Exploring geography and evolutionary history as drivers of variation in floral scent chemistry in western sessile-flowered Trillium using parsimony-constrained phylogenetics

Candela Blanco-Moreno; Kjirsten A Wayman; Alexandru M F Tomescu

doi:10.1093/aob/mcae120

. 2024 Jul 30;134(5):843–862. doi: 10.1093/aob/mcae120

Exploring geography and evolutionary history as drivers of variation in floral scent chemistry in western sessile-flowered Trillium using parsimony-constrained phylogenetics

Candela Blanco-Moreno ^1,^2,^✉, Kjirsten A Wayman ³, Alexandru M F Tomescu ⁴

PMCID: PMC11560366 PMID: 39078941

Abstract

Background and Aims

The sessile-flowered Trillium species from western North America have been challenging to distinguish morphologically owing to overlapping characters and intraspecific variation. Molecular phylogenetic analyses, currently inconclusive for this group, have not sampled multiple populations of the different species to account for this. Here, we query the diversity of floral volatile composition to understand its bearings on the taxonomy, distribution and evolution of this group.

Methods

We explored taxonomic and geographical patterns in average floral volatile composition (105 different compounds) among 42 wild populations of four sessile-flowered Trillium species and the outgroup, Pseudotrillium, in California, Oregon and Washington, by means of parsimony-constrained phylogenetic analyses. To assess the influence of character construction, we coded compound abundance in three different ways for the phylogenetic analyses and compared the results with those of statistical analyses using the same dataset and previously published statistical analyses.

Key Results

Different codings of floral volatile composition generated different phylogenetic topologies with different levels of resolution. The different phylogenies provide similar answers to taxonomic questions but support different evolutionary histories. Monophyly of most populations of each taxon suggests that floral scent composition bears phylogenetic signal in the western sessile-flowered Trillium. Lack of correlation between the distribution of populations and their position in scent-based phylogenies does not support a geographical signal in floral scent composition.

Conclusions

Floral scent composition is a valuable data source for generating phylogenetic hypotheses. The way in which scent composition is coded into characters is important. The phylogenetic patterns supported by floral volatile compounds are incongruent with previously reported phylogenies of the western sessile-flowered Trillium obtained using molecular or morphological data. Combination of floral scent data with gene sequence data and detailed morphological data from multiple populations of each species in future studies is needed for understanding the evolutionary history of western sessile-flowered Trillium.

Keywords: Floral scent, floral volatiles, gas chromatography–mass spectrometry, intraspecific variation, interspecific variation, Melanthiaceae, parsimony phylogenetics, taxonomy, sessile Trillium, western Trillium

INTRODUCTION

The genus Trillium is native to North America and eastern Asia. Its taxonomy is in flux (e.g. a prior taxonomic history as a genus in the Liliaceae; Zomlefer et al., 2001) owing to the recent discovery of rare, endemic species and the reclassification of established species in segregated genera. Currently, there are ~50 recognized species, with the greatest diversity and majority of them in North America (Meredith et al., 2022). All species of Trillium produce a solitary three-part flower and are classified into four subgenera (Lampley et al., 2022): three subgenera that include pedicellate-flowered species (subgen. Trillium, Delostylis and Callipetalon) and a fourth subgenus, Sessilia, that includes all the species in which the flower lacks a pedicel, which occur only in eastern and western North America.

In western North America, seven taxa of sessile-flowered Trillium, including subspecies and varieties, are currently recognized (Freeman, 1975; Case and Case, 1997; Case, 2002; McNeal and Ness, 2012; Meyers et al., 2015): Trillium angustipetalum (Torr.) J.D. Freeman, Trillium albidum J.D. Freeman, T. albidum subsp. parviflorum (Soukup) Chambers & Meyers (sometimes elevated to Trillium parviflorum V.G. Soukup), Trillium chloropetalum (Torr.) Howell (var. chloropetalum and var. giganteum), Trillium kurabayashii J.D. Freeman (which is sometimes included with T. angustipetalum) and Trillium petiolatum Pursh. The last of these, although a western sessile-flowered species, is not considered hereafter, because it is morphologically distinct from all other western sessile Trillium, and its floral scent composition is also distinctly different (K. A. Wayman, California State Polytechnic University Humboldt, USA, unpublished results). Additionally, T. petiolatum has been recovered consistently outside the clade that includes the other western sessile-flowered Trillium species (Kawano and Kato, 1995; Osaloo and Kawano, 1999; Farmer and Schilling, 2002; Schilling et al., 2019; Lampley, 2021). Thus, hereafter reference to sessile-flowered Trillium from western North America (western sessile Trillium) does not include T. petiolatum.

Five phylogenies of the genus Trillium that include the western North American sessile species have been published, two based on morphological characters (Kawano and Kato, 1995; Farmer and Schilling, 2002) and three on molecular data (Osaloo and Kawano, 1999; Schilling et al., 2019; Lampley, 2021). Molecular analyses recover the western sessile Trillium species as a monophyletic group, whereas morphological analyses suggest that they are a polyphyletic group. These species vary in petal colour, petal shape, ovary colour, stamen colour, stamen orientation, ratio of stamen length to ovary length, and leaf mottling. However, considerable intraspecific variation and overlapping traits among taxa make them a challenge to distinguish morphologically. Thus, chemical characters, such as those found in their sweet, fruity or spicy floral odours, might be useful where morphological differentiation is lacking.

Floral scents can be simple to complex blends of volatile organic compounds produced by diverse biosynthetic pathways. Among the typical compounds discovered in plant volatiles are aliphatic compounds, terpenes (terpenoids), aromatic compounds, nitrogen-containing compounds and organosulfur compounds (Knudsen et al., 2006). Similar to morphological features, the volatile organic compounds in floral scent represent an important set of characters with the potential to be used to describe and identify species. For instance, plant secondary compounds (including floral volatiles) can provide useful taxonomic markers, such as compounds or blends of compounds unique to particular species (Levin et al., 2003; Huber et al., 2005; Chess et al., 2008; Wayman et al., 2010; Jaeger et al., 2016). However, use of floral volatile composition for taxonomic purposes needs to take into account the ranges of intraspecific variation and interspecific overlap.

Intraspecific variation in floral scent has been attributed to geographical barriers to gene flow, pollinator-mediated selection or other ecological or environmental pressures (Delle-Vedove et al., 2017). Such variation has been documented in many flowering plants, including Gymnadenia conopsea (diel variation and geographical variation; Chapurlat et al., 2018), Oenothera harringtonii (clinal variation; Skogen et al., 2022), Abronia umbellata (related to mating system shifts; Doubleday et al., 2013) and Linanthus dichotomus (related to pollinator-mediated selection; Chess et al., 2008). Other factors have also been shown or suggested to play a role in the types and emission of floral volatiles and explain geographical variation within species: drought stress for Ipomopsis aggregata (Campbell et al., 2019); environmental and evolutionary changes for Arabis alpina (Petrén et al., 2021); and genetic drift in Lithophragma (Friberg et al., 2019).

Here, we explore, for the first time, the patterns of variation in floral scent chemistry in western sessile Trillium using parsimony-constrained phylogenetic analyses: (1) to assess the extent and the most appropriate ways in which floral scent chemistry can be used to construct informative phylogenies; and (2) to explore the potential causes of variation in floral scent chemistry. This approach is different from that adopted in a study by Wayman et al. (2023), who analysed western sessile Trillium floral scent chemistry (based on the same data as our study, but treated differently) using statistical methods (perMANOVA, non-metric multidimensional scaling and a random forest algorithm). That study found that scent compositions are consistent with current taxonomy of the western sessile Trillium species and can provide insight in cases where species delimitations are difficult based on morphology. To gain deeper insight into the two main aspects we address in this study, we test the results of phylogenetic analyses against the results of two sets of statistical analyses: the analyses published by Wayman et al. (2023), for which we provide a direct empirical counterpart, given the use of the same data; and a new set of analyses reported here.

The context of our approach, in terms of previously published analyses, is notably scarce in closely similar studies. Phylogenies constructed only from scent data are not common, and this is especially true for floral scent data. Most of the published research on floral scent compounds that involves phylogenies consists of mapping them directly on phylogenies generated using other data (e.g. Schiestl, 2010) or using the phylogenetic information as another source of data for statistical analyses (Prieto-Benítez et al., 2016). The dearth of phylogenies constructed only from scent data might be attributable to the high potential for homoplasy in floral scent chemistry. Because floral scent serves an adaptive function (in pollination or anti-herbivory) while also being potentially influenced by the geography of populations (e.g. soil chemistry, climate variables) and possibly reflecting inheritance of regulatory mechanisms for biosynthetic pathways, the weights and effects of all these factors are difficult to untangle. Among previous phylogenetic studies that have used scent data, some have constructed characters based on the synthesis pathways of scent compounds (Barkman, 2001), others have constructed characters by a combination of presence/absence or quantitative presence, depending on the compound (Williams and Whitten, 1999), and yet others have partitioned the scent composition into different numbers of discrete character states (Levin et al., 2003). Statistical analyses using floral scent composition are also infrequent (Friberg et al., 2013, 2019). Nevertheless, aside from the analyses in Trillium published by Wayman et al. (2023), statistical analyses have been used successfully in studying the scent of Antirrhinum species (Weiss et al., 2016).

In exploring the usefulness of floral scent chemistry in phylogenetic studies and potential causes of variation in floral scent chemistry, we address a series of specific questions.

Does scent bear phylogenetic signal in the western sessile Trillium and, if so, what is the unit of taxonomic importance for scent? Alternatively, does scent bear a geographical signal?
What are the best ways to construct characters for phylogenetic analyses based on floral scent chemical data such that they reflect taxonomic and evolutionary relationships?
Are floral scent chemistry-based phylogenies congruent with those based on gene sequences or morphology?
How do phylogenetic and statistical approaches using floral scent chemistry compare in their ability to address taxonomic challenges in western sessile Trillium?
What do phylogenies based on floral scent chemistry tell us about evolutionary relationships in western sessile Trillium?

MATERIALS AND METHODS

The study species, study sites and populations and the data collection procedures that produced the sample sets of scent composition used in our analyses are the same as those described by Wayman et al. (2023), apart from the additional Pseudotrillium rivale population.

Study species

The six taxa of western sessile Trillium studied, including T. albidum [subsp. albidum (TRAL) and subsp. parviflorum (TRAL/PA and TRPA)], T. angustipetalum (TRAN), T. chloropetalum [var. chloropetalum (TRCH/CH), var. giganteum (TRCH/GI)] and T. kurabayashii (TRKU) differ in their morphological traits, floral odours and geographical ranges. All species have strong floral fragrances that can be detected by nose in the field and generally have sweet odours reminiscent of banana, lemony citrus or rose (sometimes in combination with one another), occasionally with notes of cinnamon or clove (K. A. Wayman, California State Polytechnic University Humboldt, USA, personal observation).

Trillium albidum s.l. occurs over the largest geographical range, from central California to Washington, overlapping in its range with each of the other species (Fig. 1). It is the only white-petalled sessile Trillium in North America, except for uncommon white-petalled variants of T. chloropetalum. The variable morphology of T. albidum s.l. throughout its range (Fig. 2A–D) suggests that there might be multiple species included within the taxonomic concept of T. albidum. Consequently, it has received the most attention in morphological studies, resulting in the separation of two subspecies, T. albidum subsp. albidum (hereafter T. albidum) and T. albidum subsp. parviflorum (Bledsoe, 1993; Chambers and Meyers, 2011; Fig. 2D); some workers have proposed segregation of the latter as a separate species, T. parviflorum (Soukup, 1980; Case, 2002).

Fig. 1. — Geographical distribution of sites of sampled populations for each western sessile *Trillium* taxon in California, Oregon and Washington. Each site is labelled with its location code.

