In Planta Recapitulation of Isoprene Synthase Evolution from Ocimene Synthases.

Isoprene is the most abundant biogenic volatile hydrocarbon compound naturally emitted by plants and plays a major role in atmospheric chemistry. It has been proposed that isoprene synthases (IspS) may readily evolve from other terpene synthases, but this hypothesis has not been experimentally investigated. We isolated and functionally validated in Arabidopsis the first isoprene synthase gene, AdoIspS, from a monocotyledonous species (Arundo donax L., Poaceae). Phylogenetic reconstruction indicates that AdoIspS and dicots isoprene synthases most likely originated by parallel evolution from TPS-b monoterpene synthases. Site-directed mutagenesis demonstrated invivo the functional and evolutionary relevance of the residues considered diagnostic for IspS function. One of these positions was identified by saturating mutagenesis as a major determinant of substrate specificity in AdoIspS able to cause invivo a dramatic change in total volatile emission from hemi- to monoterpenes and supporting evolution of isoprene synthases from ocimene synthases. The mechanism responsible for IspS neofunctionalization by active site size modulation by a single amino acid mutation demonstrated in this study might be general, as the very same amino acidic position is implicated in the parallel evolution of different short-chain terpene synthases from both angiosperms and gymnosperms. Based on these results, we present a model reconciling in a unified conceptual framework the apparently contrasting patterns previously observed for isoprene synthase evolution in plants. These results indicate that parallel evolution may be driven by relatively simple biophysical constraints, and illustrate the intimate molecular evolutionary links between the structural and functional bases of traits with global relevance.

Introduction

Isoprene (2-methyl-1,3-butadiene, C5H8) is the most abundant naturally emitted biogenic volatile compound, first discovered in plants 60 years ago (Sanadze 1957). Many plants, especially perennial, fast-growing forest species, constitutively emit large amounts of isoprene into the atmosphere (Dani etal. 2014; Loreto and Fineschi 2015), corresponding to a yearly global emission of ∼500 Tg C (Guenther etal. 2012). Moreover, isoprene rapid oxidation, initiated by hydroxyl radicals (OH) is estimated to contribute almost half of secondary organic aerosol (SOA) (Worton etal. 2013), which plays an essential role in regulating atmospheric chemistry (Heald etal. 2009; Archibald etal. 2010; St. Clair etal. 2016).

In plants, isoprene biosynthesis is catalyzed in chloroplasts by isoprene synthase (IspS) from dimethylallyl diphosphate anion (DMADP; supplementary fig. S1, Supplementary Material online) which is formed by the 2-C-methyl-d-erythritol 4-phosphate (MEP) pathway (Schwender etal. 1997). Isoprene emission rates depend on the activity of isoprene synthase and the pool size of DMADP (Wiberley etal. 2009; Rasulov etal. 2010), which are in turn influenced by many factors, like the endogenous developmental stage of a leaf (Ahrar etal. 2015; Niinemets etal. 2015) and several environmental stimuli and constraints (Loreto and Schnitzler 2010; Sharkey and Monson 2017).

Many experiments have demonstrated that isoprene plays fundamental roles in protecting plants from environmental stresses (Vickers etal. 2009; Velikova etal. 2011; Tattini etal. 2014). There may be many mechanisms acting concurrently, whose examination is beyond the scope of this report (reviewed in Loreto and Schnitzler 2010; Sharkey and Monson 2017). Recent studies suggest that isoprene may play a general role as signaling molecule to prime and regulate stress-induced genes that activate defenses against abiotic stresses (Harvey and Sharkey 2016).

Isoprene emission could be detected across the whole plant kingdom, but in the angiosperm lineage, isoprene emitting species are scattered across the phylogenetic tree (Monson etal. 2013; Dani etal. 2014). Even though there are ∼20% of plant species emitting isoprene from diverse taxonomic groups (Loreto and Fineschi 2015), only a relatively limited number of isoprene synthase genes has been so far isolated and characterized (Sharkey etal. 2013). Several views explaining such sparse phylogenetic occurrence of isoprene have been proposed. The debate focused initially mainly on whether IspS appeared only once early in the evolution of land plants, or multiple times independently (Hanson etal. 1999; Harley etal. 1999; Lerdau and Gray 2003; Sharkey etal. 2005, 2008). With the elucidation of the general patterns of evolution of terpene synthases (Chen etal. 2011), it became clear that the isoprene emission trait must have evolved independently in the major lineages of land plants by parallel evolution (Sharkey etal. 2013). It is, however, still debated whether this model applies also to lower taxonomic levels, and in particular to the rosids, the clade from which the majority of IspS was obtained in recent years. One view proposes that a single gain of isoprene synthase function took place early during the radiation of rosids, followed by multiple losses (Sharkey 2013; Sharkey etal. 2013). Adaptive tradeoffs between metabolic costs and selective advantages due to paleoclimatic fluctuations of atmospheric CO2 concetration have been proposed to account for the repeated losses of the trait (Sharkey and Yeh 2001; Sharkey etal. 2013). A single isoprene synthase gain in rosids is also supported by the finding that IspS genes form a monophyletic group in this plant order (Sharkey etal. 2013). The other scenario proposes that the actual number of independent gains of isoprene emission trait is much higher, suggesting that IspS genes can undergo repeated gain and loss at the family and even genus level thanks to the hypothetically low number of amino acidic mutations which may be sufficient for evolution (Monson etal. 2013). Parsimony-based analyses based on the mapping of isoprene presence/absence on species phylogenies obtained with molecular markers independent from IspS genes support this conclusion (Monson etal. 2013). Under this view, the retention of the trait is postulated to require adaptive advantage(s) in a narrow range of environments and phenotypes. Other authors support this multiple gains and multiple losses view, further linking it to species richness at the genus level and the perennial growth habit of the species supporting the trait (Dani etal. 2014). As of today, it is not clear to what degree each of these scenarios explains the overall evolutionary trajectory followed by the IspS gene family in angiosperms, as the actual number of mutations required for IspS to evolve from other terpene synthases and the precise identity of such ancestral enzymes are unknown (Monson etal. 2013; Sharkey 2013; Sharkey etal. 2013; Sharkey and Monson 2017).

The crystal structure of P. x canescens IspS elucidated the overall similarities of isoprene synthases mechanism of catalysis. Like all other known Class I terpene synthases, IspS enzymes require for catalysis the coordination of a triad of magnesium ions, which in turn bind the pyrophosphate anion moiety of the DMADP substrate (Köksal etal. 2010). The amino acids coordinating the metal ions form two separate and highly conserved clusters on IspS helices D and H, called the DDXXD and DTE/NSE motifs, respectively. The DMADP substrate diphosphate anion moiety is bound by the magnesium triad on top of the substrate binding pocket, while its apolar moiety is positioned in the hydrophobic cleft which extends in the lower part of the central cavity of each IspS monomer (Köksal etal. 2010). Two conserved phenylalanine residues (F338 and F485) make Van der Waals contacts with the substrate at the bottom of the hydrophobic cleft. Like in the other Class I TPS, the catalytic mechanism involves the formation of geranyl and/or linalyl carbocations, followed by elimination of the substrate diphosphate moiety, with the diphosphate-leaving group likely serving as a general base (Köksal etal. 2010). By contrast to the majority of class I terpene synthases characterized to date, no obvious conformational change associated to the transition from the enzyme open to close (substrate-bound) state could be inferred for P. x canescens IspS, casting doubts on whether IspS may lack such mechanism of active site “capping” (Köksal etal. 2010).

