Development of human limb muscles based on whole-mount immunostaining and the links between ontogeny and evolution

The data and images that appear in all modern textbooks and atlases of human development, including the brief accounts of muscle development, are based on decades-old analyses that used simple histological sections or wax reconstructions and 3D models generated from sectioning and staining of embryos and fetuses. Much of our knowledge of limb musculoskeletal development in particular is based on work from more than a century ago (e.g. Bardeen and Lewis, 1901; Lewis, 1901; Gräfenberg, 1905; Bardeen, 1906; Lewis, 1910), with only a few exceptions (e.g. Ribbing, 1938; Cihak, 1972; O'Rahilly and Gardner, 1975). This leads to a situation in which we now have more understanding of the ontogeny of skeletal muscles for taxa such as fishes, salamanders, frogs, chicken and mice than for our own species – thanks to recent developmental techniques such as whole-mount immunostaining that have been used to study those taxa (see, e.g. Diogo et al., 2008, 2018; Grenier et al., 2009; Diogo and Abdala, 2010; Schmidt et al., 2013; Diogo and Ziermann, 2014; Ziermann and Diogo, 2014; Konstantinidis et al., 2015; Noda et al., 2017). For instance, there are databases available showing in 3D the development of certain muscles in mice (e.g. Delaurier et al., 2008), but there are no such databases available for humans. A few recent analyses have used modern techniques to study some aspects of muscle ontogeny in humans, but these have generally focused on a few muscles and/or on limited developmental stage(s) (e.g. Butler-Browne et al., 1990; Edom-Vovard et al., 1999; see also more examples of such studies in Belle et al., 2017).

Recently, one of us (Y.G.) has contributed to an effort to change this situation by performing whole-mount immunostaining on 36 human embryos and fetuses within the first trimester of gestation with over 70 antibodies (Belle et al., 2017). By combining whole-mount immunostaining, 3DISCO (three-dimensional imaging of solvent-cleared organs) clearing, and light-sheet imaging, this effort led to the generation of high-resolution 3D images of the developing peripheral nervous, muscular, vascular, cardiopulmonary and urogenital systems of human embryos and fetuses. The aim of that effort is to build an atlas of human development comprising 3D images at an unprecedented cellular resolution that are based on the direct imaging of whole-mount tissues, rather than on 3D reconstructions of sections, and can be used by developmental biologists and comparative anatomists, as well as by professors, students, physicians/pathologists and the broader public (Belle et al., 2017).

In the present paper, we provide the first detailed analysis of the ontogeny of the human upper and lower limb muscles using such 3D images. We compare our observations with those few earlier studies that focused on the development of these muscles in humans, in order to provide detailed tables summarizing the timing of appearance, as well as the splitting, fusion and/or loss of each of these muscles. In addition to the unprecedented detail offered by 3D images of karyotypically normal human fetuses and embryos, this study covers limb development all the way to fetuses that are 13 gestational weeks (GW) old. In contrast, the scarce information available for the development of most human limb muscles refers to 6-8 GW embryos. For instance, the older embryos described in the influential works of Lewis (1901, 1910), Bardeen and Lewis (1901) and Bardeen (1906), which were among the very few ones focusing on all the regions of each limb, had a crown-rump length (CR) of 18 mm, whereas the older (13 GW) fetus analyzed for the present work had CR80 mm (for details about our terminology regarding CR and GW, and how it applies to individuals described by other authors, see Materials and Methods). Even in the more detailed study of Cihak (1972), which was only focused on the development of the hand and foot, the older fetus described was only CR40 mm (10 GW).

We compare our observations with those of previous authors and we contrast the data obtained for upper versus lower limbs. This provides insight into the evolution, variations and pathologies of limb muscles and we discuss implications for the ontogeny and phylogeny of the two limbs, as well as the relationship between normal, variant and abnormal phenotypes.

Before describing the observations of the embryos and fetuses analyzed by us (Figs 3-6; Tables 1-5), which are summarized in the schemes of Figs 7 and 8, we refer the reader to the Materials and Methods section for important details of the terminology. The full list of the embryos and fetuses analyzed for the present work, their CR, and the GW that were assigned to them in the work by Belle et al. (2017) versus the present work is given in Table 1, along with the primary antibodies – myosin heavy chain, Myog and Pax7 – that were used for the immunostaining. For readers less familiar with the normal configuration and attachments of the adult human limb muscles, we show the adult muscles of the hand and foot (the main focus of this paper) in Figs 1 and 2 and further information is provided in Tables S1-S11. In the following sections, we describe key findings from our analysis, particularly focusing on muscles not previously mentioned and information that might help to resolve previously controversial or unclear issues.

