Sexual and genotypic variation in terpene quantitative and qualitative profiles in the dioecious shrub Baccharis salicifolia

Terpenoids are secondary metabolites produced in most plant tissues and are often considered toxic or repellent to plant enemies. Previous work has typically reported on intra-specific variation in terpene profiles, but the effects of plant sex, an important axis of genetic variation, have been less studied for chemical defences in general, and terpenes in particular. In a prior study, we found strong genetic variation (but not sexual dimorphism) in terpene amounts in leaves of the dioecious shrub Baccharis salicifolia. Here we build on these findings and provide a more in-depth analysis of terpene chemistry on these same plants from an experiment consisting of a common garden with male (N = 19) and female (N = 20) genotypes sourced from a single population. Our goal in the present study was to investigate quantitative and qualitative differences in terpene profiles associated with plant sex and genotypic variation. For this, we quantified leaf mono- and sesquiterpene amount, richness, and diversity (quantitative profile), as well as the composition of compounds (qualitative profile). We found no evidence of sexual dimorphism in monoterpene or sesquiterpene profiles. We did, however, find significant genotypic variation in amount, diversity, and composition of monoterpenes, but no effects on sesquiterpenes. These findings indicated that genotypic variation in terpene profiles largely surpassed variation due to sexual dimorphism for the studied population of this species.

Terpenoids encompass a group of secondary metabolites which are often produced in high amounts in most plant tissues 1 . They are typically classified based on the number of carbon atoms of a molecule, namely: monoterpenes (C 10 ), sesquiterpenes (C 15 ), diterpenes (C 20 ), triterpenes (C 30 ), and tetraterpenes (C 40 ). Due to their low molecular weight, mono-and sesquiterpenes are highly volatile components found in scents and fragrances emitted by aromatic plants 1 . Many of these volatile compounds are considered as toxic or repellent to herbivores and pathogens 1,2 . In addition, they can also play multiple other roles in plant-insect interactions. These include attraction of predatory arthropods and parasitoids [3][4][5] , attraction of insect pollinators and seed dispersers [6][7][8] , insect-insect interactions such as co-factors for bark beetle aggregation pheromones 9 , plant-to-plant communication as warning signals to neighbouring plants of herbivore presence 10,11 , and plant protection against abiotic stresses (e.g. drought or elevated temperatures 12,13 ).
A number of studies have reported substantial variation both within and among populations in terpene quantitative profiles, mainly for shrubs and trees [14][15][16][17][18][19][20] . As for other plant defensive traits involved in herbivore resistance, studies have assessed broad-sense genetic variation in terpene levels (i.e. "genotypic" effects in ecological studies or Quantitative Trait Loci 14,16,21,22 ) and in some cases addressed specific genes or groups of genes that code for focal compounds (e.g. candidate genes 23 ). Plant sex is an ecologically important form of genetic variation in dioecious plants 24 . Dioecy is frequently characterized by the presence of sexual dimorphism in various traits 25 , which includes defensive traits associated with resistance to herbivores 26 . Female plants are expected to invest more resources into reproduction than males, such that allocation trade-offs are expected to lead to decreased growth and in turn higher investment in defensive traits relative to males 27,28 . Despite mounting evidence of sexual dimorphism in traits associated with resistance to herbivory, including plant physical defences (e.g. spines 29 , leaf toughness 30 ) and secondary chemistry (e.g. phenolic compounds 28,30 or coumarins 31 ), few studies have tested for effects of plant sex on terpenes 32,33 . In addition, although there are a number of studies measuring the effects of plant sex on quantitative variation in chemical defences, including terpenes 32,33 , fewer have tested for effects on compositional variation or assessed the effects of plant sex relative to other sources of genetic variation. As a result, the degree of quantitative or qualitative variation in chemical defences between sexes (i.e. effects of sex on population genetic structure in defences) and the contribution of plant sex to variation in defences relative to total genetic variation associated with defences or that from other sources of ecologically important genetic variation are unknown. Disentangling these different sources of variation and their degree of control over plant phenotypes is important to gain a mechanistic understanding of genetic variation underlying plant chemical defences.
