Broad similarities in shoulder muscle architecture and organization across two amniotes: implications for reconstructing non-mammalian synapsids

Two study species

We collected anatomical data from six wild adult Argentine black and white tegus (S. merianae) and six wild adult Virginia opossums (D. virginiana) (Tables S1 and S2). While varanid and iguanian lizards have traditionally been used as models of plesiomorphic amniote posture and locomotion (Jenkins & Goslow, 1983; Padian & Olsen, 1984; Ritter, 1996; Blob & Biewener, 1999; Blob, 2000; Farlow & Pianka, 2000; Clemente et al., 2011; Dick & Clemente, 2016), tegus were chosen here to represent a more shallowly-nested clade of terrestrial generalists (Sheffield et al., 2011; Simões et al., 2018) of growing importance as laboratory animals (Bennett & John-Alder, 1984; Montero et al., 2004; Toledo et al., 2008; Sheffield et al., 2011). In comparison, among extant mammals, didelphid opossums are a well-established plesiomorphic model for therian development, anatomy, and locomotion (Broom, 1899; Jenkins, 1971b; Hiiemae & Crompton, 1985; Klima, 1985; Parchman, Reilly & Biknevicius, 2003; Sánchez-Villagra & Maier, 2003; Gosnell et al., 2011; Hübler et al., 2013; Diogo et al., 2016; Bhullar et al., 2019).

Similarities in life history facilitate direct comparison between Salvator and Didelphis. Both animals are opportunistic omnivores of similar size, with high growth rates and high fecundity. Both are active foragers capable of sustained locomotion, with basal metabolic rates that overlap during the warmer spring and summer months (Urban, 1965; Fournier & Weber, 1994; Toledo et al., 2008). Finally, individuals of both species are easily available; S. merianae is invasive to southeastern Florida (Pernas et al., 2012; Jarnevich et al., 2018), while D. virginiana is commonplace throughout North and Central America (McManus, 1974). These traits make tegus and opossums ideal animal models for exploring anatomical and functional adaptations for “sprawling” and “erect” limb posture in amniotes (Butcher et al., 2011; Sheffield et al., 2011; Gosnell et al., 2011).

Muscle identification, topology and architecture

Here we focused on muscles that spanned the glenohumeral joint of the shoulder. Extrinsic muscles such as m. trapezius, m. serratus, and mm. rhomboidei were not considered, nor were muscles with only soft tissue attachments such as m. dorsoepitrochlearis. Preliminary homology hypotheses were taken from the literature (Diogo et al., 2009), then expanded based on our findings from digital and physical dissections.

Digital dissection: skeletal morphology and in situ muscle topology

Two cadaveric tegus (Body Mass 0.68 and 1.04 kg) and two opossums (Body Mass 1.04 and 1.11 kg) were skinned, gutted and fixed for 24 h in 10% neutral-buffered formalin solution. Following fixation, each animal was first subjected to a baseline X-ray microcomputed tomography (µCT) scan to capture skeletal morphology. Per published diceCT guidelines, the animals were then rinsed with deionized water and completely immersed in Lugol’s iodine solution (I₃K) to increase the radiodensity of muscle and other soft tissues, rendering them visible to µCT (Gignac et al., 2016). As these were relatively large animals with few precedents in the diceCT literature, we experimented with specimen treatment, immersion times and stain concentration throughout protocol development.

One tegu and one opossum were beheaded and bisected midway between the last rib and the pelvis, while the other two individuals were stained as whole animals. Each animal was posed with one forelimb extended and maximally protracted and the other flexed and maximally retracted, in order to bracket the limits of ex vivo mobility (Fig. 2). One tegu was stained in a 10% Lugol’s solution for one week, while the other tegu and the two opossums were stained in 3% Lugol’s for three, six, and nine weeks respectively. Once a week, each specimen was removed from the iodine solution, rinsed in deionized water, and µCT-scanned to check progress. Once satisfactory contrast enhancement was achieved, each animal was imaged a final time using a Nikon Metrology (X-Tek) HMXST225 scanner (Nikon Metrology, Tokyo, Japan) at the Harvard University Center for Nanoscale Systems. µCT technique was specimen-specific for optimal image quality, but fell within the following parameter ranges: 115–130 kV, 61–130 µA, 1,000–2,000 ms exposure, tungsten target, 0.5–1 mm Cu filter, 127 µm³ voxel size.

