The dinosaurs that weren’t: osteohistology supports giant ichthyosaur affinity of enigmatic large bone segments from the European Rhaetian

Bone segments and putative giant ichthyosaurs from Europe

Large, but fragmentary bone finds from the famous Aust Cliff Rhaetic bone beds of the Bristol area (southwestern UK) were already reported in the 19th century (Stutchbury, 1850). These include what appeared to be large limb bone shafts of reptilian affinity, leading to extensive discussions in the paleontological community (Stutchbury, 1850; Sanders, 1876; Huene, 1912; Storrs, 1993; Storrs, 1994; Benton & Spencer, 1995; Galton, 2005; Naish & Martill, 2008; Redelstorff, Sander & Galton, 2014; Lomax et al., 2018). The Aust Cliff bone bed is one of a group of similar UK and continental European bone bed-type deposits formed in the Rhaetian epicontinental sea that covered much of Western and Central Europe (Sander et al., 2016; Barth et al., 2018; Cross et al., 2018; Perillo & Heijne, 2023) (Fig. S1). These bone beds yield various tetrapod fossils of both terrestrial and marine origin, often showing fragmentary preservation (Storrs, 1993; Storrs, 1994). The proposed taxonomic affinities of the large to gigantic bone shafts, hereafter less suggestively called “bone segments”, include “labyrinthodonts” (Stutchbury, 1850), dinosaurs (Sanders, 1876; Reynolds, 1946; Storrs, 1993; Storrs, 1994; Benton & Spencer, 1995; Galton, 2005) and unidentified archosaurs (Redelstorff, Sander & Galton, 2014).

The dinosaurian origin of said bone segments (hereafter ‘Dinosaur Hypothesis’) has been supported for the last decades, with Galton (2005) discussing five of the bone segments in detail and concluding that they either must represent sauropodomorph or, more likely, stegosaur long bone shaft fragments (femur, ?tibia). An inconsistency with the long bone nature of the segments would seem to be their lack of a continuous cortex and periosteal surface around their periphery. Instead, as much as two thirds of the periphery of shaft cross sections appears to consist of cancellous bone (Galton, 2005, figs. 4–6). Galton (2005) had already noticed the lack of an outer bone surface in some areas. Whereas this feature could be primary, as in a jaw bone (representing a suture surface or a surface facing the Meckelian canal), it also could result from heavy abrasion, which characterizes all Aust Cliff and other bone bed material.

Galton’s (2005) conclusion as to the stegosaurian nature of the bone segments has since been questioned by multiple workers (Maidment et al., 2008; Naish & Martill, 2008; Sander, 2013; Redelstorff, Sander & Galton, 2014; Lomax et al., 2018) due to the lack of diagnostic morphological features and stratigraphic arguments. In particular, the largest known stegosaur already occurring in the Late Triassic would be inconsistent with the known ornithischian fossil record and result in long ghost lineages (Galton, 2005; Maidment et al., 2008; Naish & Martill, 2008). Sauropods, on the other hand, would appear to be a reasonable option.

A histological test of sauropod affinities of the Aust Cliff bone segments was then conducted by Redelstorff, Sander & Galton (2014). Sampling two of the Aust Cliff specimens (BRSMG-Cb-3869 and BRSMG-Cb-3870, see Table 1) (Redelstorff, Sander & Galton, 2014) found a peculiar and previously undescribed set of histological characters (a thin cortex of fibrolamellar bone with longitudinal primary osteons and secondary osteons forming within the primary ones), inconsistent with sauropod or other sauropodomorph affinities (Redelstorff, Sander & Galton, 2014). In their primary cortex, sauropodomorph long bones show a different and rather uniform histology: laminar and plexiform fibrolamellar bone and, in the case of sauropods, almost no growth marks until late in life (Sander & Klein, 2005; Klein & Sander, 2007; Klein & Sander, 2008; Sander et al., 2011).

Following the recent find of a very large elongate and partially curved bone segment (BRSMG-Cg-2488, 96 cm long, Fig. S3B) in the Rhaetian of Lilstock (Lomax et al., 2018), also in SW England, this segment and the Aust Cliff bone segments were identified as fragments of the surangular bone derived from giant ichthyosaur jaws by Lomax et al. (2018). This interpretation by Lomax et al. (2018) was based on a morphological comparison with somewhat older giant ichthyosaurs from North America, specifically the Carnian Shonisaurus popularis from Nevada (Camp, 1980) and the Norian Shastasaurus sikanniensis (Fig. S3A) from British Columbia, Canada (Nicholls & Manabe, 2004). We term this hypothesis of the affinity of the very large Aust Cliff bone segments the ‘Giant Ichthyosaur Hypothesis’.