Baccharis salicifolia (Ruiz & Pav.) Pers. (Asteraceae) is a woody shrub for which sex is likely genetically determined 34 . Our previous work with this plant species showed genetic variation in several traits related to growth and reproduction (e.g. flower number, relative growth rate 33 ), in the emission of plant volatile organic compounds 35 , and in arthropod abundance and composition 24,33,36 , as well as sexual dimorphism in several plant traits (more flowers and higher growth rate for females compared to males) and arthropod community composition 33 . Likewise, in a recent study we also found substantial genetic variation (but not sexual dimorphism) in leaf terpene amount for this species 33 . Here we build on these recent findings and provide a more in-depth analysis of terpene variation for these same plants from an experiment consisting of a common garden with male (N = 19) and female (N = 20) genotypes sourced from a single population of B. salicifolia 33 . Specifically, we quantify leaf monoand sesquiterpene amount, richness, and diversity (i.e. quantitative profile) as well as compound composition (i.e. qualitative profile). By replicating multiple genotypes within each sex, we are able to compare the effects of plant sex vs. those due to additional genotypic variation, and in doing so provide a unique assessment of multiple sources of genetic variation not only in quantitative but also qualitative terpene expression.
Sexual dimorphism and genotypic variation in terpene quantitative profile. We found no detectable effect of plant sex on richness, diversity or amount of either mono-or sesquiterpenes ( Table 2, Fig. 1a-f). We did, however, find significant genotypic variation in diversity and amount of monoterpenes in B. salicifolia (Table 2). Specifically, there was up to 11.7-fold and 14.6-fold variation in diversity (measured as the Shannon-Weiner index, H'; range: 0.14 ± 0.24 to 1.64 ± 0.23, Fig. 1c) and amount (range: 49.69 ± 136.98 to 725.11 ± 136.98 normalized peak area per fresh weight, Fig. 1e) of monoterpenes between plant genotypes. We found no evidence of spatial autocorrelation (e.g. clustering) of monoterpene diversity and amount, but rather these two variables were homogeneously distributed throughout the study area ( Fig. S1 in the Supplementary Material). We did not find genotypic variation in richness of monoterpenes ( Table 2, Fig. 1a), or in richness, diversity, or amount of sesquiterpenes ( Table 2, Fig. 1b,d,f).

Discussion
Our findings indicated that sesquiterpenes were overall more abundant than monoterpenes in the analyzed B. salicifolia leaf samples, but only the latter exhibited significant genotypic variation in quantitative and qualitative profiles and we did not find sexual dimorphism in either terpene group. There was significant genotypic variation in monoterpene amount, diversity and composition (but not richness), with a few noticeable compounds dominating the samples (e.g. limonene and (E)-β-ocimene). In addition, we also found no sexual dimorphism in monoterpene or sesquiterpene amount, richness, diversity and composition. It therefore appears from our comprehensive analyses that plant sex is not a relevant axis of genetic variation in terpene quantitative and qualitative profiles in B. salicifolia. This, however, does not preclude the presence of sexual dimorphism in other chemical (e.g. diterpenes, triterpenes, phenolic compounds) or physical (e.g. toughness) defensive traits of potential importance to plant-herbivore or other types of interactions in this species. Our results indicated no evidence of sexual dimorphism in terpene quantitative or qualitative measures. Theory predicts that female plants should invest more energy in reproduction and defence and less in growth relative to male plants 27,28 . Our findings do not support this prediction and add to a growing number of studies reporting inconsistent patterns with either male plants being more highly defended or no difference between sexes (reviewed by Avila-Sakar & Romanow 37 ). For example, Stark and Martz 32 found no sexual dimorphism in terpene concentration in shoots of Juniperus communis. Similarly, our previous work based on the same B. salicifolia experimental plants used here indicated weak or non-detectable sexual dimorphism in plant traits associated with growth and reproduction as well as in arthropod community structure associated with this plant 33 . Although speculative, the observed lack of sexual variation in terpene chemistry for B. salicifolia could have implications for herbivore preference or performance. For example, in another previous study we found that male B. salicifolia plants had higher abundances of the generalist aphid Aphis gossypii, whereas plant sexes did not differ in  Table 2. Summary of results from linear mixed models testing for the effect of plant sex and plant genotype nested within sex in richness, diversity (H') and amount of monoterpenes and sesquiterpenes in Baccharis salicifolia plants belonging to 39 genotypes (N = 19 males and N = 20 females). We also included the block as a random factor. F-values with degrees of freedom (numerator, denominator) and associated significance levels (P-values) are shown. Significant P-values (P < 0.05) are highlighted in bold face. www.nature.com/scientificreports www.nature.com/scientificreports/ abundance of the specialist aphid Uroleucon macolai 24 . If terpene chemistry matters for herbivore preference or performance, taken together, results from that study and our current work suggest that higher A. gossypii numbers on male plants respond to plant chemical (or physical) traits other than terpenes that potentially do show differences between plant sexes (e.g. nutrients, architecture, etc.). In the case of U. macolai, this aphid could be responsive to these compounds such that a lack of sexual dimorphism in terpene chemistry would preclude concomitant variation in this aphid's abundance, or, alternatively, this aphid could also be affected but overcomes plant sex differences in other defensive traits.
