Contrasting patterns of sexually selected traits in Mediterranean and continental populations of European mouflon

Abstract The expression of sexually selected traits in highly dimorphic ungulates may be influenced by environmental quality. Variations in habitat conditions can impose different constraints on the allocation of energy resources to male life‐history traits, and possibly alter the female preferences for specific features. Here, we compared the horn growth patterns in male European mouflon Ovis aries musimon living in different habitats (Mediterranean vs. continental) but sharing a common genetic origin. We hypothesized that the expression of sexually selected traits such as horn development should be promoted in more favorable habitat conditions (i.e., Mediterranean). Using linear mixed models on data retrieved from individuals harvested under the same hunting regime, we found longer horns and greater individual variance in horn segment length in the Mediterranean population than in the continental one. Furthermore, Mediterranean rams showed no evidence of compensatory horn growth, as opposed to the continental rams. Unexpectedly, horn base circumference was greater in the continental habitat than in the Mediterranean one. The overall results suggest different patterns of investment in horns in the two populations, with seemingly stronger pressure and consequences of sexual selection on mouflon rams living in more favorable environments. Although the role of hunters' selectivity cannot be excluded a priori, our data suggest that the differences in the expression of sexually selected traits in our study populations may be influenced by environmental conditions. Because sexual selection can impose substantial fitness costs on individuals, further investigations on the trade‐offs between reproduction and survival would improve our understanding of the dynamics of mouflon populations living in different environmental conditions.

In male bovids under strong pressure of sexual selection (e.g., bighorn sheep: Coltman et al., 2002), weaponry can be energetically demanding to produce and maintain. Intuitively, the expression of this sexual trait, either in terms of absolute size, compensatory growth or individual heterogeneity, can be considered condition-dependent, that is, it can be highly variable in populations living in different environmental conditions (cf. Rowe & Houle, 1996;Zahavi, 1975).
Specifically, sexual selection is expected to be weaker under mild or harsh conditions and stronger under moderate levels of stress, which may enhance expression of genetic variation in individual condition (Hoffmann & Merila, 1999). Recent works pointed out that variation in environmental quality is important in determining the strength of sexual selection, whose effects may alter the rate of adaptation to new environments, indirectly affecting population dynamics and viability (Candolin & Heuschele, 2008;Martínez-Ruiz & Knell, 2017).
For example, qualitative environmental changes can disrupt the preference for a specific sexual trait by reducing its benefit when the link between trait value and individual quality is altered (Wong, Candolin, Lindström, 2007). In addition, the adjustment of trait expression to new environmental conditions is usually followed by phenotypic plasticity, which may change cost and benefits of sexually selected traits (Kokko & Heubel, 2008a;Kokko & Heubel, 2008b).
The strength of sexual selection should thus not be considered as a fixed, species-specific characteristic, as it may vary depending on local environmental conditions (Martínez-Ruiz & Knell, 2017).
The European mouflon Ovis aries musimon is a highly dimorphic caprid that originated in western Asia (Groves & Grubb, 2011), in areas with hot and dry summers and mild winters, but it has been introduced to many European countries (Bon et al., 1991). Despite its adaptability to different environments, the European mouflon maintains a preference for typically Mediterranean shrubby and rocky areas (Pfeffer, 1967), where it feeds on a high diversity of plants, including many species of shrubs (Marchand et al., 2013).
Notwithstanding its importance for consumptive purposes, little is known about the horn growth pattern in European mouflon and the condition-dependent investment in secondary sexual traits (but see: Garel et al., 2007;Lincoln, 1998). Here, we aim to compare the male horn growth patterns of two hunted populations introduced to Mediterranean and continental parts of Croatia. These two populations live in a very different environmental condition, but share the same founders and the same hunting regime.
Because European mouflons are highly polygynous, thus under strong pressure of sexual selection, we expect the expression of sexually selected traits to differ in males living in different habitats. Specifically, we hypothesized that the opportunity to invest in weapon growth should be greater in more favorable and stable environments, such as the Mediterranean one, compared with less favorable and more stochastic environments, such as the continental one (cf. Kokko & Brooks, 2003). Greater food availability should increase the strength of sexual selection (Janicke, David, & Chapuis, 2015), thus positively affect phenotypic and genetic variance of condition-dependent sexually selected traits (David, Bjorksten, Fowler, & Pomiankowski, 2000). Specifically, we predicted that: 1. Male horn length and base circumference size should be greater in the Mediterranean than in the continental population (i.e., stronger sexual selection should favor larger horns).
2. There should not be evidence for compensatory horn growth in males in the Mediterranean population, because relative resource allocation to a specific trait should increase under good environmental and nutritional conditions (Jobling, 2010). In contrast, compensatory growth mechanism might activate when conditions are less favorable (i.e., in the continental area), thus reducing resource allocation to a specific trait (i.e., greater sexual selection should promote a positive relationship between horn annuli in the most favorable area, and partly negative relationship in the less favorable area).
3. Individual heterogeneity in male horn growth should be greater in the Mediterranean than in the continental population (i.e., stronger sexual selection should favor differences among individuals in the most favorable area).