Considering the muscles of the pectoral girdle and arm (Fig. 3A), current tetrapod comparative anatomical data indicates that the teres minor derives from the deltoideus (reviewed by Diogo et al., 2018). Both muscles are innervated by the axillary nerve in human adults, also supporting this view. However, Lewis (1901, 1910), influenced by the evolutionary ideas of his time, suggested that the teres minor was instead derived from the primordium that gives rise to the infraspinatus (although he recognized that he did not truly see the teres minor being differentiated from the infraspinatus in the human embryos analyzed by him). Now, in the upper limbs of the CR25 mm embryos (Fig. 3A) it can be seen clearly that the teres minor derives from the deltoideus, because the teres minor is mainly formed by fibers that are continuous with those of the deltoideus and is quite distant from – i.e. more inferior than – the infraspinatus. To our knowledge, this study is the first to show clearly that the adult human configuration in which the deltoideus is close to, and often blended with, the infraspinatus is acquired only at later stages of human development (Table 2). This is consistent with observations that have been made in developmental studies of other tetrapods (see ‘Comparison with other tetrapods…’ section, below).

Concerning the muscles of the forearm (Figs 3, 4), to our knowledge our study is also the first to show that the epitrochleoanconeus muscle, which is present in chimpanzees but absent in the adult human (Diogo and Wood, 2011, 2012), is present in early human ontogeny and derived from the primordium that also gives rise to the flexor carpi ulnaris, as suggested by comparative anatomy (see recent review by Diogo et al., 2018). The epitrochleoanconeus can clearly be seen in CR25 mm, CR30.5 mm and CR33.5 mm embryos (Fig. 3B), so it is present at least until the last stages of embryonic development (Table 2). This illustrates an example of an atavistic muscle that is present in early human development and then becomes lost later in development; it would be interesting to test this muscle for apoptotic markers in the future.

Regarding the hand, the configuration seen in our embryos and fetuses, particularly in the hand of CR25 mm embryos (Fig. 3C), supports the statement by Cihak (1972) that by the end of GW7 and beginning of GW8 the flexores breves profundi (fbp) are part of the same primordium that also gives rise to the intermetacarpales, and that the dorsometacarpales (‘interossei dorsales accessorii’ of Cihak) lie at a much more posterior (‘dorsal’) level, i.e. at a level similar to that of the short extensors of the forearm (which give rise to muscles such as the extensor pollicis longus and extensor indicis; Table 2). The dorsometacarpales are part of the amniote – and perhaps the tetrapod – Bauplan (body plan) (Table 6). Cihak and colleagues studied in detail the innervation of the dorsometacarpales not only in human embryos and fetuses but also in human adults, in which, as a variation, the dorsometacarpales are present as separate muscles (Cihak, 1972). His observations indicate that they are almost always innervated mainly by the deep ulnar nerve, although in some cases their innervation is by an anastomosis formed by this nerve and the radial nerve (Cihak, 1972). Cihak thus argued that the dorsometacarpales are primarily intrinsic hand muscles, because the deep ulnar nerve innervates exclusively intrinsic hand nerves, whereas the extensor forearm musculature is innervated by the radial nerve. We therefore follow Cihak's idea in Tables 2-7, i.e. the dorsometacarpales are included in the hand musculature.

However, comparative anatomical studies (e.g. see Diogo et al., 2018) and also recent ontogenetic studies of frogs (Diogo and Ziermann, 2014) indicate that the dorsometacarpales might in fact be part of the forearm extensor musculature. Importantly, this idea is supported by the present work, because in early human ontogeny the dorsometacarpales lie near to and at the level of the short extensors (Fig. 3C). Therefore, it is possible that the contribution of the ulnar nerve to the dorsometacarpales, as well as their association with the dorsal interossei, is in fact secondary within tetrapods. In fact, regarding the foot, Cihak recognizes that normally in human embryos and fetuses the dorsometatarsales are innervated by an anastomosis made partially by the lateral plantar nerve (which innervates only foot muscles) and mainly by the deep fibular nerve (which innervates the leg extensor musculature). Thus (see also Fig. 6), the innervation patterns of the foot seem to indicate that the dorsometatarsales are primarily part of the extensor leg musculature, and not of the foot musculature as suggested by Cihak. Clearly, more detailed studies on this subject, including of the innervation patterns of both the dorsometacarpales and dorsometatarsales of embryos and fetuses of other mammalian and non-mammalian taxa, are needed to clarify whether these muscles are primarily hand/foot or instead part of the extensor forearm/leg musculature.

