woody, temperate-latitude trees, which occupy niches similar to those in Sapindales. The Geraniales, in contrast, are largely composed of herbaceous species, but they are closely related to those woody species in the Myrtales that have exceptionally high isoprene emission rates. Thus, in the first comparison (Crossosomatales versus Sapindales) ecological and growth habit tendencies are similar, but in the second comparison (Myrtales versus Geraniales) they are not; yet, phylogenetic relatedness among all four is high.
In other groups, we have highlighted a need for more studies because they occupy basal positions within the angiosperm tree. For example, the Austrobaileyales is an order with good representation of tropical woody trees (which often emit isoprene) and is a critical link between monospecific Amborellaceae, the acknowledged most basal family of angiosperms, and other groups with high frequencies of tropical woody trees, such as Piperales, Laurales and Magnoliales. Insight into the origins and diversification of isoprene emissions in the most basal angiosperms is limited by a lack of sampling in the Amborellaceae and the Austrobaileyales.
Isoprene emissions have been observed within the Acrogymnospermae (= Gymnospermae), particularly within Pinaceae. The potential for isoprene emissions in even more members of the Acrogymnospermae, however, has not been clarified. For example, within Ephedraceae, isoprene emissions are known to be present, and at significantly high rates (Geron et al. 2006), but in two other gymnosperm families, Gnetaceae and Welwitschiaceae, emission potentials have not been validated. There is unpublished evidence that these groups emit isoprene (Rasmussen, personal communication 1991), but quantitative validation is not available. Further studies into isoprene emission in these groups, which share phylogenetic affinities with emitting members of the Pinaceae and Ephedraceae, would sharpen our understanding of diversification within the angiosperms.
Emissions within family Fabaceae
We explored the phylogenetic distribution of isoprene emission in Fabaceae (= Leguminosae), which includes a large number of isoprene-emitting tropical tree species (Fig. 2). We chose Populus as the outgroup for our analysis of Fabaceae because we wanted to use a group that was basal yet close in affinity to the order Fabales, and included adequate DNA sequence data for analysis. Based on the Soltis tree, the closest basal orders would be Oxalidales, Malpighiales and Celastrales. We chose Populus (in Malpighiales) because it provided us with the largest source of sequence data. Assuming that isoprene emission was carried into the Fabaceae from ancestors, we estimate at least 16 subsequent losses and 10 secondary gains (i.e. gains that followed losses).1 One of the more active nodes of evolution with regard to isoprene emissions within Fabaceae occurs in genera that were formerly grouped together within the genus Acacia (Table 1). Using traditional taxonomy, Acacia is the second largest genus within Fabaceae, with over 1400 species (Lewis et al. 2005). More recently, however, traditional Acacia has been partitioned into five new and distinct clades: Acacia sensu stricto (Acacia s.s.), Vachellia, Senegalia, Acaciella and Mariosousa (Brown et al. 2008; Bouchenak-Khelladi et al. 2010). The state of isoprene emissions in Mariosousa, Acaciella and Acacia s.s. is uncertain. Vachellia, however, appears to lack isoprene emitters, a state that was likely ancestral. Senegalia appears to have obtained isoprene emissions secondarily, after divergence from Vachellia, and independently from the emergence of isoprene emissions in allied species in Albizia and the Inga-Zygia clade.
Table 1. Species of Acacia that have been assayed for isoprene emission and are organized according to three of the five clades recognized from chloroplast DNA analysis (following Bouchenak-Khelladi et al. 2010)
Data on the presence or absence of isoprene emission were taken from Harley et al. (2003) or the Lancaster Table at http://www.es.lancs.ac.uk/cnhgroup/iso-emissions.pdf. The question marks following some taxa in the Acacia s.s. indicate questionable affinity (see text). The isoprene emission capacity is indicated according to a quantitative scale indicated by the number of asterisks (see text).
Acacia baileyana ?
Acacia dealbata ?