The raw µCT data were reconstructed in Nikon CT Pro 3D as DICOM series, and imported into Mimics v19 (Materialise NV, Leuven, Belgium) for segmentation. Muscles and bones were identified and manually isolated with two-dimensional masks, and three-dimensional surface meshes were then computed and exported as .stl files. We used the landmark-based point registration tool in 3-Matic v11 (Materialise NV, Leuven, Belgium) to align clean bone meshes from the baseline scans with the noisier post-stain bone meshes. We used the mesh editing tools in MeshLab (ISTI-CNR, Pisa, Italy) and Autodesk MeshMixer (Autodesk Inc., San Rafael, CA, USA) to downsample and retriangulate all meshes. The refined meshes were then imported into Autodesk MudBox (Autodesk Inc., San Rafael, CA, USA), and muscle attachments (origin/insertion) were painted directly onto bone models using the PTEX mapping method (Figs. 3–6). The locations and shapes of muscle attachment areas were later manually verified in separate physical dissections.

Physical dissection: muscle architectural properties

To validate muscle attachment sites and to measure muscle architectural properties, we skinned and dissected four additional cadaveric tegus (Body Mass 1.33 ± 0.11 kg) and four opossums (Body Mass 1.35 ± 0.26 kg). Muscles were identified and photographed in situ to validate the attachment areas previously determined from diceCT scans, then carefully excised, photographed again, dipped in 1x phosphate-buffered saline, blotted dry, and weighed. Each muscle-tendon unit (MTU) was weighed (g) with and without external tendon. Each muscle was then incised along its line of action and photographed again. ImageJ (U.S. National Institutes of Health, Bethesda, MD, USA) was used to measure fascicle length (mm) and internal pennation angle (θ, °). Measurements were repeated three times each at three separate locations in the muscle. To address measurement error, we collected measurements bilaterally whenever possible (i.e., unless one of the sides was too damaged) and took the average. Finally, we calculated each muscle’s physiological cross-sectional area (PCSA)—a proxy for maximum isometric force—using a standard formula (Eq. (1); Sacks & Roy, 1982), substituting the commonly-used literature value of 0.001056 gmm⁻³ for the density of typical vertebrate muscle (Mendez & Keys, 1960). As Eq. (1) generalizes to both pennate and non-pennate muscles (Lieber & Fridén, 2000), we followed common practice in applying it uniformly across all of our muscles (Allen et al., 2010; Dick & Clemente, 2016; Cuff et al., 2016).

To facilitate comparisons between the two species, we assumed isometry within the small size range of the specimens studied and normalized each individual’s measurements to a 1 kg animal, dividing lengths by body mass^1/3, areas by body mass^2/3, and masses by body mass. Species means and standard deviations were computed for muscle mass (M_m, g), MTU mass (M_mtu, g), muscle length (L_m, mm), MTU length (L_mtu, mm), fascicle length (L_f, mm), pennation angle (θ, °), and PCSA (mm²) (Table 1). Differences in normalized mean M_m, PCSA, L_f, and θ between homologous muscles were investigated using unpaired two-sample Student’s t-tests (all variables were normally distributed based on Shapiro–Wilk testing at α = 0.05), and the results were corrected for multiple simultaneous statistical comparisons (Fig. 7; Table S3). Finally, normalized PCSA was plotted against normalized L_f to uncover gross trends in muscle function across the evolutionary bracket.

