Identification of the building blocks of ventricular septation in monitor lizards (Varanidae)

The evolution of land-living vertebrates was facilitated by the ability to breathe air with lungs (Perry and Sander, 2004). A dedicated vascular bed, the pulmonary circulation, enables the perfusion of the lungs by blood. The oxygenated blood of this circulation returns to the left side of the heart (Burggren and Johansen, 1986). However, the oxygen-rich blood from the lungs may mix with oxygen-poor blood returning from the systemic circulation if septal structures are absent in the ventricle of the heart. In mammals and in archosaurs (crocodylians and birds), the ventricle is divided into the left and right ventricle by a full ventricular septum (Poelmann et al., 2014). Reptiles and reptile-like vertebrates were ancestors of mammals and archosaurs (Laurin and Reisz, 1995). Non-crocodylian reptiles have intermediate degrees of ventricular septation (Jensen et al., 2014).

The single cardiac ventricle of extant non-crocodylian reptiles is partially divided by three septa named the ‘vertical septum’, the ‘muscular ridge’ and the ‘bulbuslamelle’. The vertical septum is a prominent sheet of trabecular myocardium positioned beneath the atrioventricular valve (Fig. 1). It separates the oxygen-rich blood from oxygen-poor blood coming from the right atrium, its function resembling that of the ventricular septum of mammals and archosaurs (Jensen et al., 2014). The muscular ridge is the most prominent septal structure, but is not involved in the separation of the inflowing blood from the atria because it is positioned to the right of the atrioventricular canal (Poelmann et al., 2014, 2017) (Fig. 1), in contrast to the ventricular septum of mammals and archosaurs, which is positioned immediately below the atrioventricular canal (Cook et al., 2017; Greil, 1903; Webb et al., 1971). During ejection, the muscular ridge, however, forms an important boundary between the pulmonary compartment, the cavum pulmonale, and the cavum venosum to separate blood flow into the aortae and the pulmonary artery, respectively (Fig. 1). The third septum, the ‘bulbuslamelle’, is positioned opposite to the muscular ridge and presses against the free edge of the muscular ridge during contraction. It has been shown experimentally that this ‘coming-together’ of the two septa limits mixing of oxygen-rich blood and oxygen-poor blood during cardiac contraction (White, 1959). The full ventricular septum of mammals, birds and crocodylians has a pronounced left-right gradient of expression of the transcription factor Tbx5 (Jensen et al. 2018; Koshiba-Takeuchi et al., 2009; Poelmann et al., 2014). Analyses of Tbx5 gradients suggested that the vertical septum of a turtle (Trachemys), but not the vertical septum of a lizard (Anolis), is homologous to the full ventricular septum (Koshiba-Takeuchi et al., 2009). Studies on other species of turtles and squamate reptiles, however, show that both the muscular ridge and the vertical septum display a gradient of Tbx5 expression (Poelmann et al., 2014). These seemingly conflicting results in non-crocodylian reptiles could be resolved when the expression of Tbx5 would identify more than one septal structure in animals without a full ventricular septum.

Because the cardiac anatomy of the last common ancestor of amniotes is likely to remain unknown, we can only provide insight into the evolution of the ventricular septum through studies of the cardiac anatomy of extant reptiles. Among these, the monitor lizards (Varanidae) have ventricular septa that are more developed than in any other lizard (Brücke, 1852; Jensen, 2019; Webb et al., 1971). Further, monitors are exceptional among lizards because their cardiac ventricle is functionally divided, such that the left side of the ventricle (cavum arteriosum and venosum) generates high systemic blood pressures and the cavum pulmonale supplies the lungs at much lower pressures (Burggren and Johansen, 1982; White, 1968). Consequently, shunting of blood streams in the monitor ventricle is almost independent of atrial filling pressures and the (arterial) afterloads, whereas in species without a functionally divided ventricle, atrial filling pressures and afterloads directly affect the shunting of blood streams in the ventricle (Hicks, 1998; Joyce et al., 2016). There are, nonetheless, minor degrees of so-called ‘washout shunting’ in the monitors (and other non-crocodylian reptiles) because the cavum venosum is a conduit for oxygen-poor blood in diastole and a conduit for oxygen-rich blood in systole (Heisler and Glass, 1985; Heisler et al., 1983; Ishimatsu et al., 1988). The mammal-like blood pressures and low levels of shunting in monitors support their metabolic rates, which are higher than in other reptiles (Thompson and Withers, 1997). The greater state of development of the septa could associate with more pronounced patterns of gene expression, in which case the monitor ventricle may be a very suitable model in which to test the evolutionary origin of the full ventricular septum.