Support for the Giant Ichthyosaur Hypothesis would seem to come from an earlier find, now lost (Fig. S3C). Huene (1912) described a 1.4 m long bone segment from Aust Cliff which he identified as the fragment of a right lower jaw of a giant ichthyosaur, including part of four elements (dentary, splenial, angular, surangular) (Fig. S3C). Huene (1912) noted that this fossil had been accessioned to the “Bristol Museum” since 1877, presumably referring to today’s Bristol City Museum and Art Gallery (BRSMG). However, Huene (1912) did not provide a specimen number, and since his 1912 study, the specimen has not been mentioned again, and it may well have been destroyed in WWII. According to Huene’s (1912) description and illustration, the specimen consists of four non-fitting parts, the penultimate of which had been sectioned transversely (Fig. S3C) at some earlier point in time before Huene’s study.

Curiously, among the putative dinosaur long bone material described by Galton (2005) from Aust Cliff, there also is a transversely sectioned specimen (BRSMG-Cb-3870, Fig. S2) of about the dimensions noted by Huene (1912) (Fig. S3C). Galton did not cite Huene, and there is a possibility that the two authors did study the same specimen. Arguing against the identity of the two specimens is the poor preservation of the Galton specimen (whereas Huene emphasized the good preservation of his material) and the fit with another segment (whereas Huene noted the lack of fits).

Finds similar to the Aust Cliff and Lilstock material have come from the epicontinental French Rhaetian localities of the Autun area (Fig. S1) and from southern France (Fischer et al., 2014; Lomax et al., 2018), as well as most recently, from the German locality of Bonenburg (Fig. S1) (Sander et al., 2016; Wintrich et al., 2017) and the Swiss Alps (Sander et al., 2022, fig. s5). Fischer et al. (2014) also had interpreted their material as ichthyosaurian but did not extend their considerations to the UK material and did not cite Huene (1912). Huene (1912), on the other hand, just described this one specimen from Aust Cliff and did not comment on the putative dinosaur leg bone shafts from the same locality nor on the French Rhaetian ichthyosaur material, all of which were known at the time.

The Late Triassic giant ichthyosaur record

Ever since the work of Charles S. Camp on Shonisaurus popularis from Berlin Ichthyosaur State Park in the Carnian Luning Formation of Nevada, USA, in the 1950s (Camp, 1980), it has been clear that Late Triassic ichthyosaurs reached body lengths of 15 m or more and must have been substantially larger than post-Triassic ichthyosaurs. The S. popularis material has been reevaluated several times since with regard to its size, skeletal reconstruction, taphonomy, and reproductive biology (Kosch, 1990; Hogler, 1992; McGowan & Motani, 1999; Kelley et al., 2022). Even larger and more complete than any of the S. popularis finds is the holotype skeleton of Shastasaurus sikanniensis (Nicholls & Manabe, 2004) from the middle Norian of British Columbia, Canada. Based on field data, this individual is estimated to have been 21 m long (Nicholls & Manabe, 2004).

It is also now acknowledged that various other ichthyosaur finds from the Late Triassic must represent animals over 10 meter in length, but most giant ichthyosaurs are represented by woefully incomplete, disarticulated, and fragmentary material from around the world (Callaway & Massare, 1989; McGowan & Motani, 1999; Sander et al., 2022; Kelley et al., 2022) which hinders the anatomical descriptive effort. In continental Europe, the fragmentary, often reworked, and poorly understood finds attributed to giant ichthyosaurs come from late Norian to Rhaetian outcrops of France (Fischer et al., 2014), the eastern Swiss Alps (Sander et al., 2022), and from a recently discovered Aust Cliff-type bone bed near the central German village of Bonenburg (Fig. S1) (Sander et al., 2016; Wintrich et al., 2017). Unlike all the other Rhaetian localities with putative giant ichthyosaurs, the Bonenburg deposit is precisely dated palynologically, ranging from late middle to early late Rhaetian in age (Schobben et al., 2019; Gravendyck et al., 2020). The Bonenburg ichthyosaur fossils include large but very short vertebral centra, a very large neural arch, and very large rib fragments (Sander et al., 2016)). In addition, the bone bed frequently yields heavily abraded fragments of thick cortical bone up to 25 cm in length (Figs. S4A, S6A), which we hypothesize to be fragments of bone segments similar to the more complete British and French specimens (Figs. S2A, S3B).

Understanding the affinity of the fragmentary Late Triassic ichthyosaurs and of the large, more obscure fragmentary finds, is important because of the absolute size of these remains, representing records of the largest animals inhabiting the Late Triassic oceans (Lomax et al., 2018; Sander et al., 2022). The fossils represent animals that far exceeded the size of any other marine tetrapods except for the largest species of baleen whales and archaeocetes (Bianucci et al., 2023). The importance of these fossils also relates to the patterns of extinction at the end of the Triassic, given that very large ichthyosaurs appear to have persisted to the late Rhaetian (indicated by the Bonenburg finds) but are lacking in the Jurassic.