Acacia decurrens ?
Biogeographic patterns suggest that Vachellia evolved in open woodlands in America during the Oligocene, approximately 16 million years ago (mya), and then subsequently migrated to Africa (Bouchenak-Khelladi et al. 2010). Once Vachellia expanded within Africa, diversification produced Senegalia (Bouchenak-Khelladi et al. 2010). The evolution of isoprene emission in Senegalia likely occurred during the relatively arid mid-to-late Miocene and subsequent Pliocene, beginning approximately 10 mya. This is the same period in which C4 savannas expanded within Africa, America and Australia, which fostered open, semi-arid habitats (Beerling & Osborne 2006). The emergence of isoprene emission may have contributed to persistence of Senegalia in the semi-arid regions of Africa, perhaps through enhanced tolerance of drought and associated abiotic stresses.
Emissions within order Fagales
We chose to focus part of our phylogenetic reconstruction on the order Fagales (Fig. 3) because it contains the oaks, which are known to exhibit isoprene emission rates among the highest recorded. Gene sequence data analysed in two independent studies have confirmed that Nothofagus, a non-isoprene-emitting genus, which had previously been included within Fagaceae, is in fact monophyletic and basal to the ‘higher’ hamamelid families including Fagaceae, Casuarinaceae, Betulaceae, Myricaceae and Juglandaceae (Manos & Steele 1997; Li et al. 2004). The basal position of Nothofagus was supported in our phylogeny. Our analysis placed Quercus and Lithocarpus within the same clade, with the potential for isoprene emission arising in both through one event. Independent origins appear to have occurred at the diversification that gave rise to Casuarina, and possibly Myrica if subsequent studies confirm emissions in the latter group. Thus, the trait appears to have been lost early in, or prior to, diversification of the order and then gained secondarily at least twice.
Emissions within genus Quercus
It is informative to take a closer look at diversification within the genus Quercus (Fig. 4), not only because some of its species are those with the highest recorded isoprene emission rates, but also because some of its species exhibit light-dependent monoterpene emissions, rather than isoprene emissions, and still others do not emit either type of compound. The phylogenetic distribution of isoprene emissions within Quercus has been analysed previously in Loreto (2002). In that analysis, 61 species were placed into phylogenetic context through a combination of traditional morphological taxonomies and gene sequence analysis. However, since that study our knowledge of phylogenetic affinities and terpene emission capacities within Quercus has increased and there is need for renewed assessment.
Consistent with the study of Loreto (2002), we conclude that isoprene emission is ancestral within Quercus, having been carried into the group during diversification from Lithocarpus. Early radiation within the genus led to two well-supported lineages – the Protobalanus-Lobatae-Quercus clade and the Cerris-Cyclobalanopsis-Ilex clade (Denk & Grimm 2010). In Fig. 4, the node at which these two clades diverged is clearly marked by the presence of low-to-high isoprene emission rates (black branches) in the Protobalanus-Lobatae-Quercus clade and lack of isoprene emissions (grey branches) in the Cerris-Cyclobalanopsis-Ilex clade. The trait appears to have been preserved in all species and amplified in some species during radiation of the Protobalanus-Lobatae-Quercus clade, but lost and not restored, at least within the Cerris-Ilex portions of the Cerris-Cyclobalanopsis-Ilex clade. (We did not include Cyclobalanopsis in the analysis shown in Fig. 4 although its position as sister group to the Cerroid and Ilicoid lines, as suggested by Denk & Grimm (2010), would not change the conclusion that loss of isoprene emissions must have occurred early during radiation of the Cerris and Ilex groups.)
