BMP10 is essential for maintaining cardiac growth during murine cardiogenesis

Cardiac development is a complex and sequential process that is composed of a series of morphogenetic events, such as cardiogenic induction and patterning, cardiomyocyte growth and differentiation, multiple cell lineage specification, and terminal differentiation and chamber maturation (for a review, see Srivastava and Olson,2000). These events are regulated by cardiogenic transcriptional factors and growth/differentiation factors. Genetic mutations of these genes lead to either early embryonic lethality or a vast majority of congenital heart defects (CHDs) that affect 1/200 live births. One of many forms of CHD is ventricular noncompaction, or spongy myocardium, which is based on its characteristic feature of abnormal ventricular myocardium with hypertrabeculation and noncompaction(Dusek et al., 1975; Ichida et al., 2001). It is a unique type of cardiomyopathy generally found in infant patients. These patients may proceed to have dilated hearts and congestive heart failure. Commonly associated defects are ventricular septal defect (VSD) and pulmonic stenosis (PS). This noncompaction of ventricular myocardium has been defined as a morphogenetic abnormality and is clearly due to defects in the ventricular trabeculation and compaction.

Ventricular trabeculation and compaction are important morphogenetic processes and are closely associated with cardiac growth regulation at mid-gestation (Rumyantsev,1991; Icardo,1984). At the final stage of cardiac looping (E9.0-E9.5),initially the endocardium penetrates through the cardiac jelly and evaginates at discrete points of myocardium to form `outpockets' directed toward the myocardium. These endocardium outpockets initiate the cardiac trabeculation. Further expansion of primitive trabecular myocardium between E9.5 and E13.5,via either myocyte recruitment or proliferation, is an important step in generating matured trabeculae. Later in development (E14.5-E15.5), trabecular myocytes in the developing myocardium undergo `compaction' and gradually become part of compact wall, papillary muscles, interventricular septum and conductive system cells (Moorman and Lamers, 1999), respectively. One of the scientific challenges is to determine the molecular mechanism by which cardiac trabeculation and subsequent compaction are regulated. Several endocardial growth factors required for the development of trabeculae, such as neuregulin(Meyer and Birchmeier, 1995)and its receptors ErbB (Gassmann et al.,1995; Lee et al.,1995), vascular endothelial growth factor (VEGF)(Ferrara et al., 1996), and angiopoietin-1 (Suri et al.,1996), have been identified. Mutant mice deficient in these genes have severe defects in ventricular trabeculation. However, more detailed analysis is still required to determine their precise role in the process of ventricular trabeculation and compaction.

Bone morphogenetic proteins (BMPs), named for their initial biological activity of inducing ectopic bone formation, belong to the transforming growth factor β (TGFβ) superfamily. They mediate a diverse spectrum of developmental events throughout evolution in species ranging between insects and mammals (for a review, see Ducy and Karsenty, 2000; Nakayama et al., 2000). BMP signals have been shown to link multiple steps of cardiac development, including cardiogenic induction and endocardial cushion formation (for a review, see Schneider et al., 2003). Unlike other BMPs, BMP10 expression is restricted to the developing and postnatal heart(Neuhaus et al., 1999). The most interesting feature of BMP10 is its transient presence in the developing trabecular myocardium. In this study, we found that BMP10 was upregulated in trabecular myocardium of genetically manipulated mutant mice deficient in FK506 binding protein 12 (FKBP12) (Shou et al., 1998). FKBP12 is able to bind type I receptors for BMP/Activin/TGFβ and possibly plays a role in preventing premature activation of type I receptors (Wang et al., 1994; Wang et al.,1996) (for a review, see Massague and Chen, 2000). FKBP12-defcient mice are embryonically lethal due to enormous overproduction of ventricular trabeculae, which severely impairs cardiac development and function. In this study, we generated BMP10-deficient mice and analyzed the biological function of BMP10 during cardiac trabeculation and uncovered a BMP10-mediated molecular pathway essential for regulating cardiac growth and function at mid-gestation.

