Digital dissection of the pelvis and hindlimb of the red-legged running frog, Phlyctimantis maculatus, using Diffusible Iodine Contrast Enhanced computed microtomography (DICE μCT)

Diffusible iodine contrast enhanced μCT scanning

One adult P. maculatus frog (15.7 g body mass), obtained from Amey Zoo (Hempstead, UK), was euthanised by Tricaine methanesulfonate (MS222) overdose (0.02% MS222, 0.04% NaHCO₃) followed by removal of the heart (compliant with primary and secondary methods of amphibian euthanasia as per procedures approved by Home Office License 70/8242). The wound in the chest was closed using non-absorbable braided silk suture (6-0) to limit internal exposure to fixative and staining solution therefore avoiding over-staining. The frog (whole and un-skinned) was fixed in 10% neutral buffered formalin (NBF) (HT501128, Sigma Aldrich) for 29 hours, at room temperature, in a darkened environment. Following the fixation process, any remaining fixative was removed by transferring the specimen in to a PBS solution where it was left soaking overnight, at room temperature. To enhance soft tissue contrast for imaging, predominantly of the muscular anatomy, the frog was stained using an aqueous Lugol’s solution (L6146, Sigma Aldrich, a.k.a. iodine potassium iodide, I₂KI). To avoid over staining, test scans were performed at regular intervals throughout the staining process. After each test scan the specimen was re-introduced to the stain. Depending on the results of the scan, various recommendations were applied to increase stain perfusion, including skinning the specimen (conducted after 3 days), increasing the stain concentration (conducted after 16.5 days), and injecting stain into body (conducted after 31.5 days) (see Table 1). The frog was placed in 70% pure ethanol (02877; Sigma Aldrich, St. Louis, MO, USA) for preservation during transport to and from the scanner. All µCT scans were conducted on the SkyScan1176 in vivo µCT scanner (Bruker microCT, Kontich, Belgium) in the Biological Services Unit of the Royal Veterinary College, Camden campus. The specimen was wrapped in cellophane and taped down for each scan to prevent drying out or movement during imaging. For each of the test scans, a section of the mid-thigh and/or pelvis was chosen for imaging since this was a time efficient way to check image quality of both bone and ample soft tissue. After a total of 39 days, the entire specimen was imaged in the final scan (17.64 µm resolution, 50 kV, 362 µA, 1 mm Al filter). The scan images were then reconstructed using NRecon software (V1.6.10.1; Bruker microCT, Kontich, Belgium), showing the subcutaneous soft tissue topography of P. maculatus (Figs. 1A–1C).

Visualisation and segmentation

The reconstructed DICE µCT scan images from the final scan were resampled (1 in 5) and then visualised and digitally segmented in Amira 6.0.1 (FEI, Hillsboro, OR, USA) (Figs. 2A–2D). The multiplanar viewer function tab was used to create a volume model of all structures. Both the magic wand and paintbrush tools of the segmentation editor function tab were used to assign voxel selections as either bone or muscle material. Voxel assignment was made on the basis of greyscale value, those of the lightest colour denoted bone or muscle, whereas black voxels denoted air space. Due to limitations of the technique (see Discussion), author’s discretion and anatomical expertise were required for selection and assignment of voxels at the boundary between two materials. Every individual bone and muscle between the 4th vertebra and the distal digits of the hindlimbs were assigned as a separate material. Muscle and bone identifications were performed with the aid of previously published descriptions of other frog species (Dunlap, 1960; Emerson & De Jongh, 1980; Přikryl et al., 2009). Once the material selections for all muscles were complete, the segmented label field data was resampled (data resampled by 50% in the Z direction) before being rendered into 3D surface meshes to produce a 3D representation of the musculoskeletal anatomy of the frog lower spine, pelvis, and hindlimb (Fig. 2D). During surface rendering the file underwent constrained smoothing to minimise the visual appearance of the voxels, providing a more even surface and therefore realistic representation of the tissues. Each material surface mesh was exported individually as an STL file. Using the software 3-Matic (Materialise Inc., Leuven, Belgium), the individually exported meshes corresponding to the bones of the spine, pelvis, hindlimb, and foot, as well as the individual muscles of the left side (spanning from the spine to the tarsometatarsal (TMT) joint) were combined to create a 3D model of P. maculatus (Fig. 2D). Additionally, viewable as a 3D PDF, the digital model presents all skeletal material and all muscles of the left side, totalling 17 bones and 41 muscles (Data S1, see Article S1 for 3D PDF user guide). The individual metatarsals and phalangeal foot bones are grouped together and referred to as the metatarsals and digits, respectively.