The loss of isoprene emissions in the Cerris-Cyclobalanopsis-Ilex clade was accompanied by the gain of light-dependent monoterpene emissions in the Ilicoid genera and at least one of the Cerroid genera (Loreto 2002). There are two possible paths by which these gains could have occurred. In the first, the gene for isoprene synthase, which is targeted through a leader peptide sequence to the chloroplast, could have undergone (1) neo-functionalization to produce a monoterpene synthase gene, also targeted to the chloroplast, in the common ancestor to the Ilicoid and Cerroid lines; (2) subsequent amplification in expression of this novel gene as the Ilicoid oaks radiated into the warmer and drier environment of the Mediterranean region; and (3) subsequent non-functionalization of this novel gene through genetic drift to produce a lineage with neither isoprene nor light-dependent monoterpene emissions as most of the Cerroid oaks (e.g. Quercus libani and Quercus cerris) radiated into the humid temperate and subtropical regions of East Asia and Western Eurasia. This possible path is thus characterized by serial events going from neo-functionalization in the ancestor to all Ilicoid and Cerroid species, followed by non-functionalization in most of the Cerroid species. In the one species from the Cerroid oaks that radiated into the Mediterranean region (Quercus suber), light-dependent monoterpene emissions have apparently emerged secondarily (Staudt et al. 2004; Pio et al. 2005). In the second possible path, the gene for isoprene synthase underwent non-functionalization in the ancestor to the Ilicoid and Cerroid oaks, to form a clade of non-emitting oaks, following divergence from Cyclobalanopsis. This would have been followed by neo-functionalization of a duplicated chloroplastic monoterpene synthase gene to produce a novel type of light-dependent monoterpene emissions in the Ilicoid line as it radiated into Mediterranean regions (Loreto 2002). This possible path is thus characterized by serial events going from non-functionalization in the ancestor to all Ilicoid and Cerroid genera, followed by neo-functionalization in the ancestor to the Ilicoid genera. Similar to the first alternative path, Q. suber, allied with the Cerroid oaks, presumably gained light-dependent monoterpene emissions secondarily as it radiated into Mediterranean regions. Both alternative paths involve three independent events; they simply differ as to the sequence of events and whether the source for light-dependent monoterpene emissions in the Ilicoid oaks was a modified isoprene synthase gene (first path) or a duplicated monoterpene synthase gene (second path).
The high degree of lability in isoprene emissions as a trait that is lost and gained frequently is strikingly demonstrated within Quercus in the results reported by Staudt et al. (2004). In a Q. suber population from northern Spain, 9% of the trees that were sampled did not emit light-dependent monoterpenes or isoprene, whereas 91% of the trees emitted light-dependent monoterpenes. The reason for loss of emissions in some of the trees is not known; although it is more likely to be due to non-functionalization of the light-dependent monoterpene synthase gene than an inability to generate adequate dimethylallyl diphosphate (DMADP) substrate, as mutations in the MEP pathway leading to DMADP tend to interfere with metabolic processes that are crucial to plant survival (Estevez et al. 2001; Floss et al. 2008; Rodriguez-Concepción 2010). The particular population that was studied is one that has experienced genetic introgression from a related sympatric species, Quercus ilex; although Q. ilex is also a light-dependent monoterpene emitter, genetic introgression cannot explain the lack of monoterpene emissions in some trees of Q. suber. However, the fact that trees within a single population have persisted for decades without light-dependent monoterpene emissions indicates that the trait can be easily lost from a population, and that its presence has not been required as an adaptation promoting tree survival and growth, at least within recent decades.
The distribution of isoprene emission within the ferns (Pteridophyta)
Our model of fern phylogeny reveals that lack of the trait was the likely ancestral condition with the potential for at least five independent gains and two subsequent losses (Fig. 5). These events include a gain and amplification of expression in Adiantum and Coniogramme, with subsequent loss in Pteris. A gain and amplification appear to have occurred within Tectaria, with subsequent loss in the Pyrossia-Drynaria-Platycerium lineage. An independent gain appears to have occurred in the Thelypteris-Cyclosorus lineage, with amplification in Thelypteris. At present, we have no hypothesis to explain why the trait appeared and became amplified to high levels of expression in some branches of a lineage, while becoming non-functional in other branches. There appears to be no clear distinction among genera with these divergent patterns in terms of ecological niche assortment or phenotype.