To investigate whether the ventricle of monitors is divided by structures that are homologous to the full ventricular septum of mammals and archosaurs, we studied heart development and function in monitors (Varanus acanthurus, V. exanthematicus, V. indicus and V. salvator) and compared them with phylogenetically distant lizards with typical lizard hearts, the brown anole (Norops sagrei) and the leopard gecko (Eublepharis macularius). To assess homology between amniote cardiac structures, we studied the expression of anole orthologues of genes associated with the formation of the ventricular septum of mammals and birds, specifically Irx1, Irx2, Tbx3, Tbx5 and Myh6 (Aanhaanen et al., 2010; Boukens and Christoffels, 2012; Bruneau et al., 1999; Christoffels et al., 2000b; de Groot et al., 1987; Hoogaars et al., 2004; Moorman et al., 1998; Yamada et al., 2000). We then related the degree of septation to electrical function. In mammals, the activating current spreads from the atrioventricular node through the atrioventricular bundle that resides in the crest of the ventricular septum (Durrer et al., 1970; van Rijen et al., 2001). This mode of activation reveals itself on the epicardial surface as an early activation near the ventricular apex (Chuck et al., 2004; Reckova et al., 2003; Rentschler et al., 2001; Sankova et al., 2012). Electrical activation is much the same in crocodylians, which have a full ventricular septum, whereas activation of the undivided ventricle in lizards and snakes proceeds in a primitive pattern from base to apex (Christian and Grigg, 1999; Gregorovicova et al., 2018; Jensen et al., 2018, 2012). We hypothesize that the advanced state of ventricular septation in monitor lizards could impact on electrical propagation.

We show in monitors that the ventricular septa express orthologues of genes associated with the formation of the ventricular septum of mammals and birds and that the pattern of electrical activation is specialized compared with other squamate reptiles. The ventricle of monitors appears to be an extreme variation of the common design of the non-crocodylian reptilian ventricle.

The monitor ventricle consists mostly of trabeculated myocardium, as is the case in other squamates (Fig. 1). In the adult monitor ventricle (V. exanthematicus and V. salvator), the trabeculations are organized into a thick-walled and left-sided systemic ‘left ventricle’ and a thin-walled and right-sided pulmonary ‘right ventricle’, or cavum pulmonale (Fig. 2A,A′, Movie 1). The ‘left ventricle’ corresponds to the so-called cavum arteriosum and cavum venosum (Fig. 1, CA and CV, respectively). In the transverse plane, the ‘left ventricle’ is circular and the ‘right ventricle’ forms a crescent at its right side (Fig. 2B,B′, Movie 2). The two ventricles are separated by the muscular ridge and the bulbuslamelle (Fig. 2B). In the apical half of the ventricle, the two septa are fused and appear as one septum. We noticed highly echogenic lamellae in the inner-most part of the cavum venosum. These lamellae correspond to the so-called bulboauricular lamellae of Greil (Greil, 1903; Jensen et al., 2013). In a large heart of Varanus salvator, these lamellae were thick and constituted the inner-most part of the muscular ridge and bulbuslamelle (Fig. 2C). They extended from the atrioventricular canal to the point where the muscular ridge and bulbuslamelle abut during contraction.

In diastole, the atrioventricular valve achieved direct contact with a prominent vertical septum (Fig. 2D, Movie 3). The vertical septum was positioned entirely within the cavity of the left ventricle (the cavum arteriosum and cavum venosum combined). Both atrial blood flows were therefore received in the left ventricle. As atrial blood enters the ventricle, the space between the muscular ridge and the bulbuslamelle widens, allowing the right atrial blood to reach the cavum pulmonale (the right ventricle). In systole, this space closes again as the muscular ridge and the bulbuslamelle abut. Then, the ‘right ventricle’ ejects into the pulmonary trunk and the ‘left ventricle’ ejects into the aorta (Fig. 2E, Movie 3). Septal structures of the monitor ventricle show functional and anatomical resemblance to the ventricular structures of four-chambered hearts of mammals and archosaurs. Next, we wanted to know whether the septal structures of monitors develop in a similar fashion to the four-chambered heart.

