Characterization of terpene biosynthesis in Melaleuca quinquenervia and ecological consequences of terpene accumulation during myrtle rust infection

Abstract Plants use a wide array of secondary metabolites including terpenes as defense against herbivore and pathogen attack, which can be constitutively expressed or induced. Here, we investigated aspects of the chemical and molecular basis of resistance against the exotic rust fungus Austropuccinia psidii in Melaleuca quinquenervia, with a focus on terpenes. Foliar terpenes of resistant and susceptible plants were quantified, and we assessed whether chemotypic variation contributed to resistance to infection by A. psidii. We found that chemotypes did not contribute to the resistance and susceptibility of M. quinquenervia. However, in one of the chemotypes (Chemotype 2), susceptible plants showed higher concentrations of several terpenes including α‐pinene, limonene, 1,8‐cineole, and viridiflorol compared with resistant plants. Transcriptome profiling of these plants showed that several TPS genes were strongly induced in response to infection by A. psidii. Functional characterization of these TPS showed them to be mono‐ and sesquiterpene synthases producing compounds including 1,8‐cineole, β‐caryophyllene, viridiflorol and nerolidol. The expression of these TPS genes correlated with metabolite data in a susceptible plant. These results suggest the complexity of resistance mechanism regulated by M. quinquenervia and that modulation of terpenes may be one of the components that contribute to resistance against A. psidii.


| INTRODUC TI ON
Terpenes, a class of secondary metabolites, play a key role in both constitutive and induced defenses in many plants. Terpene diversity arises from a large family of enzymes called terpene synthases (TPS; Degenhardt et al., 2009), which fold prenyl diphosphate substrates, such as farnesyl diphosphate (FPP) and geranyl diphosphate (GPP) into terpenes (Bohlmann et al., 1998;Köllner et al., 2004). GPP is produced through the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway in the chloroplast, whereas FPP is produced through the mevalonate (MVA) pathway in the cytosol. GPP is the precursor of monoterpenes with 10 carbon units, and FPP is used as a precursor to sesquiterpenes with 15 carbon units (Chen et al., 2011).
In contrast to studies of terpenes in plant-insect interactions, their role in fungal infections is less well known. Other than studies that investigated the effects of terpenes to beetle-vectored fungal defense in several conifer species (Keeling & Bohlmann, 2006;Zeneli et al., 2006), only few studies investigated induced terpenes with potential antifungal properties in plants in response to fungal infections alone. For example, Visser et al. (2015) compared gene expression profiles between resistant and susceptible Eucalyptus grandis in response to the necrotrophic fungus Chrysoporthe austroafricana and identified inductions of transcripts encoding a putative isoprene synthase (EgrTPS084) and a β-caryophyllene synthase (EgrTPS038) after infection in resistant plants. Maize (Zea mays) responds to infection by Fusarium graminearum by producing zealexins and kauralexins (both terpenoids) in response to Cercospora zeina (Huffaker et al., 2011;Meyer et al., 2017). Similarly, bell pepper (Capsicum annuum) accumulates the terpenoid capsidiol, which affects the growth of fungal-like oomycete pathogens Phytophthora capsici and Phytophthora infestans (Giannakopoulou et al., 2014).
The recent invasion of the exotic rust fungus Austropuccinia psidii (G. Winter) Beenken comb. nov. (Beenken, 2017;myrtle rust) in Australia has infected hundreds of woody plant species in the Myrtaceae (Carnegie et al., 2016). The rust is a serious biosecurity concern as species of Myrtaceae dominate much of the Australian flora (Ladiges et al., 2003) and include many species-rich genera, such as Eucalyptus and Melaleuca, which are widely utilized for their terpene-dominated essential oils (Boland et al., 1991;Carson et al., 2006) and are globally planted for pulp and hardwood production . This rust disease has severely affected several industries that depend on Myrtaceae (Carnegie et al., 2016) and may threaten the biodiversity of Australia, particularly along the east coast wetlands (Pegg et al., 2014). Melaleuca quinquenervia is one of the species most at risk. It is the dominant woody plant species of the east coast wetland ecosystem (Barlow, 1988;Cook et al., 2008).
Field studies on M. quinquenervia have observed foliar and stem dieback, as well as flower death, which suggests that the disease may lead to disruptive long-term consequences to ecological function of wetlands as the symptoms may impede forest regeneration (Pegg et al., 2014).
Screening studies showed M. quinquenervia was variably resistant to A. psidii (Hsieh et al., 2017;Morin et al., 2012). M. quinquenervia displays a high degree of qualitative variability in foliar essential oils (Ireland et al., 2002) described as "chemotypes" (chemical phenotypes). Different chemotypes of M. quinquenervia have characteristic terpene profiles, which have been suggested to play important roles in ecology, such as in host plant selection by insects (Padovan et al., 2010;Wheeler & Ordung, 2005). Two chemotypes have been identified (Ireland et al., 2002). Chemotype 1 occurs mainly in the southern range of the species, including Sydney (33°54′S, 151°15′E), Selection In a previous study (Hsieh et al., 2017), we observed many mRNA transcripts from defense-related genes that were upregulated after We then functionally characterized several of these TPS genes which had been induced in response to fungal infection.

