Conditional Creation and Rescue of Nipbl-Deficiency in Mice Reveals Multiple Determinants of Risk for Congenital Heart Defects

Elucidating the causes of congenital heart defects is made difficult by the complex morphogenesis of the mammalian heart, which takes place early in development, involves contributions from multiple germ layers, and is controlled by many genes. Here, we use a conditional/invertible genetic strategy to identify the cell lineage(s) responsible for the development of heart defects in a Nipbl-deficient mouse model of Cornelia de Lange Syndrome, in which global yet subtle transcriptional dysregulation leads to development of atrial septal defects (ASDs) at high frequency. Using an approach that allows for recombinase-mediated creation or rescue of Nipbl deficiency in different lineages, we uncover complex interactions between the cardiac mesoderm, endoderm, and the rest of the embryo, whereby the risk conferred by genetic abnormality in any one lineage is modified, in a surprisingly non-additive way, by the status of others. We argue that these results are best understood in the context of a model in which the risk of heart defects is associated with the adequacy of early progenitor cell populations relative to the sizes of the structures they must eventually form.


Introduction
Congenital heart defects (CHDs) are the most common of human birth defects, and a leading cause of perinatal morbidity and mortality [1]. The genetics of CHDs are complex, with only a fraction being associated with chromosomal abnormalities or Mendelian developmental syndromes [2][3][4]. As with most human birth defects, the majority of CHDs are non-syndromic (isolated), and appear to be multifactorial and most likely polygenic [5][6][7][8].
Cornelia de Lange Syndrome (CdLS) is a multisystem birth defects disorder characterized by craniofacial abnormalities, developmental delay in growth and maturation, neurological deficits and intellectual disability, limb abnormalities (particularly of the arms and hands), as well as defects in the visual, auditory, gastrointestinal, genitourinary, and cardiopulmonary systems [9][10][11][12][13]. CHDs are seen in about 30% of individuals with CdLS [14,15], with structural defects of the ventricular septum (VSD) and atrial septum (ASD) being among the most common.
The genetic cause of CdLS is, in most cases, haploinsufficiency for NIPBL (Nipped-B-homologue), which encodes a ubiquitous protein that regulates the loading of cohesin onto chromosomes [16,17]. Cohesin, a multiprotein complex originally identified by its role in sister chromatid cohesion, is now understood to play critical roles in the regulation of gene expression [18][19][20]. In support of this view, both CdLS patient cell lines and animal models of Nipbl haploinsufficiency display small but significant changes in the expression of up to 1,000 or more genes, in essentially every tissue [21][22][23][24]. Studies in mouse and zebrafish models suggest that these gene expression changes-most too small to have phenotypic consequences individually-collectively cause the structural and functional defects observed in CdLS [23,[25][26][27]. CdLS and related "cohesinopathies" [26] thus exemplify an emerging class of genetic disorders -recently termed "transcriptomopathies" [27]-in which additive or synergistic effects of quantitative variation in the levels of multiple gene products lie at the root of developmental abnormalities. Such disorders provide a unique window into the kinds of multifactorial interactions that likely underlie polygenic traits, including the majority of complex developmental disease.
Among features of CdLS that are phenocopied in Nipbl-deficient animal models are CHDs [22,23]. In particular, mouse models display a high frequency of ASD, which are both common in CdLS and among the most common CHDs in non-syndromic settings [28,29]. Elucidating the processes that lead to CHDs in such mouse models could potentially provide valuable insight into the multifactorial causation of birth defects. Yet the fact that gene expression is globally affected in such animals also makes it difficult to know a priori in which cell types, or at what developmental stages, to look for such processes. For heart defects, this is an especially challenging issue, as development of the heart involves communication and interaction among many cell types, in multiple tissues, across many stages of development [30][31][32].
Here, we address this problem by exploiting a new allelic series (Nipbl FLEX ), in which cells, and mice, can be toggled between wildtype and mutant conformations of Nipbl using Flip-Excision (FLEX) technology [33]. We use this strategy both to create and rescue Nipbl-deficiency in multiple lineages during early heart development, including the cardiogenic mesoderm, endoderm, and neural crest. Interestingly, the data do not support the assignment of CHD risk to a single cardiac lineage, but rather suggest that the cardiogenic mesoderm, the endoderm, and other lineages all participate, interacting in surprisingly non-additive, and even antagonistic, ways. We relate these findings to morphological and gene expression changes that occur in early Nipbl-deficient hearts, and propose a model in which defects arise when the ability to generate cardiac tissue during very early stages fails to keep up with the demands imposed by final heart size.

Nipbl-Deficient Mice Display Heart Abnormalities throughout Embryonic Development
In a previous study of a CdLS mouse model of Nipbl haploinsufficiency (Nipbl +/mice), we reported that about half of Nipbl +/mice display defects in closure of the atrial septum (ASD) between gestational days 15.5 and 17.5 (E15.5-E17.5) [22]. Because many of the nearly 100 genes that have been linked to the production of cardiac septal defects [34,35] influence heart development through actions at early embryonic stages, long before cardiac septa form, we also examined heart development in Nipbl +/mice at successively earlier developmental stages to determine whether earlier structural abnormalities could be found. Interestingly, even though E17.5 Nipbl +/mice do not usually display defects of the ventricular septum, we found evidence for earlier abnormalities in ventricular septation. In particular, at E13.5, 77% of Nipbl +/hearts showed incomplete fusion or complete lack of contact of the developing ventricular septum with the cardiac cushion, compared to 14% of wildtype hearts ( Fig 1A).
Studies of Nipbl-deficient tissues and cells from Drosophila, mouse, zebrafish, and man all indicate that the means by which Nipbl-deficiency causes birth defects is by subtly mis-regulating the expression of large numbers of genes [21][22][23][24][25]. Accordingly, we screened Nipbl +/hearts for abnormalities in the expression of genes known to be involved in cardiac septation. Using quantitative reverse transcription PCR (qRT-PCR) as a rapid screen, we assessed expression of 29 such genes (S1 Table) from E10-10.5, when all four heart chambers have formed and septation has just begun; through E13.5, when ventricular septation is mostly complete [30].
Hand1 and Pitx2 are genes that both play a role in left-right asymmetry. Pitx2 encodes a homeobox transcription factor critical for left-right patterning of the entire body, and is preferentially expressed in left-sided cardiac structures from as early as cardiac crescent stage [38], and Pitx2 mutations have been associated with cardiac defects, including septal defects [39][40][41][42]. Hand1 encodes a basic helix-loop-helix transcription factor that is preferentially expressed in the left ventricle, and is important for its development [43][44][45].
