Three floral volatiles contribute to differential pollinator attraction in monkeyflowers (Mimulus)

Abstract

Flowering plants employ a wide variety of signals, including scent, to attract the attention of pollinators. In this study we investigated the role of floral scent in mediating differential attraction between two species of monkeyflowers (Mimulus) reproductively isolated by pollinator preference. The emission rate and chemical identity of floral volatiles differ between the bumblebee-pollinated Mimulus lewisii and the hummingbird-pollinated M. cardinalis. Mimulus lewisii flowers produce an array of volatiles dominated by d-limonene, β-myrcene and E-β-ocimene. Of these three monoterpenes, M. cardinalis flowers produce only d-limonene, released at just 0.9% the rate of M. lewisii flowers. Using the Bombus vosnesenskii bumblebee, an important pollinator of M. lewisii, we conducted simultaneous gas chromatography with extracellular recordings in the bumblebee antennal lobe. Results from these experiments revealed that these three monoterpenes evoke significant neural responses, and that a synthetic mixture of the three volatiles evokes the same responses as the natural scent. Furthermore, the neural population shows enhanced responses to the M. lewisii scent over the scent of M. cardinalis. This neural response is reflected in behavior; in two-choice assays, bumblebees investigate artificial flowers scented with M. lewisii more frequently than ones scented with M. cardinalis, and in synthetic mixtures the three monoterpenes are necessary and sufficient to recapitulate responses to the natural scent of M. lewisii. In this system, floral scent alone is sufficient to elicit differential visitation by bumblebees, implying a strong role of scent in the maintenance of reproductive isolation between M. lewisii and M. cardinalis.

Keywords: Antennal lobe; Floral scent; Insect behavior; Olfaction; Speciation; Terpene.

Fig. 1.

Floral volatiles emitted from Mimulus lewisii and M. cardinalis. (A) Gas chromatography mass spectrometry (GCMS) analysis of floral volatiles from bumblebee-pollinated Mimulus lewisii (top, yellow) and hummingbird-pollinated M. cardinalis (bottom, red). Labels specify individual volatiles: a, α-pinene; b, sabinene and β-pinene (left and right smaller peaks, respectively); c, β-myrcene; d, d-limonene (note visible presence in both M. lewisii and M. cardinalis); e, E-β-ocimene. The unknown monoterpene at 7.92, γ-terpinene and terpinolene are not labeled due to low abundance. Mimulus cardinalis-specific volatiles include 1-octen-3-ol (f); and farnesene isomers (g and h). Notable contaminants are indicated with i. (B) Non-metric multidimensional scaling (NMDS) plot of the 12 volatiles present in M. lewisii and M. cardinalis (see supplementary material Table S1 for details); stress=0.088. Individual points represent single headspace collections of populations of each species: populations 1–5 represent M. lewisii and 6–10 represent M. cardinalis, with 1 and 6 representing the inbred lines used to determine scent composition for each species. The cluster (1*) on the right side of the plot indicates the close clustering of the M. lewisii inbred line, with nine samples in the cluster. For a list of individual populations in this figure, see supplementary material Table S1.

Fig. 2.

Responses of Bombus vosnesenskii antennal lobe neurons to gas chromatography-fractionated scent from M. lewisii flowers. (A) Depiction of gas chromatography multichannel recording (GCMR). Effluent from the GC is split such that half enters the GC's detector (flame ionization detector, FID) while the other half arrives simultaneously at the bee's antenna. (B) Rate histograms (bin=200 ms) of neural unit responses to the eluting compounds from the M. lewisii headspace extract (3 μl injection). Certain volatiles evoked significant unit responses [e.g. myrcene (myr), limonene (lim) and ocimene (oci); gray bars]. However, not all units were responsive to the eluting volatiles (e.g. units 58 and 104). (C) Unit responses for each volatile eluted from the GC. The top plot shows the chromatogram with each peak corresponding to a volatile. Only those units that demonstrated significant responses (response index, RI>2.0, RI<−2.0 s.d.) are shown (color scale, bottom plot). Note that the population responses clustered around a group of three volatiles (myr, lim and oci; outlined by a red box) within the floral headspace. Volatiles are ordered corresponding to the retention time, except for those volatiles that gave robust responses (volatiles myr, lim and oci), which were rearranged for clarity. (D) The percentage of responsive units in each ensemble was determined for each volatile in the floral headspace and plotted for each preparation. A threshold of 2 s.d. of the entire data set for each species was used to identify the volatiles that evoked the greatest activity: d-limonene, E-β-ocimene and β-myrcene. Volatiles that evoked significant unit responses are α-pinene (α-pin), sabinene (sab), β-pinene (β-pin), terpinolene (ter), γ-terpinene (γ-ter) and an unknown monoterpene (Unk. MO).