The PCR-based RNA differential display was based on the previously described protocol (Bao et al.,1998). Briefly, cellular mRNA samples were isolated from pooled FKBP12-deficient and littermate control E14.5 embryonic cardiac tissues,respectively. S³⁵-labeled cDNA fragments were amplified by PCR using a set of palindromic primers and rTh DNA polymerase and Mn²⁺and Mg²⁺ reaction buffer system (Perkin Elmer Cetus). The differentially expressed gene products revealed by the mRNA display were confirmed by Northern blot before being processed for sequencing. The mouse BMP10 full-length cDNA was obtained by RT-PCR and was confirmed by sequencing.

MicroPoly(A)Pure kit (Ambion) was used to isolate mRNA from mouse embryos or pooled embryonic hearts. The first-strand synthesis and PCR reaction were performed using RETROscript kit (Ambion) according to the manufacturer's instructions. The sets of primers are: 5′ BMP10 primer,ACCAGACGTTGGCAAAAGTCAGGC; 3′ BMP10 primer, GATGATCCAGGAGTCCCACCCAAT;5′ p57^kip2 primer, AGTCTGTGCCCGCCTTCTAC; 3′p57^kip2 primer, CTCAGTTCCCAGCTCATCACCC; 5′ FKBP12 primer,CACGTGGATCTGCCATGGAGGAA; 3′ FKBP12 primer, GTGGAAGGACTGACAGAAGCCAA;5′ GAPDH primer, GGGTGGAGCCAAACGGGTC; 3′ GAPDH primer,GGAGTTGCTGTTGAAGTCGCA.

A BMP10 genomic clone was isolated from a mouse 129SvEv genomic BAC library(RPCI-22 129 mouse library from BAC/PAC Resources, Children's Hospital Oakland). The mouse BMP10 gene contains two exons(Fig. 3A). Linearized targeting vector (25 μg) was electroporated into embryonic stem (ES) cells (CCE916 ES cell line), clones were selected in G418 and gancyclovir, DNA from the clones was analyzed by Southern blot, and targeted ES cell lines BMP10-B12 and BMP10-F8 were expanded and injected into blastocysts. Male chimeras were bred to C57BL/6J or 129SvEv females to generate F1 offspring. Mutant mice generated from both targeted ES cell lines had identical phenotype.

Embryos were harvested by cesarean section. Embryos and isolated hearts were fixed in 10% neutral buffered formalin, paraffin embedded, and sectioned(6 μm), and stained with hematoxylin and eosin. To analyze the proliferative activity of the developing heart, timed-mated females were given a single injection of tritiated thymidine (200 μCi ip at 28 Ci/mM; Amersham Biosciences Corp.). Embryos were harvested after a 3-hour labeling period, and followed by fixation (10% neutral buffered Formalin) and paraffin sectioning. Deparaffined sections were stained with Hoechst in PBS to identify the cell nucleus. The Hoechst-stained slides were coated with photographic emulsion(Polysciences, Warrington, PA, USA) and further processed for autoradiography.[³H]thymidine labeling index was the percent of labeled nuclei versus the total number of nuclei. Ink-injection analysis was performed as previously described (Winnier et al.,1999). Both whole-mount and section in-situ hybridization were performed as previously described (Franco et al., 2001). Complementary RNA probes of various cardiac markers were labeled with digoxingenin-UTP using Roche DIG RNA Labeling system according to the manufacturer's guidelines. HOP probe was a generous gift of Dr Epstein (Chen et al., 2002). These probes were hybridized with paraformaldehyde-fixed embryos or hearts for whole-mount in-situ staining or with either frozen or deparaffined sectioned samples for in-situ staining. To analyze p57^kip2 expression in the heart, we used a monoclonal antibody against mouse p57^kip2(Labvision, ms-897-P1) and a Vector staining system (Vector, PK-2200)according to the manufacturer's instructions.

