Quantitative variation in female sensory structures supports species recognition and intraspecific mate choice functions in damselflies

Males and females exchange signals prior to mating that convey information such as sex, species identity, or individual condition. Tactile signals relayed during physical contact between males and females before and during mating appear to be important for mate choice and reproductive isolation in some animals. However, compared to our understanding of visual, auditory, and chemical signals, we know little about the importance of tactile signals in mating decisions. Among North American damselflies in the genus Enallagma (Odonata: Coenagrionidae) species-specific tactile stimulation contributes to reproductive isolation between species and may also be important for intraspecific mate choice. We quantified several mechanosensory sensilla phenotypes on the female thorax among multiple sympatric and allopatric populations of two Enallagma species that occasionally interbreed in nature. Although each species differed in features of sensilla distribution within the thoracic plates, we found no strong evidence of reproductive character displacement among the sensilla traits we measured in regions of sympatry. However, substantial variation of sensilla traits was observed within populations of both species. Our results suggest that species-specific placement of female mechanoreceptors appears sufficient for species recognition, but mechanosensor variation among females within species may be important for mate choice.

isolation in some animals. However, compared to our understanding of visual, auditory, and 23 chemical signals, we know little about the importance of tactile signals in mating decisions. 24 Among North American damselflies in the genus Enallagma (Odonata: Coenagrionidae) species-25 specific tactile stimulation contributes to reproductive isolation between species and may also be 26 important for intraspecific mate choice. We quantified several mechanosensory sensilla 27 phenotypes on the female thorax among multiple sympatric and allopatric populations of two 28 2 Enallagma species that occasionally interbreed in nature. Although each species differed in 29 features of sensilla distribution within the thoracic plates, we found no strong evidence of 30 reproductive character displacement among the sensilla traits we measured in regions of 31 sympatry. However, substantial variation of sensilla traits was observed within populations of 32 both species. Our results suggest that species-specific placement of female mechanoreceptors 33 appears sufficient for species recognition, but mechanosensor variation among females within 34 species may be important for mate choice.  grasping appendages contact the female thorax before and during mating, which has led to 93 speculation that they allow females to evaluate male morphologies and identify conspecifics 94 (Jurzitza 1974(Jurzitza , 1975Tennessen 1975;Robertson and Paterson 1982;Battin 1993a,b). Each  interspecific hybrid males that take them in tandem, which shows that not only can E. anna and 120 E. carunculatum females detect large differences in male cercus morphologies, but also more 121 subtle differences such as those between conspecific and hybrid males (Barnard et al. 2017).

122
Because it appears that mesostigmal sensilla are used to mediate species recognition, they

156
Trait imaging and quantification 157 We photographed each damselfly using a Nikon D5100 camera (16.2 MP; Nikon 158 Corporation, Tokyo, Japan). We dissected the ventral thoracic cuticle from each female using 159 forceps and imaged the mesostigmal plates using scanning electron microscopy. Specimens 160 were mounted on aluminum stubs with carbon tape, sputter-coated with gold-palladium, and 161 imaged at 200X magnification and 3kV using a Zeiss NEON scanning electron microscope.

162
To avoid any potential bias during measurements, we blind-coded image files before 163 measuring traits. We measured abdomen length (abdominal segments 1-10, excluding terminal 164 appendages) on the full-body photos as a proxy for body size using the segmented line tool in 165 ImageJ (Abramoff et al. 2004). We quantified sensilla traits on the right mesostigmal plate of 166 each female damselfly unless the right plate was dirty or damaged, in which case we quantified 167 the left plate. Sensilla counts on a subset of 57 females showed that left plate and right plate 168 sensilla counts are highly correlated (r = 0.85). In cases where we quantified the left plate, we 169 flipped the image horizontally, so it was in the same orientation as a right plate. We   181 We conducted all morphometric and statistical analyses using R v. 3.4.1 (R Core Team 182 2015). We used the plate outline coordinates to calculate each plate's two-dimensional area.

183
To calculate the area of the sensilla-covered region of each plate, we generated a polygon 184 connecting the coordinates of the outermost sensilla and calculated the area within this outline. 185 We determined the proportion of each plate that is covered by sensilla by dividing the sensilla 186 area by total plate area. We calculated sensilla density in two ways. First, we divided sensilla     (Table 1). To avoid psedoreplication, for each population with N > 1, our analyses used 230 population means of trait values, so that each population was represented by a single 231 measurement. We arcsin transformed proportion data prior to analysis. We pooled data for 232 locally allopatric and fully allopatric E. carunculatum after t-tests showed that these groups did     (Table S1). However,   (Figure 2A) and the proportion of the plate covered by sensilla ( Figure 2B).

292
However, these trends were not statistically significant (sensilla number with locally allopatric 293 outlier removed: Bartlett's K 2 1 = 0.75, P = 0.39; proportion of plate covered by sensilla: Bartlett's 294 K 2 1 = 2.5, P = 0.11). KDE comparisons also did not reveal significant differences in sensilla 295 distributions between sympatric and allopatric E. carunculatum populations (Table S2). 296 However, the analysis revealed significant differences in sensilla distributions between several 297 pairs of populations that are not sympatric with E. anna ( Figure 3E). This result is consistent 298 with those described above that indicated higher variance in sensilla traits among allopatric 299 populations compared to sympatric populations. represents the average mesostigmal plate shape for the population. Asterisks indicate E. 309 carunculatum populations whose KDEs were determined to be significantly different (* indicates 310 P < 0.05, *** P < 0.001).

311
Interestingly, although mean trait values did not differ significantly between sympatric 312 and allopatric populations, sensilla traits displayed considerable variation within the populations 313 we sampled. For example, within a single population, a particular female might have twice as 314 many sensilla than another female ( Figure S2). This pattern was also observed in the E. anna 315 populations we studied.  Finally, although we did not detect a statistically significant difference between group 372 means, the small differences we observed may still have biological relevance. If gaining just one