Zehr [40] stage 24–25.

The extent of development of the heart and lungs varied considerably among our species sample at this stage. In Pantherophis guttata guttata (Fig. 13A), the tracheoesophageal septum had not formed, so that esophagus and trachea were still in continuity; this was not the case in the other species. In P. guttata guttata the right lung was longer than the left, but both had a distinct lumen distally (Fig. 10A). In Python curtus breitensteini (Fig. 13A) the right lung was longer than the left, but both featured a distinct lumen (Fig. 11A). In Causus rhombeatus (Fig. 12A; Fig. 13C) the left lung was absent.

P. guttata guttata (Fig. 13A) still had a common outflow tract. In P. curtus breitensteini (Fig. 13B) however, the outflow tract was divided into three channels: the left aorta, right aorta and the pulmonary channel. Finally, in Causus rhombeatus (Fig. 13C) only the aorticopulmonary septum had formed, dividing the outflow tract into a common aortic and a pulmonary channel. Regardless of developmental stage, the PAAs were in continuity with the two dorsal aortae, which originate from the ventricle. P. curtus breitensteini (Fig. 13B) was the only species to show two pulmonary arteries at this stage. Each artery extended to its respective lung from the sixth aortic arch; both ductus arteriosi were therefore distinct.

Zehr [40] stage 25–28.

The esophagus and trachea in P. guttata guttata (Fig. 14A) were now separated by the trachea-esophageal septum. The right lung had elongated and acquired a distinct lumen in all species (Fig. 10B, Fig. 11B, Fig. 12B). The left lung in P. curtus breitensteini (Fig. 14B) was at a comparable stage of development to the right lung. By contrast, the left lung of P. guttata guttata (Fig. 14A) was smaller than the right and had only a small lumen distally, while the right also had distinct lumina at more anterior locations (Fig. 10B). In Causus rhombeatus (Fig. 14C) the tracheal lung could be identified for the first time as a thickening of the trachea at the level of the widening of the tracheal lumen. Posteriorly, the lumen widened more, denoting the start of the right lung (Fig. 12B; Fig. 14C).

With the establishment of the definitive left aorta, right aorta and pulmonary trunk, the vascular anatomy was now similar in all species. Except for P. curtus breitensteini (Fig. 14B), no pulmonary arteries had formed, although a layer of mesenchymal tissue (Fig. 10B) around the lungs in both P. guttata guttata and Causus rhombeatus might be interpreted as the splanchnic plexus from which the distal parts of the pulmonary arteries will develop.

Zehr [40] stage 26–30.

The right lung had reached a similar stage of development in all species (Fig. 15A, B, C). In P. guttata guttata (Fig. 10C, Fig. 15A), no significant changes were observed. The left and right lungs in P. curtus breitensteini were equally developed (Fig. 15B). Causus rhombeatus and Calloselasma rhodostoma both had a tracheal lung (positioned anteriorly to the heart) and a right lung (Fig. 15C, D), but the right lung of Calloselasma rhodostoma was more obvious due to a constriction at cardiac level and luminal widening distally from this constriction (Fig. 12D); Causus rhombeatus lacked such a constriction and widening, a characteristic of a thoraconuchal lung (Fig. 12C).

At this stage, all species had one or two pulmonary arteries, arising posterior to the heart and connecting to their respective 6^th PAA, although their point of origin could not be found. So, the pulmonary artery originating at the right lung connects to right 6^th PAA. A left and right ductus arteriosus can be defined. At this developmental stage, the tracheal lung, where present, was not supplied by an obvious pulmonary artery; thus in both Causus rhombeatus (Fig. 15C) and Calloselasma rhodostoma (Fig. 15D) the pulmonary artery did not extend anterior to the heart. Possibly the tracheal lung was supplied with blood by a different vessel, or the anterior pulmonary artery was too small in diameter to be noticed or visualized reliably. If PFA did not bind with these thin vessels, it would not have appeared on the scans. Finally, Causus rhombeatus (Fig. 15C) had two pulmonary arteries that both lead to the single, right lung. Finally, Calloselasma rhodostoma (Fig. 15D) featured two left 6^th PAAs.

Zehr [40] stage 32–35.

