Deciphering an extreme morphology: bone microarchitecture of the hero shrew backbone (Soricidae: Scutisorex)

1. Introduction

The relationship between morphology and function is a central concept in the study of structural evolution. Connecting an organism's morphology to its biomechanical capabilities is critical to understanding how it interacts with its environment, as well as how organisms adapt to selective pressures and ecological opportunities. Structures with unique or extreme morphologies are problematic in this context because they often lack an obvious function and are not readily comparable to homologues or analogues in other organisms. The problem is especially severe for cryptic or extinct organisms where opportunities for the direct observation of function are limited or non-existent.

The hero shrew has a spinal column unlike that of any other known vertebrate. Most of its dorsal vertebrae, including at least half of the thoracic vertebrae and all lumbar vertebrae, have dorsoventrally expanded lateral processes whose cranial and caudal faces are covered in fingerlike bumps or tubercles (figure 1, [1–4]). These tubercles articulate with those of neighbouring vertebrae, such that the dorsal spine forms a dense column of tightly interlocking bony processes (figure 1). Scutisorex also has more lumbar vertebrae than other shrews; most soricids have five lumbar vertebrae, whereas Scutisorex thori has eight and S. somereni has 10–12 [1,2].

• Download figure
• Open in new tab
• Download PowerPoint

The function of this highly modified spine is enigmatic. Scutisorex has been the subject of intense anatomical inspection and description [1–7], but functional investigations have been inconclusive. Early reports [4] suggested that the robust spine of S. somereni allowed it to bear the weight of an adult human and escape unscathed, but no field or laboratory observations have been made of any behaviours that clearly necessitate its radical morphology. Cullinane & Bertram [8] showed that the intervertebral joints of S. somereni are highly resistant to axial torsion. They also found that the lumbar spine of S. somereni does not behave as a series of viscoelastic joints under compression, but instead as a single rigid bar. Despite observing a captive specimen alive for several months, Cullinane & Aleper [1] were unable to pinpoint an in vivo use for the vertebral column that benefits from this property. Histological work shows that the unusually modified transverse processes of the vertebrae develop endochondrally, precluding interpretations of the morphology as pathological [6].

The discovery of a second species, S. thori [2], provided a critical new data point for understanding the evolution of the Scutisorex spine. Along with having an intermediate number of lumbar vertebrae, S. thori has fewer, larger tubercles on its vertebrae than does S. somereni (figure 1). This discovery did not yield any new functional clues, but it was hypothesized that S. thori might also represent a functional intermediate between a normal spine and the extreme modification found in S. somereni. Stanley et al. [2] suggested that the robust spine might be used in flexion to pry palm leaves away from the main trunk of the plant, giving access to nutritious insect larvae, but this behaviour has not been observed. Instead, S. somereni has been found to eat mostly earthworms and is, therefore, at least a semi-subterranean forager [9], and it resembles other small fossorial mammals in having robust ribs for its body size. Yet, S. somereni does not have unusually stout limbs, and in fact has long bone proportions comparable to those of non-fossorial mammals [7].

Because Scutisorex vertebrae are so radically modified, it is difficult to compare them morphometrically to more typical vertebral columns while also capturing the likely functionally relevant parts of the morphology. We attempt to circumvent this problem by inspecting not only the external morphology of the Scutisorex spine, but also the internal morphology via centrum trabecular bone architecture (TBA). According to a set of principles commonly referred to as ‘Wolff's Law’ [10–13], the structure of trabecular bone can provide quantitative information about the physical loads experienced by a bone in vivo. The relationship between TBA, in vivo loads, and bone function has been studied in a variety of mammals, including primates [14–19], artiodactyls [20–22], rabbits [23,24], squirrels [25,26], bats [27], and murid rodents [28–31]. This work shows that direction, magnitude, and frequency of load are physically manifest in the trabecular structure of the loaded bone, and we use that phenomenon to elucidate trabecular features that might relate to the function of the Scutisorex spine.

