Striking variation of pollinator attracting scent within a highly specialized pollination system

Abstract

1 INTRODUCTION

Flowers, the reproductive organs of angiosperms, display a larger morphological diversity than equivalent structures in any other taxa, making them fascinating study objects in ecology and evolution (Harder & Johnson, 2009). This remarkable diversity has long been thought to be at least partly driven by the association between flowering plants and their pollinators (Darwin, 1862; Harder & Johnson, 2009). At a microevolutionary scale, a wealth of studies have investigated the role of pollinator-mediated selection in the evolution of visual floral traits (Caruso et al., 2019; Harder & Johnson, 2009). Despite their well-established role in attracting pollinating insects, olfactory signals remain less well-studied than visual signals (Raguso, 2008), presumably owing to their typically more complex nature. However, recent literature presents a growing number of studies that documented patterns of within and between population variation in floral scents (e.g. de Manincor et al., 2022; Eisen, Geber, et al., 2022; Friberg et al., 2019; Gfrerer et al., 2021, also reviewed in Delle-Vedove et al., 2017), as well as some studies that investigated phenotypic selection on olfactory signals (e.g. Chapurlat et al., 2019; Gfrerer et al., 2021; Gross et al., 2016; Majetic et al., 2009a; Parachnowitsch et al., 2012) and experimental evolution studies exploring the adaptative dynamics of these particular traits (Gervasi & Schiestl, 2017; Ramos & Schiestl, 2020). Floral scents are usually characterized using two main descriptors: the intensity of scent emission and scent composition. The first descriptor, the scent emission rate, is often thought to increase the distance at which the signal can be perceived by insects, and is therefore predicted to be under directional pollinator-mediated selection. This hypothesis is indirectly supported by studies documenting higher scent emission rates in self-incompatible populations compared to self-compatible ones, where selection on traits responsible for pollinator attraction is expected to be relaxed (Petrén et al., 2021; Zeng et al., 2022). A second line of indirect evidence comes from species with separate sexes (dioecious plants), where male plants commonly emit more scent than females (reviewed in Ashman, 2009, Delle-Vedove et al., 2017). Indeed, (i) sexual selection is expected to be stronger in males compared to females (Arnold, 1994; Bateman, 1948; Delph & Ashman, 2006; Moore & Pannell, 2011), at least in situations where female reproductive success is not pollen limited, and (ii) sexual selection should overlap with pollinator-mediated selection (Barbot et al., 2023; Moore & Pannell, 2011).

On the contrary, the selection acting on the second descriptor of olfactory signalling, the composition of floral scent, should be of a stabilizing type for any given pollinator species. Insect attraction is indeed thought to be usually triggered by specific ratios of ubiquitous compounds, even in highly species-specific systems (Bruce et al., 2005; Raguso, 2008), although direct experimental tests of this hypothesis remain extremely scarce (see Proffit et al., 2020). Consequently, signal composition is expected to be similar among individuals from the same population, including males and females in dioecious species (Ashman, 2009), and also among plants from populations that share the same pollinators. However, all VOCs from a given floral bouquet should not be impacted in the same way by evolutionary forces: whereas such stabilizing selection is expected to act on ratios of active VOCs (i.e. compounds detected and used by pollinators for localizing plants), inactive VOCs should be subject to other processes, including an expected strong effect of genetic drift (Mant et al., 2005; Powers et al., 2022). Consistently with this prediction, in Ophrys sphegodes, biologically active compounds were found to show less intraspecific variation than non-active compounds, an expected signature of stabilizing selection (Ayasse et al., 2000). In the many biological systems where active versus non active VOCs have not been characterized, a parallel investigation of scent composition and neutral genetic diversity should help to interpret the evolutionary significance of geographical variation in floral scents (Delle-Vedove et al., 2017; Whitehead & Peakall, 2009). The few studies that have characterized odour variation and genetic variation in parallel have, for example, documented: (i) a lack of correlation between variation in active VOCs and neutral genetic variation in a set of cultivated varieties of Ficus carica (Cao et al., 2023); (ii) greater variation in floral bouquet than in neutral genetic markers among geographical regions in Ophrys exaltata, possibly due to differences in the local preferences of specific pollinators (Mant et al., 2005) and (iii) a significant correlation between neutral genetic differentiation and odour differentiation in a wind-pollinated species, Schieda globosa, which releases by definition only non-active VOCs (Powers et al., 2022). In other words, among populations that share the same pollinators, stabilizing selection should lead to relatively low scent differentiation, in particular in highly specialized pollination systems. Any observed variation in VOCs proportions should mainly concern inactive compounds and mirror neutral genetic differentiation. In case of variation in pollination system among populations, divergent selection may lead to scent differentiation that may exceed neutral genetic variation.

In the present study, we investigated how scent varies both among populations and between sexes in the dioecious fan palm Chamaerops humilis (Arecaceae). Previous studies found (i) a significant attractivity of this scent towards a specialized pollinator (Derelomus chamaeropis, Curculionidae) through behavioural tests (Dufaÿ et al., 2003) and (ii) a high variability in scent composition among individuals and higher scent emission rates in male plants (Dufaÿ et al., 2004). However, because this variation was only documented in a common garden, using a small set of plants of unknown geographical origin, not much is known regarding the scent variation within and among natural populations or the prevalence of sexual dimorphism in scent emission rates in the wild. The predictions that we aim at testing here are threefold. First, because male traits are generally expected to be under stronger selective pressure to attract pollinators than female traits (Barbot et al., 2023; Moore & Pannell, 2011), we expect male palms to consistently display higher scent emission rates than females regardless of the population/region under scrutiny. If this is indeed the case, males should also be visited by more pollinators than females, which will be tested using pollinator counts. Second, because efficient pollination requires that insects visit both sexes, we expect the composition of the bouquet to be indistinguishable between sexes. Moreover, in this system, where pollination is achieved largely by deceit (pollinators use inflorescences as egg-laying sites, but most eggs laid in female inflorescences will not mature and hatch, Anstett, 1999; Dufaÿ et al., 2004), any detectable difference in scent composition between sexes may be exploited by pollinators and should be counter-selected (Anstett, 1999, Dufaÿ et al., 2004). In such systems, one expects stabilizing selection for inter-sex mimicry to be the main evolutionary force acting on scent composition, whereas sexual selection should impact scent emission rates (Ashman, 2009), as mentioned above. Third, C. humilis is involved in a nursery pollination with a weevil and because the pollination system is highly specific, we expect a strong stability of the attractive signal, even on a large geographical scale. A limitation of our study is that, as it stands, the identity of the active attractive compounds remains unknown. Spatial patterns of scent bouquet variation across different regions of the distribution area will therefore be discussed in parallel with the spatial distribution of neutral genetic diversity. In this context of highly specialized pollination, we expect ratios of active VOCs to be the same across all populations. Because we here investigate the whole bouquet (both active and non-active), we thus expect scent variation to co-vary with neutral genetic differentiation. To test for our three predictions, we examined the pattern of variation in scent quantity (between males and females: prediction 1) and scent composition (between males and females: prediction 2 and among populations and regions: prediction 3). We also collected pollinators and compared their abundance between male and female plants (prediction 1). Finally, we analysed neutral genetic structure among populations, to compare patterns of genetic versus scent differentiation (prediction 3).