Identification of cardiac malformations in mice lacking Ptdsr using a novel high-throughput magnetic resonance imaging technique

Background Congenital heart defects are the leading non-infectious cause of death in children. Genetic studies in the mouse have been crucial to uncover new genes and signaling pathways associated with heart development and congenital heart disease. The identification of murine models of congenital cardiac malformations in high-throughput mutagenesis screens and in gene-targeted models is hindered by the opacity of the mouse embryo. Results We developed and optimized a novel method for high-throughput multi-embryo magnetic resonance imaging (MRI). Using this approach we identified cardiac malformations in phosphatidylserine receptor (Ptdsr) deficient embryos. These included ventricular septal defects, double-outlet right ventricle, and hypoplasia of the pulmonary artery and thymus. These results indicate that Ptdsr plays a key role in cardiac development. Conclusions Our novel multi-embryo MRI technique enables high-throughput identification of murine models for human congenital cardiopulmonary malformations at high spatial resolution. The technique can be easily adapted for mouse mutagenesis screens and, thus provides an important new tool for identifying new mouse models for human congenital heart diseases.


Background
Congenital malformations are a major cause of death in childhood, and are typically characterized by lesions that do not compromise fetal survival. For instance, congenital heart disease (CHD) typically consists of lesions such as ventricular and atrial septal defects, which are compatible with fetal hemodynamics [1]. Although human genetic studies have identified some genes that cause congenital cardiac malformations, the molecular and developmental mechanisms underlying most of these defects remain largely unknown. Despite the high incidence of CHD (~1% of live births), only a handful of genes have been identified that when mutated, result in congenital heart disease [2][3][4].
The mouse is a particularly good model for studying mechanisms of cardiac diseases as its anatomy and development resembles that of the human more closely than any other genetically tractable organism. Importantly, the mouse is amenable to genotype-driven approaches such as transgenic knockouts of defined candidate genes [5], genome wide mutagenesis approaches using gene-trap [6] or transposon insertion screens [7], and high-throughput phenotype-driven screens that rely, for instance, on Nethyl-N-nitrosourea (ENU) mutagenesis [8,9]. However, high-throughput cardiovascular genomic approaches in the mouse have been hampered by the paucity of phenotyping tools that allow efficient identification of complex cardiac malformations. As mouse embryos are opaque, late developmental defects are particularly difficult to identify. For instance, cardiac septal defects or outflow tract abnormalities can only be confidently identified after 14.5 days post coitum (dpc) when cardiac and outflow tract septation are completed in normal embryos. The identification of malformations in late gestation embryos typically relies on serial histological sectioning, which is extremely labor intensive. Furthermore, this often results in the irretrievable loss of 3D information, which is essential for the interpretation of complex cardiac malformations. In addition, standard pathological analysis is not amenable to high-throughput phenotype screening protocols that are required for any mutagenesis screen aiming at the functional dissection of the developmental biology of cardiac diseases. Therefore, new technological approaches must be harnessed that allow an efficient phenotyping of heart defects and also of subtle cardiac abnormalities that are at danger of being overseen in traditional histopathology screens. This is even more important in the light of upcoming new endeavors in functional mouse genome annotation [10]. Currently, new large-scale mouse mutagenesis screens are being set up in the US and in Europe that aim to produce heritable mutations in every gene in the mouse genome [11][12][13]. To make full use of these new mouse mutant resources more precise and efficient phenotyping methods are urgently needed [12][13][14]. The success of genome wide saturation mutagenesis screens depends therefore on improved phenotyping, and new high-resolution imaging approaches for mouse mutants are one of the most important which need to be established.
We previously reported the development of fast gradientecho MRI of single mouse embryos [15][16][17]. This resulted in the acquisition of a 3D dataset in under 9 hours, with an experimental image resolution of 25 × 25 × 26 µm/ voxel. We showed that MRI is capable of accurately identifying normal embryonal structures, and cardiac and adrenal malformations in knockout mouse embryos, and we have validated this technique by performing in depth histological examinations of imaged embryos [15][16][17]. These experiments showed that single-embryo MRI could correctly identify all cardiac lesions (atrial septal defects, ventricular septal defects, outflow tract defects such as double-outlet right ventricle, and aortic arch defects) except those under 20 µm -which is below the resolution of the MRI technique. As this method images single embryos in overnight runs, it still lacks the throughput required for phenotype-driven mutagenesis screens. For instance, in a typical recessive ENU mutagenesis screen, to screen 50 ENU mutant lines using a 3-generation breeding scheme would require the analysis of ~1200 embryos [18]. We now report the development of a method of imaging up to 32 embryos simultaneously in a single unattended overnight run, at high spatial resolution. Allowing ~30 minutes per embryo, the analysis of 1200 embryos would take 75 working days for a single trained individual. We show that this high-throughput multiembryo MRI technique can be used to rapidly identify unsuspected embryonal cardiac and visceral malformations. Using this technique we could identify a novel, and hitherto unsuspected role for the phosphatidylserine receptor (Ptdsr) in controlling ventricular septal, outflow tract and pulmonary artery development. In addition, we found thymus hypoplasia in Ptdsr-deficient embryos. These findings suggest that a novel Ptdsr-mediated pathway is required for cardiac and thymus development.