Because both Pitx2 and Hand1 are associated with left heart structures, we wondered whether their elevated levels in E10-10.5 Nipbl +/hearts was due to a change in gene regulation, or a change in the proportions of left-sided versus right-sided tissue. Close inspection of hearts in which Hand1 was visualized by in situ hybridization (ISH; Fig 1C and 1D), as well as hearts imaged by optical projection tomography (OPT; Fig 1E), supports the latter explanation. As shown in Fig 1C, E10.5 Nipbl +/hearts display a relatively normal pattern and intensity of Hand1 expression (by ISH), but the right ventricle is abnormally small. This could be demonstrated consistently across embryos ( Fig 1D) and was also evident in optical cross-sections through the heart (Fig 1E).
These results suggest that the origins of heart defects in Nipbl +/mice lie earlier in development, in events that determine the relative amounts of left and right tissue that contribute to the heart. Among the most important genes upstream of Hand1 and Pitx2 is Nkx2-5 [46,47], which regulates many heart development genes [43]. Nkx2-5 can be detected as early as E7.5 in the cardiac crescent, the collection of cardiogenic mesodermal cells that later coalesce to form the early heart tube [48]. We used ISH to examine Nkx2-5 expression at this stage ( Fig 1F) and found that while the morphology of the cardiac crescent appears normal in Nipbl +/embryos, the strength of hybridization signal for Nkx2-5 is noticeably lower than in wildtype littermates, suggesting that either fewer cardiac progenitor cells are present, or these cells express Nkx2-5 at a reduced level.
The earliest-known marker of cardiogenic mesoderm is Mesp1, which acts upstream of Nkx2-5 to drive initial specification of cardiac stem/progenitor cells [49][50][51][52]. Mesp1 first appears in the lateral plate mesoderm at the primitive streak stage, about E6.5. By ISH, we observe a dramatic reduction in Mesp1 in Nipbl +/embryos at E6.5 ( Fig 1G) which we suspect, given the critical role of Mesp1 in cardiac development, is probably due to delayed onset of Mesp1 expression, rather than a loss of the capacity to express Mesp1.
Overall, these data indicate that heart development is abnormal in Nipbl +/embryos from the earliest developmental times at which cardiogenic tissue is formed. These abnormalities are reflected in reduced expression of key transcription factors at appropriate stages, and delayed growth and development of right-ventricular structures and the interventricular septum.

Generation of a Conditional/Invertible Nipbl Allelic Series Permits Tissue-Specific Creation and Rescue of Nipbl Expression Deficiency
Elucidating the mechanisms that underlie structural defects in Nipbl-deficient hearts is complicated by the fact that Nipbl is a ubiquitously expressed gene, and Nipbl-deficiency undoubtedly affects gene expression in every tissue of the embryo. Even given the data in Fig 1, which show changes in the expression of cardiac developmental genes at multiple stages of heart development, it is possible that the actual cause of ASDs lies elsewhere, especially since morphogenesis of the heart involves coordinated interaction among multiple cell types and tissue lineages. For example, the proper specification and patterning of cardiac mesoderm requires interaction with endodermal cells that provide a substrate along which cardiac progenitors migrate to form the cardiac crescent [31]. Indeed, in zebrafish, gene expression abnormalities in the endoderm appear to be the direct cause of some-but not all-cardiac abnormalities that accompany Nipbl deficiency [23]. Another lineage that contributes substantially to the heart is the neural crest, which although ectodermal in origin, contributes to the cardiac cushion and outflow tract and has been implicated in the etiology of septal defects [34].
To make it possible to investigate the roles of different cell lineages in the development of heart defects in Nipbl-deficient mice, we developed a Nipbl allelic series based on embryonic stem (ES) cells bearing a "conditional-invertible" (FLEX, or Flip-Excision [33,53,54]) gene trap in the Nipbl locus. We tested several Nipbl-gene-trapped ES cells that are available through public repositories, and ultimately selected, verified, and generated mice using ES cells bearing the Nipbl Gt(EUCE313f02)Hmgu allele (MGI: 4374347, hereafter known as Nipbl FLEX ), which is depicted in Fig 2A (see also S2 Fig). The gene-trap vector in these cells is inserted into intron 1 (14.5 kb downstream of exon 1) of the Nipbl gene, the same intron in which the gene-trap vector was inserted in the ES cells we used previously to generate Nipbl +/mice [22].
This vector introduces a FLEX cassette, which contains a β-geo reporter that reports on successful trapping, as well as flanking heterotypic recombinase target sites for both Cre and Flp recombinases [33,53,54], oriented so that exposure to either of these recombinases should lead to irreversible inversion of the trapping vector (due to excision of the cognate binding sites), loss of trapping, and loss of β-geo expression. Subsequent exposure to the other recombinase can then be used to induce a second round of irreversible inversion, restoring both trapping and β-geo expression. The different configurations of alleles in the Nipbl FLEX series, as well as the phenotypes of the different mouse lines obtained by successive rounds of recombination, are detailed in Fig 2B-2E, S2 and S3 Figs. Fig 2B illustrates salient features of Nipbl FLEX/+ mice, which were generated directly from chimeras produced using ES cells carrying the Nipbl FLEX allele. Nipbl FLEX/+ mice are, in accordance with predictions for the first (trapped) allele in the series, haploinsufficient for Nipbl and phenotypically similar to Nipbl +/mice by every measure tested: small body size, ubiquitous expression of the β-geo reporter, and reduced expression of Nipbl (assessed by qRT-PCR) ( Fig  2B). In addition, Nipbl FLEX/+ mice have a low survival rate, with only about 4% of pups surviving to weaning (4 Nipbl FLEX/+ survivors versus 104 wildtype littermate survivors across 17 litters; S3 Fig). As discussed in the next section, Nipbl FLEX/+ mice also faithfully replicate the heart defects seen in Nipbl +/mice.
When Nipbl FLEX/+ mice are crossed with transgenic Actin-FlpE ( Fig 2C) or Nanog-Cre mice (Fig 2D), to produce the genotypes we designate as Nipbl Flox/+ and Nipbl Flrt/+ , respectively, the progeny are normal in phenotype: embryos no longer express β-geo (note lack of X-gal staining), Nipbl transcript levels are restored to normal, and animals are indistinguishable from wildtype littermates in terms of size, rates of survival, and weight ( Fig 2C and 2D and S3 Fig). Finally, when Nipbl Flox/+ mice are crossed with Nanog-Cre mice to re-invert the gene-trap and generate what we refer to as Nipbl FIN/+ mice, the Nipbl FIN/+ progeny are again Nipbl-deficient, small in size, show ubiquitous β-geo expression, and survive poorly (Fig 2E legend). These results demonstrate that mice from the Nipbl FLEX allelic series can be successfully "toggled" between mutant and wildtype genotypes and phenotypes.