We examined the development of the morphology of the monitor heart in a series of sections of mangrove monitor hearts stained for a marker for cardiac muscle (cardiac troponin I). In the earliest specimen, from 10 days post-oviposition (dpo), the heart was similar to that of other reptiles and to chicken around 4 days of development; that is, myocardium was detected in the sinus venosus, sinuatrial junction, atria, atrioventricular canal, ventricle, and outflow tract (Fig. S1A,B). The ventricle was proportionally small, whereas the atrioventricular canal and myocardial outflow tract, the conus arteriosus, were proportionally large, indicating that the heart was immature (Figs S1,S2). In the atria, formation of trabeculae was just beginning and the atrial septum was a low myocardial ridge with a mesenchymal cap in the atrial roof (Fig. S1C). A non-muscular dorsal mesenchymal protrusion was present in the dorsal wall of the atrium. The atrioventricular canal was circular in cross-section and its interior was occupied by the large dorsal and ventral cushions (Fig. S1D). It was positioned to the immediate left of the body midline as defined by the position of the notochord, foregut and trachea (Fig. S1A). Trabeculation of the ventricle was prominent in the outer curvature and absent in the inner curvature (Fig. S1A). Distally, the trabeculated muscle was sponge-like, whereas proximally it was organized into about 12 sheets of trabecular muscle that were oriented approximately dorso-ventrally (Fig. S1A). The outflow tract was proportionally long, it was myocardial to the vicinity of the pericardial reflection, and its interior was dominated by mesenchyme but the lumen was not divided (Fig. S1A,B).

In the next stage investigated, 24 dpo (Fig. 3A), all three sinus horns were distinct vessels and their myocardium extended to the pericardial reflection (Fig. S1E,F). The atrial septum was developed to such an extent that the ostium primum was almost closed and secondary foramina had formed (Fig. S1F,G). Myocardium was detected within the dorsal mesenchymal protrusion (Fig. S1E). The atrioventricular canal remained positioned to the left of the body midline, but its cushions had begun to merge. The formation of the muscular ridge had begun near the base of the outflow tract, but the outer curvature of the ventricle comprised trabecular myocardium only (Fig. 3A, Fig. S1F). A shallow sulcus was found on the epicardial surface in proximity of the early muscular ridge (Fig. 3A, Fig. S1E-G). The outflow tract was myocardial only in the proximal half (Fig. S1G). Its lumen was divided by mesenchyme into three channels, the future pulmonary artery, the left aorta and the right aorta. At this stage, the proportions of the principal compartments of the heart had become similar to those of the formed heart. Nevertheless, the size of the ventricle still had to increase at the expense of the myocardial outflow, by regression of the myocardial outflow tract (Fig. S2). Furthermore, ventricular septation was not obviously different from that of other lizards.

By 32 dpo, the ostium primum was closed and specialized features had started to become apparent. By 32 dpo, the muscular ridge and the bulbuslamelle were characterized by a core of compact myocardium (Fig. 3B). Towards the apex, the compact myocardium dissolved into trabeculae (Fig. 3C). The trabecular sheets beneath the atrioventricular cushions were of approximately equal height. There was no obvious vertical septum at this stage. The bulboauricular lamellae were becoming prominent (Fig. 3B). In the next stages investigated, 56 and 73 dpo, parts of the apical trabeculae appeared to be organized in a more compact fashion (Fig. 3D,E). The vertical septum started to appear as a prominent sheet of trabeculae beneath the atrioventricular valves. Subepicardial coronaries were detected by 73 dpo, by a high density of DAPI (the erythrocytes are nucleated and DAPI positive), but the septal components appeared without coronary vasculature (Fig. 3F). At 88 and 144 dpo, compact muscle enclosed the cavum arteriosum and venosum and thus separated these parts from the cavum pulmonale, except at the gap between the muscular ridge and the bulbuslamelle (Fig. 3G,H). The bulboauricular lamellae were prominent (Fig. 3G). The atrioventricular canal remained positioned to the left of the aortic base. At these late gestational stages, the lumen of the cavum pulmonale was approximately 50% of the total ventricular lumen, but this proportion was less than 30% in two juvenile specimens (Fig. S2). In the 144 dpo specimen, a coronary artery with an approximately 50 µm diameter orifice had its origin in the sinus of the dorsal leaflet of the right aorta. Although we could detect coronary vessels in the outer compact walls, comparable vessels were not found in the compact muscle of the septa (Fig. 3H). In the hearts of two juveniles and two adult savannah monitors, the muscular ridge, the bulbuslamelle and the outer compact wall did contain coronary vessels (Fig. S3).