Multi-embryo imaging
We modified our previously described fast gradient echo magnetic resonance imaging technique [15][16][17] to image embryos embedded in four to eight layers (16 -32 embryos total) in 28 mm nuclear magnetic resonance tubes using a single quadrature driven birdcage coil (Figure 1a). Preparation and embedding of embryos typically took less than an hour. In initial experiments we imaged up to 16 embryos simultaneously in overnight runs of <9 hours, with an experimental resolution of 51 × 51 × 39 µm. Subsequently, we used a custom made optimized probe with an increased sensitivity range to enhance imaging throughput. This allowed us to image 32 embryos simultaneously (Figure 1a,b), but with a larger matrix size and an increased field-of-view in the long axis of the tube. For these experiments we imaged embryos for 12 hours, and achieved an improved experimental resolution of 43 × 43 × 36 µm. The optimized coil used for 32 embryos MRI ( Figure 1) has a sensitivity range in z-direction of close to 50 mm. The artefacts seen at both ends in the longitudinal image ( Figure 1b The data are permanently archived on DVDs, for subsequent analysis. Construction of 3D reconstructions of the heart typically takes ~4 hours, but was not necessary for the identification of cardiovascular defects.

Cardiac malformations in mice lacking Ptdsr
We next evaluated the role of multi-embryo MRI in analyzing unexplained lethality in embryos generated in collaborating laboratories. Recently, we have generated mice lacking the phosphatidylserine receptor (Ptdsr -/-) on a C57BL/ 6J background, by gene targeting in embryonic stem cells [21]. Ptdsr is a nuclear protein of unknown function, which is essential for the development and differentiation of multiple organs during embryogenesis [21][22][23][24].
Identification of septal, outflow tract, and aortic arch malformations using multi-embryo MRI Figure 2 Identification of septal, outflow tract, and aortic arch malformations using multi-embryo MRI (a -e') Images of transverse sections from 5 Cited2 -/embryos obtained using the multi-embryo technique (a-e) compared with images from the same embryos obtained subsequently using the single embryo technique (a'-e'). (a, a') Section showing left and right atria and ventricles (la, ram, live, rave). The atria are separated by the primary atria septum (pas), which is deficient at its ventral margin creating an osmium premium type of atria septal defect (ASD-P). (b, b') Section showing a ventricular septal defect (VSD) in the interventricular septum (ivs). (c, c') Section showing double outlet right ventricle, wherein the ascending aorta (a-ao) and the pulmonary artery (pa) both arise from the right ventricle (rv). The aortic valve (ao-v) is indicated. (d, d') Section showing a right-sided aortic arch (ao-a) passing to the right of the trachea (tr) and the esophagus (es). (e, e') Section showing bilateral aortic arches (ao-a) forming a vascular ring around the trachea (tr) and the esophagus (es). Also indicated are the thymus (th) and the right superior vena cava (r-svc). (f -j) Serial transverse sections through a wild-type heart obtained using single embryo MRI, demonstrating corresponding normal structures, including the systemic venous sinus (svs), left superior vena cava (l-svc), pulmonary vein (pvn), descending aorta (d-ao), mitral and tricuspid valves (mv, tv), the secondary atrial septum (sas), left and right ventricular outflow tracts (lvot, rvot), pulmonary valve (pv), and arterial duct (ad) of the pulmonary artery. Scale bars = 635 µm for multi-embryo, and 317 µm for single embryo images; axes: d -dorsal; v -ventral; r -right; l -left.
Ablation of Ptdsr function in knockout mice causes perinatal lethality, growth retardation [21,22,24] and a delay in terminal differentiation of the kidney, intestine, liver and lungs during embryogenesis [21]. In addition, Ptdsr -/embryos develop complex ocular lesions [21] as well as haematopoietic defects [24]. However, as many malformations have been described in Ptdsr mutants, none of those detected could explain the observed perinatal lethality of Ptdsr -/mice. In the process of phenotypical characterization of our Ptdsr-deficient mouse line, we frequently observed subcutaneous edema of varying sizes in Ptdsr -/embryos by gross inspection ( [21] and Figure 4). As the development of edema in various mouse mutants is frequently associated with cardiovascular defects [25] we started to investigate if this also holds true for Ptdsr-deficient mice. We examined 8 embryos lacking Ptdsr, and 8 littermate wild-type or