The chromosome-scale genome of Phoebe bournei reveals contrasting fates of terpene synthase (TPS)-a and TPS-b subfamilies

Abstract

Terpenoids, including aromatic volatile monoterpenoids and sesquiterpenoids, function in defense against pathogens and herbivores. Phoebe trees are remarkable for their scented wood and decay resistance. Unlike other Lauraceae species investigated to date, Phoebe species predominantly accumulate sesquiterpenoids instead of monoterpenoids. Limited genomic data restrict the elucidation of terpenoid variation and functions. Here, we present a chromosome-scale genome assembly of a Lauraceae tree, Phoebe bournei, and identify 72 full-length terpene synthase (TPS) genes. Genome-level comparison shows pervasive lineage-specific duplication and contraction of TPS subfamilies, which have contributed to the extreme terpenoid variation within Lauraceae species. Although the TPS-a and TPS-b subfamilies were both expanded via tandem duplication in P. bournei, more TPS-a copies were retained and constitutively expressed, whereas more TPS-b copies were lost. The TPS-a genes on chromosome 8 functionally diverged to synthesize eight highly accumulated sesquiterpenes in P. bournei. The essential oil of P. bournei and its main component, β-caryophyllene, exhibited antifungal activities against the three most widespread canker pathogens of trees. The TPS-a and TPS-b subfamilies have experienced contrasting fates over the evolution of P. bournei. The abundant sesquiterpenoids produced by TPS-a proteins contribute to the excellent pathogen resistance of P. bournei trees. Overall, this study sheds light on the evolution and adaptation of terpenoids in Lauraceae and provides valuable resources for boosting plant immunity against pathogens in various trees and crops.

Analysis of the chromosome-scale genome of Phoebe bournei reveals contrasting evolutionary fates of the TPS-a and TPS-b subfamilies. The high conservation ratio and expression level of TPS-a genes may contribute to sesquiterpenoid accumulation in P. bournei, which endows it with excellent pathogen resistance.

Introduction

Terpenoids are the largest and most diverse class of plant secondary metabolites, containing more than 25 000 compounds (McCaskill and Croteau, 1997). Some derivatives play roles in general physiology (e.g., gibberellin hormones and carotenoids), but most plant terpenoids are dedicated to mediating interorganismal interactions, stress defense, and environmental adaptations. For example, in vegetative organs, terpenoids, including monoterpenoids and sesquiterpenoids, act as toxins or repellents to herbivores or indirectly attract natural enemies of such herbivores (Martin et al., 2003; Gershenzon and Dudareva, 2007). For trees, terpenoids in the bark are the first barrier in the defense against stem borer attacks and subsequent pathomycete infections (Nerg et al., 2004; Della Rocca et al., 2020). Terpenoids are also of significant interest because of their extensive use in the pharmaceutical, cosmetic, agriculture, and food industries (Boutanaev et al., 2015). As the most chemically, structurally, and functionally diverse compounds in plants, terpenoids are primarily restricted to specific lineages or even a single species and are therefore referred to as specialized terpenoids. Lauraceae, Piperaceae, Lamiaceae, Apiaceae, Myrtaceae, Rutaceae, and Zingiberaceae are representative angiosperm families with abundant terpenoids (Ebadollahi and Sendi, 2015). Lauraceae species produce volatile organic compounds (VOCs) with high economic value in the spice and perfume industries. For instance, species in Lauraceae, such as May Chang tree (Litsea cubeba), bay laurel (Laurus nobilis), and camphor tree (Cinnamomum camphora), have distinctive scents that consist mainly of diverse monoterpenoids (Joshi et al., 2009). By contrast, Phoebe, a genus closely related to Cinnamomum, emits a scent dominated by sesquiterpenoids (Joshi et al., 2009; Xie et al., 2015; Ding et al., 2020).

Driven by selection pressure to adapt to their specific ecological niches, terpene synthase (TPS) proteins are the gatekeepers of terpenoid diversity. TPSs catalyze the cyclization and rearrangement of a few common prenyl pyrophosphate precursors (differing in the number of condensed five-carbon isoprenoid units) to generate a range of monoterpenoid (C10), sesquiterpenoid (C15), and diterpenoid (C20) scaffolds (Chen et al., 2011). In plants, the TPS family is divided into seven classes: TPS-a, TPS-b, and TPS-g, which are three angiosperm-specific subfamilies; TPS-d (gymnosperm specific); TPS-h (Selaginella spp. specific); TPS-e/f; and TPS-c. The TPS-a, TPS-b, and TPS-g subfamilies predominantly comprise synthases that catalyze the formation of monoterpenes and sesquiterpenes, whereas the TPS-c and TPS-e/f subfamilies mainly comprise diterpene synthases. Unlike the typically mid-sized diterpene synthases with 2–30 members, mono- and sesquiterpene synthases have undergone far more significant expansion (e.g., 69 members in grape (Martin et al., 2010) and 113 members in Eucalyptus (Kulheim et al., 2015)) that has resulted in diverse volatile terpenoids. Elucidation of the biological characteristics of TPS genes from an evolutionary perspective is essential for the bioengineering of metabolic pathways for specific terpenoids (Mewalal et al., 2017). Genome-wide studies of the TPS family have been conducted in the model plant Arabidopsis, the gymnosperms Picea spp. (Keeling et al., 2011), and multiple herbaceous angiosperm species (Boutanaev et al., 2015; Zhou and Pichersky, 2020a; Booth et al., 2020; Bornowski et al., 2020; Karunanithi et al., 2020). The possible molecular mechanisms and evolutionary history of terpenoids in perennial woody angiosperm plants, especially Lauraceae species, are largely unknown.

The broad-leaved evergreen tree genus Phoebe, which belongs to the Lauraceae family, comprises approximately 100 species in tropical and subtropical Asian forests (Wei and van der Werff, 2008). Phoebe timber has been widely used for high-quality furniture and buildings for centuries in Asia because of its pleasant scent, high durability, excellent decay resistance, and attractive appearance with a golden tint on the wood surface (Figure 1A). The preserved antique furniture and historic wooden buildings in the Forbidden City made from Phoebe are emblematic of these excellent wood properties. Phoebe bournei, a subtropical evergreen tree endemic to China (Figure 1A), is endangered because of the low survival rate of seedlings, intensive deforestation, and illegal logging for timber. Anti-hollow formation is highly valuable for structural wood production. In subtropical natural forests, hollow trunks are rarely observed in P. bournei trees, especially in ancient trees with a diameter of more than 100 cm (Xie et al., 2015).

Figure 1.

Characteristics of the Phoebe bournei genome.

(A) Photos of a Phoebe bournei tree. Its flowers and a piece of wood with a golden tint are shown on the lower left and right, respectively.

(B) Synteny and feature distribution of the P. bournei genome. Twelve chromosomes are ordered by size (Mb). Displayed tracks are gene density (I); repeat element coverage (II); noncoding RNA (III); gene expression in the seed (IV), bud (V), leaf (VI), stem (VII), and root (VIII); and GC content (IX). The inner plot presents syntenic blocks in the genome.

(C) Phylogenetic relationships of 18 seed plant species. The phylogenetic tree was constructed by the coalescence-based maximum likelihood method. The gymnosperm Gnetum montanum is used as an outgroup.

(D) The K_s distribution of orthologs and paralogs in five species from the magnoliids.