Specimens of Calloselasma rhodostoma or Causus rhombeatus were not available. Lung development in both P. guttata guttata (Fig. 16A) and P. curtus breitensteini (Fig. 16B) had reached a stage, at which separate faveoli became distinguishable. In P. guttata guttata (Fig. 16A) the left lung now featured a distinct lumen (Fig. 10D). Furthermore, the left lung was now in a more ventral location with respect to the right lung, compared with previous stages of this species examined.

Both P. guttata guttata (Fig. 16A) and P. curtus breitensteini (Fig. 16B) show changed spatial relations of the heart with respect to the PAAs: these are now located directly anterior to the heart. Furthermore, in P. guttata guttata (Fig. 16A) the right ductus arteriosus was becoming obliterated; in P. curtus breitensteini, the ductus was still apparent.

Zehr [40] stage 34–36.

No specimen of P. curtus breitensteini was available. The lungs were in an advanced stage of development, with distinct faveoli and tracheal rings. In Elaphe obsoleta spiloides the left lung (Fig. 17A) featured distinct faveoli and a wide lumen (Fig. 10E). In Calloselasma rhodostoma (Fig. 17B) the faveoli of the tracheal lung had reached a state of maturity apparently similar to that of the right lung. It extended until approximately halfway along the ventricle, where a gradual widening of its lumen indicated the origin of the right lung (Fig. 12E), although there was no sharp boundary between the two.

The vascular anatomy of E. obsoleta spiloides (Fig. 17A) and Calloselasma rhodostoma (Fig. 17B) were highly similar. The right ductus arteriosus was completely obliterated, although in E. obsoleta spiloides remnants of the right 4^th and 6^th PAA could still be distinguished. On the left, the ductus arteriosus persisted. Differences were found in the pulmonary artery: in Calloselasma rhodostoma it curved anteriorly, not posteriorly as it does in E. obsoleta spiloides. Additionally, at the curvature where the single pulmonary artery curves anteriorly, a second pulmonary artery branched off in a posterior direction in Calloselasma rhodostoma.

Development of the right lung.

The right bronchial bud is established early in development and develops into a mature right lung in all species. Causus rhombeatus deviates from the common plan in that it has a thoraconuchal lung in which the right lung and the tracheal lung are continuous. Already at early stages of tracheal lung development it was hard to distinguish it from the right lung. Thus, the question arises to what extent the development of right and tracheal lung are integrated. Further studies into the development of the thoraconuchal lung would therefore be of interest.

In Calloselasma rhodostoma the transition point from tracheal to right lung is demarcated by a constriction in the tracheal lumen at the axial level of the heart. However, this constriction disappeared at later stages of development. Regardless of the mode of development of the right lung, the right pulmonary artery always developed and became connected with the right branch of the 6^th PAA. As a result, this branch always persisted to some extent, as in Calloselasma rhodostoma, E. obsolete spiloides and P. guttata guttata the right ductus arteriosus regressed. In P. curtus breitensteini we lacked developmental stages sufficiently advanced to provide more definitive conclusions.

Development of the left lung.

Our data show that four developmental patterns of the left lung can be distinguished. The first (Table 5, Type 1) is characterized by absence of a left bronchial bud in both Calloselasma rhodostoma and Causus rhombeatus. However, we lacked the youngest of embryonic stages, so that we could not determine whether a left bronchial bud does not develop due to absence of tracheal bifurcation, or whether it regresses shortly after its formation. Both developmental patterns result in type 1 lung anatomy (Table 5).

The second and third developmental patterns (Table 5, Type 2a and 2b) were observed in snakes with a vestigial left lung. This condition implies inhibition of lung elongation, but not necessarily of lung maturation: in the majority of species the vestigial left lung is faveolated and vascularized [12], which are typical anatomical features of a mature lung. Moreover, a vascularized vestigial left lung may be served by a minute pulmonary artery [12], which indicates a contribution to respiration. We did not observe any such minute pulmonary artery in our 22 dao E. obsolete spiloides specimen, but it is possible that the resolution of the MicroCT scans was either insufficient to detect this artery, or its presence was obscured by nearby structures. In the snakes that possess a vestigial left lung, it is usually both faveolated and vascularised [12].

Based on the observation that the vestigial left lung may or may not be faveolated and vascularized, the vestigial nature of the left lung may result from two processes: inhibition of early elongation of the left bronchial bud, and inhibition of later faveolization and vascularization. Thus we are able to identify two developmental patterns. In both cases the left bronchial bud undergoes significant slowing shortly after appearing, followed by arrest of elongation, and then either (1) fails to become vascularized (Table 5, Type 2a), (2) or undergoes subsequent vascularization (Table 5, Type 2b). Both development patterns result in type 2 lung anatomy (Table 5).