Traditional dissection

Due to the limitations of DICE µCT to visualise tendinous material (see Discussion), musculoskeletal anatomy data were additionally obtained from a female adult frog by traditional means (specimen P. maculatus, body mass: 31.20 g, Source: Amey Zoo, Hempstead, UK). This animal had been previously euthanised using the methods described above, and fixed in 10% NBF (HT501128, Sigma Aldrich) for ∼24 hours. Each muscle of the spine, pelvis, and hindlimb (left side only) was identified, described, and removed in its entirety. Identifications were made with the aid of both the digital dissection of the 3D model (current work) as well as the previously published dissection data (Ecker, 1889; Dunlap, 1960; Emerson & De Jongh, 1980; Duellman & Trueb, 1986; Přikryl et al., 2009). The muscle names used throughout the current study are consistent with previous anuran dissection literature (Dunlap, 1960; Emerson, 1979; Emerson, 1982; Emerson & De Jongh, 1980; Duellman & Trueb, 1986; Přikryl et al., 2009).

Pelvic morphotype verification

Comparison of the traditional and digital dissection findings with descriptions of Emerson’s three morphotypes revealed that P. maculatus’s pelvis best fits the description of morphotype IIA (Fig. 10). The sacrum was dorsoventrally flattened with some lateral diapophyseal expansion, and ligaments appeared triangular in shape. Additionally, the sacro-iliac joint of P. maculatus allowed the ilia to slide anterioposterially, rotate laterally, and rotate dorsoventrally, whereas the sacro-urostylic joint was bicondylar and relatively inflexible (tested via manual manipulation).

The use of DICE μCT scans, 3D PDF, and digital dissection

Dissections revealed intricate musculoskeletal anatomy within the hindlimb and pelvic apparatus, consisting of a large number of muscles, and multiple instances of convoluted curved muscle pathways, where muscles wrapped around bony and soft tissue structures or passed through other muscles. Using DICE µCT enabled us to capture and preserve the 3D topography of the musculoskeletal system of this species in a level of detail that is challenging to achieve using traditional methods. DICE µCT has many other advantages over traditional dissection, including its non-destructive nature, opening this technique up to the anatomical investigation of specimens that cannot undergo destructive sampling. Moreover, iodine staining has been suggested to be reversible (Bock & Shear, 1972 cited in Jeffery et al., 2011), allowing repeated scanning of the same sample with optimised stain concentrations for different features (Jeffery et al., 2011).

Comparative musculoskeletal anatomy

Comparing our dissection findings from P. maculatus, with those of other anuran species (Dunlap, 1960; Přikryl et al., 2009), we found the different anatomical regions exhibited different levels of variation among species. The most variable regions were the spine and pelvis, and the tarsals and proximal foot. Whereas there were fewer examples of anatomical variation in the thigh, and the shank appeared uniform across species.

The anatomical variation of the muscles of the back and the pelvis included variation in insertion sites, some muscle belly shapes, and muscle presence/absence/fusion. While the LD in P. maculatus inserted approximately half way down the urostyle, its insertion site in other species ranges from the anterior portion, as in Pelophylax kl. esculentus, to the very tip of the urostyle, as in Ascaphus truei (Dunlap, 1960). The IL insertion in P. maculatus is also more proximal than observed in species such as Kaloula pulchra where it extends down onto the iliac shaft (Dunlap, 1960; Přikryl et al., 2009). While the CS was present as a separate muscle in P. maculatus, in X. laevis and Barbourula busuagensis the CS and LD are fused into one muscle belly (Dunlap, 1960; Přikryl et al., 2009). Compared with P. maculatus, which has a clear PY muscle, the PY of X. laevis is reduced and further, is altogether absent in Pipa pipa (Dunlap, 1960; Přikryl et al., 2009; Porro & Richards, 2017). In contrast, A. truei and Leiopelma hochstetteri possess a caudopuboischiotibialis (the ancestral trait) not present in any of the other species studied by Dunlap (1960) or indeed P. maculatus. Finally, the IE muscle of P. maculatus was narrow and cylindrical, similar to Rhaebo guttatus. In other species the IE appears broader and more fan-shaped (Přikryl et al., 2009), whereas A. truei and L. hochstetteri it is fused with the II muscle (Dunlap, 1960).

