We analyzed thin sections from right femora of 84 wild individuals from 3 species of cervids and 27 species of bovids (Table 1, see supplementary material for more information).

Some species include growth series, and others include only adult individuals (epiphyses fused). The sample of red deer (Cervus elaphus hispanicus) and roe deer (Capreolus capreolus) come from hunting, as well as from carcasses collected in the wild in Spain. Fresh carcasses of Svalbard reindeer (Rangifer tarandus platyrhynchus) were collected and sexed by researchers from Norwegian Polar Institute (NPI) during a field campaign in Svalbard archipelago (Norway) in July 2010. All skeletal remains of deer species are housed at the Institut Català de Paleontologia Miquel Crusafont (ICP). Bovid species were hunted from the 1950s until the late1970s by H. Oboussier, professor at the University of Hamburg (Germany), during several expeditions to Africa. Data (body mass, sex, site and date of death) of most of these individuals are associated (Köhler et al., 2008). The skeletal remains are currently housed at the scientific collections of the Zoological Museum Hamburg University.

Cross-sections of femora were taken at midshaft and thin sections were prepared following standard procedures of our laboratory described in García-Martínez et al. (2011), Köhler et al. (2012) and Marín-Moratalla et al. (2013). Thin sections were examined under transmitted and polarized light equipped with a digital camera (Leica DM 2500P). The software ImageJ was used to quantify and measure histological traits. Histological data were always quantified in the anterior region of the femora (Fig. 1).

The limb bones such as femur and tibia are commonly preferred in studies of bone development because they are often less subject to secondary remodelling in their internal cortices than shorter long bones. Moreover, these long bones are generally considered the best skeletal elements for skeletochronology in tetrapods because they preserve the most complete record of growth marks (García-Martínez et al., 2011, Horner et al., 1999 and Woodward et al., 2013). In this study, we conducted an analysis for the retrocalculation of missing GC or LAGs in order to explore whether a complete ontogenetic record of primary bone growth is preserved in the femora of ruminants. To carry out this analysis, we measured the antero-posterior diameter of LAG circumferences (APD-LAG) in the successive growth cycles of cortical sections (Marín-Moratalla et al., 2013). The growth curves derived from the sequences of APD-LAG were graphically overlapped among individuals of different ontogenetic stages within-species. In highly dimorphic species (e.g., C. elaphus hispanicus and R. tarandus platyrhynchus), each sex was analysed separately. We assume an annual periodicity of GC and that cortical cross-sections are similar-sized in similar-aged individuals. The analysis of retrocalculation of age was conducted in those species represented by at least two ontogenetic stages (juvenile and adult), determined by the state of fusion of epiphyses (10 species, Table 1 and supplementary material).

We used phylogenetic generalized least squares ‘PGLS’ analysis (pglmEstLambda in R package; CAIC library) to correlate the number of GC in the FLC-zonal bone of the species of our sample with their antero-posterior diameter at midshaft femur (within-species average). PGLS analysis computes the least squares regression taking into account the extent of phylogenetic nonindependence (Harvey and Pagel, 1991, Jordana et al., 2014 and Orme et al., 2009). We used a synthetic phylogenetic tree with the species of our sample based on a time-calibrated phylogenetic supertree that includes 197 extant and recently extinct species of the suborder Ruminantia (Hernández-Fernández and Vrba, 2005).

To analyse ontogenetic changes in histological features, we quantified vascular orientation, vascular density and cellular (osteocyte lacunae) density in successive growth cycles along the FLC-zonal bone, following the method described in Marín-Moratalla et al. (2014) based on Cubo et al. (2012). Cellular density was quantified as the number of osteocyte lacunae (in one focus plane) divided by the area (mm²) in a given surface of 40× micrographs. Vascular density was quantified as the number of vascular canals divided by the area (mm²) in a given surface of 5× micrographs. Vascular orientation was quantified as the proportion of one kind of vascular canal (longitudinal, circular, oblique and radial) with respect to the total number of canals in a given surface. This analysis was carried out on species composed of at least two males showing at least two GC within the FLC (7 species, see supplementary material). Females were excluded from this analysis due to their small sample size.

