Abstract
Nuclear factor of activated T cells (NF-AT) is the name of a family of four related transcription factors that may be needed for cytokine gene expression in activated lymphocytes1, 2, 3, 4. Here we report that mice with a targeted disruption of the NF-ATc gene show an unexpected and dramatic defect in cardiac morphogenesis, with selective absence of the aortic and pulmonary valves, leading to death in utero from congestive heart failure at days 13.5–17.5 of gestation. In contrast, tricuspid and mitral valve morphogenesis is normal. NF-ATc is the first transcription factor known to be expressed only in the endothelial cells of the heart. As in T cells, nuclear translocation of NF-ATc in cardiac endothelial cells is controlled by the calcium-regulated phosphatase calcineurin5,6: NF-ATc remains cytoplasmic in normal embryos cultured with cyclosporin A, an inhibitor of calcineurin. Abnormal development of the cardiac valves and septae is the most frequent form of birth defect, yet few molecular regulators of valve formation are known. Our results indicate that NF-ATc may play a critical role in signal-transduction processes required for normal cardiac valve formation.
To determine whether the NF-ATc protein had unique functions in vivo, the NF-ATc gene was disrupted in embryonic stem (ES) cells by replacing 19 amino acids (635–653) of the Rel-homology domain with a neomycin-resistance gene (neo). This resulted in a frameshift of the remaining amino acids of the protein (654–710) (Fig. 1a). Two of six ES clones, named 141 and 112, transmitted the disrupted allele to offspring and NF-ATc+/- mice were intercrossed to generate NF-ATc-/- mice. Of 181 pups born, there were no homozygotes and a slightly lower number of heterozygotes than expected, suggesting that NF-ATc is required for survival. Serial timed matings produced the expected mendelian ratios of +/+, +/- and -/- animals at embryonic day of development (e) 8.5–12.5 (Table 1). From e13.5–17.5, however, increasingly fewer live -/- animals were present, and at birth no surviving -/- embryos were found. Figure 1b shows a Southern blot analysis of e14.5 embryos representative of the three genotypes.
Figure 1: Targeted disruption of the murine NF-ATc gene.
a, At the top is shown the intron–exon (exons are boxed) structure of the 3' terminus of genomic DNA encoding a portion of the NF-ATc Rel-homology domain (RHD) (black box), isolated from a 129/sv genomic library. In the middle a replacement construct containing the neo cassette, indicating the replacement of the RHD (amino acids 635–653) with neo, is shown. The region predicted to undergo replacement is indicated with dotted lines between the wild-type locus and the targeting construct. The recombinant loci for the independently derived ES clones 141 and 112 are shown at the bottom. The probes and restriction sites used for diagnositc Southern blotting are indicated (B, BamH1; E, EcoR1; H, HindIII; S, Sal1; Sc, Sac1; N, NdeI. b, Analysis of DNA from e14.5 embryonic tissue of mice resulting from matings of NF-ATc+/- mice from line 141 or 112. As predicted from the restriction map of the wild-type locus, digestion of DNA with BamH1 generates a 13-kb fragment which is replaced by an 8.5-kb fragment when hybridized with the 3' probe. Hybridization with the 5' probe yielded the expected 13-kb band in the wild-type and 4-kb band in the 141 line but a 2-kb band in the 112 line, consistent with an additional 2-kb intronic deletion (not shown).
High resolution image and legend (23K)
By e14.5, the -/- embryos sometimes appeared pale and oedematous, with frequent pericardial effusions and abdominal haemorrhages. The fact that these animals died at e13.5–17.5 suggested a primary cardiac or haematopoietic defect7. A consistent pattern of defects was seen in embryos at e14.5 (Fig. 2) and e13.5 (not shown). There was a conspicuous absence of both the pulmonary and the aortic valve leaflets, but there was normal partitioning of the large vessels into a clearly defined right ventricular, or pulmonary, outflow and left ventricular, or aortic, outflow. There was also a defect in the superior margin of the interventricular septum which resulted in a ventricular septal defect. There was no impairment in formation of the compact layer of the ventricular myocardium, or in formation of ventricular trabeculae epicardial morphogenesis, or coronary vascularization (Fig. 2). Importantly, there were no definitive abnormalities in tricuspid or mitral-valve formation or in the inferior portion of the ventricular septum, although these abnormalities are often seen with atrioventricular valve abnormalities (Fig. 3a, b). The phenotypes of the 112 and 141 lines were qualitatively similar, although the 141 line was more severely affected: 141-/- embryos were completely impaired in aortic and pulmonic valve development, whereas 112-/- embryos had less pronounced valvular defects.
