J Chem Ecol DOI 10.1007/s10886-015-0599-1
Spatiotemporal Floral Scent Variation of Penstemon digitalis Rosalie C. F. Burdon 1 & Robert A. Raguso 2 & André Kessler 3 & Amy L. Parachnowitsch 1
Received: 20 March 2015 / Revised: 2 June 2015 / Accepted: 9 June 2015 # Springer Science+Business Media New York 2015
Abstract Variability in floral volatile emissions can occur temporally through floral development, during diel cycles, as well as spatially within a flower. These spatiotemporal patterns are hypothesized to provide additional information to floral visitors, but they are rarely measured, and their attendant hypotheses are even more rarely tested. In Penstemon digitalis, a plant whose floral scent has been shown to be under strong phenotypic selection for seed fitness, we investigated spatiotemporal variation in floral scent by using dynamic headspace collection, respectively solid-phase microextraction, and analyzed the volatile samples by combined gas chromatography–mass spectrometry. Total volatile emission was greatest during flowering and peak pollinator activity hours, suggesting its importance in mediating ecological interactions. We also detected tissue and reward-specific compounds, consistent with the hypothesis that complexity in floral scent composition reflects several ecological functions. In particular, we found tissue-specific scents for the stigma, stamens, and staminode (a modified sterile stamen common to all Penstemons). Our findings emphasize the dynamic nature
Electronic supplementary material The online version of this article (doi:10.1007/s10886-015-0599-1) contains supplementary material, which is available to authorized users. * Rosalie C. F. Burdon [email protected] 1
Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Uppsala 75236, Sweden
2
Department of Neurobiology and Behavior, Cornell University, Ithaca, NY 14853, USA
3
Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY 14853, USA
of floral scents and highlight a need for greater understanding of ecological and physiological mechanisms driving spatiotemporal patterns in scent production. Keywords Diel variation . Floral scent . Nectar scent . GC/ MS . S-(+)-linalool . Pollen odor . Staminode
Introduction Plants emit a variety of floral volatile compounds (Knudsen et al. 2006) enabling them to communicate and interact with mutualist pollinators (Raguso 2008a; Wright and Schiestl 2009) and antagonists, such as florivores (Kessler et al. 2013) and herbivores (Kessler and Baldwin 2007; Theis and Adler 2012). However, the chemical composition of floral scent is not static. Spatiotemporal variability in the identity and complexity of scent bouquets could provide critical information for the mediation of plant-animal communication because floral visitors can use subtle differences in volatiles to make foraging choices (Wright and Schiestl 2009). For instance, spatial variation in scent composition between floral tissues may inform visitors about reward location within a flower (Dobson et al. 1999; Dötterl and Jürgens 2005; Piskorski et al. 2011; Raguso 2004), whereas temporal variation in floral scent through floral development or in a circadian rhythm could inform visitors of a flower’s current status (Ruíz-Ramón et al. 2014; Theis et al. 2007). Scent emission often marks floral receptivity (Bergström et al. 1995; Raguso et al. 2003; Rodriguez-Saona et al. 2011), whereas reduced emissions often typify flowers that have been pollinated (Muhlemann et al. 2006; Theis and Raguso 2005; Tollsten 1993). Temporal variation in scent emission also is thought to match a pollinator’s activity schedules or to avoid those of antagonists (Borges et al. 2013; Dötterl et al. 2012; Jürgens
J Chem Ecol
et al. 2014). Therefore, where and when floral volatile compounds are emitted may provide clues to their ecological function (Muhlemann et al. 2014), thus suggesting a key component of studying floral scent should be to characterize its spatiotemporal variation. Using Penstemon digitalis, we highlight how spatiotemporal scent variation could provide additional information to floral visitors, and we suggest testable hypotheses for functions of particular components of the floral scent bouquet. The genus Penstemon is becoming a model system for floral biology research, in part because of its diversity and multiple evolutionary transitions from bee to hummingbird pollination (Castellanos et al. 2004; Wessinger et al. 2014; Wilson et al. 2007). Until recently, studies of scent in the genus Penstemon have been absent (Parachnowitsch et al. 2012, 2013). For beepollinated P. digitalis, Parachnowitsch et al. (2012) found natural selection to increase floral scent emission rates in a common garden population, suggesting that scent plays an important role in Penstemon reproductive ecology. Given the paucity of published estimates of natural selection on floral scent (Ehrlén et al. 2012; Schiestl et al. 2011), P. digitalis provides an opportunity to dissect floral scent variation in a microevolutionary framework. The floral bouquet of P. digitalis comprises at least 23 volatile aliphatic and aromatic compounds (Parachnowitsch et al. 2012). In particular, linalool has been identified as the target of selection, suggesting that its production could be of particular importance to plant-insect interactions in P. digitalis. Our follow-up study identified the emitted floral linalool to be the S-(+)-enantiomer and revealed that it was present in nectar (Parachnowitsch et al. 2013), suggesting both the potential for direct olfactory/gustatory signalling of reward and the protection of nectar against microbes (Raguso 2004). However, the spatial distribution of linalool beyond the nectary, as well as temporal/ developmental patterns of linalool emission, remained unexplored in P. digitalis. Furthermore, spatiotemporal patterns of the other floral volatiles emitted by P. digitalis also remained unknown. Therefore, we set out to document the spatiotemporal variation of floral volatiles in P. digitalis by measuring: (1) scent changes through floral development; (2) diel patterns of P. digitalis floral emissions to verify whether scent emission matches pollinator activity; and (3) spatial/tissue-specific differentiation in floral volatiles. Finally, we utilized these data to generate predictions about the ecological functions of particular compounds or classes in the floral biology of P. digitalis.
