Synchrotron microtomography-based osteohistology of Gansus yumenensis: new data on the evolution of uninterrupted bone deposition in basal birds

2.1.1 Samples of Gansus yumenensis

Numerous specimens of Gansus yumenensis have been excavated to date (You et al., 2006). However, no microstructural analyses have been performed despite the largely three-dimensional preservation of largely uncrushed and undistorted bones, may it be for the analysis of the bone microstructure or the ontogenetic stages. As such, we present here the first osteohistological analysis ever performed on this species.

Five midshaft samples were taken from two specimens of Gansus yumenensis in the IVPP collection: specimens IVPP V21692 and V21693 (Figure 1). Both specimens were excavated from the Changma locality (Suarez et al., 2013) by H-L. You in 2005, and most likely belong to the same individual.

Three midshaft samples from the forelimb elements of the specimen IVPP V21692 were removed: the second metacarpal (Figures 1c, 2 and 3), the third metacarpal (Figures 1d, 4 and 5) and the radius (Figures 1e, 6 and 7). Two midshaft samples were collected from the hindlimb elements of the specimen IVPP V21693: the tibiotarsus (Figures 1f, 8 and 9) and the tarsometatarsus (Figures 1g, 10 and 11). In the case of the tibiotarsus, the sample was fragmented and compressed in situ, and needed to be virtually reconstructed for this study. The overall bone microstructure and osteohistological features of the samples are listed in a table (Table S1).

2.1.2 Samples for comparison

These specimens were chosen as their physical transverse sections could be compared with the virtual cross sections obtained for Gansus yumenensis (see Analysis methods for Gansus yumenensis). They were also chosen for their diverse locomotor means and lifestyles, which in turn affects the osteohistological features in the long bones, as well as their phylogenetic position compared to Gansus yumenensis (Monfroy & Kundrát, 2021).

2.2.1 Scanning methods for Gansus yumenensis

Synchrotron-based X-ray tomographic scans were performed on the collected midshaft samples on the TOMCAT beamline (TOmographic Microscopy and Coherent rAdiology experiments) of the Swiss Light Source (SLS) at the Paul Scherrer Institute (Villigen, Switzerland) (Stampanoni et al., 2006).

The specimens were fixed with beeswax on 3.15-mm-diameter brass pins. The beam energy was set to 35 keV. The X-rays transmitted and refracted by the samples were converted by a 100-μm-thick Ce doped LuAG scintillator to visible light, magnified by a 4x Olympus microscope objective (UPLAPO4x, 0.16 numerical aperture) and recorded by a sCMOS PCO.edge 5.5 camera (Kelheim, Germany). For each scan, 4001 equiangularly spaced projections over 180 degrees have been acquired, with an exposure time of 400 ms per projection. The sample to detector distance was approximately 50 mm, the effective pixel size 1.625 μm and the covered field-of-view 4.160 x 2.860 mm².

Phase retrieved (Paganin et al., 2002) dark- and flat-field corrected projections have been reconstructed to tomographic volumes using an optimised pipeline (Marone et al., 2017), based on gridding and a direct Fourier method (Marone & Stampanoni, 2012). The scanning procedures produced for each sample image stack files of 1762 CT images.

2.2.2 Analysis methods for Gansus yumenensis

The image stacks obtained from the scans were analysed in the PaleoBioImaging Laboratory at the Center for Interdisciplinary Biosciences (TIP UPJŠ, Košice, Slovakia). We used VGStudio Max 3.0 (Volume Graphics, Heidelberg, Germany) in order to generate and describe the osteohistological features on both the virtual transverse sections and the 3D-rendered models for each sample (Figures 2-11). Some of vascular canals and osteocyte lacunae were filled in during diagenesis with minerals that show up as white inclusions in the scans, or a similar greyscale as the bone cortex. Therefore, with the “region growth” tool of VGStudio Max 3.0, they needed to be carefully processed during manual segmentation. The orientation of the canals was deciphered based on the manual segmentation of 50 virtual slices centred on the middle of the diaphysis in the samples, thereby creating a region of interest (Figures 3c.d, 5c,d, 7c,d, 9c,d and 11c,d). The osteocyte lacunae, due to their high quantity throughout the complete sample, were also analysed based on the manual segmentation of 50 virtual slices from the same region of interest, but in a more restricted fashion in the most significant quadrant of the samples (Figures 3e, 5e, 7e, 9e and 11e).

The main anatomical and osteohistological features observed in the midshaft of Gansus yumenensis are summarised in Table 1 and detailed in Table S1.

2.2.3 Linear measurement methods for Gansus yumenensis

Measurements on the various samples of Gansus yumenensis were performed to complete with the osteohistological description, using the in-built tool and estimations functions of VGStudio Max 3.0 in mm with a two-decimal accuracy. All values were measured in a clockwise manner, going from the anterior axis to the lateral axis, then to the posterior axis, and finally to the medial axis (Figure S1a).

