Phenotypic diversity is produced by changes in ontogenetic processes occurring at earlier developmental stages. Such changes may be brought about by allelic differences among individuals, differences in ontogenetic environment, or as is probably the case for the majority of phenotypic traits, both. Evolutionary biologists are accustomed to deducing ontogenetic properties based on ontogenetic outputs rather than by studying development as a process. For example, the degree of fluctuating asymmetry between paired adult structures has been assumed by many studies to be indicative of degree of developmental instability and integration [1-3], even though few studies made the effort to investigate the developmental underpinnings of asymmetry [4-6]. Comparative studies of scaling relationships between adult body parts continue to be used heavily to explore the ontogenetic basis of allometries and the evolution of shape [7-11], yet actual attempts to understand formation and growth of said parts during ontogeny are rare or absent. Similarly, considerable literature investigates the developmental costs of secondary sexual trait expression based primarily or solely on examination of adult individuals [12-15]. Here I explore the ontogenetic basis of male horn polyphenism in three closely related species of horn-dimorphic beetles and illustrate how relatively simple developmental studies can be sufficient to advance a deeper understanding of how ontogeny mediates phenotypic evolution and diversification.

In O. nigriventris, scaling relationships between body size and horn length changed significantly from the pupal to the adult stage (Fig. 2A). Specifically, both amplitude and steepness of the slope increased significantly from the pupal to the adult stage, creating a greater and more sudden disparity between minor and major male morphs (amplitude: T₆₄ = 2.76, p = 0.0075; slope: T₆₄ = 2.53, p = 0.0139). Both relative and absolute loss of horn length decreased quickly with male size (Fig. 2B, F_rel = 127; p < 0.0001; F_abs = 42; p < 0.0001). Small males commonly lost >1 mm (> 20%) of horn length compared to ~0.25 mm (< 5%) in large males. Log-log plots of pupal against adult horn length showed that hornless, minor males fell below the line expected if adult horn length was a direct, unaltered reflection of pupal horn length (indicated by gray line in Fig. 2C), whereas horned, major males appeared right on that line. This observation was backed up by regression analysis, which yielded a negative y-intercept significantly different from 0 (T₃₂ = 18.95, p < 0.0001), confirming that horn lengths of at least some males decreased from the pupal to the adult stage, and with a slope highly significantly greater than 1 (Fig. 2C; T₃₂ = 14.63; p < 0.0001), supporting that this decrease was strongest for small males.

Figure 2 Ontogenetic changes in allometric scaling between body size and thoracic horn length in male Onthophagus nigriventris. (A) Scaling relationship between body size (presented as standard deviations away from mean) and horn length in male pupae (●) (more ...)

In O. taurus, scaling relationships between body size and horn length did not change significantly from the pupal to the adult stage (Fig. 3A). Neither amplitude nor steepness of the slope increased significantly from the pupal to the adult stage. However, both relative and absolute loss of horn length still showed a significant effect of body size. As with O. nigriventris, relative horn loss decreased rapidly with male size (Fig. 3B, F_rel = 110; p < 0.0001), however, absolute horn loss first increased, reaching a peak in medium-sized males (x₀ = 4.83 ± 0.027 mm; p = 0.0001), before declining again to near zero values in large males (F_abs = 28.37; p < 0.0001). As with O. nigriventris, log-log plots of pupal against adult horn length yielded a negative y-intercept (T₅₁ = 21.98, p < 0.0001) and a slope highly significantly greater than 1 (Fig. 3C; T₅₁ = 14.96; p < 0.0001). Also as before, hornless, minor males fell below the line expected if adult horn length was a direct reflection of pupal horn length whereas horned, major males appeared right on that line.

Figure 3 Ontogenetic changes in allometric scaling between body size and head horn length in male Onthophagus taurus. (A) Scaling relationship between body size (presented as standard deviations away from mean) and horn length in male pupae (●) and corresponding (more ...)