heterozygous controls using multiembryo MRI. We found that 5 of 8 Ptdsr -/embryos had cardiac malformations, which included ventricular septal defects, double outlet right ventricle, and pulmonary artery hypoplasia ( Figure 5). None of the wild-type or Ptdsr +/embryos had cardiac malformations. These findings were confirmed on single embryo imaging ( Figure 6). Furthermore, to verify the identified cardiac defects in the Ptdsr -/mice we performed serial transverse sectioning of all analyzed embryos. In all cases, we could recognize again the same heart defects that were identified before using the multi-embryo MRI technique (Figure 7). In addition, we analyzed by multi-embryo MRI a second Ptdsr-knockout mouse line (Ptdsr tm1.1 Gbf ), which is identical to the initially analyzed Ptdsr mutant except that the loxP-flanked neomycin selection cassette was removed by breeding the original Ptdsr tm1 Gbf knockout line [21] to a CMV-Cre deleter mouse line [26]. From this Ptdsr tm1.1 Gbf knockout mouse line we analyzed 8 Ptdsr -/embryos, and as littermate controls, 3 Ptdsr +/+ and 1 Ptdsr +/embryos. We found ventricular septal defects in five out of the eight Ptdsr -/embryos (data not shown). Again we found no evidence for cardiac malformations in wild-type or heterozygous littermate control embryos.
We also observed a modest degree of thymic hypoplasia in Ptdsr -/embryos (Figure 6k). To confirm this, thymus and embryo volumes were measured in 8 Ptdsr -/embryos and 8 wild-type or heterozygous control embryos at 15.5 dpc. Embryo volume measured from the MRI datasets correlated very strongly with embryo weight (Figure 6l), and was modestly reduced in Ptdsr -/embryos (Figure 6m). The volume of the thymus was significantly reduced in Ptdsr -/embryos, even after correction for embryo volume, to 58% of the control value (Figure 6n,o). To correlate the identified cardiopulmonary malformations in Ptdsr -/embryos with expression of the Ptdsr gene during heart development, we made use of our Ptdsr gene-trap reporter mouse line [21]. Using X-Gal staining in heterozygous embryos staged between 9.5 dpc and 12.5 dpc we found specific Ptdsr expression in the heart starting at 10.5 dpc and getting more defined to the compact zone and the trabeculi from 11.5 dpc onwards ( Figure 8). Furthermore, when we analyzed Ptdsr -/hearts by histopathology at 16.5 dpc we observed a severe differentiation defect in the compact zone as well as in the trabeculi (Figure 9), thus demonstrating that Ptdsr is in addition required for heart muscle differentiation at later stages of development. Taken together these results indicate that Ptdsr plays a hitherto unsuspected role in cardiovascular development as well as in cardiac muscle differentiation.

Utility of MRI in identifying mouse models of human malformations
Our results show that it is possible to efficiently identify and quantitate relatively subtle cardiac and visceral malformations in late gestation mouse embryos using multi- embryo MRI. We have deliberately optimized our technique at late gestation in order to identify those congenital defects that allow survival through most of gestation. Importantly, these defects resemble human congenital malformations, and would provide mouse models for the study of these diseases. Our method represents a simpler alternative to the multi-coil approach published recently [27,28] in which up to eight fixed mouse embryos were imaged simultaneously, with a resolution of 200 µm. In comparison, the multi-embryo method described here has a higher experimental resolution of 43 µm. Notably, it requires substantial developmental and financial effort to equip an experimental MR-system with multiple coil and receive capability.

Role of MRI in investigating murine embryonal or perinatal lethality
Many mouse gene knockouts display late gestational lethality, but incomplete analysis and loss of 3D information consequent to histological sectioning, results in major developmental malformations being frequently missed. As shown here, Ptdsr -/embryos on a C57BL/6J background develop heart defects. This was not noted in recent reports [22,24], and emphasizes the need not only for completely examining several mutant embryos, but also repeating these examinations in different genetic backgrounds. A major advantage of MRI is the ability to easily ship fixed embryos from referring laboratories to the laboratory that performs the MRI analysis. This minimizes the expense of animal relocation, re-derivation, and breeding required to generate the embryos, and significantly reduces animal experimentation.

