INTRODUCTION

The vertebrate skeleton provides a rigid structure that supports and enables movement of the body. Like all biological structures, limb anatomy results from the conjoined effects of phylogenetic, structural and functional constraints (e.g., Gould 2002; Cubo 2004). Bone microanatomy (i.e., the organization of the bony tissues) provides crucial information on the constraints the animals are facing. The internal structure of bones, including their density, cortical thickness, and arrangement of bone tissues, allow for inferences regarding the stresses that are placed on the bones. When subjected to mechanical loading, bones undergo structural adaptations to better support the increased stress placed upon them (Ruff & Hayes 1983; Turner 1998; Ruimerman 2005; Habib & Ruff 2008; Nikander et al. 2010; Doube et al. 2011; Bishop et al. 2018). In the case of giant terrestrial animals, the greatest challenge of the skeleton is resistance to gravity: gravity exerts a downward force proportional to an animal’s body mass (Biewener 1989; Bertram & Biewener 1990; Biewener & Patek 2018). As body length doubles isometrically, body mass increases by a factor of 8 (Schmidt-Nielsen 1984), meaning that larger animals contend with relatively greater gravitational constraints compared to smaller ones. Giant quadrupeds are thus particularly challenged: some of them exhibit distinctive musculoskeletal and physiological adaptations allowing them to accommodate their massive weight, and are generally termed as graviportal (e.g. elephants, rhinoceroses, hippopotamuses, sauropodomorph dinosaurs). This term was originally proposed by Gregory (1912) and Osborn (1929), and refers to an ensemble of morphological and locomotor traits that facilitate weight-bearing. These traits include columnar limbs, relatively lengthened stylopods and shortened autopods, robust long bones (i.e., wider diaphyses for a given length), and enlarged feet with thick adipose cushions (Gregory 1912; Osborn 1929; Hildebrand 1974; Coombs 1978; Alexander & Pond 1992; Langman et al. 1995; Hutchinson et al. 2003). These features are typically associated with a high body mass (over several hundred kilograms) and limited locomotor capabilities, such as an inability to gallop. However, the concept of graviportality remains debated (Coombs 1978; Carrano 1999; Mallet et al. 2019). While extant elephants are the perfect example of graviportality, fulfilling all anatomical and functional criteria, other heavy terrestrial mammals (e.g., rhinoceroses and hippos) only partially conform to the definition. Rhinoceroses, although massive and morphologically adapted to weight-bearing, retain the ability to gallop, leading early authors to consider them mediportal (Gregory 1912; Osborn 1936), although they have since been reclassified as graviportal by others (Prothero & Sereno 1982; Eisenmann & Guérin 1984). Similarly, hippos have been variably categorized due to their capacity to trot and their semi-aquatic adaptations (Ross 1984; Alexander & Pond 1992; Carrano 1999; Stilson et al. 2016; Hutchinson & Pringle 2024). Beyond external morphology, graviportal animals also show microanatomical specializations in the limb bones (Gregory 1912; Osborn 1936; Coombs 1978; Mallet et al. 2019, 2020; Lefebvre et al. 2022; Bader et al. 2025, in press). These include the reduction or absence of a medullary cavity and increased compactness of bone tissue, which improve resistance to high compressive forces and enhance load distribution (Wall 1983; Houssaye et al. 2016, 2018; Nganvongpanit et al. 2017; Lefebvre et al. 2023; Bader et al. in press). Increased bone density and thickened cortices also improve resistance to bending and torsional loads (Currey & Alexander 1985; Oxnard 1990, 1993; Houssaye et al. 2016; Canoville et al. 2022): extant elephant display a distinctive cortical bone distribution, forming an “hourglass” or dual-cone pattern centered around the growth center (GC). In the absence of secondary bone deposition in the medullary cavity, this thickening occurs primarily where compact bone is initially laid down. The thicker cortex enhances resistance to compressive loads and likely corresponds to zones of peak stress during loading, helping maintain mechanical safety. Similar patterns have been described in extant rhinoceroses (Etienne 2023) and hippos (Houssaye et al. 2021), where reduced bone resorption near the GC preserves cortical thickness under high stresses, as indicated by trabecular anisotropy (i.e. trabeculae oriented preferentially along an axis). While some large sauropods also show cortical thickening near the GC, it is less pronounced, implying alternative weight-bearing adaptations (Lefebvre et al. 2023). Finally, the limb long bones of graviportal taxa show highly anisotropic trabeculae, oriented orthogonally to the ground (Wall 1983, Lefebvre et al. 2023; Etienne 2023) in columnar taxa such as sauropod dinosaurs and graviportal proboscideans, the trabeculae are thus oriented in a mostly parallel manner to the long axis of the limb bones, maximizing the axial distribution of the load (Lefebvre et al. 2023, Bader et al. 2025).