The lack of clear and unequivocal external morphological features in the Rhaetian European bone segments due to their fragmentary and reworked nature makes alternative approaches such as microstructure analysis (microanatomy and osteohistology) critically important for investigating the possible affinities of these fossils. Both Galton (2005) and Lomax et al. (2018) illustrated cross sections of UK fossils and discussed microanatomy (but not histology, which is not accessible without thin-sectioning). Galton compared the midshaft microanatomy of BRSMG-Cb-3869, 3870, and 4063 from Aust Cliff to that of various dinosaurs and concluded that the fossils must represent stegosaurs based on the coarse cancellous bone structure of the medullary region. Lomax et al. (2018) noted and illustrated in detail the same coarse cancellous bone structure but did not use microanatomical arguments as evidence for determining affinity, only cross-sectional shape. Histological analysis was already performed on two Aust Cliff specimens (BRSMG-Cb-3869 and BRSMG-Cb-3870) by Redelstorff, Sander & Galton (2014) (Table 1), but without considering possible ichthyosaurian affinities of the fossils.

Here we undertake a detailed and comprehensive comparison and sampling of most European Rhaetian “bone segments” and putative giant ichthyosaur jaws for histological analysis. The main aim of this study thus is to histologically test the Giant Ichthyosaur Hypothesis by searching for shared histological characters among European material of confirmed or proposed ichthyosaurian nature, on one hand, and bonafide Late Triassic giant ichthyosaurs, such as S. sikanniensis, on the other. We also compare the “bone segments” histology with other terrestrial and aquatic tetrapods that are known to have reached very large body size in the Late Triassic such as sauropodomorph dinosaurs, rauisuchians, dicynodonts, and plesiosaurs.

Materials

The material used in this study consists of bone histological samples taken from various specimens borrowed from multiple institutions as listed in Table 1. Abbreviations for these institutions are also listed in this table. In summary, there are eight sets of samples (Table 1). These include two samples (surangular, splenial) from the S. sikanniensis holotype RTMP-1994-378-0002 (Nicholls & Manabe, 2004) (Fig. S3A), one sample of the Lilstock putative ichthyosaur surangular (Lomax et al., 2018) (Figs. S2, S3B), three samples of “dinosaur bone shafts” reinterpreted as giant ichthyosaur jaw bone fragments from the Aust Cliff Rhaetic bone bed (Galton, 2005; Redelstorff, Sander & Galton, 2014; Lomax et al., 2018), two samples from a giant putative ichthyosaurian lower jaw (Fischer et al., 2014), identified as surangular by Lomax et al. (2018), from Autun, France (Figs. S2, S3B), and finally 16 cortex fragments of various sizes from Bonenburg, Germany (Figs. S4, S5A, S6A). For details on all of these samples, including sampling locations and methods, and their current identification, see Article S1.

The thin sections used for the study are either in the paleohistological collections of the IGPB or with the sampled fossils (see Table 1). Note that two of the Aust Cliff thin sections were already studied by Redelstorff, Sander & Galton (2014).

Methods

Shared histology of the British and French samples

General histological and microanatomical description

Laid down by the apposition front (Mitchell & Sander, 2014), the outer cortex of all samples from the British and French Rhaetian is characterized by compact primary bone tissue structured by wavy growth marks parallel to the outer bone surface (Figs. 1A, 2, 3A, 3B).

The primary periosteal bone matrix is a new matrix type, “intrinsic fiber matrix” or IFM. IFM is characterized by a network of bright anisotropic, intrinsic, mineralized fibers set in an isotropic matrix. IFM pertains to the woven-fibered type of bone matrices which are produced by static osteogenesis (Buffrénil & Quilhac, 2021a). It is generally accepted that the degree of organization of the collagen fibers in a bone matrix is negatively correlated with its rate of deposition (Buffrénil & Quilhac, 2021b). Because of its structure being more organized than simple woven bone but less organized than parallel fibered bone, IFM probably represents the result of an intermediate type of bone deposition between these two bone matrix types. Contrary to a normal woven-fibered bone matrix, IFM contains abundant coarse intrinsic collagen fibers, both mineralized and unmineralized, that are uniformly oriented longitudinally. Contrary to parallel fibred bone, IFM coarse fibers have different orientation showing a net or lattice-like pattern, immersed in a clearly isotropic matrix. Of particular relevance is that the fibers are intrinsic, not extrinsic (as e.g., Sharpey’s fibers).

Vascularization of the primary cortex is characterized by longitudinal vascular canals (Figs. 2, 3A, 3B). Immature primary osteons and vascular canals open up to the outer bone surface, resulting in an ornamented wavy surface (Figs. 2, 3B) in thin section. This histology correlates with distinctive longitudinal surface striations on the specimens (Fig. S3D), nicely illustrated by Lomax et al. (2018, fig. 4c, 8) for the Lilstock and Aust specimens and Fischer et al. (2014; fig. 2, s5) for the French specimens.