Although soricid trabecular bone has been studied in the context of allometric scaling of bone characteristics among mammals [32], no studies of trabecular bone have been conducted exclusively on shrews. In light of this, we have several specific aims: (i) to quantify the internal and external morphology of Scutisorex dorsal vertebrae, and their range of inter- and intraspecific variation, (ii) for comparison, to quantify internal and external morphology of the dorsal spine in the closely related, ‘unmodified’ shrew, Crocidura goliath, which is similar in size to both Scutisorex species, (iii) to determine the degree to which S. thori represents an intermediate between S. somereni and C. goliath, in both TBA metrics and external morphology, (iv) to assess the relationship between TBA variation and external characteristics of the vertebrae, including number of tubercles in Scutisorex, and (v) connect morphology and variation through the column to functional hypotheses and the limited existing behavioural data on Scutisorex.

2. Methods

(a) Specimen selection and micro-computed tomography scanning

We executed X-ray micro-computed tomography (µCT) scans of the thoracic and lumbar vertebrae of 20 specimens: 13 specimens of S. somereni, 3 specimens of S. thori (the total worldwide sample of the taxon currently available in natural history museums), and 4 specimens of C. goliath (table 1). All specimens are from the Field Museum of Natural History (FMNH), and all are adult based on epiphyseal fusion of limb bones. We scanned specimens using the 240 kV tube of the GE v|tome|x µCT scanner at the University of Chicago PaleoCT facility. Because of differences in size and spine orientation, isotropic voxel size ranged from 14 to 26 µm. In some cases, thoracic and lumbar vertebrae were scanned separately; when necessary, we scanned both at the same or similar resolutions to minimize intra-specimen variation in scan quality (electronic supplementary material, table S2). We reconstructed specimens in GE phoenix datos|x and aligned and cropped the resulting image stacks using VGStudio MAX 3.3 (Volume Graphics, 2019).

(b) Data collection: external morphology

To separate vertebral scans, we imported image stacks into ORS Dragonfly 4.0 (Object Research Systems, 2019) and exported each vertebra individually as an stereolithography (STL) mesh. We cleaned each vertebral model and separated it from any remaining pieces of neighbouring vertebrae using Autodesk Meshmixer 3.5.474. Using the ‘Units/Dimensions’ function in Meshmixer, we took four linear measurements on each thoracic and lumbar vertebra of all three included species (electronic supplementary material, figure S8). In the two species of Scutisorex, we also quantified the number of tubercles projecting from the cranial and caudal surfaces of each vertebra. To standardize and record the number of tubercles present, we opened the STL of the vertebra in Meshmixer in orthographic view, aligned the model with the craniocaudal axis of the vertebral centrum perpendicular to the plane of the screen and took screenshots of the cranial and caudal faces of the specimen. We divided the image of the vertebra into four quadrants, with a horizontal line tangent to the top of the centrum delineating dorsal and ventral sections, and a vertical line at the mediolateral midpoint of the centrum and neural arch (electronic supplementary material, figure S8). We recorded the number of tubercles in each quadrant, with a tubercle defined as a relatively small-diameter peglike or bumplike feature with a local maximum in the cranial or caudal direction on the vertebral surface, not including zygapophyses. We specify ‘relatively small-diameter’ because we excluded local maxima on the edges of the vertebral centrum (i.e. the cranial and caudal extents of the centrum itself). We also excluded transverse processes where they are clearly identifiable. Three-dimensional models of all Scutisorex thoracic and lumbar vertebrae are available for download from Morphosource (www.morphosource.org, Project #P956).

(c) Data collection: trabecular bone architecture

For our analysis of the TBA of each vertebral centrum, we sought to maximize the volume of trabecular bone we used while (i) excluding all cortical bone, (ii) as much as possible, sampling the same relative proportion of the total bone structure in each vertebra and across species, and (iii) analysing regions of interest (ROIs) from the same location in each vertebra. To keep our sampling as standardized as possible, we first aligned the vertebra in Dragonfly so that the dorsal side of the vertebral centrum paralleled the z-axis; the dorsoventral axis of the centrum paralleled the x-axis; and the mediolateral axis of the vertebra paralleled the y-axis. We then selected the largest square prism possible inside the centrum while including only trabecular bone. The resulting ROIs typically measured 0.6–1.2 mm per side (mostly 0.75–0.85 mm), and 1.0–3.25 mm in craniocaudal length.