The rest of the study species have maroon petals. Identifying a red- or maroon-flowered sessile Trillium, specifically T. angustipetalum (Fig. 2E), T. kurabayashii (Fig. 2F) or T. chloropetalum (Fig. 2H–L) to species can be problematic. Petal shape, ratio of stamen length to ovary length, and stamen orientation, which are the key characters to distinguish the species, tend to overlap among them, and one often defaults to using locality. Trillium angustipetalum occurs mainly in two disjunct regions, the Sierra Nevada foothills and the southern central Coast Ranges of California (San Luis Obispo and Santa Barbara Counties, the southernmost occurrence of Trillium in western North America), T. chloropetalum is concentrated predominantly around the San Francisco Bay Area in California, and T. kurabayashii is known mainly from the border region of north-western California and south-western Oregon. Trillium angustipetalum and T. kurabayashii are sometimes treated as the same species (McNeal and Ness, 2012).

Trillium chloropetalum plants display the greatest variation in petal morphology (shape and colour), even within a single population (Fig. 2H–L). The petals of T. angustipetalum and T. kurabayashii are more consistently red–maroon, with the occasional yellow-petalled variant, which is likely to be a pigment-free mutant (Whitkus et al., 1987; Les et al., 1989). In contrast, a single population of T. chloropetalum can have green, yellow, bronze, brown and maroon petal colours for T. chloropetalum var. chloropetalum (Fig. 2J–L) or white, pink, lavender and maroon petal colours for populations of T. chloropetalum var. giganteum (Fig. 2H, I), leading to speculation that they might be the result of hybridization. At sites where T. albidum and T. chloropetalum overlap in geographical range, a white-petalled plant is usually identified based on the appearance of the ovary and stamens (T. albidum has wider, straight stamens and greenish-white ovary and anther connective tissue; T. chloropetalum has narrower, curved stamens and maroon ovary and anther connective tissue).

Pseudotrillium rivale (PSRI), a recent segregate from Trillium (Farmer and Schilling, 2002) and, thus, closely related to the genus, was used as the outgroup in the phylogenetic analyses. Trillium petiolatum, a western species that is more closely related to the western sessile Trillium than Pseudotrillium, would theoretically provide another appropriate outgroup for polarizing characters in cladistic analyses. However, we decided against using T. petiolatum as an outgroup because its floral scent chemistry is incompletely characterized currently (K. A. Wayman, California State Polytechnic University Humboldt, USA, unpublished results), and what is known about it suggests that there are no compounds in common with the floral scent chemistry of the ingroup species, making it a poor choice for polarizing characters based on scent chemistry. Additionally, use of Pseudotrillium as an outgroup will allow for direct comparisons of phylogenetic results between the analyses presented here and future expanded analyses that would sample other Trillium species, including T. petiolatum.

Sampling sites and populations

Petal samples were collected from naturally occurring populations that range from as far south as San Luis Obispo County in California to as far north as Clark County in Washington (Fig. 1; Supplementary Data File S1). The westernmost sites extend along the Pacific Coast, and the easternmost sites extend along the western slopes of the Sierra Nevada Range. Trillium albidum was sampled from populations across California and Oregon (two populations sampled in northern Oregon might be intermediate between the two subspecies), and T. albidum subsp. parviflorum was sampled from Washington state. Samples of T. angustipetalum were collected from the Sierra Nevada foothills and the Central Coast of California. Trillium kurabayashii was sampled from the northern California coast, the Klamath Mountains and the Oregon Coast. Samples of T. chloropetalum var. chloropetalum and T. chloropetalum var. giganteum were collected from sites along the Central Coast of California. Maroon-petalled plants sampled from two sites in Napa County (NAPA) have uncertain identity as either T. angustipetalum or T. chloropetalum (Fig. 2G). Pseudotrillium rivale, a narrow endemic present only in the Siskiyou Mountains, was sampled from a population in Josephine County, Oregon. Voucher specimen information is the same as that published by Wayman et al. (2023), with the addition of Pseudotrillium specimens, and can be found in Supplementary Data File S1.

Sample collection

Plants were sampled during their peak scent-producing period, within 1–6 days of their initial blooming. Visual cues, such as petal condition and anther dehiscence, were used to determine flower stage in the field; flowers with dry/curled petals or anthers empty of pollen, insect-eaten or browned were not sampled. Sampling occurred during the months of February–June, from 2012 to 2022. At each population, petal samples from 4–27 individuals were collected, depending on population size and phenology (available plants at the correct stage), and to include the full range of morphological variation, when present. From an individual in the field, one to three petals were removed at their base and placed into the same glass vial (20–25 mL). The vials were stored in a cooler with ice or icepacks until they could be transferred to a refrigerator. For samples collected in 2012 and 2016, the flower was cut at the base of the stem and held in cold storage until the time of analysis. Samples were kept cold for ≤4 days until they could be analysed, typically within 2 days of collection.

Absorption of volatiles with solid-phase microextraction

After a petal was removed from a vial in cold storage, the petal was placed in a beaker or vial of appropriate size and enclosed with heavy-duty aluminum foil (If You Care brand, Source Atlantique, Englewood Cliffs, NJ, USA; foil was chosen owing to its lack of emission or absorption of volatile organic compounds). When more than one petal was collected from an individual owing to its smaller size, all petals were placed in the same vial for analysis. The petals remained enclosed, and volatiles were allowed to equilibrate at room temperature for 1.5–3 h. For samples analysed in 2012 and 2016, the whole flower, including leaves, was removed from the stem, placed in a beaker and covered with parafilm. After the minimum enclosure period, a solid-phase microextraction apparatus equipped with a polydimethylsiloxane fibre (100 μm, 24 gauge; Supelco, Bellefonte, PA, USA) was inserted into the enclosure to absorb the volatile organic compounds emitted from the petal(s). After a 30 min absorption period, the fibre was inserted into the injection port of a gas chromatograph–mass spectrometer (described next).

Gas chromatography–mass spectrometry analysis

A Hewlett-Packard HP G1800C (Hewlett-Packard, Palo Alto, CA, USA) was used to analyse the floral scent composition. The gas chromatograph was equipped with a Thermo Scientific TR-5MS column (Thermo Fisher Scientific, Waltham, MA, USA; non-polar phase, 5 % phenyl polysilphenylene-siloxane, length 30 m, inner diameter 0.25 mm, film thickness 0.25 μm). Samples were injected manually using a splitless injection, with inlet and detector temperatures set at 205 and 280 °C, respectively, and a helium carrier gas flow rate of 1 mL min⁻¹. The gas chromatograph oven temperature was held initially at 45 °C for 4 min, then increased at a rate of 10 °C min⁻¹ to 225 °C, then held for 1 min. Mass fragments from electron ionization were detected with mass ranges from m/z 39–450.

The gas chromatography–mass spectrometry data were processed by the Hewlett-Packard ChemStation software (v.B.02.05). The relative amount of each component was determined as a percentage of the total ion current. Identification of individual components was assisted using the NIST mass spectral search program (v.1.6) and the NIST 98 database (NIST/EPA/NIH Mass Spectral Library) and confirmed by comparison of their gas chromatography retention time and mass spectrum with authentic standards (for a complete list of compounds and associated information, see Supplementary Data File S2). When standards were not available, identifications were based on comparison of retention times with published data (Adams, 1995; El-Sayed, 2021), along with comparison of mass spectra with the NIST mass spectral library, either NIST 98 or NIST Chemistry WebBook online (NIST Mass Spectrometry Data Center and Wallace, 2022).

Calculation of floral scent composition per population

For a given population and in a given year, petal specimens of multiple individuals (n = 4–27) were collected and analysed separately to determine the chemical composition of their floral scent. Between the ingroup and outgroup taxa, a total of 105 distinct compounds were identified (of which 13 were present only in Pseudotrillium), although not all compounds were present in each sample. The gas chromatography–mass spectrometry total ion chromatogram peak areas of each individual were used to determine the percentage of each compound in its mixture of floral volatiles. Subsequently, the scent composition data acquired from individuals in each population were averaged to provide the ‘population sample’ for each population of this study. Therefore, each population sample represents the average percentages of all compounds in the combined floral scent mixtures of all individuals sampled in that population. Some populations were sampled in two years or more, in which case the sampling was conducted in different regions of that population in different years, to avoid double-counting individuals; data collected in different years were kept separate and treated as distinct sample sets. This resulted in a total of 51 sample sets representing the 42 populations sampled (Supplementary Data File S1).

The data processing protocol is different from that implemented by Wayman et al. (2023), in that each sample included in our phylogenetic and statistical analyses reflects the averaged floral scent composition over all individuals per population (totalling 51 sample sets), whereas the non-metric multidimensional scaling analysis by Wayman et al. (2023) was performed using the raw data of distinct individuals (totalling 600 sample sets).

Phylogenetic analyses

Character construction.

We used the floral scent composition to construct three sets of characters (Table 1; for character matrices, see Supplementary Data File S3) used in three separate analyses: one using ‘binary’ character coding, i.e. compounds scored exclusively as present or absent; one using ‘dominance’ character coding, i.e. compounds scored as absent, present non-dominant or dominant (this state applies only to the compound that is most abundant in each sample set); and one using a ‘ranges’ coding scheme, in which compounds were scored based on their abundance between 0 and 100 % along a partition of ten 10 % bins (Table 1).

Table 1.

Character state coding used for the phylogenetic analyses.

Character state	Binary	Dominance	Range
0	Compound is absent	Compound is absent	Compound is absent
1	Compound is present	Compound is present but not dominant	Compound composes ≤10 % of the scent
2	–	Compound is present and dominant	Compound composes 10–20 %
3	–	–	Compound composes 20–30 %
4	–	–	Compound composes 30–40 %
5	–	–	Compound composes 40–50 %
6	–	–	Compound composes 50–60 %
7	–	–	Compound composes 60–70 %
8	–	–	Compound composes 70–80 %
9	–	–	Compound composes 80–90 %
10	–		Compound composes >90 %

Open in a new tab

Additional character sets (not shown, but tested in phylogenetic analyses) include: two different multistate codings based on the contribution of each compound [(A) 0 = absence; 1 = 1 %; 2 = 1–5 %; 3 = 5–20 %; 4 = 20–50 %; 5 = >50 %; and (B) 0 = absence; 1 = <1 %; 2 = 1–5 %; 3 = 5–20 %; 4 = 20–50 %; 5 = 50–75 %; 6 = >75 %]; (C) a multistate coding based on the dominance such that each compound was scored zero if absent in a sample set, one if present in a sample set, and two only in the sample set in which it was most abundant; and (D) a continuous coding based on the gap-weighting method proposed by Thielle (1993). These analyses produced unresolved trees or trees of very low resolution [(A) 2576 MP (Most Parsimonius) trees, L (number of steps) = 549; (B) 476 MP trees, L = 502; (C) 6393 MP trees, L = 398; and (D) 1432 MP trees, L = 550] or were deemed not to represent the data in a correct way based on the original dataset (the case of C), given that the original dataset is based on percentages of the compound within the sample and not on absolute abundances.

Taxa.

We included 50 ingroup terminals plus the outgroup Pseudotrillium rivale (PSRI), as follows: 16 populations of Trillium albidum (TRAL), five of which were sampled in different years, totalling 22 terminals; two populations of uncertain identity as either T. albidum or T. albidum subsp. parviflorum (TRAL/PA); two populations of T. albidum subsp. parviflorum (TRPA); four populations of T. angustipetalum (TRAN); seven populations of T. kurabayashii (TRKU), two of which were sampled in two different years, therefore totalling nine terminals; three populations of T. chloropetalum var. chloropetalum (TRCH/CH); five populations of T. chloropetalum var. giganteum (TRCH/GI); and two populations of maroon-petalled plants sampled from two sites in Napa County (NAPA) that have uncertain identity as either T. angustipetalum or T. chloropetalum, one of which was sampled in two different years, totalling three terminals.

Analyses.