Based on his histological observations of human embryos, Cihak (1972) assumed that there is no dorsometacarpalis 1 in human development and thus that the 1st dorsal interosseous of the human hand is formed only by the fbp 2 and 3 (that is, the short deep flexors that go respectively to the ulnar side of digit 1 and the radial side of digit 2 in most tetrapods). Our detailed observations of the region of the 1st dorsal interosseous in human embryos and fetuses, and particularly of CR36 mm fetuses (Figs 4, 5), support the suggestions of Cihak (1972) on this point: the fbp 2 clearly seems to at least partially contribute to the 1st dorsal interosseus (see also Table 3). However, Cihak's conclusions about the absence of the dorsometacarpalis 1 in human development are not supported, as the muscle is consistently found in our embryos and fetuses up to CR36 mm, although at this latter fetal stage the muscle is seemingly degenerating, being completely absent in our CR51 mm fetuses (Fig. 4; Table 3). These data indicate that the dorsometacarpalis 1 is present in the embryonic, and early fetal, development of humans. Therefore, we cannot exclude the hypothesis that the intermetacarpalis 1 – also said to be completely absent in human ontogeny by Cihak – is present in early stages and then degenerates or becomes integrated in the 1st dorsal interosseous of the hand. What seems to be clear is that the first dorsal interosseous is anyway different from the other dorsal interossei of the hand because it does seem to (1) include two fbp (i.e. fbp 3, and likely fbp 2), and (2) not include a well-developed dorsometacarpalis; in contrast, the other dorsal interossei of the hand are formed by one fbp, one intermetacarpalis, and often also a dorsometacarpalis (see Table 3). That fbp 2 of the hand is included in the 1st dorsal interosseous (and thus does not contribute to the flexor pollicis brevis) could also help explain why both the deep and superficial heads of the adult flexor pollicis brevis are so blended to each other and often are mostly inserted onto the radial side of the thumb: both heads would seemingly derive (as does the opponens pollicis) from fbp 1, which precisely goes to the radial side of digit 1 in most tetrapods. In fact, our observations show that this adult configuration of the flexor pollicis brevis, including the close association between its two heads, is already seen in early stages of human development (e.g. Figs 4, 5). Comparison of the hand muscles of a CR36 mm fetus with those of an older CR42 mm fetus indicates that both the deep and superficial heads of the flexor pollicis brevis derive (together with the opponens pollicis) from a single muscle structure (i.e. fbp 1) lying just ulnar to the abductor pollicis brevis (Table 3).

Our observations also support statements by Gräfenberg (1905) that in early human ontogeny (e.g. CR25 mm) the pronator quadratus is much broader – its fibers extending even more proximally than those of the flexor carpi ulnaris as can be seen in Fig. 3B – than in adults. Data particularly from the CR42 mm and CR51 mm fetuses (Fig. 4B,C) also indicate that the ‘Henle’ muscle of the hand is likely not derived from the flexor pollicis brevis, as suggested by Cihak (1972), but instead from the oblique head of the adductor accessorius (see Fig. 4 and Table 2). This supports the idea advanced by Bello-Hellegouarch et al. (2013) that the ‘Henle’ muscle should be named adductor pollicis accessorius, a terminology we follow in the present work.

To summarize our data on the upper limb, we show that: (1) the adult human configuration in which the deltoideus is so close to, and often blended with, the infraspinatus is acquired only at later stages of human development; (2) the muscle epitrochleoanconeus is consistently present in early human ontogeny; and (3) the dorsometacarpalis 1 is present in the embryonic, and early fetal, development of humans. In addition, our study also supports the suggestions that the dorsometacarpales are part of the forearm extensor musculature, and that the ‘Henle’ muscle of the hand is likely derived from the oblique head of the adductor accessorius.

