Molecular evolution of terpene synthase underlying the diversification of isoprene emission in Fagaceae

Abstract

• Plants emit a wide range of volatile organic compounds, among which isoprene is the most abundant and atmospherically influential. Although oak species are major contributors to isoprene emission, there is considerable variation in isoprene emission capacity within the Fagaceae family.

• To unravel the evolutionary origins of isoprene emission, we investigated the molecular evolution of terpene synthase (TPS) genes across eight species within the Fagaceae.

• We identified a TPS‐b subclade in which potential isoprene synthase (IspS) activity evolved independently in two gene lineages within subgenus Quercus. Ancestral sequence reconstruction revealed that the acquisition of a diagnostic amino acid residue for IspS function arose convergently in these lineages and was subject to positive selection, suggesting adaptive evolution. Ancestral‐enzyme assays targeting the gene lineage with high gene expression revealed that the early protein primarily produced monoterpenes from geranyl diphosphate (GPP), whereas their descendants shifted substrate preference to dimethylallyl diphosphate (DMAPP), evolving into dedicated isoprene synthases.

• Our results indicate that IspS activity was not ancestral in Fagaceae, but evolved approximately 56 million years ago within the subgenus Quercus, and has been retained ever since. These findings emphasize the roles of enzyme structural innovation and regulatory shifts in the diversification of volatile terpenoid biosynthesis.

Isoprene emission in oaks originated from convergent adaptive evolution of terpene synthases, involving a substrate shift from monoterpene to isoprene production within a Fagaceae‐specific TPS lineage.

INTRODUCTION

Plants release a wide spectrum of biogenic volatile organic compounds (BVOCs), which include molecules with diverse structural and physicochemical characteristics (Kesselmeier & Staudt ¹⁹⁹⁹; Laothawornkitkul et al. ²⁰⁰⁹; Satake et al. ²⁰²⁴). These compounds, emitted from various plant tissues such as leaves, flowers, fruits and roots, enter the atmosphere in significant quantities (Antonelli et al. ²⁰²⁰). BVOCs serve as a dynamic interface between terrestrial ecosystems and the atmosphere, playing an important role in climate and atmospheric chemistry (Lee Ng et al. ²⁰¹⁷). The annual global emission of BVOCs is estimated to be around 1000 Tg C, with terpenoids—including isoprene, monoterpenes and sesquiterpenes (Guenther 1995; Guenther et al. ²⁰¹²). Among these, isoprene alone accounts for roughly half of the total BVOC, underscoring its ecological and atmospheric significance (Guenther 1995).

In the plant kingdom, isoprene emission is distributed heterogeneously across plant taxa (Harley et al. ¹⁹⁹⁹; Monson et al. ²⁰¹³; Dani et al. ²⁰¹⁴). It is particularly common in fast‐growing plants, especially hardwood trees such as oaks (Quercus spp.), Populus spp. (Miller et al. ²⁰⁰¹), and Eucalyptus spp. (Dani et al. ²⁰¹⁴). Isoprene emission has also been reported in mosses (Hanson et al. ¹⁹⁹⁹), ferns (Tingey et al. ¹⁹⁸⁷) and legumes (Harley et al. ²⁰⁰⁴), indicating its occurrence across diverse plant groups. Isoprene is synthesized by isoprene synthase (IspS), a class I terpene synthase (TPS) enzyme that converts dimethylallyl diphosphate (DMAPP) into isoprene at a key terminal branching point of the methylerythritol phosphate (MEP) pathway (Silver & Fall ¹⁹⁹⁵; Sharkey et al. ²⁰¹³).

The TPS family likely originated in ancestral land plants following their divergence from green algae (Jia et al. ²⁰²²). In angiosperms, IspS belongs to the TPS‐b subfamily (Sharkey et al. ²⁰¹³; Li et al. ²⁰¹⁷), whereas IspS in mosses belongs to the TPS‐c subfamily (Kawakami et al. ²⁰²³), suggesting that IspS has multiple evolutionary origins within the plant kingdom. In angiosperms, IspS appears to have evolved multiple times, likely from closely related terpene synthases. Among these, trans‐β‐ocimene synthases are considered a key precursor, reflecting the functional diversification of TPSs in plants (Sharkey et al. ²⁰¹³; Li et al. ²⁰¹⁷).

An intriguing and well‐studied example of the uneven distribution of isoprene emission within a plant clade can be found in the Fagaceae family (Harley et al. ¹⁹⁹⁹; Monson et al. ²⁰¹³). In this family, isoprene emission is particularly prominent in the genus Quercus (oaks), whereas other genera, such as Fagus, Castanopsis and Lithocarpus, exhibit little to no isoprene emission (Loreto 2002; Monson et al. ²⁰¹³; Tani et al. ²⁰²⁴). Intriguingly, within the genus Quercus, isoprene emission capacity exhibits significant variation, with a mix of strong isoprene emitters, such as Q. rubra and Q. robur and non‐emitters, such as Q. ilex and Q. suber (Loreto and Fineschi 2015). Based on analyses associating isoprene emission capacity with molecular phylogeny within the Fagaceae, previous studies concluded that isoprene emission was likely an ancestral trait in the genus Quercus which was subsequently lost—and not regained—in certain species within the Quercus sections Cerris and Ilex (Loreto 2002; Monson et al. ²⁰¹³). These findings lay a critical foundation for reevaluating the evolutionary history of isoprene emission in the Fagaceae. However, a systematic analysis integrating phylogenetic relationships of TPS genes, genome evolution and functional diversification of isoprene synthase in Fagaceae has been lacking to date.

To explore molecular evolution of terpene synthase underlying the diversification of isoprene emission in Fagaceae, we comprehensively identified TPS genes from eight species representing four genera within the family and examined their phylogenetic relationships, expression patterns and contributions to isoprene emission. Recent advances in genome and transcriptome sequencing have greatly expanded the availability of high‐quality data for non‐model Fagaceae species (Kremer et al. ²⁰¹²; Satake et al. ²⁰²³; Kudo et al. ²⁰²⁵), enabling systematic investigation of the evolutionary trajectory of isoprene emission in this family. Using comparative genomics, ancestral sequence reconstruction and assays of reconstructed ancestral enzymes, we show that isoprene emission was not ancestral in the Fagaceae. Instead, IspS evolved approximately 56 million years ago within the subgenus Quercus and has been retained in this lineage ever since. These results further reveal convergent evolutionary origins of IspS in two gene lineages within subgenus Quercus, shedding light on a previously unrecognized evolutionary pathway to isoprene emission in angiosperms.