We exported TIFF stacks of our hexahedral ROIs and executed binary thresholding in Fiji [33]. We acknowledge that the small size of the structures we are attempting to detect makes choice of segmentation method critical to the validity of our results, and we therefore, tested three segmenting methods to determine which would provide the most accurate binary renderings of our broader dataset. These tests are detailed in the electronic supplementary material, Section A, and figures S1–S7. Based on results of these tests, we chose to use Otsu segmenting [34] for our main body of analyses. Although Otsu segmentation could potentially result in slight underestimation of bone volume fraction (BVF) and trabecular thickness (Tb.Th), and slight overestimation of dispersal of trabecular direction away from the craniocaudal axis, it is effective at maintaining thin trabecular structures during segmenting, and less subject to user subjectivity and overestimation of trabecular parameters than global grey value threshold segmenting. Furthermore, relative resolutions (mean Tb.Th divided by scan resolution, resulting in pixels per trabecular structure; [35]) for all specimens in our study are between 1.52 and 5.48 pixels/trabecula, with an average of 3.00 px/tb. This value is not unreasonably low compared to that of other studies (see electronic supplementary material). Therefore, any error is likely to be systematic, and similar in magnitude across specimens, and will probably not affect our interpretation of bone qualities in the species relative to one another. Otsu threshold is determined using the grey value histogram for the entire stack of images to which it is applied; stacks with high and low BVF have inherently different histogram parameters, and will, therefore, be thresholded slightly differently by the Otsu procedure. However, the difference in attenuation between bone and air is large, providing a clear distinction between bone and background, and we therefore, do not expect the difference in histograms to have a notable effect on our results. We applied Otsu thresholding via the Fiji plugin Auto Threshold, v. 1.16.5 (https://github.com/fiji/Auto_Threshold) [36]. ROI image stacks, both pre- and post-Otsu thresholding, and results of alternate segmenting methods used in supplementary analyses are available on Morphosource (www.morphosource.org, Project #P956); raw data for sensitivity tests are available on the Dryad Digital Repository (https://doi.org/10.5061/dryad.8w9ghx3hf).

We calculated four structural metrics for our trabecular bone ROIs: BVF, trabecular number (Tb.N), Tb.Th, and primary trabecular orientation in the form of trend and plunge relative to the transverse plane of the vertebra. We conducted these analyses on thresholded ROI stacks in Quant3D [37,38] with a user-defined threshold of 127–255; 2049 uniform rotations; dense vectors and random rotations turned on; omit side intersecting paths turned on; and star volume and star length distributions calculated with 2000 points. We also used a centred sphere (figure 2) as the volume of interest for these calculations, to eliminate edge-related artefacts. To visualize primary trabecular orientation, we plotted trend and plunge for each vertebra using Stereonet 10 [39,40]. Raw data for these analyses are available on the Dryad Digital Repository (https://doi.org/10.5061/dryad.8w9ghx3hf).

• Download figure
• Open in new tab
• Download PowerPoint

Because our segmenting protocol is automated, it reduces user-introduced error in segmenting and subsequent error in TBA metrics. Yet, there is still a degree of uncertainty in our reported values, because slight variations in threshold can cause variation in TBA results. To estimate the uncertainty in our results, we used a subset of specimens to determine the mean deviation from our reported values when we increased or decreased the automatic Otsu threshold by 10%. We used these values to construct ranges of uncertainty for BVF, Tb.Th, and Tb.N, with the size of the range customized to reflect the volatility of the metric in response to changing threshold (electronic supplementary material, table S1). Details of this procedure are presented in the electronic supplementary material, Section B; and the image stacks with increased and decreased threshold used to determine uncertainty are available on Morphosource (Project #P956). Uncertainty ranges are applied to the species mean line for each species across all three metrics in figure 5, and are reported numerically in a separate tab of the supplementary data on the Dryad Digital Repository.