Concerning the muscles of the pelvic girdle, thigh and leg, in all the stages observed by us these muscles were similar to the description of the later stages by Bardeen (1906) and Lewis (1910). We summarize the timings of appearance, splitting, fusion and loss of these muscles in Table 4, and provide additional notes about them in that table. Here, we focus on the leg musculature that has been the subject of some controversies in the embryological literature. For instance, the fibularis tertius is often said to be derived from the extensor digitorum longus, consistent with comparative data, although some authors state that it might instead derive from the short, deep muscles of the extensor compartment of the leg (reviewed by Diogo et al., 2018). Interestingly, in a CR46 mm fetus observed by us we saw a curious configuration (Fig. 6D): the part of the extensor digitorum brevis that goes to digit 4 is continuous with a very thin muscle structure that is fibular to it and that, in terms of its distal portion, lies roughly where the fibularis tertius of adults would normally lie, i.e. just above the region of metatarsal 5. This observation could therefore be used to support an origin of the fibularis tertius from the short extensors, and in particular a homology between this muscle and part or all of the short extensor of digit 5 of other tetrapods (which is seemingly absent as a fleshy structure in adult humans because the extensor digitorum brevis goes only to digits 2, 3 and 4). However, the direction of this thin muscle structure seen in Fig. 6D and the position of its proximal portion is very different from that of the adult fibularis tertius, because the latter structure is usually proximally attached to the tendon of the extensor digitorum longus at a much more tibial level and accordingly runs more fibulo-distally, whereas the thin structure of the fetus runs basically distally in a straight line with a fibular position (Fig. 6D). Moreover, in the next stage examined by us (CR51 mm), no such thin muscle structure is found. Therefore, the overall analysis of the available data indicates that this thin structure might be a case of an anatomical variation seen in a single fetus, which would be an interesting observation by itself as almost no studies have described, or discussed, variations in embryonic/fetal stages.

Regarding the foot, Cihak (1972), Bardeen (1906) and Lewis (1910) referred to the presence of the atavistic muscle opponens digiti minimi in early human ontogeny, but they did not specify until which stage the muscle is consistently present. Our study reveals that the opponens digiti minimi of the foot is present (e.g. Fig. 6) until CR51 mm, i.e. GW11, which is striking for a vestigial muscle, although at this latter stage the muscle is reduced to a very small structure. More details about the timing of appearance of the blastema that gives rise to these two muscles, as well as of their differentiation, is given in Table 5.

Also concerning the foot, in our observations up to CR36 mm the dorsometatarsales 1, 2, 3 and 4 are clearly present, contradicting the statement of Cihak (1972) that only 2 and 3 develop in humans, although 1 and 2 are much smaller than the other ones, being almost degenerated by CR36 mm (Fig. 6A-C; Table 5). Our observations also contradict the statements of Cihak concerning the specific muscles that fuse to form each of the four dorsal interossei of the foot. For example, at CR36 mm one can still see (Fig. 6C) that the 1st dorsal interosseous clearly seems to be formed by at least two structures: a broader dorsal structure plus a smaller plantar structure, which are strikingly similar, respectively, to the broader dorsal intermetatarsales 2, 3 and 4 and smaller plantar fbp 4, 6 and 8 that seem to form the dorsal interossei 2, 3 and 4 of the foot. Therefore, our observations not only indicate that the 1st dorsal interosseous of the foot includes the fbp 3+intermetatarsalis 1 and not the fbp 3+4, as proposed by Cihak, but also that the 2nd dorsal interosseous of the foot includes the fbp 4+intermetatarsalis 2 and not only an intermetatarsalis as suggested by Cihak. Also, our observations indicate that the fbp 2 may either be reduced/lost or, more likely, correspond to the lateral head of the flexor hallucis brevis, whereas Cihak suggested that it might form the so-called ‘Henle’ muscle of the foot, which we were not able to observe in our samples but list as adductor hallucis accessorius in Table 5 following current terminology (because it topologically corresponds to the adductor pollicis accessorius of the hand; see above and Diogo et al., 2013, 2018; Diogo and Molnar, 2014).

In summary, the most important aspect concerning our observations of the development of the lower limb is that our study is the first to show that the atavistic muscle opponens digiti minimi of the foot is present until CR51 mm.