Figure 2: Histological analysis of e14.5 wild-type and NF-ATc-mutant embryos from line 141.
a, b, Show low-power magnification of comparable cross-sections through the right ventricular outflow tracts (rv) of wild-type (a) and mutant (b) embryos. c, d, Show high-power magnification of null-mutant embryos (d) and controls (c). e, f, Show low-power magnification of comparable cross-sections through the left ventricular outflow tract (lv) of wild-type (e) and null-mutant (f) embryos. High-power magnifications revealed severe defects in formation of the aortic valve leaflets (av) in the null -mutant embryos (h) when compared with controls (g). j, l, Show the ventricular septal defect (vsd) in the interventricular septum (ivs) of NF-ATc-null mutants and i, k show the situation in wild-type littermates. Ao, aorta; pv, pulmonary vein; aps, aortico–pulmonary septum; pa, pulmonary artery; cv, coronary vessels; ra, right atrium; rap, right atrial appendage; la, left atrium; lap, left atrial appendage; mv, mitral valve; edc, endocardial cushion.
High resolution image and legend (84K)Figure 3: Histological analysis of e14.5 and e11.5 wild-type and NF-ATc-/- embryos from line 141.
![Figure 3 : Histological analysis of e14.5 and e11.5 wild-type and NF-ATc|[minus]|/|[minus]| embryos from line 141. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v392/n6672/images/392186ac.tif.0.gif)
Comparable cross-sections through the atrioventricular canal region of wild-type (a, c, e) and mutant (b, d, f) embryos are shown. Mutant embryos (b) have normal tricuspid and mitral valves. c–f, Evaluation of the endocardial cushions or conotruncal ridges of the outflow tract (oft) of e11.5 wild-type (c, e) and null-mutant (d, f) embryos. Ra, right atrium; rv, right ventricle; la, left atrium; lv, left ventricle; ivs, interventricular septum; mv, mitral valve; tv, tricuspid valve; da, dorsal aorta, nt, neural tube; ms, mesenchyme; m, myocardium; e, endothelial or endocardial lining of the outflow tract lumen.
High resolution image and legend (110K)Figure 4: Developmental, temporal, and spatial expression of NF-ATc.
a, Low-power, double-immunofluorescent, confocal photomicrograph of a sagittal section through the developing heart of an e10.5 wild-type embryo shows accentuation of NF-ATc expression (red, Cy5-labelled secondary antibody) within the outflow tract (oft) and atrioventricular canal region (avc) of the heart. Colocalization of NF-ATc expression with PECAM-1 (green, FITC-labelled secondary antibody) delineated a restriction of NF-ATc expression to the endothelium or endocardium (e) of the oft and avc. Attenuated expression was also seen within the endocardium of the ventricle (v) and atrium (a). Lvr, endothelium of the liver; da, dorsal aorta; pha, pharyngeal arch arteries; m, myocardium. b, A high-power magnification of the endocardium from the avc. NF-ATc expression was not detected in the endothelial cells that had undergone mesenchymal transformation and had migrated into the endocardial cushions (indicated by arrows). c, Similar photomicrographs through the oft of a null-mutant e11.5 embryo. d, e, Cyclosporin A (CsA) blocks nuclear translocation of NF-ATc in situ. High-resolution, immunofluorescent confocal photomicrographs of cross-sections through the outflow tract of e10.0 control embryos (d) and embryos cultured in the presence of 50 
g ml-1 of CsA for 20 h (e) are shown. NF-ATc expression is detected by a Cy5-conjugated secondary antibody (red). All cell nuclei are labelled with DAPI (blue). f, An e8.5 embryo showing NF-ATc expression restricted to the endocardium of the outflow tract, ventricle, and sinus venosus (sv; future right atrium). g, A high-power magnification of the ventricular endocardium reveals the localization of NF-ATc expression in the nucleus of the endothelial cells.
