The brachial plexus – explaining its morphology and variability by a generic developmental model

Abstract

In classic anatomy teaching, the brachial plexus generally features as an enigmatic rote‐learned structure, leaving the student with a feeling of complexity. The notion of complexity may increase in dissections, where plexuses significantly differing from the standard plexus model are readily found. This raises questions: what determines the existence and prevalence of variants and to what degree should they be considered anomalous? A model linking brachial plexus morphology and its variability to causative morphological parameters which would also standardize plexus description and teaching would be beneficial. The present study aims to provide such a model by analyzing the context of plexus development and applying this model in the analysis of plexus variability in anatomical specimens. Based on a thorough literature review, a generic developmental model was formulated and different factors of variability defined. In 56 plexuses, the proposed generic principles of plexus variability were found consistent with the variations encountered. Summarized, the modeled generic principles are as follows. Brachial plexus axon bundles grow out into an environment of chemical tracer paths in which constraints and obstacles are present: the geometry of the arm bud, cartilaginous bone precursors and vessels. The overall constancy of these factors generates a gross plexus outline, while the variability in these factors gives rise to typical plexus variations. The usefulness of the model derives from the fact that the variability of the main morphologically determining factors is not random but is the expression of the possibilities of the embryological substrate. Within the model, the major plexus morphological determinant is the segmental position of the subclavian artery, which is determined by the segment level of the intersegmental artery from which it develops. Normally, the subclavian artery develops from intersegmental artery i7. However, the subclavian artery can develop from inferior or superior segmental levels, from intersegmental artery i8 or i6, and possibly also from i9 or i5. Each of these arterial variants creates a typical, morphologically distinct, predictable plexus configuration. Superimposed on these basic plexus configurations, the underlying embryological substrate may develop further variability by integrating remnants of other intersegmental arteries into the arterial network. The resulting plexus configurations are further modified by local factors, e.g. the splitting of outgrowing axon bundles around vessels. A large split in the lateral cord around a large vein or veins crossing from lateral to medial, tangentially cranially over the subclavian artery was found in 54% of the 56 investigated BP and therefore might be added to plexus teaching. The distinct plexus morphologies associated with the subclavian artery segmental levels were further found associated with, among others, typical variations in the pectoral nerves and their ansas; these associations were also modeled. The presented models could allow brachial plexus rote learning to be replaced by a more insightful narrative of formative principles suitable for teaching. Clinically, improved understanding of the relationship between plexus variability and the local anatomical environment should be relevant to brachial plexus surgery and reconstruction.

This paper presents an integrated model of the brachial plexus and its variability by considering the embryonic brachial axon outgrowth in the arm bud. The outgrowing axons encounter obstacles – the bud geometry, the precursor brachial skeleton and the preexisting vascular system. The constancy and variability of these obstacles, which are not random but are determined by the possibilities of the embryonic stratum, closely determine brachial plexus morphology and its variations.

Introduction

The brachial plexus (BP) is anatomically defined as the nerve network originating from the ventral rami of the spinal nerves C4 or C5 to T1 or T2, situated between the respective intervertebral foramina and the upper arm (Standring, 2016). In this network, nerves leaving the vertebral outlets combine and divide twice to produce the brachial nerves. To deal with this complexity, a standard anatomical nomenclature was defined (Fig. 1). The ventral rami of the nerves leaving the vertebral outlets are called ‘roots’. The roots contain the somatic and sympathetic motor and sensory axons to the periphery. Distal to the outlets, these roots combine into ‘trunks’. These trunks split into anterior and posterior ‘divisions’, which then recombine into anterior and posterior ‘cords’, again partly combining and splitting to form the terminal brachial nerves: median, ulnar, musculocutaneous and cutaneous nerves from the lateral and medial anterior cords, and radial and axillary nerves from the posterior cord. In the process, nerves leave the plexus to the scalene, serratus anterior, subclavius, latissimus dorsi, pectoral and scapular muscles, as well as to the skin of the axillary region.

Figure 1.

Standard model and terminology of the brachial plexus. The C5 and C6, and C8 and T1 nerve roots, respectively, merge into trunks. These trunks divide into anterior and posterior divisions. The posterior division branches merge into a single posterior cord (PC). The cranial two anterior division branches merge into the lateral cord (LC), which merges with the caudal anterior division branch, called the medial cord (MC), into a single nerve (the median nerve M). Therefore, both divisions ultimately merge into single nerves, but the anterior division merges in two stages, whereas the posterior division merges more or less symmetrically in one stage. DS, SS, SC, LT: dorsal scapular, suprascapular, subclavius, and long thoracic nerves. P: pectoral nerves, merged from a lateral and a medial branch. R, A: radial and axillary nerve. MCN, U: musculocutaneous and ulnar nerves. US, LS, TD: upper and lower subscapular nerve and thoracodorsal nerve. MCB, MCA: medial cutaneous brachii and antebrachii nerves.

However, within this standard view, the plexus remains a structure of enigmatic complexity, which students generally learn by rote and thereafter readily forget. Yet, the surgeon will find its variability far greater than the textbooks suggest, as reflected by the documented complexity of plexus reconstruction (Kerr, 1918; Don Griot et al. 2002; Das & Paul, 2005; Prakash et al. 2006; Singhal et al. 2007; Vilamere et al. 2008; Aggarwal et al. 2009, 2010, 2012; Yildiz et al. 2011; Adam et al. 2011; Muthoka et al. 2011; Chaudhary et al. 2012; Chaware et al. 2012; Wozniak et al. 2012; Van de Velde et al. 2013, 2016; Aragao et al. 2014; Bhosale & Havaldar, 2014; Govindarajan et al. 2014; Mian et al. 2014; Claassen et al. 2016; Emamhadi et al. 2016; Leonhard et al. 2016; Padur et al. 2016; Tubbs et al. 2016).

The functional morphologist or evolutionary biologist may wonder what evolutionary benefits determined this complexity, and the embryologist faces the question of how it grows as such. To facilitate estimation of plexus variability, the recognition of an underlying developmental pattern would be beneficial.

In the following, a generic developmental model of BP is presented that combines the content of morphology, evolutionary biology and embryology with additional findings into a synthesis that might be useful for anatomical education and that may complement the literature on BP variability for clinical purposes.

The model relates the embryonic outgrowth of axons in the arm bud to the adult morphology and variability. In development, the plexus is considered an axon front that grows out from the spine into the arm bud in a tracer‐guided environment in which spatial constraints or obstacles exist or develop simultaneously. These constraints or obstacles include the arm bud dimensions, the preexisting vascular system and the cartilaginous precursor arm skeleton. The plexus mergers and divisions emerge as the combined imprint of these generic elements and typical anatomical variability follows naturally from variations in the location and morphology of the obstacles.

Materials and methods

Generic developmental model of the formation of the brachial plexus

Following a thorough literature review on the embryology of the BP and possible influencing structures as the development of osseous, muscular and vascular structures, a model was designed to reflect the developmental causes of BP morphology and variability. This model will be presented under Results and used to interpret the dissection findings.

Model validation

For model validation, 56 BP of 35 body donors were investigated. The specimens originated from voluntary body donations to the Ghent University Body Donation Program, approved by the Institutional Ethical Committee.

Six BP in five Thiel‐embalmed specimens were dissected in detail, including the pectoral nerves. The investigated shoulders were amputated with the vertebrae C3–T3, the first three ribs, upper aorta, scapula, and two‐thirds of the humerus. Peripheral to the vertebrae, all nerves and vessels were carefully dissected to halfway up the upper arm. To gain access to the axilla, the clavicula was excised and the pectoralis major and minor were dissected from their origin at the thorax, leaving their insertions and nerves intact. The neurovascular structures were then extensively photographically documented in situ. Thereafter, the scapula was excised at the shoulder joint while the upper arm was kept with the neurovascular bundles attached. The neurovascular plexus with the upper arm was spread on a flat surface for detailed photographs. Finally, the vascular system was removed and the nerve plexus was spread out and photographed as a mere nerve network, as in the standard anatomical presentation.

Fifty BPs in 30 zinc chloride‐embalmed bodies were grossly dissected during the Ghent University medical curriculum 2017 gross anatomy dissection class. Noted were the position of the subclavian artery in passing through the anterior division and the occurrence of a split in the lateral cord through which veins passed.