Nipbl FLEX/+ and Nipbl +/-Mice Develop Similar Heart Defects
To confirm that Nipbl FLEX mice phenocopy Nipbl-deficient mice, we compared a large number of hearts from Nipbl FLEX/+ and Nipbl +/embryos (and their wildtype littermates) at E17.5. We  high-resolution (50 μm voxel diameter) magnetic resonance imaging (MRI) to scan rapidly through many specimens. Although this procedure detected heart defects with high accuracy (see Materials and Methods), we frequently confirmed defects by paraffin histology.
The results are shown in Fig 3. At E17.5, both Nipbl FLEX/+ and Nipbl +/hearts display large atrial-septal defects (ASDs) at a similar frequency, about 30% (Fig 3). This number is somewhat smaller than previously reported for Nipbl +/mice (~50%), because a later time of assessment and more stringent criteria were used here; by these criteria we observed no ASD in wildtype littermates of Nipbl FLEX/+ and Nipbl +/mice. We also examined a large number of Nipbl Flox/+ mice, and found only a single ASD among 48 hearts examined (i.e., 2%, Fig 3C). All ASDs observed in Nipbl-deficient mice were of the ostium secundum type, similar to what is observed in individuals with CdLS, when ASD is seen as an isolated cardiac defect [15]. Ventricular septal defects were not observed in this analysis, nor were arterial stenoses or obvious abnormalities of ventricular wall thickness (S4 Fig). We did note that the hearts of Nipbl-deficient mice, whether Nipbl FLEX/+ or Nipbl +/-, are noticeably smaller than those of wildtype littermates-to about the same degree that Nipbl-deficient embryos themselves are smaller than wildtypes.
To verify the domains of Cre expression in these lines, we crossed them with Td-tomato-EGFP reporter mice [61], which express a membrane-targeted tomato red fluorescent protein that is replaced by a membrane-targeted enhanced green fluorescent protein (EGFP) when Cre-mediated excision occurs. As shown in Fig 4A, Nanog-Cre embryos express EGFP in all cells of the blastocyst inner cell mass, as expected. Nkx2-5-Cre and cTnt-Cre embryos express EGFP primarily in heart at E9-10.5, with some EGFP apparent in the first branchial arch for Nkx2-5-Cre. At E13.5, hearts of Wnt1-Cre embryos express EGFP in the inner lining of the great arteries and cardiac cushion, consistent with the distribution of neural crest contributions to the heart [62]. In Sox17-2A-iCre and FoxA2-2A-iCre embryos at E8-8.5, EGFP expression is observed in endodermal derivatives (foregut and hindgut), with small numbers of EGFP+ cells in developing heart. FoxA2-2A-iCre embryos also displayed some ectodermal (floorplate) and mesodermal (somite) EGFP.
Because recombination of the Nipbl Flox allele reactivates expression of β-geo, we could also use X-gal staining to document patterns of Nipbl inactivation produced by these Cre lines ( Fig  4B-4F). These results confirm that Nanog-Cre drives ubiquitous recombination (Fig 4B), while cTnt-Cre drives recombination specifically in the heart, and solely in myocardium (note sparing of the cardiac cushion in Fig 4C). Sox17-2A-iCre drives recombination in endodermal derivatives such as foregut and hindgut at E8.5, with additional staining in the heart that, by E10.5, can be seen to correspond to endocardium and cardiac cushion (Fig 4D), a pattern complementary to cTnt-Cre. Although FoxA2 is often regarded as an endodermal marker, the pattern of recombination driven by FoxA2-2A-iCre is more complex (consistent with its broader expression pattern compared to Sox17-2A-iCre in Fig 4A), and eventually comes to include sub-portions of epi-, myo-, and endocardium, while mainly sparing the anterior heart, outflow tracts and cardiac cushion ( Fig 4E). Wnt1-Cre embryos show essentially no recombination in the heart proper at E10.5, with X-gal staining appearing by E13.5 in a pattern confined to the cardiac cushion and developing great arteries ( Fig 4F).
These results document that these Cre lines can be used to manipulate Nipbl expression in essentially all of the major tissues that contribute to, or influence the development of, the heart. Furthermore, while some show overlapping patterns of recombination, cTnt-Cre and Sox17-2A-iCre are essentially complementary to one another-the former acting within the cardiomyocyte lineage, and the latter acting within endoderm plus non-cardiomyocyte mesodermal derivatives within the heart: the endocardium, endocardially-derived cells of the cardiac cushion, and vascular endothelium (these correspond to known domains of Sox17 expression [63]). Indeed, high magnification views of Td-tomato-EGFP reporter expression at E15.5 show that cTnt-Cre (Fig 4C') recombines in the vast majority of heart cells, including cells of the atrial septum, but spares endocardium, the cushion-derived atrioventricular valves, and scattered EGFP-negative cells in the ventricular parenchyma (most likely endothelial cells lining small blood vessels). In contrast, Sox17-2A-iCre spares the myocardium, but drives recombination in endocardium, atrioventricular valves, and small blood vessels throughout the ventricles (Fig 4D'). We first crossed mice carrying Cre-expressing transgenes with Nipbl Flox/+ mice, to yield progeny in which Nipbl deficiency is introduced into specific lineages, while the rest of the embryo retains normal Nipbl expression (Fig 5). Nanog-Cre was used as a control to produce mice that were Nipbl-deficient in all tissues. When Nipbl Flox/+ ;Nanog-Cre progeny were analyzed at E17.5, 33% of hearts displayed heart defects, the vast majority of which were large ASDs ( Fig  5A). These results are similar to those observed for E17.5 hearts from both Nipbl +/and Nipbl FLEX/+ mice (Fig 3).
Next, cTnt-Cre was used to create Nipbl deficiency specifically in cardiomyocytes ( Fig 5B). In this case, large ASDs were also found at a frequency of 30%, not significantly different from that seen with Nipbl +/-, Nipbl FLEX/+ , or Nipbl Flox/+ ;Nanog-Cre progeny (Figs 3C and 5A, S1 Data). This finding suggested that Nipbl deficiency in cardiomyocytes accounts for the high incidence of heart defects seen in globally Nipbl-deficient animals.