Methods and Materials Study System Penstemon digitalis Nutt. ex Sims (Plantaginaceae) grows in open habitats including roadsides, prairies, and meadows of eastern North America, and is pollinated mainly by small to large–bodied bees (Clinebell and
Bernhardt 1998; Dieringer and Cabrera 2002; Mitchell and Ankeny 2001). The inflorescences display size ranges from 20 flowers. The fused floral corolla forms a bell-shaped gullet and is colored white with contrasting purple striping that varies in presence/absence and intensity (Parachnowitsch and Kessler 2010). Nectar is secreted from the base of the corolla tube (hereafter, the nectary). As is characteristic of the genus, a sterile fifth stamen (staminode) protrudes from the corolla tube (hereafter, the limb). Floral development progresses from bud to male-phase and then female-phase flowers (protandry). Although this species is self-compatible, bagged flowers do not produce viable seeds in our study populations, suggesting pollinators are essential for reproduction (Parachnowitsch et al. 2012). For this study, plants were haphazardly collected from three previously studied source populations (Neimi Road (NR), Whipple Farm (WF), and Turkey Hill (TH)) in Tompkins County, NY, USA. Floral scent is similar in these populations; plants emit the same 23 volatiles, although amounts and relative composition vary among populations (Parachnowitsch et al. 2012), and thus we pooled population samples for scent analysis. Because we used field-collected samples, we acknowledge that detected volatiles could be the result of plant interactions with a range of biotic and abiotic factors such as pollinators, microbes, herbivores, temperature, and humidity. However, our sampling of multiple plants and populations across years should ensure that differences in plant-interactions, while increasing noise around our estimates, are unlikely to bias our results. Plant Volatile Collection We used two different volatile collection techniques - solid-phase micro-extraction (SPME) and dynamic headspace (DH) collection - and analyzed floral emissions with gas chromatography–mass spectrometry (GC/MS). SPME was used to detect tissue-specific scent patterns within flowers (as in Irwin and Dorsett 2002; Raguso 2004) and specific floral stages in P. digitalis because it allows qualitative analyses of volatile organic compounds (VOCs) of dissected floral tissue/stages at a single time point. We mechanically removed tissues or floral structures from whole plants, producing wounds that commonly elicit volatiles associated with damage, especially aliphatic compounds derived from the octadecanoid pathway (Balao et al. 2011; Scala et al. 2013). However, these volatiles are common to all dissected samples and are easily accounted for without compromising our results. Floral samples were placed within oven-cleaned glass vials (1.8–20 ml), in which VOCs were allowed to equilibrate (see Suppl). SPME fibers (65-μ-PDMS/DVB, polydimethylsiloxane/divinylbenzene; Supelco, Bellefonte, PA, USA) then were exposed to equilibrated floral headspace for a minimum of 30 min per sample (see Suppl). Although equilibration varied between 30 and 90 min, it did not qualitatively affect the volatiles detected (data not shown).
J Chem Ecol
Hymenopteran pollinators of P. digitalis show diurnal activity patterns; therefore, flower tissue scent sampling was conducted between 10:00 to 16:00 h. Dynamic headspace was used to quantify temporal differences in whole floral bouquets of inflorescences. Inflorescences were enclosed in modified plastic drinking cups (polyethylene terephthalate, 700 ml volume, Karat® Lollicup USA, Inc., Chino, CA, USA), and pumps pulled air through ORBO-32 traps (SUPELCO® Bellefonte, PA, USA) at a flow rate of 200 ml/min (Parachnowitsch et al. 2012). The DH sampling can be used for quantitative analysis of emissions because air sampling with replacement allows calculation of a standardized rate of scent emission per floral unit. Developmental Volatile Variation To test for qualitative ontogenic differences in volatile production, we opportunistically sampled volatiles of P. digitalis during 2012–2014. Buds, flowers, and developing fruits are displayed simultaneously on an inflorescence during peak flowering. Thus, for direct ontogenic comparisons, we separated inflorescences and sampled volatiles from the different developmental stages using SPME. Flowers were initially separated into male and female-phase, however, we found little qualitative variation between phases (not shown), so we pooled samples for ‘flowers’ across sexual phase. Although flowers may differ in quantitative floral scent emission during different sexual phases, we did not use DH to measure such differences because it would have required the selective removal of different phase flowers from the inflorescence, leaving a wounded panicle of same-sex flowers. Diel Volatile Variation To determine whether P. digitalis floral scent differed between day and night, we sampled the headspace of whole plants in the field over 24 h. We collected scent from 12 plants for 8 h intervals (21:00–05:00, 05:00– 13:00, and 13:00–21:00) from two populations over two consecutive 24-h periods. Neimi Road plants were sampled 2829th June and TH sampled on the 30-31st June 2007. The number of open flowers per plant was counted and used as a covariate for emission variability. For all sampling periods, we also collected two ambient and two vegetative volatile samples to distinguish floral compounds from background contaminants. Night samples were checked for additional compounds, otherwise quantification followed Parachnowitsch et al. (2012). Within-flower Spatial Volatile Variation To identify the approximate sources of within-flower volatile production using SPME, we dissected flowers into component parts: corolla, staminode, stigma, anthers, and rewards (pollen and nectar). We divided the corolla into the nectary and limb by using sterile razor blades (see Suppl). Further dissections of the corolla limb were unnecessary because there was no detectable scent (see Results). The stigmas, anthers, and staminodes were
