Sex and origin-specific inbreeding effects on flower attractiveness to specialised pollinators

We investigate whether inbreeding has particularly fatal consequences for dioecious plants by diminishing their floral attractiveness and the associated pollinator visitation rates disproportionally in females. We also test whether the magnitude of such effects depends on the evolutionary histories of plant populations. We recorded spatial, olfactory, colour and rewarding flower attractiveness traits as well as pollinator visitation rates in experimentally inbred and outbred, male and female Silene latifolia plants from European and North American populations differing in their evolutionary histories. We found that inbreeding specifically impairs spatial and olfactory attractiveness. Our results support that sex-specific selection and gene expression partially magnified these inbreeding costs for females, and that divergent evolutionary histories altered the genetic architecture underlying inbreeding effects across population origins. Moreover, they highlight that inbreeding effects on olfactory attractiveness have a huge potential to disrupt interactions among plants and specialist moth pollinators, which are mediated by elaborate chemical communication.


Introduction
In the present study, we investigated inbreeding effects on plant attractiveness to pollinating 81 insects and its dependence on sex and population origin using the plant species Silene latifolia Pior. 82 (Caryophyllaceae) and its crepuscular moth pollinators. Natural S. latifolia populations partly suffer 83 from biparental inbreeding due to limited seed and pollen dispersal (McCauley, 1997). As inbreeding 84 reduces not only fitness (Teixeira et al., 2009), but also impairs interactions with herbivorous insects fully-opened flower per plant (see 2.3.3 for further details) using the software ImageJ 1.47t (Rueden 153 et al., 2017). 154

Flower scent 155
For characterization of flower scent, we trapped the headspace VOC of S. latifolia flowers on absorbent 156 polydimethylsiloxane (PDMS) tubing following the method of Kallenbach et al. (2014,2015). We placed 157 the plants in a spatial distance of 50 cm to one another in the greenhouse and maintained high air 158 ventilation one week prior to and during VOC collection. We selected one well-developed flower per 159 individual and enclosed it in a VOC collection unit (Fig. S5). The collection units consisted of 160 polypropylene cups with lids (50 mL, Premium Line, Tedeco-Gizeh, GE), both having holes (diameter 161 15 mm) to prevent heat and waterlogging. They were fixed via wooden sticks at the exterior of the 162 plant pot. In addition, 14 control collection units were fixed on empty plant pots and positioned 163 throughout the greenhouse. Prior use, the absorbent PDMS tubes (length 5 mm, external diameter 1.8 164 mm, internal diameter 1 mm; Carl Roth, GE), were cleaned with solvents and heat as described in 165 Kallenbach et al. (2014). Two PDMS tubes were added to each collection unit and remained in the floral 166 headspace between 9 pm and 5 pm, which is the time of peak scent emission in S. latifolia (Dötterl et 167 al., 2005). Afterwards, the PDMS tubes were removed and stored at -20 °C in sealed glass vials until 168 analysis via thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS, TD 30 -GC  169 2010plus -MS QP2020, Shimadzu, JP). 170 Trapped VOC were desorbed from PDMS tubes for 8 min at 230 °C under a helium flow of 60 mL 171 min -1 and adsorbed on a Tenax® cryo-trap with a temperature of -20 °C. From the trap, compounds 172 were desorbed at 250 °C for 3 min, injected to the GC in a 3:10 split mode, and migrated with a helium 173 flow of 1.6 mL min -1 on a VF5-MS column (30 m × 0.25 mm + 10 m guard column, Agilent Technologies, 174 USA). The GC temperature program started at 40 °C for 5 min and increased to 125 °C at a rate of 10 175 °C min -1 with a hold time of 5 min and to 280 °C at a rate of 30 °C min -1 with a hold time of 1 min. Line 176 spectra (30 -400 m/z) of separated compounds were acquired in quadrupole MS mode. An alkane 177 standard mix (C8-C20, Sigma Aldrich, GE) was analysed under the same conditions in order to calculate 178 Kovats retention indices (KI) for targeted compounds (Kováts, 1958 S7). The platform had