The study is based on the analysis of bone tissue patterns and growth marks observed in mid-shaft femoral sections with polarized light microscopy.

The material analyzed consists of a total of 48 femora, 40 belonging to H. onicensis and 8 to the extant taxon E. quercinus. The H. onicensis femora come from the fossil site “Sa Pedrera de S’Ònix” on Mallorca island, which dates from the Late Pliocene. The femora of the extant taxa sample were also analyzed and discussed in García-Martínez et al. (2011). One of these, labeled therein as “ID 15”, had been misidentified as Glis glis Linnaeus, 1766. The data sample is displayed in supplementary material, Table S1, providing information on locality, stratigraphic age, anteroposterior diaphyseal diameter (mm), number of Lines of Arrested Growth and other observations. Epiphyseal fusion is not taken into account, as many of the H. onicensis femora are represented only by proximal or distal thirds. Furthermore, the epiphyseal fusion of the H. onicensis shows no correlation with age, because epiphyses remain unfused even in very old individuals.

The fossil material belongs to the collections of the Institut Català de Paleontologia Miquel Crusafont (ICP) (Spain), whereas the extant material comes from the collections of the Museu de Ciències Naturals de Granollers. For more information see García-Martínez et al. (2011). The thin sections of both species are stored at ICP with the acronym IPS.

Once the material was prepared and ready for sectioning, the femora were embedded in epoxy resin and slides were prepared following the protocol described in García-Martínez et al. (2011). The obtained slides were observed under circularly polarized transmitted light (Leica DM 2500 P). We considered growth marks as LAGs when they can be traced around the periphery of the bone (Horner et al., 1999). The descriptions of bone tissues are based on the typological classification established by de Ricqlès et al. (1991) and de Margerie et al. (2002). We counted double LAGs, frequent in both taxa, as representing only one year.

Abbreviations:

EFS: External Fundamental System.

FLC: Fibro-Lamellar Complex.

IFS: Internal Fundamental System.

LAG: Line of Arrested Growth.

LNV: Lamellar (or parallel-fibered) Non-Vascular.

LPO: Lamellar (or parallel-fibered) with Primary Osteons.

LSV: Lamellar (or parallel-fibered) Simple Vascular.

Of the H. onicensis sample, 32 slides were sufficiently well preserved for visualization of the tissue pattern and to observe the presence of growth marks. Of these femora, only five pertained to young individuals that had not yet attained reproductive maturity, as they did not develop an EFS. On the other hand, the great majority are adult individuals presenting two or three LAGs in the EFS corresponding to an age of two–three years. We also found older exemplars, with nine individuals providing an estimated age of at least five years. From these older individuals, two exhibit more than eight, and two more than 10 LAGs in the EFS. The eight slides of the extant continental taxon (E. quercinus) included three immature individuals; the rest belongs to specimens with lines of arrested growth in their EFS. Only two femora exhibit more than four LAGs. Skeletochronological results for the whole sample are shown in supplementary material, Table S1.

Juveniles of both H. onicensis and E. quercinus exhibit relatively fast-growing tissues. In both taxa, the medullary cavity is relatively expanded, so that the preserved cortex most likely represents postnatal growth only. The tissue consists of scarcely-vascularized or avascular parallel-fibered matrix (LSV-LNV) and of limited regions of woven bone (fibro-lamellar complex [FLC]). Tissue with primary osteons (LPO, FLC) extends predominantly from the internal part of the posterior region (insertion area of the third trochanter at a later stage) along the posterior cortex, and over a small and isolated area at the right lateral part of the anterior region (Fig. 1d and h). Frequently, a resorption (cement) line delimits these areas that follow the curved margins of the primary osteons. This line is not the first LAG, though both structures may coincide. After deposition of the resorption line, growth rate slows down (parallel-fibered matrix, mainly LSV) until deposition of the first LAG. Vascularization, where present, is mostly longitudinal; some radial vessels, however uncommon, are found in both taxa. An inner cortical layer, if present, is thin and unevenly distributed, and is generally located in the anterior region. It may be deposited at a very early juvenile stage (Fig. 1), but usually it appears at a late juvenile or early adult stage.