MATERIALS AND METHODS

Genome synteny analysis

To compare the genome structures of Fagaceae species across genera, published genome sequences of eight Fagaceae species from four genera were obtained from the public databases and used for subsequent analyses (Table S1). We first performed an all‐vs‐all BLASTp search using protein sequences with a cut‐off e‐value of 1E‐5 to identify homologous relationships across the genomes. To investigate genomic synteny, we detected syntenic blocks within or between genomes using MCScanX (https://github.com/wyp1125/MCScanX), based on the results of the above BLASTp searches. The synteny blocks were visualized using SynVisio (https://synvisio.github.io). To clarify syntenic relationships across chromosomes in each species, we renumbered chromosomes according to the chromosomal order in Q. robur. Additionally, chromosomes oriented opposite to those of Q. robur were reversed. These adjusted chromosomes were used for subsequent analyses.

Species phylogeny

We employed a supermatrix approach to infer phylogenetic relationships of eight Fagaceae species, following the method outlined by Rokas et al. (2003) and Burleigh et al. (2015). Orthologous groups across the eight species were identified using OrthoFinder v2.5.5 (https://github.com/davidemms/OrthoFinder), and single‐copy orthologs conserved in all eight species were selected for phylogenetic analysis. Protein sequences within each orthologous group were aligned using MAFFT v7.520 (https://mafft.cbrc.jp), and poorly aligned regions—defined as sites with gaps in ≥30% of sequences—were removed using trimAl v1.5.rev0 (https://vicfero.github.io/trimal). The resulting alignments were concatenated into a single supermatrix. The best‐fit substitution model was determined using ModelTest‐NG v0.1.7 (https://github.com/ddarriba/modeltest). Phylogenetic tree construction was performed with RAxML‐NG v0.9.0 (https://github.com/amkozlov/raxml‐ng) under the JTT + I + G4 + F model with 500 bootstrap replicates. The resulting tree was visualized using FigTree v1.4.4 (https://tree.bio.ed.ac.uk/software/figtree).

Identification of TPS family members

We retrieved candidate genes of the TPS family based on two conserved motifs: the metal‐binding domain (PF03936) and the N‐terminal domain (PF01397). Protein sequences from eight Fagaceae species were used for a domain search against the Pfam‐A database ver 35.0 (Mistry et al. ²⁰²¹) using HMMER ver 3.3.2 (http://hmmer.org) with a cut‐off e‐value of 1E‐5. For genes with more than two isoforms, the longest isoform was selected.

To exclude candidate genes with short sequence lengths, we utilized the conserved exon‐intron arrangement patterns found in TPS genes (Trapp & Croteau ²⁰⁰¹). Across a broad range of plant species, these patterns are well‐documented and categorized into three classes: Class I (12–14 introns), Class II (9 introns) and Class III (6 introns) (Trapp & Croteau ²⁰⁰¹). Notably, Class II TPS genes are specific to gymnosperms, meaning that only Class I and Class III TPS genes are present in the angiosperm species analysed in this study. To identify putative TPS genes, we generated histograms depicting the intron count for each gene within each species. The majority of genes exhibited exon‐intron patterns consistent with either Class I or Class III (Fig. S1). Based on this distribution, we selected genes containing more than five exons.

Furthermore, some genes remained too short for detailed analysis. To address this, genes encoding proteins shorter than 400 amino acids were excluded (Fig. S2). To classify the subgroups of TPS genes, we incorporated protein sequences of well‐characterized TPS genes from S. lycopersicum and A. thaliana as references for phylogenetic analysis. These reference sequences were obtained from Table S7 in Zhou and Pichersky (Zhou & Pichersky ²⁰²⁰).

Additionally, IspS from Q. serrata (QsIspS) and IspS‐like from Q. serrata, Q. glauca and L. edulis (QsTPS9, QgIspS‐like, LeIspS‐like) were obtained from previous study (Koita et al. ²⁰²⁶) and included in the analysis. To root the phylogenetic tree, we included the ent‐copalyl diphosphate synthase sequence from Picea sitchensis as an outgroup.

Phylogenetic analysis of TPS genes

Alignment and phylogenetic tree construction were conducted using the build function of ETE3 v3.1.3 (https://etetoolkit.org). Sequence alignment was performed with MAFFT v6.861b (https://mafft.cbrc.jp), alignment trimming with trimAl v1.4.rev (https://vicfero.github.io/trimal) and phylogenetic tree construction with RAxML v8.2.11 (https://cme.h‐its.org/exelixis/web/software/raxml) using the GAMMA JTT model and 500 bootstrap replications. The resulting phylogenetic tree was visualized using the online tool iTOL (https://itol.embl.de). The analysed TPS genes were classified into five subclade groups: TPS‐a, TPS‐b, TPS‐c, TPS‐e/f and TPS‐g, based on reference genes from Arabidopsis and tomato. The chromosomal locations of these classified genes were visualized using TBtools‐II v2.2.10 (https://github.com/CJ‐Chen/TBtools‐II).

To investigate molecular evolution of IspS in Fagaceae, we reconstructed a phylogenetic tree containing only the TPS‐b subclade to which IspS belongs. We used published IspS sequences from S. lycopersicum (Zhou & Pichersky ²⁰²⁰), Q. serrata (Koita et al. ²⁰²⁶), P. alba (AB198180) (Sasaki et al. ²⁰⁰⁵) and G. max (Mostofa et al. ²⁰²⁵). The ocimene synthase close to IspS in V. vinifera (Martin et al. ²⁰¹⁰) was also included. We aligned CDS sequences of each gene using MAFFT v7.520 (https://mafft.cbrc.jp), trimmed alignment gap with trimAl v1.5.rev0 (https://vicfero.github.io/trimal) and determined best‐fit substitution model using ModelTest‐NG v0.1.7 (https://github.com/ddarriba/modeltest). Phylogenetic tree construction was performed with RAxML‐NG v0.9.0 (https://github.com/amkozlov/raxml‐ng) under the GTR + I + G4 substitution model with 500 bootstrap replicates. The resulting tree was visualized using the online tool iTOL (https://itol.embl.de). Since an adjacent subclade was located near the subclades containing previously reported IspS genes, we focused on the three subclades that include 44 Fagaceae genes.

To calculate an isoprene score (Sharkey et al. ²⁰¹³), the protein sequences of genes belonged to these three subclades were aligned by MAFFT v7.520 (https://mafft.cbrc.jp) with five previously known IspS genes ‐one gene from S. lycopersicum, Q. serrata and P. alba, two genes from G. max‐ and one ocimene synthase from V. vinifera.