Random mutation and the origins of isoprene synthase: why isoprene emission has so many independent origins
Plant terpene synthases (TPSs) represent a highly diverse set of enzymes controlled by an equally diverse set of genes (Chen et al. 2011). Many of the TPSs catalyse reactions leading to more than one type of biochemical product (Bohlmann, Meyer-Gauen & Croteau 1998), and many have more than one active site, often catalysing different reactions. The TPS gene family is subdivided into seven subfamilies, TPS-a to TPS-h (Chen et al. 2011), with isoprene synthase protein (ISPS) showing affinities to the TPS-b subfamily. [Two of the original subfamilies, TPS-e and TPS-f, were reorganized into a single subfamily in the classification of Chen et al. (2011)]. The TPS-b subfamily also includes monoterpene synthases from angiosperms (Miller, Oschinski & Zimmer 2001). The active site of ISPS contains an α-helical class I terpenoid fold (Köksal et al. 2010), which is characteristic of eudicot monoterpene and sesquiterpene synthases (Christianson 2006; Cao et al. 2010). The active site fold of poplar ISPS (Köksal et al. 2010) is shallower than those described for monoterpene synthases in the TPS-b subfamily (Hyatt et al. 2007), consistent with the binding of the C5 substrate DMADP rather than the C10 substrate geranyl diphosphate (GDP). Using protein modelling, Gray et al. (2011) showed that the volume of the active site fold of ISPS may have been reduced during evolution by replacement of smaller amino acids, in one or a few positions, with larger amino acids (e.g. phenylalanines). If true, this model would suggest that relatively few mutations are required to produce the gene for ISPS from the duplicated genes of existing TPS (Gray et al. 2011). Isoprene emission from gymnosperms, such as species in Picea, is likely due to ISPS protein derived from an ancestor within the TPS-d subfamily and may share more similarities with gymnosperm methylbutenol synthases and linalool synthases than with angiosperm ISPS (Gray et al. 2011).
The genes for ISPS from both kudzu and poplar have been transferred into Arabidopsis, and isoprene emission from transformed Arabidopsis was observed (Sharkey et al. 2005; Loivamäki et al. 2007). However, the rates of isoprene emission were approximately 1/20th the rates observed in wild-type species, and DMADP substrate concentration in Arabidopsis leaves was 6–10 times lower than in wild-type emitters (Loivamäki et al. 2007). Thus, even in those cases where a novel ISPS enzyme has appeared (in this case by ‘engineering’), isoprene emission rates are often low, most likely affected by limited DMADP substrate availability. Evidence obtained from kudzu indicates that while genetic modification to the expression of MEP pathway genes in isoprene emitters may not be required to sustain isoprene emissions, once ISPS appears, some form of adjustment (genetic or otherwise) must occur in order to increase the supply of DMADP substrate (Sharkey et al. 2005), and this does not happen solely as a consequence of the appearance of ISPS.
More recently, ISPS was transferred into non-emitting tobacco plants, and rates of isoprene emission were observed to be higher than in the previous Arabidopsis experiments; in fact, rates of isoprene emission from plants in one of the four transgenic tobacco lines were similar to those observed for Populus alba leaves (a native isoprene emitter) (Vickers et al. 2009a, 2011). The capacity for high rates of emission in this one transgenic tobacco line was supported by high rates of DMADP substrate generation (Vickers et al. 2011). These studies indicate that pre-existing phenotype matters to the potential for isoprene emissions in a lineage once a novel ISPS gene appears. Amplification of expression in the novel ISPS gene may be controlled by the pre-existing potential to generate DMADP substrate at the time of trait appearance, and this in turn may be dependent on pre-existing potential to synthesize chloroplast metabolites other than isoprene, which depend on DMADP substrate (e.g. other terpenoids, carotenoids and abscisic acid).