Septation of the monitor ventricle appears to involve compaction of trabeculae, much like it does in chicken. Therefore, we hypothesized that the monitor septa would express genes associated with the development of the full ventricular septum.

For in situ hybridization analysis of embryonic lizards, we compared monitors with Norops sagrei, which has a typical lizard heart. We focused on the stages when septal structures have formed and when the ventricle has grown to the extent that its proportion of the total cardiac mass is almost the same as that of adult animals, i.e. ∼70-80% (Fig. S2); in Varanus acanthurus this corresponded to ∼35 dpo and in Norops sagrei to Sanger stage 11 (Jensen et al., 2013; Sanger et al., 2008). In the monitors, Tbx3 was expressed in the bulboauricular lamellae from the atrioventricular canal to the vicinity of the vertical septum (Fig. 4A). This expression domain, along with the bulboauricular lamellae, extended deeper into the ventricle than it did in Norops (Fig. 4A,B). In the monitors, Tbx5 was expressed throughout the ventricle, with no gradient between the left and right sides (Fig. 4C). The myocardial outflow tract did not express Tbx5. Within the ventricle, Tbx5 was particularly rich in the bulboauricular lamellae and therefore enriched near the vertical septum and in the inner-most parts of the muscular ridge and the bulbuslamelle (Fig. 4C). In Norops, the bulboauricular lamellae were less developed and they were not enriched in Tbx5 expression (Fig. 4D). In the monitors, Irx1 and Irx2 were expressed in the bulboauricular lamellae and therefore near the vertical septum and in the inner-most parts of the muscular ridge and the bulbuslamelle (Fig. 4E,G). Irx1 and Irx2 were similarly expressed in the ventricles of the monitors and Norops, although the patterns were more pronounced in the monitors (Fig. 4E-G). Fig. 4H shows a 3D reconstruction of the myocardium in the Varanus indicus ventricle (56 dpo), which expressed Irx1, Irx2 and Tbx5. In the monitors, the inner-most parts of the muscular ridge and the bulbuslamelle associated with one large cushion each (Fig. 4C,E,G, black arrows), which persisted after hatching (Fig. S4). Similar cushions are found in other squamates with ventricles with pressure separation (Jensen et al., 2014; Jensen et al., 2010). In Norops, Irx1 expression extended beyond the domain of Tbx5 expression and into the outflow tract myocardium (Fig. 4F, Fig. S5). Fig. 4I and 4J show Irx1 and Irx2 expression in the septal region of chicken and mouse, respectively. Outside the heart, Tbx3, Tbx5, Irx1 and Irx2 were also detected in the developing trachea and bronchi (Fig. 4), and Irx1 and Irx2 in the spinal cord, homologous to the expression patterns observed in mouse and chicken (Chapman et al., 1996; Christoffels et al., 2000b; Gibson-Brown et al., 1998).

We have previously shown specific detection of Myh6 expression in the atrioventricular junction of the American alligator, using a Norops Myh6 probe (Jensen et al., 2018). In embryonic mammals and chicken, Myh6 is initially expressed in the ventricle but becomes confined to the forming atrioventricular bundle (de Groot et al., 1987; Moorman et al., 1998; Yutzey et al., 1994). We investigated whether transcripts of Myh6 associated with particularly developed bulboauricular lamellae. Across a phylogenetic diverse set of non-crocodylian reptiles, Myh6 was detected in the atria and atrioventricular canal of all species (Fig. 5). Only in monitor lizards, however, did we observe Myh6 expression in the bulboauricular lamellae in the ventricle (Fig. 5). In the leopard geckos, Myh6 was expressed in a pattern similar to that observed in Norops, i.e. readily detected in the atria (and sinus venosus), but not in the ventricle (Fig. 5). Fig. 5 shows sections of both adult and developing hearts. We then compared multiple stages of developing Norops and monitors, but only in the monitors did we detect prominent expression in the bulboauricular lamellae (Fig. S6).