Embryos were washed three times in PBS and fixed for 10 minutes in pre-chilled acetone before being treated with blocking solution containing 3%non-fat dry milk (Bio-Rad Laboratories, Hercules, CA) and 0.025% Triton X-100 for 1 hour. Directly conjugated primary antibodies were then added to a final concentration of 1 μg/ml for 12 to 18 hours at 4°C. Anti-Flk-1(PharMingen, San Diego, CA) and MF-20 monoclonal antibodies (Hybridoma bank,University of Iowa) were labeled with Alexa Fluor 488 and Alexa Fluor 647,respectively, using a monoclonal antibody labeling kit (Molecular Probes,Eugene, OR). Samples were analyzed using a Bio-Rad MRC 1024 Laser Scanning Confocal Microscope (Bio-Rad Microscopy Division, Cambridge, MA) equipped with a Krypton-Argon laser (488, 647 nm). 3D series (Z series) were obtained by imaging serial confocal planes at 512×512 pixel resolution with a Nikon 20X oil-immersion objective (2 μm intervals).

The coding region of the mouse BMP10 cDNA was subcloned into 5′ of an IRES-EGFP cassette in the retroviral vector MIEG3. Ecotropic packaging cell lines were established for MIEG3-BMP10 and MIEG3 vector control as described(Haneline et al., 2003). NIH3T3 cells were transduced with retroviral supernatants (containing approximately 1×10⁵ viral particles/ml) in a similar way to previously published methods (Haneline et al., 2003). Transduced cells were kept in DMEM containing 10% FBS for 3 days before sorting for EGFP positive cells using FACSVantage SE (Becton Dickinson). Northern blot analysis confirmed that the BMP10 transcript was detected only in NIH3T3 cells transduced with MIEG3-BMP10 (NIH3T3/BMP10) and not in cells transduced with vector control (NIH3T3/EGFP).

NIH3T3/BMP10 and NIH3T3/EGFP cells were used as feeders in our cardiomyocyte co-culture assay. We first treated feeder cells with mitomycin C(10 μg/ml) for 2 hours to prevent proliferation. After washing twice in PBS, mitomycin C-treated cells were trypsinized and re-suspended in DMEM containing 2% FBS and were mixed with freshly isolated mouse embryonic cardiomyocytes (E12.5). Cells were plated on one-well Lab-Tek chamber slides. The final plating density was 2-3×10⁵ total cells for each well. Cells were labeled with [³H]thymidine for 2 hours and fixed in methanol and processed for autoradiography and Periodic Acid-Schiff (PAS)staining. The DNA-[³H]thymidine labeling index of PAS positive cells (cardiomyocytes) was recorded and compared between experimental and control groups.

Based on the method previously described(Conway et al., 1997; Rentschler et al., 2002),embryonic hearts were dissected from E9.25 or E9.5 embryos in DMEM containing 10% FBS, and washed twice in PBS and cultured for 8, 12 or 24 hours in BMP10-conditioned or control media. BMP10-conditioned media were collected from overnight culture of BMP10 expressing cells (NIH3T3/BMP10). Control media were collected from overnight culture of control cells (NIH3T3/EGFP). Both BMP10-conditioned and control media were derived from DMEM containing 1% FBS. Conditioned media for BMP-2, -4, -5, -6 and TGFβ-1 (final concentration:50 ng/ml) and neuregulin (NRG-1, final concentration:2.5×10^–9 M) were prepared by adding each activated growth factors to DMEM containing 1% serum right before the culture to reach the final concentration (Barron et al.,2000; Rentschler et al.,2002). BMP-2, -4, -5, -6 and TGFβ-1 were from Sigma. NRG-1 was from R & D Systems.