The order Proboscidea ‌Illiger, 1811 includes some of the largest terrestrial animals to have ever lived (Sukumar 1992; Larramendi 2016). Some species reached gigantic sizes, exceeding 5 meters at the shoulder and weighing up to 22 tons (Meshram & Sonakia 2006; Larramendi 2016). Today, the group is represented by only three species of elephants (Wilson et al. 2011), which remain the heaviest living land mammals. However, early proboscideans were far smaller (Sukumar 1992; Shoshani & Tassy 2005). The order comprises over 180 extinct species (Kingdon & Happold 2013) that once inhabited every continent except Oceania and Antarctica (Cantalapiedra et al. 2021). The earliest forms, such as Eritherium ‌Gheerbrant, 2009 and Phosphatherium ‌Gheerbrant, Sudre & Cappetta, 1996, from the Paleocene of Northern Africa, weighed only 6 to 17 kg (Gheerbrant et al. 2005; Gheerbrant & Tassy 2009; Larramendi 2016). Larger taxa emerged during the Eocene and Oligocene (e.g., Barytherium ‌Andrews, 1901, Palaeomastodon ‌Andrews, 1901, Chilgatherium ‌Sanders, Kappelman & Rasmussen, 2004) weighing over 1500 kg (Larramendi 2016). From the Miocene onward, proboscideans underwent a marked size increase across all families (Tassy 1990; Cantalapiedra et al. 2021), with many species exceeding 10 tons (e.g. Deinotherium ‌giganteum ‌(Kaup, 1829), Mammuthus ‌trogontherii ‌(Pohlig, 1885)) and some approaching 20 tons, including Palaeoloxodon ‌antiquus ‌(Falconer & Cautley, 1847) and “Mammut” borsoni ‌(Hays, 1834). Extant elephants (like other graviportal taxa) exhibit distinct microanatomical adaptations in their limb long bones to support their massive body weight (Wall 1983; Houssaye et al. 2016; Nganvongpanit et al. 2017; Etienne 2023; Lefebvre et al. 2023; Bader et al. 2024). In proboscideans, gigantism was facilitated in part by the acquisition of the columnar stance, where the limbs are straight and nearly perpendicular to the ground at rest. This alignment allows an increased reliance on axial compression, enabling the support of several tons of body mass without a proportional increase in bone robustness (Hildebrand 1974). As a result of the limb orientation, extant elephants have been shown to exhibit highly anisotropic trabeculae, primarily oriented orthogonally to the ground, optimizing load transmission along the columnar limbs (Bader et al. 2025). While the microanatomical adaptations of limb bones have been documented in modern elephants (Nganvongpanit et al. 2017; Bader et al. 2025) and, to some extent, in several extinct dwarf elephantids (Köhler et al. 2021; Bader et al. in press), they remain largely unstudied across the rest of the proboscidean order. In particular, the microanatomical variations associated with the acquisition of a columnar posture are still poorly understood. Among proboscideans, the Deinotheriidae ‌Bonaparte, 1845 family was the first to display columnar limbs, and while the morphological variation of their long bones has been described (Bader et al. 2024), their microanatomy has yet to be explored.