Three different phylogenetic analyses were performed under maximum parsimony constraint using TNT (Goloboff et al., 2008) software, one for each of the three character sets. All characters were treated as unordered and were analysed using the traditional search function of TNT (Wagner trees), set to a random seed of ten, 1000 replications, and using the tree bisection–reconnection (TBR) branch-swapping algorithm set to save 100 trees by replication. A strict consensus based on all the resulting maximum parsimony trees was computed for each analysis.

A second set of three phylogenetic analyses were performed using TNT under the same parameters as for the scent trees, which included five morphological characters derived from direct observation of the populations: floral colour (no pigment, some pigment, variable pigment and fully pigmented); petal shape (ovate, lanceolate, oblanceolate, obovate and linear); stamen orientation (vertical, splayed, curved or adpressed); colour of ovary, anther, connectives and filaments (no pigment, some pigment, variable pigment and fully pigmented); and stamen length to ovary length ratio (stamens much longer than ovaries, stamens longer, stamens slightly longer and stamens almost equal to ovaries) (Wayman et al., 2023; for matrix, see Supplementary Data File S3). The three analyses in this second set had the following character samplings: (1) an analysis using only the five morphological characters; (2) an analysis using the five morphological characters plus the scent data in binary coding; and (3) an analysis using the five morphological characters plus the scent data in dominance coding. The resulting trees are very poorly resolved (for resulting trees, see Supplementary Data File S4, Figs S1–S3), and these results are mentioned only briefly in the Discussion, as extra data for the comparison among phylogenetic hypotheses.

Statistical analyses

A principal component analysis (PCA) based on a variance–covariance matrix and a cluster analysis based on UPGMA (Unweighted Pair Group Method with Arithmetic mean)clustering with Euclidean distance were performed using the original dataset of floral scent composition expressed as percentages of each compound. Both analyses were run using PAST software (Hammer et al., 2001). These statistical methods are different from those applied by Wayman et al. (2023), who used non-metric multidimensional scaling as the ordination technique, instead of a PCA, and randomForest (Liaw and Wiener, 2002) to determine the most important floral scent compounds to distinguish species. By analysing the data by means of a PCA, we obtained a taxonomic ordination that represents the differences between species and the weight of each compound in the resulting ordination in the same analysis. This allowed us to identify relevant compounds for each taxon.

The phylogenetic results were compared with these two statistical analyses. For the PCA results, the compounds with loadings of >0.1 on at least one of the first two principal components (PCs) were mapped on the strict consensus trees obtained in the binary coding analysis and the dominance coding analysis, at the nodes where they provided synapomorphies. For the cluster analysis, the clusters were colour-coded on the strict consensus trees obtained in the binary coding analysis and the dominance coding analysis.

It is important to note that several studies based on scent chemistry have taken into account correlations between scent compounds underpinned by biosynthetic pathway dynamics, to reduce the dimensionality of datasets and inform the construction of reduced sets of scent-based phylogenetic characters (e.g. Williams and Whitten, 1999; Barkman, 2001; Levin et al., 2003). Outside the realm of phylogenetics, population ecology studies have used datasets of reduced dimensionality arrived at by considering compound correlations within biosynthetic pathways (Schiestl et al., 2011; Eisen et al., 2022). We did not take the approach of reducing the dimensionality of the dataset, because of a number of considerations. First, the biosynthetic pathways for most of the specific compounds that form the Trillium floral scent or the potential interactions of such pathways are not well characterized beyond a general level of understanding. Second, one-to-one correspondence between floral volatile compounds and biosynthetic pathways is not the rule in floral scent chemistry, because numerous documented cases indicate that the same compounds can be produced by different pathways, which have originated independently several times over the course of plant evolution (Dötterl and Gershenzon, 2023). Additionally, typical patterns of compound co-occurrence might not be reciprocal in all cases, which can skew the assumptions of correlation; see Knudsen et al. (2006) for a plethora of data. For example, although α-humulene and β-caryophyllene are often associated in floral scents, where β-caryophyllene is dominant, they can sometimes be present independently of each other; sometimes β-caryophyllene is present in the absence of α-humulene or, more rarely, vice versa (at least insofar as the sensitivity of current detection methods allows for documenting the presence of these scent compounds). Likewise, 6-methyl-5-hepten-2-one, typically occurring in association with lesser amounts of 6-methyl-5-hepten-2-ol, is sometimes detected in the absence of the latter (e.g. Supplementary Data File S3).

Given that some scent compounds might be dependent on others produced in the same biosynthetic pathways, treating them as independent characters might violate a core assumption of cladistic methods. However, knowledge about the specific biosynthetic pathways that produce many of the western sessile Trillium floral scent compounds is very spotty. Additionally, the combination of all the factors discussed above introduces very high levels of uncertainty about scent compound dependence. Consequently, we consider that attempting to reduce the dimensionality of our scent chemistry dataset based on scent compound dependence hypotheses would introduce assumptions whose level of support is unknown. Therefore, the most conservative approach, minimizing potentially unsupported assumptions, is to define separate characters for each scent compound.

RESULTS

Phylogenetic analyses

The phylogenetic analyses based exclusively on scent data recovered different patterns of relationships and different amounts of resolution under the different character coding schemes (Table 2; Figs 3 and 4; Supplementary Data File S4, Fig. S4). All three analyses resolved several groups as monophyletic (e.g. Fig. 4), as follows: (1) all nine sample sets of T. kurabayashii (TRKU); (2) all four T. angustipetalum sample sets plus the three unidentified NAPA sample sets (TRAN + NAPA); (3) the two sample sets of T. albidum subsp. parviflorum plus the two unidentified sample sets of either T. albidum or T. albidum subsp. parviflorum (identified as TRPA); (4) the six T. chloropetalum (TRCH) sample sets exclusive of the GAZO and WADD populations. Additionally, the four T. albidum sample sets from the Marble Mountains (TRAL-MM) also form a monophyletic group in the trees obtained using the binary and dominance coding schemes. When the four clades above are collapsed into a single terminal each, the strict consensus obtained under the 10 % ranges coding is a complete polytomy, based on 824 maximum parsimony trees (L = 412). The binary and dominance character coding schemes resulted in strict consensus trees that are better resolved but support reversed topologies. The T. albidum sample sets form a paraphyletic grade (that also includes T. albidum subsp. parviflorum) basal to the other sample sets in the binary tree (Fig. 3A) but are recovered as polyphyletic in the dominance tree (Fig. 3B). Trillium kurabayashii is sister to all other western sessile Trillium in the dominance analysis, but is the most derived, sister to T. angustifolium + NAPA sample sets, in the binary analysis (Fig. 3A, B). The latter group is sister to a clade including all others minus T. kurabayashii (i.e. T. albidum and T. chloropetalum), in the dominance tree (Fig. 3B).

Table 2.

Results of the phylogenetic analyses.

Analysis	Number of characters	Number of most parsimonious trees	Best score	Consistency index	Retention index
Binary	105	32	331	0.314	0.705
Dominant compound	105	2	351	0.330	0.702
10 % ranges	105	824	412	0.367	0.675

Open in a new tab

Fig. 3. — Comparison of tree topologies on simplified strict consensus trees with main character state changes mapped. (A) Tree obtained from the analysis of binary coding. (B) Tree obtained from multistate coding based on dominant compound. ‘*Trillium parviflorum*’ includes the two sample sets identified as *Trillium albidum* subsp. *parviflorum* plus two sample sets from northern Oregon that cannot be assigned unambiguously to *T. albidum* or *T. albidum* subsp. *parviflorum*. GAZO and WADD refer to specific *Trillium chloropetalum* populations. Symbols on branches signify changes in state for apomorphic characters for the clade: white, change from presence to absence; grey, change from absence to presence; black, change to dominant compound.

Fig. 4. — Comparison between phylogenetic and cluster analyses. (A) Strict consensus tree of the 32 most parsimonious trees obtained using binary coding of the characters based on presence or absence of compound. (B) Strict consensus tree of the two most parsimonious trees obtained using multistate coding of the characters based on the dominant compound. In both trees, sample sets are colour coded based on the clusters in Fig. 6. The first acronym refers to the species: PSRI, *Pseudotrillium rivale*; TRAL, *Trillium albidum*; TRPA and TRAL/PA, *T. albidum* subsp. *parviflorum* and the unidentified sample sets of either *T. albidum* or *T. albidum* subsp. *parviflorum*; TRCH, *Trillium chloropetalum*; TRAN + NAPA, *Trillium angustipetalum* with the NAPA unidentified specimens; TRKU, *Trillium kurabayashii*. The second acronym refers to the specific population, and numbers at the end refer to the year when the sample set was obtained. TRAL-MM refers to the sample sets of the populations from the Marble Mountains.

Statistical analyses

The first two PC axes separate groups that are congruent with the monophyletic groups supported by the phylogenetic analyses (Fig. 5); the third PC does not add any resolution among the data. Trillium angustipetalum sample sets with the NAPA unidentified sample sets are grouped towards high positive values of PC1 (PC1) and are characterized by a scent composition with high isobutyl acetate percentage. Trillium kurabayashii sample sets are mostly grouped towards the most negative values of PC2, and their scent is characterized by a high percentage of 6-methyl-5-hepten-2-one. Trillium albidum sample sets are grouped towards negative values of PC1 and show a continuous gradient along PC2 from high positive to negative values, and present decreasing percentages of citronellol, geraniol and benzyl alcohol in their scent composition. Trillium albidum subsp. parviflorum and the unidentified sample sets (TRAL/PA) are grouped towards negative values of PC1 and the most negative values of PC2 and present high percentages of cis-linanool oxide (furanoid). Trillium chloropetalum shows a dispersed distribution in the space defined by the two PC axes.

Mapping the characters that were recovered as the most relevant in the PCA (scores of >0.1; Table 3) on the phylogenetic trees shows that these characters are diagnostic for some of the clades supported in the dominance character coding scheme (Fig. 3B). The characters are less relevant in defining clades supported in the binary character coding scheme (Fig. 3A). The topology of the latter tree is the result of multiple reversals of these characters. Consistency indices (calculated as the minimum number of character state changes divided by the observed number of character state changes) are low in general for these characters, except for cis-linalool oxide (furanoid), and retention indices [calculated as (maximum number of character state changes minus observed number of character state changes) divided by (maximum number of character state changes minus minimum number of character state changes)] are >0.5, except for β-caryophyllene, which also has a low consistency index in both analyses (Table 4). Citronellol, although recovered as highly relevant by the PCA, shows low consistency and retention indices (Table 4). On the contrary, β-elemene, which is not recovered by the PCA as relevant, is a synapomorphy for the T. kurabayashii clade, scoring one in both indices. We note that the presence of β-elemene could have been derived by a heat-induced isomerization of germacrene A in the injection port of the gas chromatograph (Bouwmeester et al., 2002). However, this does not change the fact that the compound, irrespective of its origin, is present in only one of the western sessile Trillium species.

Table 3.

Loadings of PC1 and PC2 of compounds that scored >0.1 in one of the PC loadings.

Compound	PC1	PC2
Isobutyl acetate	0.89	0.34
6-Methyl-5-hepten-2-one	0.06	−0.48
Benzyl alcohol	−0.09	0.11
Cis-linalool oxide (furanoid)	−0.06	−0.35
Linalool	−0.04	−0.1
Citronellol	−0.37	0.65
Geraniol	−0.21	0.2
β-Caryophyllene	−0.05	0.16

Open in a new tab

Table 4.

Consistency and retention indices for the characters that scored >0.1 in one of the PC loadings. Indices are shown for the binary and dominance analyses.