Fig. 2 shows the growth curves based on the sequence of APD-LAG among individuals of different ontogenetic stages (juveniles and adults). Two males of C. elaphus hispanicus (IPS 60870 and IPS 60875) and one male of Tragelaphus strepsiceros (IPS 56211) are the only ones of our sample that have probably lost their first GC (corresponding to the first year of age). Both species are the largest species, and with the longest growth period of our sample. The onset of deposition of the EFS occurs at 5–6 years of age in male red deer and at 6 years in male greater kudu (T. strepsiceros). Furthermore, the specimen IPS 60875 is the oldest of the red deer sample, with an estimated age of 10 years based on the number of GC in the whole cortical section including the EFS. However, the 4 years young deer ICP 60870 has also lost its first LAG in the anterior region of the femoral cross-section, though it is still visible in other regions. Therefore, there is some variability within-species. Expansion of the medullar cavity during bone growth and continuous remodeling of the inner cortex during life are responsible for the resorption of the first GC in these specimens. Our results therefore indicate that the largest cervids and bovids may lose their first LAG with increasing age.

We counted the number of GCs in the FLC-zonal bone until the deposition of the EFS in the adult sample (fused epiphyses), which is the same as the length of radial growth of the femoral shaft (Table 2 and Fig. 1). In each species and sex, individuals show equal numbers of GC, except for males of roe deer (C. capreolus) and red deer (C. elaphus hispanicus) in which the number of GC ranges between 3–4 and 5–6, respectively. Fig. 3 shows the correlation between the number of GC of the species of our sample and their antero-posterior diameter at midshaft femur (within-species average), for males and females separately. As expected, the number of GC increases allometrically which is evidence of the longer growth period of larger ruminants. Moreover, we see that the number of GC is influenced by environmental factors as well: species inhabiting closed environments (deer species), the blue duiker (Philatomba monticola) and most species of Tragelaphini, as well as those from energy-poor environments (the desert dweller Addax and the insular Svalbard reindeer) are above the regression line; while open habitat bovids fall below the regression line. Our results therefore indicate that species of closed habitats and of energy-poor environments show relatively delayed growth periods of the femoral cortex.

We have carried out a compilation of literature data on age of first reproduction of sampled species and we see that the number of GC within the FLC-zonal bone in adults is broadly consistent with data on age at reproductive maturity (Table 2). For instance, in our red deer sample, males show 5–6 GC and females 3 before deposition of the EFS, fitting the age at which red deer normally attains reproductive maturity: 5–7 years in males and 3–4 years in females (Clutton-Brock et al., 1982). It must be clarified, however, that for some species it is not easy to know if the data from literature relate to physiological or to reproductive maturity.

Quantitative changes in vascular orientation, vascular density and osteocyte lacunae density among successive GCs have been explored (Fig. 4). Circular canals are predominant in the FLC-zonal bone of our ruminant sample. However, we observe a general pattern of decreasing frequency of circular canals and an increasing number of longitudinal canals along the successive growth cycles from inner to outer cortex. This pattern is more evident in T. strepsiceros, Aepyceros melampus and Gazella dorcas, where the proportion is reversed in the latest growth cycles, becoming longitudinal canals more abundant than circular ones. Radial and oblique canals were less frequent in our sample. As regards vascular and osteocyte lacunae density, our results show that both features tend to decrease in the latest growth cycles. For instance, in our male red deer sample (Fig. 4F), vascular orientation changes during the third GC, cellular density drops significantly after the fourth GC and vascular density decreases in the fifth and sixth GCs. In general, these histological features tend to change significantly in the last growth cycles of the FLC-zonal bone in analysed species. Exceptionally, this trend does not occur in the Svalbard reindeer, in which the histological features remain comparatively constant (Fig. 4G).

There is controversy in the paleohistological literature about the biological meaning of the onset of deposition of the external fundamental system (EFS: Cormack, 1987), the avascular and highly-organized lamellar tissue present in the outermost cortex of the long bones of adults. The EFS is a residual periosteal growth that marks the end of the intense radial growth of the long bones shaft and that is generally linked to the skeletal asymptoty or maturity (epiphyseal fusion in mammals) (Chinsamy-Turan, 2005 and Lee et al., 2013). In the paleohistological literature, the EFS is also frequently related to sexual maturity in mammals (Klevezal, 1996, Köhler et al., 2012, Marín-Moratalla et al., 2013 and Straehl et al., 2013). However, other studies disagree with this relationship, since they maintain that sexual maturity precedes skeletal maturity in most vertebrates (Lee et al., 2013 and Martínez-Maza et al., 2014). Although we agree with the latter statement in general, we should make some considerations especially for mammals. First, skeletal maturity or skeletal asymptoty of an animal is difficult to assess from a single long bone. Epiphyseal fusion of long bones (a proxy for skeletal maturity) occurs at different rates in the different long bones in mammals (Zeder, 2006). Therefore, only the long bones that fuse their epiphyses later could be used to assess if the individual has already attained skeletal asymptoty. Second, it is not always clear when sexual maturity in the paleohistological literature refers to physiological or reproductive maturity, which are not always synchronized. For instance, the decoupling between the age at physiological and reproductive maturity is widely observed in ungulates (Mitchell and Maher, 2006). Males of ruminants may produce viable sperm at 6–24 months but they typically delay reproduction until they reach maximum body and horn or antler size (skeletal maturity), many years after the initial onset of fertility at puberty (Clutton-Brock et al., 1982, Festa-Bianchet et al., 2004, Geist, 1971, Stewart et al., 2000 and Vanpé et al., 2009).