However, subsequent experiments suggested otherwise. Use of either Sox17-2A-iCre (expressed in endoderm and non-cardiomyocyte mesodermal derivatives) or FoxA2-2A-iCre (expressed both in endodermal and multiple other derivatives) to create Nipbl deficiency also resulted in a high incidence of ASD, about 26% in each case (Fig 5C and 5D). This is nearly as high as, and not statistically significantly different from, the level caused by either global or cardiomyocyte-specific Nipbl-deficiency (S1 Data).
Finally, we used Wnt1-Cre to investigate the role of neural crest. The neural crest not only contributes cells to cardiac structures, such as outflow tracts and valves, migrating neural crest cells interact in important ways with other cells that contribute to the heart [31,64]. Smith et al. [65] have suggested that Nipbl deficiency in mice impairs the functioning of cranial neural crest, and Schuster et al., [66], using zebrafish, recently proposed a neural crest origin for heart defects in cohesinopathies. In our experiments, however, no heart defects were seen in Nipbl Flox/+ ;Wnt1-Cre progeny (Fig 5E), indicating that Nipbl deficiency in the neural crest, at least on its own, is not sufficient to produce heart defects.
Overall our results show that when cells derived from cardiogenic mesoderm, endoderm, or subpopulations of both are made deficient in Nipbl, heart defects, primarily ASD, always develop at a frequency of approximately 30%, the same incidence as observed in mice that are globally Nipbl-deficient (Figs 3 and 5A, and S1 Data). This is a striking finding, since it suggests that the effects of Nipbl-deficiency in different cardiac developmental lineages, even non-overlapping lineages, are not additive.

Restoration of Nipbl Expression to Either of Two Distinct Lineages Rescues Heart Defects
In experiments complementary to those described above, we crossed Nipbl FLEX/+ mice with Cre-expressing transgenic mice, to produce embryos that are globally Nipbl-deficient except within those lineages in which Cre recombinase acts. We focused on three Cre-expressing lines: Nanog-Cre, cTnt-Cre, and Sox17-2A-iCre.
As shown in Fig 6A, the hearts of Nipbl FLEX/+ ;Nanog-Cre mice lacked ASDs, and were phenotypically indistinguishable from wildtype, as expected for a global rescue of Nipbl expression. Interestingly, in Nipbl FLEX/+ ;cTnt-Cre hearts, the incidence of ASDs was also very low: of 19  hearts analyzed, only 1 displayed an ASD (~5%), which is significantly different from Nipbl FLEX/+ , and statistically indistinguishable from wildtype (S1 Data). Remarkably, Nipbl FLEX/+ ;Sox17-2A-iCre hearts also displayed a very low incidence of ASDs (5%, Fig 6C and  6D) that was indistinguishable from wildtype (S1 Data). Chi-square analysis indicated that, for all three types of Nipbl FLEX/+ ;Cre embryos, none was distinguishable from any other in terms of the observed incidence of heart defects, but all were significantly different from Nipbl FLEX/+ (Fig 6D).

Key Determinants of ASD Risk Most Likely Lie Elsewhere in the Embryo
The above results indicate that Nipbl deficiency in either of two non-overlapping sets of cellsthe "cTnt lineage", by which we mean descendants of cTnt-Cre expressing cells, and the "Sox17 lineage", by which we mean descendants of Sox17-Cre expressing cells)-will cause heart defects, while rescue of Nipbl deficiency in either of the same two populations rescues those defects. With respect to the creation of heart defects, the first result implies that Nipbl deficiency in either lineage is sufficient, while the second result implies that Nipbl deficiency in neither lineage is sufficient-which certainly seems paradoxical.
The problem is highlighted in tabular form in Table 1. Nipbl Flox/+ ;Cre experiments ("conditional haploinsufficiency," Fig 5) and Nipbl FLEX/+ ;Cre experiments ("conditional rescue," Fig 6) both generate embryos in which cardiomyocytes are Nipbl-deficient, and endoderm, endocardium, and vascular endothelium are not; or endoderm, endocardium, and vascular endothelium are deficient and cardiomyocytes are not. Yet opposite results, with respect to heart defects, are obtained for the same genotypes in the two types of experiments. The key to resolving this apparent paradox is to remember the additional variable that distinguishes the two classes of experiments: the rest of the embryo. In conditional haploinsufficiency experiments, all lineages outside the one in which Cre acts are wildtype. In conditional rescue experiments, all lineages outside the one in which Cre acts are Nipbl deficient.
For this difference between the genotypes of the rest of the embryo in the two experimental approaches to explain the results, a critical determinant of ASD risk would have to lie in some lineage other than those represented by cTnt and Sox17. Furthermore, in that lineage, risk would have to be conferred by being Nipbl-wildtype, while protection would have to be conferred by being Nipbl-deficient. That would explain why when the rest of the embryo is Nipblwildtype, ASDs arise when either cardiac lineage is Nipbl-deficient, whereas when the rest of the embryo is Nipbl-deficient, ASDs arise only when both are.
The possibility that an additional lineage protects against heart defects when Nipbl-deficient led us to consider ways in which essentially non-cardiac developmental events might affect the heart indirectly. One of the most penetrant phenotypes of Nipbl-deficiency is reduced body size (by~20% at birth; [22] and S3 Fig, Fig 7 below). Not surprisingly, the determinants of body size lie outside the heart. As shown in Fig 7A-7E, in the Nipbl Flox/+ and Nipbl FLEX/+ crosses described above, body size is determined by Nipbl-status outside of the cTnt and Sox17 lineages (i.e., all carriers of the Nipbl FLEX allele are small, and all carriers of the Nipbl Flox allele are normal in size). Yet it is also observed that heart size (measured as total ventricular volume) correlates strongly with body size (Fig 7F-7I). Thus, lineages outside the heart and endoderm Data: Sample Numbers) had an incidence of heart defects ranging from 0%-5% (A-D). Nipbl deficiency in the Wnt1 domain (neural crest) did not give rise to heart defects (E). CHDs occurred primarily in the form of ASDs of the ostium secundum type (yellow arrowheads); VSDs observed in A and D are of the perimembranous type. p-Values are from Chi-square analyses and are indicated for each corresponding cross. Scale bar = 500 μm. lv, left ventricle; rv, right ventricle. Restoration of Nipbl expression in different heart lineages rescues heart defects. MRI analysis of E17.5 hearts show that Nipbl FLEX/+ mice exhibit ASDs at a frequency of~30% (yellow arrowheads in A-C, and D). The incidence of heart defects was significantly reduced in mice in which Nipbl was restored in all tissues (Nipbl FLEX/+ ;Nanog-Cre), specifically in the cTnt domain (Nipbl FLEX/+ ;cTnt-Cre), or specifically in the Sox17 domain (Nipbl FLEX/+ ;Sox17-2A-iCre) (D, asterisks: Chi-square, p < 0.05). Chi-square analysis on the incidence of heart defects between the three rescued lines indicated no significant difference from one another (Chi-square, p > 0.71). Progeny are on various backgrounds depending on parental backgrounds: Nipbl FLEX/+ (mixed), wildtype (CD-1), cTnt-Cre (CD-1), Nanog-Cre (C57Bl6/J), and Sox17-2A-iCre (C57Bl6/ J); the incidence of heart defects in Nipbl FLEX/+ embryos from each of these crosses did not differ significantly (p > 0.89 by Chi-square analysis; S1 Data). Size bar = 500 μm. lv, left ventricle; rv, right ventricle apparently determine the size of the embryonic heart, just as heart and body size are known to scale together in adults [67]. Below (see Discussion), we raise the possibility that the results in Table 1 might be explained by an influence of heart size on ASD risk, with large hearts being at greater risk for defects than small ones.