a fixed location in the field and was oriented towards the setting sun. Raw 199 images were taken with a digital camera (Samsung NX1000, JP) converted to full spectrum sensitivity 200 (300-1000 nm) via removal of the sensor´s filter and fitted with a UV sensitive lens (Nikon EL 80-mm, 201 JP). We took images in the visible and in the UV part of the light spectrum by fitting an UV and infrared 202 (IR) blocking filter (UV/IR Cut, transmittance 400 -700 nm, Baader Planetarium, GE) and an UV pass 203 plus IR block filter (U-filter, transmittance 300 -400 nm, Baader Planetraium, GE) to the lens, 204 respectively. All images were taken as RAWs with an aperture of 5.6, an iso of 800 and a shutter speed 205 varying according to light conditions. 206 Images were processed using the Multispectral Image Calibration and Analysis (MICA)-Toolbox 207 plugin (Troscianko and Stevens, 2015) in Image J 1.47t (Rueden et al., 2017). They were linearised to 208 correct for the non-linear response of the camera to light intensity and equalised with respect to the 209 two light standards in order to account for variation in natural light perceived among images (Stevens 210 et al., 2007). All petals were selected for analysis, and the reproductive organs and para-corolla were 211 omitted. Linearised images were then mapped to the visual system of a nocturnal moth. As the visual 212 system of H. bicruris is unexplored, we used the tri-chromatic visual system of Deilephila elpenor L. 213 Seed set was not quantified, since the exchange of pollen within and among breeding 253 treatments and population origins could not be controlled, which would have severely confounded 254 their effects. However, we consistently removed ripe fruits (brownish colour, dry fruit wall showing 255 first signs of dehiscence) from female plants in order to prevent the escape of genetic material from 256 our experiment without impeding resource investment in fruit and seed production. responses with an access of zeroes (intensity of lilac aldehydes and pollinator visitation rates). These 274 models were additionally fitted with zero-inflation formulas. The fit of lilac aldehydes models was best 275 when including only an intercept model for zero-inflation, whereas the fit of pollinator visitation rate 276 models was best when including the same predictors and random effects in the conditional and zero 277 inflation part of the model. 278 All of the described models (Table 1) Table 2). Inflorescences of inbreds were shorter than those 289 of outbreds (p < 0.001, χ²(1DF) = 37.31, Fig. 1a). Flower number (Fig. 1b) was higher in plants from North 290 America than Europe (p = 0.005, χ²(1DF) = 8.01) and additionally depended on the interaction breeding 291 treatment × sex (p = 0.003, χ²(1DF) = 8.99). Inbred plants generally produced fewer flowers than 292 outbreds, and this effect was more severe in females (ppost < 0.001) than males (ppost = 0.011). The 293 number of flowers produced was higher in male than female plants in both inbreds (ppost < 0.001) and 294 outbreds (ppost < 0.001). The area of petal limbs (Fig. 1c) was smaller in female than male plants (p < 295 0.001, χ²(1DF) = 51.35) and reduced by inbreeding (p = 0.002, χ²(1DF) = 9.25). The expansion of the corolla 296 depended on the interaction breeding treatment × sex (p = 0.004, χ²(1DF) = 8.17). Inbreeding reduced 297 corolla expansion in female (ppost < 0.001) but not in male plants, and differences between sexes in 298 corolla expansion were consequently apparent in inbreds (ppost < 0.001) but not in outbreds (Fig. 1d). 299 Corolla expansion additionally depended on the interaction sex × origin (p = 0.009, χ²(1DF) = 6.86). 300 Flowers expanded larger in male than female plants in populations originating from North America 301 only (ppost < 0.001). 302 Breeding treatment, sex and population origin affected on the composition of floral VOC in S. 303 latifolia interactively ( Table 2). The Shannon diversity of VOC known to elicit antennal responses in H. 304 bicruris depended on the interaction breeding treatment × origin (p = 0.016, χ²(1DF) = 5.83, Fig. 1e). 