The Short Stature Homeobox 2 (Shox2)-bone Morphogenetic Protein (BMP) Pathway Regulates Dorsal Mesenchymal Protrusion Development and Its Temporary Function as a Pacemaker during Cardiogenesis*

Background: Dorsal mesenchymal protrusion (DMP) is required for cardiac septation, but its additional functions are unknown. Results: Shox2 is required for the nodal-like characteristics and development of embryonic DMP by regulating BMP/Smad4 signaling pathway. Conclusion: A Shox2-BMP genetic cascade regulates DMP development and its temporal pacemaking function. Significance: A novel function of DMP and its developmental regulatory mechanism are identified. The atrioventricular (AV) junction plays a critical role in chamber septation and transmission of cardiac conduction pulses. It consists of structures that develop from embryonic dorsal mesenchymal protrusion (DMP) and the embryonic AV canal. Despite extensive studies on AV junction development, the genetic regulation of DMP development remains poorly understood. In this study we present evidence that Shox2 is expressed in the developing DMP. Intriguingly, this Shox2-expressing domain possesses a pacemaker-specific genetic profile including Hcn4 and Tbx3. This genetic profile leads to nodal-like electrophysiological properties, which is gradually silenced as the AV node becomes matured. Phenotypic analyses of Shox2−/− mice revealed a hypoplastic and defectively differentiated DMP, likely attributed to increased apoptosis, accompanied by dramatically reduced expression of Bmp4 and Hcn4, ectopic activation of Cx40, and an aberrant pattern of action potentials. Interestingly, conditional deletion of Bmp4 or inhibition of BMP signaling by overexpression of Noggin using a Shox2-Cre allele led to a similar DMP hypoplasia and down-regulation of Hcn4, whereas activation of a transgenic Bmp4 allele in Shox2−/− background attenuated DMP defects. Moreover, the lack of Hcn4 expression in the DMP of mice carrying Smad4 conditional deletion and direct binding of pSmad1/5/8 to the Hcn4 regulatory region further confirm the Shox2-BMP genetic cascade in the regulation of DMP development. Our results reveal that Shox2 regulates DMP fate and development by controlling BMP signaling through the Smad-dependent pathway to drive tissue growth and to induce Hcn4 expression and suggest a temporal pacemaking function for the DMP during early cardiogenesis.


The atrioventricular (AV) junction plays a critical role in chamber septation and transmission of cardiac conduction pulses. It consists of structures that develop from embryonic dorsal mesenchymal protrusion (DMP) and the embryonic AV
canal. Despite extensive studies on AV junction development, the genetic regulation of DMP development remains poorly understood. In this study we present evidence that Shox2 is expressed in the developing DMP. Intriguingly, this Shox2-expressing domain possesses a pacemaker-specific genetic profile including Hcn4 and Tbx3. This genetic profile leads to nodallike electrophysiological properties, which is gradually silenced as the AV node becomes matured. Phenotypic analyses of Shox2 ؊/؊ mice revealed a hypoplastic and defectively differentiated DMP, likely attributed to increased apoptosis, accompanied by dramatically reduced expression of Bmp4 and Hcn4, ectopic activation of Cx40, and an aberrant pattern of action potentials. Interestingly, conditional deletion of Bmp4 or inhibition of BMP signaling by overexpression of Noggin using a Shox2-Cre allele led to a similar DMP hypoplasia and downregulation of Hcn4, whereas activation of a transgenic Bmp4 allele in Shox2 ؊/؊ background attenuated DMP defects. Moreover, the lack of Hcn4 expression in the DMP of mice carrying Smad4 conditional deletion and direct binding of pSmad1/5/8 to the Hcn4 regulatory region further confirm the Shox2-BMP genetic cascade in the regulation of DMP development. Our results reveal that Shox2 regulates DMP fate and development by controlling BMP signaling through the Smad-dependent pathway to drive tissue growth and to induce Hcn4 expression and suggest a temporal pacemaking function for the DMP during early cardiogenesis.