Understanding the Incomplete Penetrance of Heart Defects
One of the puzzling aspects of the above experiments is that the incidence of ASDs always seems to be either very low (5%) or about 30%, regardless of genotype. Particularly surprising is the lack of increase in incidence when the entire embryo is Nipbl-deficient, as opposed to a single lineage (e.g., Fig 5A-5C). Remarkably, a similar 30% incidence of heart defects is also seen clinically in CdLS [14,15].
We wondered whether this 30% penetrance "ceiling" is a peculiarity of the specific gene expression disturbances caused by Nipbl deficiency, or whether it reflects something about the overall state of hearts at the time that Nipbl-sensitive defects emerge. For example, is it simply that, at that stage, 30% of hearts are more labile to disturbances overall (e.g., due to embryo-toembryo variations in the intrauterine environment)?
To explore this idea, we decided to make mice doubly-heterozygous for Nipbl and Nkx2-5. As discussed previously, Nkx2-5 is a key, early cardiac transcription factor. Moreover, Nkx2-5 gene dosage is clearly important in heart development, because haploinsufficiency for NKX2-5 gives rise to congenital heart disease in man [68,69]. As shown above (Fig 1F), early expression of Nkx2-5 is reduced in Nipbl +/embryos, so it is reasonable to suspect that at least some of the cardiac phenotype of Nipbl +/embryos arises from a deficiency in Nkx2-5. Combining Nipbl and Nkx2-5 heterozygous mutations should then provide us with an opportunity to observe the effects of lowering Nkx2-5 levels even further.
For these experiments, we used a knock-in Nkx2-5 Cre allele as a null allele [55,70]. We confirmed that Nkx2-5 Cre/+ hearts do in fact express Nkx2-5 at half the wildtype level using qRT-PCR (S5 Fig), and hereafter refer to them as Nkx2-5 +/mice (it is the null state of this allele that also made it unsuitable for use in the conditional experiments in Figs 5 and 6). We crossed Nkx2-5 +/mice with our original Nipbl +/mouse line, in which Nipbl is not flanked by LoxP sites [22] (so the Cre produced by the Nkx2-5 transgenic mouse would be irrelevant), and evaluated hearts at E17.5.

Discussion
The Nipbl +/mouse-a model of Cornelia de Lange Syndrome (CdLS)-provides a unique resource for studying how the combinatorial effects of genetic variation cause birth defects, because the sole consequence of Nipbl haploinsufficiency appears to be small quantitative adjustments to the levels of expression of hundreds to a thousand genes [21][22][23]25]. As described above, Nipbl +/mice display large ASDs-one of the most common CHDs in the general population-at a frequency of about 30%, similar to the incidence of CHD in CdLS.
Not only do Nipbl +/mice display defects of the atrial septum, which normally forms and closes by E14.5 [30], they exhibit delayed closure of the ventricular septum, which normally takes place a day earlier (Fig 1A), and reduction in the size of the right ventricle several days before that (Fig 1C and 1D). Even earlier, at the cardiac crescent stage (E7.5), such mice show decreased expression of Nkx2-5; and a day before that-at primitive streak stage (E6.5)reduced or delayed expression of the earliest-known marker of cardiogenic mesoderm, Mesp1. These data suggest that the ASDs that arise in Nipbl-deficient embryos may have their origin in events as early as gastrulation. This is in good agreement with findings in nipbl-deficient zebrafish, in which the origin of several heart defects could be traced to the initial migration of cardiogenic mesoderm [23]. It also supports studies that suggest that CHDs frequently have early developmental origins [51,68,[71][72][73].
Early heart development involves interactions among all three germ layers of the embryo: the mesoderm, which produces cardiomyocytes, endocardium, epicardium, and vascular endothelium; the endoderm, which forms an essential substrate along which cardiomyocyte progenitors migrate and proliferate as they form the cardiac crescent; and the ectoderm, which is the source of neural crest cells that contribute to the cardiac cushion and outflow tract [64]. To sort out the relative contributions of different lineages to the production of ASDs in Nipbldeficient mice, we took advantage of recent improvements in gene-trap technology (Fig 2) to selectively create, or rescue, Nipbl haploinsufficiency in cardiomyocytes, endoderm and endocardium/endothelium, neural crest, or non-selectively throughout the entire embryo. The results (Figs 5 and 6) implicated both the cTnt (cardiomyocyte) and Sox17 (endoderm and endocardial/endothelial lineages), but not the neural crest. An absence of a role for neural crest is consistent with its fairly late contribution to the heart (Fig 4F), relative to the early molecular abnormalities that occur in the hearts of Nipbl-deficient mice (e.g., Fig 1F and 1G).
Surprisingly, we found that creating Nipbl haploinsufficiency in either the cTnt lineage, the Sox17 lineage, or the entire embryo produced ASDs at the same frequency (Fig 5). Even more surprisingly, we found that rescue of Nipbl haploinsufficiency in either of these two lineages in an embryo that is otherwise globally Nipbl-deficient rescued those ASDs (Fig 6). These results-in one case either lineage being sufficient to produce defects and in the other case neither lineage being sufficient to do so-imply that an additional determinant of CHD risk lies outside of both lineages (and thus, most likely, outside both the heart and endoderm). That determinant would have to be protective when the lineage responsible for it is Nipbl-deficient (as all non-Cre-expressing lineages are Nipbl-deficient in conditional rescue experiments), but it must be only partially protective in order to explain why globally Nipbl-deficient embryos do get ASDs.
