Flowering plants emit complex bouquets of volatile organic compounds (VOCs) to mediate interactions with their pollinators. These bouquets are undoubtedly influenced by pollinator-mediated selection, particularly in deceptively-pollinated species that rely on chemical mimicry. However, many uncertainties remain regarding how spatially and temporally heterogeneous pollinators affect the diversity and distribution of floral odour variation. Here, we characterized and compared the floral odours of ten populations of deceptively-pollinated Arum maculatum (Araceae), and inter-annual and decadal variation in pollinator attraction within these populations. Additionally, we transplanted individuals from all sampled populations to two common garden sites dominated by different pollinator species (Psychoda phalaenoides or Psycha grisescens), and compared pollinator attraction rates to investigate whether populations maintained odour blends adapted to a specific pollinator. We identified high within- and among-population variation in a common blend of VOCs found across the range of A. maculatum. We also observed shifts in pollinator community composition within several populations over 1-2 years, as well as over the past decade. Common garden experiments further revealed that transplanted inflorescences generally attracted the dominant local pollinator species in both transplant sites. However, one population (Forêt du Gâvre, France) appears to exclusively attract P. grisescens, even when transplanted to a P. phalaenoides-dominated site. Together, our results suggest that maintaining diverse floral odour bouquets within populations may be advantageous when pollinator communities vary over short timescales. We propose that temporally-replicated ecological data are one potential key to understanding variation in complex traits such as floral odour, and in some cases may reveal resiliency to shifting pollinator communities.

Floral odour collection and identification

Arum maculatum inflorescences open for an anthesis cycle of roughly 24h, with VOC emissions peaking in the late afternoon / early evening of the first flowering day. We therefore carried out all floral odour sampling between 18:00 at the earliest and 20:30 at the latest. We sampled dynamic headspace volatile organic compounds (VOCs) using polydimethylsiloxane (PDMS) coated Gerstel Twister® (Mülheim an der Ruhr, Germany) stir bar sorptive extraction. Inflorescences were wrapped in inert oven bags (Tangan No34 distributed by Migros, Zurich, Switzerland) cut open at least 8cm above the tip of the spathe to prevent condensation, due to the strong thermogenesis of the appendix. Twisters® were inserted in a glass tube through the oven bag at a height even with, but not contacting, the tip of the spadix. 6L of air was pumped over Twisters® at a standard rate of 200mL per minute for 30 minutes – except for five samples from Conteville, France, where sampling was carried out at the same rate over only 15 minutes. At every sampling site, at least one empty oven bag was placed approximately 5 meters away from any inflorescences, and ambient air was passed over Twisters® identically to A. maculatum inflorescences; these samples were used as controls to filter out ambient air VOCs. All samples were transported in glass vials on ice, and stored at -21°C until analysis.

Gas Chromatography

We applied 1µL of internal standard (5µg mg/mL naphthalene in dichloromethane) directly to each Twister® immediately before processing. Using a Multipurpose Sampler (Gerstel, Mülheim an der Ruhr, Germany), VOCs were thermally desorbed and separated on a HP-5MS column, 30 m x 0.25mm x 0.25µm at 40°C for 30 sec, increasing temperature by 5°C per min to 160°C, which was held for 0.01 min before increasing 3°C per min to reach 200°C, which was held for 4 min, finally increasing at 10°C per min. until reaching 250°C for 3 min.

Volatile Data Processing

We aligned peaks by retention time within each population. Major ions were recorded for each integrated peak using Agilent Chemstation software. Putative compound identifications were then derived from NIST 2.3 (library version 17) hits confirmed for the same peak in several spectra; all names used in the final analysis should be considered hypotheses. Compounds present in blank samples with a mean quantity anywhere near those within A. maculatum samples were removed prior to further analyses. Quantitative values were obtained by dividing compound peak areas by the internal standard, then multiplying by the internal standard concentration, and finally scaling based on sampling time (for the few samples run for less than 30 minutes).

Pollinator identification

On the morning after floral odour collection, all trapped pollinators were collected from within each inflorescence and preserved in 70% ethanol. All pollinators were identified to at least the suborder level. Psychodidae were further identified to species level using taxonomic information and illustrations (Ježek 1990). First, the number of antennal segments were counted. 15 segments indicated specimens were likely P. phalaenoides or P. crassipennis. 16 segments indicated either P. grisescens or P. trinodulosa - wing venation patterns were then examined for 16-segmented specimens, as P. trinodulosa has a characteristic disconnection in one branched vein. To confirm the final species identity (particularly when intact antennae were not available) and sex of all psychodids, the reproductive anatomy of specimens were also examined: Psychodid abdomens were separated, flattened, cleared in a diluted solution of potash, and mounted on a slide in glycerol beside their decapitated head, and wings laid out flat. 

The data presented here include individual-level data on 1) floral VOC emissions, in proportional format, 2) pollinator attraction at the species level for Psychodidae, and family level for other taxa, and 3) associated metadata, namely population of origin, transplant location (if applicable), and date of VOC sampling. The VOC data have been pre-filtered to remove compounds identified in blank (ambient air) samples, and compounds that consistently represented less than 1% of floral odour blends. These data may be used to reproduce all analyses and figures in our article. Analysis scripts are avalable from the corresponding author on reasonable request.