For each sample, we measured the length and width of the total bone sample and the medullary cavity with the “calliper” tool function (Figure S1b) on three virtual transverse sections that are perpendicular to the long axis of the bone sample (Figure 1c–g): at the two extremities and in the middle of the sample (slice 882). Due to the incompleteness and preservation of some samples, some measured values were affected, mainly in the tibiotarsus IVPP V21693 (Figures 8 and 9). Two length and width measurements for the total bone sample and the medullary cavity were made in transverse section for the three analysed slices and the samples in total. These values are presented in Table S2. We also measured the thickness of the total bone cortex, the ICL and the mid-cortex using the “calliper” tool function. The thickness was measured two times in four homologous quadrants (anterior, medial, posterior and lateral) on the three transverse sections of interest (Figure S1b). These values are presented in Table S3.

For each sample, we measured on the transverse sections the dimensions of all the osteonal canals using the “calliper” tool function (Figure S1c) while taking into account their orientation from qualitative identifications on both 2D virtual sections and the 3D-rendered model. For the oblique osteonal canals, we used the “polyline length” tool function as these canals exhibit more branching in transverse section. Mean values for each osteonal canal orientation type were estimated for each analysed slice. The values are presented in Table S3.

2.2.4 Stereometric data for Gansus yumenensis

From the osteohistological features selected for each sample, we used the in-built estimation function of VGStudio Max 3.0 to calculate the volume (mm³) and surface area (mm²) values for each selected feature with a three-decimal accuracy. Due to the intrinsic accuracy of the programme, many values were either too small for the programme to estimate accurately and therefore showed up with a zero value. As such, we present here the volume values for the complete sample (i.e. total sample, total bone cortex, medullary cavity, total vascular network) (Table S5). To simplify the volumetric comparisons, we calculated ratios (i.e. total bone cortex compared to the total bone sample, medullary cavity compared to the total bone sample; total vascular network compared to the total bone sample) in order to have a common denominator. These ratio values are presented in Table S5 and are plotted in bar plots (Figure 12a).

For the osteohistological features (i.e. bone cortex, medullary cavity, vascular network, canal orientations) of the regions of interest and the virtual transverse sections, we used surface area values, as the volume values were too low to estimate with VGStudio despite the three-decimal accuracy. As with the volume values, we calculate ratios (i.e. total bone cortex compared to the total bone sample, medullary cavity compared to the total bone sample; total vascular network compared to the total bone sample, canal orientations compared to the total vascular network) to simplify the comparisons with a common denominator. The surface area and estimated ratios values for the region of interest are presented in Table S6. The estimated ratio values were plotted in bar plots respectively (Figure 12b,c).

2.2.5 Comparisons with other birds

In order to provide new insight regarding the changes in osteohistology from long-tailed avialans to extant Neornithes, we compared the osteohistological features (e.g. bone tissue type, orientation and organisation of the vascular canals) from the transverse sections of the Gansus yumenensis samples (Figures 2b, 4b, 6b, 8b and 10b) with those of the other stem and crown birds presented in Material. These specimens exhibit a variety of bone adaptations, as well as various lifestyles to help refine the comparisons. The main osteohistological features of the compared specimens are summarised in Table 1 and detailed in Table S7.

From our comparisons, we perform a qualitative analysis of the relationships between Gansus yumenensis and the other birds of our study. From these comparisons, we looked at the evolution of osteohistological features, such as continuous and uninterrupted bone deposition, throughout the bird lineage (Figure 14).

Abbreviations: CDL, China Dinosaurland (Changzhou, China); DNHM and/or D, Dalian Natural History Museum (China); IMMNH, Inner Mongolia Museum of Natural History (Hohhot, China); Institutional abbreviations. AGB, Anhui Geological Museum, China; IVPP, Institute of Vertebrate Paleontology and Paleoanthropology (Chinese Academy of Sciences, Beijing); MACN, Sección Paleontología de Vertebrados, Museo Argentino de Ciencias Naturales (Buenos Aires, Argentina); MPC, Mongolian Paleontological Center, Ulaanbaatar; SGM, Hebei GEO University (Shijiazhuang City, China); SLS PSI, Swiss Light Source, Paul Scherrer Institute (Villigen, Switzerland); SNSB BSPG, Staatliche Naturwissenschaftliche Sammlungen Bayerns, Bayerische Staatssammlung für Paläontologie und Geologie (Munich, Germany); STM, Shandong Tianyu Museum of Nature (Pingyi, China); TIP UPJŠ, Technology and Innovation Park, University of Pavol Jozef Šafárik (Košice, Slovakia); UCMP, University of California Museum of Paleontology (Berkeley, USA). YPM, Yale Peabody Museum (New Haven, Connecticut, USA).