Skeletal observations

Physical dissection revealed that both the tegu and the opossum possess intra-girdle mobility. In the tegu, the coracoids (=metacoracoid sensu Vickaryous & Hall, 2006) translate craniocaudally along the sternum in a sliding, tongue-and-groove articulation, similar to the Savannah monitor Varanus exanthemicus (Jenkins & Goslow, 1983). Both the acromioclavicular joint and the clavo-interclavicular joint are able to rotate to accommodate this motion, forming an open-chain three-bar linkage. Ligaments act to constrain the extremes of mobility at all three joints. The cranioventral borders of the coracoid cartilages overlap asymmetrically, with the right coracoid lying dorsal to the left in all the individuals studied.

In the opossum, large gaps were observed in the µCT data between the clavicle and its lateral and medial articulations with the acromion and the manubrium, respectively. Physical dissection revealed radiolucent, cartilaginous elements in these regions. The medial element is particularly noteworthy: found interposed between the medial end of each clavicle and the clavicular notches of the manubrium, it is flattened and paddle-shaped, with a rounded lateral (clavicular) end and a tapered medial (manubrial) end (Fig. 5). This element appears to behave as a tensile link, as it buckles laterally when placed under compressive axial load. As with most therians, the intra-girdle mobility in the opossum occurs at the sternoclavicular and acromioclavicular joints, which provide almost unrestricted motion once the extrinsic muscles spanning the shoulder (e.g., m. pectoralis, m. latissimus dorsi, m. trapezius) are severed.

Muscle observations

Muscle specialization and architectural design

Figure 8 (alternative high-contrast color coding in, Fig. S1) is a scatter plot of PCSA (normalized to body mass^2/3) against L_f (normalized to body mass^1/3) of muscles shared between the tegu and the opossum. This functional morphospace visualizes a tradeoff between force production and working range, and allows relative muscle specialization to be compared between the two species (Lieber, 2002; Eng et al., 2008; Allen et al., 2010; Dick & Clemente, 2016; Dickson & Pierce, 2019). Homologous muscles are generally architecturally similar between the tegu and the opossum and occupy similar regions of the functional morphospace (Fig. 8). PCSA tends to differ more than L_f between the two species, with muscles in the opossum tending to exhibit greater PCSAs than their tegu homologues. This trend is reversed for the mm. deltoideus and m. coracobrachialis brevis of the tegu, which both exhibit greater PCSA. M. pectoralis may also possess a greater PCSA in the tegu, although this difference was not significant due to high variance in the opossum data.

Five of the 13 muscles show statistically significant architectural divergence that may reflect shifts in locomotor function. (1) M. latissimus dorsi and m. pectoralis possess a significantly greater working range in the opossum, but force production is similar between the animals. (2) The mm. deltoideus appear to be slightly reduced in the opossum relative to the tegu: the scapular and clavicular heads have lower force capacity, and the clavicular head has a shorter working range. (3) The scapular head of m. triceps is capable of significantly greater force production in the opossum. (4) The opossum coracobrachialis complex is reduced compared to the tegu: m. coracobrachialis longus is absent, and m. coracobrachialis brevis has a significantly lower force capacity and working range. (5) The opossum equivalents of m. supracoracoideus are adapted for significantly greater force production at the expense of working range. This relationship holds true regardless of whether we compare the tegu supracoracoideus to the opossum mm. infraspinatus and supraspinatus individually, or if the latter two are pooled together as a single muscle.