Sample collection for RNA sequencing

To evaluate expression of TPS genes in natural conditions, we analysed transcriptome data obtained from leaf and bud tissues in three Fagaceae species, Q. acutissima, Q. glauca and L. edulis. We generated data for Q. acutissima, whereas those for Q. glauca and L. edulis were obtained from the previous study (Kudo et al. ²⁰²⁵). For all data, samples were collected every four weeks spanning from 2020–2021 in Q. acutissima, 2022–2023 in Q. glauca and 2019–2021 in L. edulis in Fukuoka prefectures in Japan. For each species, we targeted three individuals and each sample was collected from three branches per individual, except for L. edulis. Given the consistent expression patterns observed among individuals in a previous study (Satake et al. ²⁰²³), we focused on only one individual for L. edulis. Study sites for Q. acutissima and Q. glauca is the biodiversity reserve on the Ito campus of Kyushu University (33°35′ N, 130°12′ E, 20 to 57 m a.s.l). L. edulis were studied at the Imajyuku Field Activity Center (33°33′ N, 130°16′ E, 84 to 111 m a.s.l). Leaf samples were collected from spring to summer in the deciduous species Q. acutissima, whereas in the evergreen species Q. glauca and L. edulis, sampling was conducted throughout the year.

RNA‐seq samples were collected and preserved in 2 mL microtubes containing 1.5 mL of RNA‐stabilizing reagent (RNAlater; Thermo Fisher Scientific, Waltham, MA, USA), then stored at −80°C until RNA extraction, following the protocol described by Kudo et al. (2025).

RNA extraction and RNA‐seq analysis

Total RNA was extracted from leaf samples following the protocol described previously (Miyazaki et al. ²⁰¹⁴). RNA integrity was assessed using the method outlined previously (Kudo et al. ²⁰²⁵). For each sample, 5–6 μg of total RNA was sent to Hangzhou Veritas Genetics Medical Institute Co., Ltd., where cDNA libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina. Transcriptome sequencing (150 bp paired‐end reads) was performed on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA). Gene expression quantification followed a three‐step pipeline: (i) quality filtering was performed using fastp v0.20.0 with default settings (https://github.com/OpenGene/fastp); (ii) RNA‐seq reads were aligned to the reference genome using TopHat v2.1.1 (https://ccb.jhu.edu/software/tophat/index.shtml); and (iii) transcript abundance was quantified using StringTie v1.3.5 (https://ccb.jhu.edu/software/stringtie).

Seasonal expression patterns of TPS genes in Fagaceae

To investigate the seasonal expression patterns of TPS genes, we extracted TPS gene transcripts from RNA‐seq data of Q. acutissima, Q. glauca and L. edulis. For each gene, we calculated the mean expression level in transcripts per million (TPM) across three individuals, except for L. edulis, for which data were available from only one individual (Kudo et al. ²⁰²⁵). TPM values were transformed using log₂(TPM + 1). Genes with a mean TPM >1 across all samples were considered to be expressed. A heatmap was generated in R ver. 4.3.1 using the ‘pheatmap’ package, with gene order arranged according to TPS clades (Fig. 6A). In addition to these three species, we also obtained expression data for IspS and IspS‐like in Q. serrata (Koita et al. ²⁰²⁶), which were used to plot seasonal changes in expression levels. To generate the heatmap, we calculated the average gene expression levels for each month by aggregating data from different individuals over a two‐year period for each species.

Fig. 6.

Seasonal expression patterns of TPS genes in leaves. (A) Heatmap showing the seasonal expression patterns of TPS genes in three Fagaceae species: Q. acutissima (blue), Q. glauca (light blue) and L. edulis (green). TPS‐c genes are not included in the heatmap, as no TPS‐c genes were consistently expressed in all three species. Grey squares indicate missing data due to leaf shedding. Genes were considered expressed if the mean of TPM values across all observation periods was ≥1. On average, 36% of TPS genes were expressed in leaves. Within the TPS‐b clade, 42% of genes were expressed in leaves, including 6 out of 14 in Q. acutissima, 10 out of 28 in Q. glauca and 2 out of 24 in L. edulis. Genes showing peaked expression during autumn to winter are indicated by black arrows. (B–D) Seasonal expression patterns of TPS genes within the subclades TPS‐b‐t1, TPS‐b‐t2, and TPS‐b‐t3. In addition to the three species mentioned above, data from Q. serrata were incorporated from a previous study (Koita et al. ²⁰²⁶). A red star indicates Q. serrata isoprene synthase (QsIspS).

Ancestral amino acid sequence inference

To infer the ancestral amino acid sequences of putative IspS in Quercus species, we analysed 95 representative nucleotide sequences, which have no ambiguity, of TPS‐b subfamily members from Fagales. These sequences include genes identified through the genomic survey of eight Fagaceae species described above, genes from Q. serrata and additional genes retrieved via BLAST searches against the database. The BLAST search identified genes from Q. suber, L. litseifolius, Castanea mollissima, Juglans regia, J. maicrocarpa x J. regida, Carya illinoinensis, Corylus avellana and Alnus glutinosa, resulting in a dataset comprising genes from 17 Fagales species (Table S3). Additionally, seven TPS‐g subfamily genes were used as an outgroup.

Sequence alignment was performed using MACSE v2 (https://www.agap‐ge2pop.org/macse), which can account for the underlying codon structure of protein‐coding sequences and generate both nucleotide (NT) and amino acid (AA) alignments. Phylogenetic reconstruction and ancestral amino acid sequence inference were conducted separately for NT and AA alignments. ML phylogenetic trees were constructed with RAxML v8.2.11 (https://cme.h‐its.org/exelixis/web/software/raxml) with 1000 bootstrap replications. The GAMMA GTR model (GTR + I + G) and the GAMMA JTT model (JTT + I + G) were used for the NT and AA alignment, respectively. The resulting ML trees exhibited compatible topology, particularly the splits of the major clusters. Two strongly supported clusters of IspS with the maximum score were identified in both trees. Ancestral amino acid sequences were inferred at each internal node using codeml in PAML (https://github.com/abacus‐gene/paml) with the F3x4 codon model for the NT alignment. We also employed GRaSP (https://github.com/charles‐abreu/GRaSP) to infer the ancestral sequences from the AA alignment and estimate the location of gaps in the ancestral sequences.