| Identification of differentially expressed TPS and terpene biosynthesis-related transcripts in M. quinquenervia
From the transcriptome profiling study of four highly resistant (HR) and four highly susceptible (HS) M. quinquenervia in response to A. psidii (Hsieh et al., 2017), we observed that the expression of F I G U R E 1 (a) Approximate geographic distribution of Chemotype 1 and 2 Melaleuca quinquenervia in Australia. Orange dots represent distribution of M. quinquenervia. Chemotype 1 occurs mostly in circled areas (black), whereas Chemotype 2 is distributed throughout the entire range of the species. The distribution of Chemotype 1 was derived from data in Ireland et al. (2002); map was produced using The Atlas of Living Australia. Seed source populations are indicated on map. (b) Major types of terpenes found in Chemotypes 1 and 2 of M. quinquenervia (figure design adapted from Padovan et al., 2010) [Colour figure can be viewed at wileyonlinelibrary.com] defense-related genes in response to A. psidii were highly dissimilar among biologically independent samples of HR. Several of these variable yet nonsignificant transcripts (compared with susceptible) encoded TPS. As such inductions may be biologically important and remembering that our samples are genetically diverse, we selected those transcripts encoding TPS (hereafter referred to as "TPS transcripts") and calculated fold changes (FC) of transcripts per million (TPM) for each genotype individually, which showed high levels of induction in one of the genotypes. Details of RNA extraction and sequencing of these samples are described in Hsieh et al. (2017).

| Phylogenetic analysis of M. quinquenervia TPS compared with other Myrtaceae
Translated amino acid sequences of all TPS transcripts observed from the M. quinquenervia transcriptome (Hsieh et al., 2017) were retrieved to align against known E. grandis TPS sequences with subfamily information . Two characterized M.  (Padovan et al., 2017) were also included in the phylogenetic analysis to assess the homology of TPS transcripts to known terpene synthases from Melaleuca.
Before aligning M. quinquenervia TPS sequences to those of E. grandis and M. alternifolia, they were first manually aligned with the characterized M. quinquenervia viridiflorol synthase gene sequences using BioEdit (Version 7.2.6;Hall, 1999) to check that sequences were not identical or potential allelic variants of each other.
Sequences were aligned using conserved diagnostic motifs as anchors. These motifs include: RRX 8 W (the cap of the catalytic pocket), RLLR (situated in the latter half of the first exon), YEACH/LEASH, RWW/RWG/RWV (intron-exon boundary), DDXXD (involved in Mg 2+ -assisted substrate binding), and (N,D)DX 2 (S,T)X 3 E. Sequences that were identical or varied only by length differences were discarded (keeping the longest).
Subsequently, we manually aligned the M. quinquenervia TPS sequences to 113 previously aligned E. grandis TPS sequences described by Külheim et al. (2015) and the five functionally characterized M. alternifolia sequences. All sequences were then truncated as described . PhyML Some sequences derived from the transcriptome were too short for use in the phylogeny and were removed if they did not cover any of the conserved motifs. Further, pairwise distances between all sequences were first estimated by MEGA7 (Kumar et al., 2016), and sequences which were too fragmented were removed. An initial tree was then estimated by the neighbor-joining (NJ) method in MEGA7 before building the phylogeny by ML in PhyML. The final tree from PhyML was visualized using FigTree (Version 1.4.3;Rambaut, 2010). In the phylogenetic tree, three transcripts (Mq50098c0s1, Mq50098c0s12, and Mq50098c0s20) were identical but were retained in the analysis because these sequences (in translated amino acids) were different before truncation. Sequence identities were determined by BioEdit (Version 7.2.6;Hall, 1999) with BLOSUM62 similarity matrix.  (Table S1). The PCR fragments were then cloned into the pCR4-TOPO vector (Thermo Fisher Scientific), which were then introduced into TOP10 Escherichia coli cells (Invitrogen) for sequencing to ensure that the correct sequences were isolated and no PCR-related errors were introduced.