Here, we present the chromosome-scale genome assembly of P. bournei and use it to address three main questions: the evolution and divergence of TPSs underlying terpenoid diversity between P. bournei and other Lauraceae species, the mechanisms of sesquiterpenoid accumulation in P. bournei, and the possible contribution of sesquiterpenoids to P. bournei adaptation. The combination of evolutionary, transcriptomic, and metabolic analyses of the TPS family in P. bournei will shed light on the evolution of metabolic diversity and enable further investigations of terpenoid-driven resilience to environmental stresses.

Results

Assembly and characterization of the Phoebe genome

Phoebe bournei (Hemsl.) Yang is a diploid (2n = 24) with an estimated genome size of 1,078.25 ± 39.81 Mb/1C, as determined by flow cytometric analysis (Supplemental Figure 1). A genome survey of P. bournei based on a 17-mer analysis suggested an estimate of 941.77 Mb for the genome size and high heterozygosity of 1.4%. PacBio long reads (117.11 Gb, more than 100×) were used for assembly with the PacBio-preferred Falcon program and then polishing with Illumina short reads. Long-range mapping was performed to assemble the contigs onto scaffolds using 272.26 Gb of BioNano Saphyr optical mapping data, which resolved the contig conflicts and produced 463 scaffolds with a scaffold N50 of 5.64 Mb (Supplemental Table 1). Subsequently, a chromosome-level assembly was constructed with 350.62 Gb of Hi-C data. As a result, 1.057 Gb of 1.065 Gb total scaffold sequences were anchored onto 12 pseudochromosomes that represented 99.20% of the final chromosome-scale genome assembly with 85 scaffolds (scaffold N50 of 97.01 Mb) and a maximum pseudochromosome length of 137.96 Mb (Figure 1B, Supplemental Table 2, and Supplemental Figure 2). Estimation of genome completeness revealed 1266 (92.07%) complete benchmarking universal single-copy orthologs (BUSCOs) in the genome (Supplemental Table 3). In addition, 92.30–93.84% of RNA sequencing (RNA-seq) reads from the root, stem, leaf, bud, and seed were mapped to the genome (Supplemental Table 4). Repetitive elements accounted for approximately 48.10% of the P. bournei genome (Supplemental Tables 5 and 6). This percentage was lower than that previously reported (68.51%) (Chen et al., 2020b) but similar to that in Cinnamomum kanehirae (48%) (Chaw et al., 2019). Among these elements, long terminal repeats accounted for 26.34% of the P. bournei genome. Copia (4.73%) and Gypsy (8.69%), which represented half of the long terminal repeats in P. bournei, were found less often in the C. kanehirae genome (6.10% and 10.40%). Gene annotation based on the integration of three strategies suggested a final set of 30 096 protein-coding genes (Supplemental Table 7), of which 29 716 genes (98.74%) were annotated with several conventional databases (Supplemental Table 8). The completeness assessment showed 1261 (91.7%) BUSCOs in all annotated proteins (Supplemental Table 9).

Figure 6.

Functional characterization of P. bournei TPS-a genes.

(A) Chromatograms of GC–MS traces of products from transient overexpression of TPS-a genes in Nicotiana benthamiana leaves. The phylogenetic tree in the left panel presents the relationships of all TPS-a genes on chromosome 8 of P. bournei. Except for that of the pseudogene PbTPS-a24, noted with an asterisk, the chromatograms of the α-copaene and β-caryophyllene standards, the empty vector (control), and the TPSs are shown in the right panel. The peaks of sesquiterpenoid products are shown in blue. Mass spectra for all detected TPS products are given in Supplemental Figure 15. The products were: 1, δ-elemene; 2, α-cubebene; 3, ylangene; 4, α-copaene; 5, β-copaene; 6, β-caryophyllene; 7, γ-elemene; and 8, Guaia-6,9-diene.

(B) Corresponding carbon skeletons are shown. Monocyclic, bicyclic, and tricyclic sesquiterpenoids are shaded in pink, yellow, and blue, respectively.

A total of 211 single-copy orthologous genes from each of 18 seed plant species were used to analyze their phylogenetic relationships. The phylogenetic topologies derived from maximum likelihood (ML) and Bayesian trees were identical, in that magnoliids were a sister clade to eudicots rather than monocots (Figure 1C and Supplemental Figure 3). In Lauraceae, C. kanehirae had a close relationship with L. cubeba, and P. bournei was closely related to Persea americana. Analysis of the whole-genome duplication (WGD) history of P. bournei showed that the synonymous substitutions per synonymous site (K_s) distribution of paralogous gene pairs in P. bournei had two peaks for WGD events at approximately K_s = 0.57 and K_s = 0.88 (Figure 1D), consistent with the results of four-fold degenerate synonymous site analysis (Supplemental Figure 4). Two peaks at similar K_s values were also identified for C. kanehirae and P. americana (Chaw et al., 2019; Rendon-Anaya et al., 2019). The K_s distribution for intergenomic orthologs indicated that the more ancient WGD event occurred approximately 144.21–146.70 million years ago, before the divergence of Laurales and Magnoliales. The more recent WGD event shared by Lauraceae plants occurred 92.90–95.39 million years ago.

Variation in tissue-specific volatile terpenoid profiles of P. bournei

The effects of environment and population divergence on terpenoid biosynthesis could be significant. Therefore, to minimize sampling biases arising from interpopulation differences in P. bournei, we analyzed VOC profiles in a collection of 297 genotypes from 33 provenances covering the entire natural habitat that were grown in a common garden for six years (Supplemental Figure 5). The results showed that sesquiterpenoids accounted for at least 85% of the volatile compounds in leaf and stem samples of all collected P. bournei genotypes, whereas monoterpenoids accounted for less than 10% (Figure 2A).

Figure 2.

Profile of volatile terpenoids in P. bournei.

(A) Relative contents of volatile compounds in the leaves and stems of 297 genotypes.

(B) Relative contents of total monoterpenoids and total sesquiterpenoids in various tissues (biological replicates: n = 3, mean ± SD). Composition data of monoterpenoids from the root xylem are shown (top inset).

(C) Scatterplot of volatile terpenoids from various tissues based on OPLS-DA. Pink triangles and colored dots represent tissues and terpenoids, respectively.

(D) Relative contents of the 50 most abundant volatile compounds in various tissues. The dot size corresponds to the percentage of each compound compared with the most abundant product in the given tissues. Hexenol, monoterpenoids, and sesquiterpenoids are represented by pink, green, and blue dots, respectively. The sampled tissues were root xylem (RX), root bark (RB), stem xylem (SX), stem bark (SB), and leaves (L).

(E) Percentages of inhibition of mycelial growth by essential oil from P. bournei leaves and the β-caryophyllene standard. The three phytopathogenic fungi are Botryosphaeria dothidea, Colletotrichum gloeosporioides, and Lasiodiplodia pseudotheobromae.