Osteohistological abbreviations. BTT, bone tissue type; CAC, canal connection; CCB, coarse cancellous bone; CIC, circumferential osteonal canal; DYO, dynamic osteogenesis; END, endosteal surface; EOC, osteonal canal opening on endosteal layer; ERR, erosional room; FBC, fibrolamellar bone complex; ICL, inner circumferential layer; LAG, line of arrested growth; LBT, lamellar bone tissue; LOC, longitudinal osteonal canals; MEB, medullary bone; MEC, medullary cavity; MID, mid-cortex; OBC, oblique osteonal canal; OCL, outer circumferential layer; OSC, osteonal canal; OSL, osteocyte lacuna; OST, osteon; PBT, non-lamellar parallel-fibred bone tissue; PER, periosteal surface; PFB, parallel-fibred bone; PGM, post-hatching growth marks; POC, osteonal canal opening on periosteal layer; PRC, primary canals; RAC, radial osteonal canal; RSL, resorption line; SEC, secondary canals; STO, static osteogenesis; TBC, total bone cortex; TVN, total vascular network; TTS, total transverse section cortex; VAN, vascular network; WBT, woven bone tissue; WPC, woven-parallel complex.

3.1.1 Bone cortex

The outline shape of the cut out samples of Gansus yumenensis IVPP V21692 and V21693 varies considerably, as well as that of the medullary cavity (Table S1). Small openings formed from old canals link the medullary cavity to the outer surface, or erosional rooms, are present in four of the five samples (Table S1). The tibiotarsus is the exception, as the bone was crushed, the lateral quadrant is absent and the total bone cortex affected by its state of preservation (Figures 8 and 9; Table S1). In general, these erosional rooms are observed either exclusively towards the proximal extremity (e.g. metacarpal II and radius) or the distal extremity (e.g. metacarpal III) (Table S1). The tarsometatarsus exhibits a small oval-shaped erosional room (Figures 10b and 11c) that runs practically along the full length of the sample (Table S1), although it is sometimes eroded due to preservation.

When looking at the ratios of the medullary cavity and the total bone cortex compared to the total sample estimated from the measurements (Tables S2 and S3), the total bone cortex represents the majority of the sample. In the case of the volume values for the complete sample, the medullary cavity ranges from 27% to 41% compared to the 58 to 100% of the total bone cortex (Figure 12a; Table S4). In the case of the surface area values of the total region of interest, the medullary cavity ranges from 13% to 41% compared to the 62 to 99% of the total bone cortex (Figure 12b; Table S6). The remaining percentages for the volume and surface area ratios belong to the total vascular network (Figure 12a,b; Tables S5 and S6).

Two layers are observed at various degrees throughout the total bone cortex of the samples of Gansus yumenensis: an ICL and a mid-cortex (Figures 2a, 4a, 6a, 8a and 10a). Both layers are generally separated by a line of resorption, whose visibility varies from sample to sample (Table S1). In most cases, the line of resorption is accentuated by cracks in which the sediments found in the medullary cavity have also infiltrated (Figures 2a, 4a, 6a, 8a and 10a). No OCL is observed throughout the samples. However, along the outer margin of the total bone cortex in the radius and tarsometatarsus samples, some osteohistological features (i.e. vascular network, organisation of the osteocyte lacunae) seem to indicate the onset of this outermost bone layer. As such, this shows that both Gansus specimens had not reached full ontogenetic maturity at their time of death, as observed in other non-ornithurine Euornithes (O’Connor et al., 2015; Wang, Hao, et al., 2019; Wang, Cau, et al., 2020), and most likely were late subadults. The mid-cortex generally makes up around 2/3 of the total bone cortex (Tables S3, S5 and S6). The ICL observed in the various samples is generally avascular, yet some radial- and oblique-oriented canals (generally under 20 total per sample) traverse and open on the endosteal surface (Figures 2b, 3c,d, 4b, 5c,d, 6b, 7c,d, 8b, 9c,d, 10c and 11c,d).

In general, the total bone cortex is relatively thin in the samples, never exceeding 2.00 mm as observed in several extant birds exhibiting a similar lifestyle (e.g. mallards, grebes, terns; Table S7) (de Margerie et al., 2002; de Margerie, 2002; de Margerie et al., 2005; Habib & Ruff, 2008). The total bone cortex is thickest in the anterior quadrant (between 0.73 and 1.99 mm) and/or the posterior quadrant (and between 1.27 and 2.66 mm) throughout the samples (Table S3). Regarding the mediolateral axis, the thickness of the total bone cortex varies between 0.59 and 2.23 mm in the lateral quadrant, and between 0.54 and 1.72 mm in the medial quadrant (Table S3). The tibiotarsus is an exception due to its incomplete and distorted preservation (Figures 1f and 8). The ICL and mid-cortex exhibit the same trend (Table S3). When comparing the zeugopodium and autopodium elements between IVPP V21692 and V21693, we observe that there is a contrast in the thickness of the sampled midshafts. The total bone cortex of the radius (between 1.34 and 1.59 mm) is thicker than that of the tibiotarsus (between 1.27 and 1.60 mm; Table S3). The opposite is observed for the autopodial elements. For the metacarpals, the thickness varies between 1.29 and 1.34 mm for the second metacarpal and between 0.84 and 0.90 mm for the third metacarpal (Table S3). In the case of the tarsometatarsus, the thickness varies between 2.00 and 2.04 mm (Table S3).