Valve formation is initiated when endothelial cells within the outflow tract and atrioventricular junction of the heart assume a mesenchymal phenotype and migrate into regional swellings of the extracellular matrix that are called the endocardial cushions8. During this process, endothelial cells downregulate expression of the endothelial-cell-specific marker platelet/endothelial cell adhesion molecule-1 (PECAM-1 or CD31) (ref. 9). We studied the histology of the outflow-tract endocardial cushions (conotruncal ridges), which are the anlage of the aortic and pulmonary valves, of the e11.5 embryos to determine whether this process occurred (Fig.3c–f). There were no qualitative differences between wild-type and -/- embryos in endothelial transformation, migration, or proliferation at this critical developmental stage. Moreover, there was normal attenuation of endothelial cell expression of PECAM-1 and initiation of expression of a4 protein in mesenchymal cells in the mutant embryos (results not shown). These results are consistent with normal endothelial activation and transformation into mesenchyme.
Whole-mount immunohistochemistry and immunohistochemical analysis of serially sectioned embryos showed no expression of NF-ATc outside the heart, except in the thymus (results not shown). To define the temporal and tissue-specific distribution of NF-ATc during cardiovascular development, monoclonal antibodies against PECAM-1 and NF-ATc were used for double immunofluorescent labelling of wild-type and mutant embryos during critical periods of valve formation. Co-immunolocalization of PECAM-1 and NF-ATc showed conclusively that NF-ATc expression is restricted to the endothelium of the heart or endocardium (Fig. 4a–c). NF-ATc was expressed at high levels in the endocardium of the outflow tract and atrioventricular canal region and at much lower levels in the endocardium lining the ventricular trabeculae. No expression of NF-ATc was detected in those endothelial cells that had transformed into mesenchyme and begun invasion of the extracellular matrix (Fig. 4a, b), indicating that NF-ATc may not be involved in maintenance of the mesenchymal phenotype. Furthermore, although PECAM expression is limited to the cell membrane (as seen by green fluorescence), NF-ATc expression is clearly nuclear (as shown by red fluorescence), suggesting that NF-ATc is in its active, dephosphorylated form. NF-ATc was not detected in endothelium outside the heart, as shown by the lack of fluorescence in the pharyngeal arch arteries, dorsal aorta and liver. There was no detectable NF-ATc protein in e11.5 141 and 112 -/- embryos (Fig. 4c), demonstrating a null mutation of NF-ATc in these embryos.
The timing of expression of NF-ATc in the heart was also very restricted. Reverse transcription with polymerase chain reaction (RT-PCR) using NF-ATc-specific primers yielded products in the e7.5 embryo (the period of initial endocardial and myocardial differentiation9,10), and at e8.5 and e12.5 (not shown). Immunohistochemistry showed prominent intranuclear NF-ATc expression that was restricted to the straight heart tube of the e8.5 embryo (Fig. 4f, g), suggesting an active role for NF-ATc at least 24 h before the first evidence of valve formation. This burst of NF-ATc expression was temporary, as, by e13.5, NF-ATc protein could not be detected by RT-PCR or by immunohistochemistry (not shown). These results support a unique temporally and tissue-restricted role for NF-ATc in the signal-transduction processes required for normal cardiac valve formation and ventricular septation.