305 Inbreeding slightly reduced Shannon diversity of these VOC in European plants (ppost = 0.013) but had 306 no effect on plants from North America. The intensity of lilac aldehyde A depended on the interaction 307 breeding treatment × sex × origin in the conditional model (p = 0.025, χ²(1DF) = 5.03, Fig. 1f). Post-hoc 308 comparisons yielded a marginally significant lower intensity of the compound in inbred than outbred 309 females in plants from North America (ppost = 0.056). Similar non-significant trends were observed for 310 the other lilac aldehyde isomers (Table S5) Table 2). Post-hoc comparisons yielded that flower visits were lower for 327 inbred females (ppost = 0.001) but higher for inbred males (ppost = 0.031) in populations from North 328 America; higher in males than females for European outbreds (ppost = 0.003), European inbreds (ppost = 329 0.027) and North American inbreds (ppost = 0.002) but higher in females than males in outbreds from 330 North America (ppost < 0.001); and higher in North American than European outbred females (ppost = 331 0.001) (Fig 2b). Both, the number of plant and flower visits depended on the interaction breeding 332 treatment × sex × origin in the zero-inflation part of the model as well (Table 2, S9). The direction and 333 magnitude of these effects did not contrast with the conditional models. 334

Discussion 335
Using an integrated methodological approach, we found that i) inbreeding diminishes several flower 336 attractiveness traits in S. latifolia. The magnitude of these effects depended partially on ii) plant sex, 337 which demonstrates that the intrinsic biological differences between males and females shape the 338 consequences of inbreeding in dioecious plant species as they are filtered through the selective 339 environment. Inbreeding effects also depended on iii) origin in a way indicating that divergent 340 evolutionary histories have shaped the underlying genetic architecture. Finally, our study showed that 341 iv) the effects of inbreeding, sex and origin on pollinator visitation rates specifically mirrored variation 342 in floral scent, which highlights the large relative importance of this attractiveness trait in shaping the 343 behaviour of crepuscular moths. 344

Inbreeding reduces floral attractiveness 345
In partial accordance with our first hypothesis, inbreeding reduced several, but not all components of 346 floral attractiveness in S. latifolia. Spatial attractiveness declined most strongly with inbreeding in 347 males and females from both origins (Fig. 1a-c). These results are in line with previous studies on and over-dominance effects at multiple loci. The diversity and quantity of floral VOC was reduced in a 351 sex and origin-specific manner in inbred relative to outbred S. latifolia ( Fig. 1e-g (2010), we observed larger instead of smaller flowers in males. This may base on the use of a size 382 estimate that accounts for variation in flower shape or the comparably large geographic range and dimorphic trait in the colour appearance of S. latifolia to crepuscular moths in the UV and blue light 385 spectrum (Fig. 1g-h). Given that moths use blue light as a major cue to start feeding on nectar (Cutler 386 et al., 1995), the lower light reflectance observed for females is another trait rendering them less 387 attractive than males. 388 The evolution of lower female attractiveness is driven by sex-specific resource allocation, i.e., 389 high costs of ovary and seeds restrict allocation to floral attractiveness in females, as well as sexual 390 selection, i.e., competition for siring success among males selects for increased attractiveness ( relative contribution of sex-specific selection and gene expression to differences in the magnitude of inbreeding depression between males and females and at predicting their feedback on sex ratios to 420 predict and handle these specific threats. 421

Evolutionary history shapes the genetic architecture underlying inbreeding effects on floral scent 422