| TPS transcripts for functional characterization
Subsequently, fragments were subcloned into the pASK-IBA37+ expression vector (IBA GmbH) by reamplifying the fragments with primers designed to have overhangs specific for the pASK-IBA37+ vector (Table S1). Additionally, some of the fragments had forward primers designed at the RRX 8 W motif so that signal peptide regions were not amplified. Heterologous expression and characterization were done following the procedures described by Külheim et al. (2015). Briefly, TOP10 E. coli cells with the pASK-IBA+ expression vector were cultured at 37℃ to a density (OD 600 ) of 0.5-0.6. The expression vector was then induced with anhydrotetracycline. The cell culture was harvested after 20-h incubation (18℃) by centrifugation, which was then disrupted by four cycles of 30-s sonication at 50% amplitude (Branson Sonifier 250) while in chilled extraction buffer (50 mM Tris-HCl pH 7.5, 5 mM Na-ascorbate pH 7.0, 5 mM MgCl 2 , 5 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride, and 10% glycerol). After centrifugation at 14,000 g to separate the cell debris from the crude protein extract, the supernatant containing the extract was transferred into the assay buffer (10 mM Tris-HCl pH 7.5, 10% glycerol, and 1 mM DTT) using an Econo-Pac 10DG column (Bio-Rad Laboratories).
Terpene synthase activity assays were done by first adding 30 μl of crude enzyme extract, 13.2 ng/μl GPP or (E,E)-FPP, 10 mM MgCl 2 , and 58 μl of assay buffer into a glass gas chromatography (GC) assay tube (Macherey-Nagel). Products from the enzymes were collected using a solid-phase microextraction fiber (SPME) of 100 μm polydimethylsiloxane (Supelco)-the SPME was exposed in the headspace of the assay tube during incubation of the mixture (45 min at 35℃). The SPME was then placed into the injector of the GC (GC-2010, Shimadzu). Terpenes were separated by the GC using an EC-5 (Grace) column (5% phenyl methylpolysiloxane, with a length of 30 m, an internal diameter of 0.25 mm, and a film thickness of 0.25 μm). Splitless injection was used at an injector temperature of 220℃; hydrogen was used as the carrier gas at 1 ml min −1 . The temperature program for monoterpenes was set as follows: Hold at 50℃ for 3 min, then ramp up to 150℃ at 7℃ min −1 , followed by another ramp to 300℃ at 100℃ min −1 , and hold for 2 min. The temperature program for sesquiterpenes was as follows: Hold at 80℃ for 3 min, then ramp up to 200℃ at 7℃ min −1 , then up to 300℃ at 100℃ min −1 , and then hold for 2 min. Terpenes were identified by a mass spectrometry (MS; GCMS-QP 2010 Plus, Shimadzu) attached to the GC and "GCMS Postrun Analysis" (Shimadzu software) with the MS library "Wiley8" (Hewlett-Packard) to determine the identity of the terpenes. The two viridiflorol synthases were further validated by authentic standards (Sigma-Aldrich) as previous GC profiles gave similarly high NIST scores for both aromadendrene and viridiflorol.