Our results showed that monoterpene diversity and amount exhibited variation among B. salicifolia genotypes. Previous work has similarly reported variation in the amount of leaf monoterpenes both between and within populations for a number of shrub and tree species [14][15][16][17][18][19]38 . Several of these studies have found that the amount of these compounds is associated with resistance to insect herbivory 1,39-41 , suggesting a defensive role in plant-herbivore interactions. On the other hand, studies reporting on intraspecific genetic variation in monoterpene diversity are much more limited. One exception is a recent study of ours where we also found that monoterpene diversity exhibited significant variation among populations of Artemisia californica distributed along a latitudinal gradient in California 19,42 . Genetic variation in terpene diversity could also be potentially important as a number of studies have shown that greater chemical diversity is associated with increased resistance against herbivores 43,44 , and may buffer populations against other sources of biotic stress (e.g. pathogens) or abiotic (e.g. temperature) stress. Although there are a number of studies reporting on phenotypic variation in secondary chemistry in the genus Baccharis (including B. salicifolia), these have not involved explicit assessments of sources of genetic variation in chemical traits [45][46][47] . In this sense, our results provide information on genotypic variation in the quantitative terpene profile for this species. The fact that we detected significant variation in these quantitative traits within a single population warrants future work assessing variation across populations and its potential biotic or abiotic correlates, as well as experimental studies investigating the influence of monoterpene amount and diversity on herbivore resistance. www.nature.com/scientificreports www.nature.com/scientificreports/ We also found significant variation in monoterpene composition among B. salicifolia genotypes. Similarly, a recent work by our group showed significant variation in monoterpene composition across populations of A. californica 19 . In addition, Thompson et al. 48 similarly reported significant variation in monoterpene composition across populations of Thymus vulgaris. It should be noted, however, that despite observing genotypic variation in monoterpene composition, we did not find evidence of distinct 'chemotypes' within the studied population as reported for other species, primarily of Mediterranean climate origin 49,50 . Genotypic variation in monoterpene composition in B. salicifolia was primarily associated with changes in the relative amounts of two major compounds (limonene and (E)-β-ocimene), which did not separate into distinct genotypic groups but rather exhibited a range of variation in relative abundances across genotypes. Previous studies have reported that these two compounds may act as repellents or toxins to herbivores in woody species 14,44,[51][52][53] . In particular, we previously found that the emission of both limonene and E-β-ocimene drastically increased after aphid herbivory 54 so unaccounted differences in herbivory on our experimental plants could have influenced (via induced responses) observed patterns of genotypic variation in constitutive terpene profiles.
We found no detectable genotypic variation in any of the quantitative or qualitative measures of variation in sesquiterpenes. With respect to quantitative measures, a number of studies have shown significant intra-specific variation in sesquiterpene amount 39,55,56 , but the magnitude of variation in these compounds appears to be lower compared to monoterpenes 14,19,57 . For example, Sampedro et al. 14 reported lower genotypic variation for sesquiterpene amount than for monoterpenes in young trees of Pinus pinaster in north-western Spain. Similarly, previous work of ours indicated that monoterpene but not sesquiterpene richness and diversity varied significantly across populations of A. californica 19 . In addition, and consistent with quantitative profiles, we found no evidence of genotypic variation in the composition of sesquiterpenes in B. salicifolia. To our knowledge, only two previous studies have tested for intra-specific variation in sesquiterpene composition and, in contrast to our study, both reported significant variation among populations 19,57 . In particular, Pratt et al. 19 found that sesquiterpene composition in A. californica significantly varied among populations distributed along the Californian coast, whereas show associated linear trends with terpenes, scaled to reflect relative magnitude of effects based on R 2 values (R 2 > 0.35, P < 0.001). This genotypic ordination displays genotypic centroids while controlling for sexual dimorphism. For monoterpenes (panel a), the PERMANOVA (controlling for the effects of sex) indicates that 23% and 25% of monoterpene composition variation is explained by male and female genotypic variation, respectively. Overall, the first two axes of ordination accounted for 45% of the genotypic variation in monoterpene composition (31% and 14% respectively). For sesquiterpenes (panel b), the PERMANOVA (controlling for the effects of sex) indicates that 20% and 16% of sesquiterpene composition variation is explained by male and female genotypic variation, respectively. Overall, the first two axes of ordination accounted for 49% of the genotypic variation in sesquiterpene composition (31% and 18% respectively). www.nature.com/scientificreports www.nature.com/scientificreports/ Moniodis et al. 57 found that sesquiterpene composition in leaves of Santalum spicatum trees significantly varied across populations distributed in arid regions of West Australia. As compared to our study of within-population variation, both of these studies assessed variation among-population, and this distinction may underlie the contrasting results.