To analyze the variety of terpenoids in P. bournei, we collected xylem and bark tissues from roots and stems, as well as leaves, from the genome-sequenced tree for further tissue-specific analysis of VOCs. There were 50 volatile compounds with a relative content above 0.5% in at least one of five analyzed P. bournei tissues, comprising 36 sesquiterpenoids, 13 monoterpenoids, and 1 aliphatic 3-hexenol (Supplemental Table 10). Consistent with data from the 297 genotypes, sesquiterpenoids dominated the volatile compounds in aboveground tissues, especially stem bark (96.17%; Figure 2B). Surprisingly, monoterpenoids accumulated significantly in the root xylem, predominantly as α-pinene (52.61%) and β-pinene (10.8%) (Figure 2B and Supplemental Table 10). In addition to these two monoterpenes, the remaining VOCs in the root xylem included 25.92% sesquiterpenoids and 5.63% monoterpenoids (Figure 2B). The results indicated that sesquiterpenoids were highly abundant in various tissues of P. bournei and represented by numerous varieties. A clear distinction of VOC profiles among tissues was revealed by orthogonal partial least squares discriminant analysis (OPLS-DA; Figure 2C and Supplemental Figure 6). Sesquiterpenes, including β-caryophyllene (21.06% and 8.19%), α-cubebene (19.76% and 40.51%), and α-copaene (10.46% and 11.92%), were the major VOC constituents in stem xylem and bark (Figure 2D and Supplemental Table 10). Hence they may contribute to the pleasant scent of Phoebe wood. VOC composition of the root bark contrasted with that in the root xylem. Overall, although the composition of sesquiterpenes varied among tissues, α-cubebene, α-copaene, and β-caryophyllene were among the most abundant sesquiterpenes in all studied tissues of P. bournei, especially the stem (Figure 2D and Supplemental Table 10).

Antifungal activity of sesquiterpenoids in P. bournei

When collecting genotypes from various natural distribution areas, we observed that no Phoebe trunks were hollowed, even among the old trees. A previous study showed that the monoterpene d-terpineol contributed to decay resistance in Cinnamomum trees (Chaw et al., 2019). P. bournei accumulated abundant sesquiterpenoids, especially β-caryophyllene. To investigate the possible antifungal activity of the sesquiterpenes, we performed in vitro fungal growth inhibition assays using the essential oil of P. bournei and a commercial standard of β-caryophyllene (Figure 2E). Botryosphaeria dothidea, Colletotrichum gloeosporioides, and Lasiodiplodia pseudotheobromae were chosen for bioactivity assays because these species are among the most widespread and important canker and dieback pathogens of trees worldwide (Kamle and Kumar, 2016; Li et al., 2016; Marsberg et al., 2017). Overall, a dose of 0.25% essential oil showed moderate inhibitory activity against the mycelial growth of all the fungi except L. pseudotheobromae (Figure 2E). Interestingly, addition of 0.25% β-caryophyllene inhibited the mycelial growth of all three phytopathogenic fungi by as much as 75% (Figure 2E).

Characteristics of the TPS gene family in P. bournei

To deepen our understanding of terpenoid biosynthesis in P. bournei, we identified genes encoding the proteins responsible for synthesizing a variety of terpenoids. In total, 72 full-length genes and 44 partial fragments encoding TPSs were identified in the P. bournei genome (Supplemental Figure 7). The coding sequences of TPS genes were further checked with transcriptomic data. Among the 72 full-length TPS genes, 10 were putative pseudogenes based on the presence of a frameshift or a premature stop codon that resulted in a truncated protein. These 10 full-length genes were also incorporated into the phylogenetic and gene expression analyses after removing the stop codons or revising the frameshifts by deleting one or two nucleotides. Using the Simple Modular Architecture Research Tool server with Protein Families Database (PFAM) domain analysis (Letunic et al., 2021), all predicted proteins encoded by the 72 genes were found to contain typical TPS domains, supporting their identification as TPS genes. However, the 44 partial TPS fragments did not contain the full-length TPS domain, indicating that the pseudogenization events of these fragments occurred a long time ago.

To define the classes of the 72 TPSs of P. bournei, we constructed a phylogenetic tree using the ML procedure with 649 full-length TPSs identified from 18 genome-sequenced species representing gymnosperms, basal angiosperms, magnoliids, eudicots, and monocots (Supplemental Table 11). TPS genes of P. bournei were grouped into six TPS clades, designated the TPS-a, b, g, e, f, and c subfamilies, with high bootstrap support values. Among these 72 TPSs, the TPS-a and TPS-b classes were the largest, with 34 and 28 members, respectively. The TPS-g and TPS-e classes had three and five members, respectively. The TPS-f and TPS-c classes had only one member each (Supplemental Table 11).

Pervasive lineage-specific gene expansion and loss contributed to the divergence of the TPS family in Lauraceae

Phylogenetic analysis of 649 full-length TPSs from 18 species showed that the largest subfamilies were TPS-a and TPS-b (Supplemental Figure 8). Each Lauraceae species except P. americana contained more than 40 copies of the TPS-a and TPS-b genes (Supplemental Table 11). The TPS-a and TPS-b subfamilies predominantly comprised sesquiterpene and monoterpene synthases, respectively. To further explore the underlying mechanism of terpenoid variation in Lauraceae species, we conducted a phylogenetic analysis of all full-length TPSs from C. kanehirae, L. cubeba, P. americana, and P. bournei. There were at least 65 ancestral TPS genes, including 24 TPS-a genes and 29 TPS-b genes, in the most recent common ancestor (MRCA) of the four Lauraceae species (Figure 3 and Supplemental Figure 9). The large number of TPSs in the MRCA may be partially attributed to two WGD events in Lauraceae species (Figure 1D and Supplemental Figure 4). The total gene numbers in the TPS-g, e, f, and c subfamilies were maintained or slightly reduced in Lauraceae species compared with their MRCA. By contrast, most of the TPS-a and TPS-b orthologs of the MRCA were lost in at least one of the four species after their divergence. This was particularly pronounced in P. americana, where most TPS genes were lost, especially for the TPS-b subfamily, leaving only 20 TPS genes (Figure 3). Contemporaneously with the loss, Lauraceae species also gained numerous TPS paralogs (Figure 3). For example, P. bournei lost nine TPS-a orthologs and specifically duplicated 19 TPS-a paralogs, resulting in 34 TPS-a copies. Similarly, C. kanehirae lost 16 but gained 28 TPS-b genes and ultimately retained 41 TPS-b copies. As a result of more gain than loss, the P. bournei genome harbored 72 TPSs with more TPS-a copies, whereas the C. kanehirae genome harbored 77 TPS genes with more TPS-b copies.

Figure 3.

Copy number changes of TPSs in four Lauraceae species.

The numbers in circles and rectangles represent the numbers of TPS genes in extant and ancestral species, respectively. Numbers on branches with plus and minus symbols represent the numbers of gene gains and losses, respectively.

Extensive tandem duplication and high retention of TPSs in P. bournei

Gene families expanded by tandem duplication, segmental duplication, and WGD provide raw genetic material for evolution. In P. bournei, only the size of the TPS-a subfamily was enlarged (i.e., 1.4 times) compared with that in the Lauraceae MRCA (Figure 3). To understand how the TPS gene family expanded in P. bournei, we first examined the chromosomal distribution of the 72 TPS genes and 44 TPS fragments. Except for 3 TPS genes and 2 TPS fragments assigned to three unattributed scaffold fragments, all TPS genes and fragments were assigned to 10 of the 12 chromosomes (Supplemental Figure 10 and Supplemental Table 12). The TPS distribution on chromosomes was heterogeneous. Chromosomes 2, 7, and 8 harbored the majority of the TPS genes. A total of 86% of the TPS genes (62/72) and 80% of the TPS fragments (35/44) were located in tandemly arrayed clusters. We also analyzed the sequence identities of TPS proteins in P. bournei. Consistent with the relationships in the phylogenetic tree, members within the same tandem cluster showed higher sequence identities than other TPS genes on the same chromosome (Supplemental Figure 11). These results indicated that tandem duplications contributed significantly to expansion of the TPS gene family in P. bournei.