Character	Consistency index		Retention index
Character	Binary	Dominance	Binary	Dominance
Isobutyl acetate	0.33	0.25	0.92	0.74
6-Methyl-5-hepten-2-one	0.25	0.33	0.85	0.85
Benzyl alcohol	0.25	0.4	0.85	0.85
Cis-linalool oxide (furanoid)	0.5	0.67	0.8	0.75
Linalool	0.2	0.29	0.82	0.77
Citronellol	0.1	0.15	0.57	0.59
Geraniol	0.2	0.22	0.67	0.56
β-Caryophyllene	0.11	0.18	0.38	0.4

Open in a new tab

The cluster analysis supports six well-defined clusters grouped based on distances of <50 (Fig. 6). The clusters are congruent with the results of the PCA. Only one of the species is found in a single cluster (cluster 2), which includes all the sample sets of T. albidum subsp. parviflorum plus the sample sets from plants that could not be assigned confidently to either T. albidum or T. albidum subsp. parviflorum (Fig. 6). All the other species or subspecies are represented by sample sets in at least two different clusters. Cluster 6 includes sample sets of T. kurabayashii exclusively, but one of the sample sets of this species is grouped in cluster 1 with other species.

In contrast to the PCA, we find differences between the results of the cluster analysis and those of the phylogenetic analyses (Fig. 4). The clusters that are recovered as clades are those that are monospecific (except for cluster 5), i.e. cluster 2 (T. albidum subsp. parviflorum plus not assignable confidently to either T. albidum or T. albidum subsp. parviflorum), cluster 4 (T. albidum from the Marble Mountains) and cluster 6 (T. kurabayashii) (Figs 4 and 6). However, the multispecific cluster 1 appears extremely dispersed on both phylogenetic trees. Clusters 3 and 5 also appear dispersed in both phylogenies and, as is the case with cluster 1, their sample sets occupy different positions in the two trees.

Comparisons of phylogenetic and statistical analyses

Comparisons of the results of phylogenetic and statistical analyses reveal the characters (chemical compounds) that support clades and distinguish groups. The compounds that emerge as most relevant for the grouping of the western sessile Trillium sample sets in the PCA (cis-linalool oxide, linalool, citronellol, geraniol, β-caryophyllene, isobutyl acetate, benzyl alcohol and 6-methyl-5-hepten-2-one; Fig. 5) might support apomorphic characters for specific clades recovered in the phylogenetic analyses. Indeed, characters based on these compounds have, in general, high retention indices (Table 4) that are consistent with synapomorphic value and provide synapomorphies for different clades (Fig. 3). In contrast, the low consistency indices of these same characters (Table 4) are probably attributable to the large number of terminals (especially given that many of them represent multiple sample sets of the same species) and not necessarily to homoplasy.

For the tree based on the dominance character coding, these characters represent mostly synapomorphies that unite sample sets of the same species (Fig. 3). This is also reflected in the higher consistency indices and the lower retention indices that these compounds have, in comparison to their respective values in the tree based on the binary character coding (Table 4). Examples include the compounds that unite each of the T. kurabayashii, T. angustipetalum, T. albidum subsp. parviflorum and T. chloropetalum sample sets, in addition to each of the two T. chloropetalum sample sets (Fig. 3B). Each of these sample sets is supported as a clade by at least one of the compounds recovered as relevant by the PCA. In contrast, in the tree based on the binary character coding, some of the same compounds represent synapomorphies for more inclusive clades (Fig. 3A) than in the dominance character coding; the consistency indices of those compounds are lower and their retention indices higher, in general, than in the dominance character coding (Table 4). In the binary tree, smaller clades are supported by character reversals. For instance, the T. angustipetalum + T. kurabayashii node is supported by a reversal in benzyl alcohol, which is otherwise a synapomorphy for the much larger clade that includes all sample sets except for T. albidum from the Marble Mountains (Fig. 3A). Likewise, the T. angustipetalum clade is supported by a reversal in 6-methyl-5-hepten-2-one, which is a synapomorphy for the clade including T. chloropetalum, T. angustipetalum, T. kurabayashii and some of the T. albidum sample sets.

The differences observed between the clusters produced by the statistical analysis and the clades supported by the phylogenetic analyses (which are taxonomically coherent) suggest that the main scent compounds that underpin similarities that define the clusters are different from the individual compounds supporting the phylogenetic nodes. It is possible that the main compounds that define the clusters are plesiomorphic in a phylogenetic sense. However, there is at least one case in which the main character defining a cluster is homoplasic in the phylogenetic results: presenting a high percentage of isobutyl acetate, which is one of the strongest similarities uniting cluster one, is homoplasic (as a change from absence to presence) in the dominance phylogenetic tree (Fig. 3B); this is also the only character in this analysis that shows a lower consistency index than in the binary analysis. In the binary tree, however, a change from isobutyl acetate absence to presence is synapomorphic (Fig. 3A), and this character also presents the highest retention index and one of the highest consistency indices in that analysis.

It is especially noteworthy that the two populations of T. chloropetalum (WADD and GAZO) that are not included in the T. chloropetalum clade recovered by the phylogenetic analyses are included in cluster 3 with the other T. chloropetalum sample sets. The only difference in scent composition between each of these two populations and the other T. chloropetalum sample sets is the absence in each of them of one of the compounds that is diagnostic of the species: isobutyl acetate is absent in the WADD population, and (E)-nerolidol is absent in the GAZO population.

DISCUSSION

Floral scents typically act as pollinator attractants or herbivore deterrents (Raguso, 2008; Schiestl, 2010; Schiestl et al., 2014). As such, they can be shaped by selective pressures exerted by pollinators or herbivores (Theis et al., 2007; Kessler et al., 2013). At the same time, their chemistry can also be constrained by the abiotic environment of the plants, such as soil chemistry and ambient temperature (Campbell et al., 2019), which can affect the availability of specific chemical elements and the chemical reaction dynamics or limit the presence of certain pollinator species, respectively. The geographical position of a population can, therefore, play a role in shaping the floral scent chemistry of that population, directly or indirectly. In contrast, because biosynthetic pathways are controlled by genetic regulatory networks whose components can be inherited, floral scent chemistry is also constrained by the evolutionary history of a population. As a result of all these constraints and determinants, the floral scent chemistry of a population might bear both a phylogenetic and a geographical signal, while also representing an adaptation to a specific set of interactions with animals.

Identifying and separating the effects of these multiple factors that influence floral scent chemistry is complicated by their independent but overlapping and potentially interacting effects. In the case of the western sessile Trillium, pollination biology and mating systems are not understood to any meaningful degree, and interactions with specialized herbivores have not been documented. Thus, by necessity, our discussion of the causes of variation in floral scent chemistry cannot factor in any type of influence of pollination or herbivory. Detecting phylogenetic and geographical signals in floral scent chemistry with direct evidence entails comparing scent-based phylogenies with phylogenies based on independent data and knowing whether and how geography-dependent factors, such as soil chemistry and climatic variables, influence scent biosynthetic pathways. However, the influence of climate and soil chemistry on floral scent chemistry has not been studied in Trillium, and a molecular phylogeny of the populations sampled for floral scent in our study has never been produced.

These above considerations do not bode particularly well for our chances of understanding the causes of variation in floral scent chemistry in western sessile Trillium species. Phylogenetic relationships derived directly from floral scent chemistry can, nevertheless, provide indirect clues on whether geography and evolutionary history played any roles in shaping the landscape of floral scent chemistry. Like the study by Wayman et al. (2023), our study has the benefit of treating multiple populations of different species independently. In this context, (1) correlations (or lack thereof) between the geographical distance of populations or species and their phylogenetic distance will, at a minimum, provide a rough measure of the strength of influences exerted by geographically controlled factors on floral scent chemistry. By the same token, (2) correlations (or lack thereof) between the phylogenetic relationships of species and populations, as supported by floral scent chemistry, and their taxonomic assignment can lend support to hypotheses that explain variations in floral scent as the result of evolutionary history. Both these types of correlations are examined below.

Geographical or phylogenetic signal in scent chemistry?

Phylogenetic signal is the tendency of related species to resemble each other more than species drawn at random from the same tree (Münkemüller et al., 2012). The results of our phylogenetic analyses based on floral scent chemistry indicate that flower scent bears phylogenetic signal in sessile Trillium species of the ‘Western clade’ (sensuLampley, 2021). This inference is supported by taxonomic coherence in the results, i.e. the congruence between the patterns of relationships among populations and the taxonomic identity of those populations: sample sets from populations identified as belonging to the same taxon generally form monophyletic groups, irrespective of the character coding scheme used (Fig. 4; Supplementary Data File S4, Fig. S4). The congruence between the clades supported by the phylogenetic analyses, the grouping of the sample sets in the PCA (Fig. 5) and the taxonomic assignment of those sample sets is also consistent with strong phylogenetic signal, as is the good resolution of the phylogenetic trees produced with scent chemistry data.

These results imply that scent data are valuable for taxonomy independent of phylogeny, when considered at the species level. This suggests that in the western sessile Trillium, the species is the unit of taxonomic importance for scent, which is also relevant at the infraspecific level (Fig. 4; Supplementary Data File S4, Fig. S4; see also discussion below). These results cannot be generalized beyond the western sessile Trillium, and they depend on the taxon addressed. For instance, in a similar study addressing the unit of taxonomic importance for scent, Levin et al. (2003) found that scent in Oenothera varied at the population or infraspecific level along more consistent patterns than at the species level.

Our results suggest that floral scent composition in the western sessile Trillium is species specific and, thus, follows certain evolutionary pathways. These results could also be interpreted as indicting some degree of intraspecific floral scent plasticity, conspicuous especially in the wide variability in scent chemistry documented among T. albidum populations (where it could be related to substrate, climate, etc. across the extensive geographical range of this species). However, only few previous studies on floral scent have found strong effects of plasticity, but have rather supported canalization, notably in the chemical composition of Lithophragma floral scent (Friberg et al., 2019). Such results suggest that ascribing population-level variation to plasticity might not be warranted. More generally, the lack of congruence in species relationships among the tree topologies supported by the different scent character coding schemes suggests a lack of phylogenetic signal in floral scent, aside from separation of individual species. Addressing whether the lack of signal is real or whether the different topologies supported under different character coding schemes reflect conflicting signals encoded in the floral scent of western sessile Trillium can be achieved only by comparisons with the relationships supported by molecular phylogenies of the same populations. Alternatively, the supraspecific-level relationships supported by the scent phylogenies might reflect a geographical signal. However, comparisons between tree topology and geographical locations of the different populations for both character coding schemes show lack of correlation (Fig. 7).

Fig. 7. — Comparison between the phylogenetic hypotheses supported by binary and dominance coding of floral scent data with the geographical distribution of the populations sampled.

Are there better and worse ways to construct characters for phylogenetic analyses based on scent chemistry?

The answer to this question has two parts. One part of the answer is related to the level of phylogenetic resolution obtained under different character coding schemes; the other part relates to the degree of congruence between the patterns of relationships supported by the different character coding schemes. To address the first part of the answer, the differences between the levels of resolution and between the topologies supported in the different character coding schemes demonstrate that character construction is important for this type of data and analyses. Our results suggest that better phylogenetic resolution (and taxonomic coherence) is obtained by codifying scent chemistry into characters at a ‘coarser grain’. Thus, both the binary coding scheme, based strictly on presence or absence of the compounds, and the dominance coding, which adds only a third character state, ‘dominant’, for each character (compound), performed much better in terms of resolution than the ‘fine-grain’ character coding into 10 % bins. This concurs with the results observed in other phylogenies based on floral scent compounds (e.g. Levin et al., 2003). Of the two coding schemes that performed best, the binary coding scheme recovered the shortest MP trees, although in higher numbers than the dominance coding and, therefore, resulted in slightly lower resolution in the strict consensus. The relatively low consistency indices of the phylogenetic trees (Table 2) might be explained by some of the characters not being informative and, in the case of the binary tree, also by the high number of character state reversals. However, retention indices are high, suggesting that there are sufficient synapomorphies supporting the clades. Irrespective of these considerations, our results suggest that the exact abundance of a compound, although it could be autapomorphic, is not as relevant for phylogenetic resolution as its presence or absence.