We speculate that this determinant may be heart size: Nipbl-deficient embryos are significantly smaller than wildtype littermates, and so are their hearts. Moreover, heart size is clearly determined by the Nipbl genotype of the rest of the body, not the genotype of the cells of the cTnt and Sox17 lineages (Fig 7). The idea that having an abnormally small heart might lower the risk of CHD may seem counterintuitive, but it makes sense if we think of heart development as a process in which a limited pool of progenitor cells must generate a large number of differentiated cells in a short period of time. Under such conditions, a shortfall in cell production might be easier to tolerate if the heart that needs to be built is a smaller one.
This view fits with much of what we know about heart development. The heart is the earliest organ system to become truly functional, and even short delays in cell production tend to be lethal (for example, a 50% deficiency in developing cardiomyocytes at E9.5, due to cardiac-specific deletion of the growth-promoting gene Yap, leads to embryo death within one day [74]). In mammals and birds, the need to rapidly remodel the initially simple linear heart tube into an elaborate, four-chambered structure requires the addition of large numbers of cells, many of which arrive by migration from the second heart field (SHF), a collection of cardiac progenitor cells that lie outside the heart tube, and which have been adapted to undergo prolonged proliferation [32,75,76]. Recent studies show the SHF provides most of the cells that drive the early expansion of the heart, including cells that add to ventricles and atria and form both the atrial septum and parts of the ventricular septum [32,34,76].
Interestingly, the SHF gives rise to almost the entire right ventricle [76], the same ventricle that is disproportionately small in Nipbl +/hearts (Fig 1). This result is consistent with a model in which the output of the SHF is reduced in Nipbl-deficient embryos. An insufficiency of cardiac progenitors would also be consistent with the reduced expression of Nkx2-5 and Mesp1 expression that we observe in early Nipbl-deficient embryos. Furthermore, it would fit with results in nipbl-deficient zebrafish, in which depletion of cardiac progenitors seems to be the result of an insufficiency of endoderm along which such progenitors migrate [23]. Interestingly, in that system, restoring the number of endoderm cells by overexpression of endodermspecific transcription factors rescues several heart defects.
One appealing feature of a model in which Nipbl deficiency drives progenitor cell numbers to a point at which a large heart cannot always form properly, but a small heart can, is that this model could potentially explain why the incidence of ASDs in conditional Nipbl-deficient mice is always~30%, regardless of whether Nipbl-deficiency occurs in the cTnt or Sox17 lineages individually or throughout the entire embryo. It may simply be that, in globally Nipbl-deficient animals, effects of reduced Nipbl in the cTnt lineage and the Sox17 lineage do interact additively, but we do not observe a stronger phenotype because of the protective effect of small heart size in such mice, which exerts a phenotypic effect in the opposite direction. Future experiments will be needed to verify this conjecture.
Overall, a consistent explanation for all of the conditional mutant phenotypes in the present study is that Nipbl haploinsufficiency, in either the cardiac mesoderm or the endoderm upon which it migrates (or both), leads to defective expansion of cardiac progenitors, with an ultimate impact on cardiac morphogenesis that depends on the demands imposed by the rest of the embryo on final heart size. Validating such an explanation will ultimately require measurements of cell numbers and proliferation rates in very early embryos that are beyond the scope of the present study. In addition, the present study did not evaluate possible contributions of epicardium (which derives from neither the cTnt or Sox17 lineages) to Nipbl mutant phenotypes, although the timing at which epicardium appears during development suggests it could not explain the early gene expression abnormalities described here (e.g., those in Fig 1F and  1G). Furthermore, since this study focused primarily on ASD, we cannot comment on whether similar multilineage interactions are likely to play a role in other kinds of cardiac defects.
Nevertheless, the present work calls attention to the fact that the embryonic lineages responsible for organogenesis can interact in ways both direct and indirect-global enforcement of scaling relationships being an example of the latter. A consequence of indirect interaction is that determinants of risk for developmental defects can easily turn up in unexpected places, as they did here. This point may need to be kept in mind as efforts continue to be made to discover the genetic causes of human CHDs.

Materials And Methods Animals
Ethics statement. All animals were handled in accordance with approved procedures as defined by the National Institutes of Health, and all animal work was approved by the Institutional Animal Care and Use Committee of the University of California, Irvine. For collection of mouse tissues, pregnant dams were humanely killed by CO 2 anesthesia followed by cervical dislocation. Noon on the day of vaginal plug appearance was designated embryonic day 0.5 (E 0.5). For young embryos (E7-E13.5), embryos were dissected in ice-cold PBS and tissue collected for processing. For E17.5 mouse fetuses, fetuses were removed from the uterus, placed on ice and immediately decapitated, then heart and other tissues removed for further study.
Genotyping. All mice and mouse embryos were genotyped by polymerase chain reaction (PCR) using genomic DNA obtained from tail biopsies or tissues taken post-dissection. Primer sequences and PCR conditions for all genotyping performed in these studies is detailed in S2 Table. Mouse lines. Nipbl +/line: The Nipbl +/mouse line (Nipbl Gt(RRS564)Byg , MGI:4332250: Nipbl +/-) was maintained on an outbred CD-1 (Charles River) background. Male Nipbl +/mice were naturally bred with females to obtain embryos used in these studies.
Nipbl Flrt/+ line: Nipbl FLEX/+ male mice were bred with Nanog-Cre mice to generate Nipbl Flrt/+ mice, in which the original Nipbl FLEX allele was recombined and Flp recognition sites (Frt and F3) remained in active conformation. To isolate the Nipbl Flrt allele from the Nanog-Cre allele, double heterozygous mice were crossed with CD-1 mice. Nipbl Flrt/+ mice are phenotypically wildtype (Fig 2D). This line has not been maintained.
Generation of mouse embryos for testing tissue-specific depletion and rescue of Nipbl expression. To generate embryos for Nipbl depletion experiments, Nipbl Flox/+ mice were naturally mated with mice carrying any one of several Cre transgenes (Nanog-Cre, Sox17-2A-iCre, FoxA2-2A-iCre, Wnt1-Cre, or cTnt-Cre). To generate embryos for Nipbl rescue experiments, cryopreserved Nipbl FLEX/+ sperm was used for in vitro fertilization (IVF) of eggs from superovulated females of indicated mouse lines heterozygous for Cre-expressing transgenes. Sperm cryopreservation and IVF procedures were carried out by the UCI Transgenic Mouse Facility using published procedures [80]. Because Cre-driver mice were not available on uniform genetic backgrounds, potential genetic background effects in these experiments were minimized by making comparisons within litters, or within groups of litters in which the distribution of genotypes across the litters was as uniform as possible.