Detection of positive selection

To investigate whether amino acid sites in putative IspS (foreground) have undergone positive selection compared with the other TPS‐b subfamily members (background), we applied the branch‐site test for positive selection (Zhang et al. ²⁰⁰⁵). This approach identifies amino acid sites where the non‐synonymous to synonymous substitution rate ratio (ω) exceeds one, indicative of positive selection. The NT alignment and MLtree used for the ancestral sequence inference were employed in this analysis. Alignment gaps were removed prior to the analyses. Previous studies have shown that gene conversion and recombination events could lead to false positives (Anisimova et al. ²⁰⁰³; Casola & Hahn ²⁰⁰⁹). To account for this, we applied the GARD method implemented in HYPHY (https://hyphy.org) to detect potential gene conversion or recombination events. No significant evidence of gene conversion or recombination was found in the genes, so subsequent analyses were conducted using the full‐length alignments.

All branches within two clades of the genes with the isoprene score of 4 were set as foreground branches and the branch‐site tests were conducted using codeml (Yang 2007). The null model was set to model = 2, NSsites = 2, fix_omega = 1 and omega = 1, and the alternative model (Model A) was with model = 2, NSsites = 2 and fix_omega = 0 (to estimate ω₂). The Bayes empirical Bayes (BEB) method was used to identify sites potentially under positive selection on the foreground branches.

Materials and reagents for in vivo enzymatic characterization

DMAPP and GPP were used as substrates, with DMAPP purchased from TargetMol (Boston, USA) and GPP from Sigma‐Aldrich (St. Louis, USA). As standard compounds, isoprene was obtained from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan), and β‐ocimene was purchased from Cayman Chemical (Ann Arbor, USA).

Construction of E. coli expression vectors

To express the IspS sequences from the extant isoprene‐emitting oak species Q. serrata (QsIspS, previously characterized by Koita et al. ²⁰²⁶) and Quercus robur (TPS‐b‐t2‐Qro1, identified in this study), as well as the ancestral sequences estimated at nodes N5, N6, N7 and N8 by PAML (denoted as N5–N8_paml) and GRaSP (denoted as N5–N8_grasp) in E. coli, their putative transit peptides at the N‐terminus were removed and the truncated CDSs were further codon‐optimized for E. coli. The optimized sequences were synthesized by GeneArt (Thermo Fisher Scientific). These sequences were amplified by PCR using KOD One® (Toyobo, Osaka, Japan) with the primer pairs shown in Table S5. The amplification program is composed of an initial denaturation at 94°C for 1 min followed by 40 cycles of denaturation at 98°C for 10 s, annealing at 51°C for 5 s and extension at 68°C for 10 s, with a final extension at 68°C for 5 min. The amplicons were individually inserted into the pET22b(+) vector (Novagen, Madison, US) by In‐Fusion reaction (Takara Bio, Kusatsu, Japan) using NdeI and BamHI sites, which enable the production of proteins with the histidine tag (6 × His). The sequence of N8_paml was constructed by introducing a single mutation to N8_grasp by PCR using KOD One® (Toyobo) and the primer pair N8_paml_Fw and N8_paml_Rv. This amplification was carried out with an initial denaturation at 94°C for 1 min, followed by 40 cycles of denaturation at 98°C for 10 s, annealing at 51°C for 5 s and extension at 68°C for 40 s, with a final extension at 68°C for 5 min. The amplicon was introduced into the pET22b(+) vector in the same process as described above.

Enzymatic characterization

Constructs ΔTP‐QsIspS‐6 × His, ΔTP‐ TPS‐b‐t2‐Qro1‐6 × His, ΔTP‐N5_grasp‐6 × His, ΔTP‐N5_paml‐6 × His, ΔTP‐N6_grasp/paml‐6 × His, ΔTP‐N7_paml‐6 × His, ΔTP‐N7_grasp‐6 × His, ΔTP‐N8_paml‐6 × His and ΔTP‐N8_grasp‐6 × His were individually introduced into E. coli BL21 (DE3) for recombinant protein production. Preparation of recombinant proteins and protein quantification were performed as previously described (Koita et al. ²⁰²⁶). A standard reaction mixture (200 μL) contained 0.5 mM of either DMAPP or GPP as the substrate, 20 mM MgCl₂ and 160 μL of the purified enzyme. For assays using DMAPP, purified enzymes were applied at concentrations of 170–198 μg mL⁻¹, whereas for assays using GPP, concentrations of 8.7–10.5 μg mL⁻¹ were used. As exceptions, QsIspS was assayed with DMAPP at 92.8 μg mL⁻¹ and with GPP at 94.8 μg mL⁻¹; additionally, ΔTP‐b22SC4_grasp/paml was assayed with GPP at 187 μg mL⁻¹. Reaction mixtures were incubated in 2‐mL glass vials (Thermo Fisher Scientific) at 37 °C for 30 min, except for QsIspS, which was incubated for 1 h. GC–MS analysis of the enzymatic products was carried out as described in Koita et al. (2026).

RESULTS

Identification of TPS genes in Fagaceae

To identify TPS genes in Fagaceae, we first obtained genome sequences of eight Fagaceae trees from four genera using public databases (Table S1). These included five species from the genus Quercus (Q. robur, Q. lobata, Q. rubra, Q. acutissima and Q. glauca), and one species each from the genera Lithocarpus (L. edulis), Castanopsis (C. hystrix) and Fagus (F. crenata) (Fig. 1; Table S1). Isoprene emission factors vary significantly among these eight species (Fig. 1; Table S2), highlighting their suitability for studying the evolution of isoprene emission.

Fig. 1.

Phylogenetic relationships and isoprene emission traits of Fagaceae species analysed in this study. The phylogenetic tree was constructed based on whole‐genome sequences from eight Fagaceae species, with bootstrap values shown at each branch. The isoprene emission traits assigned to each species were determined based on the isoprene emission factors summarized in Table S2.

To identify TPS genes in our target Fagaceae species, we conducted HMMER searches using the annotated protein sequences from the genomes of eight Fagaceae species against the metal‐binding domain (PF03936) and the N‐terminal domain (PF01397). Each TPS candidate gene was further examined and filtered based on exon‐intron structure, including the number and locations of exons (Fig. S1), by comparing with previously reported TPSs (Trapp & Croteau ²⁰⁰¹), as well as total sequence length (Fig. S2).

The rooted TPS phylogenetic tree constructed from our TPS dataset along with sequences from Solanum lycopersicum (tomato) and Arabidopsis thaliana revealed a clear distinction among the TPS‐a, TPS‐b, TPS‐c, TPS‐e/f and TPS‐g subfamilies (Fig. 2A). TPS‐a members in Fagaceae formed a distinct clade, separate from those of tomato and A. thaliana, but showing a closer evolutionary relationship to tomato TPS‐a genes (Fig. 2A). The TPS‐c, TPS‐g and TPS‐e/f subfamilies were present in all Fagaceae species, except for the TPS‐e/f subfamily, which was absent in Q. acutissima (Fig. 2B).

Fig. 2.