As noted above, the adductor hallucis accessorius of the foot probably derives from the oblique head of the adductor hallucis, as the adductor pollicis accessorius in the hand seems to derive from the oblique head of the adductor pollicis (Tables 3, 5). However, it is important to emphasize that there are cases in which there are substantial differences between the units that form, or the primordia from which derive, specific muscles of the adult human hand versus the foot muscles to which they correspond topologically. Before providing a summary of these differences, we note that many of the so-called ‘differences’ suggested by Cihak (1972) – particularly concerning muscles that are too small and/or blended in early stages to be analyzed in the necessary detail in the 2D histological images that he analyzed – seem far-stretched. In fact, Cihak himself seems confused and skeptical about his own descriptions of the fbp and intermetatarsales of the foot.

First, the flexor hallucis brevis of the foot seems to be derived from fbp 1 and possibly from at least part of fbp 2 (as also indicated by the fact that its medial and lateral heads attach, respectively, onto the tibial and fibular sides of the big toe), whereas the flexor pollicis brevis of the hand seems to be mainly derived only by fbp 1 (as also indicated by the fact that both its superficial and deep heads attach mainly on the radial side of the thumb) although a contribution of fbp 2 cannot be discarded (Tables 3, 5). Second, the 1st dorsal interosseous of the foot seems to be formed by fbp 3+intermetatarsalis 1, plus possibly a contribution of fbp 2, whereas in the hand the 1st dorsal interosseous is likely formed by fbp 2+fbp 3, although a contribution from intermetacarpalis 1 and/or a non-contribution from fbp 2 cannot be discarded (Tables 3, 5). Third, the 2nd, 3rd and 4th dorsal interossei of the foot seem to be formed by one intermetatarsalis and one flexor brevis profundus as do those of the hand, but whereas the hand ones are seemingly formed, respectively, by the fusions of intermetacarpales 2, 3 and 4 with fbp 5, 6 and 8, those of the foot are formed, respectively, by the fusions of intermetatarsales 2, 3 and 4 with fbp 4, 6 and 8. This latter difference is related to a major functional difference between the hand and the foot: in the hand, digit 3 is the axis (thence the dorsal interossei go to the radial sides of digits 2 and 3 and ulnar sides of digits 3 and 4 to abduct the digits), whereas in the foot digit 2 is the axis (thence the dorsal interossei go to the tibial sides of digits 2 and fibular sides of digits 2, 3 and 4 to abduct the digits). Accordingly, a fourth muscle difference is that whereas in the hand both lumbricals 1 and 2 (i.e. those that go to radial sides of digits 2 and 3) are innervated by the median nerve because they are radial to the axis of the hand, in the foot only lumbrical 1 (i.e. the one that goes to the tibial side of digit 2) is innervated by the medial plantar nerve (topologically similar to the median nerve in the hand) because this lumbrical is the only one that is tibial to the axis of the foot.

What is particularly remarkable about the four differences summarized in the above paragraph is that they concern differences that are not so evident in a gross anatomical observation of the adult hand versus foot muscles. That is, the muscles of the adult hand that were mentioned are topologically similar to those of the adult foot, but specific details on their ontogeny, innervation and/or attachment reveal that they are not as similar as they appear superficially. Apart from those four differences, a fifth could be added, based on observations of Cihak (1972): according to him, the contrahentes 3, 4 and 5 of the foot become completely lost during ontogeny, whereas the contrahentes 3, 4 and 5 of the hand fuse with the interossei (note that the contrahentes 1 and 2 are included in the adductor pollicis in the hand and in the adductor hallucis in the foot; Tables 3, 5).

Of course, there are other muscle differences between the human hand and foot, but these can be clearly seen in a superficial gross anatomical comparison in adults. For instance, whereas in the hand the opponens pollicis and opponens digiti minimi are present as the normal phenotype from early to adult stages, in the foot the opponens hallucis is seemingly never differentiated and the opponens digiti minimi is a vestigial muscle, present in early stages but disappearing during fetal development. Also, whereas the flexor digitorum superficialis of the adult upper limb spans all the way from the distal humerus to the digits, the flexor digitorum brevis of the lower limb lies exclusively in the foot. Lastly, there is a quadratus plantae only in the foot, and there is a palmaris brevis only in the hand (Tables 3-5).