It is important to note that although our ROIs are smaller than those examined in larger animals such as primates [14,15] and sciurid rodents [25], and in most cases do not fulfil the guidelines for applying the assumption of trabecular bone as a continuous material ([41]; electronic supplementary material, table S3), they nonetheless represent the largest possible standard sample of trabecular bone available in the vertebrae of our species of interest. Shrews are among the smallest mammals on Earth, [31] and the relatively small space within a shrew vertebral body limits the degree to which trabeculae can interact with one another rather than the cortical shell of the vertebra. This results in a structure that likely behaves less like a cortical shell filled with a cellular solid (e.g. larger mammalian bones; [42,43]), and more like a cortical shell supported by a series of internal struts (e.g. bird forelimb bones [44]). However, the metrics we collected on trabeculae and relative bone density remain relevant to understanding how shrew bones withstand force, including down to the level of individual trabeculae [45], whose shape and capability to resist bending can also be related to BVF [42].

(d) Full-column analysis

All analyses were performed in R v. 3.6.1 [46]. Trends in trabecular bone metrics were compared using goodness-of-fit for polynomial lines, estimated using the stats function lm() and polynomial degrees 1–9. For each metric (e.g. BVF), a polynomial line was fitted to the species mean for that metric through the thoracolumbar column (see electronic supplementary material, figure S10). The purpose of this line fitting was to determine the similarity in the overall pattern of change through the column among the three species. Model fit was assessed with the Akaike information criterion (AIC) and Bayesian information criterion (BIC) from the R package stats, and corrected AIC (AICc) from the package sme [47]. Because we had so few specimens of S. thori (N = 3, which nonetheless represents the entire known sample of museum specimens worldwide), we used a custom function to subsample sets of 3 S. somereni specimens (without replacement) from our total of 8, and calculate the resulting goodness-of-fit for polynomials from degree 1 to 9. We replicated this sampling 100 times to determine if the originally chosen best-fit polynomial and the differences in polynomial fits among species were robust to a small sample size.

3. Results

(a) External morphology

The total number of dorsal vertebrae ranges from 23 to 25 in S. somereni, holds at 22 for all specimens of S. thori, and ranges from 19 to 20 in C. goliath (table 1). Although C. goliath has fewer vertebrae than either species of Scutisorex, it has a similar total column length (using cumulative centrum length as a proxy for total length) because of its relatively longer caudal thoracic and lumbar vertebrae (figure 3a). Centrum width and height show considerable overlap among the three species in the cranial thoracic column, but all show peak thoracic width at slightly different vertebral positions. Caudal thoracic and lumbar centra in C. goliath also tend to be dorsoventrally taller than those of either Scutisorex species, with the exception of the caudalmost positions in the latter (L6–12; figure 3b). This mirrors qualitative observations of the dorsoventrally compressed, oval cross-section of Scutisorex centra.

• Download figure
• Open in new tab
• Download PowerPoint

Overall tubercle counts are higher on average in S. somereni than S. thori (figure 4). Total tubercle count follows a similar trajectory in both species, increasing from vertebral position 1 to position 15–17 (usually corresponding to a position between L1 and L3) and then decreasing caudally at a similar or slightly faster rate (figure 4a). Tubercles are more abundant on the caudal faces of the vertebrae than the cranial faces, but show similar patterns of increase and decrease through the column, largely mirroring the pattern of increase and decrease in total tubercle count (figure 4a; electronic supplementary material, figure S9). By contrast, dorsal and ventral tubercle counts show markedly different trajectories of change through the thoracic and lumbar spine (figure 4b). Dorsal tubercles appear around vertebral position 3–5 and peak at position 11 (T11). Ventral tubercles do not appear until position 9–10, and peak higher than the maximum for dorsal tubercles, around position 15–17 (approx. L1–L3). The number of tubercles on both the dorsal and ventral sides of the vertebrae drops off precipitously in the last few lumbar positions. This region functions as the hinge between the lumbar spine and pelvis and serves as an attachment point for increased musculature surrounding the lumbosacral joint and hind limbs (e.g. psoas major lateralis and psoas minor, [1]; figs 3 and 5). Despite the increased flexibility in this region, especially at the lumbosacral joint, there are still sufficient tubercles present on the caudalmost lumbar vertebrae to stop nearly all axial torsion at intervertebral joints.