FKBP12-deficient mice develop severe cardiac defects with the unique phenotype of hypertrabeculation and noncompaction(Shou et al., 1998),suggesting that the FKBP12-deficient mouse is a valuable mouse model to study the molecular mechanism of cardiac trabeculation. To screen for candidate genes that were abnormally expressed in FKBP12-deficient hearts, specifically in overproduced trabecular myocardium, we used RNA differential display. RNAs were isolated from pooled E14.5 hearts of FKBP12-deficient and control embryos. RNA differential displays (data not shown) were performed as previously described (Bao et al.,1998). BMP10 was identified in this gene profiling study. BMP10 expression was significantly elevated in FKBP12-deficient hearts, as confirmed by northern blot analysis (Fig. 1A, part k).

We carefully analyzed the expression pattern of BMP10 in developing and adult mouse hearts using whole-mount and section in-situ hybridization analyses (Fig. 1A, parts a-i). During cardiac development, BMP10 is expressed transiently in the ventricular trabecular myocardium from E9.0-13.5, a critical time span when cardiac development shifts from patterning to growth and chamber maturation(Fig. 1A, parts b,c,e-h). Using RT-PCR, we could detect BMP10 expression as early as E8.75, but not in E16.5 ventricles (Fig. 1A, part j). By E16.5-E18.5, BMP10 was only detectable in atria(Fig. 1A, parts d,j). This finding is consistent with previously published work(Neuhaus et al., 1999). Interestingly, in adult hearts BMP10 was present in the right atrium but not in the left atrium (Fig. 1A,part i), which was further confirmed by a quantitative RNA analysis using Clontech adult tissue mRNA array (Fig. 1B).

The mouse BMP10 gene contains only two exons. The second exon encodes a proteolytic processing site and the mature peptide. To generate a null mutation in the mouse BMP10 gene, we deleted most of the coding region(bmp10^m1) using mouse embryonic stem cell technology(Fig. 2A). Genomic Southern blot using 3′-probe (Fig. 2B) and PCR analyses were used to genotype targeted ES cells and mutant mice. RT-PCR was used to confirm that BMP10 expression was absent in BMP10-deficient embryos (Fig. 2C).

Heterozygous (bmp10^m1/+) mice were viable and fertile and were intercrossed to obtain homozygous(bmp10^m1/bmp10^m1) mutants. Genotyping analysis of 191 F₂ 129SvEv inbred and 230 F₂C57BL6/129SvEv hybrid offspring at weaning demonstrated no viable BMP10-deficient mice. The distribution of genotypes in E9.5 embryos (105 wild-type [24%], 244 heterozygotes [56%], and 87 homozygotes [20%]) and in E10.5 embryos (95 wild-type [30%], 190 heterozygotes [61%], and 28 homozygotes[9%]) suggested that BMP10-deficient mice were dying in utero between E9.5 and E10.5.

We carried out morphological and histological analyses on embryos from E8.5 to E10.5 (Fig. 3A). At E8.5-8.75 (8-12 somite pair stage), BMP10-deficient embryos appeared normal compared to wild-type and heterozygous littermate controls, suggesting that BMP10 was not required for cardiac patterning(Fig. 3A, parts a-d). At E9.0-E9.5 (15-20 somite pair stage), while having normal pairs of somites in mutant embryos, normal allantoic/umbilical connection, and normal vasculature development in mutant yolk sacs and embryos, cardiogenesis appeared arrested in BMP10-deficient embryos (Fig. 3A, parts e-p). These mutant embryos displayed cardiac dysgenesis with profound hypoplastic ventricular walls and absence of ventricular trabeculae. The development of endocardial cushions was abnormal in both outflow track (OFT) and atrial-ventricular canal (AVC) and was halted at the acellular stage. At this developmental stage, BMP10-deficient hearts exhibit rhythmic contraction, but at a significantly slower rate compared with littermate controls [43±6 beats/minute in mutant hearts (n=12)versus 96±7 beats/minute in wild-type and heterozygous littermate controls (n=15), P<0.0001]. To visualize and compare the cardiac function and circulation in mutants and controls, ink was injected into the primitive left ventricle of E9.0 and E9.5-9.75 embryos. As shown in Fig. 3B, while circulation was established in BMP10 mutant embryos at E9.0, severely impaired cardiac function and circulation was observed in mutant embryos at E9.5-9.75. By E10.0-E10.5, mutant embryos appeared to be dying(Fig. 3A, parts q-t).