For clarity, the dissection photographs were enhanced by transparent color layers using Adobe illustrator® software. Arteries and veins were colored red and purple, respectively. For the nerves, the anterior divisions were tinted blue and the posterior divisions green (except in Fig. 11C, where the medial cord was colored green for contrast with the other anterior division branches). Pectoral nerves were colored yellow.

Results

Generic developmental model of the brachial plexus

The arm bud – stages 13–15

In the early embryo, at embryological stage 13 (28–32 days of development), the arm bud has been initiated and is vascularized by a small single central (axial) artery that feeds a largely symmetric network of capillaries (Rodriguez‐Niedenfuhr et al. 2001). Axons have started to grow in broad bundles through the sclerotomes, which are half permissive and half inhibitive to axon outgrowth. Proximally‐centrally in the bud, a cartilaginous mass already condenses, the precursor to the arm skeleton. At stage 15 (35–38 days of development), the bud has tripled in length and is still vascularized by a central artery and a largely symmetric capillary network. The cartilaginous mass has elongated with the bud, and the distal end is already bifurcated as the precursor to the radius and ulna.

Development of the ‘trunks’: axon bundle convergence towards the arm bud

At about stage 15, the anterior sclerotome axon bundles grow into the arm bud base (Fig. 2A). However, sclerotomes C5–T1 are longer than the width of the bud and therefore the outer outgrowing axon bundles C5 and T1 must converge sharply to enter the bud (Figs 2B and 3A). Since all anterior sclerotome axon bundles lie in the same plane, the outgrowing converging C5 and T1 axon bundles encounter the outgrowing axon bundles C6 and C8, respectively. The latter bundles also converge to enter the bud, but to a lesser extent. As the axon bundles C5 and C6, and T1 and C8, respectively, meet, their outgrowing axons intermingle (called ‘merging’ below). For better reflection of this process, these merged axon bundles will hereafter be referred to by the constituent sclerotome numbers: C56 and C8T1, respectively. The axon bundles C56 and C8T1, i.e. the superior and inferior ‘trunks’ according to common anatomical terminology, then continue into the bud on a converging course to the bud midline. Sclerotome C7, the medial ‘trunk’, lies essentially at the level of the bud axis, so its axons grow straight out into the bud without encountering and merging with neighboring axon bundles.

Figure 2.

3D reconstructions of segmented human embryos showing the development of the anterior‐posterior (A/P) BP divisions. Cranial views sectioned at about C4. Neural tube in green, skeleton in off‐white, nerves in yellow. Specifically, the embryos illustrate the precursor arm skeleton forcing the separation of the A/P divisions from the earliest axon outgrowth into the bud (from the 3D Atlas of Human Embryology, AMC, the Netherlands, http://www.3datlasofhumanembryology.com) (de Bakker et al. 2016). (A) Stage 15 (35–38 days of development), specimen 3512. The cartilaginous precursor arm skeleton was at this stage too unspecific to segment well. However, it already forms a central mass impenetrable to the axons at the arm bud base, around which the axons grow into an anterior and a posterior wing. Arrows indicate the imprint of the skeletal cartilage on the axon outgrowth. (B) Stage 16 (38–42 days of development), specimen 6517. The outgrowing axons closely split into an anterior and a posterior course around the cartilaginous humerus skeleton, now well segmented. (C) Stage 17 (42–44 days of development), specimen 6521. The axon growth fronts have passed the humerus anterior‐posteriorly, while the shoulder is outgrowing laterally.

Figure 3.

Generic model of BP. (A) In bud development, the bud profile forces extreme central convergence of outgrowing axon bundles C5 and T1, which meet and merge with axon bundles C6 and C8, respectively, which follow a less convergent course. C7 grows straight out into the bud. All axon bundles tend to converge towards a single bundle in the thoracic outlet. (B) While converging, all axon bundles synchronously divide into an anterior (AD) and a posterior division (PD) bundle, forced by the central cartilaginous arm skeleton precursor (Fig. 2A). While dividing, the AD and PD axon bundles keep converging towards the midline. By hypothesis, without obstacles, the AD and PD axon bundle central convergence would be symmetric. (C) Cranial view of BP. The subclavian artery (SA) (and major veins) run from a thoracic course anterior to the BP to a central course in the bud. The outgrowing AD axon bundles find the SA on their path and grow around it. The outgrowing PD axons remain posterior to SA and major veins. (D) Asymmetry in the convergence of AD and PD axon bundles caused by SA. Generally, SA lies between outgrowing C7 and C8 AD axon bundles (noted as C7a and C8a). The SA body forces the outgrowing C7a bundle cranially, where it merges with the descending C65a axon bundle into the C567a bundle. C567a curves cranially over the SA body towards the bud midline, whereas C8T1a curves caudally over the SA body. Distal to the SA body, C567a and C8T1a meet and merge into the C5678T1a (C5‐8T1a) axon bundle, which continues as the median nerve. This symmetric, elongated, stepwise central merging contrasts with the short central merging of the PD axon bundles into C5‐8T1p. On its way, C567a sheds the musculocutaneous nerve (MCN). C8T1a, possibly with proximally merging strands from C567a, sheds the ulnar (UN) and caudal cutaneous nerves. (E) By hypothesis, outgrowing axon bundles may split around vessels on their path. There is a high incidence of the C567a axon bundle (lateral cord) splitting around (a) vein(s) (see Results), which pass cranially‐tangentially over SA, so that the C567a axon bundle when curving cranially over the artery may meet and split around them (see Results). (F) Vessels or vessel plexuses may obstruct the merging of axon bundles, so that the merging takes place more distally. (G) Anterior thoracic division axon strands (yellow) arising from the AD axon bundles, mainly to the subscapularis and pectoralis nerves (see Fig. 6). (H) The combination of all the above principles leads to a complex, variable neurovascular network. (I) Removal of the vascular system, as in standard BP presentations, leaves an enigmatic nerve network far more variable than the standard model suggests.

Development of the anterior‐posterior divisions

Brachial divisions

In the embryo, the shoulder‐arm musculature develops from myogenic precursor cells originating from the hypaxial sub‐region of the dermomyotome which migrate after epithelial/mesenchymal transformation in a grossly posterior and an anterior layer at the base of the growing bud (Scaal & Christ, 2004). It is generally assumed that the anterior and posterior muscle layers cause an anterior/posterior division in the outgrowing sclerotome axon bundles (Landmesser, 1978; Lance‐Jones & Landmesser, 1981; Kania & Jessell, 2003; Luria & Laufer, 2007). However, the precursor cartilaginous arm skeleton constitutes a massive object centrally/proximally in the arm bud, impenetrable by axons (Del Rio & Soriano, 2007; Hughes et al. 2009; Beller & Snow, 2014; Snow, 2014). Therefore, when entering the bud, axons have to grow anteriorly (as anterior division, AD) or posteriorly (as posterior division, PD) around it; this division starts already at stage 15. This results in a clear‐cut anterior/posterior division of all sclerotome axon bundles at the point where they meet the cartilaginous humerus (Fig. 2A,B). (The close embrace of the cartilaginous humerus by the AD and PD axons was also documented by Rodriguez‐Niedenfuhr et al. 2001.) The division caused by the cartilaginous humerus will hereafter be referred to as ‘brachial division’. However, even while dividing posterior and anterior to the humerus, the axon bundles extending into the bud continue to converge centrally (Fig. 3B).

Thoracic divisions

Superimposed on the brachial divisions, anterior and posterior branches develop, consisting of axons that split from the anterior and posterior brachial divisions, respectively, towards the anterior (pectoral, subclavius) and posterior shoulder and superficial upper trunk muscles (scapular muscles, serratus anterior and latissimus dorsi) and skin. These will hereafter be called ‘thoracic divisions’.

The basic symmetry of the brachial division

In the model of Fig. 3B, the anterior (AD) and posterior (PD) divisions develop symmetrically and converge symmetrically towards a central anterior and posterior axon bundle. However, in the real plexus, only the PD converges in this way. The AD eventually also converges, with the median nerve receiving axons from all roots, but its convergence is highly asymmetric. The following will focus on the causes of disruption of the basic plexus symmetry of Fig. 3B.