Taken together, the expression of genes associated with the ventricular septum of mammals and birds was found around the ventricular septal structures of monitors. The myocardium immediately adjacent to the vertical septum of the monitors had a molecular phenotype comparable to that of the atrioventricular bundle of mammals and birds, which led us to hypothesize that this myocardium would be activated early, as it is in mammals and birds.

Optical mapping of cardiac action potentials was performed on six ridge-tailed monitor embryos ranging from 21 to 46 dpo, spanning the period when ventricular septa are formed. Despite substantial growth of the embryo and heart during this period, the intrinsic heart rate, atrial activation time, atrioventricular delay and ventricular activation times remained essentially constant (Fig. S7). The left side of the ventricle was consistently activated earlier than the right side (Fig. 6A,C). On the left side, the earliest epicardial activation was always observed in the vicinity of the forming vertical septum. In the oldest embryos, this was seen as a site of activation relatively deep in the ventricle (Fig. 6C), which was different from the primitive base-to-apex activation of the leopard geckos (n=11; Fig. 6B). This pattern of activation could be elicited by stimulation in the apical region of the left chamber (only performed in the oldest and largest specimens; data not shown).

Monitor lizards are the only lizards we know of with a divided ventricle. Here, we establish that the dividing septa express Irx1, Irx2, Tbx3 and Tbx5. In all studied ventricles with haemodynamically distinct left and right sides, i.e. human, mouse, chicken and crocodylians, these transcription factors are expressed in the septal structures and deficiency of Tbx3 and Tbx5 are known to cause ventricular septal defects (Bakker et al., 2008; Bruneau et al., 1999, 2001; Christoffels et al., 2000b; Hoogaars et al., 2004; Jensen et al., 2018; Sizarov et al., 2011; Yamada et al., 2000). In mammals, chicken and crocodilians, Tbx3 delineates the atrioventricular conduction axis (Jensen et al., 2018) and the deeper ventricular expression of Tbx3 in the monitors was associated with a deeper initial electrical activation. Only in the monitors did the expression of these transcription factors colocalize with the expression of Myh6, which resembles the setting in four-chambered hearts, whereas Myh6 was not found in the ventricles of the six non-varanid species we investigated. Therefore, we consider it likely that conserved molecular mechanisms underlie the processes of ventricular septation in all known instances. Our results suggest that the various degrees of septation encountered in Sauropsida (reptiles and birds) and Synapsida (mammals) are variations of the same building plan inherited from their common amniote ancestor. Incidentally, we found the same transcription factors to be expressed in the trachea and bronchi, as have been observed in the developing trachea and bronchi of mouse and chicken (Chapman et al., 1996; Christoffels et al., 2000b; Gibson-Brown et al., 1998). Therefore, homologous molecular mechanisms also appear to drive the formation of the main airways despite the considerable anatomical differences in the formed respiratory systems of reptiles and mammals (Farmer and Sanders, 2010; Perry and Sander, 2004; Schachner et al., 2014).

The muscular ridge and the bulbuslamelle could both be identified by the compact organization of their myocardium. In the early stages, this myocardium blended into apical trabeculae, whereas at later stages the compact muscle of the muscular ridge and the bulbuslamelle could be followed to the apex. This strongly suggests that some trabeculae had undergone compaction, seemingly in the absence of coronary vasculature. Compaction is considered a key process of the formation of the ventricular walls and septum of mammals and birds (Captur et al., 2015; Sedmera et al., 2000), although the importance of the process may vary between species and its contribution to pathology has been debated (Anderson et al., 2017; Finsterer et al., 2017). Whereas the process of compaction is generally thought to be the coming-together of trabeculae (Anderson et al., 2017; Finsterer et al., 2017), foundational observations made in chicken suggested that compaction is the thickening of trabeculae to the extent to which the inter-trabecular spaces become obliterated (Rychterova, 1971). The trabeculae near the ventricular septa of the developing monitors had a dense organization as if the trabeculae were coming together (Fig. S1E-G), but they also appeared to be thicker like the trabeculae in the cavum pulmonale (Fig. 3D). It is possible that both processes of coming-together and thickening of the trabeculae contribute to compaction in monitors.