With respect to the patterns of relationships recovered, the binary and the dominance character coding schemes produced conflicting topologies (Fig. 3). This raises the question of which of these two topologies is closer to reflecting the natural relationships among taxa. Given that the two topologies are equally well resolved, an answer could come from comparisons with the results of other approaches, phylogenetic (based on different types of characters) and statistical, as discussed below.

Are scent chemistry-based phylogenies congruent with those based on gene sequences or morphology?

The patterns of relationships supported by our phylogenetic analyses based on scent composition are different from those of all Trillium phylogenies published to date. The phylogeny published by Kawano and Kato (1995), based on morphological characters and including other Trillium species apart from those studied here, recovers the ‘Western clade’ as polyphyletic: T. angustipetalum and T. kurabayashii are part of a large polytomy that is sister to a clade that includes T. chloropetalum and T. albidum plus five other species; in this clade, T. albidum and T. chloropetalum are members of successively more inclusive clades that also include species that are not in the ‘Western clade’ (Fig. 8). Another phylogeny, resulting from the analysis of morphological characters by Farmer and Schilling (2002), resolves the ‘Western clade’ species as paraphyletic, with T. angustipetalum sister to the clade that includes our other species of interest, and with T. kurabayashii and T. chloropetalum forming a clade that is sister to a large polytomic node that includes T. albidum and T. parviflorum (which is treated as a separate species in this analysis) (Fig. 8).

Fig. 8. — Summaries of relationships among western sessile *Trillium* species supported by previous phylogenetic studies based on characters other than scent chemistry.

Three molecular phylogenies that include our species of interest have also been published (Osaloo and Kawano, 1999; Schilling et al., 2019; Lampley, 2021). All three phylogenies resolve these species as part of a clade, not always exclusive of other Trillium species. However, none of these phylogenies supports topologies congruent with those recovered in our analyses. In the phylogeny by Osaloo and Kawano (1999), based on the matK chloroplast gene, the western sessile species are part of a large polytomic node (Fig. 8). The two other phylogenies, which were based on the nuclear ribosomal ITS region and various chloroplast and nuclear genes, have better resolution, with T. angustipetalum sister to a clade wherein T. chloropetalum is sister to the three remaining species of western sessile Trillium (Schilling et al., 2019; Lampley, 2021). In the phylogeny published by Lampley (2021), the clade formed by T. kurabayashii, T. albidum and T. albidum subsp. parviflorum is polytomic (Fig. 8), while in that of Schilling et al. (2019)T. kurabayashii is sister to the clade formed by T. albidum and T. albidum subsp. parviflorum (Fig. 8).

The differences between our results based on floral scent composition and those of phylogenetic analyses that use other types of data suggest that if flower scent reflects phylogenetic signals, it evolved along different pathways from genes and morphology. Yet some of our results suggest that these differences might not be as significant as indicated by superficial comparisons. That is because tree rooting and terminal exclusion experiments with our floral scent dataset obtain topologies that are more similar to those supported in the molecular phylogenies (Schilling et al., 2019; Lampley, 2021). For instance, we obtained such topologies that place T. angustipetalum as sister to the other western sessile species (albeit with low resolution among the latter; Supplementary Data File S4, Figs S5–S7) when using different populations of Pseudotrillium rivale as the outgroup. As another example, the high floral scent variability among T. albidum populations might result in patterns of scent composition that are incongruent with those of non-scent traits. In fact, when T. albidum is excluded from our scent-based tree searches, the phylogenetic relationships recovered among the remaining western sessile species are more similar to those supported in molecular phylogenies: (T. albidum subsp. parviflorum (T. angustipetalum (T. chloropetalum + T. kurabayashii))) (Supplementary Data File S4, Fig. S8).

How do phylogenetic and statistical approaches compare in their ability to address taxonomy?

Wayman et al. (2023) point out that considerable intraspecific variation within western sessile Trillium species and overlapping traits make them troublesome to distinguish morphologically. These considerations led them to explore whether floral scents and the chemical markers they bear can be used to distinguish the sessile Trillium of western North America, using statistical methods of comparison between populations, based on their scent composition. Their analyses of floral scent chemistry led Wayman et al. (2023) to conclude that differences in floral scent compositions are consistent with the taxonomy of the western sessile Trillium and that floral scent composition could provide taxonomic insight when species delimitations based on morphology are equivocal. Our phylogenetic results provide an independent perspective that led us to revisit some of the main questions in western sessile Trillium taxonomy, addressing whether phylogenies based on scent chemistry inform taxonomic decisions. These questions relate to the taxonomic status of T. albidum populations, the potential hybrid status of T. chloropetalum, and overall distinctions between the different western sessile species.

For T. albidum, it is still debated whether populations in Washington State represent a subspecies (referred to as T. albidum subsp. parviflorum; Bledsoe, 1993; Chambers and Meyers, 2011) or whether those populations are a separate species (T. parviflorum; Soukup, 1980; Case, 2002). Related to this, it is unclear whether populations of T. albidum from northern Oregon, which exhibit intermediate morphologies between the Washington State populations and those in southern Oregon and California, are more closely related to either of those two groups. Our phylogenies support a closer relationship with the former, because the Washington State and northern Oregon populations form a least inclusive clade, consistent with our statistical analyses and with those of Wayman et al. (2023). Although the statistical analyses (both ours and those of Wayman et al., 2023) clearly separate the Washington State populations from all populations unequivocally assigned to T. albidum, in our phylogenies the clade formed by the Washington State and northern Oregon populations is nested in successively more inclusive clades consisting only of T. albidum populations. Thus, although statistical analyses could be used to support a separate species status for the Washington State populations, the phylogenetic results are more consistent with a subspecies status, especially if all T. albidum species do, indeed, represent a single species. However, the conspecific status of T. albidum populations is not supported by the phylogenetic results, where these populations form either a grade (binary coding) or are polyphyletic (dominance coding). Of course, these questions can be answered with more certainty by population genetics studies.

Another question related to T. albidum concerns populations in the Marble Mountain Wilderness (MMW) and their possible status as a species distinct from T. albidum in California. These populations form a group that is not clearly separated from the T. albidum from California (or Oregon) in the analyses by Wayman et al. (2023) and in our PCA, both of which are inconsistent with a separate species status. However, the MMW populations are resolved as clades in the phylogenetic analyses, which could be used to support them as a separate species, like our ordination results, where they are grouped in their own cluster that is more-or-less equidistant from other T. albidum and T. chloropetalum populations. In the dominance tree (Fig. 4B), the MMW populations are most sister to a clade consisting of high-elevation T. albidum populations (YOLA, POIS and LONG). If this relationship reflects similarities owing to high-elevation adaptations (e.g. similar pollinators), then the phylogenetic position of the MMW populations would be less supportive of their separate species status.

In the broader picture of western sessile Trillium, a major question concerns the relationships between T. kurabayashii, T. angustipetalum and T. chloropetalum and their status as distinct species (Freeman, 1975; Case et al., 2002; McNeal and Ness, 2012). Consistent with a status of distinct species, populations assigned to T. kurabayashii, T. angustipetalum and T. chloropetalum each form separate clades that are not closely related in the dominance tree. However, the binary tree is less conclusive about their status as distinct species: T. chloropetalum populations form a grade that is paraphyletic to a clade where the clades formed by T. angustipetalum and T. kurabayashii are sister to each other. Of the latter two, T. angustipetalum is intermixed, at least in part, with T. chloropetalum in the statistical analyses (both ours and that of Wayman et al., 2023) and only T. kurabayashii is unequivocally distinct. Among our results, the clustering analysis is the only one that does not support T. kurabayashii as a separate species (Table 5). However, we note that it is typical of T. kurabayashii that both isobutyl acetate and 6-methyl-5-hepten-2-one are the main compounds. Of the two, the latter is more often the dominant one. However, in one of our T. kurabayashii populations (REDC), isobutyl acetate is the dominant one of the two, which probably supported its clustering with T. angustipetalum and some T. chloropetalum.

Table 5.

Comparison between how the phylogenetic and statistical analyses address the taxonomic questions raised for the western sessile Trillium taxa. (but see discussion on T. kurabayashii). NMDS (Analysis of non-parametric multidimensional scaling).

	Binary	Dominance	Ordination (PCA and NMDS)	UPGMA clustering
Placement of North Oregon populations	parviflorum	parviflorum	parviflorum	parviflorum
T. parviflorum separate species?	No (but populations form a clade)	No (but populations form a clade)	Yes (but more similar to T. albidum than to other species)	Yes
MMW populations separate species?	Yes	Yes	No	Yes
T. kurabayashii separate species?	Yes	Yes	Yes	No (one population outside cluster)
T. angustipetalum separate species?	Yes	Yes	Yes (?)	No
T. chloropetalum separate species?	No (grade)	No (two clades)	No	No
T. chloropetalum hybrid?	No	No (?)	Yes	Equivocal
Distinct varieties of T. chloropetalum	No	No	No	No
Placement of NAPA populations	angustipetalum	angustipetalum	angustipetalum/equivocal	Equivocal

Open in a new tab

A hybrid origin for T. chloropetalum was first considered by Goodspeed and Brandt (1916), and Freeman (1975) discussed the possibility of T. chloropetalum representing a hybrid of T. albidum and T. angustipetalum. This involves several potential situations: (1) T. chloropetalum is a series of hybrid populations, all of the same original parentage; (2) T. chloropetalum is a series of hybrid populations, but these have more than one hybrid origin (e.g. different sets of parents); or (3) T. chloropetalum includes a distinct species (some populations), but other populations are hybrids between this T. chloropetalum s.s. and one or more other species. In their statistical treatment, Wayman et al. (2023) observed T. chloropetalum sample sets intermixed with T. albidum and T. angustipetalum. The intermixing of T. chloropetalum with those two species is somewhat similar to our PCA results, in which, however, T. albidum and T. angustipetalum are clearly separated. These results could be interpreted as supporting T. chloropetalum as a hybrid of T. angustipetalum and T. albidum.

From a phylogenetic standpoint, hybridization can lead to two types of relationships in analyses that involve closely related species. One type is when the hybrid and the two putative parent species form a polytomy. In the other type, the hybrid is sister to one of the parent species and shares homoplasic characters with the other (Funk, 1985; McDade, 1990). Neither of these topologies indicative of hybridization is recovered between T. angustipetalum, T. albidum and T. chloropetalum in our binary tree, although the populations of the last of these are part of a grade, wherein they are positioned between populations of T. angustipetalum and those of T. albidum. In the dominance tree, populations of T. chloropetalum form two separate lineages nested within a clade that includes all T. albidum populations and is sister to the clade formed by the T. angustipetalum populations. Although one of the two T. chloropetalum lineages shares isobutyl acetate presence with T. angustipetalum, as in a hybrid origin topology, the other T. chloropetalum does not. Thus, although the phylogenetic relationships supported by floral scent composition do not provide strong evidence for a hybrid origin of T. chloropetalum, this possibility cannot be ruled out, because back-crossing and segregating individuals beyond the F1 of a hybrid swarm could also hide tree topologies diagnostic of hybridization.

To explore possible effects of the putative hybrid character of the T. chloropetalum populations on tree topology, we excluded them from the analyses. The resulting trees based on binary and dominance coding schemes support the same general topology among the three remaining species: (T. kurabayashii (T. angustipetalum (T. albidum))) (Supplementary Data File S4, Figs S9 and S10). This topology is also the one supported under the dominance coding when all taxa are included, which could be taken to indicate that the dominance coding scheme is better reflective of the true interspecific phylogenetic relationships.

The two varieties identified within T. chloropetalum, var. giganteum and var. chloropetalum, overlap considerably in the statistical analyses by Wayman et al. (2023) and in our clustering and ordination analyses (with populations of var. chloropetalum forming a tighter cluster in the latter). Consistent with these results and arguing against differentiation into varieties, populations attributed to the two varieties are intermixed in both phylogenetic trees: in the binary tree they form a paraphyletic group, whereas in the dominance tree they form two clades of intermixed varieties that are only distantly related (the GAZO and WADD populations form one of the two clades, possibly owing to each lacking a different major scent component that is found in all the populations of the other T. chloropetalum clade).