H. onicensis differs from E. quercinus in some aspects. The juvenile cortical wall is slightly broader than in E. quercinus as expected from allometry for a larger species. Primary osteons in the FLC of H. onicensis are slightly larger in diameter (Fig. 2). Unlike E. quercinus, several juveniles of H. onicensis show large resorption cavities in their FLC or LPO matrix in the place of former primary osteons (Fig. 3).

Adults of both taxa show at least one LAG. The almost entirely avascular lamellar cortex (LNV) of adults indicates a similarly slow growth rate for H. onicensis and E. quercinus. The area of the third trochanter begins to form at the medial-posterior edge approximately during the second year of life; usually, apposition in this area is accelerated over a short time (LPO-FLC). In some individuals, LAGs are irregularly spaced; however, regular apposition of double LAGs is not uncommon (Fig. 4). In both taxa, Haversian systems perforated the FLC and LPO cortex of older individuals (Fig. 1e). Contrary to E. quercinus, old H. onicensis individuals are frequent in our sample (supplementary material, Table S1) and they can double the age of old E. quercinus individuals. Related to extended longevity, old Hypnomys specimens show a very extensive and broad inner cortical layer (Fig. 1) and a higher degree of remodeling. While most “middle-aged” H. onicensis do not show any sign of resorption within their cortex but only the typical and steadily progressing resorption front at the medullary margin, young adults and very old individuals frequently show large resorption areas within the primary FLC (Fig. 3). This pattern is not found in E. quercinus.

There is a general agreement on the causes that trigger insular gigantism. The widely accepted hypothesis is that under decreased interspecific competition (low species diversity on islands results from low resource levels, McNab, 1994 and McNab, 2012), small mammals expand their dietary niche. The thus increased resource supply permits an accelerated growth leading to a larger body size, which is possible under absence of predation and favoured by natural selection under increased intraspecific competition (see McNab, 2010 and literature therein). Bone and dental histology are the most suitable tools to reconstruct life history traits in fossil vertebrates. Using bone histology, we test the hypothesis that insular gigantism results from an increase in growth rate triggered by increased resource abundance.

Several features suggest that the giant insular dormouse H. onicensis from Balearic Islands did not grow at the same rate but over a longer time period than its closest continental relative E. quercinus. Both species deposit the same tissue types at comparable ontogenetic stages. The earliest ontogenetic stage, the age until weaning, was not available, most likely because at that age juveniles do not leave their nests. After weaning, juveniles lay down a matrix of avascular or poorly vascularised parallel-fibered bone with simple vascular canals and, in small regionally limited spots, parallel-fibered bone with primary osteons or even woven bone. Late juveniles deposit poorly vascularised or avascular lamellar bone, and adults form an EFS of avascular lamellar bone with LAGs. Because bone tissue types record the rate of bone apposition (Amprino, 1947 and de Margerie et al., 2002), the similarities in bone histology between the giant insular dormouse and its small continental relative strongly suggest that they grew at similar rates. The functional implication of slight differences in the diameter and density of primary osteons between both species, however, are controversial. Both osteonal size and density have previously been used to calculate bone growth rate (de Margerie et al., 2002). The authors concluded that the bigger the cavities of primary osteons, the faster the growth rate, whereby they observed: “the osteon diameter is an indirect estimation of the osteonal fraction of the tissue, bigger osteons being closer to each other” (p. 226). Thus, both size and density of primary osteons reflect tissue growth rate (de Margerie et al., 2002). In the case of H. onicensis and E. quercinus, primary osteons in regionally limited cortical spots are slightly larger in the giant dormouse (Fig. 2). However, the primary osteons are less abundant and similar or even less closely packed at comparable age; the osteonal fraction, hence, is equal or smaller in the larger species (Fig. 2). This means that while the somewhat larger size of the primary osteons suggests a slightly faster growth rate, the osteonal density indicates a slightly slower growth rate. Considering that the study of de Margerie et al. (2002), based on an ontogenetic series of one sole species (Anas platyrhynchos), did not take into account interspecific allometry, we suggest that this feature might result from a scaling effect rather than from differences in growth rate. An allometric analysis on primary osteon size is needed.