We further reconstructed the expansion history of each gene cluster in accordance with both the gene tree and chromosome localization (Supplemental Figures 7 and 10). The most parsimonious scenarios for gene duplication, rearrangement, and loss are presented (Figure 4). In general, all 13 clusters were expanded by tandem duplications. The historical expansion events of the two large TPS-a clusters (C9 and C11) appeared to have been complex (Figure 4C). It is likely that an ancestral cluster including seven tandemly arrayed TPSs on chromosome 8 underwent a unique segmental duplication event (the segment was longer than 2 Mb), probably as a consequence of local gene duplication by unequal crossover. Subsequently, each descendant cluster experienced a series of tandem duplications and rearrangements (Figure 4C). Both the TPS-a and TPS-b subfamilies experienced an extensive expansion in P. bournei via tandem duplication. Notably, more copies of TPS-a are preserved, and more copies of TPS-b are pseudogenized in the genome, which is supported by the large numbers of pseudogenes and fragments belonging to the TPS-b subfamily (including 27 fragments and 7 pseudogenes).

Figure 4.

Phylogenetic trees and hypothetical evolutionary histories of the P. bournei TPS clusters.

(A–J) Numbers on the branches are bootstrap values. The letters T, S, and R in the schematic diagram indicate hypothetical events of tandem duplication, segmental duplication, and rearrangement, respectively. Boxes represent TPS genes that are functional (black), lost (white), or pseudogenes/fragments (gray). Full-length pseudogenes are indicated by asterisks. Frag indicates a TPS fragment. The TPS clusters numbered C1, C2, etc., are shown in Supplemental Figure 10.

Positive selection on genes of the TPS-a and TPS-b subfamilies in P. bournei

To infer the influences of selection on the expansion of the TPS-a and TPS-b subfamilies in P. bournei, we conducted molecular evolutionary analysis. The ω (d_N/d_S) values for TPS-a and TPS-b genes were estimated using Phylogenetic Analysis by Maximum Likelihood (PAML) software with the one-ratio and free-ratio models (Supplemental Figure 12). Likelihood ratio tests of the two models indicated that the one-ratio model should be rejected in favor of the free-ratio model (P < 0.001) for both the TPS-a and TPS-b subfamilies, suggesting that the branches were under various selection pressures (Supplemental Table 13). The ω values estimated by the free-ratio model were larger than one, indicating that the genes were under positive selection (Supplemental Figure 12). For all TPS-a genes in P. bournei, positive selection was detected on three branches (Clades I, II, and III, Supplemental Figure 12A). The branches under positive selection within Clade I were composed of PbTPS-a17, PbTPS-a18, PbTPS-a19, PbTPS-a22, and PbTPS-a23 from the C9 and C11 clusters. These genes were specifically expanded in P. bournei compared with the other Lauraceae species (Supplemental Figure 9). Positive selection was also detected for several branches of TPS-b genes in P. bournei (Supplemental Figure 12B).

Divergence of TPS expression in P. bournei

Gene expression divergence often occurs at or shortly after duplication, especially for tandem and dispersed duplicates (Ganko et al., 2007). To explore the functional divergence of TPSs in P. bournei, we examined the tissue-specific expression profiles of TPS genes in root xylem, root bark, stem xylem, stem bark, and leaves via transcriptomic analysis (Figure 5B). In general, the expression patterns of TPS genes in the xylem and bark tissues of the stem differed from those of the root (Figure 5B). Moreover, the TPS genes exhibited a unique expression profile in leaves compared with the other tissues. Twenty-seven of the 34 TPS-a genes showed a high expression level in root and/or stem tissue, and half of the TPS-b genes were underexpressed in all tissues we analyzed. The tissue-specific expression patterns of the TPSs were in line with the mono- and sesquiterpenoid compositions of the corresponding tissues (Figures 2D and 5B). Our results suggested that most TPS-a genes (which mainly encode sesquiterpene synthases) were constitutively expressed in stems and roots with high expression levels, whereas most TPS-b genes (which mainly encode monoterpene synthases) were expressed only under specific conditions or in specific tissues (Figure 5B).

Figure 5.

The phylogenetic relationships and expression patterns of P. bournei TPSs.

(A) Phylogenetic tree of full-length TPS proteins of P. bournei. The ML procedure using the JTT+G+F model with 100 bootstrap replicates was used for phylogenetic reconstruction. The numbers at each node in the phylogenetic tree are bootstrap values, and only bootstrap values >50% are shown. Putative pseudogenes are indicated with asterisks. The TPS-a, TPS-b, TPS-g, TPS-e, TPS-f, and TPS-c subfamilies are shaded in different colors.

(B) Tissue-specific expression patterns of TPSs in root xylem (RX), root bark (RB), stem xylem (SX), stem bark (SB), and leaves (L). Expression data are provided as log₂(fragmens per kilobase per million mapped fragments [FPKM]+1) values.

In addition, we analyzed the TPS expression profiles in P. bournei seedlings under drought and cold stresses. All 72 TPS genes exhibited divergent expression patterns either in specific tissues or in response to specific stresses (Figure 5B and Supplemental Figure 13). Furthermore, consistent with the gene expression divergence results, the cis-elements upstream of the 72 TPS genes varied greatly (Supplemental Figure 14).

Functional characterization of TPS-a genes

To better understand the functional divergence of TPS genes and verify the contribution of TPS-a genes to sesquiterpene accumulation in P. bournei, we characterized the enzymatic function of all 10 TPS-a genes from the C9 and C11 clusters, except the pseudogene PbTPS-a24, in tobacco (Figure 6). There were three paralogous gene pairs among these 10 genes: PbTPS-a20/a25, PbTPS-a21/a26, and PbTPS-a22/a23 (Figure 6A).

The sesquiterpenoids produced heterologously in tobacco leaves were similar to the major products observed in P. bournei, such as α-cubebene, α-copaene, β-copaene, β-caryophyllene, and elemene (Figures 2D and 6 and Supplemental Figure 15). For the gene pair PbTPS-a20/a25, the overexpression of PbTPS-a25, but not PbTPS-a20, in tobacco significantly enhanced the emission of β-caryophyllene, indicating the divergent functions of the paralogous genes (Figure 6). Likewise, for the gene pair PbTPS-a22/a23, the accumulation of α-cubebene, α-copaene, and β-copaene was detected in transgenic tobacco with PbTPS-a22, but not PbTPS-a23 (Figure 6). For the gene pair PbTPS-a21/a26, PbTPS-a21 catalyzed the formation of the bicyclic terpenoids β-caryophyllene and guaia-6,9-diene as major products and the monocyclic terpenoids elemene and ylangene as minor products. PbTPS-a26 exhibited a distinct enzymatic function by enhancing the emissions of the tricyclenes α-copaene and β-copaene, indicating that neofunctionalization or subfunctionalization had occurred in the PbTPS-a21/a26 gene pair (Figure 6). In summary, we identified PbTPS-a22 as the main synthase for α-cubebene and PbTPS-a21 and PbTPS-a25 as the main synthases for β-caryophyllene, the two most abundant sesquiterpenes in the stem xylem of P. bournei (Figures 2D and 6).