Finally, two populations of red sessile Trillium from Napa County thought to belong to T. chloropetalum by Freeman (1975) share some morphological characters with each of T. angustipetalum or T. chloropetalum var. giganteum, thus, they cannot be identified unequivocally to either of the two taxa. In the analyses by Wayman et al. (2023), the Napa populations are not differentiated from either T. angustipetalum or T. chloropetalum var. giganteum. However, they group well with T. angustipetalum, but not with T. chloropetalum var. giganteum in both our phylogenetic results and the PCA, which thus support their conspecificity with T. angustipetalum.

From a broad comparative standpoint, both character coding schemes based on floral scent data (binary and dominance) support the same taxonomic decisions at the species level (considered in a general sense, e.g. separation between species and subspecies, inclusion of the latter in the former), despite the different tree topologies (Table 5). Overall, the relationships supported by the phylogenetic trees are in agreement with some of the taxonomic decisions rooted in traditional ‘evolutionary systematics’ (sensuMishler, 2009) and provide answers on the taxonomic identity of putative subspecies or varieties. However, some of the decisions supported by the phylogenetic results conflict with those supported by statistical analyses of the same data (Table 5).

How do phylogenetic and statistical approaches compare in addressing evolutionary patterns?

Statistical analyses (ordination and clustering) assess overall similarity between the scent chemistry of individual populations and provide different measures of distance between populations in multidimensional space. These distances can be useful (and have been used, e.g. phenetics) in addressing taxonomic questions. However, phenotypic similarity can be attributable to shared ancestry or can represent homoplasy. As a result, measures of similarity are not directly relevant to addressing evolutionary relationships, which are reconstructed using phylogenetic methods. In contrast, cladistic methods remain problematic for identifying and visualizing reticulate speciation and become challenging when there is a history of introgression (e.g. gene trees are discordant with species trees). The utility and effectiveness of the two types of approaches are also a matter of the taxonomic level at which questions are explored (population, species, genus, etc.). Because our western sessile Trillium samples span the infra- to supraspecific levels, in exploring their relationships we compared multiple methods.

The relationships supported by the phylogenetic analyses of scent chemistry track taxonomic assignments more closely than those supported by the statistical analyses. Tree topologies in the two character coding schemes (binary and dominance) support roughly the same species-level groupings and infraspecific relationships, but are conflicting in among-species relationships. This is reflected in the overall consensus tree based on the results of the binary and dominance character coding analyses (Fig. 9), which shows that certain clades are supported irrespective of the character coding scheme. These clades are T. kurabayashii, T. angustipetalum, T. albidum subsp. parviflorum (including the northern Oregon populations), T. chloropetalum (minus the GAZO and WADD populations), the Marble Mountains Wilderness T. albidum and the high-elevation T. albidum (POIS, YOLA and LONG populations). These results: (1) suggest that floral scent composition in the western sessile Trillium is species specific and might follow certain evolutionary pathways; and (2) imply that scent data are good for taxonomy independent of phylogeny, when they operate at the species level.

Floral scent chemistry and the evolution of floral scent

The overall consensus tree of the two analyses based on the two character coding schemes (Fig. 9) also makes it obvious that it is still unclear which character coding scheme for scent data best reflects evolutionary relationships between species and subspecific-level populations. The low level of congruence between the results obtained under the two character coding schemes might reflect, in part, rooting issues and, in part, taxon sampling issues. This is suggested by the results obtained when using different outgroups or upon exclusion of the highly variable T. albidum populations, which mirror, to some extent, the tree topologies supported by gene sequence data (see above). However, because of the questions they addressed, those molecular phylogenies did not sample the same populations as our analyses, hence it is also unclear how relevant the comparisons between their results and ours are to (1) the evolutionary relationships between the western sessile Trillium and (2) the choice of the most appropriate scent character coding scheme.

More definitive answers to all these questions (e.g. bolstering arguments for genetic isolation vs. introgression between populations vs. between specific taxa) will be obtained only by broader sampling of T. albidum populations for genetic markers and by sampling of the same Trillium populations for both floral scent and gene sequence data. Comparisons with the results of model-based phylogenetic analyses might also be useful in this respect, although those analyses will also have to address at the outset the same questions of character coding. Some clarity might be contributed to both parsimony- and model-based analyses when biosynthetic pathways for different scent compounds are better understood, especially pathways that might connect some of the compounds identified in the western sessile Trillium floral scents. Morphology has so far proved unreliable as an indicator of phylogenetic relationships in Trillium (Kawano and Kato, 1995; Farmer and Schilling, 2002; see also the Materials and Methods section). However, if documented in detail and for the same populations analysed for floral scent, morphology could play an important role in interpreting specific results, especially in concert with molecular data, such as lending support to a certain scent character coding scheme over the other. Eventually, a phylogeny supported by all these data will allow for understanding the evolution of floral scent chemistry in the western sessile Trillium, with potential implications for the evolution of biosynthetic pathways and their regulation.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following.

File S1: specimen vouchers of Trillium and Pseudotrillium used for floral scent analysis. File S2: table of compounds with identification method, retention time, Kovats index, mass spectral data and presence vs. absence for each taxon. File S3: raw data of average floral volatile composition of each sample set, and matrices for phylogenetic analyses, including morphological data. File S4: phylogenetic trees.

mcae120_suppl_Supplementary_File_S1

mcae120_suppl_supplementary_file_s1.xlsx^{(14.4KB, xlsx)}

mcae120_suppl_Supplementary_File_S2

mcae120_suppl_supplementary_file_s2.xlsx^{(19.1KB, xlsx)}

mcae120_suppl_Supplementary_File_S3

mcae120_suppl_supplementary_file_s3.xlsx^{(108.7KB, xlsx)}

mcae120_suppl_Supplementary_File_S4

mcae120_suppl_supplementary_file_s4.docx^{(432.3KB, docx)}

ACKNOWLEDGEMENTS

We thank Edward Schilling (University of Tennessee, Knoxville) and Robert Raguso (Cornell University) for constructive feedback on an initial draft of the manuscript. Detailed constructive comments from two anonymous reviewers helped us to improve the manuscript considerably. Collecting permits were generously provided by Redwood National Park, California State Parks, Klamath National Forest, Shasta-Trinity National Forest, Six Rivers National Forest, Swanton Pacific Ranch and Mendocino County Resource Conservation District.

Contributor Information

Candela Blanco-Moreno, Departamento de Biología, Universidad Autónoma de Madrid, Cantoblanco, Madrid, 28049, Spain; Department of Biological Sciences, California State Polytechnic University, Humboldt, Arcata, CA 95521, USA.

Kjirsten A Wayman, Department of Chemistry, California State Polytechnic University, Humboldt, Arcata, CA 95521, USA.

Alexandru M F Tomescu, Department of Biological Sciences, California State Polytechnic University, Humboldt, Arcata, CA 95521, USA.

FUNDING

Candela Blanco-Moreno is supported by a Margarita Salas Postdoc CA1/RSUE/2021-00703 scholarship funded by the Spanish Ministry of Universities and Universidad Autónoma de Madrid.