| GC-MS profiling of foliar terpenes
Sixty-two M. quinquenervia plants were used for terpene profiling before, and 5 days' postinoculation (dpi), with A. psidii spores.
Terpenes from the seven plants used for RNAseq (Hsieh et al., 2017) were extracted with ethanol from ground leaf powder stored at −80℃, whereas terpenes from the remaining 110 samples from Peaks in the chromatograph were identified by comparing the mass spectra to those in the National Institute of Standards and Technology (NIST) MS library installed in the automated mass spectral deconvolution and identification system (AMDIS) software (Stein et al., 2002); compounds were also identified by comparison to peaks in published studies which had used the same chromatographic conditions (Padovan et al., 2010;Southwell & Stiff, 1990). To automate peak identification and quantification for all samples, we used easyGC (GitHub: https://github.com/dkain er/easyGC), which is an analysis pipeline using the PyMS python library (Kainer et al., 2017;O'Callaghan et al., 2012) that quantifies peaks from mass spectra in comparison with our IS (dodecane). Parameters for easyGC to call peaks were as follows: -W (widen peak width to capture more ions) was 13, -N (minimum ion count to allow NIST calls) was 3, and -M ("noisemult," the sensitivity threshold for a peak to be called) was 1.8. The parameters for cross-sample peak alignment were default, and the range of retention time for peak calling was set from 6.5 to 23 min, as this range included all peaks based on observations in manual identifications.

| Identification of induced TPS genes of M. quinquenervia in response to A. psidii infection and functional characterization
To identify terpene synthase genes (TPS) that may be inducible by A.  (Table S2) and were thus selected for functional characterization to identify the terpenes they produced.
Of the eight TPS candidates, we successfully characterized six terpene synthases. Product profiles from all six terpene synthases showed them to be dominated by one type of terpene ( Figure 2).
The terpene synthases were thus characterized as a 1,8-cineole synthase, a β-caryophyllene synthase, and two viridiflorol synthases. The first acyclic terpene synthase from the Myrtaceae family, a nerolidol synthase, was also identified. A terpene synthase (Mq36818c0seq1) producing β-ocimene (which is classified as TPS clade b2) was characterized without signal peptide and with GPP as substrate, although only traces of products were present in the GC profile ( Figure 2). The remaining two candidate TPS genes, Mq46296c0seq1 and Mq53793c0seq7, were also successfully isolated. Mq53793c0seq7 was inactive after attempts to subclone with a signal peptide-truncated version of the sequence and a change of substrate (to FPP), whereas Mq46296c0seq1 could not be verified by sequencing nor cloned into a vector.

| Amino acid alignment of functionally characterized TPS genes
Aligned amino acid sequences of functionally characterized TPS genes showed that sequences had conserved motifs indicative of terpene synthase functions ( Figure S1). Notably, the nerolidol synthase was shorter than other sequences, in that it lacked the common RRX 8 W motif. One of the two viridiflorol synthases characterized here, Mq43451c0s1_Vir, had only one amino acid difference (positioned at 328) to a previously characterized viridiflorol synthase, MqTPS1_Vir ( Figure S1).
In addition, we compared translated amino acid sequence identities between transcriptome-derived sequences and those derived from cloning to check for potential differences. Sequences from the six functionally characterized TPS genes were evaluated, and all TPS genes other than Mq43451c0seq1_Vir had amino acid sequence identities of 99.1%-100%. For Mq43451c0s1_Vir, the sequence that was cloned had an amino acid identity of 94.9% to the original transcriptome-derived sequence. Due to high amino acid similarities between members of the TPS family, which often encode F I G U R E 2 Gas chromatographic (GC) profile of terpene products identified from in vitro functional characterization of six candidate Melaleuca quinquenervia TPS genes which were induced in response to Austropuccinia psidii infection (5 dpi). Mq43451c0seq1 and Mq43451c0seq2 had the same profiles, and only the first is shown. The GC trace of terpene products from Mq36818c0seq1 is shown in comparison with the empty vector controls (EV). *Best hits: α-muurolene, germacrene D, germacrene D-4-ol; **bacterial enzyme products observed in Escherichia coli with an empty expression vector  Figure S2). Neither substrate nor GC methods used in this study allowed us to test whether this gene is a functional isoprene synthase.