The scarcity of proboscidean fossil material often limits their inclusion in studies requiring destructive analysis. Furthermore, practical constraints, such as the sheer size of the bones, often prevent the use of micro-CT scanning as a non-invasive method for examining their microstructure. While the bones might fit in conventional medical CT, the resolution of such devices is not high enough to provide detailed visualizations of the microanatomy. The Hungarian Natural History Museum (HNHM) collection in Budapest (Hungary) contains a unique Deinotherium ‌giganteum ‌(Kaup, 1829) femur that was previously split in two. This resulted in a longitudinally segmented bone with minimal alteration to its internal structure, providing rare and valuable insights into the microanatomy of the giant Deinotheriidae. The Deinotheriidae originated in the Oligocene of Northern Africa, and later spread across Europe and Asia during the Miocene. This family includes three genera (Chilgatherium, Prodeinotherium Éhik, 1930, Deinotherium) with species ranging from 4 to 12 tons, and is the first proboscidean family to display a columnar stance. This postural adaptation is considered as a synapomorphy of these two apical clades (deinotheres and elephantiforms), although deinotheres display distinct morphological variations in their limb bones as compared to more derived proboscideans. Notable differences, such as a larger greater tubercle of the humerus and a reduced anconeal process of the ulna, suggest potential differences in posture and weight distribution (Bader et al. 2024). These variations raise the question of whether Deinotheriidae share the same microanatomical adaptations for heavy weight-bearing as modern proboscideans and whether the acquisition of a columnar posture is reflected similarly across phylogenetically distant proboscidean taxa. The specimen studied here will thus provide a unique opportunity to further our understanding of limb anatomy evolution in proboscideans in relation to body mass. In addition to the Deinotherium femur, the collection includes another damaged proboscidean femur (probably from the genus Mammuthus ‌Brookes, 1828), providing another rare opportunity to observe the inner bone structure of a fossil proboscidean species. Since Mammuthus is more closely related to extant elephants than Deinotherium, we expect its microstructure to show greater similarity to other elephantids. This comparison will allow us to assess whether phylogenetically closer species exhibit greater microanatomical resemblance or if weight-bearing demands override phylogeny, leading to similar structural adaptations in both deinotheres and elephantids.

In this study, we aim to: 1) provide the first description of the microanatomy of a Deinotheriumgiganteum femur, taking trabecular density and orientation as well as variations in cortical thickness into account; 2) compare our observations to the microanatomy of a Mammuthus sp. femur; and 3) contextualize these findings within the broader framework of weight-bearing adaptations in proboscideans (limb posture, bone external morphology).

MATERIAL AND METHODS

Sample

Our sample is composed of two partial bones from the Palaeovertebrate Collection of the Hungarian Natural History Museum (HNHM) in Budapest, Hungary. Due to historical reasons, very little written information is now available regarding these specimens, so that only one of the two specimens could be identified at the species level.

Imaging

Digital photographs of the bones were taken using a Canon EOS 600D camera. Camera settings, including focal length, exposure, and lighting conditions, were standardized across all images to ensure consistency. Pictures were modified in Adobe Illustrator CC (23.0) to highlight the different tissues in the bones, considering the parts that were on a similar level (i.e. not broken at different depths), hence the Mammuthus femur being only highlighted in the cranial part of the diaphysis. Zoomed-in images were processed using the ‘Unsharp Mask’ filter in ImageJ (version 1.52p) to enhance sharpness (radius: 5 px, mask weight: 0.60).

Linear measurements

To highlight the differences between the two taxa and provide a quantitative reference for future studies, we measured the absolute maximal thickness of the diaphyseal cortex (AmaxT) and the corresponding diaphyseal diameter at that point (D). Using these values, we calculated the relative maximal thickness of the diaphyseal cortex (RmaxT), defined as the ratio of AmaxT to D.

RESULTS

Deinotheriumgiganteum femur

In D. ‌giganteum, the compact bone is thickest at the mid-diaphysis, with a greater thickness on the caudal side compared to the cranial side (RmaxT = 0.36, see Table 1). This compact layer tapers toward the metaphysis and epiphysis (Fig. 2A, B). The compact bone is extremely thin on the cranioproximal side of the femoral trochlea, slightly thicker in its distal portion, and thickest within the intercondylar fossa.