Fig. 3.

Identification of bioactive volatiles and their chemical class using GCMR. Analysis of the population-level neural activity in response to the different volatiles revealed that units were responsive to acyclic monoterpenes (aMO) and cyclic monoterpenes (cMO), but units were also broadly responsive to both monoterpene types (e.g. d-limonene, ocimene and myrcene) (All MO). Pie charts at the top are the percentage of units responding to the individual volatiles; letters next to the pie charts denote the individual volatiles.

Fig. 4.

AL neural responses to floral extracts, artificial mixtures and single volatiles. (A) Multichannel recording in the bee's AL allowed examination of the bee's ability to discriminate between volatile stimuli. On the right are peristimulus time histograms (PSTHs) and raster plots of a unit that showed significant responses (based on CUMSUM test) to both floral headspace and artificial mixture. (B) Response of one 16-unit ensemble to the different volatile stimuli shown, plotted as color-coded response matrices across all units (rows 1–16) and volatile stimuli (columns 1–12): M. cardinalis (MC), M. cardinalis synthetic stimulus (MC-syn; contains only MC-intensity limonene), M. lewisii (ML), M. lewisii synthetic mixture (ML-syn), ML-intensity limonene (volatile 1), 10×MC-intensity limonene (volatile 2), ML-intensity myrcene (volatile 3), ML-intensity ocimene (volatile 4), Petunia integrifolia (PI, scent 5), Oenothera speciosa (OS, scent 6) and Peniocereus gregii (PG, scent 7). In addition, hexane was tested as a negative control (Ctrl). (C) Principal components analysis of ensemble responses. Yellow circles correspond to the natural ML and synthetic ML, red circles to the natural MC and synthetic MC, and gray circles to the single volatiles. Note the clustering of the natural and synthetic ML relative to the single volatiles and other mixtures. (D) Dissimilarity indices in the ensemble firing rates in response to volatile stimuli (N=8 preparations, from as many bees). Dissimilarity indices are shown with the ML as the origin (top) or MC as the origin (bottom). Hatched bars designate the synthetic flower scents (ML-synthetic, MC-synthetic). Bars are the mean ± s.e.m.; asterisks denote a significant difference between treatments and the control (P<0.05).

Fig. 5.

Unit responses to mixtures and single volatiles. (A) The percentage of responsive units relative to the flower scents, synthetic mixtures and single volatiles. The M. lewisii scent (ML) and synthetic M. lewisii (ML-syn) evoked a significantly greater proportion of the neural units relative to the mineral oil (no odor) control or the M. cardinalis scent (two-by-two χ²: *P<0.001). n.s., not significantly different. (Bi) Units that showed similar (‘hypoadditive’; white bars), synergistic (gray bars) or suppressive (black bars) responses to the mixture (top) relative to the single volatiles (bottom) that evoked the greatest responses. For these three units (each from a different preparation), limonene elicited the greatest response. Gray bars denote the stimulus duration (500 ms). (Bii) Percentage of responsive units that showed hypoadditive, synergistic or suppressive responses.

Fig. 6.

Behavioral responses of B. vosnesenskii to floral bouquets, synthetic mixtures (hatched bars) and individual floral volatiles. (A) Total flower choices by experienced B. vosnesenskii workers trained to either natural ML or the synthetic mixture. Bars are the mean percentage of animals responding to the two-choice treatments. (B) The time the bees spent attempting to feed from the two-choice treatments. Bars are the mean percentage of total time for each treatment. Asterisks denote a significant difference between the two-choice treatments (t-test or χ²-test: P<0.05). Numbers in parentheses indicate the number of bees used in each two-choice treatment. ML-syn, synthetic mixture of three compounds.