Plants exhibited merely one general difference in attractiveness among geographic origins (Fig. 1b). However, we expected that North American populations purged genetic load linked to floral 431 attractiveness during the colonization process (i.e., interaction breeding treatment × origin). In 432 contrast to hypothesis three, the magnitude of inbreeding effects was not consistently higher in 433 European than North American populations. Instead, it was independent of origin for most 434 attractiveness components, except flower scent, and either higher or lower in European plants for 435 different scent traits (Fig. 1f-g). These findings provide no support for recent purging-events in North 436 American populations. They rather add to evidence that the magnitude of inbreeding effects is highly Pollinator visitation rates partially mirrored the above-discussed variation in flower attractiveness. 446 They depended on the breeding treatment in a highly sex-and origin-specific manner: In North 447 American populations, inbred females received significantly fewer plant and flower visits than 448 outbreds, whereas flower visits were higher in inbred males (Fig. 2). Although these findings provide 449 limited support for our fourth hypotheses, they yield interesting insight into the relative importance We conclude that the severe inbreeding effects on spatial attractiveness alone do not 452 necessarily reduce moth visitation rates because these effects were independent of plant sex and 453 origin. A trait that was negatively affected by inbreeding only in North American females, just like 454 pollinator visitation rates, was the concentration of lilac aldehyde A (Fig. 1f). The other lilac aldehyde 455 isomers exhibited similar but non-significant trends (Table S5) North American inbred females from the distance or to an immediate switch to more attractive plants 461 following first probing in our experiment. The non-significant trend for higher relative amounts of lilac 462 aldehyde in inbred than outbred males from North America (Table S5)  Manduca sexta is mediated mainly by blue receptors, but where are they located in the 519 retina? J. Exp. Biol. 198, 1909-1917 Dafni, A., Lehrer, M., and Kevan, P. G. (1997). Spatial flower parameters and insect spatial vision. Biol. Estimates were extracted from (generalised) linear mixed effects models for significant interaction effects and main effects of factors not involved in an 705 interaction (significance levels based on Wald χ²-tests denoted at top of plot). Interaction effect plots additionally indicate significant differences among breeding 706 treatments, sexes or origins within levels of other factors involved in the respective interaction (estimated based on post-hoc comparisons, denoted within plots). 707 Exact sample sizes for all traits are listed in Table 1. Significance levels: ***: p < 0.001, **: p < 0.01, *p < 0.05, • p < 0.06. 708 sizes for all traits are listed in Table 1. Significance levels: ***: p < 0.001, **: p < 0.01, and *p < 0.05. 714  The crossings were performed with five families (numbered grey circles). Females (red plants, leaf 7 pointing to left) were fertilised with pollen from males (blue plants, leaf pointing to right) from the 8 same family for inbreeding (dashed arrows), and with pollen from males from a different family for 9 outbreeding (solid arrows). Inbreeding and outbreeding were performed at distinct flowers of the 10 same female individual. inserted into the collection units and absorbed VOC for a period of eight hours. 18 Figure S6: Setup for the acquisition of digital images for flower colour analyses. The camera was fixed on a tripod positioned on an exact horizontal platform, 23 which was oriented towards the setting sun (a) to take images of flowers in the visible light spectrum (b) and the ultra violet light spectrum (c). Images included 24 an intact and fully opened flower (i) that was carefully plugged into a black ethylene vinyl acetate sheet equipped with a label (ii), a size standard (iii), a 10 % 25 polytetrafluorethylene light standard (iv) and a 99 % spectralon light standard (v). 26  were extracted for significant interaction effects from the zero-inflation part of generalised linear 56 mixed effects models (significance levels based on Wald χ²-tests denoted at top of plot). Plots 57 additionally indicate significant differences between breeding treatments, sexes or origins within 58 levels of other factors involved in the respective interaction (estimated based on post-hoc 59 comparisons, denoted within plots). Exact sample sizes for all responses are listed in Table 1.  60 Significance levels: ***: p < 0.001, **: p < 0.01, and *p < 0.05. 61 The number of pollinator visits per plant was shaped by the interaction breeding treatment × sex × 62