Our findings provide an assessment of quantitative and qualitative variation in terpene profiles in B. salicifolia and its underlying genetic sources. Additional work involving multiple populations of B. salicifolia, as well as measurements of terpenes in other plant tissues (e.g. flowers), are necessary to reach stronger conclusions about sex variation in terpene chemistry as well as assess the independent effects of different sources (e.g. sexual vs. non-sexual) of genotypic variation on terpene expression. For example, there may be genes associated with variation in terpene profiles that are linked to genes that determine sex 25 , thus appearing spuriously associated. Conducting controlled crosses with different populations to produce segregating progeny would allow for a test of sex by genotype interactions to assess the linkage between sex-related and unrelated genetic variation. In addition, further work involving population variation, e.g. along ecological gradients in herbivory or abiotic variables, would provide a useful next step for identifying relevant factors associated with genetic variation in defensive chemistry in this species. Additional work could involve experimental tests of the effects of such factors on terpene expression and its consequences for insect herbivores. As a whole, the present study points at the need of assessing the independent effects of different sources of genetically-based variation, including plant sex, concurrently shaping plant defensive traits to uncover the mechanistic basis of plant defensive phenotypes. Likewise, our findings also emphasize the importance of increasing the level of detail and comprehensiveness of analyses of chemical traits putatively associated with defences to fully describe complex chemical defensive phenotypes in plants as well as their role in herbivore resistance.

Methods and Materials
Study system. Baccharis salicifolia is a perennial, dioecious shrub widely distributed from the desert southwest of the United States and northern Mexico to South America 58,59 . It is typically found in riparian areas and mesic microhabitats in high-density monospecific stands, where multiple genotypes co-occur at small spatial scales 58,60 . In coastal southern California, B. salicifolia grows and flowers predominantly during the annual winter rains, but may also flower sporadically during the spring and fall. Notably, this species emits large amounts of volatile mono-and sesquiterpenes in all tissues (leaves, stems, flowers), and previous work suggests that these compounds confer protection against herbivores 35,54,61 and abiotic stress (e.g. drought 47 ).
This study provides a detailed analysis of terpene data from Nell et al. 33 . In that study, an experimental common garden was used to characterize sexual dimorphism and genetic variation in B. salicifolia traits and plant-associated arthropod communities. Monoterpene amount was shown to vary nearly 10-fold among 39 plant genotypes, while sesquiterpene amount did not vary significantly, and there was no sexual dimorphism in either compound class. In this study, we provide a detailed analysis of these data on chemical amount, richness, and diversity (i.e. quantitative profile) as well as compound composition (i.e. qualitative profile).
Genotype selection, propagation and common garden. We used a source population of B. salicifolia occurring in 80 ha of habitat found within the University of California San Joaquin Marsh Reserve (33.66°N, 117.85°E; Orange County, CA, USA) that was also used in previous work of ours with this species 24,33,59,60,62 . In February 2008, we collected cuttings from 20 male and 20 female plants (i.e. genotypes hereafter). To maximize variation among genotypes, we collected cuttings from wild-grown plants that were separated by approximately 900 m. Cuttings were dipped in a 20% solution of Dip 'N Grow Root Inducing Concentrate (Dip 'N Grow Inc., Clackamas, OR), planted in perlite, and kept in a greenhouse for six weeks. We then planted all cuttings in 1 L pots of soil (equal parts silica sand, redwood compost, peat moss, and pumice) where they continued to grow for two months. One male genotype did not propagate successfully and was therefore eliminated from the study.
Common garden. In May 2008, we established a common garden of B. salicifolia adjacent to the Marsh Reserve. We planted 39 genotypes and replicated each genotype 8-13 times (mean 11.5 ± 0.2; total N = 459 plants). We randomly distributed plants throughout the common garden in rows and columns with 1 m spacing between them (Fig. S2 in the Supplementary Material), and we divided the garden into 12 spatial blocks to account for soil heterogeneity. We watered plants with city water using drip irrigation emitters twice a week.