dissected from >20–100 flowers and were pooled to concentrate tissue specific volatiles. One anther sample was vortexed (~2 min) to increase pollen dehiscence. Nectar scent was difficult to obtain due to low standing nectar crops, therefore, we supplemented flowers with 3 μl of 25 % water-to-weight sucrose solution and let the artificial nectar stand in the flowers for 90 min (Raguso 2004). We collected this ‘nectar’ from >40 flowers with filter paper wicks, and compared volatiles detected to samples taken from wicks with sucrose solution only. Further details on nectar volatile analysis follow those given by Parachnowitsch et al. (2013). Volatile Quantification and Identification All volatile analyses conducted on GC/MS were done using a polar EC WAXcolumn (30 m, 0.25 mm internal diam, 0.25 μm film thickness; Alltech Associates, USA) to optimize separation and identification of a wide range of volatiles. However, due to the rapid onset of the blooming season for P. digitalis, we did not switch to an enantioselective GC-phase in order to analyze additional replicates by tracking flower parts or times for changes in the enantiomeric composition of chiral compounds. Our previous work with a chiral column focused exclusively on linalool, as it was identified as the target of selection (Parachnowitsch et al. 2013). Spatial variation and developmental stage samples of tissue-exposed SPME fibers were injected into a GC-17A equipped with a Shimadzu QP5000 quadrupole electron impact MS (70 eV) as a detector. The carrier gas was ultrapure helium with a flow rate of 1 ml/min and a split ratio of 12, with injection port temperature of 240 °C and detector temperature of 260 °C. Oven temperature increased from 40 °C to 260 °C by 10 °C per min, where it was held for 7 min, as described by Levin et al. (2001). Volatiles collected on ORBO vials were first eluted with 350 μl dichloromethane (SIGMA®, St Louise, MO, USA) with 430 ng tetralin (SIGMA®) as an internal standard and analyzed using a Varian 2200 GC/MS. Again, the carrier gas was helium, kept at a constant flow of 1 ml/min. GC oven conditions for these samples were as follows: 45 °C for 6 min, increased to 130 °C at 10 °C/min, increased to 180 °C at 5 °C/min, increased to 230 °C at 20 °C/min with a 5 min hold at 230 °C, increased to 250 °C followed by a final hold at 250 °C for 5 min. Methods for quantifying compound emissions by GC/MS were based on Parachnowitsch et al. (2012), and (2013) we used the same ion fragments and retention indices to identify volatiles. The identities of additional compounds were suggested through the use of mass spectral libraries [Wiley and NIST libraries] and were verified whenever possible using retention times and mass spectra of authentic standards. Peak areas of total ion chromatograms were integrated using Shimadzu’s GCMSolutions (SPME) or Varian® MS Data Review software (DH). Dynamic headspace samples are expressed as tetralin (internal standard) equivalents with the mean air control values subtracted (negative values were
J Chem Ecol
converted to zero). We omit compounds Unidentified 1 and (Z)-6-nonenal identified in Parachnowitsch et al. (2012) from our analysis due to lack of difference from air controls or peak irregularities. Unidentified peaks 1–4 follow Parachnowitsch et al. (2012). We have since identified Bunidentified 2^ to be (E)-4,8-dimethylnona-1,3,7-triene (E-DMNT). New unidentified peaks are numbered according to retention time, starting with 5 (see Suppl). Statistical Analyses Because SPME sampling does not easily allow for direct quantitative comparisons among samples, for the developmental stages and tissue specific VOCs we present the mean relative abundance of volatiles. Percentages were calculated by dividing each sample VOC peak area by the tissue sample’s sum of all volatile peak areas and multiplying by one hundred. Buds, flowers, corolla, staminode, stigma, and seed samples were mass standardized by dividing peak areas by dry mass of tissues. We summed compounds by biosynthetic origin [aliphatics (non-terpenoids, and terpenoids), aromatic esters and nitrogen-containing compounds] to allow comparisons across stages or tissues. Two approaches were used to explore day/night variation in scent production. First, we visualized plant differences in overall scent emission using the Random Forest classification algorithm (Ranganathan and Borges 2010). The analysis used two categories from pooled data across the two populations to increase power: (1) day (mean of morning and afternoon samples), and (2) night, after determining that the day samples did not differ substantially (not shown). We used 200 bootstrap iterations with the package VarSelRF in the statistical software program R (R Development Core Team 2009). Here, ‘out of the bag’ (OOB) probability of membership produces tree iterations excluding or retaining a replicate (based on compound emissions), estimating the likelihood of it belonging to either day or night classification. An internal error estimate is determined on group placement and thus defines a probability of group assignment. Additionally, an estimate of a decrease in accuracy is generated, indicating which compounds were most important in placing replicates into the groups (see Suppl). We interpreted the results by examining the likelihood of each replicate belonging to either day or night based on predictive strength, a mean of iteration predictions from bootstrap sampling (plotted in Fig. 2a). We interpret outliers as particularly strong or weakly scented plants as expected in natural populations. Second, we used models to determine differences in specific compound emissions during day vs. night. To assess if total emission changed over the day/night cycle, we summed peak areas per plant at each time point for the compounds detected using DH (21 compounds; see Results). To test the difference between total day and night emissions and class averages, two-tailed t-tests were performed using R function t-test (R Development Core Team 2009). Differences for individual compound emissions were
assessed using non-linear mixed effect models. We used package nlme in R to perform weighted linear mixed effect models. The use of weighted lme models corrects for increasing variance in residuals with near-zero data so that model assumptions are met, following the protocol in Zuur et al. (2010). Within these models, compound was the continuous dependent variable and time was the categorical independent variable; display (continuous) and plant (categorical) were random effects, with the random effect of population excluded as it explained little of residual variation.