Discussion

In this study, we used PacBio, Bionano, and Hi-C data to assemble a chromosome-scale genome of P. bournei. For species with high heterozygosity, a chromosome-scale genome assembly helps to reduce sequence redundancy and improve genome accuracy and completeness. Compared with a contig-scale genome (Chen et al., 2020b), the decreased sequence redundancy was mainly achieved by removing repetitive sequences. The percentages of repetitive elements in the other three Lauraceae plants with chromosome-assembled genomes ranged from 48% to 55.47% (48% for C. kanehirae, 50% for P. americana, and 55.47% for L. cubeba). The content of repetitive elements in the P. bournei genome was 48.10%, which also falls within this range. We used additional transcriptomic data from five tissues of P. bournei to correct the annotation of gene models. With the chromosome-assembled genome and the additional transcriptomic data, the quality of the P. bournei genome in our study was greatly improved compared with that in a previous report (Chen et al., 2020b). The improvements were evident in the increased average length of genes (10 606.6 versus 9 695.55 bp) and coding sequences (CDS) (1264.20 versus 1258.80 bp) and the more complete gene functional annotation (98.74% versus 95.79%) and BUSCO assessment (91.7% versus 81.1%) (Supplemental Tables 7–9).

The phylogenetic placement of magnoliids has a conflicting and undefinable phylogenetic topology. The sparse taxon sampling and limited gene numbers used for phylogenetic analysis are crucial restricting factors that result in the phylogenetic discordance of magnoliids among major Mesangiospermae lineages (Chaw et al., 2019; Chen et al., 2019, 2020c; Hu et al., 2019; Rendon-Anaya et al., 2019; Shang et al., 2020). In addition, incomplete lineage sorting during early diversification significantly impacts the elucidation of topological variation in Mesangiospermae lineages (Ma et al., 2021). The results of this study showed that magnoliids were a sister clade to eudicots rather than monocots (Figure 1C). A recent study with sufficient supporting evidence also indicated that the monocot lineage was sister to major Mesangiospermae lineages, including magnoliids and eudicots (Ma et al., 2021).

The genera Phoebe and Persea belong to the Tribe Persea and specifically accumulate sesquiterpenoids (Wei and van der Werff, 2008; Meléndez-González and Espinosa-García, 2018). By contrast, monoterpenoids are more abundant in species from other Lauraceae genera, such as Cinnamomum, Laurus, Litsea, and Cryptocarya (Joshi et al., 2009). Over the course of this project, newly assembled genome sequences became available for several Lauraceae species, including C. kanehirae, L. cubeba, and P. americana (Chaw et al., 2019; Rendon-Anaya et al., 2019; Chen et al., 2020c), which made it possible to investigate the evolution of the TPS family in Lauraceae with reliable data. The high abundance of sesquiterpenoids detected in the vegetative organs of P. bournei was attributed to the large size and high expression level of the TPS-a gene family in P. bournei (Figures 3 and 5). We found that pervasive lineage-specific expansions and contractions of the TPS family contributed to the extreme terpenoid variation within Lauraceae (Figure 4 and Supplemental Figure 9). Similar to the findings of studies on the TPS family in other angiosperm species such as tomato, rice, sorghum, C. kanehirae, and L. cubeba (Chen et al., 2011, 2020a, 2020c; Chaw et al., 2019; Zhou and Pichersky, 2020a), tandem duplication was the main pattern for TPS family expansion in P. bournei and provided raw genetic material for functional divergence and rapid biological evolution (Chantret et al., 2005; Lan et al., 2009). Interestingly, our results demonstrate that these two subfamilies of the TPS gene family (TPS-a and TPS-b) have not evolved in the same way in P. bournei (Figure 7). As shown by the MRCA analysis, 9 TPS-a orthologs were lost and 19 TPS-a paralogs were specifically duplicated, resulting in 34 TPS-a copies in P. bournei (Figure 3). By contrast, 10 TPS-b orthologs were lost and 9 TPS-b paralogs were specifically duplicated, resulting in 28 TPS-b copies in P. bournei. If we subtracted the pseudogenes among these 34 TPS-a and 28 TPS-b genes, then 31 TPS-a and 22 TPS-b genes encoding putative functional TPSs remained. Our results support 1.4 times (31/22) more putative functional TPS-a genes than TPS-b genes in P. bournei, indicating a more pronounced expansion of the TPS-a subfamily than the TPS-b subfamily after speciation.

Figure 7.

Schematic overview of the contrasting fates of the TPS-a and TPS-b subfamilies in P. bournei contributing to terpenoid variation.

Although the TPS-a and TPS-b subfamilies are both expanded by tandem duplication, more TPS-a copies are retained and constitutively expressed, whereas more TPS-b copies are lost, resulting in a high abundance of sesquiterpenoids instead of monoterpenoids. VOCs with abundant sesquiterpenoids, especially β-caryophyllene, produced in stems may confer excellent pathogen resistance to P. bournei trees.

Although the TPS-a and TPS-b families underwent similar degrees of tandem duplication based on the analysis of hypothetical evolutionary histories in P. bournei (Figure 4), the TPS-b subfamily had a higher rate of gene loss than the TPS-a subfamily. As evidenced by 27 fragments and 7 pseudogenes in the P. bournei genome, the TPS-b subfamily was contracted compared with the TPS-a subfamily. Diversifying selection is the main driver of lineage-specific expansions and contractions of gene families (McDowell et al., 1998; Alunni et al., 2007). The loss of TPS-b genes may be less harmful than that of TPS-a genes. Consistent with the gene expression divergence results, the cis-elements upstream of the 72 TPS genes varied to a great extent (Figure 5 and Supplemental Figure 14). Although the enzymatic function analysis showed that PbTPS-a25, PbTPS-a21, and PbTPS-a26 catalyzed the formation of β-caryophyllene, one of the most abundant sesquiterpenes in P. bournei (Figure 6), PbTPS-a25, PbTPS-a21, and PbTPS-a26 showed divergent gene expression patterns. Although we cannot define the function of β-caryophyllene synthase as derived either from its ancestor or from a neofunctionalization event, these three genes functionally diverged in enzymatic function and gene expression. The contrasting evolutionary mechanisms and expression patterns between TPS subfamilies have together contributed to the accumulation of sesquiterpenoids instead of monoterpenoids in P. bournei (Figure 7).

The main constituents in leaf essential oils of L. cubeba and C. kanehirae are monoterpenoids (Cheng et al., 2015; Chen et al., 2020c). By comparison, volatile analysis of avocado (P. americana) trees showed similar results to our study (Supplemental Table 10), in that sesquiterpenoids such as β-caryophyllene, α-cubebene, and α-copaene were dominant (Ogunbinu et al., 2007; Niogret et al., 2013). Recent studies of Chimonanthus praecox showed that monoterpenes were the major components of its volatile compounds (Tian et al., 2019), and expansion of the TPS-b subfamily contributed to monoterpene accumulation in C. praecox (Shang et al., 2020). Therefore, the relationship between the accumulation of sesquiterpenes and the high preservation of TPS-a genes in P. bournei is probably not coincidental. Admittedly, the number of compared genomes is limited because of the inadequacy of available genome data. Further evidence is needed from more plant species via combined analysis of TPS family evolution and metabolite profiles.