Throughout the total bone cortex of the samples, no post-hatching growth marks are observed, may they be LAGs or annuli (Figures 2, 4, 6, 8 and 10). However, this could be due to the fact that the specimens IVPP V21692 and V21693 are immature, having not reached full ontogenetic maturity at their time of death despite exhibiting a potential onset of an OCL in the radius and tarsometatarsus sample (Figures 6 and 10; Table S1).

3.1.2 Vascular network

The total vascular network, although appearing abundant in the samples (Figures 2-11), represents a small part of the total sample of each bone (Figure 12a,b; Tables S5 and S6). However, in the case of the samples of IVPP V21693, the conservation of the total vascular network was affected by the preservation of the bone, either partially in some quadrants (e.g. tarsometatarsus on medial–posterior limit and middle of the lateral quadrant) (Figures 10 and 11) or almost completely (tibiotarsus) (Figures 8 and 9). Because of this, despite the resolution of the scans, the canals were very hardly visible or altogether absent (Table S6). The canals are mainly observed within the mid-cortex (Figures 2b, 3c,d, 4b, 5c,d, 6b, 7c,d, 8b, 9c,d, 10b and 11c,d), with some that open on the inner surface (generally under 20 in total) and the outer surface (generally under 10 in total). Regarding the surface area ratios (Table S6), the total vascular network varies between 4% and 7% in the region of interest.

The main orientation of the vascular network in the bone cortex of the samples of Gansus yumenensis IVPP V21692 and V21963 is longitudinal (Figures 2-11; Tables S1 and S4). The dimensions (i.e. length and width) of these longitudinal canals in transverse sections vary considerably, from 0.02 × 0.02 mm to 0.30 × 0.10 mm (Table S4). The highest number of canals, as well as the largest ones, are observed and measured in the tarsometatarsus (Table S4). Longitudinal canals, the length of which exceeds 20 mm in cross section, are few in number. Regarding the ratios based on the surface area values of the longitudinal canals compared to the total vascular network in the region of interest, the highest ratio value estimated with certainty is found in the tarsometatarsus, for a value of 77% (Figure 12c; Table S6). The tibiotarsus is an exception, with an estimated ratio value of 88%. However, this value is affected by the preservation of the bone as well as the number of canals measured with certainty (Figure 12c; Table S6).

Regarding the canal connections and other non-longitudinal canals, the most abundant connections throughout the samples have oblique-oriented, followed by circumferential-oriented canals (Tables S1 and S4). Radial canals are generally few in number, never exceeding more than eight throughout the analysed slices, and generally serve as connections between the total vascular network and the medullary cavity (Tables S1 and S4). When we look at the surface area values (Table S6), we observe that the oblique canals have the highest values throughout the total bone cortex (between 0.585 and 1.605 mm²), while the circumferential canals exhibit the lowest values (between 0.112 and 1.096 mm²). The ratio values for the non-longitudinal canals compared to the total vascular network follow the same trend (Table S6). When we look at the distribution of the canal connections and other non-longitudinal canals in the regions of interest (Figure 12c), clear differences between the fore- and hindlimbs are observed. In the metacarpals, radial canals exhibit the highest ratio value compared to the total vascular network. This trend is not observed in the tarsometatarsus, where oblique-oriented canals exhibit the highest ratio value. Overall, we observe higher values for the estimated ratios of the vascular canals in the tarsometatarsus than in the metacarpals. The ratio values of the tarsometatarsus exhibit a similar distribution to what is observed for the radius (Figure 12c). Even if little to no circumferential canals are observed in the tibiotarsus (Figures 8 and 9; Table S4), the distribution of the ratio values for the non-longitudinal canals compared to the total vascular network is more reminiscent of the distribution observed for the canals of the second metacarpal than that of the third metacarpal (Figure 12c).

The highest surface values of canal connections are observed in the bones of the forelimb samples of IVPP V21692 compared to those in the hindlimb samples of IVPP V21693 (Figure 12c; Table S6). The radius sample exhibits the highest values for both the surface area and the ratio of the canal connections compared to the total vascular network, with 4.650 mm² for a ratio of 22% (Figure 12c; Table S6). The lowest values are found in the tibiotarsus (Table S6). However, these values can be explained by the effects of preservation on both the bone and the total vascular network. As the total vascular network is mainly located within the mid-cortex of the bone samples, there are little to no canal connections observed within or bordering the ICL (Figures 3c,d, 5c,d, 7c,d, 9c,d and 11c,d; Table S6).

3.1.3 Osteocyte lacunae

Regarding the osteocyte lacunae throughout the samples of Gansus yumenensis IVPP V21692 and V21693, more similarities than differences are observed (Table S1). The lacunae in the ICL are in general circular to ovoid in transverse section (Figures 2b, 4b, 6b, 8b and 10b). In the 3D-rendered models (Figures 3e, 5e, 7e, 9e and 11e), their overall shapes are elongated oval along the proximodistal axis. The osteocyte lacunae in the ICL exhibit a parallel organisation to the circumference of the total bone cortex, with highly variable distance between each lacuna (Figures 2b, 3e, 4b, 5e, 6b, 7e, 8b, 9e, 10b and 11e).