Terpene analyses. In November 2011, we collected two fully expanded (undamaged) sun-exposed leaves from half of the replicates (5-6) for each genotype (N = 215 plants; Fig. S2). For terpene extraction, we immediately weighted the collected leaves and placed them in small pieces into 2 ml n-hexane (99.9% purity), sonicated them for 10 min and allowed them to soak at room temperature for seven days 19 . We then poured off the extracts and stored them at −80 °C. For the terpene analysis, we added 10 μL of an internal standard solution (0.13 μL mL −1 m-xylene in n-hexane) to 90 μL of each sample extract. We injected the samples (4 μL) onto a gas chromatograph (GC, ThermoFinnegan TraceMS+, Waltham, MA, USA) with a mass spectrometer (MS) detector that was fitted with a 30 m × 0.25 mm × 0.25 μ film thickness DB-5 fused silica column. The GC was operated in splitless mode with helium as the carrier gas (flow rate 1 mL min −1 ). The GC oven temperature program was: 1 min hold at 50 °C, 5 °C min −1 ramp to 180 °C, 20 °C min −1 ramp to 290 °C, and 1 min hold at 290 °C. The MS was operated in electron ionization mode at 70.0 eV and we collected data between 50-650 m/z. We identified mono-and sesquiterpenes using a NIST Mass Spectral Library and comparing their Kováts indices (Table 1), calculated relative to the retention times of a series of n-alkanes (C 8 -C 20 , Sigma-Aldrich, Merck KGaA, Darmstadt, Germany) analysed under the same chromatographic conditions, with those reported in the literature 46,63 . It is important to note that, although our Kováts indices matched well with those previously reported 45,63 , terpene compounds should be considered as 'putative' until confirmation with standards. For each plant, we estimated the (2019) 9:14655 | https://doi.org/10.1038/s41598-019-51291-w www.nature.com/scientificreports www.nature.com/scientificreports/ amount of mono-and sesquiterpenes by using normalized peak areas per fresh weight. The normalized peak area per fresh weight of each compound was obtained by dividing their integrated peak area by the integrated peak area of the internal standard and then dividing this value by the leaf fresh weight. To assess the relative abundance of terpenes across plant genotypes and sexes, we also calculated mono-and sesquiterpene diversity for each plant using the Shannon-Weiner index: H' = −Σ(P i log[P i ]), where P i is the relative amount of a given terpene divided by the total terpenes in each plant. Finally, we also recorded the total number of mono-and sesquiterpene compounds (i.e. richness).
Statitical analyses. Sexual dimorphism and genotypic variation in terpene quantitative profiles. We ran linear mixed models including plant sex and plant genotype nested within sex as fixed factors to test for sexual dimorphism and genotypic variation in richness, diversity and amount of monoterpenes and sesquiterpenes (i.e. quantitative profile). We also included block as a random factor. We ran all analyses with PROC MIXED in SAS 9.4 (SAS Institute, Cary, NC) 64 . We log-transformed all variables to achieve normality of residuals, and reported least square means ± SE in the original (untransformed) scale as descriptive statistics.
Sexual dimorphism and genotypic variation in qualitative terpene profiles. We tested for the effect of plant genotype on mono-and sesquiterpene composition (i.e. qualitative profile) separately using data on the relative amount of individual compounds for each type of terpene. We used a permutational multivariate analyses of variance (PERMANOVA) 65 including plant genotype as a fixed factor, constrained by plant sex to control for any effects of sexual dimorphism. A PERMANOVA is analogous to an ANOVA, but partitions similarity matrices between treatments and uses permutation tests with pseudo F-ratios. The PERMANOVA was based on 10,000 permutations using the 'vegan' package 66 in R software 67 . To visualize the results of this analysis, we used pairwise Bray-Curtis dissimilarities as input to a principal coordinates analysis. The result of this analysis was then visualized in two dimensions, where each point reflected the genotype centroid. We selected influential terpenes based upon R 2 > 0.35 (P < 0.001) for associations with the first two ordination axes (using 'envfit' in vegan) and displayed using biplot arrows with length scaled to R 2 values.
We used the same procedures described previously using PERMANOVA to test for sexual dimorphism on terpene composition (qualitative profile) using genotype least square means. We visualized sexual dimorphism (with ordination) in terpene composition with the two sex centroids as well as the mean values for each male and female genotype displayed on the ordination plot. We selected influential terpenes based upon R 2 > 0.60 (P < 0.001) for associations with the first two ordination axes.