Compared with the other TPS subfamilies, the TPS-a and TPS-b genes exhibited rapid expansion in angiosperms (Chen et al., 2011), especially in Lauraceae species (Supplemental Figure 8). Why did this rapid expansion of TPS-a and TPS-b genes occur in angiosperms? One possible reason is driven by functional requirements (Nam et al., 2004). Perennial trees are subjected to many more abiotic and biotic stresses during their long life cycle than are annual plants. The diseases caused by herbivores and pathogens are devastating to forests, especially in modern forestation areas with a single or closely related species (Ploetz et al., 2017; Aylward et al., 2019). Many tree species in Lauraceae accumulate essential oils with specific scents, each of which exhibits a unique pattern of monoterpene and sesquiterpene composition (Guo et al., 2016; Chaw et al., 2019; Chen et al., 2020c). These volatile monoterpenes and sesquiterpenes play an essential role in pollinator attraction and defense against herbivory and pathogens (Pichersky and Raguso, 2018). Lauraceae trees dominate many subtropical forest sites in terms of abundance and coexisting species diversity (Ye et al., 2019). The chemical dissimilarity between cohabitant species could help reduce the probability of specialist biological enemies being transmitted between host plants (Pichersky and Raguso, 2018).

Defense against biological enemies is the best-established function of plant terpenoids (Pichersky and Raguso, 2018). However, the assignment of specific roles to specific terpenes has been difficult. Phoebe trees show good decay resistance and rarely have hollowed or cankered trunks in natural forests (Li and Min, 2020). The genus Persea is closely related to the genus Phoebe but is mainly distributed in America (Wei and van der Werff, 2008). In contrast to those in P. bournei, both the TPS-a and TPS-b subfamilies in P. americana were mostly contracted, with only 12 and 3 genes preserved in the genome, respectively (Figure 3). The loss of functional TPS genes in Persea may increase the risk of pathogen infection. In fact, epidemic laurel wilt disease has killed over 300 million red bay trees (Persea borbonia) in the eastern United States and currently poses an economic threat to avocado (P. americana) (Ploetz et al., 2017). Laurel wilt is caused by the fungus Raffaelea lauricola, a symbiont of an invasive Asian ambrosia beetle (Xyleborus glabratus Eichhoff) that prefers native lauraceous hosts such as Phoebe and camphor trees in China (Fraedrich et al., 2015). Despite the presence of the sesquiterpenes α-copaene and α-cubebene, which accumulate to high levels in P. bournei and are kairomones for attracting the beetle X. glabratus, there are no reports of laurel wilt in Phoebe trees in China (Ploetz et al., 2017).

Interestingly, β-caryophyllene and its oxide have been confirmed to inhibit the growth of a wide range of microbes and function as allelochemical components to induce resistance via jasmonic acid signaling (Chang et al., 2008; Silva et al., 2019; Karunanithi et al., 2020). β-caryophyllene, which is synthesized by PbTPS-a21 and PbTPS-a25, exhibited high antipathogenic activity against three common woody plant pathogenic fungi (Figure 2E), suggesting that the abundance of sesquiterpenoids, especially β-caryophyllene, in the stem plays a vital role in the adaptation and pathogen defense of P. bournei (Figure 7). Notably, PbTPS-a26 was the synthase for α-copaene, whereas its paralogous protein PbTPS-a21 was sub- or neofunctionalized, synthesizing β-caryophyllene instead (Figure 6). The divergence of multifunctional duplicates of PbTPS-a21/a26 might provide proof of the “coevolutionary arms race” hypothesis that assumes terpenoids diverge as a result of changes in their biotic partners and enemies (Pichersky and Raguso, 2018).

The genome-level comparison showed pervasive lineage-specific tandem duplication and contractions of the TPS family that contributed to the extreme terpenoid variation within Lauraceae. We demonstrated that the TPS-a and TPS-b subfamilies experienced contrasting fates during the evolution of P. bournei. The abundant sesquiterpenoids produced by TPS-a proteins in the stem contributed to the excellent pathogen resistance of P. bournei trees. Although the ecological functions of each specific terpenoid in Phoebe have not yet been elucidated, there is excellent potential for selecting effective antipest/pathogen compounds from Phoebe. In conclusion, this study sheds light on the evolution of terpenoid divergence in Lauraceae and the contribution of sesquiterpenoids to adaptation in P. bournei.

Materials and methods

Plant material

Seeds of P. bournei collected from 33 provenances in southern China were planted in a nursery in Qinyuan (27° 38′ 0.67″ N, 119° 02′ 36.48″ E), Zhejiang Province, China, in 2011 (Supplemental Figure 5). A total of 297 individual trees from the nursery were sampled in the summer for VOC analysis. A six-year-old tree was selected for genome sequencing in 2017; this tree was the offspring of a tree from the largest population in Zhenghe (27° 26′ 56.7″ N, 118° 36′ 49.7″ E), Fujian Province, China. Different tissues of the sequenced individual were collected in the summer for VOC and gene expression analyses.

Flow cytometry analysis

The genome size of P. bournei was measured by flow cytometry. Fresh leaves were chopped with a razor blade in a Petri dish with 1 ml Tris–MgCl₂ isolation buffer (200 mM Tris, 4 mM MgCl₂·6H₂O, 0.5% Triton X-100 [pH 7.5]) and 500 μl 1% polyvinyl pyrrolidone (PVP). The cell nuclei suspension was filtered through a 50-μm nylon filter into a 1.5-ml centrifuge tube and incubated on ice for 5 min. After 5 min of centrifugation at 1000 g, the supernatant was discarded. Then, 100 μl of precooled isolation buffer was added for resuspension. The nuclei were stained by adding 150 μl of precooled 50 μg ml⁻¹ propidium iodide in the dark for 5 min at 4°C. The DNA content of at least 10 000 stained nuclei was analyzed on a CytoFLEX Flow Cytometer (Beckman Coulter, USA). The genome size of P. bournei was estimated using Populus trichocarpa as the internal standard.

Genome sequencing and assembly

The DNA library was constructed with a DNA Template Prep Kit, DNA Polymerase Binding Kit, and MagBead Kit (PacBio, Menlo Park, CA, USA). The 20-kb DNA library for long reads was sequenced on a PacBio RS II instrument using the optimal combination of P6 polymerase and C4 reagent. The raw data were filtered for quality control. DNA libraries for the genome survey, BioNano, and Hi-C analyses were sequenced on the Illumina HiSeq X Ten platform with 150-bp paired-end reads. PacBio long-read sequences were assembled using FALCON v.1.8.7 (Chin et al., 2016). The consensus sequences of the primary assembly were then corrected and polished using GenomicConsensus v.2.3.0 (https://opensourcelibs.com/lib/genomicconsensus) with PacBio reads and Pilon v.1.22 (Walker et al., 2014) with Illumina reads. The PacBio assembly was scaffolded using BioNano data and gap closed with PBJelly v.15.8.24 (English et al., 2012). A chromosome-level assembly was achieved with Hi-C data. Genome completeness was assessed using BUSCO v.3.0.2 and the Embryophyta 10 dataset and by alignment of RNA-seq datasets from the root, stem, leaf, bud, and seed tissues of P. bournei with STAR v.2.7.3 (Dobin et al., 2013). More details on genome assembly are described in the Supplemental notes.