| Study areas and populations
The study was conducted in two areas of Croatia, hosting two separate populations of European mouflon. The northern Dinaric mountains, hereafter "Mediterranean population", and central Croatia, hereafter "continental population" (Figure 1). These populations show major differences in environmental and climatic features (Table 1), with milder and more stable (i.e., more favorable, for European mouflon) environmental conditions in the Mediterranean habitat than in the continental habitat (Seletković, Tikvić, Vučetić, & Ugarković, 2011). Both populations were introduced in the 1980s: Notably, the same source population from Brijuni islands was used for introduction (Kusak & Krapinec, 2010), thus the study populations share the same genetic origin. Today, both populations appear to be numerically stable with similar population sizes, at ca. 300 individuals (Anonymous, 2019a).
Population size was estimated by visual counts, performed by hunter during field observations twice a year (in spring and in autumn).
Chamois can be found only in the Mediterranean area and red deer in the continental one.  (Geist, 1966), whereas horn measurements were taken on both horns. The absence of directional asymmetry in Caprinae species is evident (von Hardenberg et al., 2004) and since most horns had some wear on the tip, we considered only the longest horn to avoid errors in data analysis.

| Data collection
We used a flexible ruler to measure base circumference and each annulus from the first (L1) to the tenth (L10), depending on the animal's age. Only complete segments were considered for analysis.
Measurements were taken only once by the same experienced game warden using a standardized methodology (Merchant, Hoefs, Nette, Kale, & Janssen, 1982). The hunting regime is the same in both areas, as individuals are hunted in accordance with pre-defined game management plans (J. Tomljanović, personal communication), and regulated by the Croatian hunting act (Anonymous, 2019b).

| Data analysis
We created two datasets to be used in the analyses. The first (full) dataset consisted of the horn length measures for all 514 individuals, used for the analysis of total horn length and base circumference, and to assess compensatory growth. The second (reduced) dataset included only individuals from cohorts present in both populations, that is, cohorts (year of birth) between 1987 and 1999 (n = 183 for the Mediterranean and n = 83 for the continental population) to test for individual heterogeneity within the same cohorts.
To test for the influence of the environmental quality, hereafter referred to as "habitat" (Mediterranean vs. continental), on the total horn length and base circumference (response variables), we fitted separate linear mixed models (LMMs) for both response variables.
The predictors included in both models were habitat (2-levels fixed nominal variable), individual age at death and cohort (factors with random intercepts fitted for each age and cohort classes).
To explore the potentially different patterns of compensatory growth between the two populations, we first selected from the full dataset individuals that had at least 5 years of age at the time of harvest (n = 169 for the Mediterranean and n = 44 for the continental population). We then fitted a linear model to regress the cumulative length of horn segments L3-L5 (response variable) against the length of the second horn segment (L2) corrected for cohort and age, with population fitted as categorical predictor. Since all models were conceptually developed prior to analyses, following the methodological framework used in previous studies (Alpine ibex: Toïgo et al., 2013;chamois: Corlatti et al., 2015;Kavčić et al., 2018), we did not use any model selection procedure.
Lastly, to investigate the importance of individual heterogeneity on log-transformed horn segment length variation (response variable), for each population, we fitted a linear model with cohort and age of the segment as fixed variables, while animal identity was fitted as a factor with random intercepts for each individual. Fixed factors accounted for the variation of horn segment lengths between different cohorts and variation between horn segment lengths between segments grown in different animal ages. The individual random factor was included to assess the correlation among repeated measures of horn annuli within the same individual, that is, the intraclass correlation coefficient (ICC). ICC provides information on the degree of similarity of the annuli length within individuals, relative to the difference in annuli length among individuals. In simpler words, ICC allows to measure repeatability that is the level of individual clustering in horn annuli length: the higher the ICC value, the greater the clustering (i.e., the higher the difference among individuals in terms of horn annuli), thus the more variance is explained by individual heterogeneity. To estimate this variance, we calculated the conditional and marginal R 2 statistics developed by Nakagawa and Schielzeth (2013) for each population model. The difference between the two statistics represents the variance explained by the individual random factor within each population. In addition, for each area we estimated the "adjusted" repeatability, that is, the repeatability (ICC) values obtained from the fitted models after accounting for fixed effects (Stoffel, Nakagawa, & Schielzeth, 2017 (Barton, 2013) for modeling and estimation of R 2 statistics.
Repeatability of models was estimated using the function rpt in the "rptR" package (Stoffel et al., 2017).

| RE SULTS
The effect of population, corrected for age and cohort, was significant for total horn length (F = 134.28; p < .001) and base circumference (F = 31.84; p < .001). Specifically, horns were longer in the  (Figure 2).
We did not find a significant effect of population and early horn growth (L2) on the sum of segment length from L3 to L5 (p > .05, Figure 3), thus no significant compensatory growth in either popula- tion. Yet, we found a weak signal for partial horn growth compensation in the continental area, suggesting that rams in this population may have invested less energy in growing large horns than did the Mediterranean ones.