Based on Populus x canescens IspS crystal structure and alignment of isoprene synthase proteins from rosids, a core of four conserved amino acids distinguishing isoprene synthase from phylogenetically closely related terpene synthases in angiosperms was identified (Sharkey etal. 2013). Two of the residues (F338 and F485) were suggested to constrain the size of the binding pocket, but the relevance of this hypothesis for IspS function was not experimentally tested (Gray etal. 2011; Sharkey etal. 2013). Since then, these amino acids have been considered as essential markers while screening for enzymes having exclusive isoprene synthase activity from sequence identity. The other two amino acids of the IspS diagnostic tetrad (S446 and N505) (Sharkey etal. 2013) are preferentially found in isoprene synthases, but recent identification of novel IspS genes with amino acid variants indicate that these positions display some degree of variation (Ilmén etal. 2015; Oku etal. 2015). All the validated IspS genes so far have been identified in rosids, with the only exception of one gene isolated from asterids (Ilmén etal. 2015). More importantly, to date no report on the identification of isoprene synthase genes from monocotyledons exists.

The Arundinoideae is a small subfamily of the Poaceae. Out of the six taxa from four genera investigated, the majority were found to emit isoprene (Hewitt etal. 1990; Kesselmeier and Staudt 1999; Ahrar etal. 2015). The most highly emitting species among them is Arundo donax (giant reed), a perennial fast-growing weed, which can grow up to 6–8 m in height in mediterranean climate regions all over the world. Arundo donax is used as an important biomass and biofuel crop as it may produce >35 tons ha−1 year−1 of dry biomass with low agronomic input, and a yield in bio-ethanol that may reach 50% of the dry matter (Angelini etal. 2009).

In this study, we identified and functionally validated an isoprene synthase gene from A. donax, which, to the best of our knowledge, is the first isoprene synthase characterized from monocot species. Phylogenetic reconstruction indicates that monocots and dicots isoprene synthases are most likely the result of parallel evolution. Evolutionarily informed site-directed mutagenesis and characterization of total volatile emission in Arabidopsis thaliana as heterologous host support the pivotal role of a single amino acid of the diagnostic tetrad in IspS evolution and suggest the likely ancestors of IspS enzymes. Based on these results we propose a model reconciling in an unified conceptual framework the apparently contrasting views previously proposed for isoprene synthase evolution in rosids. The mechanism of neofunctionalization of active site size reduction demonstrated in this study for IspS seems to be general, as the very same amino acidic position is implicated in the parallel evolution of different short-chain terpene synthases from both angiosperms and gymnosperms.

Results

Identification of a Putative IspS from A. donax

Several terpene synthase genes were identified in the transcriptome of A. donax (Sablok etal. 2014) based on sequence homology to poplar IspS, and the best hit (AdoIspS) was selected for further confirmation. The identified AdoIspS genomic locus (GeneBank accession No. KX906604) codes for an open reading frame of 581 amino acids with a predicted molecular mass of 69 kD. The first 16 amino acids were identified by the TargetP program as the putative transit peptide for chloroplast import (Emanuelsson and Nielsen 2000). As already reported for other IspS genes, attempts to express a functional protein in E. coli were not successful despite removal of the transit peptide (Sasaki etal. 2005; Ilmén etal. 2015; Oku etal. 2015). An overexpression construct containing the full-length cDNA of putative AdoIspS gene under the control of CaMV 35S promoter (p35S:AdoIspS-WT) was, therefore, transformed into Col-0 wild-type Arabidopsis plants. Isoprene emission was detected in the majority of a set of 120 independent T1 transgenic lines overexpressing AdoIspS compared with Col-0 wild-type plants (supplementary fig. S2, Supplementary Material online). Expression of the AdoIspS transgene was confirmed by semi-quantitative RT-PCR in 12 lines emitting isoprene (supplementary fig. S3, Supplementary Material online), indicating that isoprene emission was due to stable expression of the transgene into the Arabidopsis genome.

To conclusively demonstrate that isoprene emission was due to the AdoIspS transgene, PCR amplification with primers specific to the AdoIspS transgene and isoprene emission measurement were conducted in parallel in the T2 segregating population of the highly emitting single copy line AdoIspS-44 (supplementary fig. S4, Supplementary Material online). Isoprene emission above background level was detected only in individuals positive for the AdoIspS transgene (n = 78) and not in the negative ones (n = 30). The null hypothesis of random association between isoprene emission and transgene presence in the analyzed sample was rejected at P < 0.001 (two tailed Fisher’s exact test; P = 2.27E-27). These results demonstrate that the AdoIspS gene is the first isoprene synthase gene identified so far from a monocot species, and that the predicted 16-aa transit peptide is sufficient for chloroplast targeting of the mature protein (Emanuelsson and Nielsen 2000).

The AdoIspS gene (3,764 bp in length; see fig. 1) comprises seven exons and six introns alike all monoterpene synthases in angiosperms (Trapp and Croteau 2001). The major differences compared with kudzu and aspen IspS genes are a shorter first exon (consequence of the shorter transit peptide) and a greater length of the second intron. The multiple sequence alignment of the C-terminal domain (Class I terpenoid synthase fold) has an overall average of 144 amino acids differences out of 330 aligned positions, while the number of differences of AdoIspS from the other validated IspS ranges from 177 to 195, indicating that this is the most divergent protein in the alignment (fig. 1). Modeling of the tertiary structure based on the crystal structure of P. x canescens IspS (Köksal etal. 2010) (PcIspS, 39.3% identity with AdoIspS) confirmed a very similar folding of the two proteins (Cα RMSD = 0.19 Å comparing 522 residues), with typical alpha-helix rich structure (supplementary figs. S5 and S6, Supplementary Material online). In addition, the distribution of electrostatic potential on the surface and active site of the two proteins are largely conserved (supplementary fig. S6, Supplementary Material online). Structural alignment allowed the reliable identification of functionally homologous amino acids in A. donax and P. tremuloides proteins.

. 1.

Comparison of AdoIspS to selected isoprene synthases from dicots. (A) Gene structure of Isoprene synthase from Arundo donax (AdoIspS), kudzu (Pueraria montana, PmoIspS) and aspen (Populus tremuloides, PtrIspS). Exons correspond to gray boxes, introns to white ones. (B) Multiple sequence alignment of the C-terminal domain (Class I terpenoid synthase fold) for one selected IspS protein per family. Orange shading indicates residues of the diagnostic tetrad (bold numbers refer to aligned positions of the AdoIspS sequence), while light blue shading indicates the RXR, DDXXD and DTE/NSE motifs. Dots represent residues which are identical to those in AdoIspS.

Tissue- and Stress-Specific Expression Patterns of AdoIspS

Semi-quantitative rtPCR was performed to elucidate the expression patterns of AdoIspS in different organs of A. donax. AdoIspS was highly expressed in leaf blade, while its expression dramatically decreased in leaf sheath (supplementary fig. S7, Supplementary Material online). Only traces of expression were detected in emerging rhizome bud, and there was no expression in root, node and internode.