Results Overall, 57 volatile compounds were detected during floral development, with the number, composition and intensity of volatiles differing by stage (Fig. 1a), through time (Fig. 2) and by the plant tissue investigated (Fig. 1b). These volatiles included aliphatic hydrocarbons, alcohols, acids, aldehydes and ketones, aromatic esters, and nitrogen-containing compounds (see Suppl). Although our two different sampling techniques (SPME and DH) overlapped in the compounds detected, they also differed in sensitivity allowing for detection of different compounds. Developmental Volatile Variation In general, most volatiles emitted throughout the floral development of P. digitalis were aliphatic compounds, with terpenoids being the most common. Non-terpenoid compounds dominated the bud and fruit stage, whereas terpenoids dominated floral emissions (Fig. 1). Dominant among the 11 volatiles identified from floral buds were (Z)-3-hexenyl isovalerate, 1-octen-3-ol, and (Z)-3hexenyl acetate (59 %) (Fig. 1a and Suppl), the biosynthesis of which involves the action of a lipoxygenase. Additionally, the mushroom-scented 3-octanol (10 %) and 3-octanone (5 %) were detected. The terpenoid α-bergamotene also constituted a large portion of the bud bouquet (20 %). Flower tissues produced the richest diversity and highest relative abundance of sweet-scented terpenoids (90 %, Fig. 1a), when compared with static emissions from buds and fruits. Floral scent was dominated by the monoterpene (E)-β-ocimene (33 %) and the sesquiterpenes (E,E)-α-farnesene (21 %) and α-bergamotene (21 %), with lower amounts of heptanoic acid (5 %). Additionally detected were, (Z)-β-ocimene (2 %), (Z)-β-farnesene (4 %), (Z,E)- α -farnesene (5 %), S-(+)-linalool (3 %), ycadinene (3 %), trace amounts of α-copaene, and unidentified 15 (2 %) (Fig. 1b). SPME analyses of cut flowers revealed compounds similar to those collected using DH techniques, along with 10 additional flower-specific compounds (see Suppl). The P. digitalis fruits emitted α-bergamotene and germacrene D (11 %), along with 1-octen-3-yl acetate (42 %) and 3-octanone (2 %), methyl salicylate (1 %), and
J Chem Ecol
a Monoterpene Sesquiterpene Other aliphatic
Fruit 20
Flower 11
Bud 11
Aromatic ester N-containing Percentage of volatile class produced per floral development stage Unidentified
b 100
Volatile emission (%)
80
9
Corolla nectary
2 Corolla limb
7 Stigma
2
Staminoid
26
Anther & Pollen
2
Nectar
60
40
20
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
Fig. 1 (a) Pie charts showing the relative abundance of different volatile organic compound (VOC) classes produced at each stage of Penstemon digitalis reproduction. (b) Relative abundance of floral scents produced within tissues of P. digitalis flowers and their detailed composition. The number of compounds are embedded within each pie. Compound numbers are as follows: 1. 3-methyl-1-butanol, 2. (Z)-β-ocimene, 3. (E)-β-ocimene, 4. 3-octanone, 5. octanal, 6. (Z)-3-hexenyl acetate, 7. hexan-1-ol, 8. 3-octanol, 9. nonanal, 10. unidentified alkenal 8, 11. 1-octen-3-yl acetate, 12. α-copaene, 13. 2-butanone, 14. (Z)-3-hexenyl
isovalerate, 15. unidentified 11, 16. unidentified, 12. unidentified 13, 18. pentadecene, 19. S-(+)-linalool, 20. 1-octanol, 21. linalyl acetate, 22. α-bergamotene, 23. undecanal. 24. unidentified alkenal, 25. nonanol, 26. isovaleric acid, 27. (Z)-β-farnesene, 28. unidentified terpenoid (suspected lavender lactone: C 7 H 10 O 2 ), 29. unidentified 16, 30. dodecanal, 31. (Z,E)-α-farnesene, 32. (E,E)-α-farnesene, 33. unidentified alkene 17, 34. methyl nicotinate, 35. hexanoic acid, 36. unidentified nitrogen containing, 37. methyl cinnamate. Numbers within the pies represent the total number of VOCs in the sample
six unidentified compounds (unknowns 5, 6, 10, 14, 18, 19; see Suppl).