In the third species, O. binodis, results were yet again different. The scaling relationships between body size and horn length changed significantly from the pupal to the adult stage, however, unlike the previous two species all adults beetles had much shorter horns then their corresponding pupae regardless of final body size (Fig. 4A). As a consequence, y-intercepts of linear regressions on pupal and adult allometries differed significantly (T₉₈ = 31.1, p < 0.0001). More importantly, allometric slopes also changed significantly, but in a direction opposite to that observed in the previous two species. Adult O. binodis had a slightly but significantly shallower slope than the pupae from which they had eclosed (T₉₈ = 2.64, p = 0.0098). Loss of horn tissue also showed a pattern different from that observed in O. nigriventris or O. taurus (Fig. 4B). Specifically, relative horn loss only showed a very moderate, if any, decrease with adult size (F_rel = 4.8; p = 0.033; ns after corrections for multiple comparisons). In contrast, absolute horn loss exhibited a moderate, yet significant, increase with adult size (F_abs = 11.25; p = 0.0016). Log-log plots of pupal against adult horn length yielded a negative y-intercept significantly different from 0 (T₄₉ = 4.38, p < 0.0001), further confirming that horn lengths changed from the pupal to the adult stage, but with a slope indistinguishable from 1 (Fig. 4C; T₄₉ = 1.026; p = ns), indicating that this change was similar for all males regardless of their size. Combined, these data suggest that pupal remodeling does affect male shape but not male dimorphism in this species. Most strikingly, however, this analysis failed to support the existence of a male dimorphism in horn expression in the first place, even though this species has been the subject of several previous studies into the biology of alternative male morphs [4,5,15]. To help resolve this contradiction I measured adults using the same landmarks used by previous studies, as illustrated in Figure 5 (J. Tomkins, pers. communication). Rather than using the scutellum as a posterior boundary, earlier studies used an imaginary line drawn perpendicular to the body axis at the level of two lateral concavities (μ in Fig. 5B), or depressions, that characterize the adult prothorax. No such concavities exist in pupae and this measurement can only be made on adults. Use of this alternative posterior landmark recovered a dimorphic scaling relationship very similar to those published previously (Fig. 5C). This suggested that male dimorphism in O. binodis may not be a function of horn growth per se, but rather a function of the degree to which lateral aspects of the pupal prothorax retract, or "cave in", around the medial prothorax. To further examine this possibility, I calculated the length of the part of the thorax, c, that was unaffected by lateral concavity formation for each individual (Fig. 5b). If lateral concavity formation is independent of body size, i.e. the same proportion of the prothorax participates in concavity formation regardless of final adult size, then c should increase isometrically with male size. In contrast, if large, major males have larger horns because the lateral prothorax is further caved-in than in small males, c should scale with body size with a slope < 1, and possibly even a negative slope. In support of the latter hypothesis, c was found to first increase and reach a peak in medium-sized males size before declining again rapidly in larger males (Fig. 5D; log normal regression: r² = 0.42, p < 0.0001). In contrast, a linear regression of c on body size failed to yield a significant fit (r² = 0.06; p = ns). Separating small and large males in the analysis further confirmed this observation. c was found to increase significantly with body size in males less than 6.4 mm in thorax width (F = 42.74; p < 0.0001), but to decrease significantly with body size in males larger than that (F = 12.72; p < 0.001). This supports the hypothesis that male dimorphism found in O. binodis is not the product of dimorphic growth of horn tissue but instead results from body size-dependent caving-in, or retracting, of the lateral prothorax, exposing a relatively larger thoracic projection in large males compared to their smaller male counterparts.

Figure 4 Ontogenetic changes in allometric scaling between body size and thoracic horn length in male Onthophagus binodis. (A) Scaling relationship between body size (presented as standard deviations away from mean) and horn length in male pupae (●) and (more ...)

Figure 5 Dorsal view of a male O. binodis as (A) pupa and (B) corresponding adult. Note presence of pronounced lateral concavities, or depressions, in the prothorax of the adult but not pupae (marked by arrows on left side of the adult). a indicates pupal and (more ...)

Onthophagus beetles have emerged as a promising model system for exploring the interplay between genetic, developmental, and ecological factors in the genesis of phenotypic diversity [16,17,21,46]. Part of their attractiveness stems from the enormous morphological diversity in secondary sexual traits found between species, sexes, and alternative morphs within sexes, providing opportunities to explore, within a very narrow phylogenetic distance, topics such as the costs and limits of trait elaboration [8,23], developmental and life-history tradeoffs related to secondary sexual trait expression [10-12,43], or the frequency of independent evolutionary events necessary to explain extant patterns of diversity [13,14]. Importantly, the majority of earlier studies implicitly or explicitly assumed that the expression of secondary sexual traits in adult beetles is the consequence of differential activation of growth of these structures during immature development [8,10-12,16,37-40,42,43]. Here I present results on the developmental origin of alternative male dimorphism in three Onthophagus species that may help refine this assumption.

In two of the three species studied here, O. nigriventris and O. taurus, pupal remodeling of horn length occurred in a size dependent manner. Smaller and medium sized-males lost both relatively and absolutely greater amounts of horn tissue compared to their larger male counterparts, causing males to change pupal to adult proportions in a size dependent manner. In large males of both species, pupal proportions were largely maintained in the eclosing adults, whereas smaller and medium-sized males were more extensively remodeled. In O. nigriventris, size-dependent pupal remodeling of male horn length significantly altered the scaling relationship between horn length and body size, causing adult allometries to exhibit greater disparity between alternative male morphs and to transition from one morph to the other over a narrower range of body sizes compared to the pupae from which they eclosed. Allometric differences in O. taurus were in the same direction as in O. nigriventris but were not significant. Overall, however, loss of pupal horn tissue, and remodeling of pupal proportions prior to adult eclosion, were considerably less severe than what has been previously documented for female Onthophagus and the development of sexual dimorphisms, where in many cases fully horned pupae have been shown to molt into entirely hornless adults lacking any indication of the earlier existence of a horn [21,44,45]. These results suggest that pupal remodeling can play a moderate, but not dominant, role in the development of some male dimorphisms. Specifically, in species in which prepupal horn growth is already dimorphic, pupal remodeling appears to have the capacity to further exaggerate male dimorphism. Furthermore, given that the relative loss of horn tissue was highest in small to medium-sized males, pupal remodeling in these species also has the potential to further reduce the production of intermediate morphologies from the population of phenotypes generated by dimorphic growth alone.