The mammalian embryonic heart develops from a tubular structure with a single circulation to a four-chambered structure with a dual circulation. Simultaneously, its beating pattern undergoes transformation from a peristaltic motion to a sophisticated synchronous contraction as a result of maturation of the cardiac conduction system (CCS) 4 that consists of the sinoatrial node (SAN) and atrioventricular node (AVN) as well as His-Purkinje tracts (1). Myocytes of the SAN and AVN have unique electrophysiological features characterized by the slower upstroke and smaller action potential (AP) amplitude and presence of the diastolic depolarization compared with that of working myocardium of the atrium and the ventricle. The AV junction that lies at the merging point of cardiac chambers is critically involved in both septum and CCS developmental processes (2). It comprises the dorsally positioned second heart field-derived dorsal mesenchymal protrusion (DMP) that gives rise to the future vestibular spine (2) and ventrally positioned endocardial cushion tissue as well as AV canal-derived myocardium including AVN precursors (3,4), The outgrowth and fusion of DMP and endocardial cushion tissue are essential for proper formation of AV septation (2,5). During cardiogenesis in mice, in the initial simple heart tube at embryonic day 8.5 (E8.5), the ring-like AV canal, which connects single atrium and ventricle, functions to generate the delay between the atrial and ventricular contraction (6). However, the definitive AVN becomes morphologically distinguishable only at late gestation stage (3,6). The AV conduction network, particularly the AVN in adults, receives propagation initiated by the SAN and gives a * This work was supported, in whole or in part, by National Institutes of Health pause to allow the occurrence of ventricular contraction after atrial contraction, ensuring proper blood flow in the heart (1,2). On the other hand, the second heart field cells expand and invade into the atrial cavity to form the DMP at E10.5. This mesenchymal cell population has been well characterized for its importance in the development of AV septum and pulmonary vein (2,5).
SHOX2 is closely related to the short stature homeobox gene SHOX in humans (7)(8)(9). We and others have reported previously that Shox2 is expressed in the sinus venosus and the SAN in a specific manner in the developing heart of both mice and humans (10 -12). Targeted inactivation of Shox2 in mice has revealed an essential role for Shox2 in the development of multiple organs including the heart (10,11,13,14). Shox2 is shown to be related to CCS disorders including Holt-Oram syndrome in mouse models and Shox2 mutation results in a failed differentiation of the SAN cells and bradycardia (10,11,15). Moreover, Shox2 mutation also impaired Bmp4 expression in the SAN (15). We have recently reported Shox2 expression in the developing AV junction (16), suggesting a potential role for Shox2 in AV junction development.
Bone morphogenetic proteins (BMPs), belonging to the TGF␤ superfamily, bind to BMP receptors to mobilize the canonical Smad-dependent transcriptional activities or to function through non-canonical pathways including mitogenactivated protein kinases as well as PIK3 and RAS (17)(18)(19). Numerous studies have demonstrated the fundamental roles of BMP signaling in multiple cellular processes including proliferation, apoptosis, and differentiation during cardiogenesis (20), but its function in CCS development, particularly in the pacemaker tissues, appears to be under-studied. It was reported that Bmp2 and Bmp4 are expressed in the embryonic AV canal and are required for the specification of AV myocardium (21,22). In addition, Bmp4 is also expressed in the developing SAN overlapping with Shox2 and is a direct target of Shox2 (15). However, although inactivation of Bmp4 in the embryonic heart caused AV septal defects due to severe hypoplasia, no CCS defects were reported (21). In the developing AV junction, it was reported that the BMP-activated Smad dependent pathway induces the expression of working myocardium repressor genes such as Tbx2 to maintain the conduction cell identity (23). Conditional deletion of BMP receptor Alk3 in the second heart field from where the DMP derives resulted in impaired DMP development and AV septal defect (5), indicating an essential role for BMP signaling in DMP development. However, the upstream genetic regulators of BMP signaling and its role in AV conduction axis development remain unknown.
In this study we identified Shox2 expression, overlapping with that of pacemaker-specific genes including Hcn4 and Tbx3 (24 -26), in the early developing DMP that also manifests a nodal-like electrophysiological character, suggesting a potential pacemaking function for the embryonic DMP during cardiogenesis. Using genetic ablation and transgenic rescue approaches, we present evidence that Shox2 plays an essential role in DMP development and differentiation by regulating BMP signaling. Disruption of BMP signaling results in similar DMP defects as that in Shox2 mutants, and transgenic activation of BMP signaling in Shox2 mutant background rescues the DMP phenotype.
We further present evidence demonstrating that BMP signaling acts through a Smad-dependent pathway to regulate Hcn4 expression and an early fate of the DMP.