The medullary area is occupied by trabecular bone from the distal epiphysis up to just below the mid-diaphysis. The midshaft remains largely devoid of trabecular bone, though some breakage may have occurred. However, the absence of trabecular bone in this region appears to be natural rather than taphonomic.

The trabeculae are thin and numerous, with a higher density in the distal epiphysis compared to the shaft. In the mid-diaphysis, the trabeculae exhibit high anisotropy, oriented proximodistally and running parallel to the compact bone (Fig. 3a, e). In the distal diaphysis, the outermost layer of trabecular bone consists of highly anisotropic but thinner trabeculae, whereas the innermost region contains thicker, sparser, and mostly isotropic trabeculae (Fig. 3b). Just above the distal metaphysis, the trabeculae are thin and highly anisotropic, oriented from the cranial side to the distocaudal side. This orientation aligns orthogonally to the ground when the femur is at rest, reflecting the slight curvature of the bone in a normal standing position (Fig. 3f). The innermost and caudal regions of the distal diaphysis contain predominantly isotropic trabeculae, while the outermost parts exhibit anisotropic trabeculae oriented orthogonally to the compact bone layer (Fig. 3c). In the cranioproximal portion of the trochlea, the trabeculae are anisotropic and aligned parallel to the articular surface, whereas in the distal trochlea, they are oriented orthogonally to the articular surface. Overall, trabecular anisotropy is more pronounced in the cranial part of the distal metaphysis and epiphysis than in the caudal part (Fig. 3c, d).

Mammuthus sp. femur

In Mammuthus sp., the visible portion of the femur reveals a thick compact bone layer at mid-diaphysis, which tapers distally toward the distal epiphysis but shows a similar thickness proximally (RmaxT = 0.11, see Table 1), extending toward the greater trochanter (Fig. 2C). The distal epiphysis is filled with trabecular bone, with fractured trabeculae in the distal metaphysis, suggesting that this area was originally also filled with trabecular bone.

A thick layer of trabecular bone, approximately one-quarter of the shaft’s diameter, is present along the outermost parts of the diaphysis, while the inner regions remain largely free of trabecular bone. In the proximal metaphysis, distal to the greater trochanter, the trabecular layer is notably thin but thickens again closer to the greater trochanter itself.

In the proximal metaphysis, the trabeculae are thick and arranged in large lattice-like structures, forming arching patterns from the cranioproximal side of the shaft to the distocaudal side (Fig. 4a). In the proximal and mid-diaphysis, the trabeculae remain thick and highly anisotropic, with those in the outermost regions oriented parallel to the compact bone. In the inner regions, many trabeculae appear broken, but the intact ones are oriented from proximocaudal to craniodistal, forming lattices of thick and widely spaced trabeculae (Fig. 4b, c). Just below the mid-diaphysis, distal to the growth center (GC), the trabeculae maintain a similar thickness but oriented in the opposite way, with the inner trabeculae arranged from proximocranial to distocaudal. The trabeculae in the outermost regions remain parallel to the compact bone (Fig. 4d, e). Above the distal metaphysis, the trabecular tissue is largely fragmented. However, the remaining trabeculae are much thinner and show a much lower anisotropy, suggesting that the original metaphysis was filled with isotropic or lowly anisotropic trabeculae oriented proximodistally (Fig. 4e, f). The fractured lateral portion of the trochlea reveals isotropic trabeculae in the inner part of the structure, while the outermost trabeculae are anisotropic and aligned parallel to the compact bone (Fig. 5). On the medial side, the broken region of the trochlea primarily consists of isotropic trabeculae, with a faint orthogonal orientation to the compact layer in the most proximal part (Fig. 5).

DISCUSSION

General pattern and weight-bearing adaptations

This study provides preliminary insights into the microanatomy of the femur of D. ‌giganteum and Mammuthus sp., suggesting that both taxa share similar microanatomical patterns adapted to massive weight-bearing. Both specimens exhibit femora with a relatively thick cortex, a medullary area filled with trabecular bone except in the mid-diaphysis, and highly anisotropic trabeculae oriented orthogonally to the ground. These characteristics are expected in graviportal animals (Houssaye et al. 2016; Lefebvre et al. 2023), including proboscideans (Bader et al. 2025, in press).