It has been suggested that cardiac trabeculation is the consequence of interactions between the developing myocardium and the `outpocketing'endocardium at E9.0-E9.5 (Icardo,1984). To better understand if the lack of ventricular trabeculae in the BMP10-deficient heart was due to the failure of endocardial outpocketing (i.e. initiation of trabeculation) or the failure of a myocardial response (i.e. further expansion of trabecular myocardium through either myocyte recruitment or growth), we used confocal microscopy to analyze the structural relationship between endocardium and myocardium in BMP10-deficient hearts at E9.5 (Fig. 4). Endothelial receptors for vascular endothelial growth factor (Flk1)(Millauer et al., 1993) and angiopoietin (Tek/Tie2) (Dumont et al.,1994) (data not shown) were normal in BMP10-deficient endothelial cells throughout the developing cardiovasculature compared with wild-type and heterozygous littermate controls at E9.5(Fig. 4A,B). MF20 staining(anti-myosin heavy chain) was also maintained in the BMP10-deficient myocardium despite overall hypoplasia (Fig. 4C,D). These observations confirmed that BMP10 was not required for differentiation of either endocardium or myocardium. Furthermore, the developing endocardium was in normal proximity to the ventricular wall and the primitive ventricular trabeculae were indeed formed in the BMP10 mutants(Fig. 4C,D). These findings strongly suggested that BMP10 was not crucial to the initiation of cardiac trabeculation, nor myocyte recruitment, but was essential to the further growth of both ventricular wall and trabeculae.

Surprisingly, TUNEL assays showed no significant cellular apoptosis in E9.5 BMP10 mutant hearts, despite a severely thinned ventricular wall (data not shown). This finding suggested that a defect in myocyte proliferation was the main reason for the hypoplastic wall seen in BMP10-deficient hearts. We used[³H]thymidine labeling to determine the proliferative activity in BMP10-deficient hearts. [³H]thymidine labeling index in the mutant hearts was significantly lower than that in littermate controls(10.1%±2, n=5, in mutants versus 29.0%±3, n=4,in controls, P<0.001). These results further suggested that a marked reduction in proliferation of BMP10-deficient cardiomyocytes was responsible for the hypoplastic ventricular wall and failure of normal ventricular trabeculation.

Among multiple cell cycle regulators, p57^kip2 expression is first detectable in the developing heart at E10.5 by in-situ hybridization and RT-PCR and is restricted to ventricular trabeculae(Kochilas et al., 1999). Therefore, p57^kip2 is considered a key negative regulator involved in cardiac cell cycle exit within the developing ventricular trabeculae during chamber maturation (Kochilas et al.,1999). Immunohistochemistry staining revealed that p57^kip2 was upregulated and ectopically expressed throughout the ventricular wall in BMP10-deficient hearts at E9.5 compared with littermate controls (Fig. 5A, parts a-c). FKBP12-deficient hearts (E13.5), which have elevated BMP10 levels and an overproduction of trabeculae, exhibited significantly lower p57^kip2expression in trabecular myocardia compared with littermate controls(Fig. 5A, parts d,e),suggesting a mechanistic relationship between p57^kip2 and BMP10. RT-PCR analysis confirmed this observation(Fig. 5B) and further suggested that BMP10 regulated p57^kip2 at the transcriptional level. Together these data show that elimination or elevation of BMP10 expression jeopardized the regulation of ventricular growth and chamber maturation.