In the mid-cortex, the shapes of the osteocyte lacunae are generally more circular than ovoid in transverse section compared to the ICL (Figures 2b, 4b, 6b, 8b and 10b). The main shape of the lacunae in 3D is also elongated oval along the proximodistal axis, with some small globular lacunae (Figures 3e, 5e, 7e, 9e and 11e). In some bone samples, some of the lacunae are more elongated along the proximodistal axis than others, as well as more flattened along the inner to the outer margin of the total bone cortex. In the metacarpals, some of the osteocyte lacunae in the mid-cortex are elongated along the circumference of the bone in transverse section (Figures 3e and 5e), an orientation which is rarely or not observed in the other samples. Compared to the ICL, the lacunae of the mid-cortex exhibit two types of organisation. The first is generally parallel organisation along the circumference of the total bone cortex, as observed in the ICL. However, the distribution of the lacunae becomes more random when they surround the canals. For some of the longitudinal canals, they form a circle that is centred on the canal itself, a clear indication of an osteon (Figures 2b, 4b, 6b, 8b and 10b). Although no OCL is formed throughout the analysed samples of Gansus yumenensis, we observe along the periosteal surface of the radius and tarsometatarsus samples that the osteocyte lacunae start to exhibit a less random organisation than what is observed throughout the mid-cortex (Figures 6 and 10). As such, it could suggest the onset of an OCL. If so, this would indicate that the two Gansus specimens had not reached full somatic maturity, but were in the late subadult stage at their time of death.

Regarding their general distribution throughout the mid-cortex, the osteocyte lacunae exhibit a more or less even distribution through the total bone cortex, with little to no decrease in density from the inner to the outer margin (Figures 3e, 5e, 7e, 9e and 11e). The tibiotarsus is an exception (Figure 9e), but the apparent decrease can be explained by the state of preservation of the sampled midshaft (Figures 8 and 9).

3.1.4 Bone tissue type

From our analysis of the osteohistological features (Table 1; Table S1), we determine the bone tissue type in the ICL and mid-cortex layers of the samples of Gansus yumenensis based on the nomenclature and determination methods established in previous studies, using the formation principles of the total bone cortex to determine the bone tissue type (Marotti, 2010; Prondvai et al., 2014; Stein & Prondvai, 2014). We cannot use the birefringence of collagen fibres and apatite crystallites as typically done in other studies that use ground sections and polarised light microscopy (Prondvai et al., 2014).

Throughout the samples, the ICL is built on the mid-cortex, which served as a pre-existing firm surface (Figure 13). This layer is also characterised by little to no observed vascular canals. The osteocyte lacunae of the nearly avascular ICL exhibit a parallel organisation along the circumference of the samples, as well as uniform orientation along the long axis of the samples (Figures 3e, 5e, 7e, 9e and 11e). This organisation is indicative of dynamic osteogenesis (sensu Marotti, 2010; Prondvai et al., 2014; Stein & Prondvai, 2014) (Figure 13). From these elements, we can define the bone tissue type of the ICL as most likely being non-lamellar parallel-fibred, exhibiting some slight erosion due to medullary resorption (Figures 2-11). However, we cannot overlook the possibility that the ICL may exhibit lamellar bone tissue, a specific form of parallel-fibred bone tissue (Prondvai et al., 2014; Stein & Prondvai, 2014), which would require further description of the collagen fibres and the orientation of the lamellae through other means than synchrotron analysis and virtual models (i.e. physical cross sections, see Notes on synchrotron analysis).

For the mid-cortex, similar observations can be made as in the ICL, yet associated with other characteristics that indicate a different bone tissue type. Part of the matrix is built on a pre-existing surface surrounding the osteons (Figure 13). The osteocyte lacunae surrounding the osteonal canals exhibit a mainly parallel organisation along the circumference of the total bone cortex, with a uniform orientation along the long axis of the samples (Figures 3e, 5e, 7e, 9e and 11e), both characteristic of dynamic osteogenesis (Marotti, 2010; Prondvai et al., 2014; Stein & Prondvai, 2014) (Figure 13). Therefore, part of the bone tissue in the mid-cortex is parallel-fibred. In between the osteonal canals, the organisation of the osteocyte lacunae becomes more random both along the circumference and the long axis of the samples (Figures 2b, 3e, 4b, 5e, 6b, 7e, 8b, 9e, 10b and 11e), indicative of static osteogenesis and therefore of woven bone tissue (sensu Marotti, 2010; Prondvai et al., 2014; Stein & Prondvai, 2014) (Figure 13). The presence of the two types of bone tissue in the mid-cortex form a woven-parallel complex associated with a dense total vascular network of primary osteons, which can therefore be characterised as a fibrolamellar complex (sensu Prondvai et al., 2014) (Figure 13). The slow decrease in osteocyte lacunae density towards the outer surface of the bone cortex, more explicit in the radius and tarsometatarsus samples, could be indicative of a decrease in bone deposition prior to ontogenetic maturity, and the onset of the formation of an OCL as observed in other non-ornithurine euornithians (O’Connor et al., 2015; Wang, Hao, et al., 2019).