Vascular canals and primary osteons are arranged in appositional circumferential rows demarcated by closely spaced growth marks (GM) that vary in number (Figs. 1A, 2, 3A, 3B). Growth marks appear as depositional layers of periosteal primary bone and run around the periosteal vascular canals (Figs. 1A, 3A, 3B), appearing to embrace them. The GM vary in thickness (Figs. 1A, 3A vs 3B) and show alternating light-dark coloration. The differential coloration seems to be related to differences in intrinsic fiber density and orientation. Vascularization as observed in longitudinal sections does not show anastomoses between vessels, with vessel cross sections rarely showing shapes more complex than an elongated ellipsoid (Fig. 3C).

The Haversian substitution front is diffuse in that the outer cortex shows scattered evidence of secondary remodeling through small resorption cavities and secondary osteons within primary ones (Figs. 1A, 1D), or even mature secondary osteons (Fig. 3A). Appositional rows of primary osteons may follow or precede rows of primary osteons with secondary ones within them, or rows of secondary osteons may even be intercalated with rows of purely primary osteons (Figs. 1A, 1D, 3E). The deep cortex, i.e., the part of the cortex that is fully within the Haversian substitution front, can again be subdivided in an outer template deep cortex and an inner, completely remodeled area, where none of the primary pattern of vascularity is preserved (Fig. 2). This situation was already described in detail by Redelstorff, Sander & Galton (2014, fig. 4). The thickness of these two subzones of the deep cortex varies between samples (Fig. 2).

As noted above, the template deep cortex is so named because it preserves most or some of the original primary vascular architecture (Figs. 2A–2E, 3E). This is because of the peculiar pattern that Haversian substitution is initiated from existing primary vascular canals, i.e., existing vascular pathways were reused. This results in a predominance of secondary osteons within primary ones. Of further relevance is the generally small diameter of the secondary osteons, which is comparable to the diameter of primary osteons (Figs. 1A, 1D, 2F, 3A, 3AE), unlike what seen in other amniote taxa (see Discussion). The primary osteons thus clearly influence the course of the secondary ones, even leading to rows of exclusively secondary osteons forming complete Haversian tissue, templated by the primary rows of osteons (Fig. 3E). The templating we observed is different from normal Haversian substitution in amniotes in which the cutting cones of secondary osteons show little regard for preexisting structures (Mitchell, 2017).

Both, primary and secondary osteons, have a high number of lamellae and a small vascular canal, which results in a rather low average porosity for the entirety of the sections (between 17% and 13%) (Figs. S7A–S7E), possibly indicating an osteosclerotic state of the cortex. Porosity generally decreases from the deep cortex toward the outer cortex, except for BRSMG-Cg-2488 R-101, which shows rows of widened vascular canals at the transition between deep and outer cortex.

Osteon cross sections vary consistently between circumferential rows, sometimes horizontally flattened, sometimes more vertically (Figs. 2, 3E), indicating modulations resulting from variations in growth rate (Woodward, 2019). Migratory and incipient osteons (Skedros, Sorenson & Jenson, 2007; Mitchell, 2017) are present, but secondary osteons within primary ones represent the majority of osteons in the template cortex.

Further inward from the template cortex, the regular deep cortex can be seen as resulting from complete secondary reconstruction (Figs. 2A–2D). This part of the cortex is characterized by more chaotically arranged secondary osteons that have obliterated the primary vascular architecture by several cycles of secondary osteon formation. The result is normal Haversian tissue which marks the full effect of the Haversian substitution front (e.g., Fig. 2D).

The boundary of the perimedullary region, i.e., the resorption front, is also diffuse (Figs. 2A–2D). Here, the deep cortex becomes more and more affected by larger resorption cavities lined only by a few lamellae. Porosity in the perimedullary region is between 65 to 85% (Figs. S7A–S7E). This signifies an increasing imbalance between secondary bone deposition and resorption activity and initiates the formation of secondary trabeculae (Figs. 2A–2E). Through the activity of the resorption front, the perimedullary region is rich in resorption cavities replacing bone tissue with some secondary osteons and transitioning to a medullary area of secondary trabecular bone (Figs. 2A–2E).

In cases, where the resorption front has overtaken the Haversian substitution front, interstitial areas of primary tissue consisting of IFM are visible (Fig. 3D) between the secondary trabeculae. The percentage of interstitial primary tissue decreases inwards but is patchy. Our histological observations are consistent with the descriptions and illustrations of cross-sectional microanatomy given by Galton (2005), Fischer et al. (2014), and Lomax et al. (2018), who all note that there is only a very small open medullary cavity surrounded by an extensive zone of inward-decreasing trabecular density (Galton, 2005, figs. 4–6; Fischer et al., 2014, fig. s5; Lomax et al., 2018, fig. 6). Fischer et al. (2014, fig. s5) interpret this open medullary cavity as the dental groove, however.