Developmental AD/PD symmetry distortion: the AD axon plane transects the subclavian artery

The subclavian artery (SA) leaves the thorax anterior to the plexus axon plane and then takes a central course in the arm bud. In doing so, the SA runs obliquely through the AD axon plane, thus forming an oblong obstruction in the path of the outgrowing anterior division axon bundles (Fig. 3C). Embryologically, the SA generally develops from the seventh intersegmental artery (i7) [further noted as SA(i7)]. After the division and fusion of the sclerotomes into vertebra precursors, i7 lies lateral to cervical vertebra 7 and, by hypothesis,  retains this position as it initially develops in SA(i7). This positions SA(i7) in the arm bud midway between the outgrowing AD C7 and AD C8T1 axon bundles (noted as C7a and C8T1a) (Padget, 1954). The outgrowing C8T1a axon bundle, called the ‘medial cord’ in the common anatomical terminology, then passes caudally and the C56a and C7a axon bundles cranially to SA(i7) on their converging paths. C7a is thus forced upwards by the arterial body, whereas C56a descends on its converging path, which results in the outgrowing C7a and C56a axon bundles meeting and merging cranially at the arterial surface, producing a merged C567a axon bundle (called the ‘lateral cord’ in the classic nomenclature). The C567a bundle then continues to follow a converging, caudally directed path curving over the arterial body. Eventually, the descending C567a and ascending C8T1a axon bundles meet and merge to form the C5678T1a axon bundle (further noted as C5‐8T1a), i.e. the merger of all AD axon bundles into a single axon bundle, which continues as the median nerve. Thus, the obstruction from the arterial volume causes a highly elongated, stepwise merging of the AD axon bundles.

However, before this final merger, large axon bundles divide caudally and cranially towards the periphery (this could be called the cranial‐caudal brachial division). Caudally, axon bundles divide that run parallel to the humerus: the medial brachial and antebrachial cutaneous nerves as well as the ulnar nerve. These will derive from C8T1(T2)a, unless axon strands of C567a manage to pass over the arterial body to merge with them, which happens frequently with the ulnar nerve axon bundle (Kerr, 1918). Cranially, a dividing axon bundle forms the musculocutaneous nerve, generally dividing from C567a before its merger with C8T1a.

In contrast to the AD, the PD axon plane does not transect the SA. The PD axon bundles can therefore symmetrically converge and merge into a single axon bundle significantly more proximal than the AD. In this way, the asymmetric AD/PD merging pattern of the standard BP model arises (Figs 1 and 3D).

Developmental BP variability associated with vessels

The outgrowth of axons in an environment of variable vascular obstacles generates BP variability at major and minor scales. At the largest scale are BP variations associated with the SA segmental position; at smaller scales, splits of axon bundles around smaller vessels occur.

BP variations associated with the SA segmental level

Embryological considerations

Generally, the SA develops from i7 (Fig. 4A). However, SA may also develop from adjacent intersegmental arteries (Padget, 1954). When developing from i6 instead of i7, SA(i6) will lie at the level of vertebra C6 and will likely be positioned between the outgrowing axon bundles C6 and C7, instead of between C7 and C8 as with SA(i7). Similarly, SA(i8) should lie at vertebra level C8, between sclerotome axon bundles C8 and T1. In principle, SA might also develop from i9 or from i5. However, the likelihood of SA developing from i9 or i5 should be (much) smaller than SA developing from i8 or i6. Moreover, SA developing so eccentric to the bud would likely be associated with further aberrant vascularity.

Figure 4.

BP morphologies associated with subclavian artery (SA) segmental levels, by hypothesis determined by the intersegmental artery from which it develops. Left: embryonic schematics. Right: adult form. (A) Standard BP anatomy – SA developing from intersegmental artery i7 [noted as SA(i7)]. Axon bundles C7a and C8a grow out cranial and caudal to SA(i7), respectively. (B) SA developing from i6. C6a and C7a grow out cranial and caudal to SA(i6), respectively, producing a mirror image of the BP in (A), reflected around the SA. (C–F) BP variations in which no two ‘cords’ in the anterior division exist. (C) SA(i8). C8 and T1 merge proximal to where the anterior division axon plane transects SA(i8). The resulting C8T1a axon bundle grows out cranial to SA(i8), but further distally, the anterior division grows out superficial to SA(i8). (D) SA(i8) as in (C), but distally the anterior division grows out deep to SA(i8), so that the brachial artery becomes positioned superficial to the median nerve. (E) C8 and T1 not merging proximal to SA(i8). C8a and T1a grow out cranial and caudal, respectively, to SA(i8) and merge distal to where the anterior division axon plane transects SA(i8). (F) SA(i9) would lie very caudal to the entire plexus, which might reflect in aberrant arterial arborations. The likelihood of SA developing from i9 should be much smaller than SA developing from i8.

BP associated with SA(i6)

By hypothesis, SA(i6) should lie between the outgrowing C7a and C6a axon bundles, which would pass it caudally and cranially, respectively. A mirror image of the standard BP model then would arise (Fig. 4B). C8T1a and C7a would meet over the caudal arterial surface and merge, and the resulting C78T1a bundle would only merge with C56a distal to the arterial body. The result would be a thin lateral cord and a thick medial cord, with normal brachial nerves arising from them. Possibly, the ulnar nerve might receive more axons from C7a than in the standard BP[SA(i7)] configuration.

BP associated with SA(i8)

Depending on the point of merging of C8 and T1, two variants may be hypothesized.

1. SA(i8) caudal to C8T1a: By hypothesis, generally the C8 and T1 axon bundles would merge proximal to the point where SA(i8) passes the AD axon plane (Fig. 4C). In that case, the merged C8T1a axon bundle would grow out cranial to SA(i8), so that SA(i8) would lay caudal to the entire AD. Two main BP configurations may then be hypothesized. (i.a) If SA(i8) in the proximal bud would run central, the AD would distally grow superficial over it (Fig. 4C). This would reorganize the BP axon bundle mergers in a predictable way. The outgrowing axons of C8T1a would be forced cranially over the artery body towards the C7a axon bundle. These bundles would then merge quite proximally into a single C78T1a axon bundle. The C56a axon bundle, descending on its anatomical normal course, would merge with the C78T1a axon bundle somewhat more distal creating a single broad C5‐8T1a axon bundle. Vessels obstructing axon paths might delay any of these mergers. The resulting broad C5‐8T1a axon bundle is then a homologue of the merging of all PD axon bundles, so that the underlying symmetry of AD and PD becomes evident, as modeled in Fig. 3B. The broad C5‐8T1 axon bundle may further find vessels on its path, around which the axons bundles may split and remerge (Fig. 5A). While such a plexus looks quite different from the standard model, it would contain a normal number of axons, from which normal thoracic and brachial nerves would atypically divide. (i.b) If the caudal SA(i8) would run somewhat anterior in the bud, the AD may grow deep to SA(i8), so that the brachial artery would become positioned superficial to the brachial (median) nerve(s) (Fig. 4D).

2. C8a and T1a merging distal to SA(i8): Only in the rare case, axon bundles C8 and T1 would not merge proximal to the SA(i8) body and would pass SA(i8) cranially and caudally, resp. C8a and T1a would then merge distal to the SA(i8) body, meaning that the ‘inferior trunk’ from the standard BP model would not be formed (Fig. 4E).

Figure 5.

(A) Superimposed on BP variability associated with the SA segmental position are the effects of obstructions from lesser vessels, including splits around vessels and delayed mergers of axon bundles. (B,C) BP variability resulting from the location of the merging of C5 and C6 or C8 and T1 axon bundles relative to the points where the AD/PD division occurs. (B) C56 and C8T1 formation proximal to the AD/PD division point. (C) AD/PD division occurring proximal to C56 and C8T1 formation.

BP with SA(i9)

As stated above, the likelihood of SA(i9) developing would be much smaller than SA(i8) developing. However, the BP associated with SA(i9) would be similar to SA(i8) in Fig. 4C,D, likely with a more aberrant arterial network.

BP variations associated with intersegmental artery remnants

Miller (1939) describes a SA that runs caudal to the BP while producing a large branch that enters the BP between C7a and C8a, as the SA(i7) in the standard BP model would. Miller hypothesized that this variant resulted from an SA(i9) associated with a remnant of i7 (see Discussion). More generally, arterial variants causing BP variability might embryologically result from remnants of intersegmental arteries prevailing in their original anatomical position in the BP.