The muscular part of the human ventricular septum has been described by anatomists as having distinct components, specifically an inlet part (dorsal/posterior and a-trabecular), an apical part (surrounded by trabeculae), and an outlet part (ventral/anterior and without trabeculae) (Anderson and Becker, 1980; Van Mierop and Kutsche, 1985; Wenink, 1981). Molecular markers that distinguish these components remain to be found. Comparative anatomical studies concluded that the outlet part corresponds to the muscular ridge (Benninghoff, 1933; Poelmann et al., 2014; Van Mierop and Kutsche, 1985; Webb, 1979). Indeed, the muscular ridge of non-crocodylian reptiles expresses Tbx5 in a gradient, as does the ventricular septum of mammals, chicken and crocodylians (Jensen et al., 2018; Koshiba-Takeuchi et al., 2009; Poelmann et al., 2014). Because the muscular ridge is not positioned beneath the atrioventricular junction it cannot be equated wholesale to the muscular part of the full ventricular septum (Benninghoff, 1933; Poelmann et al., 2014; Van Mierop and Kutsche, 1985; Webb, 1979). A gradient of Tbx5 is also present on the ventral-most part of the bulbuslamelle. However, in mammals and birds this part seems to contribute to the right atrioventricular valve rather than the septum (Jensen and Moorman, 2016; Jensen et al., 2013; Lamers et al., 1995). The dorsal-most part of the bulbuslamelle is thought to contribute to the ventricular septum (Benninghoff, 1933; Greil, 1903; Jensen et al., 2014; Webb, 1979), but this part was not enriched for any of the investigated transcripts and did not have a gradient of Tbx5. Therefore, when comparing the muscular ridge and the bulbuslamelle of monitors with the ventricular septum of mammals and birds, there is a shared expression of key transcription factors but a divergence in the anatomy of the myocardial tissues that express these genes. In reptiles, the vertical septum is beneath the atrioventricular junction and the adjacent myocardium was enriched in transcripts associated with the atrioventricular bundle of mammals and archosaurs (Bakker et al., 2008; Jensen et al., 2018). The early electrical activation near the vertical septum corroborates this interpretation. The final component of the full ventricular septum is the membranous septum (Cook et al., 2017; Greil, 1903; Jensen et al., 2014; Webb, 1979). The membranous septum develops from mesenchyme of the atrioventricular cushions and of the proximal part of the two mesenchymal ridges of the myocardial outflow tract (Anderson et al., 2014, 2019; Asami, 1969). Two similar mesenchymal ridges can be found in the developing outflow tract of reptiles, which apparently persist in monitors and pythons as large cushions of connective tissue at the point where their muscular ridge and bulbuslamelle abut (Greil, 1903; Jensen et al., 2014, 2010; Poelmann et al., 2017). If, in a reptile setting, these cushions were to fuse, not only would the ventricle be fully divided, but the membranous septum would also be closely aligned with the arterial wall that separates the pulmonary artery from the aorta (Fig. S4C).