Both, the largest Aust Cliff bone segments (BRSMG-Cb-3869) and the Lilstock specimen (BRSMG-Cg-2488 R-101), show conspicuous cavities in the cortical bone (Figs. 2A, 2C, 3F). There are two obvious ones in the latter and one obvious and a second possible one in the former (Figs. 2A, 2C, S7A, S7C). Based on the sampling location, these open cavities represent the nutrient canals extending inwards at a low angle from the elongate foramen (Figs. S2A, S3B–S3C) opening in caudal direction (already described by Huene, 1912 and identified as part of the fossa surangularis by Lomax et al., 2018) on the bone surface. On the outward margins of the cavities (those facing the periosteal surface), both samples show primary tissue and simple vascular canals (Fig. 3F). Both BRSMG-Cb-3869 and BRSMG-Cg-2488 R-101 show signs of resorption along the inner and lateral margins of the nutrient canals, indicating microanatomical drift related to the growth of the bone enclosing the canal.

Histology of indeterminate cortical fragments from Bonenburg, Germany

The largest cortical fragment from Bonenburg (WMNM P88133), the thin sections produced from WMNM P-uncatalogued (probably derived from cranial material), and the numerous smaller unidentified cortical fragments (for which no precise anatomical placement is possible) share the same primary bone tissue and overall general features (Figs. 4, S5, S6). WMNM P88133 has a primary cortex rather similar to the previously discussed samples from Europe. However, observation of histology of this and most of the other Bonenburg samples is hampered by a nearly opaque outer diagenetic zone >2 mm wide (Figs. 4A, 4B). The remaining bone tissue is very well preserved. Macroscopically, the outer bone surface bears fine longitudinal striations (Figs. S4, S5A, S6A), of the same kind noted above for the Aust and Lilstock specimens (Fig. S3D).

As in the other specimens, vascularization is strictly longitudinal (Figs. 4A, 4B, S5B, S5C, S6B, S6C). Simple vascular canals and primary osteons are arranged in surface-parallel rows which may be enhanced by GM bordering and embracing the vascular canals (Figs. 4B, S6C). An external fundamental system does not appear to be present. The bone matrix, in which the vascular canals and primary osteons are set, is IFM, and together they form PIFT (Figs. 4C, S5C, S6C, S6D).

The GM show alternations of differently colored IFM but do not show an appreciable pattern in spacing, while they appear to show differences in fiber density (Figs. 4C, S6C, S6D). Under cross-polarized light, it is possible to observe clearly bright coarse fibers in the paler yellow areas, with a reduction of their presence corresponding with increased darkness in areas of darker brown color (Figs. 4B, 4C, S5C, S5D). The darkest GM seem to be made up by fewer intrinsic fibers (Fig. 4C).

Osteocyte lacunae are numerous in the IFM with mainly plump and irregularly shaped ones throughout tissue, while more flattened ones are scarcer and present only in centripetal lamellae of osteons (Fig. 4D). Osteocyte lacunar density and size is greater in the primary bone matrix compared to the lamellar bone of the osteons (Figs. 4D, S5D).

The largest fragment (WMNM P88133) is characterized by a continuous gradient in bone compactness from the inner more cancellous area to the outermost cortex (Fig. 4A), resulting from the advancement of the Haversian substitution and resorption fronts. A wide and diffuse Haversian substitution front is detectable toward the center of the section, evidenced by the interruption of the semicircular GM (Fig. 4B). Elsewhere in the section, secondary osteons develop preferentially within primary ones, conserving the primary arrangement of the osteon rows (Figs. 4B, 4C). Rarely, there are secondary and primary osteons showing an infill of centripetal layers of anisotropic woven bone (Fig. 4D) alongside osteons showing presence of intrinsic fibers similar to the ones of the surrounding IFM.

In the deep cortical area, there is a high amount of resorption cavities eroded into the compact bone which consist of primary bone only partially replaced by secondary remodeling. Resorption cavities are lined by lamellar bone resulting in an increasingly cancellous condition.

The longitudinal section (Fig. 4E) is dominated by simple longitudinal canals with a very limited degree of anastomosis. It is possible to observe diffuse strands of thin, dark fibers of variable length, distributed mainly longitudinally (Fig. 4E). Sometimes, the fibers intersect each other next to the edges of the vascular openings (Fig. 4E). The borders of secondary osteons are lined by bright lamellar bone, darker bone tissue, and by both types together in an alternating fashion.