BP variability caused by axon bundles splitting and remerging around vessels

As the results will show, outgrowing axon bundles may split and remerge around vessels in their path, which may cause additional BP variability superimposed on the variability associated with the SA segmental level. A split in the C567a axon bundle (lateral cord) around a vein is modeled in Fig. 3E. When a vein tangentially crosses the SA cranially at the point where the outgrowing C567a axon bundle (lateral cord) curves cranially over the SA body, the axon bundle finds the vein in its growth path. By hypothesis, a high likelihood then exists that the C567a axon bundle will split to grow bilaterally around the vein. Past the vein, the split axon bundle remerges into a single axon bundle. However, splits in axon bundles around vessels may equally occur elsewhere in the plexus.

PD asymmetries caused by vessels

The PD is not transected by the subclavian artery or veins. Nevertheless, it is transected by smaller arteries and veins, which may branch into small plexuses, which with their connective tissues might also form obstructions to outgrowing axon bundles. By Hypothesis, such arteries, veins or vascular plexuses may locally obstruct the merging of axon bundles so that this merging takes place more distal. As a result, PD axon bundles might not symmetrically merge, as modeled in Fig. 3F.

Trunk asymmetries caused by vessels

By the same principle, vascular obstructions might off‐set the merging of the C5 and C6 or C8 and T1 axon bundles, respectively, more distally than generally happens.

Division proximal to trunk formation

Delayed merging of the sclerotome axon bundles C5–C6 or C8–T1 may cause the brachial division to occur proximal to the merging of C5–C6 or C8–T1 (formation of trunks) (Fig. 5B,C).

Variability in pectoral nerve origins and nerve loops (ansas) associated with BP(SAi) variations

According to Lee (2007), the pectoral axons arise from the anterior spinal nerves C5–T1 as follows. C7 provides the greatest number of axons to the pectoral nerves, followed by C6. C5 only provides axons in 50% of cases, C8 in 97% and T1 in 77%, but the combined C8T1a bundle always provides pectoral axons. The classic BP representation of the pectoral nerves is a medial and a lateral nerve associated with a loop (ansa) bridging the medial and lateral cord (Fig. 1). Recent studies (supported by the present study) significantly modified this view: predominantly three nerve bundles were found – medial, middle and lateral – with multiple branches (Aszmann et al. 2000; Loukas et al. 2006; Lee, 2007; Porzionato et al. 2012; Riggio & Bordoni, 2013; Mian et al. 2014; Tubbs et al. 2016). The model presented here (Fig. 6) focuses on the pectoral nerve variability that arises from BP variants associated with the SA segmental position, as well as from pectoral axon strands (PAS) dividing from the main BP axon bundles more or less proximally. It is assumed that during development the PAS grow out from the AD axon bundles within a thin anatomic layer, within which they branch out. Within this layer, PAS branches likely meet and merge with branches from PAS arising from neighboring AD axon bundles. In this way, a network of pectoral nerve branches arises in which ansas naturally form, with pectoral branches from C7a merging with branches from C56a and from C8T1a. This model predicts the following morphologies. (i) When all PAS arise proximal to the C567a bundle (lateral cord), two loops (ansa’s) are created ‐ a C56a‐C7a axon strand loop spanning the artery and a C7a‐C8T1a loop (Fig. 6A); (ii) when the PAS arise from the merged C567a bundle (lateral cord), the C56a and C7a axon strands may merge into a single nerve branch or retain some individually stranded aspect. However, the only visually prominent ansa remaining will be the C7a‐C8T1a loop that bridges the artery (Fig. 6B); (iii) with SA running caudal to the plexus, the mirror image of Fig. 6B may occur, in which the PAS from C7, C8 and T1 arise more or less individually or merge into a single strand, while forming an ansa bridging the C56a and C78T1a space (Fig. 6C); (iv) when, with a caudal SA, all PAS emerge distal to the merging of all anterior division axon bundles, the axon strands of C5, C6, C7, C8 and T1 may arise more or less individually from the same location and in the extreme case even as a single trunk, from which all pectoral nerves divide (Fig. 6D); (v) in contrast, when SA passes between C8a and T1a (Fig. 4E) and the PAS arise sufficiently proximally, then four individual strands from C56a, C7a, C8a and T1a may form three ansas (Fig. 6E).

Figure 6.

Model of variations and ansas in the pectoral nerve origins. The pectoral nerves consistently get axons from C56a, C7a and C8T1a. These pectoral axon strands (PAS) combine into cranial, middle and caudal pectoral end nerve branches. PAS generally arise distal to the AD/PD division point and proximal to the origin of the musculocutaneous nerve. For simplicity, only configurations with all PAS arising at the same proximal‐distal level are modeled here. Further variability results when PAS arise at different proximal‐distal levels. (A) When the PAS arise proximal to the merger of C56a and C7a, then C7a would generally provide strands combining with the C56a strand (2) and with the C8T1a strand (3), forming two loops bridging axon bundles C56a and C7a, and C7a and C8T1a, respectively. Additional loops may be created, e.g. in the middle branches (Fig. 11A,B). (B) When the PAS divide distal to the formation of C567a, the strands 1, 2 and 3 arise from C567a more or less individually, and only one ansa results, bridging SA (Fig. 11C). This is the configuration in the standard BP model of Fig. 1. (C) With SA(i8) positioned caudal to the plexus, axon bundles C8T1a and C7a may merge proximal to the merging with C56a. If PAS arise from C78T1a proximal to the merger with C56a, a single ansa is formed spanning C78T1a and C56a. This ansa does not span SA (Fig. 9). (D) Same SA as in (C) but with PAS arising distal to the merger of all axon bundles, i.e. from the C5‐C8T1a axon bundle. In this case no ansa will exist and all pectoral nerves may arise from a single axon trunk. (E) SA(i8) positioned between C8a and T1a. If the PAS arise proximal to the merger of any axon bundles, four PAS will be present, of which the branches would likely create three ansas, spanning C56a and C7a, C7a and C8a, and C8a and T1a, respectively.

Model summary

The model states that the BP develops as a neurovascular entity in which vascular variations cause BP variability (Fig. 3H). Removal of the vasculature leaves an axon bundle network of which the variability, by removal of its causes, remains enigmatic (Fig. 3I).

Dissection results

General overview of BP and vascular structures

The passage of the subclavian artery (SA) through the AD

In 48 of the 56 BP (86%), the SA passed through the AD between axon bundles C567a and C8T1a, as in the standard BP model. The axon bundles at the vertebral foramen were cordlike, as were, after merging, the C56 and C8T1 axon bundles. However, over the arterial body the axon bundles, especially C567a, flatten and broaden, and the C567a and C8T1a merger may occur via multiple axon strands over a broad area (Figs 7B,C and 8C). In six of 56 BP (10%) the SA passed caudal (medial) to the AD. One of these cases, modeled in Fig. 4C, is shown in Fig. 9. In this specimen, all AD axon bundles merged into a single bundle as an almost exact homologue of the merging of all PD axon bundles into a single bundle. From the resulting single AD bundle, the normal anterior brachial nerves divided. This case illustrates the underlying symmetry of AD and PD, as modeled in Fig. 3B. In three of the six caudal SA (5% of BP), a large arterial branch from SA passed between C8T1a and C567a, which merged distal to this arterial branch, just as, in the standard model, C8T1a and C567a merge distal to the SA itself. An example is shown in Fig. 10A,B. The fact that a caudal SA with an arterial branch in the same position occurred three times suggests that this variation is not random but results from an underlying embryological stratum. Miller (1939) suggested that such an arterial branch would be a remnant of i7, which would explain its position between C8T1a and C567a (see Discussion). In the specimen in Fig. 10A,B, all PD axon bundles merged normally into a single bundle (C5‐8T1p), but this bundle split again more distally and remerged around the circumflex scapular artery – a variation not found in the literature. In two BP (3.5%), the SA passed between C56a and C7a instead of between C8T1a and C7a. The lateral cord then consisted only of C56a and the medial cord of C78T1a, creating a mirror image of the standard BP (Fig. 4B). In both cases, proximal to the passage of SA through the AD, large arterial and venous branches passed between the C7 and C6 axon bundles superficially to deeply, isolating these axon bundles from each other (Fig. 10C,D). In one case (not shown) a small axon strand passed from C78T1a to C56a proximal to the SA passage, whereas the main merger of C78T1a and C56a occurred distal to the artery. This corresponds to a split C78T1a axon bundle around the SA.

Figure 7.