Role of MRI in high-throughput phenotype driven screens
MRI screens performed at late gestation would be expected to identify genes that affect later aspects of development, and identify hypomorphic and haploinsufficient alleles of genes that affect earlier steps of development. Published data from genome-wide ENU mutagenesis screens in the mouse indicate that ~30% progeny carry a heritable recessive phenotype, making 3-generation recessive screens the method of choice for identifying developmental malformations [18]. At least 24 3 rd generation progeny per 1 st generation mutant are typically screened, resulting in a >78% probability of identifying at least one fully penetrant recessive homozygous mutant. A typical recessive screen (e.g. 50 -100 first generation mutants per year) would require the analysis of ~1200 -2400 embryos per year. Our results show that multi-embryo MRI is eminently suitable for such throughput. Although, screening of 32 embryos overnight requires typically about 16 hours of data analysis, multi-embryo MRI mutagenesis screens can be easily performed at a reasonable scale if multiple operators analyze the data in parallel. As the data are Identification of cardiac malformations in Ptdsr -/embryos using multi-embryo MRI permanently stored on DVDs, and can be analyzed easily by commercially available software, they can be without difficulty disseminated to specialists for further and more detailed analysis. Another powerful application of multiembryo MRI will likely be the investigation and screening of potentially teratogenic drugs.

Functional role of Ptdsr in heart development
Our results presented here indicate a new, and hitherto unsuspected role for the phosphatidylserine receptor in controlling ventricular septal, outflow tract, pulmonary artery, and thymus development. This finding suggests that a novel Ptdsr-mediated pathway is required for Cardiac malformations and thymus hypoplasia in Ptdsr -/embryos Figure 6 Cardiac malformations and thymus hypoplasia in Ptdsr -/embryos. (a-c) Transverse and oblique (through the plane of the ascending aorta) sections, and 3D reconstruction (left-ventral oblique view) of a heart of a wild-type embryo at 15.5 dpc. The left and right ventricles (lv, rv) are separated by the interventricular septum (ivs). The left and right atria (la, ra), and the trachea (tr) are also indicated. The ascending aorta (a-ao) arises from the left ventricular outflow tract (lvot), via the aortic valve (ao-v), and continues on as the aortic arch (ao-a), which joins the descending aorta (d-ao). The pulmonary artery (pa) arises from the right ventricular outflow tract (rvot), and continues as the arterial duct (ad), which joins the descending aorta.
(d-f) Corresponding images of a Ptdsr -/embryo, showing a smaller heart with a ventricular septal defect (VSD). The aorta arises from the right ventricle. The pulmonary artery is small and its connection to the descending aorta (arterial duct) could not be identified. (g-i) Corresponding images of another Ptdsr -/embryo, showing a ventricular septal defect (VSD). The aorta overrides the VSD resulting in a double-outlet right ventricle. (j, k) Coronal sections of Ptdsr +/+ and Ptdsr -/embryos, showing the two lobes of the thymus (th). The arterial duct of the pulmonary artery in the Ptdsr -/embryo is narrowed. (l) Correlation between embryo weight and volume. Scattergram of embryo weight versus embryo volume measured from multi-embryo MRI datasets for 16 embryos using Amira. The co-efficient of regression (r) is indicated. (m, n,) Absolute embryo and thymus volumes (µl) were measured from the MRI datasets from 5 wild-type (wt), 3 heterozygote (h), and 8 Ptdsr -/-(m) embryos at 15.5 dpc. There was no significant difference in the wild-type and heterozygote data, which were therefore pooled together (wt/h). cardiac and thymus development. Recently, we have demonstrated that in contrast to previously reported hypothetical Ptdsr functions, the Ptdsr protein is not required for the clearance of apoptotic cells [21]. Moreover, detailed analysis of apoptosis induction and apoptotic cell clearance in Ptdsr +/+ and Ptdsr -/embryos during heart development did not reveal any difference in the number and location of apoptotic cells between the genotypes (J.B., A.D.G. and A.L. unpublished observations). This further excludes that Ptdsr has any function in apoptotic cell clearance and points to other developmental mechanisms that are affected by Ptdsr ablation. The neural crest plays an important role in the development of the cardiac outflow tract, aortic arches, and the thymus [29]. As Ptdsrdeficient embryos lack intestinal ganglia [21] which are also derived from the neural crest, these results suggest that Ptdsr -/mice may have an underlying neural crest defect. Importantly, dysfunction of these Ptdsr-mediated pathways during development could also potentially result in heart defects in humans.