The results of our histological study on a ruminant sample show that the number of GC within the FLC-zonal bone of adult femora (until the deposition of the EFS) is broadly consistent with data on average age at reproductive maturity obtained from literature (see Table 2). In a few cases, obvious discrepancies exist and different reasons could be behind these inconsistencies. First, it is not always clear whether data on age of sexual maturity from literature correspond to physiological or reproductive maturity. This could explain the difference of almost 2 years between our results on addax females and data on age at maturity from Kingdon (1997).

Second, inter-populational variability may explain some of the discrepancies as well. For instance, we counted a single GC in Nanger granti, while Estes (1991) reported an age at reproductive maturity of 3 years for this gazelle. N. granti typically dwells in arid to semi-arid habitats (Estes, 1991). In these ecosystems, water restriction affects reproductive patterns (Cain et al., 2006) and gazelles attain maturity according to the water availability. Under favourable conditions, gazelles may attain reproductive maturity earlier than when water is restricted (Baharav, 1983). The N. granti individual analysed in this study comes from Lolkisale Conservation Area (Tanzania), characterized by a tropical savannah climate where water is less limited than in the typical habitat of this species.

Last, bone remodelling is another important factor to take into account for understanding the discrepancies between our results from bone histology and data from literature on age at reproductive maturity. Our results show that certain large species (i.e. red deer males and the greater kudu male) have lost their first GC due to expansion of the medullar cavity and endosteal remodeling during ontogeny. This suggests that large ruminants might not preserve a complete ontogenetic record of primary bone growth, becoming necessary the use of different ontogenetic stages to determine the age of the specimen, as is common in other large non-mammalian vertebrates (Woodward et al., 2013). This could also be the case for the large antelope Kobus ellipsiprymnus (> 225 kg), in which we counted 3 GC in the FLC-zonal bone of two males; however, this species generally attains reproductive maturity at 5 years (Owen-Smith, 1993). Nevertheless, we could not conduct retrocalculation of age in this species because we only have adult specimens.

The age at reproductive maturity is often the life history trait most sensitive to ecological conditions (Clutton-Brock et al., 1982 and Gaillard et al., 2000). Specifically, predation is considered to exert strong selection on age at first reproduction (Stearns, 1992). Under conditions of high mortality rates (i.e. high predation risk), species invest in rapid growth and early maturation that entails higher probabilities of surviving to first reproduction (Festa-Bianchet et al., 2006, Gaillard et al., 2000 and Stearns, 1992). Conversely, in populations under low extrinsic mortality environments (low predation pressure), selection favours delayed age of reproductive maturity, allowing an investment in growth and survival (Ricklefs, 2007 and Stearns, 1992). Similarly, the length of radial growth of the femoral cortex (or the number of GC in the FLC-zonal bone) of our ruminant species also appears to show this dependence on ecological factors, specifically on predation. Our study shows that among our ruminant sample, deer species and some bovid species such as the blue duiker (P. monticola), addax (Addax nasomaculatus) and most of tragelaphine antelopes (Tragelaphus imberbis, T. scriptus, T. spekii, T. strepsiceros), display a relatively long growth period (relative to their size). These species, except for addax and the insular Svalbard reindeer, dwell in wooded and closed habitats, which are environments characterized by lower predation pressure than open habitats (Clutton-Brock et al., 1982, Köhler, 1993, Owen-Smith, 1993 and Wronski et al., 2009). Addax dwells in desert, habitat characterized by resource limitation and scarcity of predators (Mallon and Kingswood, 2001). Similar selective pressures also characterize the habitat of the arctic reindeer from Svalbard archipelago (Aanes et al., 2000). Therefore, our study suggests that the transition from fast-growing FLC bone to slow-growing lamellar bone (EFS) in the femoral cortex of ruminants records the fundamental life history trade-off between growth and reproduction (Stearns, 1992). Hence, this transition between main bone tissue types is a good proxy for reproductive maturity, one of the most important life history traits that determine the species’ fitness.