Results in the third species, O. binodis, however, suggest an additional dimension to the role of pupal remodeling. Initial results suggested that pupal remodeling of horn length also occurs in this species, causing pupal proportions to change as pupae become adults, but without a strong dependence on adult size. Instead, pupal horn length changed similarly in all males regardless of final size. Most importantly, however, horn length-body size measurements failed to reveal an actual male dimorphism in both pupae and adults, even though this species has been studied in detail for its male dimorphism by previous studies [10,13,37,38,43], and visual inspection of specimens leaves the strong impression of a significant discontinuity in horn expression when comparing males of different body sizes. Repeat measurements of the same individuals using the same posterior landmark used in earlier studies recovers a male dimorphism similar to what has been previously documented [43]. This alternative, original landmark is formed by two lateral concavities, or depressions of the prothorax, and an imaginary line that can be drawn perpendicular to the body axis at the level of the posterior-most extent of these depressions (Fig. 5), whereas the landmark used initially in this study was formed by the posterior border of the horn-bearing segment, the prothorax. The crucial difference between these two measurements lies in the fact that if horn length is measured using the lateral concavities as a landmark, length of horns becomes a function of not only the extent to which the horn extends forward, but also of the degree to which the lateral depression indent the prothorax. Measuring horn length using the posterior edge of the prothorax as a boundary is independent of the presence or absence of lateral depressions. Subsequent analysis showed that large males indeed allocate much larger proportions of their prothorax to the formation of lateral concavities compared to males of smaller body sizes (Fig. 5D). The complete absence of these concavities in pupae suggests that male dimorphism in O. binodis is not the product of dimorphic prepupal growth of horn tissue but instead may result entirely from body size-dependent caving-in, or retracting, of the lateral pupal prothorax, which ultimately exposes a much larger thoracic horn in large males compared to their smaller male counterparts. More generally, this suggests that dimorphisms, such as those found in O. binodis, do not necessarily require dimorphic growth of the trait of interest, but may be generated partly or entirely by dimorphic remodeling, or sculpting, of the surrounding body regions.

The data presented here illustrate that alternative male morphologies, at least among the three Onthophagus species studied here, are shaped by a minimum of two developmentally dissociated processes: growth of horn tissue during the prepupal stage, followed by remodeling during the pupal stage. In species in which prepupal horn growth is already dimorphic, pupal remodeling appears to have the capacity to further exaggerate male dimorphism and selectively remove intermediate phenotypes from the distribution of phenotypes generated by prepupal growth alone. In contrast, in species such as O. binodis male dimorphism may not require dimorphic prepupal growth and instead may be shaped entirely by dimorphic pupal remodeling. Most importantly, adult morphology alone appears to provide few clues, if any, as to the relative contributions of prepupal growth and pupal remodeling to the expression of alternative male morphs, thus underscoring the importance of developmental studies in efforts aimed at understanding the evolution of adult diversity patterns.

Figure 1 Species used in the present study. (A) Onthophagus nigriventris, (B) O. taurus, and (C) O. binodis. Shown for each species are large horned (major) males (top) and small, hornless (minor) males (bottom) as pupae (left) and corresponding adults (right). (more ...)

Laboratory colonies of all three species were derived from field populations. O. taurus was collected from pastures around Bloomington, IN, and O. nigriventris and O. binodis were collected from pastures near Waimea, Hawaii. All species were maintained and reared as described previously [44]. Early third instar larvae of each species were transferred from their natural brood ball into 12-well plates to monitor larval development as described previously [61]. First to second-day pupae were measured, weighed, and then returned to their artificial brood ball until adult eclosion. At this stage the pupal epidermis still fully lined the pupal cuticle (Moczek, unpublished) and pupal size measurements are therefore a direct reflection of growth completed prior to the pupal stage. Eclosing adults were retained in brood balls for an additional 3–4 days to allow the adult cuticle to fully harden, then weighed, killed, stored in ethanol, and re-measured as described below.

APM is responsible for planning and executing all work involved in this paper.

I thank B. Heather Bleakley, Harald Parzer, and two anonymous reviewers for helpful comments on earlier versions of this manuscript, and Joseph Tomkins for help in replicating earlier measurement approaches in O. binodis. C. Jacobson and J. Micantin expertly located and collected beetles in the field. Funding for this study was provided through NSF Grants IOS 0445661 and IOS 0718522 to the author.