Dissections illustrating neurovascular BP morphology. Axon bundles are denoted by constituent axon root numbers, from cranial to caudal, and a division letter (‘a’ for AD, ‘p’ for PD). (A) (Specimen S1L) BP with spinal cord, main vessels and upper arm. Scapula and muscles excised, some muscle tissue preserved at pectoral end nerves. The subclavian artery (sa) passes between axon bundles C7a and C8T1a. A vein (v) crosses medially, cranial‐tangential over SA (white arrow) and passes through a split in C567a (yellow arrow). For clarity, the vein (v) was retracted, but in the anatomic position the angle between (v) and C567a is small, which should explain the large length of the split. The pectoral nerves (pn, yellow) derive from C65a, C7a and C8T1a and form two loops bridging C56a and C7a, and C7a and C8T1a, respectively, as modeled in Fig. 6A. sv: subclavian vein. cn, un, mn, mcn, rn, an, sn: cutaneous, ulnar, median, musculocutaneous, radial, axillary and scapular nerves, respectively. C5–T1: brachial nerve roots. (B) (S2L) Specimen with intact cervical spine and upper ribs. The gross neurovascular outline is as in (A). However, the vein (v) leaves the cranial SA surface more distally than in (A) and passes cranially to the C567a axon bundle without passing through it. Cranially at SA, C567a becomes broad and flat (white arrow) and caudally branches into multiple thin axon strands before merging with the C81a. Between C56p and C7p, vein branches (v_b) and an artery (a) pass. These may have obstructed a more proximal merger of C56p and C7p into C567p, causing the PD asymmetric merger shown in Fig. 8C. Artery (a) passes just proximal to the C567p merger point. Similarly, one of the vein branches (v_b) passes just proximal to the C567a merger point, possibly obstructing a more proximal merger C56a and C7a. The pectoral nerves are further detailed in Figs 8C and 11B. (C) S2R, right shoulder of specimen in (B). A vein (v) passes through C567a where the musculocutaneous nerve (mcn) divides. This has been interpreted as mcn shedding a branch to the median nerve (mn). However, if the C567a split were not present, a normal C567a morphology with a normally dividing mcn would be apparent. Similar to (B), C567a and, to a lesser degree, C7a fan out over SA into axon strands. Similar to (A), the pectoral nerves (yellow) arise separately from C56a, C7a and C8T1a and form two loops, while a branch of the pectoral axon strand from C8T1a rejoins the C567a axon bundle. (D) Another example of a vein (v) passing through a large split in C567a, resulting in a proximally split median nerve (mn).

Figure 8.

BP with vascularization removed. (A) Specimen of Fig. 7A, anterior division (AD) with posterior division (PD, green) covered (dark grey area). sa: subclavian artery passage space. v: split in C567a by vein passage. The caudal branch of C567a becomes broad and flat near the merger with C8T1a and sheds an axon strand, part of which continues to the ulnar nerve (white arrow). The pectoral axon strands (PAS) derive from C56a, C7a and C8T1a and combine into four main nerves. The C8T1a PAS arises as proximal as the AD/PD separation point (yellow arrow); some of its axons continue into the cutaneous nerve (cn) (blue arrow). (B) PD of (A), posterior side, photograph flipped left–right for easy comparison with (A) and (C). The AD is covered in dark grey. Axon bundles C56p, C7p and C8T1p merge symmetrically into C5‐8T1p. (C) Specimen of Fig. 7B. Blue: AD; green: PD; yellow: pectoral nerves. In contrast to (B), the PD axon bundles merge asymmetrically. First, C8T1p and C7p merge. Further distally, C78T1p merges with C56p. Hypothetically, the more distal merger of C78T1p and C56p might be caused by vessels (v, a) in Fig. 7B, obstructing a more proximal merger. A thin axon strand (green arrow) splits distal to the main AD/PD division point from C8T1a and joins the radial nerve (rn). However, in the anatomical position, this strand runs closely parallel to C8T1p, so it may have been separated from C81p by vascularization. The AD has the standard anatomical BP morphology. C567a and C81a flatten and broaden at their merger into the median nerve (mn). From C567a, a thin strand passed to the ulnar nerve (un) (blue arrow), as in (A) above, confirmed by further dissection; such strands are well documented (e.g. Kerr, 1918). The PAS arise separately from C56a, C7a and C8T1a. The latter PAS arises quite distally (yellow arrow), whereas in (A) this PAS arises very proximally (A, yellow arrow). These PAS merge and divide to form a network from which five nerve branches arise towards the pectoral region (see Fig. 11B). an: axillary nerve, mcn: musculocutaneous nerve.

Figure 9.

Subclavian artery (sa) passing caudal to BP, modeled in Fig. 4C (Scapula removed for clarity). (A) sa (red) and subclavian vein (sv, purple) pass caudal to the AD to a course between AD (blue) and PD (green) (sa and sv are not in the anatomical position at the vertebrae; further distally, their position is anatomical). (B) Plexus after sa and sv removal. All AD axon bundles merge into a single bundle (5‐81a), homologous to the PD axon bundle merger (5‐81p). This suggests that without SA passing through AD, both AD and PD would merge symmetrically, as modeled in Fig. 3B. The pectoral nerves (yellow) arise from two strands from C56a and C78T1a, respectively. The strands partially merge, forming a single ansa as modeled in Fig. 6C, and divide into all pectoral nerves. cn, un, mn, pn, an, rn, tdn: cutaneous, ulnar, median, pectoral, axillary, radial and thoracodorsal nerves, respectively.

Figure 10.

Effect of the subclavian artery (sa) segmental level on BP morphology. (A,B) SA passing caudal to the AD (blue), dividing in three branches. ba: brachial artery, retracted cranially from a caudal course to show the underlying structures. pa: arteria profunda brachii. ca: arteria circumflexa scapula. ba runs superficial to the brachial nerves, as modeled in Fig. 4D. By hypothesis, SA would derive from i8 and ca would be a remnant of i7 integrated at its normal anatomical position in the arterial network. (A) C567a and C8T1a (blue) merge around the ca‐pa branch to form the median nerve (mn) as in the standard BP model. (B) SA retracted, showing that the PD axon bundles proximally merge, then split around ca, and distal to ca merge again to form the radial nerve, more or less as a homologue of the median nerve formation in the AD. (C,D) SA passing between C6a and C7a, as modeled in Fig. 4B. (C) Major arteries, veins and axon bundles in situ. Blue: AD. Green: PD. Red: arteries. Purple: veins (v). Yellow: ansa pectoralis. An arterial branch and several veins separate C6 and C7, including the posterior division, proximal over a significant distance and, by hypothesis, force a very proximal merger of the C7a and C8T1a axon bundles into a single axon bundle (781a in D). The SA passage between C6 and C7 creates a thin lateral cord (C56a axons only) and a large medial cord (C78T1a axons). D. AD after removal of vasculature, showing the thin lateral cord formed by C56a and the proximally merged C78T1a axon bundle. This axon bundle also has a small split (white double arrow), through which a small artery passed branching from SA.

For completion, it may be mentioned that extensive connective tissues (sheaths) envelop the plexus axon bundles. Especially near the roots, axon bundles that merge tend to be interconnected by such connective tissues well proximal to the actual merger of their axons.

Lateral cord (C567a) splitting around a vein

In 30 of the 56 BP (54%), a vein passed through a split in the C567a axon bundle cranial to the SA. No split in C567a was observed without a vessel passing through. In five cases the split was present bilaterally and in 10 unilaterally; in the remaining cases only one shoulder was available. The split length ranged from a few centimeters, with a small vein passing through, to 10 cm or more, with multiple vein branches passing through a single large split. Examples are shown in Figs 7A,C,D and 11A,C. In four specimens, the split occurred just proximal or at the level of the C5‐8T1a merging distal to the passage of SA and resulted in a split in the median nerve (Fig. 11A). In two cases the median nerve split extended distally up to 10 cm before the median nerve halves merged into a single nerve. As illustrated in the figures, the vein(s) passing through the C567a split seem to be the collector vessel(s) in continuity with the brachial artery companion veins, passing cranially‐tangentially over the SA while receiving venous branches from the proximal anterior shoulder region. The split seems associated with the point where the vein(s) leave the cranial arterial surface towards the subclavian vein. When crossing the SA too distally, the vein(s) may pass cranial to the merging or merged C5‐8T1a bundle and no split occurs (Fig. 7B). In the contralateral BP of this body, a vein did pass through C567a (Fig. 7C). The high prevalence (54%) of veins passing through splits in C567a was not found in the literature and is here presented as possible new data.