• Download figure
• Open in new tab
• Download PowerPoint

(b) Internal morphology: trabecular bone architecture, interspecific differences, and polynomial model fitting

TBA metrics vary differently across the dorsal vertebral column, and their patterns also differ among species. BVF in both Scutisorex species decreases in the first five vertebral positions before undergoing a slight increase and then remaining relatively constant from position 10 onward (figure 5a). Although BVF follows similar trajectories in both species, and the two have some overlapping values, as well as overlapping areas of uncertainty, overall BVF is higher in S. somereni than in S. thori. In C. goliath, BVF generally decreases through the whole column, and BVF values in C. goliath are lower than those of either species of Scutisorex, such that there is almost no range overlap between C. goliath and S. thori from position 7 onward, and nearly no overlap in uncertainty around the species means. Tb.N generally decreases through the column in C. goliath and S. thori, whereas in S. somereni, Tb.N increases from position 4 to 14, followed by a decrease to the end of the column (figure 5b). In terms of absolute value, means for C. goliath and S. thori are quite similar to one another through the whole column, but S. somereni has consistently higher mean Tb.N values from position 9 onward, with a local maximum at position 14, just one position cranial to the peak in tubercle number in that species (figure 4a). Tb.Th is higher in Scutisorex than in C. goliath (figure 5c), especially in the more caudal thoracic and lumbar vertebrae. The mean Tb.Th also mostly decreases through the vertebral column in C. goliath, whereas it increases in S. thori and S. somereni.

• Download figure
• Open in new tab
• Download PowerPoint

Polynomial line fits were chosen using BIC because it made the most consistent choices for the subsampled S. somereni data (electronic supplementary material, table S4). Because BIC exacts a stronger penalty than AIC for additional model parameters, it consistently chose the same or lower-degree polynomial models (table 2; electronic supplementary material, table S5), and is therefore a more conservative choice than AIC or AICc with our small sample size. According to BIC, C. goliath and S. thori exhibit more similar patterns in BVF through the dorsal vertebral column, whereas for Tb.N, all three species have slightly different patterns: second degree for S. thori, first degree for C. goliath, and fourth degree for S. somereni. Crocidura goliath and S. somereni follow a linear trajectory for Tb.Th, whereas S. thori has a more complex pattern (figure 5; electronic supplementary material, figure S10). Chosen polynomial models for S. somereni are robust to subsampling combinations of three specimens (electronic supplementary material, table S4), indicating that differences among the three species are unlikely to stem from insufficient sampling.

Primary trabecular direction is largely aligned with the craniocaudal axis of the vertebral centrum in both thoracic and lumbar vertebrae of S. somereni and S. thori, and, to a lesser degree, the thoracic vertebrae of C. goliath (figure 6). Trabecular orientation in the lumbar vertebrae of C. goliath is much more irregular, and has no strong association with anatomical directions (figure 6).