Phylogenetic analysis and summary of TPS subclade genes. (A) Molecular phylogenetic tree of TPS genes in the eight Fagaceae species. TPS genes from Solanum lycopersicum and Arabidopsis thaliana were used as references to define five TPS subclades (TPS‐a, −b, −c, −e/f and ‐g). The three TPS‐b genes in tomato including isoprene synthase were shown by a red asterisk. The ent‐copalyl diphosphate (ent‐CPP) synthase gene from Picea sitchensis was used as the outgroup. The branches whose bootstrap values were 0.9 or more were shown by purple circles. (B) Number of TPS genes in each subclade of Fagaceae genomes. The heatmap shows the number of TPS genes assigned to each category, with warmer colours representing higher gene counts. TPS gene counts from A. thaliana and S. lycopersicum are also included for comparison.

The TPS‐b subfamily represents the largest gene group within Fagaceae species (Fig. 2A,B). The three TPS‐b genes in tomato form a distinct clade from other tomato and Arabidopsis TPS‐b genes (red asterisk in Fig. 2A). These three tomato genes are known to synthesize acyclic monoterpenes, with one of them, TPS47, encoding a cytosolic isoprene synthase (Zhou & Pichersky ²⁰²⁰).

Many genes in the TPS‐a and TPS‐b subfamilies are arranged in tandem gene arrays on chromosomes 8 and 9, respectively (Fig. 3A), suggesting their origin from tandem gene duplication events. Overall, the genomic locations of these genes were largely conserved across Quercus, Lithocarpus and Castanopsis (Fig. 3B). However, in Fagus, significant genome rearrangements were observed, indicating a distinct evolutionary trajectory (Fig. 3B).

Fig. 3.

Genomic distribution of TPS genes in eight Fagaceae species. (A) Chromosomal distribution of TPS genes in the twelve chromosomes of Q. robur. The IspS candidate genes with the highest isoprene score (4) are marked by black stars. (B) Collinearity plot of TPS genes across the twelve chromosomes of eight Fagaceae species. Grey vertical lines connect other orthologous genes across the eight species. Only connections between chromosome pairs with significantly enriched conservation of synteny are shown. Coloured inverted triangles indicate the locations of TPS genes classified into different subclades. Diamonds mark regions where multiple TPS genes (two or more) are clustered. Coloured lines represent collinear genomic blocks between species that include TPS genes, highlighting conserved syntenic relationships.

Isoprene score of TPS‐b genes in Fagaceae

To investigate the evolutionary history of IspS in Fagaceae, we focused on the TPS‐b subfamily, which includes the functionally characterized IspS from tomato. We also retained previously characterized IspS and ocimene synthase from three non‐Fagaceae species—Populus alba (Sasaki et al. ²⁰⁰⁵), Vitis vinifera (Martin et al. ²⁰¹⁰) and Glycine max (Mostofa et al. ²⁰²⁵)—as well as from one Quercus species (Q. serrata; Koita et al. ²⁰²⁶) and included these sequences in subsequent analyses. Based on the phylogeny of TPS‐b subfamily including these sequences, we designated the clade containing the tomato IspS as TPS‐b‐target1 (TPS‐b‐t1) and the clade containing IspS of Q. serrata (QsIspS) and P. alba (PaIspS) as TPS‐b‐target3 (TPS‐b‐t3) (Fig. 4). Between the TPS‐b‐t1 and TPS‐b‐t3, we identified a distinct clade, which we designate as TPS‐b‐target2 (TPS‐b‐t2) (Fig. 4). The IspS from G. max clustered with ocimene synthase from V. vinifera, outside the TPS‐b‐t2 and TPS‐b‐t3 clades. Both TPS‐b‐t2 and TPS‐b‐t3 consist predominantly of genes from Fagaceae, suggesting that these clades are potentially involved in the evolution of IspS in Fagaceae.

Fig. 4.

Phylogenetic tree of TPS genes in the TPS‐b Subclade. Molecular phylogenetic tree of TPS genes classified within the TPS‐b subclade. The phylogeny was inferred with RAxML‐NG v.0.9.0 under the GTR + I + G4 substitution model, using 500 bootstrap replicates to assess node support. A TPS‐g gene from F. crenata (TPS‐g‐Fc1) was used as an outgroup. Three previously reported isoprene synthase (IspS) from species other than Fagaceae were added for comparison, one gene from Populus alba (PaIspS: Sasaki et al. ²⁰⁰⁵) and two genes from G. max (GmTPS8 and GmTPS23: Mostofa et al. ²⁰²⁵). The ocimene synthase close to IspS in V. vinifera (VvCSbOci: Martin et al. ²⁰¹⁰) was also included. Five previously reported IspS with high isoprene synthase activity—score 3 in S. lycopersicum and score 4 in Q. serrata, P. alba, G. max—are marked with red stars. The IspS candidate genes from Fagaceae species with the highest isoprene score (4) are marked by black stars. The 107 terminal nodes composed of genes from Fagaceae species are collapsed into a triangle.

Sequences of genes included in these three clades, together with the previously characterized IspS and ocimene synthase, were aligned and used to calculate an isoprene score (Fig. 5). The isoprene score serves as a useful screening tool to identify enzymes with exclusive IspS activity based on sequence information and the structural analysis of IspS (Sharkey et al. ²⁰¹³). This score is calculated using the four conserved amino acid residues (known as the diagnostic tetrad) that are characteristic of IspS. Two of the residues (F338 and F485 in Fig. 5) were suggested to restrict the size of the substrate‐binding pocket and play a key role in determining enzyme specificity (Sharkey et al. ²⁰¹³).

Fig. 5.

The multiple sequence alignment with isoprene scores. Multiple sequence alignment of TPS genes from the three TPS‐b subclades: TPS‐b‐t1, TPS‐b‐t2 and TPS‐b‐t3. Four conserved amino acid positions used to compute isoprene scores are highlighted by red and yellow shading; red means conserved amino acid at this position and yellow means another. The previously reported isoprene synthases (IspS) are indicated by red stars, and Fagaceae candidates with the highest isoprene score (4) in this study are marked with black stars.