3.2.1 Bone tissue type

When comparing the bone tissue type of the extinct and extant specimens used in this study, several similarities are observed in Gansus yumenensis with the crownward bird species (Wilson & Chin, 2014; Brusatte et al., 2015; O’Connor et al., 2015; Bailleul et al., 2019; Wang, Hao, et al., 2019; Wang, Cau, et al., 2020), as well as with the stem birds found as of the basal pygostylians (de Ricqlès et al., 2003; Chinsamy, Marugán-Lobón, et al., 2020), potentially the Jeholornithiformes (O’Connor et al., 2014; Wang, Huang, et al., 2020). The bone tissue type observed in the mid-cortex of the crownward bird species as well as the Enantiornithes is a fibrolamellar complex, which is surrounded by a nearly to completely avascular and generally thin ICL and OCL (when present) of non-lamellar parallel-fibred or lamellar bone tissue (Figure 14; Table 1; Table S7). These various bone tissue types are characterised by the shape (e.g. globular, ovoid, long oval along the proximal-distal axis) and orientation (e.g. parallel organisation in parallel-fibred bone, random organisation in woven bone) of the osteocyte lacunae, as well as the density of the total vascular network (i.e. nearly avascular in parallel-fibred/lamellar bone tissue, dense neurovascular network in woven bone tissue/fibrolamellar complex) and the long-range fibre orientation (Table 1; Figures 13 and 14c; Table S7) (Prondvai et al., 2014).

Only a few comparative specimens are questionable on the bone tissue type, such as Archaeopteryx lithographica BSPG 1999 I 50 (Erickson et al., 2009), Archaeopteryx albersdoerferi SNSB BSPG VN-2010/1 (Kundrát et al., 2019) and Archaeorhynchus spathula IVPP V20312 (Wang & Zhou, 2017). In the case of the stem long-tailed avialan, either the conclusions made from the analysis are questionable due to the sample used (i.e. minute fragments for Archaeopteryx lithographica BSPG 1999 I 50), or no true conclusions can be made because of the ontogenetic stage of the studied specimen (i.e. juvenile, so unclear if we are observing only woven bone in transverse section or the onset of a fibrolamellar complex). In several specimens, such as the long-tailed avialan Jeholornis sp. STM 2–51 (O’Connor et al., 2014), the enantiornithines Mirusavis parvus IVPP V18692 (Wang, Hao, et al., 2019) and Mirarce eatoni UCMP 139500 (Atterholt et al., 2021), and the non-ornithurine euornithians Archaeorhynchus spathula IVPP V20312 (Wang & Zhou, 2017) and Hollanda luceria MPC-b100/205 (Bell et al., 2010), the mid-cortex exhibits a mixture of both parallel-fibred and woven bone tissue (Table 1; Table S7), forming a structure reminiscent of a woven-parallel complex (Figure 13) (Prondvai et al., 2014). As this structure is located only in the mid-cortex (not within or drifting from the medullary cavity), yet no true dense vascular network is observed in the cross sections of the samples (Figure 13), we suggest that these species potentially exhibited the onset or a form of a fibrolamellar bone complex prior to the Neornithes, which is characterised by a generally dense and well-vascularised fibrolamellar bone complex in the mid-cortex (Table 1; Table S7).

The changes in bone tissue types in the various extinct and extant specimens of this study could be indicative of a mosaic pattern in the evolution of the bird lineage (Figure 14; Table 1; Table S7). This mosaic pattern is most likely associated with the body size and the ontogenetic stage (generally woven bone tissue in juvenile, parallel-fibred to fibrolamellar bone tissue in subadults and adults) (Figure 14; Table 1; Table S7), as well as physiological (i.e. metabolic activity) and biomechanical adaptations (i.e. muscle insertions in relation to locomotion, overall organisation of the neurovascular network in the bone to relieve stress on the bone). All of these factors are associated with increased flight capabilities from the stem long-tailed avialans to the basal pygostylians before reaching more active flapping flight reminiscent of extant Neornithes as of the extinct Ornithothoraces (Figure 14c) (Tobalske, 2016; Kundrát et al., 2019; Schwarz et al., 2019; Monfroy & Kundrát, 2021), as well as lifestyles of these extinct stem birds (i.e. terrestrial, aerial and/or aquatic) (Habib & Ruff, 2008; Lee & Simons, 2015; McGuire et al., 2020).