Not only adults but also early juveniles of Hypnomys (tentatively aged as “postweaning before the first winter” because of their close-to-adult size and the lack of LAGs, Fig. 1d and h) are much larger than Eliomys of comparable age (see supplementary material, Table S1 for sample size). The size differences between both species at the earliest available ontogenetic stage, raises the question of whether a faster growth rate or a prolonged growth period (or both) accounts for the larger size of juvenile Hypnomys. As discussed above, both taxa deposit the same tissue type from the earliest preserved juvenile stage (innermost cortical wall of youngest individuals) until old adults. This excludes differences in growth rate for almost the entire growth period. Though we cannot rule out that a faster intrauterine growth rate accounts for a larger size of Hypnomys after birth, the supposition of a longer gestation period seems more appropriate given the allometric scaling relationship between gestation time and body size (Lindstedt and Calder, 1981).

A similar line of reasoning applies to the differences in thickness of the cortical wall between both species. Because the cortical wall of terrestrial vertebrates has to withstand strains and stresses generated by weight bearing, locomotion, and other activities, variance in thickness can result from differences in life style or may simply reflect body size allometry. Some authors suggested a fossorial life style for Hypnomys (Bover et al., 2010, but see Quintana and Moncunill-Solé, 2014 for a different interpretation). Nevertheless, the differences in thickness between the cortical walls of Hypnomys and Eliomys are only modest at comparable ages, which are most parsimoniously accounted for by simple body size allometry. Accordingly, the slightly larger juvenile cortical area in H. onicencis suggests a somewhat longer growth period (at a similar growth rate) until attainment of maturity as expected from the correlation between body size and age at maturity (Lindstedt and Calder, 1981).

The many LAGs found in the EFS of the oldest H. onicensis individuals that double those of E. quercinus incontrovertibly document an important increase in longevity. The attainment of fairly old ages in H. onicensis led to an increased incidence of osteoporotic resorption cavities in the cortical area of older individuals, especially in those older than 10 years (Fig. 3a). The relatively high frequency of large resorption cavities at very young ages might instead indicate nutritional problems, an explanation congruent with the notion of resource limitation as a principle characteristic of insular environments (McNab, 1994 and Palkovacs, 2003).

Life history traits correlate with body size, and one might argue that an increase in body mass must be associated with an increase in age at maturity and longevity. Testing whether the observed increase in longevity is predicted from allometry is problematic, because extant dormice taxa are all small, which makes it impossible to plot a regression line based on dormice that extends up to giant Hypnomys. Including in that plot other rodents such as murids does not make sense, since dormice have an exceptionally slow life history for their body mass, requiring comparison within the taxon Gliridae. However, the important increase in longevity from 5.5 years in 82 g Eliomys (Tacutu et al., 2013) to more than 10 years in closely related 200 g H. onicensis contrasts with the unchanged longevity (max. four years) between 20 g Mus musculus and 200 g Rattus rattus, (Tacutu et al., 2013). This strongly suggests that the doubling of longevity from E. quercinus to H. onicensis exceeds by far what is expected from size scaling. Thus, taking together the similarity in growth rate and the important increase in longevity, we conclude that gigantism in Hypnomys does not result from accelerated growth but from a deceleration of the pace of the life cycle, that is a delay in key life history features (age at maturity and longevity).

Finally, the frequent appearance of evenly spaced double LAGs in many individuals is intriguing (Fig. 4) and only exceptionally observed in other small mammals such as murids (personal observation MK). We speculate that this pattern is related to dormancy in dormice, reflecting two periods of dormancy, cold winter and hot dry summer. More research, particularly experiments with fluorescent markers, is needed to understand the formation of double-spaced LAGs in dormice.

Our histological study on long bone tissues of the giant insular dormouse H. onicensis in comparison with that of the continental closest relative E. quercinus provides evidence of an unchanged (in relation to the closest mainland relative) postnatal growth rate. This was likely associated with changes in time rather than in rate of foetal growth during gestation, leading to a larger size in early juveniles. Increased cortical thickness, in combination with a similar growth rate, suggests that Hypnomys attained sexual maturity somewhat later than Eliomys. Skeletochronology provides evidence of an extended maximum life span of more than 10 years, doubling the age of the oldest Eliomys individuals. These results do not lend support to the widely accepted hypothesis that insular giants evolved through faster growth rates in an environment with abundant resources.