While F338 and F485 of the diagnostic tetrad residues are considered highly specific to IspS, the other positions are considered redundant (Li et al. ²⁰¹⁷) with valine (V446) found in place of serine in some angiosperm families (Li et al. ²⁰¹⁷). Therefore, we treated both serine and valine as conserved residues at this position and denoted it as S/V446. The isoprene score ranges from 0 to 4, with a score of 4 representing full conservation of the diagnostic tetrad and thus a high likelihood of exclusive IspS activity. Indeed, the isoprene scores of previously characterized IspS from P. alba, G. max and Q. serrata were the maximum value of four, whereas that of tomato IspS was three (Fig. 5). The multiple sequence alignment revealed that the isoprene score for Fagaceae sequences in the TPS‐b‐t1 clade was consistently one (Fig. 5). In the TPS‐b‐t2 and TPS‐b‐t3 clades, sequences from the species in the subgenus Quercus (Q. robur, Q. lobata, Q. rubra and Q. serrata) were found to have an isoprene score of 4 (indicated as red or black stars in Fig. 5). Specifically, Q. robur contained three, Q. lobata had two and Q. rubra had one sequence with an isoprene score of 4 (Fig. 5). In contrast, the maximum isoprene score for sequences from Lithocarpus, Castanopsis and Fagus was only three, with a lack of F338, an amino acid considered crucial for isoprene synthase activity (Fig. 5; Sharkey et al. ²⁰¹³). These patterns in isoprene score align with isoprene emission capacities, where Q. robur, Q. lobata, Q. rubra and Q. serrata are known to exhibit high isoprene emission, while other species show low or no emission (Fig. 1; Table S2).

Within the twin clades TPS‐b‐t2 and TPS‐b‐t3, Fagus sequences were found only in TPS‐b‐t2, whereas sequences from other species were present in both clades (Fig. 5). Most genes within TPS‐b‐t2 and TPS‐b‐t3 were located on chromosome 8, with the exception of F. crenata (Fig. 3B; Fig. S3). Interestingly, two genes in Q. robur and one in Q. lobata, all with an isoprene score of 4, were located on chromosome 3 (Fig. 3B; Fig. S3), suggesting chromosomal translocation events.

Seasonal expression patterns of TPS genes

To investigate the magnitude and seasonality of TPS gene expression, we compared seasonal gene expression patterns across four Fagaceae species grown under the same study site in Fukuoka, Japan. The dataset comprises two deciduous species, Q. serrata and Q. acutissima, and two evergreen species, Q. glauca and L. edulis. For the isoprene‐emitting species, we used previously reported transcriptome data from Q. serrata (Koita et al. ²⁰²⁶). For the isoprene low‐emitting species, we newly generated transcriptome data from leaf and bud tissues of Q. acutissima, collected monthly over a two‐year period. In addition, we incorporated previously published transcriptome datasets from two other isoprene low‐emitting and non‐emitting species, Q. glauca and L. edulis, which were sampled over a two‐year period at the same study site as Q. serrata and Q. acutissima (Kudo et al. ²⁰²⁵).

Seasonal expression profiles in leaves revealed that most expressed genes exhibited peak expression in summer, although a small fraction of genes showed peaked expression during autumn–winter (e.g. Qa12 and Qg1 in the TPS‐a clade highlighted as black arrows in Fig. 6A). Notably, expression patterns differed markedly among the three target TPS‐b subclades (Fig. 6B–D). In leaves, only genes in the TPS‐b‐t3 clade showed clear and consistent summer expression peaks (Fig. 6D), whereas expression levels in the TPS‐b‐t1 and TPS‐b‐t2 clades were low in all species (Fig. 6B,C). These results suggest that TPS‐b‐t3 genes are primarily involved in terpenoid biosynthesis in leaves during summer. Furthermore, the expression pattern of Q. serrata IspS (QsIspS) within the TPS‐b‐t3 clade also showed strong expression (Fig. 6C,D), providing additional support for the functional role of this clade in isoprene biosynthesis. In contrast, in bud tissues, expression levels of genes within the TPS‐b‐t3 clade were consistently low, and genes in the other two clades, TPS‐b‐t1 and t2 were relatively high (Fig. S4B–D), implying a tissue‐specific regulation of TPS gene expression.

Convergent evolution and positive selection in IspS candidates of Fagaceae

To determine whether the genes with an isoprene score of 4 were gained or lost during evolution, we employed the ancestral sequence reconstruction method. This approach relies on statistical principles to infer the sequences of putative ancestral genes based on the amino acid or nucleotide sequences of extant species. For this analysis, we used nucleotide sequences from 17 Fagales species belonging to the TPS‐b clade (Table S3) and reconstructed putative ancestral sequences using codeml implemented in PAML (https://github.com/abacus‐gene/paml).

The estimated ancestral sequences of the last common node of the three target clades (node N2 in Fig. 7A) had an isoprene score of 2, with the conserved residuals S/V446 and F485 (Fig. 7B). Subsequently, N505 was gained in the common ancestor of the TPS‐b‐t2 and TPS‐b‐t3 clades (node N4 in Fig. 7A,B), prior to their duplication. Following this duplication, F338 was acquired independently in each of the TPS‐b‐t2 and TPS‐b‐t3 clades (nodes N6 and 8 in Fig. 7A,B), specifically in the lineage leading to the three species within subgenus Quercus including Q. robur, Q. lobata and Q. rubra.

Fig. 7.

Results of ancestral sequence reconstruction of Fagaceae TPS genes using PAML. (A) Maximum likelihood phylogeny of the TPS gene set used for ancestral reconstructions. The phylogeny was inferred with RAxML v8.2.11 under the GAMMA GTR substitution model, using 1,000 bootstrap replicates to assess node support. Internal nodes N1–N9 mark the bifurcation points analysed. Branch colours denote the inferred isoprene score—blue (2), purple (3) and red (4). Black stars indicate Fagaceae IspS candidates, and a red star marks Q. serrata IspS (QsIspS). The three focal clades (TPS‐b‐t1, TPS‐b‐t2 and TPS‐b‐t3) are highlighted in distinct colours for ease of reference. (B) Inferred amino acid states at the diagnostic tetrad and the corresponding isoprene score for each ancestral node (N1–N9). The posterior probability (PP) of the most likely codon/amino acid state was as follows: N1 – S/V446: GCA (Ala, PP = 0.263) → GTA (Val, PP = 0.600); F485: CTT (Leu, PP = 0.523) → TTT (Phe, PP = 0.405). N4 – T505: AAA (Lys, PP = 0.774) → AAT (Asn, PP = 0.956). N6 – F338: ACT (Thr, PP = 0.999) → TTT (Phe, PP = 0.970). N8 – F338: ACT (Thr, PP = 0.996) → TTT (Phe, PP = 0.990). (C) and (D) Enzyme assays of the ancestral sequence products using DMAPP and GPP as substrates. The clades of each sequence are highlighted by the same colours as in (A). The reaction products were detected at m/z 67 for isoprene (C) and m/z 93 for monoterpenes (D) as the substrate. The cis or trans configurations of β‐ocimene were identified based on spectral annotation. The chromatograms, except for the β‐ocimene standard, are plotted on the same scale. Asterisks represent peaks of impurities. [Correction added on 06 July 2026, after first online publication: Figure 7 has been updated in this version.]