The majority of the muscles of the thigh appear uniform among species, however variation is observed in the presence/absence of the AL, number of heads present in the CR and ST, and fusion of some muscle bellies. While P. maculatus shares the presence of AL with Discoglossus pictus, B. busuagensis and Pelobates fuscus, the AL is absent in, Bombina orientalis, B. bombina, B. pachypus, Alytes obstetricans, Spea hammondii, Megophrys montana, X. laevis, P. pipa, Rhinophrynus dorsalis, Atelopus cruciger, Polypedates leucomystax, and Chiromantis xerampelina (Dunlap, 1960; Přikryl et al., 2009). Furthermore, while in P. maculatus, and most other species, CR is one muscle belly, in specialised aquatic species, the CR muscle is divided into two incomplete heads (Dunlap, 1960). Finally, A. truei and L. hochstetteri exhibit fusing and combinations of multiple thigh muscles (SA and ST; AL and PEC; QF and GE) that were not observed in P. maculatus (Dunlap, 1960).

While muscles of the shank were not variable, multiple differences in muscle proportions, attachment sites, presence/absence and splitting/fusion are evident in the tarsals and proximal foot. Like most other species, P. maculatus had two distinct heads of the TiAL muscle, which were of roughly equal size. Dunlap (1960) reported variation in the proportions of the two heads, and the position at which the muscle bellies of TiAL diverge. The TaP muscle of P. maculatus is similar to that of other Ranids but is smaller than seen in A. truei and L. hochstetteri (Dunlap, 1960). The EBS muscle splits into multiple heads in all species however the digit upon which the middle head of this muscle inserts is variable among species (Dunlap, 1960). As in the other areas of the hindlimb, the highly specialised aquatic species, such as P. pipa and X. laevis, lacked muscles which were present in P. maculatus, such as the EBS, AP, and TPP and D. Additionally, while in P. maculatus, the PP remains separate from FDBS, in A. truei and L. hochstetteri these two muscles are fused (Dunlap, 1960).

Overall, P. maculatus had a similar gross muscular anatomy to R. catesbeiana and R. guttatas but differed from those species that exhibit ancestral traits (leopelmids) and those that are highly specialised (pipids) (Dunlap, 1960; Přikryl et al., 2009). The highly specialised, P. pipa and X. laevis lacked muscles which were present in P. maculatus, such as the PY (which is only reduced in Xenopus (Porro et al., 2017), CI, EBS, AP, and AL. Whereas, A. truei and L. hochstetteri exhibit fusing and combinations of multiple muscles (SA and ST; AL and PEC; OE; QF and GE; PP and FDBS; IE and II) that were not observed in P. maculatus (Dunlap, 1960).

Diversity of hindlimb/pelvis muscle morphology in relation to diversity in function

In the pelvic region, there is strong evidence linking variation frog musculoskeletal anatomy to diversity in locomotor style (Emerson & De Jongh, 1980). The region differentiates after metamorphosis and thus the variation seen in this musculature seems to relate to expansion of locomotor capacity as froglets develop (Fabrezi et al., 2014). Indeed, differences in the LD, CI, CS, and IL muscles contribute to functional differences in lateral bending and gliding of pelvis associated with walking and swimming, respectively (Emerson & De Jongh, 1980). The II, IE, and TFL are derived muscles also likely to influence locomotor mode (Přikryl et al., 2009). For example, IE length correlates with jumping (Fabrezi et al., 2014), thus may be expected to have a similar morphology among jumpers. However, IE morphology of P. maculatus differs from other jumpers (e.g., Ranids) owing to its cylindrical shape (see above). Given that the IE may play an important role in bringing the leg upwards and forwards (Kargo & Rome, 2002), perhaps its shape represents a modification to assist with the swing phase of walking/running.