3.2.2 Post-hatching growth marks

No post-hatching growth marks are observed throughout the total bone cortex of the samples of Gansus yumenensis. As such, this species is the fourth recorded non-ornithurine euornithian to exhibit uninterrupted bone deposition (Figure 14), along with Iteravis huchzermeyeri (O’Connor et al., 2015), Yanornis martini (Wang, Hao, et al., 2019) and Khinganornis hulunbuirensis (Wang, Cau, et al., 2020). Uninterrupted bone deposition pattern is also observed in almost all extant Neornithes (Figure 14). Some exceptions are observed, such as LAGs in the long bones of the kiwi and moa of New Zealand (Bourdon et al., 2009; Legendre et al., 2014), the recently extinct elephant bird of Madagascar (Hansford & Turvey, 2018; Chinsamy, Angst, et al., 2020), the recently extinct dodo (Angst et al., 2017) and the psittacid Amazona amazonica (de Ricqlès et al., 2001). Uninterrupted growth is also observed in the extinct ornithurines such as Hesperornis regalis (Chinsamy et al., 1998; Wilson & Chin, 2014) and Ichthyornis dispar (Chinsamy et al., 1998) (Figure 14). Post-hatching growth marks (LAGs, annuli or a combination of both) are observed in non-avian dinosaurs such as troodontids (Weishampel et al., 2004; Gao et al., 2012; Shen et al., 2017), as well as in most extinct bird specimens (Figure 14; Table 1; Table S7) such as Jeholornithiformes (O’Connor et al., 2014; Wang, Huang, et al., 2020), basal pygostylians (Chinsamy, Marugán-Lobón, et al., 2020) and Enantiornithes (O’Connor et al., 2014; Wang et al., 2017; Atterholt et al., 2018; Wang, Hao, et al., 2019; Wang, O’Connor, et al., 2019). They are also observed in some non-ornithurine euornithians, such as Archaeorhynchus spathula (Wang & Zhou, 2017), Hollanda luceria (Bell et al., 2010) and Patagopteryx deferrariisi (Chiappe, 2002) (Figure 14).

In the case of the long-tailed avialan Archaeopteryx, may it be A. lithographica BSPG 1999 I 50 (Erickson et al., 2009) and A. albersdoerferi SNSB BSPG VN-2010/1 (Kundrát et al., 2019), no post-hatching growth marks are observed. In the case of A. lithographica BSPG 1999 I 50, the absence of post-natal growth marks is highly questionable, as the analysed samples were minute fragments and therefore are not fully reliable. In the case of A. albersdoerferi SNSB BSPG VN-2010/1, the absence of post-natal growth marks is most likely as the specimen died in its early ontogenetic stage (Kundrát et al., 2019) (Figure 14c; Table S7). Growth marks are observed in the Jeholornithiformes, mainly from the outer margin of the mid-cortex to the OCL (Figure 14c; Table S7). Their high concentrations and distribution (i.e. little space between each line) are reminiscent of an external fundamental system (EFS) (Ponton et al., 2004), which is also observed in the adult samples of Confuciusornis sanctus (Figure 14c; Table 1; Table S7). In the case of the Enantiornithes, post-natal growth marks are also observed, mainly from the outer half of the mid-cortex to the OCL (when present) (Figure 14; Table 1; Table S7). However, compared to the long-tailed avialans and the basal pygostylians, the number of LAGs and/or annuli observed decreases, in many case never exceeding three–four. The spatial distribution between these post-natal growth marks is also greater, so no EFS is observed (Figure 14c; Table 1; Table S7).

From our analyses of the samples of Gansus yumenensis V21692 and V21693, we provide further evidence for the widespread presence of uninterrupted bone deposition in non-ornithurine euornithians during the Early Cretaceous (Figure 14), as well as a progressive loss of post-natal growth marks from the long-tailed avialans to the Euornithes.

4.2.1 Notes on the osteohistological features

In the radius and tarsometatarsus samples of Gansus yumenensis, the high values for the total bone cortex to total sample ratio along with the total vascular network to total sample ratio could indicate adaptations to mechanical stresses during locomotion (Figure 12s; Table S3). The organisation and distribution of the vascular network in the radius are probably associated with mechanical stresses related to flight, while in the tarsometatarsus it could be associated with mechanical stresses related to swimming and foot-propelled diving (Ji et al., 2006; You et al., 2006; Habib & Ruff, 2008). Differences in the total vascular network along the anteroposterior axis are observed between the various samples of Gansus yumenensis, offering further evidence for adaptations to mechanical stresses related to locomotion (aerial, terrestrial and aquatic).

Similar observations are made in the long bones of extant semi-aquatic birds with similar lifestyles to that of Gansus yumenensis, such as mallards (de Margerie et al., 2002; de Margerie, 2002; de Margerie et al., 2005) and grebes (de Margerie et al., 2005) (Table 1; Table S7). Both are characterised as bimodal species with drag-based hindlimb diving as well as flight for locomotion (Habib & Ruff, 2008). Most of the fossil species are compressed during preservation along the mediolateral or anteroposterior axis, rendering comparisons difficult to make. The organisation of the total vascular network appears to be more or less uniformly distributed within the samples of the Ornithurae specimens, specifically within Hesperornis regalis YPM 1491 (Wilson & Chin, 2014) and the mallard Anas platyrhynchos (de Margerie, 2002; de Margerie et al., 2005). This type of vascular distribution in the bones of the hindlimbs can be considered as an adaptation to swimming and foot-propelled diving (Habib & Ruff, 2008; Habib, 2010), although no comparative quantitative correlations between the density and distribution of the canals in long bones with locomotion types have been made to date.