To determine if cardiogenic pathways were affected by the loss of BMP10, we evaluated the expression patterns of key cardiac transcriptional factors and chamber-restricted markers. The expression of cardiac chamber-specific markers MLC2v and MLC2a (Franco and Icardo,2001) were normal in BMP10-deficient hearts at E8.5 (data not shown) and E9.5 (Fig. 6A-D),which further confirmed our observation that BMP10 was not required for cardiac patterning and cardiogenic differentiation. While the expression of cardiogenic transcriptional factors NKX2.5(Schott et al., 1998) (for a review, see Harvey et al.,1999) and MEF2C (for a review, see Black and Olson, 1999) was normal in BMP10-deficient embryos at E8.5-E8.75 (data not shown), these molecules were dramatically downregulated in the BMP10-deficient hearts at E9.25-E9.5 (Fig. 6E-H). Whole-mount insitu hybridization revealed that the downregulation of NKX2.5 was restricted to the heart, as NKX2.5 expression in the branchial arches and developing gut remained normal at E9.5 (data not shown). ANF and Chisel are known downstream targets of NKX2.5(Durocher et al., 1996; Palmer et al., 2001). Both Chisel and ANF were significantly downregulated(Fig. 6I-L). Another NKX2.5 downstream target, HOP (Chen et al.,2002; Shin et al.,2002), which is a novel homeodomain protein that serves a negative feedback for NKX2.5 function, remained normal in the mutant hearts at E9.5(Fig. 6M,N). In addition,cardiogenic transcriptional factor eHand also remained normal in BMP10-deficient hearts, while dHand was slightly downregulated (data not shown). These finds suggested that NKX2.5 and MEF2C were regulated by BMP10.

To further analyze the biological activity of BMP10, we generated BMP10-expressing NIH3T3 cells using a retroviral gene delivery system(Fig. 7A, part a). The expression of BMP10 in these cells was confirmed by northern blot analysis(data not shown). Using BMP10-expressing cells as feeders, we established a co-culture system in which freshly isolated wild-type embryonic cardiomyocytes were laid on top of mitomycin C pre-treated (non-mitotic) BMP10-expressing cells or control cells. BMP10-expressing feeder cells maintained a higher[³H]thymidine labeling index in embryonic cardiomyocytes after 48 hours of co-culture compared with control co-cultures(Fig. 7A, part b), indicating that BMP10 has growth-promoting activity for embryonic cardiomyocytes. This growth-promoting activity of BMP10 was also confirmed using BMP10-conditioned media in both cell culture and organ culture(Fig. 7B).

When we applied BMP10-conditioned media to isolated BMP10-deficient hearts in culture, we found BMP10-conditioned media rescued BMP10-deficient hearts(Fig. 7B, part b). The rescued BMP10-deficient hearts displayed significant improvements in cardiac size,[³H]thymidine labeling (Fig. 7B, part d, and Fig. 7C), and cardiac performance as measured by the improved heart rate (Fig. 7D). This experiment further delineated the essential role for BMP10 in cardiac development.

Mutant mice deficient in neuregulin or its receptor ErbB developed severe defects in ventricular trabeculation(Meyer and Birchmeier, 1995; Gassmann et al., 1995; Lee et al., 1995). To determine if the strong effect of BMP10 on myocardium was via a neuregulin–ErbB-mediated pathway, we cultured BMP10-deficient hearts in the neuregulin (NRG-1)-containing media(Rentschler et al., 2002). Unlike BMP10-conditioned medium, NRG-1 medium was not able to rescue the growth-deficiency phenotype of BMP10-deficient hearts (data not shown). This observation suggested that BMP10 was either downstream of, or irrelevant to,the neuregulin–ErbB-mediated pathway.

Importantly, when E9.5 mutant hearts were cultured in BMP10-conditioned media, both NKX2.5 (Fig. 8A)and MEF2C (data not shown) expression was restored, and p57^kip2expression was significantly downregulated in these rescued hearts(Fig. 8B). This observation further indicated the close relationship between BMP10 and cardiogenic transcriptional factors NKX2.5 and MEF2C and cell cycle regulator p57^kip2.

In summary, our studies demonstrated that the presence of BMP10 in the developing ventricle at E8.75-E9.0 is crucial in maintaining the proliferative activity of the embryonic cardiomyocytes by preventing premature activation of the negative cell cycle regulator p57^kip2 and in maintaining the required expression level of two very important cardiogenic factors, NKX2.5 and MEF2C, in the developing heart at mid-gestation(Fig. 9).