The femur of D. ‌giganteum exhibits a distinctive cortical bone distribution, characterized by thickening around the growth center, resulting in an hourglass-shaped pattern. While this feature is less prominent in Mammuthus, it remains discernible. Similar distributions have been observed in modern elephants (Nganvongpanit et al. 2017; Bader et al. 2025) and other large-bodied taxa, including extant rhinoceroses (Etienne, 2023) and hippopotamuses (Houssaye et al. 2021). In these species, the growth center is surrounded by highly anisotropic trabeculae, suggesting that this area is subjected to high stresses. Retaining a thick cortical layer around the growth center would then provide an additional adaptation to massive load support: instead of resorbing the bone, as is usual for non-graviportal species, the cortical bone around the growth center suggest a limited bone resorption resulting in a thick cortical layer capable of supporting massive loads. In both D. ‌giganteum and Mammuthus, the compact bone is much thicker in the diaphysis, where it forms the cortical bone, than in the epiphysis where it is limited to a thin layer. The trabecular tissue occupies a large portion of the diaphysis, leaving only the central part free, which corresponds to the growth center. This trabecular filling is typical of large quadrupeds (Houssaye et al. 2016), where an increased trabecular volume enhances load distribution along the bone, thereby improving weight-bearing capacity (Currey, 2002). This organizational pattern is consistently observed in other graviportal animals, including rhinoceroses, sauropod dinosaurs, and modern elephants (Etienne 2023; Lefebvre et al. 2023; Bader et al. 2025; Bader et al. in press).

Additionally, the trabeculae in the diaphysis are highly anisotropic, oriented parallel to the compact bone (i.e., proximodistally, orthogonally to the ground), which optimizes axial load distribution (Currey 2002, Barak et al. 2008). In modern elephants, this trabecular arrangement reflects the immense mechanical demands and the load application angle associated with a columnar stance, where the limb’s long axis remains nearly perpendicular to the ground during standing and slow walking (Ren et al. 2010). Given the columnar posture of both Deinotherium and Mammuthus, the trabecular orientation in their femora similarly reflects limb orientation and weight-bearing adaptations. In the distal epiphysis, the trabeculae are less anisotropic than in the diaphysis, exhibiting a predominantly isotropic arrangement in the innermost part of the bone, and slightly increased anisotropy in the outermost part. This pattern reflects the diverse mechanical constraints to which the knee is subjected: the distal femoral epiphysis is part of the knee joint articulation, which imposes complex strain patterns beyond simple proximodistal loading. As a result, the trabeculae do not exhibit a strong preferential orientation. In the distal part of the trochlea, the trabeculae are orthogonal to the compact layer, facilitating optimal load distribution. Conversely, in the proximal part of the trochlea, which is not in direct contact with the tibia, the trabeculae are parallel to the compact layer, reflecting the different orientation of the mechanical strain.

Intergeneric variation

While the femora of D. ‌giganteum and Mammuthus sp. share an overall similar organization, we observed clear differences between the two species.

The hourglass-shaped distribution is clearly visible in D. ‌giganteum but only slightly visible in Mammuthus sp. This variation may be linked to a number of factors: the two taxa might exhibit differences in posture and/or weight distribution. Conversely, if D. ‌giganteum and Mammuthus sp. share a similar posture and weight distribution pattern along their craniocaudal axis, the observed difference might reflect different adaptive responses (e.g. relying more on thick cortex rather than on trabecular filling). In extant elephants, the hourglass-shaped distribution is more pronounced in the forelimb than in the hindlimb, the latter bearing approximately 40% of the body mass (Ren et al. 2010). The marked hourglass-shaped distribution in D. ‌giganteum might then indicate a different weight distribution or adaptive strategy compared to that of extant elephants. Conversely, it is also possible that this pattern is more pronounced in the forelimb of D. ‌giganteum as well: in this case, both D. ‌giganteum and extant elephants would share the same overall pattern (hourglass-shaped distribution more pronounced in the forelimb) although in D. ‌giganteum, this distribution would then be more accentuated across all limb bones. In Mammuthus sp., the less pronounced hourglass shape is unsurprising, as this species is more closely related to extant elephants. Additionally, mammoths generally have shorter hind limbs (Larramendi 2016), resulting in a greater weight load on the forelimbs and reduced strain on the hind limbs. D. ‌giganteum, on the contrary, displays particularly high hips relative to their shoulder height (Larramendi 2016), thus increasing the relative weight on the hind limb, which might explain the more marked hourglass-shaped distribution in the femur, allowing for better resistance to mechanical loading.