LITERATURE CITED

Adams RP. 1995. Identification of essential oil components by gas chromatography/mass spectrometry. Carol Stream: Allured Publishing Corporation. [Google Scholar]
Barkman TJ. 2001. Character coding of secondary chemical variation for use in phylogenetic analyses. Biochemical Systematics and Ecology 29: 1–20. [DOI] [PubMed] [Google Scholar]
Bledsoe KE. 1993. Morphological and cytological variation in Trillium albidum Freeman (Liliaceae). M.S. Thesis, Oregon State University, Corvallis. [Google Scholar]
Bouwmeester HJ, Kodde J, Verstappen F, Altug IG, de Kraker JW, Wallaart TE.. 2002. Isolation and characterization of two germacrene A synthase cDNA clones from chicory. Plant Physiology 129: 134–144. [DOI] [PMC free article] [PubMed] [Google Scholar]
Campbell DR, Sosenski P, Raguso RA.. 2019. Phenotypic plasticity of floral volatiles in response to increasing drought stress. Annals of Botany 123: 601–610. [DOI] [PMC free article] [PubMed] [Google Scholar]
Case FW. 2002. Trillium. In: Flora of North America Editorial Committee, eds. Flora of North America, Vol. 26. Oxford: Oxford University Press, 90–117. [Google Scholar]
Case FW, Case RB.. 1997. Trilliums. Portland: Timber Press. [Google Scholar]
Chambers KL, Meyers SC.. 2011. Nomenclatural changes for some taxa in the Oregon flora. Journal of the Botanical Research Institute of Texas 5: 619–623. [Google Scholar]
Chapurlat E, Anderson J, Ågren J, Friberg M, Sletvold N.. 2018. Diel pattern of floral scent emission matches the relative importance of diurnal and nocturnal pollinators in populations of Gymnadenia conopsea. Annals of Botany 121: 711–721. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chess SK, Raguso RA, LeBuhn G.. 2008. Geographic divergence in floral morphology and scent in Linanthus dichotomus (Polemoniaceae). American Journal of Botany 95: 1652–1659. [DOI] [PubMed] [Google Scholar]
Delle-Vedove R, Schatz B, Dufay M.. 2017. Understanding intraspecific variation of floral scent in light of evolutionary ecology. Annals of Botany 120: 1–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
Dötterl S, Gershenzon J.. 2023. Chemistry, biosynthesis and biology of floral volatiles: roles in pollination and other functions. Natural Product Reports 40: 1901–1937. [DOI] [PubMed] [Google Scholar]
Doubleday LAD, Raguso RA, Eckert CG.. 2013. Dramatic vestigialization of floral fragrance across a transition from outcrossing to selfing in Abronia umbellata (Nyctaginaceae). American Journal of Botany 100: 2280–2292. [DOI] [PubMed] [Google Scholar]
Eisen KE, Ma R, Raguso RA.. 2022. Among‐ and within‐population variation in morphology, rewards, and scent in a hawkmoth‐pollinated plant. American Journal of Botany 109: 1794–1810. [DOI] [PubMed] [Google Scholar]
El-Sayed AM. 2021. The Pherobase: database of insect pheromones and semiochemicals. http://www.pherobase.com (20 November 2023, date last accessed). [Google Scholar]
Farmer SB, Schilling EE.. 2002. Phylogenetic analyses of Trilliaceae based on morphological and molecular data. Systematic Botany 27: 674–692. [Google Scholar]
Freeman JD. 1975. Revision of Trillium subgenus Phyllantherum (Liliaceae). Brittonia 27: 1–62. [Google Scholar]
Friberg M, Schwind C, Raguso RA, Thompson JN.. 2013. Extreme divergence in floral scent among woodland star species (Lithophragma spp.) pollinated by floral parasites. Annals of Botany 111: 539–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
Friberg M, Schwind C, Guimarães PR, Raguso RA, Thompson JN.. 2019. Extreme diversification of floral volatiles within and among species of Lithophragma (Saxifragaceae). Proceedings of the National Academy of Sciences of the United States of America 116: 4406–4415. [DOI] [PMC free article] [PubMed] [Google Scholar]
Funk VA. 1985. Phylogenetic patterns and hybridization. Annals of the Missouri Botanical Garden 72: 681–715. [Google Scholar]
Goloboff PA, Farris JS, Nixon KC.. 2008. TNT, a free program for phylogenetic analysis. Cladistics 24: 774–786. [Google Scholar]
Goodspeed TH, Brandt RP.. 1916. Notes on the Californian species of Trillium. Berkeley: University of California. [Google Scholar]
Hammer O, Harper DA, Ryan PD.. 2001. PAST: paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4: 4. [Google Scholar]
Huber FK, Kaiser R, Sauter W, Schiestl FP.. 2005. Floral scent emission and pollinator attraction in two species of Gymnadenia (Orchidaceae). Oecologia 142: 564–575. [DOI] [PubMed] [Google Scholar]
Jaeger DM, Runyon JB, Richardson BA.. 2016. Signals of speciation: volatile organic compounds resolve closely related sagebrush taxa, suggesting their importance in evolution. The New Phytologist 211: 1393–1401. [DOI] [PubMed] [Google Scholar]
Kawano S, Kato H.. 1995. Evolutionary biology of Trillium and related genera (Trilliaceae) II. Cladistic analyses on gross morphological characters, and phylogeny and evolution of the genus Trillium. Plant Species Biology 10: 169–183. [Google Scholar]
Kessler D, Diezel C, Clark DG, Colquhoun TA, Baldwin IT.. 2013. Petunia flowers solve the defence/apparency dilemma of pollinator attraction by deploying complex floral blends. Ecology Letters 16: 299–306. [DOI] [PubMed] [Google Scholar]
Knudsen JT, Eriksson R, Gershenzon J, Ståhl B.. 2006. Diversity and distribution of floral scent. Botanical Review 72: 1–120. [Google Scholar]
Lampley JA. 2021. A systematic and biogeographic study of Trillium (Melanthiaceae). PhD Thesis, University of Tennessee, Knoxville. doi: https://doi.org/ 10.1111/j.1442-1984.1995.tb00137.x [DOI] [Google Scholar]
Lampley JA, Gereau R, Floden A, Schilling EE.. 2022. A revised subgeneric classification of Trillium (Parideae, Melanthiaceae). Phytotaxa 552: 278–286. [Google Scholar]
Les DH, Whitkus R, Bryan FA, Tyrell LE.. 1989. Biochemical basis of floral color polymorphism in a heterocyanic population of Trillium sessile (Liliaceae). American Journal of Botany 76: 23–29. [Google Scholar]
Levin RA, McDade LA, Raguso RA.. 2003. The systematic utility of floral and vegetative fragrance in two genera of Nyctaginaceae. Systematic Biology 52: 334–351. [DOI] [PubMed] [Google Scholar]
Liaw A, Wiener M.. 2002. Classification and regression by randomForest. R news 2: 18–22. [Google Scholar]
McDade L. 1990. Hybrids and phylogenetic systematics I. Patterns of character expression in hybrids and their implications for cladistic analysis. Evolution 44: 1685–1700. [DOI] [PubMed] [Google Scholar]
McNeal DW, Ness BD.. 2012. Trillium. In: eds. Jepson Flora Project, Jepson eFlora, https://ucjeps.berkeley.edu/eflora/eflora_display.php?tid=9561 (20 November 2023, date last accessed). [Google Scholar]
Meredith C, Frances A, Highland A, et al. 2022. The conservation status of Trillium in North America. Hockessin: New Mexico BioPark Society; Albuquerque: Mt. Cuba Center. [Google Scholar]
Meyers SC, Jaster T, Mitchell KE, Hardison LK.. 2015. Flora of Oregon, Vol. 1. Fort Worth: Brit Press. [Google Scholar]
Mishler BD. 2009. Three centuries of paradigm changes in biological classification: is the end in sight? Taxon 58: 61–67. [Google Scholar]
Münkemüller T, Lavergne S, Bzeznik B, et al. 2012. How to measure and test phylogenetic signal. Methods in Ecology and Evolution 3: 743–756. [Google Scholar]
NIST Mass Spectrometry Data Center, and W. Wallace 2022. Mass spectra. In Linstrom PJ, Mallard WG. eds. NIST Chemistry WebBook, NIST Standard Reference Database Number 69 [online database]. Gaithersburg: National Institute of Standards and Technology, https://doi.org/ 10.18434/T4D303 (2021–2022, date last accessed). [DOI] [Google Scholar]
Osaloo SK, Kawano S.. 1999. Molecular systematics of Trilliaceae II. Phylogenetic analyses of Trillium and its allies using sequences of rbcL and matK genes of cpDNA and internal transcribed spacers of 18S–26S nrDNA. Plant Species Biology 14: 75–94. [Google Scholar]
Petrén H, Toräng P, Ågren J, Friberg M.. 2021. Evolution of floral scent in relation to self-incompatibility and capacity for autonomous self-pollination in the perennial herb Arabis alpina. Annals of Botany 127: 737–747. [DOI] [PMC free article] [PubMed] [Google Scholar]
Prieto-Benítez S, Millanes AM, Dötterl S, Giménez-Benavides L.. 2016. Comparative analyses of flower scent in Sileneae reveal a contrasting phylogenetic signal between night and day emissions. Ecology and Evolution 6: 7869–7881. [Google Scholar]
Raguso RA. 2008. Wake up and smell the roses: the ecology and evolution of floral scent. Annual Review of Ecology, Evolution, and Systematics 39: 549–569. [Google Scholar]
Schiestl FP. 2010. The evolution of floral scent and insect chemical communication. Ecology Letters 13: 643–656. [DOI] [PubMed] [Google Scholar]
Schiestl FP, Huber FK, Gomez JM.. 2011. Phenotypic selection on floral scent: trade-off between attraction and deterrence? Evolutionary Ecology 25: 237–248. [Google Scholar]
Schiestl FP, Kirk H, Bigler L, Cozzolino S, Desurmont GA.. 2014. Herbivory and floral signaling: phenotypic plasticity and tradeoffs between reproduction and indirect defense. The New Phytologist 203: 257–266. [DOI] [PubMed] [Google Scholar]
Schilling EE, Floden A, Lampley J, Patrick TS, Farmer SB.. 2019. A new species in Trillium (Melanthiaceae) from Central Georgia and its phylogenetic position in subgenus Sessilium. Systematic Botany 44: 107–114. [Google Scholar]
Skogen KA, Jogesh T, Hilpman ET, Todd SL, Raguso RA.. 2022. Extensive population-level sampling reveals clinal variation in (R)-(−)-linalool produced by the flowers of an endemic evening primrose, Oenothera harringtonii. Phytochemistry 200: 113185. [DOI] [PubMed] [Google Scholar]
Soukup VG. 1980. A new Trillium (Liliaceae) from the northwestern United States. Brittonia 32: 330–333. [Google Scholar]
Theis N, Lerdau M, Raguso RA.. 2007. The challenge of attracting pollinators while evading floral herbivores: patterns of fragrance emission in Cirsium arvense and Cirsium repandum (Asteraceae). International Journal of Plant Sciences 168: 587–601. [Google Scholar]
Thielle K. 1993. The holy grail of the perfect character: the cladistic treatment of morphometric data. Cladistics 9: 275–304. [DOI] [PubMed] [Google Scholar]
Wayman KA, de Lange PJ, Larsen L, Sansom CE, Perry NB.. 2010. Chemotaxonomy of Pseudowintera: sesquiterpene dialdehyde variants are species markers. Phytochemistry 71: 766–772. [DOI] [PubMed] [Google Scholar]
Wayman KA, Reilly MJ, Petlewski AR.. 2023. Taxonomic insights from floral scents of western North American sessile‐flowered Trillium. American Journal of Botany 110: e16255. [DOI] [PubMed] [Google Scholar]
Weiss J, Mühlemann JK, Ruiz-Hernández V, Dudareva N, Egea-Cortines M.. 2016. Phenotypic space and variation of floral scent profiles during late flower development in Antirrhinum. Frontiers in Plant Science 7: 1903. [DOI] [PMC free article] [PubMed] [Google Scholar]
Whitkus R, Bryan FA, Les DH, Tyrrell LE.. 1987. Genetic structure in a heterocyanic population of Trillium sessile (Liliaceae). Plant Species Biology 2: 67–73. [Google Scholar]
Williams NH, Whitten WM.. 1999. Molecular phylogeny and floral fragrances of male euglossine bee-pollinated orchids: a study of Stanhopea (Orchidaceae). Plant Species Biology 14: 129–136. [Google Scholar]
Zomlefer WB, Williams NH, Whitten WM, Judd WS.. 2001. Generic circumscription and relationships in the Tribe Melanthieae (Liliales, Melanthiaceae), with emphasis on Zigadenus: evidence from ITS and trnL-F sequence data. American Journal of Botany 88: 1657–1669. [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

mcae120_suppl_Supplementary_File_S1

mcae120_suppl_supplementary_file_s1.xlsx^{(14.4KB, xlsx)}

mcae120_suppl_Supplementary_File_S2

mcae120_suppl_supplementary_file_s2.xlsx^{(19.1KB, xlsx)}

mcae120_suppl_Supplementary_File_S3

mcae120_suppl_supplementary_file_s3.xlsx^{(108.7KB, xlsx)}

mcae120_suppl_Supplementary_File_S4

mcae120_suppl_supplementary_file_s4.docx^{(432.3KB, docx)}

[CIT0001] Adams RP. 1995. Identification of essential oil components by gas chromatography/mass spectrometry. Carol Stream: Allured Publishing Corporation. [Google Scholar]

[CIT0002] Barkman TJ. 2001. Character coding of secondary chemical variation for use in phylogenetic analyses. Biochemical Systematics and Ecology 29: 1–20. [DOI] [PubMed] [Google Scholar]

[CIT0003] Bledsoe KE. 1993. Morphological and cytological variation in Trillium albidum Freeman (Liliaceae). M.S. Thesis, Oregon State University, Corvallis. [Google Scholar]

[CIT0004] Bouwmeester HJ, Kodde J, Verstappen F, Altug IG, de Kraker JW, Wallaart TE.. 2002. Isolation and characterization of two germacrene A synthase cDNA clones from chicory. Plant Physiology 129: 134–144. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0005] Campbell DR, Sosenski P, Raguso RA.. 2019. Phenotypic plasticity of floral volatiles in response to increasing drought stress. Annals of Botany 123: 601–610. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0006] Case FW. 2002. Trillium. In: Flora of North America Editorial Committee, eds. Flora of North America, Vol. 26. Oxford: Oxford University Press, 90–117. [Google Scholar]

[CIT0007] Case FW, Case RB.. 1997. Trilliums. Portland: Timber Press. [Google Scholar]

[CIT0008] Chambers KL, Meyers SC.. 2011. Nomenclatural changes for some taxa in the Oregon flora. Journal of the Botanical Research Institute of Texas 5: 619–623. [Google Scholar]

[CIT0009] Chapurlat E, Anderson J, Ågren J, Friberg M, Sletvold N.. 2018. Diel pattern of floral scent emission matches the relative importance of diurnal and nocturnal pollinators in populations of Gymnadenia conopsea. Annals of Botany 121: 711–721. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0010] Chess SK, Raguso RA, LeBuhn G.. 2008. Geographic divergence in floral morphology and scent in Linanthus dichotomus (Polemoniaceae). American Journal of Botany 95: 1652–1659. [DOI] [PubMed] [Google Scholar]

[CIT0011] Delle-Vedove R, Schatz B, Dufay M.. 2017. Understanding intraspecific variation of floral scent in light of evolutionary ecology. Annals of Botany 120: 1–20. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0012] Dötterl S, Gershenzon J.. 2023. Chemistry, biosynthesis and biology of floral volatiles: roles in pollination and other functions. Natural Product Reports 40: 1901–1937. [DOI] [PubMed] [Google Scholar]

[CIT0013] Doubleday LAD, Raguso RA, Eckert CG.. 2013. Dramatic vestigialization of floral fragrance across a transition from outcrossing to selfing in Abronia umbellata (Nyctaginaceae). American Journal of Botany 100: 2280–2292. [DOI] [PubMed] [Google Scholar]

[CIT0014] Eisen KE, Ma R, Raguso RA.. 2022. Among‐ and within‐population variation in morphology, rewards, and scent in a hawkmoth‐pollinated plant. American Journal of Botany 109: 1794–1810. [DOI] [PubMed] [Google Scholar]

[CIT0015] El-Sayed AM. 2021. The Pherobase: database of insect pheromones and semiochemicals. http://www.pherobase.com (20 November 2023, date last accessed). [Google Scholar]