Conclusions
Our results validate the utility of multi-embryo MRI for high-throughput identification of murine models for human congenital cardiac malformations, and using this technique we have shown that Ptdsr is essential for normal cardiac development. Further experiments are needed to define exactly in which pathways Ptdsr is involved during heart development. We expect that multi-embryo MRI will be an important technology for future phenotypedriven mouse mutagenesis screens. The technology can be easily implemented at standard MRI imaging centers, thus allowing by collaboration with individual researchers or mouse mutagenesis centers, a high-throughput functional genetic dissection of mechanisms underlying cardiac development and congenital heart diseases.

Embryo preparation
Embryos were fixed in 4% paraformaldehyde at 4°C for 1 week, and then embedded in 1% agarose (Seakem) containing 2 mM gadolinium-diethylenetriamine pentaacetic anhydride (Gd-DTPA, Magnevist, Schering UK) in 28 mm nuclear magnetic resonance tubes (Figure 1). The left forelimb was removed from each embryo to facilitate the identification of the left side. In addition, embryos had other limbs and/or tails removed before embedding so that each embryo in a given layer (of four embryos) could be unequivocally identified.

Magnetic resonance imaging
Single embryo imaging was performed as described previously [15][16][17]. For multi-embryo imaging, we used the same 11.7 Tesla (500 MHz) vertical magnet (Magnex Scientific, Oxon, UK). This was interfaced to a Bruker Avance console (Bruker Medical, Ettlingen, Germany) equipped with a shielded gradient system with a maximal gradient strength of 548 mTesla/m (Magnex Scientific, Oxon, UK), and quadrature-driven birdcage type coils with an inner diameter of 28 mm (Rapid Biomedical, Würzburg, Germany). Compressed air at room temperature was used to reduce the heating induced by the gradients. A 3D spoiled gradient echo sequence (echo time 10 ms), a π/2 excitation pulse with rectangular pulse shape, (π/2 = 100 µs), was used with a short repetition time (30 ms) to obtain strong T 1 contrast. A matrix size of 512 × 512 × 768 (bandwidth: 130 Hz/pixel) at a field of view of 26 × 26 × 30 mm achieved an experimental resolution of 51 × 51 × 39 µm when imaging up to 16 specimens. In case of 32 embryos, a matrix size of 608 × 608 × 1408 at field of view of 26 × 26 × 50 mm, yielded an experimental resolution of 43 × 43 × 36 µm. The total experimental time was ~8.75 hours for 16 embryos, and ~12.3 hours for 32 embryos (typically overnight runs) whereby each phase encoding step was averaged four times.

Data reconstruction and analysis
The raw MR data were reconstructed into a stack of 1024 (for 16 embryos), or 2048 (for 32 embryos) 2D TIFF files (16 bit pixel resolution, 2 or 4 GB total size) using purpose-written software as described previously [16]. The TIFF files were analyzed using Amira 3.1(TGS Europe, Mérignac Cedex, France). 3D reconstructions were performed using the Image Segmentation Editor, and tis-Analysis of Ptdsr expression in the embryonic heart Figure 8 Analysis of Ptdsr expression in the embryonic heart (a, b) Staining of heterozygous Ptdsr-βgeo-embryos [21] using X-Gal at 10.5 dpc (a) and 11.5 dpc (b). (a) At 10.5 dpc Ptdsr expression can be seen throughout the heart. (b) Transverse sections of X-Gal stained embryos at 11.5 dpc showed an increased expression of Ptdsr in the myocardial wall and a beginning decrease of the expression in the trabeculation. Scale bar = 100 µm in (b).
sue volumes for morphometric analysis were measured using the Measure Tissue Statistics tool available in Amira 3.1. The probability (p) of a Type I error was calculated using a 2-sample equal variance 2-tailed t-test in Microsoft Excel.

Histology
Embryos were dehydrated in ethanol, embedded in paraffin wax, and sections were stained with hematoxylin and eosin.

X-Gal staining of embryos
Embryos were dissected free of extraembryonic membranes and then fixed in 4% paraformaldehyde at 4°C. Expression of Ptdsr was detected by staining the embryos overnight in X-Gal according to standard protocols. The embryos were postfixed in 4% paraformaldehyde and processed for documentation or histology.