• Download figure
• Open in new tab
• Download PowerPoint

4. Discussion

(a) Evidence of elevated axial compressive load in Scutisorex

The TBA of the centra of both Scutisorex species indicates that they are subjected to greater axial compressive loads than the vertebrae of C. goliath. However, as with its external morphology ([2]; this study), S. thori is intermediate in trabecular morphology between S. somereni and C. goliath in several respects. Both Scutisorex species have higher BVF than C. goliath through almost the entire dorsal vertebral series, but the mean values for S. thori are between those of S. somereni and C. goliath, and overall trends in BVF through the column are more similar between S. thori and C. goliath (figure 5a and table 2). BVF is an effective predictor of Young's modulus and strength in human trabecular bone, with higher BVF generally corresponding to stronger and stiffer bone [48–50]. Nevertheless, BVF does not fully describe the structure of trabecular bone, and is a better predictor of mechanical properties when considered with other microstructural characteristics [48]. Besides having higher BVF than C. goliath, both Scutisorex species have higher Tb.Th (but with slightly different trends through the column; table 2). Higher Tb.Th also correlates to higher compressive strength in the lumbar spine of humans [51]. Further, increased cyclical bone loading has been shown to result in increased BVF and Tb.Th in mouse vertebrae [52], an indication that the higher values in Scutisorex likely relate to more frequent, higher-magnitude loads. In contrast with the gradient in BVF (lowest in C. goliath, moderate in S. thori, highest in S. somereni), the pattern in Tb.Th is much more similar in the two species of Scutisorex, with C. goliath alone having a much lower average through the column (figure 5c). This is an even more intriguing pattern in light of pattern of change in Tb.N (figure 5b and table 2). Scutisorex thori and C. goliath change through the dorsal column with almost identical values and patterns, whereas S. somereni has higher Tb.N than both S. thori and C. goliath, especially in the caudal thoracic and cranial lumbar positions. This pattern is of special interest because a given value for BVF can be achieved with several different configurations of trabecular structure (e.g. fewer but thicker trabeculae versus more numerous but thinner trabeculae). Two- and three-dimensional finite-element models have shown that a change in BVF accomplished primarily by altering Tb.N has a greater effect on Young's modulus and bone strength than the same change in BVF accomplished via alteration of Tb.Th [53–55]. As such, despite its intermediate BVF, S. thori could have trabecular bone mechanical properties slightly closer to those of C. goliath because its Tb.N is closer to that of C. goliath. Likewise, because change in Tb.N has a greater effect on the mechanical properties of trabecular bone than change in Tb.Th [54,55], if S. thori exhibited the same BVF, but a Tb.N closer to that of S. somereni and Tb.Th more similar to that of C. goliath, we might expect its trabecular bone to have mechanical properties more similar to S. somereni.

Based on the findings outlined above, we expect that the vertebral trabecular bone of S. somereni and S. thori are both stronger and stiffer under in vivo forces than that of C. goliath. However, the anisotropic qualities of bone must be accounted for when considering loading schemes. Anisotropic materials behave differently when forces are applied from different directions. In the vertebrae of a quadruped mammal, the primary load is craniocaudal compression, which allows the animal to maintain its posture [56]. To withstand deformation under this load, we might expect the primary direction of trabecular elements in the vertebral centrum to be along the craniocaudal axis [57]. We found that the primary trabecular orientation in Scutisorex is much more closely aligned with the craniocaudal axis of the vertebral centrum than in C. goliath (figure 6). This could be due in part to the low BVF values and small ROI size in C. goliath: some ROIs capture only a single trabecular element, making assessment of the continuous trabecular structure difficult [41]. However, if low BVF is the reason for the difference in measured orientation, we might expect that S. thori, which has an intermediate BVF, to have an intermediate tendency for craniocaudal trabecular orientation. This does not appear to be the case (figure 6). It is possible that the interlocking tubercles on the vertebrae of Scutisorex spp. also play a role in the signal detected here, restricting the primary trabecular direction through restriction of the overall intervertebral range of motion. Further in vitro or in silico tests will be required to determine the degree to which range of motion of an intervertebral joint, and craniocaudal alignment of trabeculae in its component vertebrae, are correlated.

(b) Whole-bone morphology and adaptations for stereotyped spinal flexion

Scutisorex somereni has a number of specialized muscles for fine control of axial compression in the spine, including the m. intertubercularis ([1], fig. 2), which connects tubercles on adjacent vertebrae, and the m. tuberculoventralis, which connects the ventral tubercles of a vertebra to the ventral side of the centrum on a more cranial vertebra [1]. Additionally, under compression, S. somereni intervertebral joints lose their viscoelastic qualities and begin to behave mechanically as an immobile, solid unit [8]. If the joint bends, it does so almost exclusively in dorsoventral flexion. This restriction of intervertebral motion to flexion in the sagittal plane relies on robust hypaxial musculature [1] and a suite of whole-bone characteristics that resist bending, torsion, and extension.

