Early constraints in sexual dimorphism: survival benefits of feminized phenotypes

Study species and general procedures

Adult common kestrels (hereafter kestrels) show marked SD in size and plumage coloration with males being smaller and more conspicuously coloured than females (Massemin et al., 2000). As in many dimorphic birds, fledglings are more similar to adult females than to adult males. Yet, within such a female appearance, kestrel fledglings also exhibit SD in size and coloration with males being smaller and showing greyer coloration in their rump and lesser blackish appearance than female fledglings. These phenotypic differences in size and coloration have served to traditionally assign sex to fledglings (e.g. Village et al., 1980; Dijkstra et al., 1990). Nonetheless, there are individuals that are difficult to sex due to their sexually intermediate phenotypes in both size and plumage coloration (Village et al., 1980). Plumage coloration in kestrels is melanin based (Fargallo et al., 2007a) and has been found to genetically covary with body mass in males but not in females, indicating sexual differences in phenotypic trait integration (Kim et al., 2013).

The study was performed during the breeding seasons of 2004–2012, in Segovia (central Spain), where most kestrels breed in nest boxes (Fargallo et al., 2009). Nests were frequently visited to determine laying date (the day the first egg was laid), clutch size (range 3–7), hatching date (of the first chick in each brood) and brood size at fledging (range 1–7). Morphological and chromatic variables for determining SD were recorded at the age of 26 days when fledglings were measured (tarsus length, wing length and body mass), plumage coloration was estimated (see below) and a blood sample was extracted for sex determination (see below). Also, on this day fledglings were individually marked with metallic rings in order to identify them in the following years as recruits into the population. To estimate local recruitment adults were captured using nest-box traps during the chick-rearing phase.

Fledgling coloration

Plumage coloration of fledglings was measured over 7 years (2004–2007 and 2009–2011) in 324 females and 337 males (n = 661). An experiment carried out in 2004 affected coloration (Fargallo et al., 2007a), for which reason only fledglings from the control group were considered in this study. coloration was assessed following previously published procedures (Fargallo et al., 2007a,b). Briefly, the greyness of rump and uppertail coverts (hereafter rump greyness; RG) was estimated as the percentage of grey with respect to brown covering the whole rump (0% = completely black-barred brown and 100% = completely black-barred grey, Fig. S2a). The repeatability of this measurement is high (Fargallo et al., 2007a,b). Digital photographs were taken from the back side of the bird while it was held with the wing and the tail outstretched (i.e. at an angle of 90° between the objective of the camera and the surface of the body) using a Nikon D70 camera (objective: 18–70 mm AF-S Nikkor DX). All photographs were taken under a sunshade in the morning. Images were imported into Photoshop CS3, where four colour trait variables were measured: (i) width of black terminal tail bands (TTB, Fig. S2b), (ii) width of black superior tail bars (TB, Fig. S2b), (iii) width of black rump bands (RB, Fig. S2b) and (iv) wing blackness (WB, Fig. S2c). All measurements were conducted by the same person (ILR). In a subset of 30 photographs belonging to 30 individuals, all colour variables were measured twice to determine repeatability. A metric scale was set for all pictures prior to measurements by equalizing pixel distance to measurement units (cm). All measurements were performed at real scale. To compute bar and band widths the ‘ruler tool’ was used. Measurements of the TTB were performed in the two central tail feathers, whereas TBs and RBs were measured in four different feathers on the left side of the body. Band and bar widths were calculated as the mean value of these feather measurements and were highly repeatable in all cases (TTB: R = 0.95, F_1,60 = 40.40, P < 0.0001; TB: R = 0.90, F_1,60 = 18.00, P < 0.0001; RB: R = 0.91, F_1,60 = 20.38, P < 0.0001). To estimate WB, the proportion of black coloration (spots and bars) was calculated in the covert feathers of a known area of the wing. Black spots and bars are easily distinguishable from the brown background of the wing plumage, and thus, they were outlined and their area determined as the number of pixels occupied using the ‘magic wand tool’ (Fig. S2c). WB measurements were highly repeatable (R = 0.97, F_1,60 = 58.53, P < 0.0001). As WB and RG are proportions, they were transformed using arcsine transformation.