EXPERIMENTAL PROCEDURES
Animals-The generation of Shox2 ϩ/Ϫ , Shox2-LacZ, Shox2-Cre, pMes-Noggin, and pMes-Bmp4 mice has been described previously (13,16,27,28). Smad4 floxed mice and R26R-LacZ reporter mice were purchased from The Jackson Laboratory. Bmp4 floxed mice (29) were kindly provided by Dr. Brigid Hogan. All described mouse lines were crossed onto CD-1 background. Wild-type CD-1 mice were obtained from Charles River. To harvest Shox2 null embryos at E13.5 or older, pregnant females were administrated with the ␤-adrenergic receptor agonist from day 7.5 of gestation by supplementing drinking water with 200 g/ml isoproterenol. Approval from the Institutional Animal Care and Use Committee of Tulane University was acquired for all the procedures used in this study.
Histology, X-Gal staining, in Situ Hybridization, and Immunohistochemistry-Standard hematoxylin/eosin staining was performed on 10-m sections from paraffin-embedded samples. X-gal staining on cryosections or whole embryo was performed following standard procedures. For in situ hybridization, embryos were harvested in ice-cold diethyl pyrocarbonatetreated PBS and fixed in 4% paraformaldehyde, PBS overnight at 4°C. After dehydration through graded ethanol and paraffin embedding, samples were sectioned at 10 m and subjected to an in situ hybridization procedure as described previously (13). Immunohistochemistry was conducted on frozen-sectioned or paraffin-sectioned samples. For cryostat sectioning, embryos were fixed in Z-fix (Anatech) at room temperature for 2-3 h. After rinsing through 15 and 30% sucrose, PBS, and OCT embedding, samples were cryosectioned at 8 m. A standard immunofluorescence procedure without antigen recovery was conducted. For paraffin sections, embryos were fixed in Z-fix at 4°C for 24 h. After dehydration through graded ethanol and paraffin embedding, samples were sectioned at 8 m. A standard immunofluorescence procedure with antigen recovery was carried out. The following antibodies were used: anti-␤ galactosidase Isolation of Embryonic Cardiac Myocytes and Electrophysiology-The SAN, atrium, and ventricle of embryonic hearts from Shox2 LacZ/ϩ or Shox2 LacZ/LacZ mice at designated ages were dissected in prewarmed Tyrode's solution that contained 140 mM NaCl, 5.0 mM HEPES, 5.5 mM glucose, 5.4 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , with pH adjusted to 7.4 with NaOH, according to the method described previously (30 -33). The isolation of the DMP followed the method reported previously for the dissection of AV junction from neonatal heart (33). Briefly, after removal of atrial wall and entire inflow tract including inferior and superior vena cava, the AV junction was exposed. To validate successful isolation of the DMP, X-gal staining was performed to show Shox2-positive tissue positioned at the top of the AV junction. Subsequently, the anterior outflow tract and most of the ventricle were cut off to obtain the desired AV junction tissue. The dissected heart tissues were transferred quickly to "low Ca 2ϩ " digestion solution (30) containing collagenase type II (200 units/ml, Invitrogen) and excised into smaller pieces. The digestion was performed using a cell culture incubator at 37°C for 30 min. Digested tissue was agitated to gain suspension of the dissociated cardiac myocytes. Cells were plated on collagen-coated coverslips and cultured overnight. Attached ␤-galactosidase positive myocytes were labeled by C 12 FDG substrate (Invitrogen) with a concentration of 33 M for 40 min in culture medium or culture medium containing 10 ng/ml BMP4 (R&D Systems) at 37°C as described previously (34). The fluorescence of labeled cells was identified using a Nikon eclipse Ti-S microscope.
A coverslip with attached myocytes was placed in the measuring chamber, and the fluorescence of ␤-galactosidase-positive cells was verified under an upright microscope (Olympus BX50WI). Cells were recorded in an extracellular Tyrode's solution that contained 140 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 5.5 mM glucose, and 5.0 mM HEPES at pH 7.4. Whole-cell patch recordings were made using the Axon MultiClamp 700B amplifier (Axon Instruments). Glass electrodes (3-5 megaohms) were pulled on a Sutter puller. Internal solution contained 120 mM potassium glutamate, 120 mM KCl, 4 mM NaCl, 10 mM HEPES, 0.2 mM EGTA, 4 mM MgATP, 14 mM phosphocreatinine disodium salt, and mM 0.3 Tris GTP (pH 7.2). Data were digitized at a rate of 10 KHz using Clampex 10.3 and analyzed using Clampfit 10.3 (Axon Instruments). Recordings were done at 36.0 Ϯ 0.5°C. All reagents used in intra-and extracellular solution were purchased from Sigma. The value of the maximum rate of rise of the AP upstroke (dV/ dtmax) was determined using Clampfit 10.3.