The gain of F338 occurred independently in both the TPS‐b‐t2 and the TPS‐b‐t3 clades, despite requiring two nucleotide substitutions for the transition from ACT (T) to TTT (F). To assess the robustness of this inference to the choice of ancestral sequence reconstruction method, we additionally employed GRaSP (https://github.com/charles‐abreu/GRaSP) to infer the ancestral sequences and reached the same conclusion that the gain of F338 occurred independently in both the TPS‐b‐t2 and TPS‐b‐t3 clades (Fig. S5A,B). Although the inferred ancestral sequences for the distant ancestor at nodes N1–N3 differed slightly between PAML and GRaSP (Data S1), the evolutionary pattern across the remaining nodes was consistent (Fig. S5A,B). This concordance indicates that our inference is robust to methodological differences and provides increased confidence in the reconstructed evolutionary history. Overall, these results suggest that the acquisition of isoprene‐producing function occurred through the stepwise accumulation of mutations, potentially driven by convergent molecular evolution in the TPS‐b‐t2 and TPS‐b‐t3 clades.

We next tested for evidence of positive selection in the clades that acquired F338 (predefined as foreground branches) by applying the branch‐site model in codeml (Zhang et al. ²⁰⁰⁵; Yang ²⁰⁰⁷). The analysis revealed significant evidence of positive selection in the foreground branches (likelihood ratio test, P = 1.027 × 10⁻⁶) with four residuals identified as candidates under positive selection (Bayes Empirical Bayes posterior probability >0.95) (Table S4; Fig. S7). Notably, the residue corresponding to F338 was included among these candidates, suggesting that the convergent gain of this residue in the TPS‐b‐t2 and TPS‐b‐t3 clades was likely driven by positive selection.

In vitro enzyme assays demonstrated that the IspS evolved from monoterpene synthase

To evaluate the evolutionary trajectory of enzymatic function, ancestral TPS sequences reconstructed using PAML and GRaSP were artificially synthesized (Data S1) and subjected to enzymatic characterization using an Escherichia coli expression system. Because genes within the TPS‐b‐t3 clade exhibited high expression levels, we first focused our analyses on this clade. Enzyme assays using purified recombinant proteins were conducted with either DMAPP (for IspS activity) or GPP (for mono‐TPS activity) as a substrate. Based on the PAML reconstructions, the ancestral enzyme at node N7 exhibited no IspS activity but showed mono‐TPS activity, predominantly producing trans‐β‐ocimene (Fig. 7C,D). In contrast, the enzyme at node N8 exhibited IspS activity only (Fig. 7C,D). Importantly, ancestral proteins reconstructed using both PAML and GRaSP displayed identical enzymatic activities at nodes N7 and N8 (Fig. 7C,D; Fig. S5C,D), despite minor differences in the inferred amino acid sequences between the two methods (with 6 and 1 amino acid differences at nodes N7 and N8, respectively). We further confirmed that QsIspS, the IspS sequence from the extant isoprene‐emitting species Q. serrata (Koita et al. ²⁰²⁶), exhibited IspS activity exclusively, consistent with the activity of the ancestral sequence at node N8 (Fig. 7C,D). These results indicate that IspS with an isoprene score of 4 in the clade TPS‐b‐t3 evolved from β‐ocimene synthases, supporting the functional shift during evolution.

We also performed the same analyses for the ancestral sequences of the TPS‐b‐t2 clade at nodes N5 and N6. The ancestral enzyme at node N5 exhibited both IspS and mono‐TPS activities, producing trans‐ and cis‐β‐ocimene, whereas no detectable enzymatic activity was observed for the enzyme at node N6 (Fig. 7C,D). The lack of activity at node N6 was attributable to low production levels of the recombinant protein in E. coli (Fig. S6). Because Q. serrata does not possess an IspS within the TPS‐b‐t2 clade, we instead examined IspS candidate sequence from Q. robur (TPS‐b‐t2‐Qro1), a present‐day enzyme downstream of node N6, and attempted its biochemical characterization. However, insufficient recombinant protein production similarly precluded functional assessment (Fig. S6). Consequently, it remains unclear whether proteins within the N6 clade specifically acquired IspS activity through the gain of residue F338.

DISCUSSION

Using comparative genomics, ancestral sequence reconstruction, and in vitro assays of ancestral enzymes, we found that IspS with an isoprene score of 4 arose independently twice in the subgenus Quercus—once in the TPS‐b‐t3 clade and once in the TPS‐b‐t2 clade—within the TPS‐b subfamily in Fagaceae. Because the subgenus Quercus diverged from its common ancestor roughly 56 million years ago (Zhou et al. ²⁰²²), we infer that the onset of substantial isoprene emission in Fagaceae to the Eocene. Moreover, in all species we examined (Q. robur, Q. lobata and Q. rubra), once the IspS activity in the TPS‐b‐t3 clade evolved, it has been retained. This contrasts with rosids, where a single early gain of isoprene synthase was followed by multiple independent losses (Shang et al. ²⁰²⁴). Fagaceae therefore provide a complementary pattern—multiple, convergent gains coupled with long‐term retention—highlighting diverse evolutionary routes to isoprene emission in angiosperms.

The early TPS‐b‐t3 proteins preferred GPP and mainly produced β‐ocimene, whereas their descendants evolved to favour DMAPP and became dedicated isoprene synthases in the subgenus Quercus. The shift would correlate with pocket‐narrowing substitutions—most notably the substitution from Thr to Phe at the 338 residue—that exclude the larger GPP substrate. These results confirmed the previous view in which the ocimene synthases are ancestral TPS for which IspS evolved by active site size reduction following acquisition of F338 (Gray et al. ²⁰¹¹). The combination of branch‐level and site‐level likelihood tests detected positive selection at this residue, indicating adaptive convergence towards isoprene emission. We also identified three additional residues under positive selection, suggesting that other sites may contribute to isoprene emission, although the effects of these sites on substrate specificity are currently unknown. Isoprene provides a threefold heat shield—membrane stabilization, ROS scavenging and Ca²⁺‐mediated heat‐shock signalling (Zuo et al. ²⁰²⁵)—mechanisms that likely protect photosynthesis and sustain growth in Quercus species with thin, deciduous leaves under rising temperatures. Elucidating the adaptive significance of these protective effects, and their direct relationship to variation in isoprene emission capacity among species and populations, will be an important direction for future research.