Molecular sex determination

Fledglings were sexed using a polymerase chain reaction (PCR) amplification and agarose electrophoresis based on the technique used by Fridolfsson & Ellegren (1999). The method is based on the detection of a constant size difference between CHD1W and CHD1Z introns. DNA was isolated from the red blood cells using the Qiagen QIAamp kit and the manufacturer's protocol (Qiagen, Hilden, Germany). PCR amplification was run using a particular set of primers (2550F and 2718R) as proposed by Fridolfsson & Ellegren (1999).

Phenotypic sexual resemblance

Phenotypic sexual resemblance (PSR) was defined as the probability that an individual will be correctly sexed according to its phenotype. PSR was calculated first using a discriminant analysis in order to identify the best combination of phenotypic traits distinguishing sex. Variables used were both morphological (tarsus length, wing length and body mass) and chromatic (RG, WB, TTB, TB and RB). After that, posterior classification probabilities for each group was computed based on Mahalanobis distances of a given case (individual) to each group centroid, that is, the point that represents the means for all individuals for a group (sex in our case as determined by molecular sexing) in the multivariate space defined by the variables in the model. In general, the nearer a case is from the centroid of its group and the further away from the other group centroids, the more likely it is that it belongs to that group. Therefore, the higher the value of PSR, the more likely it is that the fledgling was correctly sexed according to these phenotypic variables.

Local recruitment

As a proxy of survival we used post-fledgling survival, which is calculated as the binary probability of recruitment and has been previously used for the study of common kestrels (Kim et al., 2013; Terraube et al., 2015). We also recorded age at recruitment defined as the age at first-time initiation of reproduction. Recruits were monitored during the 3 years following the birth of chicks, as almost all kestrels in our population recruit between the ages of 1 and 2 years, but some individuals recruit in their third year after fledging. For this reason, local recruitment was based on the ring codes of breeding kestrels captured from 2005 to 2012 of the individuals born between 2004 and 2009 (excluding 2008).

Statistical procedures

SD and phenotypic integration of sexually dimorphic traits

The degree of SD varied quantitatively among different characters of coloration and size (Table S2). Both plumage pattern coloration (PC1) and the width of the black TTB (PC2) differed significantly between sexes (Table 1). Overall, females presented blacker coloration in wings, wider bars in the rump and tail and less greyness in rumps than males (higher values of PC1), while males presented wider TTBs (higher values of PC2) than females. Furthermore, PC1 was significantly explained by tarsus length (Table 1). Individuals with longer tarsi also showed higher values of PC1, that is, a more female-like phenotypic coloration (Table 1). PC1 was not significantly explained by wing length and body mass and no significant interactions with sex were observed (all P > 0.2). PC2 was significantly dependent on wing length and tarsus length (Table 1). Individuals having longer tarsi also showed wider TTBs, with this effect being more pronounced in males as concluded from the significant interaction between sex and tarsus length (Table 1). PC2 was not significantly explained by body mass or the interactions sex × body mass and sex × wing length (all P > 0.2).

Phenotypic divergence between sexes: PSR

Except for tarsus length (P = 0.36), the remaining variables (wing length, body mass and the five colour traits) contributed significantly to discriminating the sexes (Table 2). The discriminant function correctly classified 89% of cases, being correct in 90.43% of females and 86.65% of males (Fig. 1). This difference was statistically significant (GLMM, F_1,596 = 6.4, P = 0.012, estimate ± SE = 0.6514 ± 0.26): males were more likely to be incorrectly classified than females. Similarly, PSR also differed between sexes, with females having a higher PSR (mean ± SD, females = 0.87 ± 0.21, males = 0.83 ±0.25; GLMM, F_1,596 = 5.1, P = 0.024, estimate ± SE= 0.3271 ± 0.11).