Figure 11.

Examples of pectoral nerves, illustrating the models of Fig. 6. (A) Specimen with scapula removed. The pectoral axon strands (PAS) (yellow) derive separately from C56a, C7a and C8T1a and form two ansas, illustrating Fig. 6A. A C7a PAS branch combines with the C56a PAS into one ansa. The other C7a PAS branch combines with the C8T1a PAS into another ansa, around SA (sa, red). The median nerve is proximally split by a large vein plexus (purple), which divides the entire lateral cord, even proximal to the merging of C7a with C56a. mcn, mn, un, cn, pn: musculocutaneous, median, ulnar, cutaneous and pectoral nerves. sv: subclavian vein. (B) Pectoral nerve network of Fig. 8C, in detail. White arrows indicate where the PAS divide from the main axon bundles, which, for clarity, were covered distal to the PAS separation point. The layout is not anatomical; the nerves are spread to show the network they form. Single PAS arise from C56a and C8T1a. From C7a, multiple PAS arise or branch out. Similar to (A), these strands together form two large ansas (a1 and a2) but also several more distal loops from merging sub‐branches. (C) Pectoral neurovascular complex in situ. Two strands (yellow arrows) divide from C567a (blue) and a single strand from C8T1a (green) (black arrow), illustrating Fig. 6B (C567a and C8T1a were partly retracted to show the origin of the C8T1a PAS beneath the SA). The distal C567a PAS splits and forms a loop around an arterial branch (left white arrow). The distal branch of this loop combines with the C8T1a PAS to form an ansa around SA (right white arrow). In total, from these PAS at least five nerve branches emerge to the pectoral region. A vein (v), tangential to the SA, passes through C567a, although through a much shorter split than in (A).

Splits in other axon bundles caused by vessels

Whereas the passage of a vein through C567a occurred in the majority (54%) of plexuses, other splits of axon bundles around vessels were only incidentally found. For example, in Fig. 10A,B the PD split and remerged around a large artery branch. In Fig. 10C,D, where the artery passed between C56a and C78T1a, the C78T1a axon bundle split around a small artery branch.

Asymmetric mergers of axon bundles

By model hypothesis, vessels between axon bundles may cause the merger of these axon bundles to occur more distally. Conversely, large vessels may force axon bundles closer together so that they merge more proximally. The dissection results provided a number of examples of symmetric and asymmetric mergers that might be associated with vasculature. In Fig. 8B, the PD merged symmetrically. In contrast, the merging of PD in Fig. 8C was asymmetric, with C8T1p and C7p merging proximally, and C78T1p and C56p merging more distally. However, two vessels passed between C78T1p and C56p (Fig. 7B). Hypothetically, these vessels may have caused C78T1p and C56p to merge more distally, creating an asymmetric posterior division merger. A large example of vessels separating axon bundles is shown in Fig. 10C,D. Here a large arterial branch and multiple veins separated the plexus, including the posterior division, between C56 and C78T1. Hypothetically, these vessels also forced C7 towards C56 and caused these axon bundles to merge quite proximally into a common C567 axon bundle.

The pectoral nerves

In the six Thiel‐embalmed specimens, all pectoral axon strands arose proximal to the C567a–C8T1a merger into the median nerve, and distal to the mergers of C5–C6 and C8–T1, as well as distal to the AD‐PD separation points. The case modeled in Fig. 6A was present in four of the six Thiel BPs. The PAS arose from C56a and C7a before their merger into C567a (Figs 7A,C and 11A,B). The PAS from C7a split and combined with PAS branches from C56a and C8T1a into two ansas bridging C56a and C7a, and C7a and C8T1a, respectively. Moreover, the loops produced central branches, some strands of which anastomosed into secondary loops (Fig. 11B). The case modeled in Fig. 6B was found in one of the six BP (Fig. 11C). Two separate strands arose from C567a. The proximal strand split into a lateral and a middle branch. The distal strand split into two branches, which formed a loop around the thoracoacromial artery. The distal of these branches also formed a loop (ansa) around SA with PAS branches from C8T1a. In the process, these loops shed central and medial branches. The case modeled in Fig. 6C, was found in one of the six BP (Fig. 9B). The SA ran caudal to the plexus; very proximally, C7a, C8a and T1a merged into C78T1a. Proximal to the merger of C78T1a and C56a, a large PAS arose from C78T1a and a lesser PAS from C56a. These PAS combined to form an ansa while further branching out to produce all pectoral nerves.

Discussion

In the above, a generic model was presented of the general BP morphology and variability. The model considered embryological axon outgrowth through the anterior sclerotomes (Hughes et al. 2009), guided by genetic factors and chemical tracers that regulate the overall axon organization, within an obstacle course formed by the limb bud geometry, the cartilaginous upper arm skeleton, the preexisting vascular system, and the anterior and posterior arm bud muscle precursors. The combination of these factors produces a variable network of axon bundle mergers and divisions that becomes enigmatic when viewed without the vascular and other contexts. The standard BP anatomy emerges from this generic model as a statistically prevailing configuration. Past studies did associate the arterial system with BP variability (Miller, 1939; Aggarwal et al. 2009, 2010, 2012; Yang et al. 2009; Claassen et al. 2016; Tubbs et al. 2016), but to our knowledge no studies so far considered the systematic involvement of the venous system. For instance, Bergman’s Comprehensive Encyclopedia of Human Anatomic Variation makes no mention of the venous system in relation to BP variability (Tubbs et al. 2016).

Miller revisited

Miller (1939) is one of the few authors who systematically investigated the vascular involvement in BP variability in embryological and evolutionary perspective. Although the present results and models generally concur with Miller’s conclusions, there are some points the present study may improve upon.

1. Miller observes ‘many interesting BP variations involving the venous system’ but finds ‘only 4% of the regions studied showed BP anomalies associated with veins’ and chooses not to report these ‘since the course of veins is lacking in definite pattern’. However, in the present results, the venous system was the cause of the most systematic of all variations: a split lateral cord in 54% of investigated BP. Moreover, the morphology can be systematically understood ‐ not by describing individual veins, but by the generic context. Whereas individual veins may be variable, the anatomical planes they drain are constant. The common characteristics of the veins causing split lateral cords are that they are generally in continuity with the companion vein of the brachial artery and that they collect additional branches from (the musculature of) the anterior side of the upper humerus. In doing so, these veins pass the SA cranially‐tangentially from a lateral to a medial course to the subclavian vein. By hypothesis, at that location these veins are in the path of the outgrowing C567a axon bundle (see ‘Embryological considerations’).

2. Miller associated the intersegmental arteries with the SA position within the plexus. A subclavian artery developing from i7 would run between C7a and C8a, SA(i6) between C6a and C7a, SA(i8) between C8a and T1a, and SA(i9) would run caudal to the AD. This interpretation prevailed, so that SA caudal to the AD are still systematically classified as deriving from i9 (e.g. Yang et al. 2009). However, in the present model, additional morphologies are proposed for SA(i8). The C8 and T1 axon bundles would generally merge proximal to where the AD axon bundle plane intersects the SA(i8). Then the AD axon bundles would grow out cranial to SA(i8), with the possible variations proposed in Fig. 4C,D. Since the likelihood of SA developing from i9 should be (much) smaller than the likelihood of SA developing from i8, with the present hypothesis, most SA running caudal to the AD would derive from i8 and not i9. Only in rare cases would C8 and T1 not merge proximal to where the AD axon plane intersects SA(i8). Only then would C8a and T1a likely merge distal to the SA(i8) so that SA(i8) would pass between C8a and T1a (Fig. 4E).

3. In the present study, three SA caudal to the AD were found with an arterial branch running between C7a and C8a (Fig. 10A,B). Around this arterial branch, the AD showed the standard BP pattern of axon bundles C8T1a and C567a merging into the median nerve. Miller reports one similar case, hypothesizing that the arterial branch was a remnant of i7, developing with a caudal SA deriving from i9. However, in view of the possibility that SA(i8) may also run caudal to the plexus (see 2), it is possible – or likely – that the variants found in the present study derived from SA(i8) instead of SA(i9), in combination with a remnant of i7. For both variants, the embryological derivations from the intersegmental arteries are schematized in Fig. 12. These derivations suggest that the likelihood of SA(i8) combining with a remnant of i7 would be much greater than the likelihood of SA(i9) combining with a remnant of i7 with total suppression of i8.