Compared with the pelvis, evidence for interspecific muscle structure-function relationships in the hindlimb is relatively sparse. As stated above, the thigh region appears more conserved compared with the pelvis and distal hindlimb among species, including P. maculatus. Despite conservation of thigh muscular traits among modern taxa, A. truei and L. hochstetteri (representing “primitive” taxa) exhibit fused musculature (see above). In absence of further modelling analysis (e.g., Kargo & Rome, 2002), we can only speculate on the functional consequence of fused muscles. Perhaps separate versus fused muscles allows for a greater variation in muscle moment arms, thus increasing the functional workspace of the limb hence enhancing the limb’s ability to perform multiple tasks.

Apart from the above discussion based on a small number of species and focused anatomical regions, we are far from a clear and complete understanding of how interspecific variation in muscle morphology relates to locomotor function. In light of broad patterns relating limb, body, and skeletal morphology to locomotor style (Zug, 1978; Emerson, 1988) as well as performance, ecology, and phylogeny (Moen, Irschick & Wiens, 2013), we also expect relationships to emerge with muscular morphology. Although certain features of the musculoskeletal system seem to vary independently of ecology/function (Fratani, Ponssa & Abdala, 2018), we expect that other traits may exhibit morphologically small differences amounting to profound effects on locomotion. In particular, small changes in muscle moment arm distances have a strong impact on muscle function (Lombard & Abbott, 1907; Kargo & Rome, 2002). For example, there is great diversity in subtle aspects of bone shape (e.g., ridges on ilia; Reilly & Jorgensen, 2011) which may alter muscle origins/insertions significantly enough to change muscle moment arms. Moreover, muscles are not mechanically independent; the function of a single muscle depends on the current configuration of the joints, thus is dependent on the action of other muscles (Lombard & Abbott, 1907; Kuo, 2001; Kargo & Rome, 2002). Consequently, analyses that treat all muscles independently might overlook synergistic effects of small changes in one muscle with respect to other muscles. Regardless, the species currently sampled for detailed muscle analysis are likely too few to disentangle function/ecology from phylogenetic effects. Fortunately, recent workers have assembled a vast archive of frog digital anatomy (open Vertebrate; “oVert”) and are currently making it available for exploration and study (Watkins-Colwell et al., 2018). Thus, future work can build upon our observations as well as past work to accumulate sufficient intraspecific data to apply rigorous comparative/phylogenetic methods (e.g., Moen, Irschick & Wiens, 2013; Fratani, Ponssa & Abdala, 2018) towards elucidating muscle morphology-function relationships across Anura.

Gracilis major and Semimembranosus myology

The split found in the SM and GR (major) is unique among the dozens of muscles of the leg and pelvis. Although this tendinous “septation” is well-characterised in all other anurans observed (Dunlap, 1960; Duellman & Trueb, 1986; Přikryl et al., 2009), neither the developmental mechanism nor the biomechanical significance of this muscle structure is known. One possible explanation is incomplete fusion of two evolutionary/developmentally distinct muscles which appears as intramuscular separation. However, Přikryl et al. (2009) note that the hindlimb extensor muscles in modern frogs have not changed over the course of evolution; thus, we speculate that separation may not be a developmental/evolutionary artefact, but rather may have functional significance. For example, tendinous separations between muscle bellies are not unique to anurans; separated muscles have been described in salamanders (Ashley-Ross, 1992; Walthall & Ashley-Ross, 2006) and cats (Bodine et al., 1982). In those cases, the divided muscle bellies receive separate innervation (Francis, 1934; Bodine et al., 1982) suggesting that two regions of a single muscle could act independently. If the partitions of the SM and GR were found to have separate innervation, we propose a potential mechanical function as follows. Both the SM and GR muscles span from the pelvis to the knee and insert into the aponeurosis covering the knee. The SM in particular has been recorded to function both as a femur retractor and a knee flexor and, while it does not act on the knee joint, GR is also dual functional, acting to either retract or adduct the femur (Přikryl et al., 2009). We speculate a separation could act to partition the portions of muscle designated for each role, allowing the animals to fine tune hindlimb motion. Further anatomical investigation determining innervation and/or spatial partitioning of fibre types would be required to better characterise these muscles. Subsequently, modelling analyses could be performed to assess the biomechanical impact of intramuscular separation.