Correcting for multiple comparisons

As the number of simultaneous statistical tests increases, so too does the chance of incorrectly rejecting the null (false positive/type I error), purely as a function of probability (Pike, 2011). It is standard practice to correct for this issue of multiple comparisons when evaluating a large number of hypotheses at the same time. Among the available correction methods, the Benjamini–Hochberg procedure controls the False Discovery Rate (FDR), or the proportion of incorrectly-rejected nulls (Benjamini & Hochberg, 1995). It is more powerful than the commonly-used Bonferroni and Holm corrections, and is considered more appropriate for situations where the number of hypotheses being tested is large relative to the number of samples (Horn & Dunnett, 2004). In the uncorrected analysis, 2/13 muscles differ significantly in θ, 5/13 in PCSA, 7/13 in L_f, and 9/13 in M_m. After applying the Benjamini–Hochberg correction, m. subscapularis no longer differs significantly in L_f between the tegu and the opossum, but all other differences remain significant at α = 0.05.

Broad similarities in muscle topology and architecture

Pectoral girdle musculature is markedly similar between the tegu and the opossum. Based on physical and digital dissection, we found strong evidence for anatomical homology in 13 out of 18 muscles across the two species, including resurrecting a formerly-dismissed correspondence between the lizard m. scapulohumeralis anterior and the therian m. teres minor (Text S1). Muscle topology follows skeletal morphology: while the proximal shoulder muscle attachments have migrated with the modification of the pectoral girdle bones in the opossum, they remain recognizable between the two taxa; further, muscle attachments are directly comparable on the long bones of the forelimb, which differ little between the tegu and the opossum. These findings bolster a growing body of evidence from the extant phylogenetic bracket for a common set of amniote pectoral girdle and shoulder muscles (Abdala & Diogo, 2010; Lai, Biewener & Pierce, 2018), while adding new architectural data for these structures.

With our data we can go beyond gross morphology and ask: does internal architecture parallel topological similarity between muscle homologues? We find that, once scaled to body mass, the architectural similarities between homologous muscles outnumber the differences (Fig. 7). Muscle mass is significantly different in many cases, but is frequently accompanied by significant differences in fascicle length, resulting in fewer instances of significantly different PCSA. Mean pennation angle is very similar, differing significantly between the tegu and the opossum in only a couple of cases. The fact that PCSA and pennation are more similar between the two species than muscle mass and fascicle length suggests that differing aspects of muscle function might be under differing selective pressures. Muscle PCSA relates to the force requirements for body support and locomotion (Eng et al., 2008). Pennation is linked to PCSA and force capacity for a given volume of muscle, but may also enable dynamic gearing of mechanical output (Azizi, Brainerd & Roberts, 2008). These functional traits almost certainly interact with one another within the confines of the shoulder and forelimb, and a balance between competing functional priorities may explain the architectural similarity observed between the morphologically-conservative amniotes studied here.

A possible explanation for the broad parallels shown here is that both the tegu and the opossum have retained many elements of the amniote last common ancestor’s (LCA) pectoral musculoskeletal system largely unmodified. These affinities echo the findings of prior comparative work, in particular that of Jenkins and colleagues (Jenkins & Weijs, 1979; Jenkins & Goslow, 1983), who compared muscle gross anatomy and function between the Savannah monitor Varanus exanthemicus and the Virginia opossum, hypothesizing a set of “functional equivalences” between lizard and therian shoulder muscles based on similarities in muscle activation patterns. Another possibility is that these myological characteristics evolved convergently, and are simply representative of a smaller-bodied terrestrial generalist phenotype. Both conservation and convergence have their issues: conservation is made less likely by the 320 million-plus years since the amniote LCA (Benton, 2009), while convergence is difficult to assess without prior knowledge of plesiomorphic states in the sauropsid and synapsid lineages. More extensive collection of architectural data from other extant generalist amniotes would be helpful in determining between conservation and convergence: Sphenodon is an early-diverging lepidosaur that attains comparable adult body sizes to S. merianae and D. virginiana (Feldman et al., 2016), and would be a useful point of comparison in inferring the plesiomorphic lepidosaur condition. Larger varanid lizards could provide a perspective from a different extant locomotor generalist; while the pectoral limb anatomy (Jenkins & Goslow, 1983) and pelvic limb architecture (Dick & Clemente, 2016) of varanids are well-studied, the architecture of the pectoral limb has not been published on. Among extant therians, quolls, tasmanian devils, hyraxes, and some of the more terrestrial civets are all locomotor generalists of similar size whose muscle architecture has yet to be studied. Finally, the morphologically-specialized (Jenkins, 1971a) extant monotremes invite comparison as similarly-sized phylogenetic intermediates between tegus and opossums. While a musculoskeletal model for the short-beaked echidna Tachyglossus aculeatus already exists (Regnault & Pierce, 2018), our ability to interpret this animal’s joint mobility and muscle moment arms will improve once its muscle architecture is described as well.