unidentified 3 significantly differed through time. The majority (18 of 21) of compounds had higher emissions during day (Fig. 2b), with only two compounds showing significantly increased night emission (methyl salicylate, P=
Spatiotemporal Floral Scent Variation of Penstemon digitalis Rosalie C. F. Burdon 1 & Robert A. Raguso 2 & André Kessler 3 & Amy L. Parachnowitsch 1
Received: 20 March 2015 / Revised: 2 June 2015 / Accepted: 9 June 2015 # Springer Science+Business Media New York 2015
Abstract Variability in floral volatile emissions can occur temporally through floral development, during diel cycles, as well as spatially within a flower. These spatiotemporal patterns are hypothesized to provide additional information to floral visitors, but they are rarely measured, and their attendant hypotheses are even more rarely tested. In Penstemon digitalis, a plant whose floral scent has been shown to be under strong phenotypic selection for seed fitness, we investigated spatiotemporal variation in floral scent by using dynamic headspace collection, respectively solid-phase microextraction, and analyzed the volatile samples by combined gas chromatography–mass spectrometry. Total volatile emission was greatest during flowering and peak pollinator activity hours, suggesting its importance in mediating ecological interactions. We also detected tissue and reward-specific compounds, consistent with the hypothesis that complexity in floral scent composition reflects several ecological functions. In particular, we found tissue-specific scents for the stigma, stamens, and staminode (a modified sterile stamen common to all Penstemons). Our findings emphasize the dynamic nature
Electronic supplementary material The online version of this article (doi:10.1007/s10886-015-0599-1) contains supplementary material, which is available to authorized users. * Rosalie C. F. Burdon [email protected] 1
Plant Ecology and Evolution, Department of Ecology and Genetics, Evolutionary Biology Centre, Uppsala University, Uppsala 75236, Sweden
2
Department of Neurobiology and Behavior, Cornell University, Ithaca, NY 14853, USA
3
Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY 14853, USA
of floral scents and highlight a need for greater understanding of ecological and physiological mechanisms driving spatiotemporal patterns in scent production. Keywords Diel variation . Floral scent . Nectar scent . GC/ MS . S-(+)-linalool . Pollen odor . Staminode
Introduction Plants emit a variety of floral volatile compounds (Knudsen et al. 2006) enabling them to communicate and interact with mutualist pollinators (Raguso 2008a; Wright and Schiestl 2009) and antagonists, such as florivores (Kessler et al. 2013) and herbivores (Kessler and Baldwin 2007; Theis and Adler 2012). However, the chemical composition of floral scent is not static. Spatiotemporal variability in the identity and complexity of scent bouquets could provide critical information for the mediation of plant-animal communication because floral visitors can use subtle differences in volatiles to make foraging choices (Wright and Schiestl 2009). For instance, spatial variation in scent composition between floral tissues may inform visitors about reward location within a flower (Dobson et al. 1999; Dötterl and Jürgens 2005; Piskorski et al. 2011; Raguso 2004), whereas temporal variation in floral scent through floral development or in a circadian rhythm could inform visitors of a flower’s current status (Ruíz-Ramón et al. 2014; Theis et al. 2007). Scent emission often marks floral receptivity (Bergström et al. 1995; Raguso et al. 2003; Rodriguez-Saona et al. 2011), whereas reduced emissions often typify flowers that have been pollinated (Muhlemann et al. 2006; Theis and Raguso 2005; Tollsten 1993). Temporal variation in scent emission also is thought to match a pollinator’s activity schedules or to avoid those of antagonists (Borges et al. 2013; Dötterl et al. 2012; Jürgens
J Chem Ecol
et al. 2014). Therefore, where and when floral volatile compounds are emitted may provide clues to their ecological function (Muhlemann et al. 2014), thus suggesting a key component of studying floral scent should be to characterize its spatiotemporal variation. Using Penstemon digitalis, we highlight how spatiotemporal scent variation could provide additional information to floral visitors, and we suggest testable hypotheses for functions of particular components of the floral scent bouquet. The genus Penstemon is becoming a model system for floral biology research, in part because of its diversity and multiple evolutionary transitions from bee to hummingbird pollination (Castellanos et al. 2004; Wessinger et al. 2014; Wilson et al. 2007). Until recently, studies of scent in the genus Penstemon have been absent (Parachnowitsch et al. 2012, 2013). For beepollinated P. digitalis, Parachnowitsch et al. (2012) found natural selection to increase floral scent emission rates in a common garden population, suggesting that scent plays an important role in Penstemon reproductive ecology. Given the paucity of published estimates of natural selection on floral scent (Ehrlén et al. 2012; Schiestl et al. 2011), P. digitalis provides an opportunity to dissect floral scent variation in a microevolutionary framework. The floral bouquet of P. digitalis comprises at least 23 volatile aliphatic and aromatic compounds (Parachnowitsch et al. 2012). In particular, linalool has been identified as the target of selection, suggesting that its production could be of particular importance to plant-insect interactions in P. digitalis. Our follow-up study identified the emitted floral linalool to be the S-(+)-enantiomer and revealed that it was present in nectar (Parachnowitsch et al. 2013), suggesting both the potential for direct olfactory/gustatory signalling of reward and the protection of nectar against microbes (Raguso 2004). However, the spatial distribution of linalool beyond the nectary, as well as temporal/ developmental patterns of linalool emission, remained unexplored in P. digitalis. Furthermore, spatiotemporal patterns of the other floral volatiles emitted by P. digitalis also remained unknown. Therefore, we set out to document the spatiotemporal variation of floral volatiles in P. digitalis by measuring: (1) scent changes through floral development; (2) diel patterns of P. digitalis floral emissions to verify whether scent emission matches pollinator activity; and (3) spatial/tissue-specific differentiation in floral volatiles. Finally, we utilized these data to generate predictions about the ecological functions of particular compounds or classes in the floral biology of P. digitalis.