Genome annotation and functional prediction

Repetitive sequences in the P. bournei genome were explored by homolog searching and ab initio prediction. For the first method, two databases, Repbase (Bao et al., 2015) and PGSB-REdat (Spannagl et al., 2016), were used for annotation with RepeatMasker v.4.0.8 (Bedell et al., 2000), RepeatProteinMask (Bedell et al., 2000), and Tandem repeats finder (Benson, 1999). For the second method, a de novo repeat database was built with RepeatModeler v.1.0.9 (Bedell et al., 2000) and then applied to the P. bournei genome with RepeatMasker. The repeat-masked genome was then subjected to protein-coding gene prediction. Three strategies were used to identify gene models. First, the ab initio and evidence-directed predictor Augustus (Stanke and Waack, 2003) was trained on Arabidopsis thaliana and used to predict gene models along the P. bournei genome. Second, protein-homology-based prediction was performed with GeMoMa (Keilwagen et al., 2016) and GeneWise (Birney and Durbin, 2000) using homolog protein sets of six representative plants, A. thaliana, Oryza sativa, Populus trichocarpa, Amborella trichopoda, Eucalyptus grandis, and Vitis vinifera, which were downloaded from Phytozome. Third, transcriptome data were used for annotation with Program to Assemble Spliced Alignments (PASA) (Haas et al., 2008). Finally, the three sets of gene models were merged and filtered using the EVidenceModeler (EVM) pipeline to generate the final gene models for the P. bournei genome. The gene annotation was processed by eggNOG-mapper (Huerta-Cepas et al., 2017) with the Non-Redundant Protein Sequence Database (NR), Swisspro, Trembl, InterProScan, Gene Ontology, Eukaryotic Orthologous Groups of protein (KOG), and Kyoto Encyclopedia of Genes and Genomes databases. BUSCO v3.0.2 analysis of identified genes was performed using the Embryophyta 10 dataset.

Comparative genomic analyses

To elucidate the phylogenetic positions of Phoebe and magnoliids on the tree of life, we selected P. bournei, P. americana, C. kanehirae, L. cubeba, Liriodendron chinense, and Piper nigrum as representatives of magnoliids and compared these with five eudicots and five monocots, including the relatively early evolved monocot Zostera marina and the basal eudicots Aquilegia coerulea, Macleaya cordata, and Nelumbo nucifera. Gnetum montanum and Amborella trichopoda were used as outgroups. Single-copy orthologous genes were identified using OrthoMCL v.2.0.9 (Li et al., 2003). Protein sequences were aligned with MAFFT v.7.407 (Katoh and Standley, 2013). Conserved alignments were subsequently filtered with Gblocks v.0.91b (Talavera and Castresana, 2007). A single-copy gene phylogenetic tree was constructed by the ML method using RAxML v.8.2.12 (Stamatakis, 2006). Then, all single-copy orthologous gene trees were integrated into a coalescent species tree with ASTRAL v.5.6.1 (Zhang et al., 2018a) using the PROTGAMMAILGF model and 500 bootstrap replicates. In addition, all conserved alignments were concatenated into a superalignment matrix. The ML species trees based on the supermatrix were constructed with RAxML v.8.2.12 (Stamatakis, 2006), and a Bayesian phylogenetic species tree was constructed using MrBayes v.3.1.2 (Ronquist and Huelsenbeck, 2003).

Whole-genome duplication analysis

The longest transcript of each gene for representative plants was selected for WGD analysis. For the analysis of intraspecies paralogous genes, we performed a self-to-self alignment with BLASTP v.2.6.0+ (-Evalue 1e-10 -max_target_seqs 5), and then collinear blocks were identified using MCScanX (Wang et al., 2012). Synonymous substitution rates (K_s) of codons and four-fold degenerate synonymous site distances between anchored genes were calculated with KaKs_Calculator v.2.0 (Wang et al., 2010) and a custom Perl script, respectively. For analysis of interspecies orthologous genes, the K_s values of different species pairs were calculated separately based on single-copy orthologous genes drawn from the OrthoMCL analysis. The ggplot2 and ggridges packages were used to compute and draw the kernel density estimate. The times of WGD events were calculated based on a synonymous substitution rate of 3.02E−9 synonymous substitutions per year for the basal angiosperms (Cui et al., 2006).

Determination of volatile components

The volatile components from five tissues, including the root xylem, root bark, stem xylem, stem bark, and leaf of P. bournei, were extracted by modified headspace solid phase microextraction (SPME) and determined by gas chromatography–mass spectrometry (GC–MS) (Deng et al., 2004; Cecchi and Alfei, 2013; Chen et al., 2020c). The SPME fiber used for volatile component extraction was 100-μm polydimethylsiloxane (Sigma–Aldrich, St. Louis, MO, USA) and was conditioned before use as recommended by the manufacturer. A 1.2-g fresh sample was minced and placed in a 20-ml vial sealed with a poly tetra fluoroethylene (PTFE)-silicone septum, and the fiber was then inserted into the vial for 1 h at room temperature. After the extraction, the SPME fiber was exposed to the hot injection port of the Trace GC/ISQ GC–MS (Thermo Fisher, USA) equipped with a TM-5MS column (30 m × 0.25 mm × 0.25 μm; Thermo Fisher) at 250°C for 3 min. The helium carrier gas flow rate was 1 ml min⁻¹. The program setting for the oven was as follows: held at 40°C for 5 min, gradually increased to 170°C at 3°C min⁻¹, held at 170°C for 5 min, gradually increased to 280°C at 10°C min⁻¹, and held at 280°C for 5 min. The volatile compounds were identified by spectral comparison against the National Institute of Standards and Technology MS Library v.2.2 and interpreted manually with an MS matching score above 800. The retention index was calculated using a homologous series of n-alkanes (C7–C30). The relative quantification of volatile terpenoids in P. bournei was performed by normalizing the peak area to whole compounds within each sample. Multivariate Data Analysis (SIMCA-P) software was then used to perform the OPLS-DA. In addition, β-caryophyllene was identified by comparing retention times and mass spectra with those of its commercially available authentic standard (product number 22075; Sigma–Aldrich). Three biological replicates each of the five tissues were used for metabolite profile analysis.

Fungal growth-inhibition assays

Isolates of B. dothidea, C. gloeosporioides, and L. pseudotheobromae were kindly provided by Yongjun Wang from Zhejiang A&F University and Shuaifei Chen from the China Eucalypt Research Centre, Chinese Academy of Forestry. To evaluate antifungal activities, 5 μl each of the leaf essential oil and the commercial β-caryophyllene standard were dissolved in 15 μl of dimethyl sulfoxide and then diluted in 1980 μl of distilled deionized water to obtain a final concentration of 0.25%. Diluted dimethyl sulfoxide was used as a negative control. Mycelial plugs (6 mm in diameter) from the edges of each culture were incubated on potato dextrose agar (PDA) plates (90 mm in diameter). Cultures were incubated at 28°C and 70% relative humidity for 48–72 h. Tests were repeated in quadruplicate. The antifungal index was calculated as described by Chang et al. (2008).

TPS family identification

Nonredundant sequence sets of TPS proteins were collected from 18 seed plant genomes (Supplemental Table 11) via HMMER (Eddy, 2011) using the PFAM models PF03936 and PF01397 with an e-value cutoff of 1 × 10⁻⁵. The predicted proteins were checked with the InterPro (Mitchell et al., 2019) and SMART (Letunic et al., 2021) programs to classify them into protein families and confirm the presence of typical TPS domains in their protein structures. Sequences with more than 350 amino acids were regarded as putative functional proteins and retained for further analysis. With these criteria, the TPS gene numbers we identified in A. thaliana and L. cubeba were the same as those previously reported (Zhou and Pichersky, 2020b; Chen et al., 2020c). However, the gene number in C. kanehirae was lower than that identified previously because we removed TPS fragments based on the integrity of their conserved domains (Chaw et al., 2019).