| Correlation analysis of TPS gene expression changes among susceptible and resistant M. quinquenervia to A. psidii
We hierarchically clustered the overall expression pattern of putative TPS transcripts and characterized TPS genes in highly susceptible (HS) and highly resistant (HR) M. quinquenervia plants (Figure 4).
In addition, to assess whether samples with the same chemotype correlated with each other in changes of TPS gene expression after A. psidii infection, we evaluated the correlation of gene expression FC of TPS transcripts among the eight samples ( Figure S3).
Overall, we observed that M. quinquenervia plants from each chemotype expressed more of the TPS genes responsible for its dominant terpenes at 5 dpi compared with the other chemotype, regardless of resistance and susceptibility status (Figure 4). For example, Chemotype 2 plants (all highly susceptible, HS1, HS2, and HS3) expressed the 1,8-cineole synthase gene (Mq40374c0seq1_Cin) and the viridiflorol synthase gene (Mq43451c0seq2_Vir) more strongly at 5 dpi than Chemotype 1 plants, which had one highly susceptible (HS4) plant in addition to highly resistant (HR) plants ( Figure 4; Table   S2).

| Differences in terpene concentration between resistant and susceptible M. quinquenervia plants of Chemotypes 1 and 2 after A. psidii infection
We calculated the proportional increase and decrease of terpene concentrations for resistant and susceptible plants in both chemotypes. In M. quinquenervia Chemotype 1, the most abundant terpene observed was nerolidol (Table S3) (Table S3; Figure S4). Rust infection, | susceptibility, or interaction of both did not significantly alter terpenes in Chemotype 1 plants (Table 1).
In contrast, Chemotype 2 susceptible and resistant plants showed  (Table   S4; Figure 6). Overall, susceptible (SUS) plants had more monoterpenes and sesquiterpenes both before and at 5 dpi, as well as a higher proportional increase, than resistant (RES) plants (Table 2).
Both SUS and RES plants of Chemotype 2 in general contained relatively more 1,8-cineole than viridiflorol (Table 2). Rust infection significantly increased the concentration of β-caryophyllene in M.

| Integrated investigation of terpenes and TPS gene expressions from highly susceptible and highly resistant M. quinquenervia after A. psidii infection
Combining the results from terpene profiling, RNASeq differential gene expression, and functional characterization of TPS genes from the seven highly susceptible (HS) and highly resistant ( (Table S2).

| D ISCUSS I ON
The role of terpenes in defense of woody trees against fungal infection is complex partly because of the background variation in constitutive foliar terpenes within a single species. This variation is extreme within Myrtaceae with chemotypic variation occurring widely (Keszei et al., 2008;Padovan et al., 2014). Recent attempts to correlate terpene profiles with resistance to rust in Eucalyptus concluded that a deeper study of terpenes was necessary (Yong et al., 2019).

| Functional identification of terpene synthases through transcriptome and metabolite profiling
The combination of metabolite and transcriptome profiling in this study allowed us to identify induced expression of TPS genes in response to A. psidii. Six new TPS genes were functionally characterized ( Figure 2), which increased the number of existing TPS genes that had been characterized in Myrtaceae by 30%. Furthermore, some of these TPS genes showed expression profiles that correlated with metabolite data, which are rare in comparative studies of transcriptome and metabolite profiling of terpenes. We also observed that although no particular chemotype contributed resistance to A. psidii, Chemotype 2 susceptible plants contained significantly more constitutive terpenes characteristic of Chemotype 2 than Chemotype 2 resistant plants. These results contrast with those of Yong et al.
(2019), who found that combinations of several terpenes correlated to resistance in E. globulus and E. obliqua although the significant terpenes differed between the species.
Six new TPS genes have been functionally characterized in this study ( Figure 2). Terpene synthase genes from species of Myrtaceae have been shown to be difficult to functionally characterize Külheim et al., 2015). Despite Myrtaceae being a large family of around 5650 species with each species estimated to have between 25 and 120 TPS genes, only 15 of them had been functionally characterized before this study Külheim et al., 2015;Padovan et al., 2010Padovan et al., , 2017Sharkey et al., 2013).