These differences support the emerging idea that the topological similarities between the hand and foot of tetrapods, such as humans, are mainly secondary (see recent reviews by Diogo et al., 2013, 2018; Diogo and Molnar, 2014; Sears et al., 2015; Miyashita and Diogo, 2016). This idea is further supported by the fact that the order of developmental appearance of the hand muscles is markedly different from that of the corresponding foot muscles (Tables 6, 7). As an illustrative example, whereas the lumbricales are the first muscles to differentiate in the hand, together with the contrahentes (Table 6), in the foot the lumbricales differentiate only after most other foot muscles are already differentiated (Table 7). Thus, these developmental data and evidence from comparative anatomy and from the evolutionary history of human limb muscles (see Tables 6, 7) indicate that several of the muscles that seem to be topologically similar in the human upper and lower limbs actually appeared at different evolutionary times; appear in a markedly different ontogenetic order; derive from different primordia; and/or are formed by the fusion of different developmental units in each limb.

As summarized in Tables 2-5 and also noted above, various atavistic muscles that were present in the normal phenotype of our ancestors are present as the normal phenotype during early human ontogenetic stages and then disappear or become reduced and completely fused with other muscles, thus not being present/distinguishable in human adults. These include the upper limb muscles epitrochleoanconeus (Fig. 3), dorsoepitrochlearis, contrahentes 3-5 (Fig. 4) and dorsometacarpales 1-4 (Figs 3-5), and the lower limb muscles contrahentes 3-5, dorsometatarsales 1-4 (Fig. 6) and opponens digiti minimi (Fig. 6). These muscles are present in some other tetrapods, as shown in Tables 6 and 7, which summarize the comparisons with other limbed vertebrates. Of all these muscles, only the dorsometacarpales often remain in adults, fused with other muscles: all the others are normally completely absent in human adults. Fascinatingly, all these atavistic muscles are found both as rare variations of the normal adult population and as anomalies in individuals with congenital malformations such as those associated with trisomies 13, 18 and 21, reinforcing the idea that such variations and anomalies can be related to delayed or arrested development (e.g. Wood, 1864, 1865, 1866, 1867a,b, 1868, 1870a,b; Macalister, 1866, 1867, 1875; Testut, 1888; Loth, 1912, 1931; Barash et al., 1970; Bersu and Ramirez-Castro, 1977; Ramirez-Castro and Bersu, 1978; Pettersen, 1979; Pettersen et al., 1979; Stark et al., 1979; Aziz, 1979, 1980, 1981a,b; Bersu and Optiz, 1980; Dunlap et al., 1986; Bergman et al., 1988; Smith et al., 2015; Diogo et al., 2016).

The occurrence of such atavistic muscles in human embryonic/fetal development can be used to support the idea that there are at least some cases of parallelism between phylogeny and development (i.e. ‘phylo-devo’ parallelism described by Diogo et al., 2015). One interesting question is: what are the functional consequences, and evolutionary implications, of the presence of such atavistic muscles as the normal phenotype in human embryos, and as the variant or abnormal phenotype in later human development? This subject was discussed by Diogo et al. (2015, 2016). A major implication is that the presence of such muscles as the normal phenotype of human embryos supports an ‘ontogenetically constrained’ (internalist) view of evolution, as proposed by authors such as Gould (1977, 2002) and Alberch (1989), rather than an ‘adaptationist’ (externalist) evolutionary view. When these atavistic muscles are present postnatally, either as a variation of the karyotypically normal population or as abnormalities within the karyotypically abnormal population, they seem to be mainly functionally neutral, not providing any type of major functional advantage or disadvantage (e.g. Diogo et al., 2015, 2016).

Other examples of ‘phylo-devo’ parallelism provided by the developmental data compiled in the present paper concern the comparison of the order in which the muscles of a certain region/module of a limb (e.g. forearm or leg) appear in evolution versus the order in which they appear in development (Tables 6, 7). An illustrative example is the fact that four of the last nine forearm muscles to differentiate (within a total of 22 forearm muscles that are differentiated) in human ontogeny are also among the last ones to have appeared in evolution: the flexor pollicis longus and extensor pollicis brevis, which appeared only during human evolution, and the extensor carpi radialis brevis and extensor carpi radialis longus, which appeared only in mammalian evolution (Tables 6, 7). Illustrative examples seen in the lower limb are that the sartorius was acquired in amniotes and is accordingly already differentiated at very early stages of human development, whereas muscles that evolved much later, such as the gemellus inferior and superior and the obturator internus (only acquired in the last common ancestor of marsupial and placental mammals), appear much later in human ontogeny (Tables 6, 7).