Methods and Materials Study System Penstemon digitalis Nutt. ex Sims (Plantaginaceae) grows in open habitats including roadsides, prairies, and meadows of eastern North America, and is pollinated mainly by small to large–bodied bees (Clinebell and
Bernhardt 1998; Dieringer and Cabrera 2002; Mitchell and Ankeny 2001). The inflorescences display size ranges from 20 flowers. The fused floral corolla forms a bell-shaped gullet and is colored white with contrasting purple striping that varies in presence/absence and intensity (Parachnowitsch and Kessler 2010). Nectar is secreted from the base of the corolla tube (hereafter, the nectary). As is characteristic of the genus, a sterile fifth stamen (staminode) protrudes from the corolla tube (hereafter, the limb). Floral development progresses from bud to male-phase and then female-phase flowers (protandry). Although this species is self-compatible, bagged flowers do not produce viable seeds in our study populations, suggesting pollinators are essential for reproduction (Parachnowitsch et al. 2012). For this study, plants were haphazardly collected from three previously studied source populations (Neimi Road (NR), Whipple Farm (WF), and Turkey Hill (TH)) in Tompkins County, NY, USA. Floral scent is similar in these populations; plants emit the same 23 volatiles, although amounts and relative composition vary among populations (Parachnowitsch et al. 2012), and thus we pooled population samples for scent analysis. Because we used field-collected samples, we acknowledge that detected volatiles could be the result of plant interactions with a range of biotic and abiotic factors such as pollinators, microbes, herbivores, temperature, and humidity. However, our sampling of multiple plants and populations across years should ensure that differences in plant-interactions, while increasing noise around our estimates, are unlikely to bias our results. Plant Volatile Collection We used two different volatile collection techniques - solid-phase micro-extraction (SPME) and dynamic headspace (DH) collection - and analyzed floral emissions with gas chromatography–mass spectrometry (GC/MS). SPME was used to detect tissue-specific scent patterns within flowers (as in Irwin and Dorsett 2002; Raguso 2004) and specific floral stages in P. digitalis because it allows qualitative analyses of volatile organic compounds (VOCs) of dissected floral tissue/stages at a single time point. We mechanically removed tissues or floral structures from whole plants, producing wounds that commonly elicit volatiles associated with damage, especially aliphatic compounds derived from the octadecanoid pathway (Balao et al. 2011; Scala et al. 2013). However, these volatiles are common to all dissected samples and are easily accounted for without compromising our results. Floral samples were placed within oven-cleaned glass vials (1.8–20 ml), in which VOCs were allowed to equilibrate (see Suppl). SPME fibers (65-μ-PDMS/DVB, polydimethylsiloxane/divinylbenzene; Supelco, Bellefonte, PA, USA) then were exposed to equilibrated floral headspace for a minimum of 30 min per sample (see Suppl). Although equilibration varied between 30 and 90 min, it did not qualitatively affect the volatiles detected (data not shown).
J Chem Ecol
Hymenopteran pollinators of P. digitalis show diurnal activity patterns; therefore, flower tissue scent sampling was conducted between 10:00 to 16:00 h. Dynamic headspace was used to quantify temporal differences in whole floral bouquets of inflorescences. Inflorescences were enclosed in modified plastic drinking cups (polyethylene terephthalate, 700 ml volume, Karat® Lollicup USA, Inc., Chino, CA, USA), and pumps pulled air through ORBO-32 traps (SUPELCO® Bellefonte, PA, USA) at a flow rate of 200 ml/min (Parachnowitsch et al. 2012). The DH sampling can be used for quantitative analysis of emissions because air sampling with replacement allows calculation of a standardized rate of scent emission per floral unit. Developmental Volatile Variation To test for qualitative ontogenic differences in volatile production, we opportunistically sampled volatiles of P. digitalis during 2012–2014. Buds, flowers, and developing fruits are displayed simultaneously on an inflorescence during peak flowering. Thus, for direct ontogenic comparisons, we separated inflorescences and sampled volatiles from the different developmental stages using SPME. Flowers were initially separated into male and female-phase, however, we found little qualitative variation between phases (not shown), so we pooled samples for ‘flowers’ across sexual phase. Although flowers may differ in quantitative floral scent emission during different sexual phases, we did not use DH to measure such differences because it would have required the selective removal of different phase flowers from the inflorescence, leaving a wounded panicle of same-sex flowers. Diel Volatile Variation To determine whether P. digitalis floral scent differed between day and night, we sampled the headspace of whole plants in the field over 24 h. We collected scent from 12 plants for 8 h intervals (21:00–05:00, 05:00– 13:00, and 13:00–21:00) from two populations over two consecutive 24-h periods. Neimi Road plants were sampled 2829th June and TH sampled on the 30-31st June 2007. The number of open flowers per plant was counted and used as a covariate for emission variability. For all sampling periods, we also collected two ambient and two vegetative volatile samples to distinguish floral compounds from background contaminants. Night samples were checked for additional compounds, otherwise quantification followed Parachnowitsch et al. (2012). Within-flower Spatial Volatile Variation To identify the approximate sources of within-flower volatile production using SPME, we dissected flowers into component parts: corolla, staminode, stigma, anthers, and rewards (pollen and nectar). We divided the corolla into the nectary and limb by using sterile razor blades (see Suppl). Further dissections of the corolla limb were unnecessary because there was no detectable scent (see Results). The stigmas, anthers, and staminodes were