| Myrtle rust induces TPS and alters terpene concentrations in M. quinquenervia
To our knowledge, this is the first study in Myrtaceae that showed induced expression of TPS genes with biotrophic fungal infection F I G U R E 3 Phylogenetic analysis of TPS transcripts of Melaleuca quinquenervia (pink branches) with TPS genes of Eucalyptus grandis and characterized TPS genes of M. alternifolia (green branches). The tree was constructed by maximum likelihood in PhyML with 100 bootstraps. Nodes with bootstrap values ≥95% and ≥80% are noted by black circles (•) and white diamonds (◇), respectively. The scale bar represents 0.2 substitutions per site [Colour figure can be viewed at wileyonlinelibrary.com] ( Figure 4; Table S2). Induction of TPS genes or changes of terpene concentrations in response to biotic and abiotic stress has rarely been shown in Myrtaceae (Webb et al., 2014). Apart from a recent study where some TPS-encoding genes and terpenes were observed to be upregulated in E. grandis upon challenge by Leptocybe invasa (gall wasp; Oates et al., 2015) and another study which found induced expression of isoprene synthase-and β-caryophyllene synthase-encoding genes (EgrTPS084 and EgrTPS038, respectively) in E. grandis in response to C. austroafricana infection (Visser et al., 2015), there has been little to no evidence of induction of terpenes in Eucalyptus or Melaleuca in response to wounding, herbivory, or even application of methyl jasmonate (MeJA; Henery et al., 2008). Henery et al. (2008) noted that it was possible that herbivory and MeJA could still induce TPS genes in Eucalyptus without increasing the amount of constitutive terpenes contained in the leaves due to volatile losses. This is the pattern found through studies of monoterpene synthase activity assay and terpene profiling in Pinus ponderosa, P. contorta, and Abies concolor (Litvak & Monson, 1998). However, (2018) found that headspace volatiles in M.

Bustos-Segura and Foley
alternifolia were minor in undamaged plants. Henery et al. (2008) argued that evergreen trees such as Myrtaceae are more likely to have constitutively expressed terpenes for defense against insect herbivores, whereas deciduous trees tend to use induction of terpenes for such defense. However, in contrast to Henery et al. (2008), we were able to identify multiple TPS genes that were induced in M. quinquenervia, in response to fungal attack, suggesting that these plants may have utilized induction of terpenes as one of the modes of defense.
Several TPS genes had changes in gene expression that correlated with changes in concentration of respective terpenes (Table S6) (Tables S2 and S5). One possible explanation for this discrepancy may be transcriptional variation. Nie et al. (2006) showed that correlation between transcript and corresponding protein explained <30% of variation in corresponding protein abundance through the Desulfovibrio vulgaris model and that certain functional gene categories showed greater variation. A second explanation could be that the initial enzyme product had undergone further modifications such as hydroxylation or glycosylation (Aharoni et al., 2004). For example, certain types of terpenes may be more likely to be glycosylated than other types-the study of Vitis vinifera berries by Wen et al. (2015) observed that free-form nerol was less abundant than its glycosylated-bound form (neryl glycoside) compared with linalool, where its free form was more abundant than its glycosylated form (linaloyl glycoside). Furthermore, glycosylation has been suggested to function to stabilize terpenes as inactive, nontoxic forms for plants to store at high concentrations. Plants could then react to the pathogen through reactivation of the terpene by glycosidases and de-compartmentation to bring the toxic terpene form in contact with the pathogen (Vogt & Jones, 2000;Wittstock & Gershenzon, 2002). A recent study found that two families of UDPglycosyltransferase (UGT) genes were significantly associated with terpene concentration in a genome-wide association study in E. polybractea (Kainer et al., 2019). All Eucalyptus secretory cavities contain glycosylated terpenes although these are predominately monoterpene derivatives. The two gene families identified in E. polybractea (UGT76G1 and UGT85A2) have been shown to glycosylate a wide range of terpenes in vitro (Caputi et al., 2008). Although we would expect levels of homology between Eucalyptus and Melaleuca, we did not find UGT76G1 and UGT85A2 homologues in the M. quinquenervia transcriptome.