| D ISCUSS I ON
Our data suggest that habitat characteristics may affect the patterns of sexually selected traits in European mouflon. A potential for environment-dependent pressure of sexual selection may be indicated by different investments in horn growth between populations. Mediterranean rams had longer horns, showed no compensatory horn growth, and had greater individual heterogeneity in horn segment length compared with the continental ones. As the strength of artificial selection due to hunting pressure appears similar in both areas, our results suggest that the effects of sexual selection may be stronger in more favorable environments, thereby supporting our predictions.
Despite the shared genetic origin, mouflon rams in the Mediterranean population apparently invested more energy in weapon length (i.e., horn length and annual segments) than rams in the Continental population, which seemingly traded shorter horn lengths for slightly larger base circumferences. These data are in agreement with Krapinec et al. (2013), who found thicker base circumferences in the continental population and longer horns in the Mediterranean population. These authors suggested that different forage availability was the most important factors shaping horn growth between two study sites. Indeed, lower quality and availability of food resources are known to cause decline in horn size in European mouflon (Garel et al., 2007). This may result from a tradeoff between investment in survival (natural selection) and reproduction (sexual selection). In fact, allocation of energy to secondary sexual traits in the Caprini tribe is under strong influence of habitat conditions and varies with resource availability (Faliu, Cugnasse, Auvray, Orliac, & Rech, 1990;Festa-Bianchet et al., 2004;Garel et al., 2007;von Hardenberg et al., 2004). The greater horn length found in the Mediterranean rams may thus support prediction (a) of greater pressure of sexual selection in this population than in the continental one, owing to relatively better environmental conditions.
It remains unclear why continental rams invested more in horn circumference than the Mediterranean rams, especially because male horn base typically correlates with male horn length in wild sheep (Pigeon, Festa-Bianchet, Coltman, & Pelletier, 2016).
The compensatory rates did not differ significantly between the two populations. Our data therefore support previous studies on wild sheep species, where compensatory mechanism was weak or absent (Festa-Bianchet, 2017). Interestingly, however, we found a and traits, such as body mass (Mason, Apollonio, Chirichella, Willis, & Stephens, 2014). In turn, this suggests a more conservative strategy of energy allocation in the continental population. In contrast, in the Mediterranean population some rams may afford to invest conspicuously in early horn growth, avoid compensatory growth, and because sexual selection normally favors large male size through rapid early growth (Andersson, 1994), they may possibly gain a reproductive advantage over smaller-horned males (Coltman et al., 2002).
As a consequence of seemingly different patterns of compensatory growth, we found greater individual heterogeneity in horn segment length was in the Mediterranean population than in the continental one. This suggests higher frequency of "extreme" phenotypes (e.g., "low-quality" and "high-quality" individuals) in the former than in the latter population. In line with prediction iii), given that the investment in horn growth is energetically demanding (Zahavi, 1975), the opportunity to express extreme sexually selected traits is expected to increase in habitats of higher quality.  (Mysterud, Trjanowski, & Panek, 2006), and thus contribute to explain our results. Even if the hunting regime adopted in two study areas was legally the same, we cannot exclude the possibility that the hunters might have had slightly different preferences toward larger or smaller rams. Nonetheless, preliminary data suggest that no significant difference in selectivity (i.e., age at culling vs. early horn growth, cf. Douhard et al., 2016) exists between the two populations. Furthermore, our suggestion that habitat quality may be the major driver of the observed differences is supported by previous studies showing that forage supply had a positive impact on horn volume, but only in the continental population (Krapinec et al., 2013). Additionally, the effect of founders' morphological traits might also play a role in shaping contrasting horn growth patterns in the study populations; no data are available to clarify whether this stochastic effect might have had an impact on our results.
Sexual selection can impose substantial fitness costs at the individual level, especially in low-quality habitats (Janicke et al., 2015; Martinossi-Allibert, Rueffler, Arnqvist, & Berger, 2019), with consequences for the dynamics of populations. We acknowledge that our study does not explicitly investigate the role of potential environmental drivers on the pattern of sexually selected traits. Future research should thus aim to elucidate the fine-scale mechanisms linking sexual traits and environmental constraints, thus the tradeoffs between reproduction and survival in mouflon living in different environmental conditions. This, in turn, should allow to better understand the consequences of such trade-offs for the dynamics and thus the management of populations.

F I G U R E 3
Relationship between log-transformed second horn segment length L2 and log-transformed cumulative length of horn segments L3-L5 in the Mediterranean and continental European mouflon populations between 1985 and 2012

ACK N OWLED G M ENTS
We want to express our appreciation to Josip Tomljanović for providing samples for our analyses and Hunting club "Jarebica-Senj." The article processing charge was funded by the German Research Foundation (DFG) in the funding programme DEAL.

CO N FLI C T O F I NTE R E S T S
We have no competing interests.

AUTH O R S ' CO NTR I B UTI O N
KK wrote all drafts of the manuscript. LC conceptualized the framework and revised all drafts of the manuscript. TS did the statistical analyses. NB obtained and arranged raw data. NS supervised all stages of this work, from data collection to data analysis, and participated in revising the manuscript.