Histochemistry, In Situ Hybridization, and Microscopy
Whole mount X-gal staining on E7.75-13.5 mouse embryos for detecting β-galactosidase activity was performed as described [81]. Histological evaluations were performed on tissue fixed in 4% paraformaldehyde (PFA) in phosphate buffer or neutral buffered formalin (VWR 16004-126). Hematoxylin and Eosin Y (H&E) staining was performed on 20 μm cryosections (for E13.5 hearts) or 7 μm paraffin sections (for E17.5 hearts) using standard techniques. E17.5 hearts were positioned at a canonical angle for analysis in histological sections and MRI studies: atria superior to ventricles, with the apex of the heart inferior and the left atrium and ventricle to the right; sections were assessed from most ventral to most dorsal.
Whole mount ISH, X-gal, and fluorescent images, as well as all images of paraffin-sectioned hearts, were obtained using a Discovery V8 stereomicroscope equipped with Axiovision software (Zeiss). Confocal microscopy was performed on 30 μm sections of E15.5 hearts. Images were taken every 5 μm using a 20x 0.75NA Olympus objective on an Olympus Fluoview FV1000 microscope, and processed and stitched using Python and ImageJ Software as described [83].

Magnetic Resonance Imaging (MRI)
Torsos from E17.5 embryos containing intact cardiopulmonary organs (heart and lungs) were fixed in neutral buffered formalin for a minimum of 1 wk at 4°C. Intact torsos, or torsos with ribs removed, were rinsed with phosphate buffered saline (PBS; 2 x 5 min.) and soaked in 2.5 mM Gadoteridol (Gd-HP-DO3A, a.k.a. ProHance, Braccoo Diagnostics Inc., Princeton, NJ) in PBS for 12 h at 4°C. After soaking, samples were embedded in 2% low melting point (LMP) agarose (Sea Plaque GTG Agarose, FMC Bioproducts) and stacked in a 20 mm diameter glass tube. The chamber was then filled with perfluoro-polyether Galden-D (Inland Vacuum Industries, Churchville, New York) to limit tissue dehydration as well as susceptibility effects at the surface of the specimen. Imaging was performed in a vertical bore 11.7T (500 MHz) Bruker AVANCE imaging spectrometer with a microimaging gradient insert and 20 mm birdcage RF coil. Images were acquired with a 3D RARE protocol (TR/TE effective , 250 ms/16 ms) with a RARE factor of 4 [84], voxel size of 50 3 μm 3 with typical image matrix: 512x320x320 and field of view (FOV): 25.6x16x16 mm 3 , and number of acquisitions = 10-18. The FOV and matrix sizes were modified to accommodate differing numbers of samples and sample sizes. Scanning was performed at 15°C to minimize noise.
Each heart from a scan was isolated as a file of 80 3 pixel 3 using ImageJ (v. 1.45s). Heart scans were evaluated for the presence or absence of ASD, VSD and/or PTA using the volume viewer plug-in of ImageJ. Hearts were scored positive for ASD when: 1) a minimum of two consecutive sections (100 microns) lacked atrial septum; 2) gap in atrial septum tissue was greater than 3 pixels (150 microns); and 3) growth of both atrial septum walls (septum primum and septum secundum) were stunted (observed as flat) at both the superior and inferior ends of the developing atrial septum at the region of the atrio-ventricular valves. Hearts were scored positive for VSD when ventricular septum was not continuous in one or more sections of the MRI scan. Hearts were scored positive for PTA when the aorta and pulmonary artery were observed as fused. Ventricular volumes were calculated from MRI data using the outline function of ImageJ: The outlined ventricle area was calculated for each section and multiplied by section thickness (section thickness = 1 pixel or 50 μm); total ventricular volume for each heart was calculated as the sum of these volumes. Heart defects detected by MRI were confirmed by paraffin sectioning and histological staining. Chi-square test was used to calculate p-values for the frequency of heart defects (ASD, VSD, and/or PTA) at E17.5 for all crosses. Mann-Whitney U test was used to calculate p-values for measurements of the great vessels, ventricle wall thickness, and ventricular volume at E17.5.
Optical Projection Tomography E10.5 embryos were fixed in 4% PFA, washed thoroughly in PBS and embedded in LMP agarose. Agarose blocks were affixed to metal mounts with cyanoacrylate Krazy Glue, trimmed, dehydrated through three changes of methanol for 8-12 h, and then immersed in 2:1 benzyl benzoate:benzyl alcohol mixture until optically clear (minimum 6 h). Specimens were scanned using an optical projection tomography scanner (OPT 3001M; Bioptonics, United Kingdom), under ultraviolet light using a GFP1 filter (exciter 425 nm/40 nm, emitter LP475 nm). Tomographic reconstruction was carried out using NRecon v. 1.6.1 (Skyscan, Belgium) to generate a series of bitmap images. Three-dimensional reconstructions were rendered from stacked bitmaps using ImageJ v. 1.41o (http://resb.info.nih.gov/ij) and Amira v. 5.2.2 (Visage Imaging, USA).