Histology and microanatomy of Shastasaurus sikanniensis holotype jaw bone samples

Both of the jaw bone samples from the holotype of S. sikanniensis show poor histological preservation, which hides most of the discernable features in the areas where the bone tissue is most altered. This is particularly evident in the surangular. Poor preservation of birefringence is accompanied by a dark brown stain of the tissue, making it nearly opaque. However, with sufficiently bright illumination, the salient features, in particular the presence of IFM, can be discerned (Figs. 5B–5D). Both the surangular and splenial histology are characterized by highly spongious secondary bone tissue (porosity ∼82% and ∼60%, respectively), dark brown in color under the crossed polarizers (Fig. 5B). Towards the outer bone surface, which appears to be compromised by preparation (see below), there are interstitial areas of primary tissue characterized by distinctive IFM, with an outwards increase in frequency. Although no obvious dense cortical bone is present, a decrease in porosity is detectable toward the outer bone surface of both samples (respectively 64% and 43% porosity) with smaller longitudinal vascular cavities and higher compactness. The vascularization, consisting of large Haversian canals and resorption cavities, is strictly longitudinal (Figs. 5A, S3D, S7E, S7F). This is also seen with the naked eye on the outer bone surface which shows regular fine striations (Fig. S3E) as observed in the European specimens.

The secondary osteons visible in thin section have only few centripetal lamellae. On the outer bone surface, the presence of osteons half cut open indicates the removal of tissue due to taphonomic or diagenetic causes or harsh preparation (Figs. 5A, S3D). It is not possible to determine the presence of secondary osteons within primary ones in the trabecular bone. It is debatable, though, whether the absence of secondary osteons within primary ones is genuine or simply related to the lack of enough compact bone tissue in the sampled location of the jaw bones.

Rejection of the “Dinosaur Hypothesis”

Although we do not question the ichthyosaurian status of the Lilstock and Autun specimens based on their morphology (Fischer et al., 2014; Lomax et al., 2018), the morphological information provided by the more fragmentary specimens (BRSMG-Cb-3869, BRSMG-Cb-3870, BRSMG-Cb-4063 from the UK and WMNM P88133 from Bonenburg, Germany) is insufficient for recognizing their systematic affinities (beyond excluding certain identifications, like as ichthyosaur or plesiosaur long bones). Thus, a comprehensive histological comparison was needed. Based on the histological evidence obtained from sampling bonafide (i.e., S. sikanniensis) and putative Late Triassic giant ichthyosaurs, we regard as relevant four histological features (Table 2), three of which had already been noted in the histological study of two Aust Cliff samples by Redelstorff, Sander & Galton (2014). IFM had not been reported by these authors, but was recognized by us in the same Aust Cliff thin sections.

All four features (Table 2) are present in the Lilstock and Autun ichthyosaur jaw specimens from the European Rhaetian. The most distinctive feature, IFM, is present in the jaw of the type specimen of the best-known giant ichthyosaur, S. sikanniensis. The uniqueness of IFM thus provides strong support for the “Giant Ichthyosaur Hypothesis”. Redelstorff, Sander & Galton (2014) concluded that the histology of the Aust Cliff bone segments did not resemble the histology of any dinosaur long bones known at the time. This statement still holds true, especially since the histology of virtually all dinosaur clades, and especially all large-bodied ones, is known by now. We thus can confidently reject the Dinosaur Hypothesis.

Testing other possible affinities using histology

To test for the presence of a similar combination of features and to further test the “Giant Ichthyosaur Hypothesis”, we performed extensive histological comparisons, considering a “Non-dinosaur Hypothesis”, addressing known large or giant Late Triassic tetrapods, both terrestrial and aquatic. The results of this comparison are summarized in Table 2 and discussed in more detail below. We found that the unique combination of histological features of the Aust Cliff bone segments and German cortical fragments, combined with their large size (Fig. S3B), thick cortices, and shaft-like shape, rules out affinities with any other Late Triassic giant tetrapod, dinosaurian or non-dinosaurian, other than giant ichthyosaurs.

IFM and PIFT and possible analogs

Although PIFT has not been explicitly described in the literature, it is not uncommon to see published micrographs seemingly showing this type of bone tissue or similar ones. A brief, but probably incomplete, list of examples includes a wide variety of amniotes: the rib sample of a large Nothosaurus specimen (Klein, Canoville & Houssaye, 2019, fig. 4n, o) (Figs. S8A, S8B), the femur of Simosaurus (Klein & Griebeler, 2016, fig. 5) (Figs. S8H, S8I), various bones of the thalattosaur Askeptosaurus (Klein et al., 2023) (Figs. S8D, S8E), the rib of the thalattosuchian crocodylomorph Metriorhynchus (Buffrénil, Quilhac & Cubo, 2021 fig. 10.2f), and the humerus of the ornithopod dinosaur Telmatosaurus (Buffrénil & Quilhac, 2021a, fig. 8.6a). Some of the more suggestive cases noted above are discussed in Article S2 and figured in Fig. S8.