[CIT0016] Farmer SB, Schilling EE.. 2002. Phylogenetic analyses of Trilliaceae based on morphological and molecular data. Systematic Botany 27: 674–692. [Google Scholar]

[CIT0017] Freeman JD. 1975. Revision of Trillium subgenus Phyllantherum (Liliaceae). Brittonia 27: 1–62. [Google Scholar]

[CIT0018] Friberg M, Schwind C, Raguso RA, Thompson JN.. 2013. Extreme divergence in floral scent among woodland star species (Lithophragma spp.) pollinated by floral parasites. Annals of Botany 111: 539–550. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0019] Friberg M, Schwind C, Guimarães PR, Raguso RA, Thompson JN.. 2019. Extreme diversification of floral volatiles within and among species of Lithophragma (Saxifragaceae). Proceedings of the National Academy of Sciences of the United States of America 116: 4406–4415. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0020] Funk VA. 1985. Phylogenetic patterns and hybridization. Annals of the Missouri Botanical Garden 72: 681–715. [Google Scholar]

[CIT0021] Goloboff PA, Farris JS, Nixon KC.. 2008. TNT, a free program for phylogenetic analysis. Cladistics 24: 774–786. [Google Scholar]

[CIT0022] Goodspeed TH, Brandt RP.. 1916. Notes on the Californian species of Trillium. Berkeley: University of California. [Google Scholar]

[CIT0023] Hammer O, Harper DA, Ryan PD.. 2001. PAST: paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4: 4. [Google Scholar]

[CIT0024] Huber FK, Kaiser R, Sauter W, Schiestl FP.. 2005. Floral scent emission and pollinator attraction in two species of Gymnadenia (Orchidaceae). Oecologia 142: 564–575. [DOI] [PubMed] [Google Scholar]

[CIT0025] Jaeger DM, Runyon JB, Richardson BA.. 2016. Signals of speciation: volatile organic compounds resolve closely related sagebrush taxa, suggesting their importance in evolution. The New Phytologist 211: 1393–1401. [DOI] [PubMed] [Google Scholar]

[CIT0026] Kawano S, Kato H.. 1995. Evolutionary biology of Trillium and related genera (Trilliaceae) II. Cladistic analyses on gross morphological characters, and phylogeny and evolution of the genus Trillium. Plant Species Biology 10: 169–183. [Google Scholar]

[CIT0027] Kessler D, Diezel C, Clark DG, Colquhoun TA, Baldwin IT.. 2013. Petunia flowers solve the defence/apparency dilemma of pollinator attraction by deploying complex floral blends. Ecology Letters 16: 299–306. [DOI] [PubMed] [Google Scholar]

[CIT0028] Knudsen JT, Eriksson R, Gershenzon J, Ståhl B.. 2006. Diversity and distribution of floral scent. Botanical Review 72: 1–120. [Google Scholar]

[CIT0029] Lampley JA. 2021. A systematic and biogeographic study of Trillium (Melanthiaceae). PhD Thesis, University of Tennessee, Knoxville. doi: https://doi.org/ 10.1111/j.1442-1984.1995.tb00137.x [DOI] [Google Scholar]

[CIT0030] Lampley JA, Gereau R, Floden A, Schilling EE.. 2022. A revised subgeneric classification of Trillium (Parideae, Melanthiaceae). Phytotaxa 552: 278–286. [Google Scholar]

[CIT0031] Les DH, Whitkus R, Bryan FA, Tyrell LE.. 1989. Biochemical basis of floral color polymorphism in a heterocyanic population of Trillium sessile (Liliaceae). American Journal of Botany 76: 23–29. [Google Scholar]

[CIT0032] Levin RA, McDade LA, Raguso RA.. 2003. The systematic utility of floral and vegetative fragrance in two genera of Nyctaginaceae. Systematic Biology 52: 334–351. [DOI] [PubMed] [Google Scholar]

[CIT0058] Liaw A, Wiener M.. 2002. Classification and regression by randomForest. R news 2: 18–22. [Google Scholar]

[CIT0033] McDade L. 1990. Hybrids and phylogenetic systematics I. Patterns of character expression in hybrids and their implications for cladistic analysis. Evolution 44: 1685–1700. [DOI] [PubMed] [Google Scholar]

[CIT0034] McNeal DW, Ness BD.. 2012. Trillium. In: eds. Jepson Flora Project, Jepson eFlora, https://ucjeps.berkeley.edu/eflora/eflora_display.php?tid=9561 (20 November 2023, date last accessed). [Google Scholar]

[CIT0035] Meredith C, Frances A, Highland A, et al. 2022. The conservation status of Trillium in North America. Hockessin: New Mexico BioPark Society; Albuquerque: Mt. Cuba Center. [Google Scholar]

[CIT0036] Meyers SC, Jaster T, Mitchell KE, Hardison LK.. 2015. Flora of Oregon, Vol. 1. Fort Worth: Brit Press. [Google Scholar]

[CIT0037] Mishler BD. 2009. Three centuries of paradigm changes in biological classification: is the end in sight? Taxon 58: 61–67. [Google Scholar]

[CIT0038] Münkemüller T, Lavergne S, Bzeznik B, et al. 2012. How to measure and test phylogenetic signal. Methods in Ecology and Evolution 3: 743–756. [Google Scholar]

[CIT0039] NIST Mass Spectrometry Data Center, and W. Wallace 2022. Mass spectra. In Linstrom PJ, Mallard WG. eds. NIST Chemistry WebBook, NIST Standard Reference Database Number 69 [online database]. Gaithersburg: National Institute of Standards and Technology, https://doi.org/ 10.18434/T4D303 (2021–2022, date last accessed). [DOI] [Google Scholar]

[CIT0040] Osaloo SK, Kawano S.. 1999. Molecular systematics of Trilliaceae II. Phylogenetic analyses of Trillium and its allies using sequences of rbcL and matK genes of cpDNA and internal transcribed spacers of 18S–26S nrDNA. Plant Species Biology 14: 75–94. [Google Scholar]

[CIT0041] Petrén H, Toräng P, Ågren J, Friberg M.. 2021. Evolution of floral scent in relation to self-incompatibility and capacity for autonomous self-pollination in the perennial herb Arabis alpina. Annals of Botany 127: 737–747. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0042] Prieto-Benítez S, Millanes AM, Dötterl S, Giménez-Benavides L.. 2016. Comparative analyses of flower scent in Sileneae reveal a contrasting phylogenetic signal between night and day emissions. Ecology and Evolution 6: 7869–7881. [Google Scholar]

[CIT0043] Raguso RA. 2008. Wake up and smell the roses: the ecology and evolution of floral scent. Annual Review of Ecology, Evolution, and Systematics 39: 549–569. [Google Scholar]

[CIT0044] Schiestl FP. 2010. The evolution of floral scent and insect chemical communication. Ecology Letters 13: 643–656. [DOI] [PubMed] [Google Scholar]

[CIT0045] Schiestl FP, Huber FK, Gomez JM.. 2011. Phenotypic selection on floral scent: trade-off between attraction and deterrence? Evolutionary Ecology 25: 237–248. [Google Scholar]

[CIT0046] Schiestl FP, Kirk H, Bigler L, Cozzolino S, Desurmont GA.. 2014. Herbivory and floral signaling: phenotypic plasticity and tradeoffs between reproduction and indirect defense. The New Phytologist 203: 257–266. [DOI] [PubMed] [Google Scholar]

[CIT0047] Schilling EE, Floden A, Lampley J, Patrick TS, Farmer SB.. 2019. A new species in Trillium (Melanthiaceae) from Central Georgia and its phylogenetic position in subgenus Sessilium. Systematic Botany 44: 107–114. [Google Scholar]

[CIT0048] Skogen KA, Jogesh T, Hilpman ET, Todd SL, Raguso RA.. 2022. Extensive population-level sampling reveals clinal variation in (R)-(−)-linalool produced by the flowers of an endemic evening primrose, Oenothera harringtonii. Phytochemistry 200: 113185. [DOI] [PubMed] [Google Scholar]

[CIT0049] Soukup VG. 1980. A new Trillium (Liliaceae) from the northwestern United States. Brittonia 32: 330–333. [Google Scholar]

[CIT0050] Theis N, Lerdau M, Raguso RA.. 2007. The challenge of attracting pollinators while evading floral herbivores: patterns of fragrance emission in Cirsium arvense and Cirsium repandum (Asteraceae). International Journal of Plant Sciences 168: 587–601. [Google Scholar]

[CIT0051] Thielle K. 1993. The holy grail of the perfect character: the cladistic treatment of morphometric data. Cladistics 9: 275–304. [DOI] [PubMed] [Google Scholar]

[CIT0052] Wayman KA, de Lange PJ, Larsen L, Sansom CE, Perry NB.. 2010. Chemotaxonomy of Pseudowintera: sesquiterpene dialdehyde variants are species markers. Phytochemistry 71: 766–772. [DOI] [PubMed] [Google Scholar]

[CIT0053] Wayman KA, Reilly MJ, Petlewski AR.. 2023. Taxonomic insights from floral scents of western North American sessile‐flowered Trillium. American Journal of Botany 110: e16255. [DOI] [PubMed] [Google Scholar]

[CIT0054] Weiss J, Mühlemann JK, Ruiz-Hernández V, Dudareva N, Egea-Cortines M.. 2016. Phenotypic space and variation of floral scent profiles during late flower development in Antirrhinum. Frontiers in Plant Science 7: 1903. [DOI] [PMC free article] [PubMed] [Google Scholar]

[CIT0055] Whitkus R, Bryan FA, Les DH, Tyrrell LE.. 1987. Genetic structure in a heterocyanic population of Trillium sessile (Liliaceae). Plant Species Biology 2: 67–73. [Google Scholar]

[CIT0056] Williams NH, Whitten WM.. 1999. Molecular phylogeny and floral fragrances of male euglossine bee-pollinated orchids: a study of Stanhopea (Orchidaceae). Plant Species Biology 14: 129–136. [Google Scholar]

[CIT0057] Zomlefer WB, Williams NH, Whitten WM, Judd WS.. 2001. Generic circumscription and relationships in the Tribe Melanthieae (Liliales, Melanthiaceae), with emphasis on Zigadenus: evidence from ITS and trnL-F sequence data. American Journal of Botany 88: 1657–1669. [PubMed] [Google Scholar]

PERMALINK

Exploring geography and evolutionary history as drivers of variation in floral scent chemistry in western sessile-flowered Trillium using parsimony-constrained phylogenetics

Candela Blanco-Moreno

Kjirsten A Wayman

Alexandru M F Tomescu

Abstract

Background and Aims

Methods

Key Results

Conclusions

INTRODUCTION

MATERIALS AND METHODS

Study species

Fig. 1.

Fig. 2.

Sampling sites and populations

Sample collection

Absorption of volatiles with solid-phase microextraction

Gas chromatography–mass spectrometry analysis

Calculation of floral scent composition per population

Phylogenetic analyses

Character construction.

Table 1.

Taxa.

Analyses.

Statistical analyses

RESULTS

Phylogenetic analyses

Table 2.

Fig. 3.

Fig. 4.

Statistical analyses

Fig. 5.

Table 3.

Table 4.

Fig. 6.

Comparisons of phylogenetic and statistical analyses

DISCUSSION

Geographical or phylogenetic signal in scent chemistry?

Fig. 7.

Are there better and worse ways to construct characters for phylogenetic analyses based on scent chemistry?

Are scent chemistry-based phylogenies congruent with those based on gene sequences or morphology?

Fig. 8.

How do phylogenetic and statistical approaches compare in their ability to address taxonomy?

Table 5.

How do phylogenetic and statistical approaches compare in addressing evolutionary patterns?

Fig. 9.

Floral scent chemistry and the evolution of floral scent

SUPPLEMENTARY DATA

ACKNOWLEDGEMENTS

Contributor Information

FUNDING

LITERATURE CITED

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases