A Subpopulation of Smooth Muscle Cells, Derived from Melanocyte-Competent Precursors, Prevents Patent Ductus Arteriosus

Background Patent ductus arteriosus is a life-threatening condition frequent in premature newborns but also present in some term infants. Current mouse models of this malformation generally lead to perinatal death, not reproducing the full phenotypic spectrum in humans, in whom genetic inheritance appears complex. The ductus arteriosus (DA), a temporary fetal vessel that bypasses the lungs by shunting the aortic arch to the pulmonary artery, is constituted by smooth muscle cells of distinct origins (SMC1 and SMC2) and many fewer melanocytes. To understand novel mechanisms preventing DA closure at birth, we evaluated the importance of cell fate specification in SMC that form the DA during embryonic development. Upon specific Tyr::Cre-driven activation of Wnt/β-catenin signaling at the time of cell fate specification, melanocytes replaced the SMC2 population of the DA, suggesting that SMC2 and melanocytes have a common precursor. The number of SMC1 in the DA remained similar to that in controls, but insufficient to allow full DA closure at birth. Thus, there was no cellular compensation by SMC1 for the loss of SMC2. Mice in which only melanocytes were genetically ablated after specification from their potential common precursor with SMC2, demonstrated that differentiated melanocytes themselves do not affect DA closure. Loss of the SMC2 population, independent of the presence of melanocytes, is therefore a cause of patent ductus arteriosus and premature death in the first months of life. Our results indicate that patent ductus arteriosus can result from the insufficient differentiation, proliferation, or contractility of a specific smooth muscle subpopulation that shares a common neural crest precursor with cardiovascular melanocytes.


Introduction
The ductus arteriosus (DA) is a normal fetal artery linking the aortic arch to the left pulmonary artery in mammals (bilateral in birds), and allowing the blood to bypass the lungs in utero (reviewed in [1]). The blood circulation of amniotes then changes dramatically at birth. Once the lungs ventilate, blood flow reverses as pulmonary resistance decreases. The higher oxygen levels induce onset of DA closure through functional constriction of its muscular tunica media. This process initiates proliferation, migration, extracellular matrix production, and, through resultant hypoxia, the apoptosis of the cells forming the DA.
After birth, full closure of the DA can fail (patent ductus arteriosus or PDA), potentially leading to cardiac failure. In most cases, maintenance of an increased volume load to the left heart causes dilation of the left cardiac cavities associated with a progressive pulmonary vascular remodeling due to excess volume in the pulmonary artery bed. This compromises postnatal health by leading to respiratory complications such as pulmonary hypertension and edema. PDA is a frequent problem affecting premature children, with a prevalence greater than 40% in infants with a birth weight of 1.5 kg or less [2].
The final steps of DA closure include a drop in the levels of vasoactive hormones such as prostaglandin E2 (PGE2), which plays a critical role in maintaining the necessary open status of the DA during fetal development, followed by remodeling of the tunica media under the influences of hypoxia. In PDA due to prematurity, closure can usually be achieved by treatment with cyclooxygenase inhibitors such as indomethacin or ibuprofen, which block prostaglandin synthesis [3]. Such drug-induced closure is more difficult to obtain in cases of PDA in at-term newborns: the rate of success in these cases is only about 30% [3]. A properly closed DA ultimately undergoes physiological fibrosis to evolve into the ligamentum arteriosum (LigA).
In humans, mutations in TFAP2B (encoding the AP-2b transcription factor) and MYH11 have been identified in cases of isolated, inherited PDA [4,5], but most cases are believed to be multifactorial in origin and indeed often present as part of a syndromic spectrum, suggesting that the condition is associated with a complex genetic inheritance. A number of PDA mouse models have been produced, in which lack of DA closure generally leads to death within three days of birth (P3). Single gene mutations in Hpgd, Ptger4, Ptgs2, Foxc1 or Myh11 result in the absence of, or delay in, DA closure [6,7,8,9,10,11]. In some cases, affected mice may be rescued by indomethacin injection [7]. Double mutant mice, in which Cox-1 and Cox-2 are disrupted, also exhibit PDA [12]. A deletion encompassing 24 genes between the Dgcr2 and Hira loci on chromosome 16 leads to death with the presence of an open DA, as does a nested deletion comprising Tbx1 and three other genes [13]. Deletion of the equivalent chromosomal region in humans is associated with the complex DiGeorge syndrome, of which PDA is a common feature [14,15]. The somatic ablation of myocardin (Myocd), using Wnt1::Cre or Pax3::Cre mice, leads to the failure of DA closure [16]. These two promoters are active in early postmigratory neural crest cell derivatives, comprising the smooth muscle cells (SMC) of the DA.
The heart and its associated great arteries are composed for the most part of cells derived from the mesoderm, but with a critical contribution from neural crest cells [17]. Vagal neural crest cells (VNCC) can differentiate into numerous derivatives including neurons and glial cells of the enteric, autonomic and somatic nervous systems, SMC, mesenchymal cells and melanocytes (Mc). The tunica media of the DA, which like other pharyngeal archremodeled arteries consists mainly of SMC, is derived from these VNCC [18,19,20] [21,22,23].
Recently, Mc have been observed in the heart, in the valves and septa and in the DA/LigA [22,24,25,26]. Mouse lines producing the Cre recombinase under the control of the tyrosinase promoter (Tyr::Cre) [27] have been used to show that the DA/LigA contains cells derived from VNCC: a very small number of Mc (pigmented and recombined cells), SMC2 (non-pigmented and recombined cells, estimated to constitute 10-20% of the DA/LigA cell population) and SMC1 (non-pigmented and non-recombined cells, estimated to make up to 80-90% of the DA/LigA cell population) [22]. The tyrosinase promoter [28] is transiently active around day 9 of gestation (E9) in Mc precursors, called melanoblasts (Mb) and SMC2 precursors of the DA [22].
Activation of Wnt/b-catenin signaling favors the Mc fate in multipotent progenitors [29,30]. Here, we generated mice that produce a constitutively active form of b-catenin (ctnnb1Dex3) in cells having expressed tyrosinase, to both explore the SMC2/Mc bipotency of a subset of VNCC and to evaluate the anatomical consequences of cell fate alterations in the DA.

Melanoblasts replace a subpopulation of smooth muscle cells in ctnnb1Dex3 mice
Recombinant ctnnb1Dex3 mice were generated by producing an activated form of b-catenin in cells of the Tyr::Cre lineage. LoxP sequences had previously been inserted into introns 2 and 3 to flank the third exon of the gene encoding b-catenin (ctnnb1), for which the allele is also known as ''f3'' [31]. Exon 3 encodes serine and threonine residues involved in phosphorylation and degradation of the protein. The resulting modified b-catenin is reportedly more stable [32]. We verified in vitro that the deletion of exon 3 substantially increases protein activity (by five-fold) in B16 melanoma cell lines (Figure 1), as has been demonstrated in vivo for recombined cells of the intestinal epithelium [31].
Hemizygous Tyr::Cre/u mice were then crossed with heterozygous or homozygous floxed ( = non-recombined) ctnnb1Dex3/+ or ctnnb1Dex3/ctnnb1Dex3 mice to produce Tyr::Cre/u; +/+ (WT) and Tyr::Cre/u; ctnnb1Dex3/+ (ctnnb1Dex3) siblings (Table S1). These WT and ctnnb1Dex3 mice were initially backcrossed on a Dct::LacZ background, to visualize Mb in the DA. Dct is a member of the tyrosinase gene family and encodes the enzyme dopachrome tautomerase (also known as tyrosinase-related protein 2 or Trp2) involved in eumelanin synthesis. In Dct::LacZ/u reporter mice, the transgene LacZ, under the control of the Dct promoter, is expressed in Mb, retinal pigmented epithelium cells, and in the forebrain, but not in vascular smooth muscle cells. Tyr::Cre/u; +/+; Dct::LacZ/ u ( = WT-Dct) and Tyr::Cre/u; ctnnb1Dex3/+; Dct::LacZ/u ( = ctnnb1Dex3-Dct) hearts were isolated, fixed and stained with Xgal on embryonic day (E)18.5. Whole-mount observation revealed that there were visibly more Mb in the mutant than the WT DA ( Figure 2A-D). Transverse sections confirmed that Mb were scarce in the WT-Dct DA, consistent with a previous study [22], but were numerous (60623 Mb per section) in ctnnb1Dex3-Dct DA ( Figure 3A, B).
The numbers of SMC2+Mb per section were very similar in the DA of WT-Rosa and ctnnb1Dex3-Rosa (around 70) ( Figure 5). However, the numbers of SMC2 alone were 62629 versus 563 per section in WT-Rosa and ctnnb1Dex3-Rosa, respectively, while the numbers of Mb were 565 versus 5,80063,400 per 100 sections in WT-Rosa and ctnnb1Dex3-Rosa, respectively. These various observations show that on a ctnnb1Dex3 background, Mb replaced most of the SMC2 in the DA. In other words, a subset of VNCC is bipotent for SMC2 and Mc, and responds to signaling through bcatenin.
ctnnb1Dex3 mice have a greatly dilated left atrium, which develops after birth As expected and according to specific crosses, Tyr::Cre/u; ctnnb1Dex3/+ mice were produced in a Mendelian ratio (72 u/u; ctnnb1Dex3/+ versus 67 Tyr::Cre/u; ctnnb1Dex3/+ live births). However, all ctnnb1Dex3 mice died between 4 and 18 weeks of age ( Figure 6A). A distinct Tyr::Cre mouse line, Tyr::CreB, was used   to see if the site of integration of the transgene or other cisregulatory differences would affect the outcome [27]. Nonetheless, all Tyr::CreB/u; ctnnb1Dex3/+ mice died within a similar age range (between 3 and 12 postnatal weeks; Figure 6B), indicating that the death of these animals was independent of the Tyr::Cre mouse line used.
The main clinical sign presented by ctnnb1Dex3 mice was prostration 24 hours prior to death. Such prostrated, heterozygous ctnnb1Dex3 mice and their wild-type siblings were sacrificed and autopsied. While never observed in the WT controls ( Figure 7A, C), a major dilation of the left atrium was observed macroscopically in all mutant mice ( Figure 7B, D) and a lesser dilation of the left ventricle was observed in about half of the mice ( Figure 7D), Micro-computed tomography (CT) analysis after FenestraH injection was performed on WT and ctnnb1Dex3 mice ( Figure 7E-H and movies S1, S2). This revealed a rightward shift of the whole heart in live mutant animals, with a clear dilation of the left atrium ( Figure 7G, H).
WT and ctnnb1Dex3 hearts were examined at P1, P10 and P28. At P1, WT and ctnnb1Dex3 hearts were comparable, indicating that the enlargement of the left atrium was not associated with a congenital malformation due to a developmental defect ( Figure 8A, D). At P10, the ctnnb1Dex3 left atrium was visibly dilated ( Figure 8B, E). By P28, the expansion of the left atrium was substantial ( Figure 7B, D and Figure 8C, F). Some dilation of the left ventricle, not hypertrophy, was also visible in a limited number of mutants by P28 ( Figure 7D and Figure 8F).
Echocardiographic examination also showed substantial enlargement of the left atrium, associated with thrombi of various sizes ( Figure 9). Thrombus formation was subsequent to chamber dilation and aggravated over time ( Figure 9C, D). In one case, atrial myocardial rupture and pericardial blood effusion was observed on ultrasound analysis, and led to the death of the mouse. Atrial dilation was unlikely to be due to mitral valve dysfunction, because the Doppler mitral inflow pattern was normal (data not shown). Similarly, left-ventricular failure was not the cause, because the fractional shortening was normal. These data led to the hypothesis that death was due to the accumulation of large thrombi in the left atrium and/or its rupture.

The ductus arteriosus is not fully closed in ctnnb1Dex3 mice
When the DA does not fully close after birth, a part of the systolic left-ventricular stroke volume goes directly into the pulmonary artery (left-to-right shunting), leading to a progressive overload of the pulmonary circulation by an increase in pulmonary pressure. Simultaneously, this volume overload triggers the progressive dilation of the left cardiac cavities. Surprisingly, in ctnnb1Dex3 mutant mice, this enlargement affected essentially the left atrium, the left ventricle being more modestly and not systematically dilated. Ultrasound analysis demonstrated that the DA remained open in postnatal ctnnb1Dex3 mice, which is never the case in WT mice ( Figure 10A, B). Ultrasound ( Figure 10C) and color Doppler analyses ( Figure 10D) showed blood flow back through the patent foramen ovale ( Figure 10E) from the right to the left atrium of all ctnnb1Dex3 mice, but not WT mice.
The death of the ctnnb1Dex3 mice thus seemed to result from the failure of DA closure and increased pulmonary pressure, leading ultimately to retrograde blood flow through the foramen ovale from the right to the left, driving the progressive dilation of the left atrium and thrombus formation.  After birth, the very few Mc in the DA normally differentiate, produce melanin and remain in the LigA. The number of pigmented cells was substantially higher in ctnnb1Dex3 LigA than WT LigA by four weeks ( Figure 11A, B). Histological analysis revealed that the ctnnb1Dex3 LigA was not fully closed ( Figure 11C-F). The surface areas of the intimal cushion and lumen, when present, were larger in ctnnb1Dex3 than WT mice ( Figure 11G, H). Moreover, blood was observed in the canal and large numbers of pigmented cells were present in the tunica media of the ctnnb1Dex3 LigA ( Figures 11E, F). Only ctnnb1Dex3 mice presented a PDA, as well as disorganized, fibrotic lung alveolar structures ( Figure 12). These observations suggested that abnormally high pulmonary pressure may have been the result of the PDA and cause lung damage in ctnnb1Dex3 mice as compared to WT mice.

ctnnb1Dex3 mice can be partially rescued by indomethacin
In humans, indomethacin is widely used to treat PDA, by inhibiting the cyclooxygenases that participate in prostaglandin biosynthesis. As a proof of concept, pregnant Tyr::Cre/Tyr::Cre; +/ +; Dct::LacZ/Dct::LacZ females that had been crossed to ctnnb1Dex3/+ males, and thus carrying litters with both WT-Dct and ctnnb1Dex3-Dct embryos, were injected with indomethacin at E18.5 and compared with mock-injected controls. Embryos were removed from three indomethacin-injected females, 4 hours after injection. Five WT-Dct and five ctnnb1Dex3-Dct hearts were   Indomethacin induced the closure of the DA as expected, in WT-Dct mice, but also by means of the remaining SMC in ctnnb1Dex3-Dct DA.
Newborn ctnnb1Dex3 pups were also directly subcutaneously injected with indomethacin at birth, and this significantly improved their survival rate and indeed cured about twenty per cent of animals (n = 29; Figure 13E). Therefore, ctnnb1Dex3 mice could be partially rescued by indomethacin treatment, suggesting that one of the primary causes of death in ctnnb1Dex3 mice was the failure of full DA closure at birth. However, we cannot exclude the possibility that the left atrium is structurally abnormal in ctnnb1Dex3 mutants, possibly contributing to the death of the animals.
b-catenin does not lead to overall Cox-2 induction in the DA b-catenin can directly induce the expression of Ptgs2 (encoding cyclooxygenase-2 or Cox-2) and stabilizes its mRNA by interacting with AU-rich elements of the 39-UTR in vitro [33,34]. Cox-2 in turn is known to catalyze the formation of PGE2, and elevated levels of PGE2 are associated with an open DA.
One possible explanation for the failure of DA closure in ctnnb1Dex3 mice is that unduly high levels of Cox-2 were induced by the augmented b-catenin signaling. Therefore, we verified that Cre-recombined, truncated ctnnb1Dex3 b-catenin mRNA was detectable in ctnnb1Dex3 DA but not in WT DA, as determined by extracting total RNA from DAs and performing semiquantitative RT-PCR ( Figure 14A, B). b-catenin was visible in both cytoplasm and nuclei, as assessed by immunofluorescence, in both ctnnb1Dex3-Dct and WT-Dct Mb; as expected, more Mb were visible in the former on sections ( Figure 14C-H). Nonetheless, the amount of Ptgs2 (Cox-2) mRNA in ctnnb1Dex3 DA was similar to that in WT DA ( Figure 14A), indicating that increased b-catenin signaling in ctnnb1Dex3 DA did not result in massive induction of Cox-2. Therefore, Cox-2, although a target  ctnnb1Dex3-mi mice lacking melanocytes still die of heart failure Mitf-deficient mice (mi = mitf vga9/vga9 ) have no Mc and a white coat, resulting from a recessive null allele for the microphthalmiaassociated transcription factor. In mi mice, Mc are genetically ablated during development at around E11.5, which is after specification of the common SMC2/Mc precursor at about E9.0, but well before birth [22]. To evaluate the contribution of Mc to the full closure of the ctnnb1Dex3 DA at birth, ctnnb1Dex3-mi (Tyr::Cre/u; ctnnb1Dex3/+; mitf vga9/vga9 ) mice were produced. As anticipated, the ctnnb1Dex3-mi mice lacked cutaneous Mc and were white, like mi mice (data not shown). Surprisingly, ctnnb1Dex3-mi mice showed clinical signs similar to ctnnb1Dex3 mice and died at similar ages ( Figure 15A). Left atrium enlargement was observed in ctnnb1Dex3-mi mice at P28 ( Figure 15B, C). PDA was observed at P2 in ctnnb1Dex3-mi, but not in mi mice ( Figure 15D, E).
Additional mice were generated in order to estimate the numbers of Mb, through b-galactosidase activity, and SMC through SMA immunoreactivity, in transverse sections of the DA. Tyr::Cre/u; ctnnb1Dex3/+; mi vga9/vga9 ; Dct::LacZ/u (ctnnb1Dex3-mi-Dct) and Tyr::Cre/u; +/+; mi vga9/vga9 ; Dct::LacZ/u (mi-Dct) mice were compared to ctnnb1Dex3-Dct mice (cf. Figures 2 and 3). No Mb were observed in the DA of ctnnb1Dex3-mi-Dct or mi-Dct, in contrast to the ctnnb1Dex3-Dct DA. However, the numbers of SMC were similar in ctnnb1Dex3-Dct and ctnnb1Dex3-mi-Dct DA and reduced relative to the DA of mi-Dct, such that no cellular compensation for SMC was observed in ctnnb1Dex3-mi-Dct DA ( Figure 15F).
The general heart morphology appears disrupted in ctnnb1Dex3-mi mice compared to mi mice ( Figure 15B, C), as this is the case for ctnnb1Dex3 mice compared to WT mice ( Figure 7C, D). A conjunction of events may explain these phenomenona, which are potentially linked, but are not yet understood. These abnormalities could be due to the general blood flow defect in the mutant mice, leading to abnormal pressures on the chambers, thus affecting the morphology. However, it could be associated with melanocytes. Indeed, Mc are present in the heart, and not only in the DA [25,26]. A careful analysis allowed the visualization of a general increase in Mc numbers in different parts of ctnnb1Dex3 hearts. It includes the foramen ovale ( Figure 10E), and the mitral, tricuspid and aortic valves (not shown). In wild-type and ctnnb1Dex3 hearts, no Mc was found in the pulmonary valves. All ctnnb1Dex3 Mc of the heart would have a different expression pattern compared with WT Mc, leading to molecular and cellular modifications. Besides the consequence on the DA, these modifications do not appear dramatic on ctnnb1Dex3 heart. The valves did not present major defects and the Doppler mitral inflow pattern was normal. However, mi mice do not have Mc in the heart due to the lack of Mitf. Mitf is not only expressed in Mc, it is also expressed in cardiomyocytes and is important to regulate cardiac growth and hypertrophy [35]. On C57BL/6 background, mi hearts are not dramatically affected. The conjunction of the lack of Mitf in the heart with the disappearance of an increased number of Mc may explain the different heart morphology and a potential earlier death of ctnnb1Dex3-mi mice, although this hypothesis remains elusive.
In order to confirm that b-catenin does not play a major role in the cardiac Mc lineage after its segregation from SMC2, Dct::Cre/ u; ctnnb1Dex3/+ mice were also produced, and found to be normally viable, with no evident enlargement of the left atrium when sacrificed at one-year old. In these mice, the activated bcatenin is produced specifically in Mb after specification. In another approach, Tyr::CreERt2/u mice [36] were crossed with ctnnb1Dex3/+ mice to produce a tamoxifen-inducible CreERt2 in Tyr-expressing cells. Tamoxifen was administered at E18.5, well after the segregation of the SMC2 and Mc lineages, and all double heterozygous mice were phenotypically normal: viable, again with no enlargement of the left atrium. We counted Dct/Trp2-positive cells, corresponding to Mc, in DA sections at P2 in both tamoxifen-and mock-induced, unrecombined Tyr::CreERt2/u; ctnnb1Dex3/+ mice. The numbers were very small in both cases, similar to that in WT mice (not shown). Repeating these experiments in both WT and ctnnb1Dex3 backgrounds using mi mice, as expected from the results in Figure 15F, failed to detect any Dct/Trp2-positive cells. The mice were viable and with no cardiac defect.
In conclusion, the DA of ctnnb1Dex3-mi and ctnnb1Dex3 mice did not fully close, demonstrating that the presence or absence of differentiated Mc was irrelevant to the onset of PDA. Rather, the absence of a significant proportion of SMC (SMC2 population) was associated in both cases with the failure of full DA closure and death; indomethacin injection reduced or prevented these manifestations (see Figure 13).

Discussion
We have demonstrated here that the increased b-catenin activity modifies the fate of a fraction of the smooth muscle cells (SMC2) of the DA into Mc, leading to postnatal PDA. We showed that the presence of numerous Mc in the DA is not the primary cause of the failure of the DA to close. Instead, the substantial reduction in number of SMC2 prevents the closure. ctnnb1Dex3 mice die within a few months, unlike other mutant mice presenting a PDA, which die in the first days after birth.

SMC2 and Mc have a common precursor
b-catenin can favor the specification of Mb, but it does not induce Mc proliferation [30,37]. However, b-catenin has been reported to induce cell proliferation of SMC [38,39]. If SMC2 and Mc had two different precursors, we would expect to observe more SMC2, rather than an increased proportion of Mc in our model system. Conversely, if SMC2 and Mc shared a common precursor, we would expect fewer (or no) SMC2 and an increase in the number of Mc. We observed this latter scenario. The production of more Mc seems to occur at the expense of SMC2 in ctnnb1Dex3 mice. Indeed, we showed that the number of SMC2 missing in the mutant DA was the same as the number of Mb appearing (see Figure 5). Moreover, the same disappearance of SMC2 was observed in mi mice, in which Mc disappear after specification (See Figure 15). These observations suggest that SMC2 and Mc have a common precursor.
However, we cannot formally exclude the existence of other mechanisms, and that the disappearance of SMC2 and presence of high numbers of Mc are not directly linked. For instance, invasion of the DA by Mb may induce the loss of the SMC2 population by an unknown mechanism.

Absence of SMC2 leads to PDA
The ctnnb1Dex3 mice display incomplete closure of the DA, an enlarged left atrium and die several weeks after birth. The death of these animals is certainly related to thrombus formation. Mc are found in the heart including the left atrium [22,24,25,26,40]. The Four hours later, embryos were isolated, fixed, X-gal stained, transversally sectioned through the DA and counterstained with eosin. We treated three pregnant females and sectioned ten embryonic hearts (five WT and five mutants). The ductus arteriosus was closed in all cases. Note that the numbers of Dct+ cells derived from ctnnb1Dex3-Dct embryos obtained from pregnant mothers injected with indomethacin or mock-injected were similar. (E) Kaplan-Meier curves of WT and ctnnb1Dex3 newborn pups treated or mock-treated with indomethacin (6 mg/kg body weight indomethacin within 12 hours of birth). Ultrasound analysis was performed on treated versus non-treated animals during the second and third months, which associated survival of treated ctnnb1Dex3 to the size of the left atrium (not shown). Indomethacin-treated ctnnb1Dex3 mice survived significantly longer than mock-treated mice (p,0.009). Note similar results were obtained when ctnnb1Dex3 mi mice were treated with indomethacin or mock. Scale bars, (A, B) = 100 mm, (C, D) = 50 mm. doi:10.1371/journal.pone.0053183.g013 enlargement of the left atrium may theoretically be an indirect consequence of an abnormality in the left atrium itself, due to the increased Wnt signaling in Mc, which may affect the expression of various genes and/or have haemodynamic effects. The likelihood of this hypothesis was small considering that ctnnb1Dex3-mi mice do not have Mc in the heart but present the same phenotype as the ctnnb1Dex3 mice. However, we cannot exclude the possibility that  There is no significant difference between survival of ctnnb1Dex3-mi and ctnnb1Dex3 populations. Macroscopic view of (B) mi and (C) ctnnb1Dex3mi hearts at P28. Note the enlargement of the ctnnb1Dex3-mi left atrium (cf. Figure 8). Transverse eosin-stained sections of (D) mi and (E) ctnnb1Dex3-mi DA at P2. Note that the ctnnb1Dex3-mi DA is not closed (cf. Figure 11). (F) Numbers of SMA-positive and LacZ-positive cells were evaluated after SMA, LacZ and DAPI staining in the DA of mi-Dct ( = Tyr::Cre/u; +/+; mi vga9 /mi vga9 ; Dct::LacZ/u), ctnnb1Dex3-Dct ( = Tyr::Cre/u; ctnnb1Dex3/+; +/+; Dct::LacZ/u) and ctnnb1Dex3-mi-Dct ( = Tyr::Cre/u; ctnnb1Dex3/+; mi vga9 /mi vga9 ; Dct::LacZ/u) mice at E18.5. For each genotype, the number of cells were estimated from 5-10 sections per embryos using 3 mice. mb = melanoblast. Scale bar (D, E) = 50 mm. *: p-value ,0.05, **: pvalue ,0.01, ns = non significant. doi:10.1371/journal.pone.0053183.g015 these mutant mice (with or without Mc) may present undetected weaknesses of the left atrium per se.
Nonetheless, to examine a possible effect from within the differentiated Mc lineage, we conducted some complementary experiments to assess these possibilities. For example, inducing the production of the activated b-catenin within Tyr-expressing cells, after the segregation of SMC2 from their common precursors with Mc (at E18.5), yielded a normal cardiac phenotype, as did the expression of the activated b-catenin in Mb specifically (Dctexpressing cells).
These approaches, together with our observations in mi mice, indicate that neither the induction of Wnt signaling in Tyror Mitfexpressing Mb, nor the abnormally large number of Dct/Trp2expressing Mb in the ctnnb1Dex3 DA, nor both, are sufficient to induce the cardiac phenotype observed. The main cause of the PDA is presumably associated with the absence of SMC2 from the DA. The SMC2 population therefore prevents, rather that the Mc population favors, the development of PDA.
An induction of the prostaglandin pathway by the activated beta-catenin is not the primary cause of PDA Fetal PDA is controlled by many factors, the most important of which are proportionally low fetal blood oxygen partial pressure [41] and cyclooxygenase-mediated products of arachidonic acid metabolism (primarily PGE2 and prostacyclin) [42]. Several mutant mouse lines present post-natal PDA as a result of targeting genes of the prostaglandin pathway, acting on contractility (Ptgs2, Pgdh and EP4).
At the molecular level, b-catenin had been shown to induce Ptgs2/Cox-2 expression [33]. Thus, it could have been argued that in our PDA model, the expression of b-catenin in the differentiated, more numerous Mc could lead to the PDA. However, we did not observe an increase in Ptgs2/Cox-2 expression level in the ctnnb1Dex3 DA, as shown in Figure 14. Moreover, this hypothesis is unlikely because the ctnnb1Dex3-mi DA does not contain cells producing the activated b-catenin, namely SMC2 and Mc, but still presents a PDA.
Relevance for human PDA ctnnb1Dex3 and ctnnb1Dex3-mi mice have a smaller number of SMC in the DA than WT mice, and present a dilated left atrium, with moderate left ventricular enlargement in about half of the mice. They suffer a high mortality rate, and this was due to thrombosis and left atrium rupture, rather than the left ventricular failure observed in the human clinical context. These characteristics are thus not the standard characteristics of human PDA. However, although indomethacin is an effective treatment for PDA due to prematurity in humans, it is less effective in term infants, as was also the case for ctnnb1Dex3 mice.
It had not previously been observed that a reduction of smooth muscle cell numbers was correlated with incomplete DA closure in humans. It could theoretically be possible to look for abnormally low numbers of SMC in the DA of term infants with indomethacin-unresponsive PDA, in particular. However, it is difficult (or indeed impossible) to determine objectively whether the numbers of SMC in DA from such patients differ from normal cardiac-healthy, age-matched individuals. There is substantial variability in the number of SMC in various human DA; furthermore there is no ethically appropriate control for determining reference values for normal human DA. To compound things, a very large cohort of normal and abnormal DA would be necessary to compensate for population variability, given the possible genetic differences and effects of the age of the patients. On the other hand, in C57BL/6 mice (a congenic mouse line) at a defined stage, the number of smooth muscle cells in the DA does not differ between individuals.
In human neonates displaying a PDA, left atrium dilation is only observed when the blood velocity of the shunt is high. In these conditions, the left-atrium-to-aortic-root ratio (LAARR) is greater than 1.7; most PDA patients have a LAARR of 1.5. In ctnnb1Dex3 mice, we found that the LAARR ratio was greater than 5 and in many cases around 8 (data not shown), indicating a very high output through the DA and/or a higher compliance of the atrial tissue than in humans. Most of the mouse strains with a high blood flow PDA die shortly after birth, such that in vivo evaluation is very difficult. By inactivating the p45 subunit of the transcription factor Nfe2, Echtler et al. [43] studied the contribution of platelets to DA closure in mice. In their homozygous mouse model, they report a lack of DA closure with high blood flow through the channel; they estimated that about 40% of the cardiac output passes through the PDA. These animals show typical features of pulmonary remodeling (in the network of collagen fibers along vessels) and an increase in right ventricular pressure (augmentation of right ventricular wall thickness). Nevertheless, macroscopic analysis of left ventricular sections did not reveal either dilation or hypertrophy. The absence of left ventricular dilation in cases of high flow PDA may be a particular aspect of mouse physiology.
Infants with PDA who are not sufficiently responsive to indomethacin treatment may have an unidentified molecular cause. However, the physiological read-out could be a previously unsuspected cellular defect.

SMC1 cannot naturally compensate for the lack of SMC2
The main cell components of the tunica media of the DA are the SMC1 and SMC2 populations, which are involved in its closure. In ctnnb1Dex3 mice, the number of SMC2 was reduced in favor of Mc, but although the number of SMC1 remained normal, this was not sufficient to allow the DA to close fully. It is surprising that the SMC1, although numerous, were unable to allow full closure through some sort of compensation mechanism, such as proliferation, in the absence of SMC2. Steric constraints were not imposed by the presence of Mc because, in Mc-deficient ctnnb1Dex3-mi mice, the number of SMC1 remained unaffected and the DA still did not fully close. This latter observation even suggested that the number of SMC1 cell divisions is controlled during development independently of the number of SMC2 cells.
The failure of DA closure associated with the absence of SMC2 could be due to the low total SMC count, or to SMC2 having an essential role in the development of the mature DA (despite SMC1 and SMC2 being histologically indistinguishable). Other differences between SMC1 and SMC2 have not been described. It would be valuable to characterize these two cell types at the molecular level to resolve this question. One possibility could be to look at MYH11 expression, which has been shown to be fully specific of SMC [44]. Potentially, the production/localization of MYH11 could be different in SMC1 and SMC2.
In principle, compensation could occur either at the cellular level, as indicated above, or at a molecular level. Thus, for SMC located in the DA, numerous proteins could contribute to molecular compensation. For instance, a major decrease of Cox-2 production in mutant SMC1 could allow the closure of the DA, although we found no evidence for such a reduction in Cox-2 levels. Other limited molecular compensations may have taken place in SMC1, which would not be sufficient to prevent PDA. SMC1 and SMC2 may also both be involved in the closure of the DA, but fulfilling different functions. This scenario is consistent with the partial closure of the DA observed in mutant mice where there is about 80% of the WT number of SMC (SMC1). It is nevertheless surprising that the absence of only 20% of the cells (SMC2) is fatal. Hyperstimulation of SMC1 with indomethacin in late gestation ctnnb1Dex3 embryos showed that SMC1 do retain the capacity to close the DA. Moreover, indomethacin treatment of ctnnb1Dex3 newborns with PDA prolonged their life span and even cured some of the animals. Ultrasound analysis revealed that in the cured animals, both the DA and the foramen ovale were closed, and the left atrium was of normal size. However, in the partially cured mice, although the DA was closed, the foramen ovale remained open, which correlates with their enlarged left atria. These differences could be related to the timing of injection/ closure of the DA. Another explanation could be related to the increased number of melanocytes in the ctnnb1Dex3 foramen ovale, which might interfere with its closure ( Figure 10E).

Function of melanocytes in the DA and in the heart
Melanocytes are consistently found in all heart valves except pulmonary valves [26]. Moreover, a few Mc are also consistently found in the DA. These observations argue that the presence of Mc in the heart is regulated rather than accidental. Nevertheless, the function of these cardiac Mc remains poorly understood, as these cells do not appear to be essential for general cardiac morphogenesis and physiology. Indeed, mice lacking Mc (e.g. Kit W/Wv or Mitf vga9/vga9 mice) or displaying strong pigmentation (e.g. Tyr::Nras Q61K ) in cardiac valves and septa are viable, with apparently normal hearts [26]. Moreover, we report here that the lack of either pigmented or unpigmented differentiatied melanocytes in the DA does not lead to an observable phenotype. Thus, cardiac Mc would most likely play a subtle role, which would become critical under stress conditions. Cardiac Mc could be involved in atrioventricular (AV) valves development from the endocardial cushions. Valve development requires remodeling of the extra-cellular matrix [45] and it has been shown that classic Mc produce and secrete metalloproteases [46]. This function of the Mc in the AV valves could continue into adulthood. Indeed, Mc and interstitial cells could be involved in tissue homeostasis and could affect the mechanic properties of the AV valves, as it has been shown that pigmentation stiffens murine tricuspid valve leaflet [47].
Mc of the atrium and pulmonary veins have been shown to contribute to atrial arrhythmia triggers [40]. Deregulation of intracellular calcium and reactive oxygen species (ROS) levels have been described in patients with atrial fibrillation. Dct participates to the regulation of these levels in Mc and mice lacking cardiac Mc do not develop arrhythmias. As we know that skin Mc have a buffer role for various UV-induced physical and chemical stresses, we could hypothesise that cardiac Mc could play a buffer role for calcium and ROS-induced stresses.
The function of the Mc of the DA remains even more elusive, since we showed here that they do not play a role per se in the closure of the DA. Their persistence in the LigA after birth suggests that they could maybe play a role in the homeostasis of the region. In zebrafish, numerous Mc have been described surrounding the aorta in the adult, whereas they were absent during development, suggesting a later role [48]. To sum up, it is still not possible to draw conclusions about the physiological role, if any, of Mc in the DA in a healthy and non-stressful environment. However, another descendant from the same precursor, the VNCC-derived SMC2 population, is irreplaceable in preventing PDA by promoting closure of the DA at birth.