The femoral cortex is relatively thinner in Mammuthus sp. (RmaxT = 0.11) than in D. ‌giganteum (RmaxT = 0.36). This variation might be due to multiple factors: as with the cortical thickness distribution, differences in posture and limb proportions could result in a relatively lower mechanical load on the hindlimb of Mammuthus. Additionally, if the Mammuthus specimen studied here is indeed Mammuthus ‌primigenius (c. 6000 kg; Larramendi 2016), it would be much lighter than D. ‌giganteum (c. 12 000 kg; Larramendi 2016). As a result, Mammuthus would experience both lower absolute and relative weight constraints on its femur, leading to less pronounced weight-bearing adaptations. Conversely, this more slender build of the Mammuthus specimen might also suggest differences in overall weight distribution between the species, or different trade-offs between bone robustness (defined as the ratio of bone maximal length to the minimal diaphyseal circumference) and maximal cortical thickness. Indeed, in extant elephants Elephas ‌maximus ‌Linnaeus, 1758 display slightly more robust long bones than Loxodonta ‌africana ‌Cuvier, 1825 despite being lighter. However, L. ‌africana shows slightly thicker cortices, which might compensate the lower robustness of the bones (Bader et al. 2025). In both L. ‌africana and E. ‌maximus, RmaxT falls between that of D. ‌giganteum and Mammuthus sp., with values around 0.20-0.25 (measurements taken from Bader et al. 2025: fig. 6). This “intermediate” maximal relative thickness in extant elephants as compared to the two fossil taxa studied here highlights the range of possible trade-offs and adaptations to heavy weight-bearing in proboscideans. However, as only the femur is considered in this study, without the other long bones, our understanding of weight distribution and microanatomical adaptation in D. ‌giganteum and Mammuthus sp. remains incomplete.

Another observed variation concerns the orientation of trabeculae in the center of the femoral shaft. In Mammuthus, the trabeculae around the growth center are oriented cranio-caudally, with a proximal inclination above the growth center and a distal inclination below it. This pattern is reminiscent of the trabecular arches observed around the growth center in extant elephant femora (Bader et al. 2025). Such an organization might function similarly to architectural arches, enhancing load distribution (Currey 2002). As such, we can hypothesize that pronounced arches might act as a compensating mechanism for a thinner cortex in Mammuthus, by distributing the load along the shaft. In Deinotherium, the presence of trabecular arches is not observable due to the incompleteness of the specimen and the presence of plaster fillings in the central part of the shaft. As a result, we cannot conclude on the possible link between the hourglass-shaped distribution of the cortex and the orientation of the trabeculae.

Microanatomical variations and morphotypes

The two femora analyzed in this study provide valuable new insights, particularly in the case of D. ‌giganteum. While all proboscideans from the Miocene onwards exhibit columnar limbs, there is considerable variation in the external morphology of their bones (Bader et al. 2024) and in their limb proportions (Larramendi 2016; Belyaev et al. 2025). Graviportal proboscidean limb bones can be categorized into two main morphotypes: one characterized by relatively thin shafts and narrow epiphyses (including elephantids and deinotheres) and another with thicker shafts and larger epiphyses (including mammutids and gomphotheres) (Bader et al. 2024). Although these morphotypes are not absolute categories, they represent two biomechanical strategies for weight-bearing in graviportal proboscideans, and a such raise the question of potential associated microanatomical variations.