dissected from >20–100 flowers and were pooled to concentrate tissue specific volatiles. One anther sample was vortexed (~2 min) to increase pollen dehiscence. Nectar scent was difficult to obtain due to low standing nectar crops, therefore, we supplemented flowers with 3 μl of 25 % water-to-weight sucrose solution and let the artificial nectar stand in the flowers for 90 min (Raguso 2004). We collected this ‘nectar’ from >40 flowers with filter paper wicks, and compared volatiles detected to samples taken from wicks with sucrose solution only. Further details on nectar volatile analysis follow those given by Parachnowitsch et al. (2013). Volatile Quantification and Identification All volatile analyses conducted on GC/MS were done using a polar EC WAXcolumn (30 m, 0.25 mm internal diam, 0.25 μm film thickness; Alltech Associates, USA) to optimize separation and identification of a wide range of volatiles. However, due to the rapid onset of the blooming season for P. digitalis, we did not switch to an enantioselective GC-phase in order to analyze additional replicates by tracking flower parts or times for changes in the enantiomeric composition of chiral compounds. Our previous work with a chiral column focused exclusively on linalool, as it was identified as the target of selection (Parachnowitsch et al. 2013). Spatial variation and developmental stage samples of tissue-exposed SPME fibers were injected into a GC-17A equipped with a Shimadzu QP5000 quadrupole electron impact MS (70 eV) as a detector. The carrier gas was ultrapure helium with a flow rate of 1 ml/min and a split ratio of 12, with injection port temperature of 240 °C and detector temperature of 260 °C. Oven temperature increased from 40 °C to 260 °C by 10 °C per min, where it was held for 7 min, as described by Levin et al. (2001). Volatiles collected on ORBO vials were first eluted with 350 μl dichloromethane (SIGMA®, St Louise, MO, USA) with 430 ng tetralin (SIGMA®) as an internal standard and analyzed using a Varian 2200 GC/MS. Again, the carrier gas was helium, kept at a constant flow of 1 ml/min. GC oven conditions for these samples were as follows: 45 °C for 6 min, increased to 130 °C at 10 °C/min, increased to 180 °C at 5 °C/min, increased to 230 °C at 20 °C/min with a 5 min hold at 230 °C, increased to 250 °C followed by a final hold at 250 °C for 5 min. Methods for quantifying compound emissions by GC/MS were based on Parachnowitsch et al. (2012), and (2013) we used the same ion fragments and retention indices to identify volatiles. The identities of additional compounds were suggested through the use of mass spectral libraries [Wiley and NIST libraries] and were verified whenever possible using retention times and mass spectra of authentic standards. Peak areas of total ion chromatograms were integrated using Shimadzu’s GCMSolutions (SPME) or Varian® MS Data Review software (DH). Dynamic headspace samples are expressed as tetralin (internal standard) equivalents with the mean air control values subtracted (negative values were
J Chem Ecol
converted to zero). We omit compounds Unidentified 1 and (Z)-6-nonenal identified in Parachnowitsch et al. (2012) from our analysis due to lack of difference from air controls or peak irregularities. Unidentified peaks 1–4 follow Parachnowitsch et al. (2012). We have since identified Bunidentified 2^ to be (E)-4,8-dimethylnona-1,3,7-triene (E-DMNT). New unidentified peaks are numbered according to retention time, starting with 5 (see Suppl). Statistical Analyses Because SPME sampling does not easily allow for direct quantitative comparisons among samples, for the developmental stages and tissue specific VOCs we present the mean relative abundance of volatiles. Percentages were calculated by dividing each sample VOC peak area by the tissue sample’s sum of all volatile peak areas and multiplying by one hundred. Buds, flowers, corolla, staminode, stigma, and seed samples were mass standardized by dividing peak areas by dry mass of tissues. We summed compounds by biosynthetic origin [aliphatics (non-terpenoids, and terpenoids), aromatic esters and nitrogen-containing compounds] to allow comparisons across stages or tissues. Two approaches were used to explore day/night variation in scent production. First, we visualized plant differences in overall scent emission using the Random Forest classification algorithm (Ranganathan and Borges 2010). The analysis used two categories from pooled data across the two populations to increase power: (1) day (mean of morning and afternoon samples), and (2) night, after determining that the day samples did not differ substantially (not shown). We used 200 bootstrap iterations with the package VarSelRF in the statistical software program R (R Development Core Team 2009). Here, ‘out of the bag’ (OOB) probability of membership produces tree iterations excluding or retaining a replicate (based on compound emissions), estimating the likelihood of it belonging to either day or night classification. An internal error estimate is determined on group placement and thus defines a probability of group assignment. Additionally, an estimate of a decrease in accuracy is generated, indicating which compounds were most important in placing replicates into the groups (see Suppl). We interpreted the results by examining the likelihood of each replicate belonging to either day or night based on predictive strength, a mean of iteration predictions from bootstrap sampling (plotted in Fig. 2a). We interpret outliers as particularly strong or weakly scented plants as expected in natural populations. Second, we used models to determine differences in specific compound emissions during day vs. night. To assess if total emission changed over the day/night cycle, we summed peak areas per plant at each time point for the compounds detected using DH (21 compounds; see Results). To test the difference between total day and night emissions and class averages, two-tailed t-tests were performed using R function t-test (R Development Core Team 2009). Differences for individual compound emissions were
assessed using non-linear mixed effect models. We used package nlme in R to perform weighted linear mixed effect models. The use of weighted lme models corrects for increasing variance in residuals with near-zero data so that model assumptions are met, following the protocol in Zuur et al. (2010). Within these models, compound was the continuous dependent variable and time was the categorical independent variable; display (continuous) and plant (categorical) were random effects, with the random effect of population excluded as it explained little of residual variation.