The intranuclear location of NF-ATc in endothelial cells is consistent with the previously demonstrated regulation of NF-ATc activity in response to calcium fluxes in lymphocytes, where sustained increases in intracellular calcium levels result in activation of the calcium- and calmodulin-dependent phosphatase calcineurin. This protein is needed for the dephosphorylation and nuclear translocation of NF-AT6. Calcineurin is a target for the immunosuppressive drug cyclosporin A. To test whether this signalling pathway is present in cardiac endothelial cells, we collected wild-type embryos at e8.5, cultured them for 12 h in media, and then added cyclosporin A at 50 g ml-1. Embryos were cultured for an extra 5 or 20 h and three-channel confocal immunofluorescent microscopy was used to identify NF-ATc (monoclonal antibody 7A6, red fluorescence), PECAM-1 (monoclonal antibody 390, green fluorescence) and a nuclear stain (blue fluorescence). These studies (Fig. 4d, e) showed that embryos treated with cyclosporin A do not show nuclear localization of NF-ATc in cardiac endothelial cells. In untreated cultured embryos (Fig. 3d) and wild-type embryos in vivo (Fig. 4a, b), NF-ATc protein is primarily nuclear (pink staining), whereas in embryos treated with cyclosporin A most NF-ATc protein is cytoplasmic (Fig. 4e). As in NF-ATc mutants, inhibition of nuclear translocation of NF-ATc by cyclosporin A did not prevent initial endothelial cell activation, transformation to a mesenchymal phenotype (as shown by loss of PECAM-1 expression and migration into the extracellular matrix (Fig. 4d, e).
The heart and cardiovascular system are particularly sensitive to gene perturbations, with defects often leading to embryonic death7,11, 12, 13. Cardiovascular system defects, most of which involve abnormal development of the cardiac valves and membranous septum14, occur in nearly 1% of newborns. However, null mutations in mice most often affect the formation of the compact ventricular myocardium15,16 and less frequently affect the trabecular myocardium17, 18, 19 and epicardium of the heart15,20. The phenotype seen here is similar to that seen in Sox-4-deficient mice21. However, the spectrum of defects in the Sox-4-/- mice suggests an abnormality in cranial neural crest function during early heart morphogenesis, an abnormality which is unlikely in NF-ATc-/- mice because of their circumscribed defect in semilunar valve morphogenesis. Nevertheless, the regional and temporal distribution of Sox-4 and NF-ATc and their obvious importance in outflow tract ontogeny justifies further investigation of potential interactions between these proteins.
Extensive semilunar valve insufficiency can cause in utero death in human embryos22. Thymic hypoplasia is also seen22, supporting the expression of NF-ATc in the embryonic thymus. These results may explain the significant increase in fetal death seen when cyclosporin A is administered to pregnant rats between e8 and e14 (ref. 23). Indeed, normal embryos cultured with cyclosporin A develop congestive heart failure (our unpublished observations). Examination of non-viable human embryos for NF-ATc gene mutations and caution in the use of cyclosporin A during pregnancy seem warranted.
Similar mechanisms are thought to regulate the formation of the semilunar and atrioventricular valves8. However, the selective defects in the semilunar valves seen here imply the existence of some essential mechanisms that are unique to aortic and pulmonary valve formation. There is differential gene expression in the outflow and inflow portions of the heart24 and use of an insertion mutation has identified a locus on mouse chromosome 13 that is required for normal conotruncal development25. However, all of these genes are primarily expressed by the myocardium. Our results are the first to indicate distinct, endothelial-dependent, molecular pathways controlling semilunar valve formation. This conclusion concurs with the clinical spectrum of congenital heart disease: where it is rare to detect simultaneous defects in atrioventricular and semilunar valve formation26.
The restriction of NF-ATc expression to the endothelial lining of the heart is unique in development, as NF-ATc is the only endocardial-specific, and only the second endothelial-specific27, transcription factor to be described. Selective endothelial proliferation on the arterial face of the valve cusps28, leading to semilunar valve formation, may result from an 'inductive' interaction between the endocardium and the underlying mesenchyme of the endocardial cushions. In an in vitro model system designed to observe the invasion of atrioventricular canal endothelium into a collagen gel8,29, endothelial 'transformation' was shown to depend upon increases in levels of intracellular free calcium30. One possible explanation for the phenotype seen might be that NF-ATc mediates calcium-dependent initiation of transformation. However, this is not the case as NF-ATc-/- endothelial cells delaminated from the endocardial layer and populated the endocardial-cushion extracellular matrix (Figs 3, 4). Thus, NF-ATc must be involved in events that are essential for valve formation after, and independently of endothelial transformation. These events do not involve the production of transforming growth factor 
proteins 1–3, which are expressed at normal levels in NF-ATc-/- heart (our unpublished observations). Future experiments will focus on identifying the upstream targets and downstream effectors of NF-ATc.