To our knowledge, all graviportal proboscidean specimens available for microanatomical analysis belong to either deinotheres or elephantids, i.e. from the same “thin” morphotype, so that no direct comparison has yet been made between the two morphotypes. Interestingly, the “thin” morphotype does not follow the phylogeny, as it evolved convergently in both elephantids and deinotheres (Bader et al. 2024). As such, this raises the question of whether this morphological similarity is associated with shared microanatomical adaptations. While limited by the scarce material, the present study provides partial insights: our findings suggest that in both D. ‌giganteum and Mammuthus, trabecular bone adapts in a broadly similar manner to support massive body weight, reminiscent of what is observed in extant elephants. These results suggest that in both deinotheres and elephantids, the “thin” morphotype is associated with overall similar microanatomical features. However, we observed a clear difference in relative cortical thickness, which is much greater in the femur of D. ‌giganteum. This variation may be linked to differences in absolute body mass, weight distribution or posture, but might also reflect distinct biomechanical strategies; further comparisons including complete bones of both species are needed to conclude on the link between external morphology and bone inner anatomy in relation to limb bone morphotype.

CONCLUSION

This study highlights key microanatomical adaptations in the femora of Deinotherium and Mammuthus, revealing a shared structural organization associated with graviportal weight-bearing. Both taxa exhibit thick cortical bone in the diaphysis, a medullary area largely filled with trabecular tissue, and highly anisotropic trabeculae aligned with axial loading. These patterns are similar to those observed in extant elephants, where they allow for efficient load distribution and mechanical resistance; our results thus highlight overall similar microanatomical adaptations to massive weight support among graviportal proboscidean taxa. Despite these global similarities, the femur of D. ‌giganteum displays a more pronounced hourglass-shaped cortical distribution and greater relative cortical thickness than that of Mammuthus, suggesting potential differences in weight distribution and/or posture, or potentially different adaptive features of the bone microanatomy.

Acknowledgements

We warmly thank Erzsébet Latrán for providing the camera, and Mara Bader and János Magyar (HNHM, Budapest) for their assistance in moving heavy material. We would also like to thank Márton Szabó (HNHM, Budapest) for his help in the curatorial work. We are grateful to Ursula B. Göhlich (NHM, Vienna), Vlad Codrea (UBB, Cluj-Napoca), Ștefan Vasile (UniBuc, Bucharest), Márton Venczel (Muzeul Ţării Crişurilor, Oradea), and Attila Virág (ELTE, Budapest) for the useful consultations and for sharing literature. Many thanks to László Makádi (SARA, Budapest), Gábor Papp, Angéla Matuszka, Attila Vörös, János Szabó and Tibor Kecskeméti (HNHM, Budapest) for sharing their historical knowledge and helping us investigate the origin of bones. We are also grateful to two anonymous reviewers for their useful comments and suggestions, as well as Emmanuel Côtez (MNHN, Paris) and Sylvain Charbonnier (MNHN, Paris) for editorial work.

Credit statement

Conceptualization CB; Data curation CB, MG, MS; Formal Analysis CB; Investigation CB, MS; Methodology CB; Project Administration CB; Validation CB, MG, MS; Visualization CB; Writing – Original Draft Preparation CB; Writing – Review & Editing CB, MS.

Data availability

The 3D models are accessible on MorphoSource:

Deinotheriumsp001 (HNHM-V.79.166.) – 10.17602/M2/M630968;

HNHM-PAL 2025.5.1 – 10.17602/M2/M723153.

Fig. 1. — 3D model reconstruction of the Deinotherium ‌giganteum ‌(Kaup, 1829) femur (HNHM-V.79.166.) in distal (A), caudal (B) and cranial (C) views and of the Mammuthus sp. femur (HNHM-PAL 2025.5.1.) in cranial (D) and medial (E) views. Dotted line indicates the boundary between the two reconstructed fragments of HNHM-V.79.166. Abbreviations: Cr, cranial; M, medial; L, lateral; P, proximal. 3D models by Camille Bader. Scale bars: 10 cm.10.5281/zenodo.19604425