IFM, PIFT and ossified tendons

IFM remarkably resembles extrinsic fibers bundles seen in metaplastic bone tissue of osteoderms (Scheyer & Sander, 2004) and longitudinal fiber bundles of ossified tendons of various dinosaurs (Horner, Woodward & Bailleul, 2016; Surmik et al., 2023). However, we introduced the new terms IFM and PIFT in order to set this clearly periosteal matrix and tissue apart from metaplastic tissues. Nevertheless the similarity of PIFT with metaplastic bone tissue would suggest a shared osteogenic process.

Aside from the similarity between IFM and longitudinal extrinsic fibers, it is possible to observe further similarities with ossified tendons. The longitudinal strands of unmineralized fibers are in a herringbone pattern in both cases (compare Figs. 1E, 4E, S5D and, e.g., Horner, Woodward & Bailleul, 2016 fig. 2g, Surmik et al., 2023 fig. 2f), and there are numerous irregular, sometimes elongate cell lacunae somewhat resembling fibrocytes in shape (Horner, Woodward & Bailleul, 2016). Finally, it is possible to see the presence of occasional centripetal coarse fibrous bone in the Haversian canals of Meleagris gallopavo tendon, figured by Adams & Organ (2005, figs. 2c, d).

The GM in our specimens find close similarity with the structures reported by Horner, Woodward & Bailleul (2016) as regions of varying primary orientation and density of the fibers (Horner, Woodward & Bailleul, 2016 fig. 2d-f). The hypothesis proposed by Horner, Woodward & Bailleul (2016), that the variable color of similar structures is related to the density and orientation of fibers observed in longitudinal sections in ossified tendons, fits our observations (Figs. S6C, S6D) and explains the appearance of such marks. Contrary to what was reported by Horner, Woodward & Bailleul (2016) for ossified tendons, the GM are identifiable as classical cycles of periosteal apposition, given the clear primary origin of these structures in relation to the spatial distribution of periosteal vascular canals and nutrient foramina, and the presence of osteocyte lacunae.

In conclusion, our literature review suggests that IFM is a novel matrix type that has not been reported before in the osteohistological literature. This leads to the question if a bone tissue formed of IFM may be viewed as an apomorphy of a clade of giant ichthyosaurs. This hypothesis would have to be tested by phylogenetic analysis incorporating histological characters, which may well find IFM as a non-unique synapomorphy, resulting from parallel osteogenetic processes. Alternatively, IFM could be mapped on an amniote cladogram and may show up as a synapomorphy.

Template remodeling, osteons within osteons, and unmineralized fibrous matrices

The phenomenon of template remodeling, resulting from the formation of secondary osteons within primary ones, and secondary osteons within secondary osteons (i.e., concentric osteons) appears to be a distinctive shared feature of the histology of the European samples investigated here. The unifying feature of all of these types of secondary remodeling is the reuse by the basic structural unit (bone remodeling unit) of preexisting vascular pathways, be they primary or themselves the result of previous remodeling activity. This reuse of existing pathways is unusual for secondary osteons which in dinosaurs, mammals, and most other amniotes show little regard for the primary histology (Mitchell, 2017). The diameter of secondary osteons in amniotes in general (Mitchell, 2017) is usually substantially greater than that of the primary osteons, resulting in the obliteration of primary features of histology (e.g., Rogers & Kulik, 2018, fig. 4c and several others) and not in templating. We here emphasize the unusual nature of the secondary osteons within primary ones as a pattern we observe. Whereas this pattern may be a special feature of the specimens investigated here, it also could result from the strictly longitudinal orientation of the primary osteons or from the underlying propensity of reuse of vascular pathways. Although we are currently uncertain which of these mechanisms is at work, this does not diminish the systematic value of the secondary osteons within the primary ones.

As already noted, bone remodeling involving pre-existing primary or secondary osteons has been reported in various aquatic tetrapod taxa, such as in the long bones of plesiosaurs (Sander & Wintrich, 2021) and in temnospondyls (Konietzko-Meier et al., 2018). Klein et al. (2015) describe “secondarily widened primary osteons” (Klein et al., 2015 figs. 7, 8, s5) in various placodonts (Sauropterygia). With this term, they refer to the normal transformation process of compact bone to spongy bone. Specifically, these authors note that the resorption activity leading to cancellous bone, i.e., the resorption front of Mitchell & Sander (2014), originates from pre-existing vascular canals. In this way, there is a similarity to secondary osteons within primary osteons. The difference, however, is that in our material the secondary osteons within the primary ones do not lead to cancellous bone, but the tissue remains compact. The transformation to cancellous bone occurs deeper in the cortex. An interesting report is that by Surmik et al. (2023) of what appear to be secondary osteons within primary osteons in ossified tendons of ornithischian dinosaurs (Surmik et al., 2023 fig. 2d). Although confirmation of the presence of this feature would require direct observation of the Surmik et al. (2023) sections, this potential occurrence may be informative on the underlying mechanism of vascular architecture reuse.