Three-dimensional Reconstruction and Volume Rendering-Twenty 10-m X-gal-stained slices from each E12.5 embryo and 30 10-m X-gal-stained slices from each E14.5 embryos of wild-type and Shox2 mutants were loaded into an Amira software (Version 5.4.3, Visualization Sciences Group US) to carry out three-dimensional-reconstruction and volume rendering following procedures described previously (35). Three embryos from each stage and genotype were used for three-dimensional reconstruction. Student's t test was applied to determine the significance of difference.
Terminal Deoxynucleotidyltransferase dUTP Nick End Labeling (TUNEL) Assay-A TUNEL assay was performed to detect apoptosis using the In Situ Cell Death Detection kit from Roche Applied Science following a described method (36). Four samples of each genotype were tested.
Chromatin Immunoprecipitation-About 10 -12 hearts isolated from one litter of E13.5 wild-type mouse embryos were subjected to chromatin immunoprecipitation (ChIP) assay using a kit from Invitrogen. Previously described anti-Smad1/ 5/8 antibodies (37) (Santa Cruz) were used to immunoprecipitate chromatin fragments containing the Smad binding ele-ment (SBE) sequence, with normal rabbit immunoglobulin G that came with the kit used as a control. A set of primers, 5Ј-CTGGTGATGGGCATGTAGTG-3Ј and 5Ј-TAGACT-GCTGAAGGGCATGA-3Ј, covers the described SBE in the first intron of the Hcn4 locus amplified a 291-bp fragment, and a control set of control primers, 5Ј-AGCAACCCTTGGTTGT-GAGT-3Ј and 5Ј-CTGCTATCACCCTCTCAGCA-3Ј, covers a 259-bp DNA sequence 5 kb downstream from the SBE site. Amplified DNA products were confirmed by correct size on 3% agarose gel and sequencing.

RESULTS
Genetic Profile and Electrophysiological Properties of Shox2expressing DMP-To reveal comprehensive Shox2 expression patterns and to dissect genetic network in Shox2-expressing cells during cardiogenesis, we created a Shox2-LacZ allele and a Shox2-Cre allele by gene knock-in strategy (16). Shox2 LacZ/ϩ or Shox2 Cre/ϩ mice are indistinguishable from their wild-type littermates, but homozygous mice die in mid-gestation stage and develop phenotypes identical to that in Shox2 Ϫ/Ϫ mice. Although Shox2 expression in the SANs, venous valves, and sinus horns and other developing organs was confirmed by the LacZ reporter allele, we also identified a novel Shox2 expression site that was not reported previously; that is, the developing AV junction from embryonic day 11.5 (E11.5) up to E18.5, at which time the expression became down-regulated as compared with persistent strong expression in the SAN (Fig. 1, A-C) (16). This Shox2 expression domain in the AV junction was further confirmed by in situ hybridization assays (Fig. 1D) and by X-gal staining performed on R26R-LacZ;Shox2 Cre/ϩ mice (Fig. 1E). Because the AV junction myocardium consists of DMP and AVN, we set out to determine the identity of this Shox2-expressing tissue by cell lineage tracing using the Shox2-Cre allele compounded with the R26R-LacZ reporter and performed an immunohistochemistry assay on the expression of pacemaker markers hyperpolarization-activated, cyclic nucleotide-gated 4 (Hcn4) and Tbx3 as well as myocardium marker Cx40. At E18.5, when a definite AVN becomes distinguishable (3) and a residual Hcn4 expression was found in the LacZ-positive domain, Tbx3 expression was not detectable (Fig. 1, G-I). However, adjacent to the LacZ-positive domain on the ventral side, there was a domain positive for both Hcn4 and Tbx3, representing the maturing AVN ( Fig. 1, G-I). At this stage, Cx40 expression was also absent in both domains despite its expression in the adjacent atrial septal tissue (Fig. 1J). At postnatal day 7 (P7), the expression of Hcn4 and Tbx3 was absent within the LacZpositive domain but was persistent in the AVN (Fig. 1, K-M). At this time, consistent with the previous report (38), Cx40 expression was found in the LacZ-positive domain but was completely absent in the AVN (Fig. 1N), suggesting that the Shox2-expressing cells have differentiated into working myocardium and the AVN has become matured at this stage. These observations indicate that the Shox2-expressing cells are of DMP origin, which is further confirmed by Shox2 expression in the second heart field derivative in the AV junction (Fig. 1F). Because Hcn4 is expressed in the early developing DMP, we wondered if the Shox2-expressing DMP possesses the pacemaking genetic profile during early cardiogenesis. We conducted double immuno-staining assay on ␤-Gal and Hcn4 in the Shox2 LacZ/ϩ DMP. Indeed, Hcn4 expression overlapped with that of ␤-Gal in the AV junction ( Fig. 1O and see Fig. 5, A-C). In addition, the expression of Tbx3, the lack of working myocardium marker Cx40 and Cx43 as well as the lack of endocardium marker Tie-2 demonstrated that the Shox2-expressing DMP possesses the genetic feature of the cardiac pacemaker at early developmental stage ( Fig. 1, P-R; and data not shown).