Figure 12.

Hypothesis of development of BP variant in Fig. 10A,B, in which SA runs caudal to the AD, while C56a and C7a merge distal to a large arterial branch to form the median nerve as in the standard plexus. In Fig. 10A,B, this arterial branch gave rise to the a. circumflexa scapulae (CS) and the a. profunda brachii (PB). (A) Present proposed development from the intersegmental artery network. SA develops from i8 and combines with a remnant of i7 to produce the configuration in (C). (B) Development schematic of suppressed branches to produce this variant (the sixth aortic arch develops into the ductus arteriosus and is only postnatally suppressed in the ligamentum arteriosum). (D,E) Development proposed by Miller (1939), in which SA would develop from i9 and combine with a remnant from i7. However, the likelihood of i9 combining with i7, with complete suppression of i8, seems much smaller than of i8 combining with i7 with suppression of i9. 1–6: aortic arches. i6–i9: intersegmental arteries 6–9.

Features that the generic developmental BP model would explain

The presented BP model should provide insight in the factors underlying large and small BP variations.

1. On a large scale, variations in the SA segmental level may fundamentally affect the visual aspect of the plexus. The prevailing position (86% in this study) of the SA is between axon bundles C7a and C8T1a. However, the artery may be positioned between C7a and C56a (two cases, or 3.5%). This mirrors the BP around the artery: the lateral cord will be thin, consisting of C56a only, and the medial cord will be thick, consisting of C78T1a, whereas C7a may fuse with C8T1a very proximally. When the artery runs entirely caudal to the plexus, the C7a and C8T1a axon bundles will likely merge very proximally; more distally, all axon bundles will tend to fuse into a single bundle (Figs 4D,E and 9). In standard terminology, one cord is then considered absent (Aggarwal et al. 2012). However, even with a caudal SA, arterial branches may yet create other morphologies (Fig. 10A,B).

2. The proximal‐distal location of the points of merging of axon bundles can be delayed by vascularization, causing variability (Fig. 3F). An obvious example is the delayed merging of C8T1a and C567a in the standard BP configuration because of the obstruction formed by the SA body (Fig. 3D). However, smaller vascular obstructions may cause asymmetries in the merging of the axon bundles of the posterior division, or of the C5 and C6 or C8 and T1 axon bundles (e.g. Adam et al. 2011).

3. Variability in the merging location of C5 and C6 or C8 and T1 into trunks (by vascular or other causes) may cause variability in the AD/PD division, in that the AD/PD division separation points may be located proximal or distal to the points of trunk mergers (Fig. 5B,C).

4. Additional variability arises from axon bundles splitting around vessels. In more than half (54%) of the 56 investigated BP, a vein passed through a split in the C567a axon bundle. Thus, in the present sample a split C567a axon bundle was the statistically most prevalent morphology, suggesting that it should be included as a clinically relevant feature in standard BP anatomy teaching. The fact that a split C567a axon bundle mostly occurs unilaterally, suggests that such vessel passages are incidental, being the result of small anatomical variations (a vein passing tangentially to the SA more or less distally), which may lead to visually quite different outcomes (a vein passing through a nerve or not). Arterial passages through BP axon bundles are less frequent. Miller reports in 480 bodies arteries passing through median nerves in 15 cases (seven bilaterally), through lateral cords in eight cases (two bilaterally) and four other passages. In the present study, in the six Thiel specimens, two axon bundle splits around arteries were found, shown in Fig. 8 (A–B: artery passing through PD; D: small artery passing through C78T1a).

5. On a smaller scale, the pectoral nerves arise from the anterior division axon bundles as a variable axon strand network, discussed in the paragraph below.

Variability in pectoral nerves and ansas

Although a small sample, the pectoral nerve dissections in the six Thiel‐embalmed BP illustrated three main model variations of the formation of pectoral ansas (Fig. 6A–C). Loukas et al. (2006) report ansas spanning the SA in all their 30 dissected specimens. However, the specimen in Fig. 9, modeled in Fig. 6C, demonstrates that pectoral nerve variations without ansa spanning the SA do exist. Moreover, extrapolating these results, the model predicts that variations without any ansa, consisting of only a single pectoral nerve trunk, may also exist (Fig. 6D). At the other end of the spectrum, theoretically possible but likely rare, are PAS arising separately from C56a, C7a, C8a and T1a, which could form three ansas (Fig. 6E). The model also includes the variants where some PAS arise proximally and others distally; these were not represented in Fig. 6. An example is the PAS from C8T1a, arising very proximally in Fig. 7A, and very distally in Fig. 8C. In conclusion, the models of Fig. 6 may provide a basis for the systematic interpretation of pectoral nerve variability near their origin at the plexus.

Embryological considerations

The systematic occurrence of ansas spanning adjacent axon bundles and additional anastomoses between PAS branches, as shown in Fig. 11B,C, supports the following model assumptions. After dividing from the AD, the PAS axons grow out proximally in a thin anatomical plane and branch into substrands that take diverging courses. In this anatomic plane, substrands diverging from adjacent axon bundles, would meet and merge so that ansas and secondary anatomoses naturally form. To the degree that vasculature passes through the pectoral axon plane, anastomoses may form around vessels, as in Fig. 11C.

Limitations in standard anatomic BP terminology which the present model may improve upon

Essentially, the BP is the local bundling of individual axons. While axon bundles may merge or divide, underlying this is the continuity of the individual axons stretching between the cord and their end‐effectors. The standard anatomical terminology does not reflect this underlying continuity, but divides the BP in discrete and even ill‐defined zones. Where exactly do the trunks commence? Where and why does C7 change from a root to a trunk without any defining landmark? Moreover, the standard nomenclature is only consistent with the main BP variant of Fig. 1 and is ill‐suited to describe BP variability, which basically consists of different axon bundling configurations. For example, if the SA runs caudal to T1, all anterior division axon bundles might merge into a single bundle (Fig. 9). The standard BP terminology provides no proper means to describe such variation, which causes universal confusion, as testified by descriptions such as ‘the anomalous absence of a cord’ (Aggarwal et al. 2012). However, in the present model, this ‘anomaly’ is merely a different but morphologically logical path that axons take within the specific vascular context. As another example, a very distal merging of roots C5 and C6 or C8 and T1 has been interpreted as the ‘absence of a trunk’ (Adam et al. 2011). In fact, in such traditional reporting, variations are seen as anomalous to the degree that they cannot be described by the rigid standard BP terminology. In the present model, variations are merely different axon bundling configurations, logical within the local context and with a likelihood of occurrence depending on the likelihood that such a context would arise. Therefore, in the present paper, the use of standard anatomical terminology was systematically avoided and axon bundles were determined by the spinal nerve numbers of their constituent axons. Although the present paper does not claim that such nomenclature would be fully consistent, it did allow a comprehensive description of the wide range of BP variations presented here.

Variations reported in literature that the model would explain

In addition to the detailed investigations by Kerr (1918), numerous brachial plexus variations have been reported (see Introduction). In these reports, the venous vasculature was typically removed, leaving only the main arteries, if any. The present model demonstrates that without the full vasculature, the clinical and morphological value of reported variations remains incomplete. Extrapolating from the present BP model and the above dissection results, the nature and causes of a number of reported variations are interpreted in Table 1. These variations consist of splits in the lateral cord (567a bundle) which are, according to current dissection evidence, generally caused by a vein passing through this bundle, and the non‐standard merging of sclerotome axon bundles (e.g. ‘absence’ of trunks) as would be caused mainly by a variant course of the subclavian artery and possible proximal branches.

Table 1.

Brachial plexus variations in literature interpreted in the present model.