Phylogenetic and molecular evolutionary analyses

Protein sequences of the 649 hypothetical TPS genes from 18 plants were aligned with Multiple Alignment using Fast Fourier Transform (MAFFT) (Katoh and Standley, 2013). An approximate ML tree of aligned protein sequences was constructed using FastTree (Price et al., 2010). The protein sequence alignments of Lauraceae TPSs and P. bournei TPSs were performed with MUltiple Sequence Comparison by Log-Expectation (MUSCLE) (Madeira et al., 2019). The corresponding coding sequence alignment was performed using RevTrans 2.0 (Wernersson and Pedersen, 2003). The ML tree was constructed using PhyML 3.0 (Guindon et al., 2010) with smart model selection (Lefort et al., 2017). Phylogenetic trees were modified with MEGA X (Kumar et al., 2018). We used the same method to construct phylogenetic trees of TPS proteins. However, the protein members used to construct each phylogenetic tree were different, which caused alignment variation, and the model used for tree building by smart model selection may have been different. As a result, the phylogenetic relationships may differ between Figures 4 and 5 and Supplemental Figures 7 and 12, especially for clades with low bootstrap values. To evaluate selection pressures among the TPS-a and TPS-b subfamilies of P. bournei, branch models of CODEML in PAML were used to estimate ω values under two assumptions: a one-ratio model that assumes the same ω ratio for all clades and a free-ratio model in which the ω ratio is assumed to vary over all branches. Likelihood ratio tests were performed to determine which model best fit the data by comparing twice the difference in the log-likelihood values between pairs of models using a χ² distribution (Yang and Nielsen, 2000).

Ancestral genes were identified according to a previous report (Kim et al., 2006). First, we constructed a phylogenetic tree with protein sequences of all full-length TPSs from C. kanehirae, L. cubeba, P. americana, and P. bournei (Supplemental Figure 9). Subsequently, we identified nodes that led to species-specific clades and predicted that these nodes represented the divergence points between the four species based on two criteria: (a) the bootstrap value was higher than 50%; and (b) the relationship among the four species-specific clades was consistent with the species tree, as presented in Supplemental Figure 3. If these two criteria were confirmed simultaneously, we designated the clades defined by such nodes as orthologous groups derived from a MRCA gene of the four species. However, gene losses could have occurred; therefore, some TPS genes may be present in the common ancestor but later lost in several descendant species. Thus, more orthologous groups were also identified based on their sister group relationships to the MRCA clade.

The hypothetical expansion histories of TPS clusters were reconstructed by reconciling both the gene tree and the positions of genes within clusters, according to previous reports (Lan et al., 2009; Ren et al., 2014). Individual genes were assigned to tandem gene clusters if the genes or fragments were closely distributed on the same chromosome; the distances between two TPS genes/fragments were mostly less than 2.7 Mb, except for TPS fragments 11 and 16. Moreover, those genes or fragments were grouped together with a bootstrap value higher than 50 on the phylogenetic tree (Supplemental Figure 7). The tandem duplication events of each cluster were deduced according to the relationship of the cluster-specific phylogenetic tree shown in Figure 4. We assumed a rearrangement event occurred if two gene copies were derived from a tandem duplication but were not located tandemly in the same chromosome region.

Expression analysis of TPS genes

Samples were harvested from leaves, stem xylem, stem bark, root xylem, and root bark of the sequenced P. bournei treelet and frozen quickly in liquid nitrogen before RNA extraction. One-year-old seedlings of P. bournei were hydroponically cultured with (drought-stressed condition) and without (nonstressed condition) 10% polyethylene glycol PEG 6000. Leaf and root samples were collected for RNA extraction at 0 h, 1 h, 1 day, and 3 days after treatment. Three biological replicates were included for each sample. Two pedigrees (WP and WY) collected from the southern and northern edges of the P. bournei distribution area were considered cold-sensitive and cold-resistant, respectively. Seedlings of these two pedigrees were treated with cold stress (Zhang et al., 2018b). Library construction was performed following the manufacturer’s protocols. The Illumina HiSeq X Ten platform was used for 150-bp paired-end sequencing. The raw data were filtered with fastp v.0.20.0 (Chen et al., 2018) and then mapped to the P. bournei genome using HISAT2 v.2.1.0 (Kim et al., 2015). Gene expression levels were calculated using featureCounts v.1.6.4 (Liao et al., 2014), and differentially expressed genes were detected using the R package edgeR v.3.26.8 (Robinson et al., 2010) based on the criteria |log₂(fold-change)| > 1 and false discovery rate <0.05.

TPS gene cloning and functional characterization

TPS genes in the sequenced P. bournei were amplified with specific primers (Supplemental Table 14) and introduced into the pK2GW7 vector with a 35S::PbTPS overexpression cassette by Gateway-assisted vector construction (Han et al., 2020). The empty and overexpression vectors were transformed into Agrobacterium tumefaciens GV3101 and infiltrated into the leaves of tobacco (Nicotiana benthamiana) for transient expression analysis. After 48 h of growth, the transformed leaves were collected for headspace SPME/GC–MS analysis. The extraction of tobacco VOCs was performed using a procedure similar to that described above but with a few modifications (Wu et al., 2020). In brief, the injector temperature was set to 250°C, and the column flow rate was 1.5 ml min⁻¹. The program setting for the oven was as follows: held at 50°C for 2 min, gradually increased to 180°C at 3°C min⁻¹, held at 180°C for 1 min, gradually increased to 230°C at 10°C min⁻¹, held at 230°C for 5 min, increased to 250°C at 20°C min⁻¹, and held at 250°C for 3 min. The retention index was calculated using a homologous series of n-alkanes (C7–C30). The authentic standards of β-caryophyllene and α-copaene were obtained from Sigma–Aldrich. The experiments were performed with three biological replicates.

Graphical and statistical software

Most graphs were plotted with the R package ggplot2. Analysis of variance was performed in R using the base package. Final figures were assembled using PowerPoint and Adobe Illustrator.

Data availability

The raw sequencing data used in the genome assembly of P. bournei have been deposited in NCBI under BioProject accession number PRJNA627308. RNA-seq data from different tissues of P. bournei are available under BioProject accession number PRJNA628065. The genome assembly has been deposited into the Sequence Archive of China National GeneBank DataBase with accession number CNP0002030.

Funding

This work was supported by the Zhejiang Science and Technology Major Program on Agricultural New Variety Breeding (grant 2021C02070-10), the National Natural Science Foundation of China (grants 32171828 and 32101545), and the State Key Laboratory of Subtropical Silviculture (grant ZY20180204).

Author contributions

Z.T., Q.Y., X.H., and J.Z. designed the research. X.H., S.Z., L.L., X.L., and L.C. performed the sampling. S.H., S.L.C., M.S., Y.W., C.L., E.L., H.H., and X.H. performed the experiments. Q.Y., G.M., X.H., and S.H. analyzed the data. Q.Y., X.H., J.Z., and Z.T. wrote the manuscript. All authors read and approved the final manuscript. X.H. and J.Z. contributed equally.

Published: July 16, 2022