However, it is important to note that even within the same anatomical region there are a few clear examples that do not match a ‘phylo-devo’ parallelism. For instance, within the pectoral girdle muscles the rhomboid complex evolved later than the serratus anterior and levator scapulae but in human ontogeny it differentiates before the split between those two muscles (Tables 6, 7). Also, within the flexor compartment of the leg, the plantaris is the last muscle to differentiate in human ontogeny but in evolution it evolved before the soleus and the flexor hallucis longus (Tables 6, 7).

Even within a single region/module what one sees is in reality a mosaic, with different muscles of a certain region at a certain developmental stage resembling adult configurations that were acquired in very different phylogenetic nodes and thus at very different geological times. An illustrative example concerns the pectoral and arm muscles. The axial pectoral muscles (as defined by Diogo and Abdala, 2010) at very early developmental stages (CR9-10.5 mm) resemble the adult human configuration in the sense that there is a subclavius but seemingly no levator claviculae, but concerning other muscles they resemble a more ancestral mammalian configuration (Diogo et al., 2018) as there is only a rhomboid complex, a levator scapulae and a serratus anterior (Tables 2, 6). Regarding the pectoralis major and minor and infraspinatus and supraspinatus, those early stages resemble again the adult human condition as there is, for instance, no panniculus carnosus, but the presence of a dorsoepitrochlearis and of the teres minor resembles the basal marsupial/placental condition (Diogo et al., 2018), as does the presence of three clearly separate deltoids (as seen, for example, in Fig. 3; see also Tables 2, 6). With respect to the presence of only two major anterior arm muscle masses – the ‘biceps/coracobrachialis’ and the ‘brachialis’ described by Lewis (1901, 1910) – those early stages actually seem to resemble topologically the humeroantebrachialis and coracobrachialis of adult amphibians (Diogo et al., 2018).

However, such comparisons also call attention to biases – particularly seen when Haeckel's ‘recapitulation’ theory (see Diogo et al., 2015) was widely accepted – in trying to establish parallels between the embryonic/fetal configuration found in humans and configurations found in adult stages of other animals. For instance, this topological similarity has been used to suggest that the coracobrachialis gave rise to the coracobrachialis+short head of the biceps brachii of amniotes, and that the humeroantebrachials gave rise to the brachialis and the long head of the biceps brachii of amniotes (see reviews by Lewis, 1901, 1910 on these topics). But, in fact, such similarity might not reflect a true homology because human embryos do not have a coracoradialis muscle as adult amphibians do, and it was this latter muscle that likely gave rise to part of the biceps of amniotes, a point that was neglected by most authors in the past (see Diogo and Tanaka, 2012). Moreover, Lewis (1901, 1910) noted that during early human ontogeny the short and long heads of the biceps brachii actually originate closely together from a similar region of the scapula, near the point of origin of the coracobrachialis, and that it is the subsequent growth of the scapula that results in the proximal attachment of the two biceps heads being so far apart in adults, thus supporting the idea that the long head of the biceps does not derive from the humeroantebrachialis/brachialis. That is, apart from comparative anatomy, a detailed, unbiased look at human ontogeny per se also reveals that the anterior arm muscles of early human developmental stages do not directly correspond to those of adult amphibians.