To estimate the general stress responses of AdoIspS in A. donax at molecular level, young seedlings grown in hydroponic solution were treated with either 150 mM NaCl, 15% PEG 6000, 500 µM CdS04, or a 42 °C heat shock, and the entire shoots were collected to examine the expression pattern of AdoIspS at different time points in different stress conditions, (fig. 2), respectively. The salt treatment did not result in significant variation of the AdoIspS expression levels compared with untreated controls at any time points (fig. 2). PEG and heavy metal treatments induced the upregulation of AdoIspS at specific time points (30′ for both treatments; 11 h and 3 h for PEG and heavy metal treatments, respectively; fig. 2). Heat shock caused the largest transcriptional response of AdoIspS, resulting in a significant upregulation of the gene at all time points tested, up to a 8-fold with respect to unstressed levels (fig. 2).

. 2.

Time course of AdoIspS expression in response to 150 mM NaCl (A), 15% PEG 6000 (B), 500 µM CdS04 (C), and 42 °C heat shock (D) treatments of Arundo donax shoots. The relative expression of AdoIspS in stressed versus unstressed plants was calculated for different lengths of the stress treatment (0.5, 1.5, 3.0, 6.0, 11.0 and 24.0 h) and normalized to the mean of internal control AdoActin at the same time points. Data represent the mean ± SD of three independent experiments. Expression differences were analyzed with Student’s t-test, correcting for multiple tests using Holm’s method. For each time point, the black stars indicate statistically significant differences between AdoIspS expression means of treated and untreated samples (*statistically significant at P < 0.05; **statistically significant at P < 0.01).

Distinctive Features of A. donax IspS with Respect to Dicot IspSs

Phylogenetic reconstructions with representatives of all major classes of angiosperm terpene synthases using both maximum likelihood (ML) and Bayesian inference (BI) confirmed that AdoIspS belongs to class b terpene synthases (supplementary figs. S8 and S9, Supplementary Material online). AdoIspS is basal with respect to the majority of the other known isoprene synthases, and forms a monophyletic clade with two putative terpene synthases from Sorghum bicolor with unknown function (Sb06g002820 myrcene/ocimene synthase, putative; and Sb01g039090 lyase/magnesium ion binding). To gain a more precise relationship of AdoIspS with all isoprene synthases validated to date, ML and bayesian reconstructions were carried out on a larger set of class b terpene synthases and some class g terpene synthases as outgroup (fig. 3). The placement of AdoIspS is in both cases basal to the major isoprene synthase group (closely related to poplar IspS), but differs depending on the phylogenetic method employed. In the ML tree (fig. 3), AdoIspS is at the base of TPS-b Clade 2, while in the BI tree it forms a separate clade basal to both Clade 1 and Clade 2 (part of the clade of basal IspS, supplementary fig. S10, Supplementary Material online). The ML topology corresponds to the one published earlier (Sharkey etal. 2013) and to the one observed in the reconstruction with all TPS classes, thus lending support to the positioning of AdoIspS at the base of Clade 2. Analogously to AdoIspS, the isoprene synthase gene from Casuarina equisetifolia (Oku etal. 2015) and the myrcene/isoprene synthase gene from Humulus lupulus (Sharkey etal. 2013) do not belong to the main IspS clade, but to TPS-b Clade 1, confirming that they likely originated from independent events of parallel evolution (fig. 3 and supplementary fig. S10, Supplementary Material online).

. 3.

Phylogenetic reconstruction of class b terpene synthases and some representatives from class g with maximum likelihood. The branches of validated proteins with exclusive IspS activity are marked in red, while the Myrcene synthase with minor IspS activity from Humulus lupulus is marked in orange. The two names in red indicate IspS enzymes that most likely are the result of parallel evolution with respect to the major clade of “standard” IspS (all from the rosids but for the single fabid protein from Ipomoea batatas). For clarity, only SH-like aLRT support values >0.5 are shown.

AdoIspS shared with isoprene synthases and other class I terpene synthases the functionally relevant motifs previously identified (Degenhardt etal. 2009): the two conserved metal-binding motifs (DDXXD and DTE/NSE) and the RxR motif, involved in stabilization of the released pyrophosphate (fig.1). As confirmed by structural alignments with Populus x canescens IspS (PcIspS; Köksal etal. 2010), only three out of the four amino acids that together are considered diagnostic of IspS activity (Sharkey etal. 2013) (hereafter indicated as “IspS diagnostic tetrad”) were conserved in AdoIspS. In the AdoIspS model, F310, S418 and F457 cluster close to each other and contribute to the definition of the substrate binding site of the enzyme, while S477, the last amino acid of the H-α1 loop, is separated from the others at the top of the active site (supplementary fig. S11A and B, Supplementary Material online). As the IspSs identified so far from species/genera belonging to the same family do not differ in their diagnostic tetrad (supplementary table S1, Supplementary Material online), we derived a family-specific diagnostic tetrad profile, which equally weighs the validated variations among families and mapped it onto the phylogram representing the known taxonomy of the considered families (fig. 4). AdoIspS diagnostic tetrad is identical to that of Anacardiaceae (Ilmén etal. 2015) and differs from the standard one of Salicaceae by the presence in the fourth position of a serine (S477) instead of an asparagine (N505 in Populus species; see protein structural alignment in fig. 1). Excluding from computation the H. lupulus gene, which is primarily a myrcene synthase with secondary IspS activity (Sharkey etal. 2013), both first and third positions of the diagnostic tetrad were invariantly F residues, while the other 2 positions allowed a limited number of variants (supplementary table S1, Supplementary Material online). The ranking in decreasing order of conservation among positions was: Pos1 = Pos3 > Pos2 > Pos4. Only four amino acidic combinations (FSFN, FVFN, FSFS, FVFK) were found to be present in obligate isoprene synthases across the 10 families for which sequences are available. With the exception of the very divergent tetrad from C. equisetifolia (FVFK), the first three tetrads were repeatedly found in more than one (sometimes distantly related) families, indicating parallel and/or homoplasious evolution. IspS sequence information was used to search by Blast for additional IspS proteins from monocots. Two bona fide IspS proteins from Poaceae species Phragmites australis (Genbank accession number GEKX01061072) and Phyllostachis edulis (moso bamboo; accession PH01005131G0060; http://www.bamboogdb.org/; last accessed April 18, 2017) were identified which have the same diagnostic tetrad found in A. donax. Curated searches in the two fully sequenced palm genomes available to date (Phoenix dactylifera and Elaeis guineensis) failed to identify homologues of IspS synthase proteins with any of the diagnostic tetrads identified so far.

. 4.

The extended version of the IspS diagnostic tetrad based on families. (A) Phylogram representing the known taxonomy of the angiosperm families from which IspS genes have been validated. The four diagnostic tetrads are color-coded based on their sequence. (B) Family-based sequence logo of all IspS validated so far.