Mice
Crosses are summarized in Table S1. Tyr::Cre, Tyr::CreERt2, Dct::Cre, b-catenin ctnnb1Dex3 (also known as f3), b-catenin ctnnb1Dex2-Dex6 (also known as f2-6), Dct::LacZ, Rosa26R, and Mitf vga9 mice were used in this study [27,31,36,49,50,51,52,53,54]. Tyr::Cre mice produce the enzyme Cre, which recognizes and intramolecularly recombines LoxP sequences in cells expressing the tyrosinase promoter at any point throughout life. In Dct::LacZ reporter mice, the transgene LacZ, producing the enzyme bgalactosidase, is under the control of the dopachrome tautomerase (Dct) promoter. Rosa26R floxed reporter mice activate the LacZ transgene to produce b-galactosidase only within any recombined cells and all of their descendants. This work was carried out in accordance with the Policies of the French Committee of Ethics. Mice were maintained in the SPF mouse colony of the Institut Curie according to French and EU law and is fully accredited by the French Direction of Veterinary Services (C 91 471 108, february 16th, 2011). Animal surgery and experimentation are authorized by the French Direction of Veterinary Services to LL C 91-642, July 20th, 2012) and BF (# 31-205, 2011).

Mouse embryo and tissue preparation, X-gal staining and immunohistochemistry
Mouse embryos and tissues were collected and rinsed in cold PBS and fixed by incubation for 15-20 minutes in PBS containing 0.25% glutaraldehyde at 4uC. They were rinsed twice in cold PBS and incubated for 16 hours in 30% sucrose/PBS at 4uC, then for 5 hours in sucrose (30%)/OCT(50%)/PBS at 4uC, embedded in OCT compound (TissueTek, 4583, Sakura Finetechnical Co. Ltd. Tokyo, Japan) and sectioned. PBS containing 0.25% glutaraldehyde was injected into the left ventricles of adult mice after anesthesia. Hearts were then removed and treated as described above. For X-gal staining, 8 mm-thick sections were treated as previously described [36]. For immunofluorescence, 8 mm-thick sections were treated as previously described [55]. Whole heart samples placed on agarose gels were observed from the ventral side. All specimens of DA and LigA were sectioned vertically for X-gal staining and immunofluorescence analysis. Antibodies directed against beta-galactosidase and alpha-smooth muscle actin (SMA) were purchased from Abcam (ab9361) and Sigma (A5228), respectively.
Determination of the intimal cushion (ic) and lumen (l) areas DA/LigA were sectioned and stained with eosin. The areas of the intimal cushion and the lumen were determined using ImageJ (http://rsbweb.nih.gov/ij/) version 1.37.

Echocardiography and micro-CT
For routine examination and color Doppler analysis, transthoracic echocardiograms were performed in a Sonos 5500 on mice anesthetized with 2% isoflurane as described previously (Philips Ultrasound, USA) [56]. To visualize the DA and foramen ovale, high-resolution imaging was performed with a 30 MHz probe on a Vevo 660 (Visualsonics, Canada) with the same anesthetic regimen. Due to the rightward shift of the heart of ctnnb1Dex3mice, non-conventional incidences were employed. Mice were processed as previously described for micro-CT [57].

Indomethacin treatment
An indomethacin (Sigma) stock solution was prepared as a 10 mg/ml solution in ethanol and stored at 220uC. Working solutions were prepared extemporaneously as 500 ng/ml solutions in PBS. E18.5 pregnant females were injected intraperitoneally with 10 mg/kg body weight indomethacin and the embryos harvested 4 hours later. Newborn mice were injected subcutaneously with 6 or 12 mg/kg body weight indomethacin within 12 hours of birth, using a microsyringe (Fisher-Bioblock, France W23305). No difference in long-term survival was observed between animals treated with the two doses of indomethacin.

(AVI)
Movie S2 Reconstructed truncal regions from CT scan pictures of ctnnb1Dex3 (Tyr::Cre/6; ctnnb1Dex3/+) at the truncal level at P28. Note in Tyr::Cre/u; ctnnb1Dex3/+; mice, the left atrium is so enlarged that it affects the angular orientation of the heart by displacement. (AVI)