| Chemotype did not affect resistance to myrtle rust, but higher concentration of terpenes could induce spore germination
Chemotype differences alone did not seem to affect the growth of A. psidii on M. quinquenervia ( Figure 5). This finding was consistent with a study from Florida where M. quinquenervia is an invasive weed (Rayamajhi et al., 2010). Both Chemotype 1 (dominant in nerolidol) and Chemotype 2 (dominant in viridiflorol) M. quinquenervia in Florida showed spore development of A. psidii which was classified in the genetic cluster of C4. The C4 biotype is considered as the same "Pandemic biotype" as the biotype identified in Australia through microsatellite marker genotyping (da S. Machado et al., 2015;Rayamajhi et al., 2010;Sandhu et al., 2016;Stewart et al., 2018).
We have previously shown that in M. alternifolia, chemotype had a small effect on myrtle rust susceptibility with plants containing higher levels of 1,8-cineole being more susceptible (Bustos-Segura et al., 2015). This is similar to susceptible M. quinquenervia in that Chemotype 2 produced significantly higher concentrations of total terpenes and constitutive terpenes such as viridiflorol and those in the "cineole cassette" (α-pinene, limonene, and 1,8-cineole) at control and 5 dpi compared with resistant M. quinquenervia (Table 2; Figure S2; Fahnrich et al., 2011). Rayamajhi et al. (2010) also detected higher concentrations of total terpenes and limonene, myrcene, and β-caryophyllene in susceptible M. quinquenervia compared with resistant M. quinquenervia, although it was unclear which chemotype was studied. We can speculate that higher concentrations of these constitutive terpenes in Chemotype 2 susceptible plants may have acted as a stimulant to rust spore germination. The diffusion assay study by French (1961) found that limonene strongly stimulated the germination of Puccinia graminis var. tritici (stem rust) uredospores, with an activity rating of 128 at 10 −3 dilution, which was higher than twelve other types of terpenes. α-Pinene was also rated as stimulatory (rating of 73 at 10 −3 dilution; French, 1961). Similarly, the study by Rodríguez et al. (2011) demonstrated that downregulated expression of a (+)-limonene synthase gene (CitMTSE1) in transgenic Citrus sinensis (orange) resulted in decreased limonene, which in turn made C. sinensis less likely to be infected by Penicillium digitatum (green mold), thereby suggesting that limonene accumulation may be involved in successful fungal development. In addition, Droby et al. (2008) also examined stimulatory effects of citrus volatiles and found that limonene strongly enhanced the germ tube elongation of P. digitatum, followed by α-pinene which moderately stimulated P. digitatum growth. One possible reason for the stimulatory effects of these terpenes on A. psidii, as French (1961) postulated, may be that rust uredospores required a stimulatory substrate to activate certain components necessary for germination, since the rust themselves seemed to have lipophilic self-inhibitors. Therefore, in our study, limonene and α-pinene may have promoted the germination of A. psidii uredospores. However, the molecular mechanisms of stimulatory effects of terpenes on rust spores would require further study.  (Naidoo et al., 2014) or induced defenses such as reinforcement of the cell walls as shown through induction of genes related to cell wall biosynthesis in Syzygium luehmannii (Tobias et al., 2018) and cuticular wax composition and structure in various Eucalyptus spp. (Santos et al., 2019). M. quinquenervia may utilize a combination of these constitutive or inducible defense-related factors in addition to volatile compounds to achieve effective resistance against the rust fungus A. psidii, and such a combination may not be identical among individuals of M. quinquenervia.

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
RNAseq data were deposited to the NCBI short-read archive (SRA) with the identifier SRP095052 and BioProject accession number PRJNA357284. All remaining data for this study can be found in the supplemental material.