PSR and local recruitment

Recruitment probability was higher in males than in females (males = 12.1% and females = 9.5%; Table 3). The relationship between probability of recruitment and PSR differed significantly between sexes as concluded by the significant PSR × sex interaction (Table 3). More female-like females (higher PSR) showed higher probabilities of recruitment than less female-like females (Table 3, Fig. 2). In addition, recruited males tended to be more female-like (lower PSR) than nonrecruited males (Table 3, Fig. 2). When working only with recruited individuals the model showed that the age of recruitment was higher in males than in females (Table 3). More feminized individuals in both sexes tended to recruit at older ages, as concluded by the significant interaction observed between PSR and sex (Table 3, Fig. 3).

Early SD from an integrated perspective

At 26 days old, 89% of fledglings express their sex through an integration of at least seven traits including body size and plumage coloration. In both sexes, individuals with longer tarsi showed a more female chromatic appearance. In the case of the TTB, larger individuals presented wider bands. Such phenotypic integration may have evolved as a result of natural selection favouring certain combinations of different quantitative traits rather than a single trait (Pigliucci & Preston, 2004). Similar results have been shown in several taxa, including common kestrels, in which genetic covariation has been identified (Armitage & Siva-Jothy, 2005; Kim et al., 2013; Kim & Velando, 2015), thus reinforcing the evidence that melanin-based coloration covaries with other phenotypic traits such as size, body mass, morphology, physiology and behaviour. Our results also evidence that morphological traits can be differently integrated with colour traits within sexes as it has been found in this and other bird species (Kim et al., 2013; Fargallo et al., 2014). In general, larger, but not heavier, individuals showed more feminized chromatic phenotypes, indicating a correlation between chromatism and size in which individuals of intermediate sexual chromatism (11%) were also intermediate in body measurements. Interestingly, these intermediate sexual phenotypes were more frequently observed in males than in females.

Fitness benefits associated with phenotypic sexual resemblance

The main finding of our study is that individual survival is related to the extent to which an individual phenotype resembles the sex to which it belongs (PSR). In common kestrels it has been reported that most individuals do not exceed 50 km of dispersal distance (Terraube et al., 2015), therefore, leading local recruitment as a good estimator of individual survival. Our results show that although more feminized females achieved higher survival, the higher values in males were exhibited by less masculinized individuals, suggesting a selection on juvenile males towards more feminized phenotypes. In agreement with this, male fledglings showed lower PSR values than females and a higher proportion of incorrect sex classifications based on phenotype. Moreover, age of recruitment was also influenced by PSR in both sexes with more feminized phenotypes recruiting at older ages. A longer period of time and more kestrel cohorts are required to complete data on the individual lifetime reproductive success of our population. In other species, however, recruitment at early ages can influence individual fitness positively (Charmantier et al., 2006; Dugdale et al., 2011; Kim et al., 2011; Zhang et al., 2015), negatively (Auld & Charmantier, 2011; Kim et al., 2011) or have no effect (Froy et al., 2013; Pardo et al., 2013). Such data in our population would give us information on whether different reproductive tactics could be phenotypically identified during the fledging period and how adult SD is derived from juvenile phenotypes.

Two classical hypotheses highlight the fitness benefits associated with a female-like appearance, although only one has demonstrated this phenotype to affect fitness (Berggren et al., 2004). The first posits that more cryptic female-like phenotypes increase survival by reducing predation risk (Selander, 1965), which may be especially significant for juveniles during their first winter (Rohwer & Butcher, 1988). The second states that when male–male competition involves aggressive interactions, female-like individuals avoid agonistic behaviour of males and gain access to resources during winter, thus increasing the probability of survival (Berggren et al., 2004). These two hypotheses are not mutually exclusive and they may both explain our results. Although the winter biology of common kestrels has been scarcely investigated, they are known to be territorial and aggressive against conspecific intruders during this period (Wiklund & Village, 1992). Information on their sexual hierarchy of dominance when competing for hunting territories is also lacking. In the closely related American kestrel Falco sparverius, it has been recorded that adult females (which are also the larger and more cryptic sex) obtain better hunting territories than males and maintain them over the winter, indicating a higher capacity than males in scramble competition (Ardia & Bildstein, 1997). Also, in the common kestrel, female nestmates show a capacity superior to males when fighting for food (Fargallo et al., 2003). We posit that if female phenotypes signal higher competitive faculties, juveniles with female-like phenotypes may benefit when occupying hunting territories during winter, since diminishing costly interactions via signalling should be adaptive (Rohwer, 1982).