In Vivo Analysis of Functional Relevance of AdoIspS Diagnostic Tetrad Residues

In order to dissect the contribution of each amino acid of the IspS diagnostic tetrad to isoprene emission, we created by site-directed mutagenesis (1) substitutions of each of the four residues to alanine (F310A, S418A, F457A and S477A), (2) the double mutation F310A/F457A, (3) the S477N mutation (which replaces the A. donax residue with the most common one present in Salicaceae) and (4) a total of additional 36 mutations independently substituting F310 and F457 with each of the remaining 18 amino acids (saturating mutagenesis). Transgenic lines with any of the single alanine mutants dramatically decreased total isoprene emission by 69–98% with respect to WT IspS levels (fig. 5). In addition, isoprene emission of the S477N mutant was almost totally abolished, indicating that S477 cannot be functionally replaced by its most common amino acid counterpart from dicots. Strikingly, replacement of the first phenylalanine residue with alanine (F310A) caused a dramatic change in total volatile emission, with a reduction of isoprene emission to <10% of WT AdoIspS and the preponderant production of monoterpenes, demonstrating neofunctionalization of the mutant enzyme to a monoterpene synthase. By contrast, the F457A mutation had a negative effect on both monoterpene and isoprene emission, as indicated by the strong reduction observed in the F310AF457A double mutant (fig. 5).

. 5.

Volatile emissions from transgenic Arabidopsis lines transformed with WT AdoIspS construct and the six independent site-directed mutageneses F310A, S418A, F457A, S477A, S477N, F310AF457A. (A) Isoprene emission. (B) Monoterpene emission. Letters on top of the bars are the result of one-way analysis of variance (ANOVA) tests. Different letters indicate statistically different means based on Tukey’s test with P < 0.05.

The three-dimensional models of AdoIspS WT and F310A protein structure show that the size of the active site of the F310 mutant enzyme is increased by comparison to that of the WT protein, where the protrusion of F310 aromatic ring causes a significant reduction of the space to accommodate the substrate (fig. 6). To systematically evaluate invivo the role of active site reduction in determining enzyme activity, we performed the saturating site-directed mutagenesis of each phenylalanine residue (F310 and F457).

. 6.

Superposition of structural models of AdoIspS WT and F310A mutant. Modeling was carried out using the SWISS-MODEL server based on the crystal structure of Populus x canescens. The resulting models were superposed using the DeepView program. Three of the amino acids of the WT diagnostic tetrad (F310, S418 and F457) are shown in green, while the F310A mutation is shown in orange. Part of the surface of the F310 A mutant active site is shown as a pink network. The aromatic ring of F310 (present in the WT enzyme) clearly protrudes in the middle of the active site cavity, thus causing its size reduction.

In general, isoprene emission was more tolerant at both positions to substitution of residues with apolar and large side chains, with the exception of F457N, which emitted amounts of isoprene similar to those of WT AdoIspS (fig. 7). In addition, the concomitant or exclusive invivo neofunctionalization to monoterpene synthase was significantly detected only in F310 mutants (especially by alanine replacement and to a minor extent also glycine, serine and cysteine; fig. 7). The majority of mutations abolished both isoprene and monoterpene emission (fig. 7).

. 7.

Relative volatile emissions from transgenic Arabidopsis lines transformed with WT AdoIspS and 38 saturating mutagenesis mutants for diagnostic residue F310 and F457. (A) Relative isoprene emission. (B) Relative monoterpene emission. Relative emission level was calculated as the ratio of net emission (obtained by subtracting background: Emission level of Col-0 for isoprene emission and WT AdoIspS_79 for monoterpene emission, respectively) divided by the isoprene emission of the internal reference (WT AdoIspS_79 in the same PTR-ToF-MS run on the same day). Average of the six top ranking emitters either for isoprene or monoterpene emission out of 30 T2 transgenic lines measured were used for this analysis. Letters on top of the bars are the result of one-way analysis of variance (ANOVA) tests. Different letters indicate statistically different means based on Tukey’s test with P < 0.05. The absolute values corresponding to 100% average (± SD) emissions of isoprene and monoterpene were, respectively, 355.2 ± 106.5 and 428.6 ± 94.5.

To determine the specific monoterpenes emitted by transgenic plants overexpressing p35S:AdoIspS neofunctionalized mutants, GCxGC-ToF-MS analysis of the spectrum of terpenes using the fragments of mass 93, 121 and 136 Da identified several peaks with large quali-quantitative differences with AdoIspS-WT lines (supplementary figs. S12–S15, Supplementary Material online). The large majority of the F310A volatile compounds were acyclic monoterpenes and none of them resulted from carbocation hydroxylation with the exception of negligible amounts of linalool produced also in the control line (Degenhardt etal. 2009). The monoterpene emitted in largest quantity in the neofunctionalized AdoIspS lines was ocimene, which is below detection level in transgenic lines expressing WT AdoIspS (fig. 8 and supplementary fig. S15, Supplementary Material online). While the presence of neo-allo-ocimene isomers in both the commercial standard and the samples from mutant lines are almost certainly due to conversion from beta-ocimene due to thermal desorption of volatiles from the SPME fiber (Chamorro etal. 2014) (supplementary figs. S12 and S13, Supplementary Material online), alpha-ocimene as well as the other unidentified ocimene isomers are bona fide minor products of the mutant enzyme. The only other acyclic monoterpene, beta myrcene, was found in minor amount, while the cyclic monoterpenes gamma terpinene and terpinolene amounted together to <10% of the total monoterpenes (fig. 8). The major monoterpenes emitted were the same in all neofunctionalized lines (supplementary fig. S15, Supplementary Material online). Thus, the F310A, F310G, F310C, and F310S mutations cause the same invivo neofunctionalization of AdoIspS, converting it into a functional ocimene synthase.

. 8.

Relative amounts of the major monoterpenes emitted by F310A AdoIspS. Acyclic monoterpenes are indicated in blue, while cyclic monoterpenes in different hues of orange and red. The unidentified monoterpene (gray) is not visible because present in extremely low amounts.

Discussion

AdoIspS Is the First Isoprene Synthase Gene Identified from Monocots

Several isoprene synthase genes have been characterized in dicots, especially from the rosids (Sharkey etal. 2013). The AdoIspS gene is the first IspS characterized from a monocot species. From a functional point of view, several features of AdoIspS are similar to IspSs from dicots. The AdoIspS transgene is both necessary and sufficient to confer the ability to synthesize isoprene to the nonemitter Arabidopsis (Sharkey etal. 2005; Sasaki etal. 2007; Vickers etal. 2009). Isoprene synthases catalyze the formation of isoprene only when localized in the chloroplast (Vickers etal. 2011). Thus, the chloroplast transit peptide of the monocot A. donax is sufficient for the correct targeting of the mature peptide to the chloroplast also in Arabidopsis. No emission of other volatile compounds besides isoprene was observed in Arabidopsis plants transformed with WT AdoIspS, indicating that in the physiological conditions tested the encoded protein is a highly specific isoprene synthase like poplar and kudzu IspS (Sharkey etal. 2005; Loivamäki etal. 2007; Schnitzler etal. 2010). The genomic structure and the number of exon–introns are highly conserved to those of dicots, and also expression of the gene in A. donax is detected nearly exclusively in photosynthetic tissues, and especially the leaf blade (Sasaki etal. 2005; Sharkey etal. 2005). Like dicot IspS (Sasaki etal. 2005), also AdoIspS expression is mainly responsive to heat stress, but, to a lower extent, also to heavy metal and drought/osmotic stresses, which matches responses of isoprene emission to stress observed in nature (Loreto and Schnitzler 2010). Given the difficulties we encountered in expression of a functional protein in bacteria, the in-depth comparison of AdoIspS enzymatic activity with other IspS awaits validation of a suitable expression system of the tagged enzyme. However, the results presented demonstrate that AdoIspS shares all major functional features with dicot IspSs.