In fact, any type of possible ‘phylo-devo’ parallelism between the order in which muscles appeared in evolution versus the order in which they appear in human ontogeny is further called into question when one takes into account not only a certain region/module of a limb, but instead the limb as a whole. One illustrative example is that various adult human autopod (hand/foot) muscles were already present in the last common ancestor of extant tetrapods or amniotes, but ontogenetically they appear after the girdle and stylopod (arm/thigh) muscles, which appeared much later in evolution (Tables 6, 7). This is obviously related to the fact that the autopods appear in development much later than the more proximal regions of the limbs. Of course, one could argue that this order of events resembles the order of events seen in phylogeny, because proximal structures, such as those of the girdles, evolved before (i.e. early in gnathostome evolution) those of the autopods (which mainly appeared during the transition from fish to tetrapods). However, such a resemblance does not apply to individual muscles, because none of those muscles associated with the girdles in basal gnathostomes directly corresponds to those girdle muscles found in modern humans. As can be seen in Tables 6 and 7, the reality is that within a limb as a whole the ontogenetic order of muscle appearance is more related to modularity (i.e. being part of a certain region/module) and to the proximo-distal development of the limb, than to the order of evolutionary events. Thus, when one takes into account the limb as a whole, the data contradict (1) the ‘recapitulation’ of Haeckel, (2) the notion of ‘phylo-devo’ parallelism proposed by authors such as Gould (1977) and defended by Diogo et al. (2015) on the basis of comparisons within only a certain region/module of a limb, and (3) also the older ideas from Van Baer (see Diogo et al., 2015) that ontogeny normally proceeds from the general to the peculiar. For instance, muscles such as the rhomboideus major and minor are very peculiar and derived phylogenetically (being only found in a very small subset of placental mammals) but are already differentiated in CR14 mm human embryos, whereas muscles such as the fbp, intermetacarpales and abductor digiti minimi of the hand are clearly part of the ‘general’ tetrapod Bauplan but only appear after that embryonic stage, at CR15 mm or even at CR16 mm (Tables 6, 7).

In summary, a major contribution of the developmental observations made in the present work and their comparisons with what is known from the development and evolution of other tetrapods, is that overall there is no strong ‘phylo-devo’ parallelism between the order in which muscles appeared in evolution versus the order in which they appear in human ontogeny. We therefore hope that this work will not only contribute to the understanding of limb muscle development in humans and in tetrapods in general, but also pave the way for, and stimulate, other researchers to undertake deeper and broader discussions on the links between the upper and lower limbs, between atavisms, variations and anomalies, and between phylogeny and evolution.

For details on the embryos used for the analysis presented in the paper, see Table 1, which also lists the primary antibodies used: myosin heavy chain (MHC), myogenin (Myog) and Pax7 (though note that only MHC images are shown in the figures). It is important to emphasize that the present work concerns simply the analysis – using Imaris – and comparison and labeling (using Microsoft PowerPoint) of the 3D images obtained in the project described by Belle et al. (2017) that showed embryonic and fetal limb muscles. Therefore, for any details regarding the methods used by these authors to perform the immunostaining of muscles and to create 3D images from it, readers should refer to the materials and methods information provided in that paper. The identification of the limb muscles seen in those 3D images is based on our extensive experience of identifying these muscles both in later human developmental stages and in early developmental stages of many other animals (see, e.g. Diogo et al., 2008, 2018; Diogo and Abdala, 2010; Diogo and Ziermann, 2014; Ziermann and Diogo, 2014). More details about the specific methodology we use are extensively described in a recent book (Diogo et al., 2018).

As our aim is that our work is used not only by developmental biologists and comparative anatomists, but also by students, professors, physicians and the broader public in general, we used a crown-rump length (CR)/gestational weeks (GW) chart that is highly consensual within and used by pediatricians in the USA, including the fact that the transition from an embryo to a fetus is considered to occur at 10 GW (the full chart is given here: https://www.babymed.com/fetus-crown-rump-length-crl-measurements-ultrasound). Therefore, some of the numbers concerning GW used in Belle et al. (2017) paper are slightly different from those used in the present paper, because following the tradition of comparative embryological works we give preponderance to CR length, more than to GW, as is also reflected in Figs 3-6 and Tables 2-7. Accordingly, it should be noted that within the two measures given by Bardeen and Lewis (1901) – which were then applied by Lewis (1901, 1910) and Bardeen (1906) – their ‘NB length’, rather than their ‘VB length’, corresponds to the CR length of current works and thus of the present work. Also, for our data to be also useful to students, professors, researchers and physicians focused on human anatomy, variations and pathologies, we apply here the official human anatomical nomenclature to the embryonic and fetal stages observed by us, e.g. the anterior and posterior regions of the body and limbs correspond respectively to the ‘ventral’ and ‘dorsal’ commonly used in non-human model organisms (for more details, see Diogo et al., 2016; Diogo and Molnar, 2016).

We want to particularly thank Morgane Belle and Alain Chedotal for allowing us to use the images produced in their project, and also for providing extra information about the individuals observed by us and about the techniques used to stain their muscles.