Quantitative RT-PCR
RNA extraction was performed using an Aurum Total RNA kit (BioRad, USA) according to the manufacturer's instructions. RNA was reverse-transcribed into cDNA using oligo dT and random hexamers with Superscript II reverse transcriptase (Invitrogen). Quantitative real-time PCR amplifications were performed in an iQ5 iCycler real time PCR detection system (Bio-Rad) using the iQ SYBR Green supermix (Bio-Rad). qRT-PCR was performed with primer sets given in S1 Table, using the following cycles: an initial cycle at 95°C for 3 min, followed by 40 cycles of 95°C for 10 s each, 59°C for 30 s, and 72°C for 30 s, followed by melt-curve analysis from 61-95°C in 0.5°C increments. mRNA expression was normalized to expression of B-2-microgoblulin (B2m) or glyceraldehyde 3-phosphate dehydrogenase (GAPDH), and the relative expression of target genes was obtained by the 2 -ddCt method [85]. A range of 3-14 biological replicates (individual hearts between the ages of E10-13.5, brain or kidney at E17.5) were assessed in technical duplicates or triplicates for each gene. Statistical significance was determined using Student's t test (with Bonferroni correction where indicated).   Fig. Heart development genes showing no significant change in expression in Nipbl+/heart versus wildtype controls. Q-RT-PCR was performed as described in Materials and Methods, and S1 Table. Values for each tested gene were normalized to expression value of B-2-microgoblulin (B2M) in the same PCR run, and relative expression of tested genes was obtained by the 2 -ddCt method as described in Materials and Methods. 3-14 biological replicates (individual hearts) were assayed in duplicate for each gene (see S1 Data). Data are expressed as mean (± SEM); statistical significance was determined using Student's t-test with Bonferroni correction. No gene tested showed a significant difference in Nipbl +/hearts versus wildtype controls, except Nipbl, which was tested at E10.5 (52%, P<0.001). (TIF) S2 Fig. Conditional-invertible gene modifications of the Nipbl FLEX allelic series. A. Schematic of EUCE313f02 (Nipbl FLEX ) allele from which the Nipbl FLEX/+ mouse line and subsequent Nipbl Flox/+ , Nipbl Flrt/+ , and Nipbl FIN/+ mouse lines are derived using Cre and Flp DNArecombinases). Top: The SA-βgeo-pA (rsFlp-Rosa-bgeo) cassette in the Nipbl FLEX allele traps Nipbl expression after Exon1, and instead expresses the β-geo reporter gene resulting in a null allele. Middle: Flp recombinase (right path) inverts the SA-βgeo-pA cassette at frt and F3 sites, simultaneously excising the heterotypic recognition targets, and locking the cassette against reinversion, resulting in the Nipbl Flox allele. The same principle is adaptable to Cre recombinase, using the LoxP and lox5171 sites, resulting in the Nipbl Flrt allele (left path). Either (Flp or Cre mediated) inversion activates normal splicing between the endogenous splice sites (Exon 1 to Exon 2), skipping the inverted SA-βgeo-pA cassette, thereby repairing the mutation and resulting in a phenotypically-wildtype allele. Bottom: Subsequent Cre (right path) or Flp (left path) recombinase-mediated inversion repositions the SA-βgeo-pA cassette, activating the gene-trap again, and re-introducing the mutation, resulting in the phenotypically null Nipbl FIN allele. Each allele can be distinguished by genomic PCR, in which primers are designed within the rsFlp-Rosa-βgeo cassette (across the β-geo reporter and the arm of cassette outside of the rates. The weights of Nipbl FLEX/+ mice overlapped within the first few days of life regardless of whether they survived to weaning or not. Growth of un-genotyped non-survivors (green Xs) was similar to Nipbl FLEX/+ non-survivors (blue filled boxes) within the first 12 days of life, suggesting that pups that failed to survive were likely Nipbl FLEX/+ . Some pups that were logged as having been born could not be genotyped because tissue could not be recovered. B. Growth curves of Nipbl Flox/+ mice compared to wildtype littermates from P1 to P21, from 5 litters. Data are pooled by genotype. Data show overlapping growth and similar sizes of wildtype and Nipbl Flox/+ mice. For purposes of clarity, only upper error bars for Nipbl Flox/+ mice (SD, blue) and lower error bars for wildtype (SD, black) are shown. C. Growth curves for Nipbl Flrt/+ mice compared to wildtype littermates from P1 to P21, from 4 litters. Data show overlapping growth and similar weights of wildtype and Nipbl Flrt/+ mice. For purposes of clarity, only upper error bars for Nipbl Flrt/+ mice (SD, green) and lower error bars for wildtype (SD, black) are shown. (TIF) S4 Fig. Nipbl FLEX/+ hearts do not display pulmonary stenosis, aortic stenosis, or ventricular hypertrophy. To assess stenosis, diameters of the aorta and pulmonary artery were measured from images using Axiovision software (Zeiss). 90% of pulmonary stenoses are isolated at the valvular region, so diameters were measured just distal to the semilunar valves (supravalvular). To confirm that any stenoses observed did not involve the pulmonary trunk/aorta, serial sections were followed by the observer to the bifurcation. To assess hypertrophic cardiomyopathy, ventricular wall thickness was measured and the ratio of the right to left wall thickness was calculated for each sample, and different genotypes were compared. Diameter of the ventricular wall (left and right myocardium, excluding trabeculae) was measured in images of H&E stained paraffin sections in the region that was midway through the atrioventricular valve (ventral-dorsal) and midway through the ventricular septum (superior to inferior). A. The supravalvular diameter of the great arteries of paraffin sectioned wildtype and Nipbl FLEX/+ hearts were measured as indicated by black bar. B. Both the pulmonary artery diameter and the aorta diameter (C) were significantly smaller in Nipbl FLEX/+ hearts (N = 11) compared to wildtype littermates (N = 9) (Mann-Whitney U, P<0.05). D. The thickness of the ventricular walls (excluding trabiculae) were measured as indicated by black bar. Ventricular wall thickness was significantly reduced in Nipbl FLEX/+ hearts (N = 11) compared to wildtypes (N = 9) on both the right (E) and left (F) walls (Mann-Whitney U, P<0.025). G. A ratio of the right to left ventricle wall thickness revealed no significant difference between wildtypes and Nipbl FLEX/+ mice (Mann-Whitney U P>0.05). Differences observed in C, D, E, and F likely due to the overall smaller body and heart size observed in Nipbl FLEX/+ embryos (see results and Fig 7). LV, left ventricle; PA, pulmonary artery; RV, right ventricle. (TIF) S5 Fig. Nkx2-5 Cre/+ hearts have half the normal level of Nkx2-5 expression. Q-RT-PCR was performed on E10.5 Nkx2-5 +/+ (N = 7) and Nkx2-5 Cre/+ (N = 5) hearts. Nkx2-5 Cre/+ hearts had approximately half the Nkx2-5 expression of Nkx2-5 +/+ littermates. Nkx2-5 mRNA expression was normalized to expression of B2m, and the relative expression of Nkx2-5 was obtained by the 2 -ddCt method; data are expressed as mean ± SEM. PCR primers and conditions are described in Materials and Methods, and S1 Table. Asterisk: Student's t-test, P = 0.001. (TIF) Irvine is a designated Research Center of Excellence for CdLS. The authors thank Michelle Digman and Jenu Chacko for assistance and advice with confocal microscopy and Sheetal Jotwani for assistance in OPT image analysis. We thank the many dedicated undergraduate researchers who assisted with mouse husbandry and genotyping and Mona Yazdi, Ceyda Yaramanoglu, Tritia Schostak, Harrison DiStefano, Phoebe Yam, David Razo, Salvador Deniz, Qumber Ali, Phoebe Valdes, and Louise A. Villagomez for technical support and assistance with data acquisition for this project. We thank the UC Irvine Transgenic Mouse Facility (TMF) for assistance with IVF experiments. The TMF is a shared resource funded in part by the Chao Family Comprehensive Cancer Center support Grant (P30CA062203) from the National Cancer Institute, and by strategic partnerships with various UCI offices and schools. We also thank the Biological Imaging Center of the Beckman Institute at Caltech; the Canadian Foundation for Innovation, Alberta Innovates, and the University of Calgary; and University of Southern California Translational Imaging Center for support.

Author Contributions
Conceptualization: RS SK BH SEF ADL ALC.