The fourth developmental pattern (Table 5, Type 3) leads to a fully developed, albeit relatively short left lung. The question arises whether the left lung is shorter than the right lung because it develops slower, as suggested previously [15], [63], or because its growth is arrested early (truncation). The earliest stage of Python curtus breitensteini in our sample (Fig. 13B) shows that the left lung is shorter than the right lung. Since already at this early stage the two lungs in this species are subequal, our data are most easily explained, in view of previous studies in Thamnophis radix and Natrix natrix [15], [24], by assuming that the left lung has a lower growth rate than the right lung, as opposed to truncation at a later stage of development. It is important to note that availability of this rare embryonic material is limited, and so it was not possible to make any quantitative determination of growth rates. It is also possible that the left lung bud initiated its development at a later stage (also a kind of heterochrony) or was smaller from the beginning (perhaps due to differences in cell allocation between the left and right buds). Further studies on a larger series of python embryos is needed to examine these issues.

To summarize, we propose that a stepwise change in developmental mechanisms resulted in the obliteration of the left lung. First, in basal macrostomata (represented by Python curtus breitensteini in this study), slowing of left lung elongation leads to a left lung that is shorter than the right lung. Secondly, in more advanced snakes (represented by Pantherophis guttata guttata and Elaphe obsolete spiloides), the arrest of elongation (truncation) later in left lung bud development results in a vestigial, faveolated left lung. Finally, failure to faveolate results in a vestigial left lung lacking faveoli. See Fig. 2 for a mapping of these events onto our phylogeny.

Development of the tracheal lung.

Our results suggest that the tracheal lung arises as an outgrowth of respiratory tissue from the trachea. In the adult, the tracheal rings are open on the dorsal side so that the lumen of the tracheal lung and trachea communicate. Since tracheal rings are typically C-shaped [12], it is possible that respiratory tissue evaginates from the dorsal cleft in the trachea. A follow-up study investigating the specifics of tracheal lung development would be of great value. Tbx5 gene expression studies could be of special interest, given its role in formation of the tracheal cartilage in mice [5].

We show that no distinct pulmonary artery develops in relation to the tracheal lung; instead it might be supplied by the tracheal artery, which in humans branches of the inferior thyroid artery [64]. In snakes, however, it usually branches of the carotid artery [65]–[70], which does not correlate with the anterior pulmonary artery described here and in the literature (S1 Table). We, therefore, suggest that the anterior pulmonary artery is either an anteriorly relocated posterior pulmonary artery (as we propose for Calloselasma rhodostoma), or a branch of a posterior pulmonary artery (as we propose for Hydrophis elegans).

The tracheal lung and associated pulmonary artery were previously proposed to be an anteriorly relocated left lung and pulmonary artery, respectively [61]. This was disproved [71] when the presence of a ligamentum arteriosum between pulmonary trunk and left aorta was highlighted in snakes lacking a left pulmonary artery, which agrees with our results; the anterior pulmonary artery can therefore not be analogous with the left pulmonary artery.

We here propose a new hypothesis: descent of the heart, cephalic folding and elongation of the neck during development cause a change in pulmonary artery location with respect to the heart. While the lungs remain in a relatively stable location, the heart will pass the location where the pulmonary arteries attach to the lung. An embryonic posterior pulmonary artery will thus become relocated to an anterior position. Our observation in 20 dao Calloselasma rhodostoma is thus easily explained. In this species, the pulmonary artery curved posteriorly in early stages, but anteriorly in late stages. Additionally, the posterior branch was but a branch off the anteriorly curving pulmonary artery. This resembles our dissected Trimeresurus heart, and the same applies to Causus rhombeatus, whose adults are reported to exhibit two anterior pulmonary arteries [54]: it is a flipped version of the embryonic pulmonary artery anatomy found in this study.

Some snakes however, with have respiratory tissue along the entire longitudinal axis of the body [14], will have respiratory tissue posterior to the heart even after the descent of the heart, such as Hydrophis (Fig. 8). In this species the primary pulmonary artery in this species is likely not the anterior one but the posterior one. Further research into the development of such secondary pulmonary arteries is warranted.