Further analyses of midshaft samples from long bones of extant birds and other flying amniotes are required to further elaborate and define adaptations to mechanical stresses, as done in previous studies (Lee & Simons, 2015; McGuire et al., 2020), specifically with the use of in-depth observations and volume analyses of 3D-rendered models from which we could try to divine the evolution of these locomotor adaptations in the extinct bird lineages, as well as better understand the flight capability of the various bird species (i.e. more adapted for gliding or true flapping flight).

4.2.2 Notes on the vascular network

The predominant canal orientation in Gansus yumenensis is longitudinal, associated with different quantities and densities of the non-longitudinal canals exhibiting various orientation varying from sample to sample (Figure 12c; Table S6). The high canal connections to total vascular network ratio values, associated with the relatively thin total bone cortex, support the hypothesis of high metabolic needs for fast growing bone tissue, as observed in extant birds (Montes et al., 2007; Montes et al., 2010; Werner & Griebeler, 2014; Legendre et al., 2016; Monfroy & Kundrát, 2021).

We also observe in the samples of Gansus yumenensis that the total vascular network exhibits a noticeable decrease from the inner to the outer margin of the total bone cortex (Figures 2-11), an osteohistological feature, which is also observed in other non-ornithurine euornithians (O’Connor et al., 2015; Wang, Hao, et al., 2019; Monfroy & Kundrát, 2021) (Figure 14c; Table 1; Figure S2; Tables S1 and S7). In the case of the extinct and extant Ornithurae, a slight decrease in density from the endosteal surface to the periosteal surface can be observed whether an OCL is present or not, yet the total vascular network is generally more uniformly distributed throughout the total bone cortex compared to the non-ornithurine euornithians (Figure 14c; Table 1; Tables S1 and S7).

The decrease in the density of the total vascular network towards the periosteal surface of the bone cortex suggests a decrease in growth rate in Gansus yumenensis, as observed in other non-ornithurine Euornithes (Figure 14c). However, our comparisons with the other bird specimens must be considered carefully, as some of description made previously could have most likely been affected by the state of preservation of the specimens, as well as the ground preparation of the transverse sections. In the case of our virtual analysis of the samples of Gansus yumenensis IVPP V21692 and V21693, the synchrotron 2D transverse sections are closer to true 2D slice of the bone cortex compared to the 100-micron-thick ground physical sections which can change the appearance of the analysed features.

Further 3D analysis of bone samples from various extant and extinct specimens with virtual models would be required so as to see whether qualitative and quantitative correlations can be established between osteohistological features (e.g. density and orientation factors of the total vascular network) and physiological/biomechanical adaptations (e.g. cortical drift associated with periosteal resorption, biomechanical constraints with muscle attachments, preferential directions of stress and strain) (Enlow, 1962a; Enlow, 1962b; Enlow, 1963; Maggiano et al., 2015). These quantitative correlations could also provide insight regarding biomechanical questions such with the development of powered flapping flight amongst early birds.

4.2.3 Notes on the ontogenetic stage

Throughout the samples of Gansus yumenensis IVPP V21692/V21693, no distinct OCL is observed throughout the total bone cortex. However, the decrease of the vascular network and the change in the organisation of the osteocyte lacunae towards the outer margin of the mid-cortex suggest the onset of an OCL (Chinsamy et al., 1995; Ponton et al., 2004), specifically in the radius and tarsometatarsus samples (Figures 6, 7, 10 and 11; Table S1). Based on these observations, we do not necessarily expect the deposition of future growth marks, in a similar fashion to extant birds. Associated with the presence of a relatively thin ICL along the inner margin of the total bone cortex (Figures 2-11), this indicates that Gansus yumenensis was closed to ontogenetic maturity, and most likely was a late subadult when it died.

This is further supported by other osteohistological analyses performed on other non-ornithurine Euornithes closely related to Gansus yumenensis (Figure 14a,b), such as Iteravis huchzermeyeri (O’Connor et al., 2015), Yanornis martini (Wang, Hao, et al., 2019) and Khinganornis hulunbuirensis (Wang, Cau, et al., 2020). These specimens do not exhibit an OCL, as observed in specimens of Ornithurae (Figure 14c; Table 1), which likely suggests ontogenetic immaturity (Chinsamy et al., 1995; Ponton et al., 2004; O’Connor et al., 2015). No LAGs are observed in these species, and the descriptive analyses suggest that the deposition of future growth marks is not expected, contrary to other non-ornithurine Euornithes from the Cretaceous such as Patagopteryx deferrariisi (Chiappe, 2002; Starck & Chinsamy, 2002), Archaeorhynchus spathula (Zhou et al., 2013) and Hollanda luceria (Bell et al., 2010) (Figure 14c).

Further sampling for osteohistological analyses and comparisons would be required to complete our knowledge and categorise the ontogenetic stages in extinct euornithians.