The IspS gene from A. donax fills a large gap in the taxonomic sampling of validated IspS, as it is to date the single-product IspS gene from the most taxonomically divergent lineage with respect to the main IspS rosids clade. It has been previously hypothesized that five to six independent origins of IspS may have occurred in angiosperms, one of them being in commelinids (encompassing both Poales and Arecales orders; Sharkey etal. 2013). The separation of AdoIspS from the main rosids IspS clade supports parallel IspS evolution in Poaceae and dicotyledons. Despite divergent evolution cannot be completely ruled out, a parallel evolution scenario seems to be more consistent with the small number of phylogenetically related species emitting isoprene in Poales (Sharkey etal. 2013), as well as with their highly derived position among monocots (Chase etal. 2016). The recent isolation of Casuarina equisetifolia IspS, a species from the rosid order fagales, constitutes an independent evidence of parallel evolution of single-product IspS based on the clearly polyphyletic origin of fagales IspS (Monson etal. 2013; Oku etal. 2015). Taken together, these results suggest that isoprene emission independently evolved at least three times in angiosperms, twice in rosids and once in Poaceae.

A Possible Role for S418 and S477 in Carbocation Stabilization and Active Site Closure

Our revision of the IspS diagnostic tetrad using a family-based normalization captures the relevant variation of amino acid combinations among the angiosperm single-product IspS known so far, indicating that positions 2 and 4 are subject to a more relaxed selection with respect to specification of IspS activity. Lack of full conservation has been recently interpreted as indication of accessory function for these residues (Ilmén etal. 2015). Relative emissions of AdoIspS S418A and S477A mutants invivo demonstrate the essential function of S477, suggesting a possibly less fundamental, although relevant, role of S418 in IspS activity.

The region corresponding to the serine triplet centered around AdoIspS position 418 has been recently identified (Schrepfer etal. 2016) as part of a triad of amino acids responsible for a mechanism of induced fit substrate ionization in taxadiene synthase (TXS). Only minor movements of S418 and neighbouring residues can be inferred by the structural model obtained from PcIspS and this mechanisms of substrate ionization is in contrast to the proposed role of DMADP diphosphate as the general base for carbocation formation (Köksal etal. 2010). The carbonyl oxygen atom of S418 is, however, in closest proximity to the C31 and C4 methyl groups of the substrate analogue DMASPP, suggesting that it may contribute to the stabilization of the geranyl and/or linalyl carbocation (Köksal etal. 2010). Consistent with the possible involvement of the S triplet in carbocation stabilization, the region homologous to AdoIspS S417-L422 has been shown to contribute to product specificity in monoterpene synthases (Kampranis etal. 2007), as suggested by the deformation of the G1/G2 helix-break motif observed in several TPS upon transition to the closed conformation (Baer etal. 2014). This model is compatible with the Ser-Val variability observed in isoprene synthases, indicating that side chain size may be more relevant than polarity for optimal positioning/function of position 418 alpha carbonyl oxygen.

S477 is the last amino-acid of the H-α1 loop, which is involved in capping of the active site of several TPS (Whittington etal. 2002). Despite mutation of AdoIspS S477 structural homologue in MdAFS1, an apple α-farnesene synthase, abolishes K+ dependence and phenocopies the corresponding AdoIspS mutation (Green etal. 2009), a primary role of S477 in K+ coordination seems unlikely because this ion enhances, but is not essential for IspS activity (Köksal etal. 2010). Bornyl synthase K512 residue, the structural homologue of AdoIspS S477, donates a hydrogen bond to the diphosphate group of the substrate (Whittington etal. 2002), indicating that S477 may instead contribute to diphosphate positioning/sensing during transition to the closed conformation.

These observations suggest that S477 may be part of a mechanism of substrate sensing and conformational remodeling, associated to transition to closed conformation, ancestrally shared by several TPS enzymes (Whittington etal. 2002). Crystallographic data indicate that in isoprene synthases the transition from open to close conformation may not happen. However, it cannot be excluded that structural changes may have been prevented invitro by molecule packing in the crystal lattice (Köksal etal. 2010), leaving open the possibility of a so far undetected close state transition. Additional studies will, thus, be required to elucidate whether this hypothesis is supported by invivo evidence, and the eventual significance of this mechanism in IspS function.

A Pivotal Role of F310 in the Evolution of Short-Chain TPS

Saturating mutagenesis of positions 310 and 457 uncovered several amino acids compatible with detectable levels of isoprene emission, but only one of the possible combinations among them (F310/F457) has been found in naturally occurring single-product IspS enzymes to date. This result is consistent with the fact that only a minority of the possible mutational paths in the fitness landscape are accessible to Darwinian selection of protein function due to pleiotropy (Weinreich etal. 2006). The high proportion of substitutions abolishing or drastically reducing both isoprene and ocimene synthase emission (indicating that they affect the overall structure/stability of the enzyme) supports this view.

Previous studies predicted that both F310 and F457 residues could decrease the size of the substrate binding pocket and make van der Waals contacts with the substrate DMADP, thus determining isoprene production instead of monoterpene synthesis from GDP (Sharkey etal. 2005; Köksal etal. 2010). Our results demonstrate invivo that only position 310 plays a fundamental role in AdoIspS substrate specificity determination by active site size modulation. In the lack of data from purified enzymes, the relative entity of the ocimene and isoprene enzymatic activities underlying the dramatic changes in volatile emission observed in transgenic plants harboring the AdoIspS F310A mutant is currently unknown. Due to the limited pool of isoprene and ocimene precursors in Arabidopsis, neofunctionalization and the consequent activation of a novel sink for GDP may alter in nonlinear ways the amounts of AdoIspS F310A substrates, which in addition could compete with each other for active site binding. Strikingly, also single mutations of AdoIspS F310 homologue in widely divergent cineole synthases from plants (Salvia officinalis) and fungi (Hypoxylon sp.) shifted invitro substrate specificity from GDP to FDP, causing neofunctionalization from monoterpene to sesquiterpene synthase activity (Kampranis etal. 2007; Shaw etal. 2015). Salvia officinalis cineole synthase N338A mutant, the structural homologue of AdoIspS F310A, displayed a nearly 200-fold decrease of the catalytic efficiency of the enzyme when using FDP instead of GDP as only substrate (Kampranis etal. 2007). Unfortunately, neither enzymatic activities were tested with different ratios of both substrates, nor was the impact of this change assessed invivo in terms of total volatile emission. The actual enzymatic underpinnings of the preponderant ocimene emission from AdoIspS F310A, thus, remain an open question awaiting the establishment of a suitable expression system. Irrespective of the enzyme kinetics details, these results strongly support that the very same amino acidic position is involved in substrate specificity determination of different short-chain TPS-b (hemiterpene, monoterpene and sesquiterpene synthases) and that in the case of isoprene synthases this single mutation has a large effect on the total emission of volatiles invivo. This plasticity determinant is in common between the monocotyledonous and dicotyledonous lineages. Additional evidences, however, indicate that the pivotal role of this position in plants likely extends even further, to the gymnosperm lineage. F310, in fact, is structurally homologous to the phenylalanine hypothesized to determine substrate specificity of a conifer methylbuthenol synthase (Gray etal. 2011), strongly supporting that the independent evolution of hemiterpene synthases in gymnosperms and angiosperms relied on the parallel acquisition of a phenylalanine residue in the same position of the class I terpenoid synthase fold helix D. Analogously, substrate and product specificity has been demonstrated to rely on this very same residue in the parallel evolution of cineole synthases in plants and fungi (Kampranis etal. 2007; Shaw etal. 2015). By contrast to the double function in substrate preference and product specificity of cineole synthases, the saturating mutagensis of F310A indicate that in AdoIspS the F310 residue is involved in determination of product specificity as a consequence of substrate preference, as the same products are present in all neofunctionalized mutants.