To identify the physiological properties of the Shox2-expressing DMP myocytes, we used the whole-cell patch clamp technique to characterize the AP properties of the Shox2-posi-tive cells in the DMP from the embryonic hearts. First, recordings were performed on the myocytes from the SA node, the atrium, and the ventricle of E14.5 embryos to establish distinct AP patterns. The typical firing pattern of each measured cell group is shown in Fig. 2. These correspond to similar characteristics described in numerous previous studies (39 -41). In brief, the nodal-like APs of SAN cells show the slow upstroke (dV/dtmax), the small AP amplitude, and the presence of the diastolic depolarization ( Fig. 2A). However, the APs of contracting atrial and ventricular myocardium show fast upstroke, large AP amplitude, and weak or no diastolic depolarization  . F, a two-dimensional diagram showing the action potential amplitude and the maximum rate of rise of the AP (dV/dtmax) for each cell type studied. In E14.5 SAN and DMP cells, action potential amplitudes were smaller than 75, and dV/dtmaxs were smaller than 4. In E14.5 atrial and ventricular cells, action potential amplitudes and dV/dtmaxs were much larger. In E18.5 DMP cells, action potential amplitudes and dV/dtmaxs were considerably increased. The red-lined region will be revisited in Fig. 4D. (Fig. 2, B and C) (40). Subsequently, the AP of the DMP cells was examined. Based on the slow maximum rate of rise of the AP upstroke (dV/dtmax Ͻ 5 V/s), the smaller AP amplitude, and prominent diastolic depolarization in phase 4, the APs of E14.5 Shox2-positive DMP myocytes (Fig. 2D and Table 1) show similar properties to those of the SAN myocytes we observed ( Fig.  2A and Table 1), consistent with previously reported electrophysiological characteristics of the SAN cells (30). By contrast, at E18.5 when Shox2 and Hcn4 were down-regulated in the DMP-derived tissue (Fig. 1, C and H), the APs of the DMP myocytes displayed an increased maximum rate of rise of the AP and enlarged AP amplitude. The AP amplitude and the maximal rate of rise of the AP from each cell type were summarized in Fig. 2F. These firing properties may indicate a transitional time point of DMP toward a working myocardium fate at the neonatal stage.
Enhanced Apoptosis Underlies Hypoplastic DMP in Shox2 Mutants-The detection of Shox2 expression in the developing DMP prompted us to examine potential phenotype in Shox2deficient mice. Because the majority of Shox2 mutants died around E12.5, to harvest mutant embryos at relatively late developmental stages we administered the ␤-adrenergic receptor agonist isoproterenol, which stimulates cardiac beating, to pregnant Shox2 heterozygous females to prevent earlier embryonic lethality of Shox2 Ϫ/Ϫ mice. Histological analyses revealed a hypoplastic AV junction in Shox2 mutants beginning at E12.5 as compared with littermate controls (Fig. 3, A, B, F, and G). To quantify the volume reduction in Shox2 mutant DMP, threedimensional rendering was conducted based on X-gal staining of both Shox2 LacZ/ϩ and Shox2 LacZ/LacZ mice. We identified a 34.4% volume decrease at E12.5 and 64.5% volume decrease at E14.5 in the Shox2-expressing tissue on the right side of the AV junction in Shox2 LacZ/LacZ animals compared with heterozygous animals (Fig. 3, C-E and H-J).