Phlyctimantis maculatus pelvic morphotype

Given recent findings that P. maculatus are walkers/runners as well as hoppers, jumpers, and climbers (Porro et al., 2017), we evaluated whether the pelvic morpohotype represented walker or jumper traits. Pelvic dissection revealed that P. maculatus’s anatomy was consistent with the Type IIA morphotype defined by Emerson (1979). The sacrum was dorsoventrally flattened as opposed to the cylindrical sacral shape of Type IIB jumping species, yet lacked the extreme laterally flared diapophyseal expansion and broad ligamentous cuff of the Type I pelvis, as seen in aquatic, swimming species such as P. pipa (Emerson, 1979; Emerson, 1982; Přikryl et al., 2009).

However, some of the key features of the Type IIA morphotype were subtler in P. maculatus compared with other walking and hopping, and burrowing species. For example, R. guttatus, B. busuagensis, D. pictus S. hammondii, Rentapia hosii, A. obstetricans, Anaxyrus (Bufo) americanus, A. boreas are walking, hopping, or burrowing species that exhibit broader lateral flaring of the sacral diapophysis and a more obvious bow-tie shape sacrum (Přikryl et al., 2009; Reilly & Jorgensen, 2011). Although dorsoventrally flattened, the sacrum of P. maculatus demonstrated less flaring and more posterior lateral projection of the sacral diapophyses, similar to the skeletal morphology of A. truei and A. montanus (Reilly & Jorgensen, 2011). Perhaps the typical Type IIA morphotype features are less prominent in P. maculatus because of its more arboreal ecology and use of both walking and jumping behaviour (as opposed to hopping).

Furthermore, P. maculatus also exhibited slightly dorsally ridged ilia and urostyle. Having prominent ridges along the dorsal surface of the ilia is a trait commonly seen among the sagittal-hinge Type IIB morphotypes (Emerson, 1979; Emerson, 1982; Reilly & Jorgensen, 2011). Nonetheless, Reilly & Jorgensen (2011) reported the same set of features in other representatives of Hyperoliidae (Kassina senegalensis and Hyperolius lateralis), summarising that the pelvic girdle design of this family comprises an expanded sacrum, iliac ridges, and half urostyle ridge. They also categorised this family as walkers, hoppers, and arboreal jumpers. P. maculatus is therefore consistent with their description of hyperoliid frogs. Furthermore, manual manipulation of the ilio-sacral joint prior to and following muscle dissection suggested a capacity for lateral rotation, some anteroposterior sliding, and sagittal bending. Since lateral rotation of this joint is only freely permitted by the Type IIA morphotype (by definition Emerson, 1979; Emerson, 1982), we felt justified in classifying the pelvis of P. maculatus as such. It should also be noted that while the degree of sacral diapophyseal expansion is a useful trait in distinguishing between the sagittal hinge or lateral bender morphotype, the extent of sacral diapophysis lateral expansion is highly variable among anurans (Jorgensen & Reilly, 2013).

Given the extent of subtle variation observed in the musculoskeletal anatomy of anurans in general, it seems likely that Emerson’s categories represent the archetypical morphotypes within each behavioural group (walking/hopping, jumping, and swimming). Those species that use multiple locomotor behaviours possess a subtle blend of pelvic characteristics. Thus, rather than fitting discrete morphotypes, frogs more likely span a continuum of pelvic morphologies depending on the combination of locomotor behaviour expressed by a given species, as suggested previously (Soliz, Tulli & Abdala, 2016).

Limitations and caveats of DICE μCT

DICE µCT has proved an excellent method to present the 3D topography of the musculoskeletal system, allowing the visualisation of complex 3D interactions between muscles and other structures not possible with traditional techniques. However, there are some notable caveats associated with DICE µCT that should be discussed here.