The substrate actually used by some TPS enzymes can be up to one isoprenoid unit shorter than the one potentially fitting into the active site cleft (Köksal etal. 2011). This suggests that current substrate preference is the relatively recent result of one/few key mutations preventing accommodation of a larger, ancestral substrate for which the overall active site cavity architecture/size was optimized in the course of evolution (Pazouki and Niinemets 2016). As the architecture and maximal size of the active site depends from many residues, these features are expected to evolve slowly and to constrain the effect of mutations of plasticity residues, which, by definition, are major switches for evolutionarily fast neofunctionalization (Yoshikuni and Keasling 2007). Analogously to the “molecular ruler” mechanism of prenyl transferases (Wang etal. 2016), thus, we propose that, in class I TPS, plasticity towards different substrates relies mainly on the geometric constraints imposed by α-helical modularity and by the distance from the site of coordination of the magnesium triad. The side chain of any amino acid like F310, laying exactly two helical turns below (i.e., seven positions before) the first aspartic residue of the DDXXD motif, is exposed to the active site hydrophobic cleft, and defines, according to its size, the possibility to accommodate in a conformation compatible with catalysis DMADP versus GDP or GDP versus FDP as substrate. This structural modularity confers to terpene synthases a surprising capacity to revert to original functions, which is not always shared by other enzymes (Kaltenbach etal. 2015).

Based on this general mechanism, the preponderant production of ocimene in all AdoIspS F310 neofunctionalized mutants compared with the very low amounts of myrcene and other monoterpenes confirms the analogy between isoprene and ocimene reaction chemistries (Sharkey etal. 2013). This in turn points to ocimene synthases as the most likely ancestral TPS from which IspS evolved by active site size reduction following acquisition of F at position 310 (Gray etal. 2011) (fig. 9). Given the intrinsically high capacity of TPS to evolve new functions (Christianson 2008), discerning past evolutionary relevance from evovability potential of plasticity residues identified exclusively by structurally informed choices is generally difficult. The choice of the residues to mutagenize in this study, on the contrary, was evolutionarily informed, as it relied on the conservation of the four residues that are preferentially (diagnostic tetrad positions 2 and 4) or exclusively (diagnostic tetrad positions 1 and 3) present together in isoprene synthases, but in no other known terpene synthase (Sharkey etal. 2013). The saturating mutagenesis and functional characterization invivo of the two F residues, thus, amounts to the closest experimental recapitulation of the mutational landscape faced during IspS gene evolution by plant species lacking the trait, in line with Jensen’s hypothesis and many directed mutagenesis experiments (Khersonsky and Tawfik 2010). The strong invivo reduction of the original isoprene emission, maintained in the F310A lines as promiscuous function, indicates strong negative trade-offs in the evolution of IspS and rapid fixation of the trait, suggesting that the adaptive driver(s) of trait evolution may have concomitantly selected in favor of the new function and against the original one (Khersonsky etal. 2006; Khersonsky and Tawfik 2010). As the mutagenesis has been carried out on an extant IspS gene, a limitation of the approach used in this study is that it cannot take into account mutations with functional relevance that may have happened after the gain of IspS function. Several other evidences, however, independently corroborate derivation of isoprene synthases from ancestral ocimene synthases. IspS active site volume is as large as that of other monoterpene synthases and it was demonstrated that GDP competes with DMADP for binding to IspS active site (Köksal etal. 2010, 2011), indicating, for the reasons provided above, that the ancestors of IspS probably were monoterpene synthases (Dani etal. 2014) using GDP as substrate. In addition, the catalysis of IspS is inefficient compared with monoterpene synthases, further implying that IspS specificity for DMADP may be a derived trait (Lerdau and Gray 2003). Finally, evolution of IspS from ocimene synthases ancestors is also supported by the derived position of all known IspS genes with respect to other terpene synthases in the TPS-b clade (Sharkey 2013; Sharkey etal. 2013) (fig. 3) and by the monophyly of isoprene and ocimene synthases in the TPS-b2 clade (Sharkey etal. 2013).

. 9.

Model of gain/loss of isoprene synthase function in angiosperms. Solid arrows indicate experimentally supported processes, dashed arrows postulated ones. (A) IspS genes are proposed to have evolved multiple times from ocimene synthases (TPS-b clade), but the rate of first recruitment from the pool of pre-existing ocimene synthases is currently unknown. It is postulated that primary gains are relatively rare (dashed black line), but once an IspS gene has been acquired in a clade, single mutations can switch the enzyme substrate specificity back and forth between isoprene and ocimene synthase by maintaining active site overall structure, potentially leading to repeated losses and secondary gains of the trait in the same clade (blue arrow “Reversion” and black arrow “Secondary gain”). Additional mutations can lead to either divergence or can hit functionally fundamental amino acids leading to sub-functionalization (or complete protein inactivation). (B) Molecular mechanism for the evolution of isoprene synthases supported by the results of this work. Both primary and secondary gains of isoprene emission are associated with a mechanism of active site reduction (depicted as a shallower cavity in the enzyme) by replacement of the seventh position before the first D of the DDXXD motif (AdoIspS position 310) with the bulky phenylalanine residue (F). The reduction of the size of active site prevents the efficient use of geranyl pyrophosphate (GDP) as substrate for the production of ocimene, causing instead the conversion of dimethylallyl pyrophosphate (DMADP) to isoprene. Additional mutations leading to putative structural changes may be required for the first recruitment (dashed arrow) of an Ocimene synthase (OciS) from the pool of TPS-b genes present in each species. Enzymatic activity can potentially revert to ocimene synthesis by a second substitution to the same position with a smaller residue, which is preferentially alanine (A) in AdoIspS.

With the notable exception of F457N, the tolerance of isoprene activity to amino acid replacements at position 310 and 457 is similar. F457, however, is basically unable to promote significant invivo neofunctionalization. The different plasticity of the two residues (Yoshikuni and Keasling 2007), indicates that acquisition of an F at position 310 most likely constituted a more relevant constraint in the evolution of IspS activity compared with appearance of F457. This conclusion is supported by two evidences. First, the secondary neofunctionalization to isoprene synthase activity detected in a myrcene synthase from hops is associated to the exclusive presence of the structural homologue of F310, but not of F457 (Sharkey etal. 2013). Second, relative occurrence in TPS-b monoterpene synthases of F in correspondence of position 457 is higher than for position 310, indicating a lower selectivity of position 457 towards the amino acid side chain.