In the craniocaudal view, the vertebrae with the largest number of tubercles (caudal thoracic through cranial lumbar positions) resemble an I-beam in cross-section (figure 2). This structure has a large second moment of area that is exceptionally capable of resisting bending perpendicular to its flange, here formed by the dorsoventral expansion of the transverse process. This expansion of the transverse process is relatively larger in S. somereni than S. thori, resulting in not only more points of contact between adjacent vertebrae (more tubercles; figure 4) to interlock and resist axial torsion, but also a beam cross-section with a greater concentration of material away from the neutral axis, and potentially better resistance to deformation under mediolateral bending.

The increased number of dorsal tubercles in the caudalmost thoracic vertebrae of both species, along with an additional v-shaped joint between the neural spines of adjacent vertebrae in the caudal thoracics [4,5] hint at a greater need to resist extension in this region using only the dorsal side of the column. This region also shows a local peak in Tb.N in both Scutisorex species (figure 5), suggesting additional centrum strength in compression. The automatic interdigitation of dorsal tubercles would be useful for resisting both extension and shear if the animal were undergoing an external, ventrally directed load positioned at the top of the kyphosis (like the load proposed in [2]), but it is unclear how that load would be transmitted through the limb girdles and the limbs themselves, which do not appear to be specialized to resist extreme forces [7].

Although we cannot yet identify a specific behaviour associated with the spinal morphology of Scutisorex, our results suggest that spinal flexion, and the vertebral compression associated with flexion, are important parts of the animal's lifestyle and/or locomotion. Scutisorex somereni has been variously described as having a ‘decidedly unsoricid’ body motion [1], and a ‘ ‘trotting’ (rather than ‘crawling’) gait under an extraordinarily hunched posture’ [58]. Passive bending of the spine also suggests a much greater range of motion in sagittal flexion than in sagittal extension or lateral bending [1]. How these qualitative descriptions translate into vertebral kinematics, and how spinal structure plays into the animal's unique locomotion, are yet to be fully investigated. The posture, gait, and spinal musculature of S. thori are poorly known compared to those of S. somereni, and may provide additional support for the hypothesis that the S. thori spine is functionally intermediate.

5. Conclusion

Determining the function of extreme morphologies is a quintessential challenge in organismal biology. We have shown here that the plastic mechanical properties of trabecular bone provide insight into the in vivo use of unusually modified bones, despite a lack of clear external homologies in functional structures. Our synthesis of data on the internal and external structure of Scutisorex vertebrae with the existing literature reveals that: (i) Scutisorex vertebrae likely undergo axial compressive forces that are higher in frequency and/or magnitude than the forces acting on C. goliath vertebrae, (ii) besides being intermediate in external morphology, S. thori is probably also functionally intermediate between S. somereni and C. goliath, based on TBA and external tubercle morphology, and (iii) there is correspondence between peaks in Scutisorex tubercle number and TBA, particularly Tb.N. Because our understanding of the respective functions of trabecular bone and cortical bone in very small mammals is limited, and behavioural information on Scutisorex is scant, much remains to be discovered about this enigmatic genus of shrews. Our work provides a starting point for additional tests of how Scutisorex and other small mammals experience force, as well as a new approach for tracking the evolution of function in morphologically extreme vertebrates.

Data accessibility

Trabecular bone architecture data and external vertebral measurements, including tubercle counts, as well as raw data from segmentation sensitivity and uncertainty tests, are available on the Dryad Digital Repository: https://doi.org/10.5061/dryad.8w9ghx3hf [59]. A μCT scan regions of interest (ROIs), and 3D STL models of Scutisorex vertebrae used in this study, are available for download at the online repository Morphosource (https://www.morphosource.org/Detail/ProjectDetail/Show/project_id/956).

Competing interests

We declare we have no competing interests.

Authors' contributions

S.M.S. and K.D.A. conceived the study; S.M.S. designed the study, selected specimens, performed μCT scans, and carried out segmentation and data analysis; K.D.A. consulted on data collection and analysis; S.M.S. and K.D.A. wrote the manuscript.

Funding

S.M.S. was supported by a National Science Foundation Postdoctoral Fellowship in Biology (grant no. DBI-1811627).

Footnotes

Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.4938126.