Is early SD adaptive in kestrels?

Since in both sexes more feminized phenotypes achieved higher survival, why does early dimorphism exist in kestrels? Sexual dichromatism in juveniles has been scarcely studied and its adaptive significance is unknown. It has been suggested to play a signalling role in offspring–parent interactions to gain parental investment (Johnsen et al., 2003), or within juvenile flocks to mediate encounters between unfamiliar individuals (Tringali & Bowman, 2012). Although appealing, neither of these mechanisms may be at work since both predict individual variation in the expression of coloration, but not sexual dichromatism. Moreover, there is no evidence supporting differential parental investment related to offspring sexual phenotype in common kestrels (Fargallo et al., 2003; Laaksonen et al., 2004; Vegara & Fargallo, 2008), and this species does not form juvenile flocks. In relation to size, smaller males are more efficient hunters and are preferred by kestrel females (Hakkarainen et al., 1996). However, before small size can play a favourable role in adult mate selection and foraging, males must compete for territories and survive as juveniles during winter. A female-like appearance could provide competitive benefits to males during this period. Therefore, males would face a trade-off between enhancing survival by exhibiting more feminized phenotypes (larger size) or enhancing reproductive success by producing less feminized phenotypes (smaller size).

Interestingly, some chromatic characteristics of the fledgling male phenotype are similar to that expressed by adult females. For example, male fledglings showed wider TTBs than female fledglings and greyer coloration on the rump. These characters are associated with indicators of quality in fledglings and adult females (Vegara & Fargallo, 2008; Vergara et al., 2009). Furthermore, smaller males (more male-like fledglings) are those showing a higher chromatic similarity (greyer rumps) with adult females. Therefore, we propose that an adult female plumage appearance could provide benefits to smaller males within this hierarchical system of dominance. The question remains as to why juvenile females do not adopt this phenotypic strategy. The chromatic phenotype produced by adult females and juvenile males may entail costs. In common kestrels, the expression of two heritable characters, grey coloration and immune response, are negatively correlated (Kim et al., 2013). Selection on the immune response is stabilizing, with individuals showing intermediate immune responses selected, however in the case of rump coloration the selection was disruptive, favouring the greyest and brownest phenotypes, although the greyest individuals with the lowest immune responses were those with the highest survival rates (Kim et al., 2013). We therefore propose that larger and more competitive female fledglings do not incur the costs of producing expensive chromatic signals of quality (grey coloration), since size is in itself a signal that can be evaluated by individuals in contest situations (Maynard Smith & Parker, 1976). On the other hand, male fledglings may adopt two different strategies: being larger and browner (more fledgling female-like) or being smaller and greyer (more fledgling male-like). Although the benefits of the former strategy seem to be a trend for better winter survival, the benefits of the latter strategy remain to be investigated. We propose that males adopting the second strategy might obtain benefits during reproduction due to female preference for smaller males with higher hunting abilities.

In summary, our study raises the possibility that outside of sexual selection, selective pressures acting in youth may constrain the sexual divergence of adult traits due to a trend to increase fitness through more feminized phenotypes at juvenile stages. We propose that variation in adult SD may be influenced by differential phenotypic strategies adopted in early life. This is therefore an essential perspective from which to focus investigations on SD that highlights the importance of the ontogenetic studies on size and chromatism. Moreover, winter behaviour in the acquisition of resources including territories and the change in adult SD over time at the population level are two main questions that should be investigated to determine the true influence of juvenile phenotypes on the evolution of SD.