A Unified Model for IspS Evolution

Taken together, the results presented above reconcile the apparently contrasting evolutionary patterns previously observed (Monson etal. 2013; Sharkey 2013; Sharkey etal. 2013) in the unified conceptual framework summarized in figure 9. On one hand, as suggested by the relatively low number of independent gains estimated in this study, primary acquisition of IspS activity seems to be a relatively complex and rare event. All amino acid replacements at position 310 produced low amounts of monoterpenes derived from hydroxylation of geranyl or linalyl cation intermediates (Degenhardt etal. 2009). Even mutation F310N, replacing Phe with the Asn responsible of coordination of the water molecule which hydroxylates the substrates in both plant and fungal cineole synthases, does not produce hydroxylated products, being instead basically inactive. This indicates that, by contrast to cineole and BPP synthases, reduced hydroxylation ability due to lack of the cavity necessary for water coordination at the bottom of the active site is ancestral for isoprene as well as ocimene synthases (Kampranis etal. 2007; Köksal etal. 2010). The existence of clear evolutionary constraints related to the overall active site cavity volume and the ability to catalyze hydroxylation and efficient cyclization of carbocation intermediates and the phylogenetic evidences, thus, support that all rosids and possibly fabids IspSs are the likely result of a single, primary gain of IspS function. On the other hand, the potential ease of invivo interconversion between ocimene and isoprene synthases by single mutations demonstrated in this study provides a rationale contributing to explain the repeated evolutionary loss and especially re-gain of IspS activity at the family and genus level (Monson etal. 2013) after any primary acquisition of the trait (fig. 9).

As discussed earlier, the insights provided in this study through an evolutionarily informed functional analysis of the residues preferentially or exclusively present in isoprene synthases are a good starting point, but for sure not an exhaustive elucidation of the factors contributing to IspS evolution. As recently pointed out, the selective drivers underlying evolution of the trait as of today are still unknown and their identification will be definitely required to dissect the details of lineage-specific conservation or loss of isoprene emission (Sharkey and Monson 2017). The findings of this work, by providing a mechanistic basis for the structural constraints controlling IspS and more generally TPS evolution by neofunctionalization, and identifying the likely ancestors of IspS genes, will hopefully help to stimulate novel research lines focusing on the possible adaptive trade-offs that may exist between emission of evolutionarily sister volatile compounds, like isoprene and ocimene.

Isoprene synthase is the single enzyme responsible for the yearly emission worldwide of ∼600 Tg of isoprene, the most abundant biogenic volatile organic compound and possibly the one with the largest impact on global atmospheric chemistry (Sindelarova etal. 2014). Stunningly, the molecular mechanism at the root of the evolution of such hugely important ecological phenomenon has been traced back to the mutation of a single amino acid responsible of active site size and substrate specificity modulation. The elucidation of the likely evolutionary path of isoprene synthase provides a relevant example of how parallel evolution may be driven by relatively simple biophysical constraints and underlines the power of the emerging paradigm of evolutionary biochemistry to define the intimate molecular evolutionary links between the structural and functional bases of traits with global relevance (Harms and Thornton 2013).

Materials and Methods

Plant Materials, Growing Condition, and Stress Treatments

Rhizomes and cohorts of A. donax cuttings (collected in Sesto Fiorentino, Florence, Italy 43°49′01.8″N 11°11′57.0″E) were used in this study. The plants generated from rhizomes were grown under the same condition as described previously (Sablok etal. 2014). Arundo donax cuttings were used for different stress treatments under the same growing condition used in earlier studies (Fu etal. 2016). When the new shoots reached the five-leaf stage, they were subjected to different stress treatments. The stress treatments performed were: water limitation (addition of 15% PEG 6000 to the basal hydroponic solution), salt (addition of 150 mM NaCl), heavy metal (addition of 500 µM CdS04), heat (42 °C) for 24 h. The entire shoots (treated and nontreated) were collected separately at different time points (0.5, 1.5, 3, 6, 11 and 24 h after inducing the stress), snap-frozen in liquid N2 and stored at −80 °C till use. At least three plants per biological replica and three biological replicas per treatment and respective controls were sampled. For volatile compound analyses, transgenic and wild-type Arabidopsis plants (Col-0) growing at 23 °C in the growth chamber under standard long-day condition (16 h light/8 h dark) were used.

RNA Extraction and RT PCR Analyses

Total RNA was extracted from different organs of A. donax under different stress conditions with Trizol (Invitrogen) according to the manufacturer’s instruction, and treated with Dnase I (Rnase-free) (Sigma–Aldrich) to eliminate genomic DNA contamination. One µg of total RNA was reverse transcribed with Superscript III reverse transcriptase (Invitrogen). The synthesized cDNA was used as a template for semi-quantitative and real-time PCR analyses. Semi-quantitative rtPCR was performed for different organs of A. donax using the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene as reference. For stress treatments, real-time PCR was carried out with Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen) in a Bio-Rad C1000 Thermal Cycler detection system using a putative A. donax actin (Fu etal. 2016) and Arabidopsis AtActin2 (Lim etal. 2014) as a reference gene for A. donax and transgenic Arabidopsis lines, respectively. All reactions were performed in triplicates and the 2−ΔΔCT method was used to calculate fold changes. Significance of expression difference was calculated on Δct averages of three biological replicates based on the Student’s t-test using Holm’s correction for multiple tests (n = 6, α = 0.05). Primers used for PCR are listed in supplementary table S2 in Supplementary Material online.

Cloning and Plant Transformation

The full-length cDNA of AdoIspS was amplified with primers AdoIspS_For and AdoIspS_Rev (supplementary table S2, Supplementary Material online) using Phusion High-Fidelity DNA Polymerase (NEB). The purified amplicons were cloned into Gateway vector pENTR/D-TOPO (Invitrogen) to yield pENTR_AdoIspS. A total of 42 mutations were introduced into AdoIspS WT CDS using the Quikchange Site-Directed Mutagenesis Kit (Stratagene) in the A. donax pENTR_AdoIspS plasmid: The four amino acids corresponding to the “IspS diagnostic tetrad” previously identified (Sharkey etal. 2013) (F338, S446 [identified based on its being in “the middle of a triple serine” and being “most commonly Val or Ile in other proteins,” but erroneously indicated as S445 in the original publication], F485 and N505 of PcIspS) are F310, S418, F457 and S477 in AdoIspS. Saturating mutagenesis was applied for both F310 and F457, (respectively, mutations F310A, F310C, F310D, F310E, F310G, F310H, F310I, F310K, F310L, F310M, F310M, F310N, F310P, F310Q, F310R, F310S, F310T, F310V, F310W, F310Y; and F457A, F457C, F457D, F457E, F457G, F457H, F457I, F457K, F457L, F457M, F457M, F457N, F457P, F457Q, F457R, F457S, F457T, F457V, F457W, F457Y). Furthermore, S418 and S477 were independently mutagenized to alanine (S418A and S477A), and the monocot-specific S477 residue was replaced also with asparagine (S477N); finally, the double mutant F310A/F457A was constructed. All these 42 pENTR clones with IspS diagnostic tetrad mutations plus wild-type pENTR_AdoIspS (AdoIspS-WT) were further recombined in the destination vector pK7WG2 (Karimi etal. 2002) to generate the final constructs through LR reaction with LR clonase II (Invitrogen). These constructs were transformed independently into Agrobacterium tumefaciens strain GV3101-pMP90RK by electroporation and further transformed into Arabidopsis thaliana Col-0 ecotype by the floral dip method (Clough and Bent 1998). Transgenic plants were obtained by screening sterilized seeds on solid MS (Murashige and Skoog) medium supplemented with 50 mg l−1 of kanamycin. The primer sequences used for construct preparation are listed in supplementary table S2 in Supplementary Material online.