Anatomical differences reflect locomotor transformation

Of the five muscles not directly shared between the tegu and the opossum, all have known origins in the amniote pectoral musculoskeletal system. M. subcoracoideus and m. coracobrachialis longus are probably plesiomorphic for amniotes and secondarily lost in therians; they are present in amphibians and monotremes as well as the tegu, but absent in the opossum (Walthall & Ashley-Ross, 2006; Diogo et al., 2009; Gambaryan et al., 2015). The remaining three muscles are found in the opossum and not the tegu, but are not neomorphic. M. supraspinatus and m. infraspinatus have been shown to be differentiated developmental homologues of the lizard m. supracoracoideus (Romer, 1944; Cheng, 1955). Meanwhile, m. teres major is found in mammals, crocodile-line archosaurs, turtles, and at least one lizard, and is suggested to be an amniote character present in the LCA and secondarily lost in bird-line archosaurs as well as most lepidosaurs (Abdala & Diogo, 2010). Both of the muscles that are lost in therians (m. subcoracoideus and m. coracobrachialis longus) originate on the procoracoid and metacoracoid in the amniotes that possess them. Meanwhile, m. supraspinatus, m. infraspinatus, and m. teres major all take origin on the scapula, which in therians has uniquely expanded to comprise almost the entirety of the pectoral girdle. The synapsid fossil record preserves a marked trend towards coracoid reduction and scapular expansion, extending well into crown mammals (Romer, 1922; Romer & Price, 1940; Jenkins, 1971a; Luo, 2015). The shifting proportions of these skeletal elements and the accompanying dimensional and positional changes to the associated muscles are likely strongly linked to postural and locomotor evolution in synapsids.

Comparative study of architecture has the potential to reveal aspects of locomotor specialization (Allen et al., 2014; Böhmer et al., 2018). Hence, architectural differences in muscles that are shared between the tegu and the opossum may also be interpreted in light of the postural and locomotor differences between these animals. The longer fascicles of the opossum’s m. latissimus and m. pectoralis give these muscles a greater working range, which may accommodate the longer strides and greater pectoral girdle mobility of therians (Sereno & McKenna, 1995). The smaller PCSAs of the opossum’s clavicular and scapular deltoids compared to the tegu may signify a decreased reliance on these muscles for limb protraction during swing phase, since girdle mobility may be a greater factor than glenohumeral joint mobility in the therian stride compared to other amniotes (Fischer et al., 2002; Baier & Gatesy, 2013). The unusual, folded morphology of the tegu’s clavicular deltoid may work in conjunction with longer parallel-fibered fascicles to increase the muscle’s working range beyond what would be achievable with a conventional parallel-fibered muscle, and may again reflect a greater role in humeral protraction for the deltoids of a “sprawling” animal.

The triceps complex is notably more massive in the opossum, but its architecture suggests different functional specializations between the scapular and humeral heads. The opossum m. triceps scapularis exhibits a ~50% increase in PCSA over its tegu counterpart but has a similar mean fascicle length. By comparison, the two humeral heads of the triceps have the same PCSA in both animals, but with significantly longer fascicles in the opossum. The increased PCSA of the opossum m. triceps scapularis suggests adaptation to resist larger ground reaction flexor moments at the elbow, as a result of shifting to an erect forelimb posture. Meanwhile, the longer fascicles of the opossum m. triceps humeralis likely provide a wider working range for the muscle. Taken together, these features provide evidence for a functional differentiation within the triceps complex, with the scapular head becoming specialized for force control in therians, and the humeral heads becoming specialized for position control of the zeugopod.