The importance of BMP10 to cardiac development is reflected by its temporal and spatial pattern of expression. It is known that cardiac cell cycle withdraw is associated with ventricular chamber maturation, myocyte terminal differentiation and lineage specification. The fact that trabecular myocytes have lower proliferative activity than that of compact wall suggests that trabecular myocytes initiate the process of terminal differentiation earlier(Rumyantsev, 1991; Icardo, 1984). In this regard,cardiac growth and chamber maturation has to be regulated by a temporally and spatially controlled process and a balance between proliferation and terminal differentiation. During trabeculation (E9.0-13.5 in the mouse), trabecular myocardium is required to expand in order to generate sufficient numbers of myocytes to contribute to future ventricular septum, papillary muscles and conductive system cells; at the same time, these cells are also required to reduce proliferative activity for terminal differentiation and lineage specification. Although it is still not clear about the identities of negative growth signals, p57^kip2, a negative cell cycle regulator, is one of the intracellular candidates and is initially found in trabecular myocardium at E10.5 (Kochilas et al.,1999). Interestingly, cardiac defects have not been reported in p57^kip2-deficient mice (Zhang et al., 1997; Yan et al.,1997), suggesting that there are other negative regulators involved, in a cooperative way with p57^kip2, in the cardiac cell cycle withdraw and terminal differentiation. In fact, there is evidence that p57^kip2 and other negative cell cycle regulators (e.g. p21^cip1 and p27^kip1) cooperate to regulate cell cycle exit and cell differentiation of skeletal muscle cells, lens fiber cells and placenta trophoblasts (Zhang et al.,1998; Zhang et al.,1999). Our data support the suggestion that BMP10 provides a positive growth signal for cardiomyocytes that antagonizes negative regulators such as p57^kip2. Perturbation of this balance would lead to either cardiac hyperplasia and hypertrabeculation and noncompaction (e.g. elevation of BMP10 expression in FKBP12-deficient hearts) or the cardiac hypoplasia seen in BMP10-deficient hearts.

Another important finding in our studies is that BMP10 was specifically required to maintain the expression level of cardiogenic transcription factors NKX2.5 and MEF2C during mid-gestation. Both NKX2.5 and MEF2C have been shown to be critical for cardiac patterning in the early phase of cardiogenesis (for a review, see Harvey et al.,1999; Black and Olson,1999). BMP10-deficient hearts have normal cardiac patterning,which is consistent with the normal expression of NKX2.5 and MEF2C before E9.0. However, the rapid downregulation of NKX2.5 and MEF2C in E9.5 mutant hearts may help to explain the severely impaired cardiac function seen in BMP10-deficient hearts, since many of the NKX2.5 and MEF2C downstream gene products are crucial for cardiac function.

A related observation is that the downregulation of NKX2.5 and MEF2C is not restricted to the ventricular chamber in which BMP10 is expressed. In fact, in a similar way to other BMPs, BMP10 is a secreted peptide that can function as an autocrine and paracrine growth and differentiation signal, which would lead to a broader effect during cardiac development. The lack of seeded mesenchymal cells in endocardial cushions in BMP10-deficient hearts and the relative normal cushion structure in rescued hearts suggest that BMP10 is able to contribute to the epithelial–mesenchymal transformation, a key step in endocardial cushion development. However, we cannot yet exclude the likelihood that the cushion defect in BMP10 mutant heart is secondary to severely impaired cardiac growth in vivo.