Variation as reported in the literature	Reference and figure in publication	Interpretation in present model
Fused musculocutaneous and median nerve	Fig. 2 in Chaudhary et al. (2012)	According to the present hypothesis, all these variations seem to describe a split of the C567a axon bundle (lateral cord) around a vessel, probably a vein, as occurred in 54% of BP in this study. Removing the vein leaves an oblong split, illustrated in Fig. 7A,C,D. In all papers cited here, the veins were removed, so that dissection results only showed the split C567a axon bundle. The variability between these cases would consist of the relative number of axons, i.e. the relative cross‐section of the branches of C567a of the split, the length of the split and the location of the axon bundle splitting and merging points delimiting the split
Middle trunk of anterior division dividing in two branches	Fig. 2 in Chaware et al. (2012)
Anomalous branching pattern of lateral cord	Fig. 1 in Das & Paul (2005)
Unusual connection between musculocutaneous and median nerve	Fig. 2 in Emamhadi et al. (2016)
Variant branches of lateral cord	Fig. 1 in Padur et al. (2016)
Extra branch between median and musculocutaneous nerve	Bergman’s Comprehensive Encyclopedia of Human Anatomic Variations (Tubbs et al. 2016), p. 595–596, Figs 53.6, C and 53.7, C	Artery passing through split in C567a (lateral cord)
Proximal merging of C7, C8 and T1 ‐ only two trunks	Aggarwal et al. (2010)	According to the present hypothesis, these variations are associated with a caudal course of the subclavian artery, as modeled in Fig. 4C,D. The arterial body volume would force sclerotome bundles T1 and C8 towards C7, causing very proximal fusion of these bundles
Uncommon variation in the trunks	Govindarajan et al. (2014)
Single cord anomaly (Proximal merging of C5a, C6a, C7a, C8a, T1a)	Aggarwal et al. (2012)	The single cords here reported seem to be an extreme variant of sclerotome axon bundles fusing very proximally in association with a caudal running subclavian artery, as modeled in Fig. 4C,D
Artery entering plexus between C6 and C7 instead of between C7 and C8	Fig. 5 in Aggarwal et al. (2009)	Artery causing a proximal merger of C8T1a and C7a, as in Fig. 10C,D
Bilateral absence of lower trunk	Fig. 1 in Aragao et al. (2014)	This concerns very distal mergers of the C8a and T1a sclerotome axon bundles, which run independently and only merge at the point of formation of the median nerve. The vascular system was removed in the presented data, but one could speculate that (a) large vessel(s) ran between C8a and T1a, preventing a more proximal merger of these bundles

Embryological considerations

The frequent venous passages through the C567a axon bundle (lateral cord)

The development of the arm bud vascularization progresses from proximal to distal. From the originally undifferentiated capillaries, an arterial and venous collector tree develop. The axonal outgrowth into the bud starts at stage 15 (35–38 days of development) and by stage 16 (37–42 days of development) the axon growth front has already reached the point where the ulnar, median and musculocutaneous axon bundles divide (Rodriguez‐Niedenfuhr et al. 2001). Therefore, the frequent splitting of the C567a axon bundle around a vein should occur at about the end of stage 15. At that time a definite axillary venous layout may not yet have developed. Therefore, one could speculate that at this embryological stage of axon outgrowth, multiple splits of axon bundles around small vessels might occur, but that only the splits around the vessels (veins) that are retained in the further development are preserved in the adult form.

Why no anastomoses seem to exist between AD and PD – a hypothesis

In the adult, the separation point of the anterior‐posterior divisions is located quite proximally near the vertebrae, and the anterior and posterior division axon bundles run closely adjacent in the subclavian passage before separating to pass anterior and posterior to the humerus. The question then is why axon strands passing between the anterior and posterior division in the subclavian region do not occur or are rare (only one small strand was found, arising close to and distal to the main division point, see Fig. 8C). The standard anatomical notion is that the divisions result from neuron regions in the cord that project their axons onto the anterior (flexor) or posterior (extensor) brachial muscle layers (Landmesser, 1978). However, this does not explain why the divisions should run so cleanly juxtaposed in the subclavian passage. Clearly, axons could intermingle in that region and still project onto their anterior or posterior muscle layers.

The following hypothesis is proposed. In the embryo, the precursor shoulder/arm cartilaginous skeleton is proximally located at the bud near the spine. Therefore, the AD/PD separation, forced by the central skeletal cartilage, occurs close to the spine (Figs 2 and 13A,B). However, with the outgrowth of the shoulder in the fetus and postnatally, the humerus translates laterally. In the process, the brachial nerves elongate, but the point where the AD/PD actually divide, does not move laterally to the same degree as the humerus (Fig. 13C). As a result, between the AD/PD separation point and the laterally translated humerus, the AD/PD axon bundles become closely juxtaposed in the subclavian BP passage. Although the SA generally runs between them, they may be in contact at their edges. Moreover, when the course of the SA is caudal to the plexus, the AD and PD may have no real physical separation in the subclavian passage at all. The absence of AD/PD anastomoses in the subclavian passage would be explained by the fact that the axon network layout would already be grossly finalized prior to the lateral shoulder growth. Axons growing from the cord into the bud at later stages would follow the existing bundles, without crossing over between them, even if these bundles come in contact upon later lateral shoulder growth. This being said, a small axon strand dividing from the radial nerve proximal to its passage posterior to the humerus and joining the ulnar nerve at the distal third of the upper arm was found in one of the Thiel specimens.

Figure 13.

Model of the position of the anterior and posterior (A/P) brachial division separation point and A/P course from embryo to adult. H: humerus. SC: spinal cord. (A,B) In the embryo, the cartilaginous humerus (H) forces the A/P division separation point proximal at the arm bud. (A) Schematic. (B) Embryo stage 16 (37–42 days of development) (specimen 6517, 3D Atlas of Human Embryology, AMC, the Netherlands, http://www.3datlasofhumanembryology.com ). Neural tube in green, skeleton in off‐white, nerves in yellow. (C–E) With lateral shoulder outgrowth, the humerus, as a division‐separating obstacle, translates laterally. The nerves lengthen and in the narrow subclavian passage the A/P divisions, previously separated, proximally come in close contact. Blue arrows: A/P separation points. Green arrows: points where the A/P bundles diverge from a parallel course in the thoracic outlet to an anterior and posterior course, respectively, around the humerus. (D) Stage 23 embryo (56–60 days of development), specimen 9226, right‐hand side BP. (E) Same specimen, different viewing angle, left and right BPs.

Evolutionary considerations and variability – when is a plexus anomalous?

The exit of axons through intervertebral foramen is shared by all vertebrates and upper and lower limb plexuses exist in all tetrapods. Therefore, the outgrowth of spinal axon bundles into nerve complexes was evolutionarily resolved well before the mammalian genesis. The many BP variations reported from dissections of persons who during their lives apparently had normal function, demonstrate that the formation of the brachial plexus is a robust process. The present model suggests that the likelihood of interface conflicts – such as compressions – of nerves with the arterial (and muscle/tendon/ligament) system is embryologically minimized by development chronology. The plexus axon bundles grow out in a vascular environment, through which they navigate in conformity with local circumstances. Thus the layout of the peripheral nervous system flexibly and robustly adapts to the environmental variability. A different path of the SA may cause a visually totally different plexus layout, which nevertheless is equally functional and generates the same number of axons and peripheral nerves. According to this model, the standard anatomical BP outline of Fig. 1 is not in the first instance due to the evolutionary optimization of the axon outgrowth. Indeed, the axonal outgrowth itself, as shown in Fig. 2, seems to have little room for fundamental variability. This is testified by the fact that even in hypovascularized buds, normal axon outgrowth initially occurs (Bates et al. 2003). Instead, the real determinant of the standard anatomical BP morphology seems to be the anatomically fairly constant course of the subclavian artery, such that C7a generally passes it cranially and C8T1a caudally. In other species, the general BP layout may be different due to a different subclavian arterial course (Miller, 1939).

In conclusion, in the present model the distinction between ‘normal’ and ‘anomalous’ BP morphologies is replaced by a likelihood that a certain context arises that will produce a certain BP morphology. Hereby the formation of the BP appears to be a robust process that even with an aberrant vascular outlay will likely generate a functional BP morphology, however visually different it may be from the anatomically prevailing morphology.

Teaching the generic developmental BP model in the classroom

For medical education, the generic model offers the advantage that the actually occurring plexus variability is incorporated in the model. For teaching the model, Figs 3 and 4 would suffice, with a few chosen slides of axon bundle splits around vessels. The standard model can still be discussed as the most prevailing variant, but students need not learn it by rote, as Figs 3 and 4 provide a narrative that constructs it. Fig. 6 could also be discussed to exemplify how anatomic variability arises. After teaching the generic BP model, in gross anatomy dissection classes students will not be puzzled by BP variations, but will expect to find them and be challenged to understand them in contextual terms. Students will also be more mindful of the vascular context. For these reasons, the generic developmental model should prepare students better for, e.g. a surgical career, than would the teaching of the standard BP model.

Conflict of interest

There is no conflict of interest with this study.

Data availability statement

Data available on request from the authors.