Both m. supracoracoideus and its derivatives m. supraspinatus and m. infraspinatus are thought to stabilize the glenohumeral joint during locomotion (Jenkins & Weijs, 1979; Jenkins & Goslow, 1983). In therians, a rotator cuff comprising m. supraspinatus, m. infraspinatus, m. subscapularis, and m. teres minor pulls on the humerus from opposing directions. The resulting compression of the humeral head against the glenoid fossa has been shown to be an important source of dynamic stability in ball-and-socket glenohumeral joints (Lippitt & Matsen, 1993; Hsu et al., 2011), and is thought to be particularly important in intermediate poses of the shoulder when the capsule and ligaments are relaxed and unable to contribute to joint stability. The greater PCSAs of the m. supraspinatus and m. infraspinatus could thus serve to stabilize the humerus as it undergoes the rapid, rhythmic oscillations generated by the highly-tuned therian neuromuscular system (Jenkins & Goslow, 1983; Ross et al., 2013). By comparison, the smaller PCSA and longer fascicles of m. supracoracoideus may reflect a greater reliance by other amniotes on ligaments rather than muscle force (Haines, 1952) to stabilize the shoulder, and may facilitate protraction of the humerus as it moves through the long, horizontal arcs typical of “sprawling” locomotion (Jenkins & Goslow, 1983; Baier & Gatesy, 2013).

Reconstructing synapsid musculoskeletal evolution

Whether conservation or convergence is ultimately responsible for the architectural similarities found here, it is apparent that selective pressures have either maintained or resulted in a distinct terrestrial generalist architectural phenotype shared by the tegu and the opossum. As non-mammalian synapsids are typically reconstructed as terrestrial generalists, the architectural parameters of the tegu and opossum may provide realistic “bookends” for the phylogenetically- and morphologically-intermediate non-mammalian synapsids. Here, we illustrate this by estimating shoulder musculature PCSA in Massetognathus pascuali, a Triassic traversodontid cynodont for which a phylogenetically-bracketed musculoskeletal reconstruction has been published (Lai, Biewener & Pierce, 2018). Using a scaling equation to predict total body mass based on the circumferences of the humeral and femoral diaphyses (Campione & Evans, 2012), we estimate the body mass of a particular Massetognathus pascuali individual (MCZVP 3691) as 1.437 kg, with upper and lower bounds given by mean percent prediction error (PPE) as 1.806 kg and 1.069 kg respectively (Table S4). This prediction closely matches the tegus (1.33 ± 0.11 kg) and opossums (1.35 ± 0.26 kg) dissected in the present study.

Assuming geometric similarity, we can scale mass-normalized PCSAs from both extant animals by Massetognathus’ predicted body mass^2/3 to arrive at architectural estimates reflecting contrasting postural paradigms (Fig. 9). Error around estimated PCSA is calculated by scaling by the upper and lower bounds of the body mass prediction. By taking the mean of the tegu and opossum values, we can obtain a third, intermediate estimate that might better represent an extinct transitional form. As expected from the broad similarities in PCSA previously discussed, Salvator-like, Didelphis-like, and intermediate values tend to fall fairly close to one another. In seven of the 16 muscles reconstructed both the Salvator-like and the Didelphis-like estimate are included within the upper and lower bounds of the intermediate estimate (thick black lines, Fig. 9); they fall just outside the bounds of the intermediate estimate in another five muscles. Altogether, this suggests that discrepancies between Salvator-like and Didelphis-like estimates of PCSA are generally proportionately small, and tend to be either closely comparable to or dominated by uncertainty around estimated body mass. This provides confidence in the intermediate estimate, and at the same time illustrates the importance of conducting sensitivity testing in future work, so as to ascertain the relative contributions of the bracket and predicted body mass towards uncertainty in muscle parameter estimation.