If the cushions of the muscular ridge and the bulbuslamelle fused, the right atrial blood would only fill the cavum venosum and would not be able to reach the cavum pulmonale behind this hypothetical membranous septum. Interestingly, in developing mammals and archosaurs, the atrioventricular canal undergoes a pronounced right-ward expansion during the process of ventricular septation whereby the right atrium remains in contact with the right ventricle (Jensen and Moorman, 2016; Jensen et al. 2013; Lamers et al., 1995). This process includes the formation of the right atrioventricular valve. Failed expansion results in congenital malformations, such as tricuspid atresia and double inlet left ventricle, which have been reported in multiple mammalian species (Jensen and Moorman, 2016; Michaëlsson and Ho, 2000). In squamate reptiles with an undivided ventricle, the atrioventricular canal is positioned to the left of the body midline (Jensen et al., 2013) and does not undergo a rightward expansion (Jensen and Moorman, 2016). That the atrioventricular canal of monitors remains left-sided, shows that the right-wards remodelling of the atrioventricular canal associates with full ventricular septation only. Consequently, the cavum venosum of monitors serves as the conduit for systemic venous blood to the cavum pulmonale in ventricular diastole and serves as the conduit for pulmonary venous blood in systole as it does in other squamates (Fig. 2D-E). So-called washout shunting is therefore inherent to the monitor heart (Heisler and Glass, 1985). The wash-out of systemic venous blood to the systemic arterial circulation, or so-called right-to-left shunting, is greater than left-to-right shunting in monitors (Heisler and Glass, 1985; Heisler et al., 1983; Ishimatsu et al., 1988), consistent with our finding that the cavum pulmonale is smaller than the cavum venosum and cavum arteriosum.

Anatomists and physiologists have long emphasized that the heart of monitors can be seen as a conceptual stage between the hearts of non-crocodylian reptiles and the fully septated hearts of crocodylians, birds and mammals (Acolat, 1943; Brücke, 1852; Greil, 1903; Jensen et al., 2014). The primitive (left-sided) position of the atrioventricular canal concomitant with the specialized state of deep left ventricular activation corroborates this notion. Comparing lizards, crocodylians, birds and mammals, we find considerable overlap in the functional anatomy, electrophysiology, and molecular phenotype of the septation of the ventricles of amniotes. Our results suggest that Sauropsida and Synapsida inherited precursor structures from their common amniote ancestor. Particular features evolved within each lineage (Cook et al., 2017), but we propose that they could be seen as variations on a common design. In the special case of monitor lizards, we suggest that ventricular septation relies on evolutionarily conserved building blocks and results in a convergence of function despite a divergence of appearance.

Of the savannah monitor (Varanus exanthematicus, Bosc 1792), we used one adult for echocardiography (body mass ∼2 kg), hearts of two adults (unknown body mass) and two juveniles (body mass 18 and 22 g) for histology stained with Picro-Sirius Red. We used the heart of one 11-year-old water monitor (V. salvator, 10 kg), visualized with MRI, that was donated to us (B.J.) after the animal died from senescence. From six fertilized eggs of ridge-tailed monitor (Varanus acanthurus, Boulenger 1885) that were incubated at 30°C and 80-90% humidity, we investigated hearts isolated on the following dpo: 15, 21, 25, 32, 39 and 46. From 11 fertilized eggs of leopard gecko (Eublepharis macularius, Blyth 1854) that were incubated similarly, hearts were isolated on the following days: 7, 11, 15 (n=2), 18, 26 (n=2), 32, 40, 47 and 53. We further made use of a previously published developmental series of the mangrove monitor (Varanus indicus, Daudin 1802) (Gregorovicova et al., 2012). The hearts of this series were used for immunohistochemistry of heart muscle only, as initial tests showed that the tissue preservation was not good enough for the detection of most antigens and gene transcripts by in situ hybridization. Of the brown anole (Norops sagrei, Duméril and Bibron 1837), we investigated embryos of Sanger stages (Sanger, Losos, and Gibson-Brown, 2008) 7 (n=1), 9 (n=1), 11 (n=2), 16 (n=1), and one adult, approximately 1 year old. The handling of the adult brown anole complied with Dutch national and institutional guidelines (Amsterdam UMC, The Netherlands) and with the Institutional Animal Care and Use Committee of the University ratified approval registered as ‘DAE101617’. All data were generated within the time limits of DAE101617.

One adult savannah monitor (V. exanthematicus) was anaesthetized with isoflurane and intubated for manual ventilation with room air containing 1.5% isoflurane. Images were obtained at room temperature while the lizard was placed on its back on a heating pad set at 25°C. We used a Vivid 7 echocardiographic system (GE Healthcare) with an 11 MHz phased array paediatric transducer operating at a frame rate of 60 Hz. Two-dimensional images were recorded in the three principal planes of the body (sagittal, transverse and frontal) and stored for off-line evaluation on a workstation (EchoPac Dimension 06, GE Healthcare).