Inferring from these past studies, it is reasonable to conclude that the evolution of isoprene emission in plants requires either (1) mutations that affect both the appearance of ISPS and up-regulation of the capacity for leaves to produce DMADP substrate; or (2) a pre-existing phenotype that enables amplification of expression of ISPS once it appears. As stated in the study by Sharkey et al. (2005): ‘The trait of isoprene emission appears to be easily acquired by plants, although the regulation of DMADP concentration may have to be altered to support the high rates of isoprene emission seen in some plants.’ We agree.
Non-random selection for isoprene emissions: why isoprene emissions only persist in some plant lineages
The evidence drawn from phylogenetic distributions of isoprene emission suggests that it has disappeared frequently from lineages during past diversification. There are two underlying evolutionary forces that might explain loss of the trait. Firstly, isoprene emission may incur a cost to growth and associated fitness in certain phenotypes and/or environments. In the presence of a cost to fitness, and without a balanced adaptive benefit, any random mutations that cause non-functionalization of the ISPS gene would provide a selective advantage, favouring loss of the trait. Secondly, if the trait is neutral in its effect on fitness, then the accumulation of mutations leading to ISPS non-functionalization may occur through genetic drift.
We propose a model whereby the mutations giving rise to ISPS occur relatively frequently, but the range of environments or phenotypes within which isoprene emissions are adaptive or within which its expression is capable of being amplified, is relatively narrow. Thus, the determinant as to whether the trait exists in a lineage reflects the balance between a relatively high frequency of origin and a relatively low frequency of retention. Furthermore, the availability of niches in which isoprene emissions are adaptive is likely to change over time, either due to a changing environment (e.g. change in atmospheric CO2 concentration or climate), changes in potential plant migration patterns (e.g. due to continental drift or uplift of topography), or evolutionary changes in plant phenotype (e.g. greater or lesser capacity to produce DMADP substrate).
What are the environmental conditions or functional phenotypes that could determine the persistence or loss of isoprene emissions? There is clear evidence that isoprene emission fosters tolerance of abiotic stresses. The exact nature of the tolerance, and the relative importance of different stresses, continues to be debated (Loreto & Schnitzler 2010). In one study, tolerance of extremely high leaf temperatures and photon flux densities by isoprene-emitting poplar leaves was shown to be greatest when trees were grown at a low (190 ppmv) atmospheric CO2 concentration (Way et al. 2011). This observation was used to suggest that the adaptive advantage provided by isoprene emission was enhanced during past epochs with relatively low atmospheric CO2 concentrations (e.g. post Miocene) (Way et al. 2011). When isoprene-emitting species are grown in an atmosphere of elevated CO2, emission rates are suppressed, most likely due to limited availability of chloroplastic pyruvate, one of the initial substrates required for DMADP biosynthesis (Rosenstiel et al. 2003; Monson et al. 2007; Wilkinson et al. 2009; Possell & Hewitt 2011; Trowbridge et al. 2012). When grown at elevated atmospheric CO2 concentrations, the photosynthetic processes of C3 plant leaves are typically more tolerant of abiotic stresses (Morgan, Ainsworth & Long 2003; Leakey et al. 2009), especially high-temperature stress (Wang et al. 2012), meaning that isoprene emissions may have less of an adaptive margin during epochs with higher atmospheric CO2 concentration. There are also certain types of pre-existing plant phenotypes that may amplify or mute selection for isoprene emission once it appears in a lineage. For example, it has been suggested that plants with certain phloem-loading phenotypes, which favour high leaf sugar concentrations in the cytosol of leaf mesophyll cells, might be better able to support the high rates of DMADP synthesis required to amplify the expression of isoprene emissions once it appears (Logan, Monson & Potosnak 2000; Kerstiens & Possell 2001). This latter point may be relevant to the puzzle as to why isoprene emission occurs so commonly in lineages of trees, but so uncommonly in lineages of herbs – these growth habits tend to differ in their phloem-loading mechanisms. Thus, a complex set of interactions exists among changes in the earth's climate and atmospheric CO2 concentration, pre-existing and novel plant phenotypes, and geographic congruence between trajectories of lineage radiation and the location of those abiotic stresses most likely to favour, or not, the persistence of isoprene emissions, once the trait appears.