Our results also suggest that ancestral sequences reconstructed for TPS‐b2‐t2 clade may represent bifunctional enzymes with latent isoprene synthase activity. Although these proteins could possess the structural capacity to catalyse isoprene formation, their overall physiological contribution is likely limited by low expression levels. Such genes in the TPS‐b2‐t2 clade may function as a backup or stress‐responsive reservoir of catalytic potential, becoming physiologically relevant only under specific environmental conditions or regulatory states. From an evolutionary perspective, this scenario is consistent with an intermediate stage in which bifunctional terpene synthases retain ancestral mono‐TPS activity while harbouring latent isoprene synthase capacity. Subsequent evolutionary shifts in expression regulation, rather than catalytic innovation alone, may have been required for the emergence of highly expressed, specialized IspS genes in the TPS‐b2‐t2 clade.

Terpenoids represent the largest and most diverse class of natural products in terrestrial plants (Chen et al. ²⁰¹¹). Comparative and functional genomics can chart the continuous evolutionary innovation of terpenoid‐biosynthetic pathways with unprecedented resolution. Such insight will deepen our understanding of how chemical diversity has shaped plant adaptation and diversification on the earth.

AUTHOR CONTRIBUTIONS

YI and AS designed research; YI, SNK, NT, SK, RM, KY, SI, HH, JK and AS performed research; YI, SNK, NT, SK, RM, KY, SI, HH, JK and AS analysed data; TT and NT provided genome data; YI and AS wrote the paper with inputs from all the authors.

FUNDING INFORMATION

This study was supported by Japan Society for the Promotion of Science (no. JP23H04965 to AS, KY, RM, SI and HH; JP23H04966 to AS, JP23H04967 to KY, RM).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Data S1. This is an image file showing multiple sequence alignment of ancestral sequences of Fagaceae TPS genes using PAML and GRaSP. The four conserved amino acid residues used for calculating isoprene score were shown by black frame.

Data S2. This is a fasta file of multiple sequence alignment of 102 nucleotide sequences used for ancestral sequence inference and detection of positively selected residues.

Data S3. This is a fasta file of multiple sequence alignment of 102 amino acid sequences used for ancestral sequence inference and detection of positively selected residues.

Data S4. This is a fasta file of codon‐optimized sequences used for E. coli expression.

Table S1. Summary statistics for the genomes of the eight analysed Fagaceae species.

Table S2. Isoprene emission factors reported in previous studies.

Table S3. List of the sequences used for ancestral amino acid sequence inference and detection of positively selected residues.

Table S4. Candidate sites under positive selection identified under branch‐site model.

Table S5. Primers used in this study.

Fig. S1. Distribution of intron numbers in TPS candidate genes across eight Fagaceae species.

Fig. S2. Histogram of amino acid lengths of TPS genes in eight Fagaceae species. Cells with diagonal lines represent all TPS genes, while filled cells indicate those retained after filtering out genes with low intron numbers (fewer than five). Genes shorter than 400 amino acids were excluded from subsequent analyses.

Fig. S3. Chromosomal locations of TPS‐b‐t2 and TPS‐b‐t3 subclade genes in Fagaceae species. In seven species (excluding Fagus crenata), all TPS‐b‐t2 genes were located on chromosome 8. In Quercus robur and Q. lobata, TPS‐b‐t2 genes were found on both chromosomes 8 and 3. In contrast, TPS‐b‐t2 genes in F. crenata were located on chromosome 9. A collinearity block linking the distal region of chromosome 8 in Q. robur (which harbours TPS‐b‐t2 genes) and chromosome 9 in F. crenata is shown in Fig. 3B, indicating a conserved genomic region. Coloured segments of arrows represent gene exons.

Fig. S4. Seasonal expression patterns of TPS genes in buds. (A) Heatmap showing the seasonal expression patterns of TPS genes in three Fagaceae species: Quercus acutissima (blue), Q. glauca (light blue), and Lithocarpus edulis (green). TPS‐c genes are not included in the heatmap, as no TPS‐c genes were consistently expressed in all three species. Genes were considered expressed if the mean of TPM values across all observation periods was ≥1. On average, 41% of TPS genes were expressed in buds. Genes showed peaked expression during autumn to winter are indicated by black arrows. (B–D) Seasonal expression patterns of TPS genes within the subclades TPS‐b‐t1, TPS‐b‐t2, and TPS‐b‐t3. In addition to the three species mentioned above, data from Q. serrata were incorporated from a previous study (Koita et al. ²⁰²⁶) and red star shows Q. serrata isoprene synthase (QsIspS).

Fig. S5. Results of ancestral sequence reconstruction of Fagaceae TPS genes using GRaSP. (A) Maximum‐likelihood phylogeny of the TPS gene set used for ancestral reconstructions. The phylogeny was inferred with RAxML v8.2.11 under the GAMMA GTR substitution model, using 1,000 bootstrap replicates to assess node support. Internal nodes N1–N9 mark the bifurcation points analysed. Branch colours denote the inferred isoprene score—blue (2), purple (3) and red (4). Black stars indicate Fagaceae IspS candidates, and a red star marks Q. serrata IspS (QsIspS). The three focal clades (TPS‐b‐t1, TPS‐b‐t2 and TPS‐b‐t3) are highlighted in distinct colours for ease of reference. (B) Inferred amino acid states at the diagnostic tetrad and the corresponding isoprene score for each ancestral node (N1–N9). Key amino acid changes were as follows: N4 – F485: Leu → Phe. N6—F338: Thr → Phe. N8—F338: Thr → Phe. (C) and (D) Enzyme assays of the ancestral sequence products using DMAPP and GPP as substrates. The clades of each sequence are highlighted by the same colours as in (A). The reaction products were detected at m/z 67 for isoprene (C) and m/z 93 for monoterpenes (D) as the substrate. The cis or trans configurations of β‐ocimene were identified as described (Koita et al. ²⁰²⁶). The chromatograms, except for the β‐ocimene standard, are plotted on the same scale. Asterisks represent peaks of impurities.

Fig. S6. SDS‐PAGE of ancestral TPS proteins reconstructed using PAML (N5‐N8 paml) and GRaSP (N5‐N8 grasp), and the extant Quercous robur IspS (TPS‐b‐t2‐Qro1) expressed in Escherichia coli.

Fig. S7. Multiple sequence alignment of TPS‐b and TPS‐g subfamily members from 17 Fagales species. Four candidate sites under positive selection are highlighted by red. The 245th site corresponds to F338, one conservative amino acid residue for isoprene synthases.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in DNA Data Bank of Japan at https://ddbj.nig.ac.jp/search/entry/bioproject/PRJDB35868, reference number PRJDB35868.