Optical mapping was performed on V. acanthurus and Eublepharis macularius. Shell and membranes were removed and the embryos were placed in a dish containing ice-cold reptilian Ringer solution (specific Ringer solutions for Anolis adopted from Jensen et al., 2012); in mmol/l: NaCl 95, Tris 5, NaH₂PO₄ 1, KCl 2.5, MgSO₄ 1, CaCl₂ 1.5, glucose 5, pH adjusted to 7.5 with HCl). The heart and adjacent posterior body wall structures were isolated and stained in 2.5 mmol/l di-4-ANEPPS (Invitrogen) for 10 min. Contractions were inhibited with cytochalasin D (0.2 μM, Sigma) and the torso was pinned to the bottom of the dish containing oxygenated Ringer solution at 28°C. Imaging was performed from both the ventral and dorsal surfaces at 0.25-1 kHz. Ventricular pacing was performed in the three oldest specimens of both species using a platinum electrode at 125% of the intrinsic rate and twice the diastolic threshold (Sedmera et al., 2003). Data acquisition and analysis were performed using an Ultima L high-speed camera and bundled software (Sankova et al., 2012). Epicardial activation maps were then constructed separately for the ventral and dorsal (as well as lateral, when available) ventricular surfaces in sinus and stimulated rhythm.

For V. acanthurus, the hearts and surrounding tissues were fixed for 1 day in freshly made 4% paraformaldehyde in PBS (0.9% NaCl) and then washed twice in PBS before being stored in 70% ethanol. For in situ hybridization, we used previously published probes for Anolis mRNA (Tbx3, Tbx5, Ttnnt2, Myh6; Jensen et al., 2018; Jensen et al., 2012) and new probes based on the following coordinates using UCSC Genome Browser on Lizard May 2010 (Broad AnoCar2.0/anoCar2) Assembly; Irx1 (chrUn_GL343292:777,311-778,155), Irx2 (chrUn_GL343292:1,436,684-1,440,677). Tests with Anolis Bmp2 and Gja5 (Cx40) (also from Jensen et al., 2012) did not show specific stain in Varanus. In situ hybridizations were performed as described previously (Moorman et al., 2001) and performed on a series of sections of 8, 10 or 12 µm thickness that contained the entire heart of specimens of all V. acanthurus. The data displayed in Fig. 4I,J were generated in connection with the study by Christoffels et al. (2000a). For immunohistochemistry, sections of the embryonic hearts of V. indicus were cooked under pressure for 5 min in antigen retrieval solution (H-3300, Vector Laboratories). Myocardium was bound with a mouse polyclonal antibody to cardiac troponin I (Millipore, RRID: AB_11212281, 1:250). The antibody was detected with a fluorescently labelled secondary donkey-anti-mouse antibody (Invitrogen, RRID: AB_2556542, 1:200) conjugated to Alexa Fluor 488. Subsequently, sections were mounted using glycerol-PBS (1:1) containing SYTOX Blue (Invitrogen, S11348, 1:40.000) or DAPI (Sigma-Aldrich, D9542, 1:1000) for nuclear staining.

For the MRI acquisition of the heart of the V. salvator, a 3.0T MR scanner (Ingenia; Philips Medical Systems, Best, The Netherlands) was used. For signal reception, a 32-channel head coil was used. For the 3D MRI acquisition, a T1 weighted turbo-field echo (TFE) pulse sequence was used with the following imaging parameters: TR/TE=6.7/2.9 ms, FA=20°, acquired voxel size=0.5×0.5×0.5 mm, NSA=2. Total scan duration was 7 min 23 s. The presented image is a thin MinIP (minimal intensity projection) magnetic resonance image reconstruction (5 mm slice thickness).

Reconstructions were made in Amira software (FEI) version 5.5 as described previously (Soufan et al., 2003). In the few cases in which a section was damaged (less than 10% of all images in a given preparation), its image was replaced by the nearest intact sister section. Images were aligned using the Align tool, including automatic alignment supplemented by manual fitting. Once aligned, signal was annotated using a manually set threshold. Surface files were made of label files that were resampled to a voxel size of approximately 10×10×10 µm.

We received valuable help from Corrie de Gier-de Vries with in situ hybridization and immunohistochemstry, and from Jaco Hagoort with Amira reconstructions.