Given the complexity of these interactions, the most intriguing question to be answered would appear to be not ‘why isoprene emissions has appeared so many times independently within plant lineages’, but rather ‘why has it been lost so many times independently’. Is it because random patterns of evolutionary radiation have caused some clades to assort into niches where the trait loses its adaptive premium and genetic drift leads to eventual loss? If so, then we would expect to discover patterns of niche assortment that correlate with the presence or absence of the trait within lineages. This is not always the case. In groups such as the Fabaceae, where so many loss events appear to have occurred, many species occupy similar tropical or subtropical forest niches, whether they are emitters or not. Alternatively, is it because the trait has a positive effect on fitness in a relatively narrow range of niches and phenotypes; and in the absence of the trait radiating into those niches or arising in those phenotypes, genetic drift (or weak negative selection) causes the trait to disappear as frequently as it appears? If few mutations are required to create the trait, and even fewer mutations are required to lose the trait, there is a narrow window of opportunity for the trait to ‘find’ a situation that favours adaptive persistence, before genetic drift causes it to be lost.
Most investigations to date on the topic of isoprene emission from plants have focused on its potential adaptive benefit to the photosynthetic processes of the leaf; in other words, ‘why plants emit isoprene’. However, one question has remained in the background and has not been explicitly addressed during decades of investigation: ‘why do only some plants emit isoprene?’ In order to address this question, investigators will need to move beyond descriptions of plant function in emitting species, and consider the differences that exist between emitting and non-emitting species in phenotype, ecological niche and phylogenetic history. We have shown that there is substantial evidence suggesting that isoprene emission has emerged, and been lost, many times independently within plant lineages. An explanation of these patterns requires consideration of several inferences: (1) few mutations are required for the trait to evolve; (2) the adaptive benefit of the trait varies depending on several factors, including climate, atmospheric CO2 regimes, phenotype and patterns of geographic migration; and (3) adaptive benefits are likely to occur in a relatively narrow range of combinations of these factors. Given these conditions, it is possible to explain patterns of frequent trait emergence and loss in the phylogenetic record. We hope that by bringing these perspectives into the arena of debate we can stimulate future research that will further clarify the ‘why’ questions underlying the distribution of isoprene emissions with regard to taxonomy, plant growth habit, ecological niche and geography.
We are grateful to the helpful comments provided by Dr Peter Harley, Dr Hardeep Rai and an anonymous reviewer. The authors have no conflicts of interest to declare.
NOTE ADDED IN PROOF
In a recent study (Welter et al., 2012, Tree Physiology, doi:10.1093/treephys/tps069), populations of the North African oak species, Quercus afares, which have hybrid origins from ancestral isoprene-emitting and light-dependent, monoterpene-emitting species, have been shown to have secondarily evolved complete suppression of isoprene emission and modification of monoterpene emission capacity as it radiated into cooler, less arid ecosystems. These results are consistent with our hypothesis of high evolutionary lability and frequent shifts that lead to gains or losses of isoprene and light-dependent monoterpene emissions.
The numbers of gains and losses that we report are dependent not only on our inferred phylogenies, but also on the incomplete sampling within lineages for the presence or absence of isoprene emissions. However, given acceptance of the phylogenetic models, the estimated number of state transitions should be conservative. If, for example, in the future, isoprene emissions are discovered in a genus of otherwise non-emitting species, and we have designated that genus as having undergone a transition from emitting to non-emitting, then the new observation can only lead to a greater or equal number of estimated state transitions, not a lesser number.