Results Overall, 57 volatile compounds were detected during floral development, with the number, composition and intensity of volatiles differing by stage (Fig. 1a), through time (Fig. 2) and by the plant tissue investigated (Fig. 1b). These volatiles included aliphatic hydrocarbons, alcohols, acids, aldehydes and ketones, aromatic esters, and nitrogen-containing compounds (see Suppl). Although our two different sampling techniques (SPME and DH) overlapped in the compounds detected, they also differed in sensitivity allowing for detection of different compounds. Developmental Volatile Variation In general, most volatiles emitted throughout the floral development of P. digitalis were aliphatic compounds, with terpenoids being the most common. Non-terpenoid compounds dominated the bud and fruit stage, whereas terpenoids dominated floral emissions (Fig. 1). Dominant among the 11 volatiles identified from floral buds were (Z)-3-hexenyl isovalerate, 1-octen-3-ol, and (Z)-3hexenyl acetate (59 %) (Fig. 1a and Suppl), the biosynthesis of which involves the action of a lipoxygenase. Additionally, the mushroom-scented 3-octanol (10 %) and 3-octanone (5 %) were detected. The terpenoid α-bergamotene also constituted a large portion of the bud bouquet (20 %). Flower tissues produced the richest diversity and highest relative abundance of sweet-scented terpenoids (90 %, Fig. 1a), when compared with static emissions from buds and fruits. Floral scent was dominated by the monoterpene (E)-β-ocimene (33 %) and the sesquiterpenes (E,E)-α-farnesene (21 %) and α-bergamotene (21 %), with lower amounts of heptanoic acid (5 %). Additionally detected were, (Z)-β-ocimene (2 %), (Z)-β-farnesene (4 %), (Z,E)- α -farnesene (5 %), S-(+)-linalool (3 %), ycadinene (3 %), trace amounts of α-copaene, and unidentified 15 (2 %) (Fig. 1b). SPME analyses of cut flowers revealed compounds similar to those collected using DH techniques, along with 10 additional flower-specific compounds (see Suppl). The P. digitalis fruits emitted α-bergamotene and germacrene D (11 %), along with 1-octen-3-yl acetate (42 %) and 3-octanone (2 %), methyl salicylate (1 %), and
J Chem Ecol
a Monoterpene Sesquiterpene Other aliphatic
Fruit 20
Flower 11
Bud 11
Aromatic ester N-containing Percentage of volatile class produced per floral development stage Unidentified
b 100
Volatile emission (%)
80
9
Corolla nectary
2 Corolla limb
7 Stigma
2
Staminoid
26
Anther & Pollen
2
Nectar
60
40
20
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37
Fig. 1 (a) Pie charts showing the relative abundance of different volatile organic compound (VOC) classes produced at each stage of Penstemon digitalis reproduction. (b) Relative abundance of floral scents produced within tissues of P. digitalis flowers and their detailed composition. The number of compounds are embedded within each pie. Compound numbers are as follows: 1. 3-methyl-1-butanol, 2. (Z)-β-ocimene, 3. (E)-β-ocimene, 4. 3-octanone, 5. octanal, 6. (Z)-3-hexenyl acetate, 7. hexan-1-ol, 8. 3-octanol, 9. nonanal, 10. unidentified alkenal 8, 11. 1-octen-3-yl acetate, 12. α-copaene, 13. 2-butanone, 14. (Z)-3-hexenyl
isovalerate, 15. unidentified 11, 16. unidentified, 12. unidentified 13, 18. pentadecene, 19. S-(+)-linalool, 20. 1-octanol, 21. linalyl acetate, 22. α-bergamotene, 23. undecanal. 24. unidentified alkenal, 25. nonanol, 26. isovaleric acid, 27. (Z)-β-farnesene, 28. unidentified terpenoid (suspected lavender lactone: C 7 H 10 O 2 ), 29. unidentified 16, 30. dodecanal, 31. (Z,E)-α-farnesene, 32. (E,E)-α-farnesene, 33. unidentified alkene 17, 34. methyl nicotinate, 35. hexanoic acid, 36. unidentified nitrogen containing, 37. methyl cinnamate. Numbers within the pies represent the total number of VOCs in the sample
six unidentified compounds (unknowns 5, 6, 10, 14, 18, 19; see Suppl).
unidentified 3 significantly differed through time. The majority (18 of 21) of compounds had higher emissions during day (Fig. 2b), with only two compounds showing significantly increased night emission (methyl salicylate, P=