The aim of the DICE µCT technique is for the stain to disperse through the soft tissue, increasing contrast (Figs. 2A–2C), as such, poor diffusion or too low a stain concentration results in poor contrast enhancement. Here, we skinned the specimen in order to assist with stain perfusion however a lack of published methodologies for amphibian staining means we cannot comment on how effective removal of the skin was for this purpose. On the other hand, poor soft tissue contrast can also result from overstaining. Furthermore, a fine balance needs to be struck in order to avoid distortion of the sample due to extreme tissue shrinkage.

Tissue shrinkage is a commonly reported caveat of DICE µCT (Vickerton, Jarvis & Jeffery, 2013; Cox & Faulkes, 2014; Buytaert et al., 2014; Gignac et al., 2016; Bribiesca Contreras & Sellers, 2017). Muscle volume shrinkage due to contrast enhanced staining has been previously reported ranging from 10–56% (Vickerton, Jarvis & Jeffery, 2013; Buytaert et al., 2014; Bribiesca Contreras & Sellers, 2017). The extent to which shrinking occurs in stained specimens increases with higher concentrations of I₂KI and can be reduced by using lower concentrations over a longer duration as was implemented in this study (Vickerton, Jarvis & Jeffery, 2013; Gignac & Kley, 2014). Contrast enhanced staining is therefore a time consuming process and despite us using low concentrations of Lugol’s solution here to avoid it, shrinkage of the muscle tissue was observed in our frog. We do not believe the level of shrinkage in our specimen to be detrimental to the overall results of our anatomical assessment since origins, insertions, and pathways are unlikely to have been affected significantly (Cox & Faulkes, 2014). While muscle volumes are not the focus here, it should be noted that measuring the specific level of shrinkage is an important consideration in those studies reporting quantitative muscle data (for example, see Bribiesca Contreras & Sellers, 2017 for their comparison of dissected vs stained muscle volumes).

A further limitation associated with the DICE methods used here is the inability to visualise tendinous or ligamentous tissues. Important structures such as the knee aponeurosis, muscle tendons, or plantar fasciae were therefore indistinguishable in the scan data and excluded from the 3D PDF. This caveat was also encountered by Porro & Richards (2017) who suggest using agents that bind to collagen as an alternative. Descamps et al. (2014) demonstrate that phosphotungstic acid (PTA) has a preference for binding to collagen and connective tissues whereas phosphomolybdenic acid (PMA) provides good contrast for the visualisation of cartilage using CT. When using any chemicals health and safety precautions must be adhered to, not all laboratory spaces are suitable to conduct the aforementioned procedures.

Finally, obtaining the highly detailed results of DICE µCT is time consuming, making analysis of several specimens of one species impractical. Consequently, studies such as this assume low intra-species variation. Moreover, the subsequent segmentation of the µCT scan data requires anatomical expertise and relies on the user’s discretion to appropriately define voxel material. The final 3D model is therefore best defined as a 3D representation of the anatomy of an example specimen.

Despite the limitations, DICE µCT as a technique has already begun to revolutionise the way anatomy is visualised and studied. With further use in a wider range of taxa, the protocols are likely to improve and become more standardised as a wider knowledge base for troubleshooting is generated. Even though this is one of the earliest uses of the technique in anurans, the results obtained in this study were remarkable and facilitated a deeper understanding of the gross anatomy of P. maculatus.

Future work

Recently, anatomical reconstructions created using DICE µCT have been used as the foundation for musculoskeletal models to investigate structure-function relationships of the locomotor musculoskeletal system (Charles et al., 2016a; Charles et al., 2016b; Allen et al., 2017). We propose to use our 3D digital model to generate an anatomically accurate musculoskeletal model of P. maculatus, allowing us to explore the mechanical effect of the complex curved muscle trajectories, and test our speculative hypotheses regarding the implications of separate innervation in the same muscle belly (such as seen in the SM and GR). Furthermore, our musculoskeletal model can be applied to questions regarding evolutionary adaptations. For example, by altering muscle attachment sites, and/or skeletal proportions to mimic those of extinct species, we can explore the functional significance of such adaptations.