While the Salvator-Didelphis intermediate seems to provide a reasonable estimate for the bulk of the shoulder musculature, a different approach is clearly needed for the m. supracoracoideus/mm. infraspinatus + supraspinatus group, where the opossum-like estimate exceeds the tegu-like estimate by a factor of approximately four (Fig. 9). The fourfold difference in normalized PCSA reflects the smaller relative size of this muscle in the tegu (0.85 g) compared to the opossum (2.32 g). However, as shown in the Table S5, the normalized ratio of PCSA: muscle origin area is similar between the tegu m. supracoracoideus (0.18) and its homologues mm. infraspinatus + supraspinatus in the opossum (0.15). Moreover, mm. infraspinatus and supraspinatus PCSAs have been shown to covary with the size of their areas of origin in the short-nosed bandicoot Isoodon fusciventer, a marsupial with pectoral skeletal anatomy comparable to Didelphis (Martin et al., 2019). Considering mm. supracoracoideus, infraspinatus, and supraspinatus are all flat, broad, pennate muscles with fleshy origins that occupy well-delineated areas on the scapulocoracoid or scapula (Fig. 3), relative origin area may be a reasonable proxy for estimating PCSA in Massetognathus. Lai, Biewener & Pierce (2018) reconstructed Massetognathus with an m. infraspinatus on the infraspinous fossa of the scapula, as well as an m. supracoracoideus/incipient m. supraspinatus on the anterior coracoid (procoracoid sensu Vickaryous & Hall, 2006) (Fig. S2B). The area of origin for m. infraspinatus in Massetognathus resembles that of the same muscle in the opossum (Figs. S2B–S2C), whereas the area of origin for the incipient m. supraspinatus in Massetognathus resembles that of the tegu m. supracoracoideus (Figs. S2A–S2B). Taking the PCSA: muscle origin area ratios for the opossum m. infraspinatus and the tegu m. supracoracoideus, we are able to reconstruct the PCSAs of mm. infraspinatus and incipient supraspinatus in a 1.437 kg Massetognathus as 29.40 mm² and 18.86 mm² respectively, resulting in a combined PCSA of 48.26 mm². This is a plausible estimate for Massetognathus, falling in between the PCSAs of the tegu (35.50 mm²) and the opossum (130.84 mm²). An alternative, higher estimate (65.12 mm²) using the proportions of the tegu m. supracoracoideus and the opossum m. supraspinatus for the mm. infraspinatus and supraspinatus (respectively) of Massetognathus also falls between the tegu and the opossum.

We note several caveats for reconstructing non-mammalian synapsid myology using this approach. While evidence exists that PCSA and L_f for many limb muscles scale at or close to isometry in varanid lizards (Dick & Clemente, 2016), proximal limb muscles scale with positive allometry in extant mammals and certain crocodiles (though not alligators) (Alexander et al., 1981; Allen et al., 2014). Not all non-mammalian synapsids were comparable in size to tegus and opossums; many, such as the Permian caseids and the Permo–Triassic dicynodonts, may have been up to three orders of magnitude larger (Reisz & Fröbisch, 2014; Sulej & Niedźwiedzki, 2019; Romano & Manucci, 2019), and architectural allometry may have to be taken into account. More studies of architectural allometry within extant squamates and therians are needed to determine the salience of body size to muscle functional design. Further, the ultimate mechanical consequences (e.g., for muscle force, strain, and work production) are difficult to determine without modeling the musculoskeletal system as a whole. With that being said, general agreement among most of these estimates of PCSA provides confidence that fossil muscle architecture can be empirically inferred within reasonable limits, and the robustness of parameter estimates for reconstructing locomotor evolution can be assessed in future work by performing sensitivity testing on musculoskeletal models incorporating inferred architecture.