The occurrence of a common unusual feature in bone tissue formed by different processes (e.g., periosteal apposition in mandibles and long bones vs. metaplastic ossification of ossified tendons and osteoderms) suggests a common constraint as explanation. In the process of bone resorption, osteoclasts are unable to act on the mineralized bone matrix until the organic protective layer of bone lining cells is removed by cambial cells (Zylberberg, 2021). It has also been hypothesized that sites characterized by non-mineralized structures are less attractive or accessible to osteoclasts (Aaron, 1980; Aaron, 2012; Jones, Boyde & Ali, 1984). The widespread presence of non-mineralized fibers in a bone tissue may significantly inhibit the progression of the basic structural units. The absence of unmineralized fibers in the osteonal bone matrix may thus induce primary osteons to serve as preferential “highways” for osteoclast activity, especially during the initial resorptive phase (i.e., the resorption front), thus explaining the occurrence of abundant secondary osteons within primary osteons and template remodeling.

Alternatively, a difference between the regulatory signals emanating from osteocytes in the outer cortical matrix and those in the osteonal bone matrix may be the primary driver of osteoclast regulation and attraction. Osteocyte regulatory activity is known to be influenced by mechanical loading during development and appears to vary with lacunar shape (van Oers, Wang & Bacabac, 2015). Therefore, it is plausible that the numerous and highly heterogeneous lacunar spaces observed in the matrix may have played a critical additional role.

Implications of PIFT for growth rate, gigantism, and feeding behavior

Several of the features we described are commonly associated with fast growth rates, the most common being a histology dominated by a woven-parallel complex (most commonly fibrolamellar bone), a high degree of vascularization, a high rate of remodeling with multiple generations of osteons, and a high number of osteocyte lacunae, both irregular and spindle-shaped (Buffrénil & Quilhac, 2021b). The presence of numerous open canals in the outer cortex and a well-vascularized outer periosteal surface indicates for all bones sampled that the animals were actively growing at the time of death. The presence of unmineralized fibers in the cortex could be related to rapid mineralization of the osteoid layer laid down by the periosteum as well as to the presence of fibrocytes (Buffrénil & Quilhac, 2021a). The latter cell type would be rather unusual in the formation of periosteal bone, however. The occasional presence of woven bone as infill of osteons may be another feature indicating rapid bone deposition.

Only limited hypotheses regarding developmental stages are possible, due to the absence of enough data to constraint precise taxonomic identity and the homologous location of the thin sections, except possibly for BRSMG-Cg-2488 R-101 and BRSMG-Cb-3869 (see Materials and Methods). In BRSMG 4063, the extension of dense Haversian tissue up to the outer cortex (Fig. 2D) may indicate a mature individual (Buffrénil & Quilhac, 2021b). It is equally likely, though, that the sample could originate from a region subject to high mechanical stress or mineral recycling (Buffrénil & Quilhac, 2021b). The size and number of the cortical vascular canals of BRSMG-Cg-2488 R-101 could indicate a younger, more actively growing individual compared to BRSMG-Cb-3869 (Buffrénil, Quilhac & Cubo, 2021). Before more samples and specimens become available to better understand the biology of these animals, we refrain from further speculations.

The similarity between the bone matrices of giant ichthyosaur mandibles, ossified tendons, and osteoderms invites speculations on the biomechanical properties of the former (as already done by Horner, Woodward & Bailleul (2016) with the hadrosaur nasal). For example, the largest bone segment from Aust Cliff has been suggested to belong to an animal in the size range of modern blue whales (Lomax et al., 2018). Although the feeding strategy of these giant ichthyosaurs remains unknown, it is reasonable to assume that their large jaws were adapted to withstand significant stress associated with hunting and feeding underwater, similar to the feeding behavior of blue whales, which actively process thousands of liters of seawater in one gulp (Goldbogen et al., 2011). Given the high tensile strength of mineralized ossified tendons, it is possible that these large ichthyosaur jaws were selected to withstand similar stresses, either during simple opening, as in baleen whales, or during potential ramming behavior, as observed in odontocetes such as killer whales. At the same time, the high amount of unmineralized fibers in the longitudinal axis of the mandible would have provided some flexibility in different bending planes (Horner, Woodward & Bailleul, 2016). The high rate of remodeling, typically associated with bones subjected to loading, is another factor supporting this hypothesis. It is possible that the presence of specialized soft tissues, such as muscle and connective tissue, likely played an important role in the development of this peculiar histology (Organ & Adams, 2005; Klein, Christian & Sander, 2012; Horner, Woodward & Bailleul, 2016). The occurrence of specializations for buccal processing of large amounts of water (relative to body size) is not isolated within Ichthyosauromorpha (Fang et al., 2023) and is expected in the evolutionary context of achieving giant sizes in marine environments (Sander et al., 2021).