Changes in vascular orientation, vascular density and osteocyte lacunae density along the FLC-zonal bone of ruminants have also been explored in this study. Our results show an important ontogenetic variation of these histological features. Circular canals predominate during early bone development, but later the number of circular canals decreases while that of longitudinal ones increases. Moreover, vascular and cellular (osteocyte lacunae) density tend to decrease significantly during bone development. Several studies have revealed that these histological features offer insights into relative periosteal growth rates (Bromage et al., 2009, Castanet et al., 2000, Cubo et al., 2012, de Margerie et al., 2002, Montes et al., 2010 and Mullender et al., 1996). According to them, the histological changes that we quantified during the development of the FLC-zonal bone of our ruminant sample reflect the slowdown of growth as maturity approaches (Lee et al., 2013). Actually, the most significant variation in the histological features along the FLC-zonal bone occurs within the last growth cycles before deposition of the EFS, indicating that there is an important slowdown of growth shortly before reproductive maturity. This trend could be related with the onset of physiological maturity (Clutton-Brock et al., 1982, Georgiadis, 1985 and Vanpé et al., 2009). For instance, males of red deer (C. elaphus hispanicus) usually reach sperm maturity during their third year of age (Garde et al., 1998), coinciding with the age (third GC) when major changes in the histological features of our male deer appear (see Fig. 4F). Likewise, bushbucks (T. scriptus) reach physiological maturity at 1.5 years (Apio et al., 2007), and the significant changes in vascular orientation and density during their second year could be related to this physiological change (see Fig. 4A). Oddly, the arctic reindeer from Svalbard archipelago (R. tarandus platyrhynchus) does not show this important ontogenetic variation in histological features of the FLC-zonal bone (see Fig. 4G). This could be related to physiological adaptations to their extreme environment and would require further research.

The important ontogenetic variation in the histological features of FLC-zonal bone in ruminants has major implications for comparative studies on bone histology. Our results point out the need to standardize ontogeny for comparisons of bone tissues of different individuals or species, both quantitatively and qualitatively. Some studies quantify histological features in the innermost part of the primary bone cortex because this is the region formed during the highest growth rate phase (Cubo et al., 2012, Legendre et al., 2014 and Marín-Moratalla et al., 2014), thus maximizing growth rate differences between individuals or species. Our study shows that large ruminants might not preserve a complete ontogenetic record of primary bone growth. Hence, in these species we run the risk of comparing features of different chronology. We further show that important changes in histological features occur later in bone development, during the last growth cycles before deposition of the EFS. Therefore, in ruminants with a long growth period, i.e. 4 or more annual cycles of radial growth in the femoral cortex such as red deer, there are no important differences in histological features between the first and the second year. Other studies, however, quantify histological traits in the outer third of the bone cortex before the EFS (Stein and Werner, 2013). However, our results indicate that in ruminants this zone is formed after a significant drop in the rate of periosteal growth; therefore, it represents an ontogenetic period of minimum growth rate, minimizing differences between groups.

Our study suggests that the transition between the fast-growing FLC tissue and the slow-growing EFS in femoral cortical bones of ruminants records the fundamental life history trade-off between growth and reproduction. Accordingly, the number of growth cycles in the FLC-zonal bone of femurs of ruminants is a good proxy for the onset of reproductive maturity. Large ruminants such as red deer, however, might not preserve a complete ontogenetic record of primary bone growth, making the use of ontogenetic series necessary for skeletochronological studies. Moreover, our study also provides evidence of an important ontogenetic variation in the FLC-zonal bone of ruminants. Circular canals that are predominant during the early juvenile stage decrease in frequency during bone development along with an increase in longitudinal canals. Additionally, vascular and cellular (osteocyte lacuna) densities tend to decrease during the development of FLC-zonal bone as well. Significant changes in these histological features tend to occur during the last growth cycles and could be related to the important slowdown in growth associated with physiological maturity. Finally, our study suggests that the innermost cortex should be the preferred zone for comparing histological features of primary bone tissue between individuals or groups, because this zone is formed during the highest postnatal growth rate, hence maximizing differences in this parameter between groups.