Methods
Disruption of the NF-ATc gene. An NF-ATc genomic clone was isolated by screening a 129/sv genomic library with a complementary DNA fragment corresponding to the NF-ATc Rel-homology domain. A 2-kilobase (kb) fragment from this clone was replaced with a thymidine kinase (TK) promoter neo-poly(A) cassette from pMC1neopoly(A) (Stratagene) by transferring a PstI–NdeI fragment containing 6 kb of 5' NF-ATc sequence, the neo cassette, and 2 kb of 3' NF-ATc sequence to a plasmid encoding the TK drug-resistance gene. ES cell clones transfected with the targeting construct were selected for drug resistance in 180 g ml-1 G418 and 2 M gancyclovir. Homologous recombinants were identified by digesting genomic DNA from drug-resistant clones with BamH1 and hybridizing with the 0.8 kb Asp 718/BamH1 3' probe. Offspring were genotyped by Southern blot analysis of tail or yolk-sac DNA. Two different primers, indicated in Fig. 1, were used to amplify a 275-base-pair (bp) fragment from the wild-type allele or a 350-bp fragment from the line 141 mutant allele.
Histological analysis and immunohistochemistry. Whole embryos were fixed in buffered formalin or 4% paraformaldehyde and embedded in paraffin. Sagittal and transverse sections were cut, and stained with hematoxylin–eosin. For immunohistochemistry, embryos were snap-frozen in optimal cutting temperature (OCT) compound (Miles Laboratories); sections (of 5–6 m) were thaw-mounted on 0.05% poly-L-lysine-coated slides and post-fixed with 1% paraformaldehyde. Tissue sections were pretreated with 2.5% goat serum and incubated overnight at 4 °C with purified tissue culture supernatant of a rat monoclonal antibody against murine PECAM-1 at 25 g ml-1 in Tris-buffered saline (TBS) and a monoclonal antibody ascites (7A6; a gift from G. Crabtree) specific for the amino terminus of NF-ATc at 1:250 dilution. A control isotype antibody and no primary antibody were used as controls. Sections were then rinsed in TBS and incubated for 1 h in a mixture of fluorescein isothiocyanate (FITC)-conjugated goat secondary antibody against a rat protein (1:200 dilution) preabsorbed against mouse antigens, and a biotin-conjugated goat secondary antibody against mouse IgG1 (1:200 dilution, Southern Biotechnical). Sections were rinsed in TBS, incubated with Cy5-conjugated streptavidin (Jackson ImmunoResearch) for 15 min, washed and mounted in vectashield (Vector) antiquench media with DAPI at a concentration of 0.75 g ml-1.
Confocal microscopy. Immunostained specimens were optically sectioned using a computer-interfaced, laser scanning microscope (Leica TCS 4D). This equipment is fitted with a 488 nm/568 nm/647 nm krypton–argon laser, allowing, simultaneous analysis of fluorescein and rhodamine chromophores, with the option of rescanning for Cy5 fluorescence (in the presence of a 600-nm band-pass filter) and ultraviolet-excitable fluorochromes. Images were acquired at a resolution of 1,024 
1,024 pixels and compiled on an image processing (Os9) workstation; subsequent figures were composed on a Macintosh 7600.120 PowerPC using Photoshop 4.0 (Adobe).
Whole-embryo culture. Postimplantation mouse embryos were cultured as described31. Briefly, embryos collected from ICR mice at e8.5 were dissected free from the decidua and Reichardt's membrane; the yolk sac was left intact and cultured (three embryos per 10 ml media) in 60% serum and 40% Tyrodes at an initial O2 concentration of 20% for 12 h, followed by increasing concentrations of O2 to 40% at 24 h and 90% after 36 h with 5% CO2 and balance N2. Cyclosporin A was added at 50 g ml-1.
4 integrins are essential in placental and cardiac development.