Fig. 2. — Distribution of the bony tissues in the femur of: A, B, Deinotherium ‌giganteum ‌(Kaup, 1829) (specimen HNHM-V.79.166); B, lateral view; C, Mammuthus sp. (specimen HNHM-PAL2025.5.1), medial view. Abbreviations: Ca, caudal; Cr, cranial; P, proximal. Photos by Camille Bader and Martin Segesdi. Scale bars: 10 cm.10.5281/zenodo.19604427

Fig. 3. — General and zoomed-in views of trabecular orientation in the femur of HNHM-V.79.166. of Deinotherium ‌giganteum ‌(Kaup, 1829) in medial (A) and lateral (B) views. Abbreviations: Ca, caudal; Cr, cranial; P, proximal. The red lines indicate the direction of the trabeculae in highly anisotropic areas. Photos by Camille Bader and Martin Segesdi. Scale bars: A, B, 10 cm; a-f, 1 cm.10.5281/zenodo.19604429

Fig. 4. — General and zoomed-in views of trabecular orientation in the femur of HNHM-PAL 2025.5.1 of Mammuthus sp. in medial view. Abbreviations: Ca, caudal; P, proximal. The red lines indicate the direction of the trabeculae in highly anisotropic areas. Photos by Camille Bader and Martin Segesdi. Scale bars: A, B, 10 cm; a-f, 1 cm.10.5281/zenodo.19604432

Fig. 5. — Zoomed-in views of trabecular orientation in the femur of HNHM-PAL 2025.5.1. of Mammuthus sp. in cranial view: A, lateral border; B, medial border. Abbreviations: M, medial; P, proximal. The red lines indicate the direction of the trabeculae in highly anisotropic areas. Photos by Camille Bader and Martin Segesdi. Scale bars: 1 cm.10.5281/zenodo.19604434

Fig. 6. — Supplementary Figure S1: remaining fragments of the HNHM-V.79.166. Deinotherium ‌giganteum ‌(Kaup, 1829) specimen (femur). A-C, greater trochanter; D, E, proximal half of the diaphysis; F, 3D model of a modern elephant with corresponding areas highlighted. Abbreviations: Ca, caudal; L, lateral; M, medial; P, proximal. Scale bars: 5 cm. photos by Camille Bader and Martin Segesdi, 3D model and illustration by Camille Bader.10.5281/zenodo.19604436

Fig. 7. — Supplementary Figure S2: remaining fragments of the HNHM-V.79.166. Deinotherium ‌giganteum ‌(Kaup, 1829) specimen (tibia). A-C, proximal part of the diaphysis and proximal epiphysis; D, 3D model of a modern elephant tibia with the corresponding area highlighted. Abbreviations: Cr, cranial; L, lateral; M, medial; P, proximal. Scale bars: 5 cm. photos by Camille Bader and Martin Segesdi, 3D model and illustration by Camille Bader.10.5281/zenodo.19604438

Fig. 8. — Supplementary Figure S3: The HNHM-V.79.166. Deinotherium ‌giganteum ‌(Kaup, 1829) specimen mounted in exhibition. The picture was provided by the Library of the Hungarian Natural History Museum.10.5281/zenodo.19604440

Fig. 9. — Supplementary Figure S4: comparison of the femur mounted in exhibition (A, newsreel clip from 1961.05.01) and the reconstructed 3D model of the broken specimen HNHM-V.79.166 (B). The video is accessible at https://filmhiradokonline.hu/watch.php?id=15855 [accessed March 18, 2025].10.5281/zenodo.19604442

Table 1. — Linear measurements taken on the two femora of Deinotherium giganteum (Kaup, 1829) and Mammuthus sp. Abbreviations: AmaxT, absolute maximal cortical thickness; D, corresponding diaphyseal diameter; RmaxT, relative maximal cortical thickness defined as the ratio of AmaxT to D. Species Specimen ID AmaxT (cm) D (cm) RmaxT Deinotherium ‌giganteum HNHM-V.79.166 4.86 13.51 0.36 Mammuthus sp. HNHM-PAL 2025.5.1 1.08 9.45 0.11