In addition to the cardiac-specific BMP10, other BMPs (e.g. BMP-2, -4, -5,-6 and -7) and TGFβ family members and receptors (for a review, see Schneider et al., 2003), which have much wider expression patterns during embryonic development, are also found in the developing heart. These BMPs appear to form an important network that controls several crucial morphogenetic events during cardiac development. Obviously, these BMPs are not able to compensate for the loss of BMP10 in vivo. Probably, this is partially due to their differences in temporal and spatial pattern of expression and partially due to their differences in ligand-receptor specificity. During cardiac development, BMP2/4 is more restricted to the myocardium adjacent to the endocardial cushion region(Nakajima et al., 2000). BMP-5, -6 and -7 have overlapping expression patterns and redundant functions in cardiac development (Solloway and Robertson, 1999; Kim et al.,2001). Mice deficient in both BMP-6 and BMP-7 develop outflow tract and valvo-septation anomalies (Kim et al., 2001), while mice deficient in both BMP-5 and BMP-7 have defects in cardiac looping and ventricular chamber formation(Solloway and Robertson,1999). In our on-going rescue experiments using different BMPs(e.g. BMP-2, -4, -5 and -6, and TGFβ-1), we clearly observed the differences in their ability to rescue the BMP10-deficient hearts in culture(data not shown). This finding suggests a complicated BMP-receptor network in the developing heart. It is consistent with recent data in cardiac conditional knockout of type I receptor for BMP (Alk3)(Gaussin et al., 2002). Cardiac myocyte-specific deletion of Alk3 leads to major defects in endocardial cushion and ventricular wall but has very little effect on cardiac trabeculation. BMP10 expression is also normal in these mutant mice. Interestingly, NKX2.5 and other cardiogenic transcription factors are not altered in Alk3 mutant hearts, suggesting that Alk3 is not the receptor mediating BMP10 signaling.

Recently, analysis of the NKX2.5 promoter showed that NKX2.5 expression in both early and late cardiac development is mediated by multiple conserved Smad binding sites (Liberatore et al.,2002; Lien et al.,2002). This finding further suggests that BMP signaling is required beyond the initial step of cardiogenic induction. Previously, BMP2/4 has been shown to mediate cardiac induction via activation of NKX2.5 in chick embryos at an early stage (Schultheiss et al., 1997) (for a review, see Srivastava and Olson, 2000). Our data supported the observation from NKX2.5 promoter analysis. BMP10 is a strong candidate that is responsible for maintaining the expression of NKX2.5 during mid-gestation, possibly via a Smad-NKX2.5 pathway.

Our studies have demonstrated an important genetic pathway that regulates ventricular trabeculation and chamber maturation. The regulation of BMP10 expression may be an important mechanism for normal cardiac trabecular growth and compaction in mice. Genetic investigation of human patients with noncompaction myocardial defects may demonstrate a role for BMP10 in this pathological condition as well.

A recent related study by Pashmforoush et al.(Pashmforoush et al., 2004)has indicated that Nkx2.5 can act as a negative-feedback regulator of BMP10 expression at later developmental stages. Mutant mice with a late onset cardiomyocyte-specific Nkx2.5 null mutation, generated via a conditional cre/loxP knockout strategy, have an abnormal myocardium with increased ventricular trabeculation. Significantly, BMP10 expression is ectopically upregulated in these conditional Nkx2.5-deficient hearts. These findings further support a role for BMP10 in ventricular maturation and underscore the deleterious effects of abnormal BMP10 regulation during the pathogenesis of ventricular non-compaction in murine models. Further studies will be required to determine the link between BMP10 and the etiology of ventricular non-compaction in humans.

We wish to thank Dr Shaolian Jing and Mr William Carter of Indiana University Mouse Core for their superb assistance in generating BMP10 knockout mice, Dr Loren Field for his generous support and valuable discussion and Drs Michael Rubart-Von Der Lohe, Kishore Pasumarthi and Hisako Nakajima for their generous help in this study. This study was supported in part by National Institute of Health HL70259 (W.S.), HL63196 (M.Y.), HLOK04071 (L.H.), HL60714(S.J.C.), Showalter Research Foundation (W.S. and L.H.), Muscular Dystrophy Association (W.S.), the Riley Children's Foundation (W.S., M.Y., L.H.).