Genotyping

Segregation analysis was carried out on T2 segregating plants transformed with p35S:AdoIspS (AdoIspS-44 line). T2 seeds stratified at 4 °C for three days were sown directly in pots filled with commercial soil GS90L. Four-week-old plants were used for both isoprene emission measurement (see below) and genomic DNA extraction with the CTAB method (Doyle 1987). DNA quality and presence of the AdoIspS transgene were assayed by PCR using primer pairs AtActin2_For + AtActin2_Rev and AdoIspS_Rev + pK7WG2_2S, respectively. Thermal cycling was carried out as follows: 95 °C for 2 min, 35 cycles at 94 °C for 40 s, 60 °C for 30 s and 72 °C for 2 min. Primer sequences are listed in supplementary table S2 in Supplementary Material online.

PTR-MS Measurements

One young and mature rosette leaf for each genotype (Col-0 and transgenic lines) was cut and transferred into a 20-ml glass vial containing 300 µl of distilled water. The vials were kept open for at least 30 min and immediately closed with metal lids sealed with silicon septa (Agilent), and then incubated for 3 h at 30 °C with light intensity of 130–150 µmol m−2 s−1. The measurement was done with a commercial PTR-ToF 8000 apparatus from Ionicon Analytik GmbH, Innsbruck (Austria). The data of each measurement were extracted and converted to parts per billion (ppb) following the methodology previously described (Cappellin etal. 2011), and the concentration was converted to flux on a mass basis (nmol ) by using the corresponding individual leaf fresh weight.

Volatile Extraction Conditions and Analysis

The headspace-solid phase microextraction (SPME) method was applied to extract volatile compounds with slight modification from the former description (Matarese etal. 2014). One gram of leaves pretreated as for PTR-MS measurement was collected and ground with liquid nitrogen, then mixed with 0.3 g of NaCl and 2 ml of freshly prepared citrate-phosphate buffer (0.1 M Na2HPO4, 50 mM citric acid, pH 5.0) in 20 ml headspace glass vials. Each sample was spiked with 50 μl of 2-octanol at 1 mg l−1 as an internal standard. Three independent biological replicates for each genotype were analyzed to generate each data point.

The headspace was sampled using 2 cm DVB/CAR/PDMS 50/30 μm fiber from Supelco (Bellefonte, PA) equipped with a Gerstel MultiPurpose Autosampler (GERSTEL GmbH & Co. KG, Mülheim an der Ruhr Germany). Samples were kept at 60 °C for 20 min and then extracted for 35 min at 60 °C. The volatile and semi-volatile compounds were desorbed in the GC inlet at 250 °C for 3 min in splitless mode and the fiber was reconditioned for 7 min at 270 °C. The GC × GC system consisted of an Agilent 7890 A (Agilent Technologies, Santa Clara, CA) gas chromatography equipped with a Pegasus IV time-of-flight mass spectrometer (Leco Corporation, St. Joseph, MI). A VF-Wax column (100% polyethylene glycol; 30 m ×0.25 mm ×0.25 μm, Agilent J&W Scientific Inc., Folsom, CA) was used as a first-dimension (1D) column and a RXI-17Sil MS-column 1.50 m ×0.15 mm ×0.15 μm (Restek Bellefonte) as a second-dimension (2D) column.

The GC system was equipped with a secondary oven and nonmoving quadjet dual-stage thermal modulator. The injector/transfer line was maintained at 250 °C. Oven temperature program conditions were as follows: initial temperature of 40 °C for 4 min, ramped at 6 °C min−1 to 250 °C, where it remained for 5 min. The secondary oven was kept 5 °C above the primary oven throughout the chromatographic run. The modulator was offset by +15 °C in relation to the secondary oven; the modulation time was 7 and 1.4 s of hot pulse duration. Helium (99.9995% purity) was used as carrier gas at a constant flow of 1.2 ml min−1. The MS parameters included electron ionization at 70 eV with ion source temperature at 230 °C, detector voltage of 1,317 V, mass range of m/z 35–450 and acquisition rate of 200 spectra per second.

GCxGC-ToF-MS data alignment and processing

ChromaTOF software version 4.32 was used to perform baseline correction, chromatogram deconvolution and peak alignment. The baseline offset was set to 0.8 and signal to noise (S/N) ratio was set at 100. Required match (similarity) to combine peaks was set to 650. Library (NIST and Wiley) selection was conducted for molecular weights between 35 and 350 with five library hits to return. Mass threshold was set at 50 and minimum similarity to assign compound name was set at 700. Identification of some volatile compounds was performed by injecting pure reference standards (beta ocimene). The semiquantitative analyses were based on peak area ratios relative to the internal standard 2-octanol.

Sequence Alignment and Phylogenetic Reconstruction

To elucidate the relationship of A. donax IspS with respect to known angiosperm terpene synthases and dicotyledonous isoprene synthases, we aligned the AdoIspS protein and reference IspS sequences from Salicaceae, Fabaceae and Myrtaceae to TPS proteins from three previously published datasets (Sharkey etal. 2013; Ilmén etal. 2015; Oku etal. 2015). A second dataset was created by aligning AdoIspS and all other IspS proteins functionally validated to date to a selection of TPS-b and TPS-g representatives used for the identification and classification of Eucalyptus terpene synthases (Külheim etal. 2015). Multiple sequence alignments were carried out with Mafft v.7 using the highly accurate L-INS-i method (Katoh and Standley 2013), removing short sequences to avoid their potentially deleterious effect on the accuracy of phylogenetic reconstruction. To remove potentially disordered/low-homology regions, the multiple sequence alignments encompassing the remaining proteins were processed with GBLOCKS v. 0.91b (Talavera and Castresana 2007), using the options: maximum number of contiguous nonconserved positions = 8, minimum length of a block = 5, and retaining positions with <50% of gaps. We used the SMS (Smart Model Selection; http://www.atgc-montpellier.fr/sms/; last accessed April 18, 2017) web server to identify the optimal model of amino acid substitution for each alignment. Maximum likelihood (ML) phylogenetic reconstruction was carried out with PhyML v.3.0 (Guindon etal. 2010). ML tree heuristic search was carried out with the nearest-neighbor interchanges (NNI) strategy, using both branch length and topology optimization. Branch support was calculated using the approximate Likelihood-Ratio Test (aLRT). Terpene synthase subfamilies were identified according to (Külheim etal. 2015). Bayesian inference (BI) phylogenetic reconstruction was performed with MrBayes v3.2 (Ronquist etal. 2012), using two independent runs, each with one cold and three heated chains over 5,000,000 generations. Trees were sampled every 5,000 generations and posterior probabilities of splits were obtained from the 50% majority rule consensus of the sampled trees, discarding 25% as burn-in. Tree editing was carried out with the iTOL v.3.0 program (Letunic and Bork 2011). Both alignments and the corresponding ML and BI trees are available from TreeBASE (accession numbers S19973 and S19974).

Structural modeling was carried out with the SwissModel server (Arnold etal. 2006).

Supplementary Material

Supplementary data are available at Molecular Biology and Evolution online.

Click here for additional data file.