To reveal the potential cellular defects underlying the hypoplastic DMP in Shox2 mutants, we examined cell proliferation rate and apoptosis in the AV junction at E12.5. Although the ratio of Ki67-positive cell in the DMP of Shox2 mutants (13.38%) was slightly lower as compared with wild-type controls (14.7%), statistical analysis determined a lack of significance between the controls and mutants (p ϭ 0.38) (Fig. 3,  K-M). However, a TUNEL assay revealed an 8-fold elevation of cell apoptosis in the developing DMPs of Shox2 mutants compared with littermate controls (Fig. 3, N-P). This observation was further confirmed by detection of ectopic activation of cleavage Caspase-3 in the mutant AV junction (Fig. 3, Q and R), indicating the contribution of elevated apoptosis to the reduced DMP tissue volume in Shox2 mutants.
Shox2 Deficiency Leads to Defective DMP Differentiation and Aberrant Electrophysiological Properties-Because ablation of Shox2 causes defective cell differentiation of the SAN (11), particularly in the sinoatrial junction domain, 5 and reduced Shox2 expression level is accompanied by down-regulation of Hcn4 in the DMP at late gestation stage (Fig. 1), we sought to determine if Shox2 regulates Hcn4 expression in the developing DMP within the AV junction by immunohistochemistry. At E11.5, the level of Hcn4 within the Shox2-expressing domain appeared comparable between Shox2 LacZ/ϩ and Shox2 LacZ/LacZ embryos (Fig. 4, A-F). However, at E12.5, along with the hypoplastic phenotype described above, Hcn4 expression domain and intensity were reduced significantly in the mutant DMP (Fig. 4, G-L). In addition, in the E12.5 mutant DMP, we detected strong ectopic expression of Cx40 as compared with controls (Fig. 4, M-R). Similar reduced Hcn4 and ectopic Cx40 expression patterns were also observed in the mutant DMP at E14.5 as compared with controls (Fig. 5, A-H). However, the expression of podoplanin, which was shown to be required for SAN development (42), remained unaltered (data not shown). These observations suggest that the fate of DMP as potential temporary pacemaking cells in Shox2 mutants is deviated and likely adopts a working myocardial fate precociously. Indeed, the APs of the DMP cells from E14.5 Shox2 mutant demonstrated significant differences compared with the wild-type counterpart, characterized by increased maximum rate of rise of the AP and AP amplitude, and it was further illustrated by a changed firing pattern (Fig. 6, A-D, and Table 1). On the other hand, the AP properties of mutant atrial and ventricular myocytes were relatively unchanged (Fig. 6, E-H, and Table 1).
BMP Signaling Is Disrupted in the DMP of Shox2 Mutants-Bmp2 and Bmp4 are expressed in the embryonic AV canal, and Smad-mediated BMP signaling activates Tbx2 expression in the AV junction (3,23). Because Shox2 is known to regulate Bmp4 expression in the developing SAN (15), we wondered if both Bmp2 and Bmp4 represent downstream genes of Shox2 in the developing DMP as well. At E11.5, Bmp2 expression domain was found restricted primarily to the developing AV canal of wild-type embryos, complementing that of Shox2, and was found not altered in Shox2 mutant (Fig. 7, A and B). On the other hand, Bmp4 was expressed within the Shox2 expression domain of the DMP region and the wall of the left sinus horn (Fig. 7C). In Shox2 mutants at this stage, Bmp4 expression was down-regulated dramatically within these regions (Fig. 7D). At E12.5, the expression of both Bmp2 and Bmp4 was detected in 5 W. Ye and Y. Chen, unpublished data.

TABLE 1 Characteristics of action potentials in multiple types of myocytes
Data are the mean Ϯ S.E. n indicates the cell number; APA, AP amplitude; dV/dtmax, maximum rate of rise of action potential; V/S, voltage/second; MDP, maximum diastolic potential. Number of cells 3  7  7  4  3  9  5  8   the control DMP but was dramatically reduced in Shox2 mutants (Fig. 7, E-H). Accompanied with this down-regulation of Bmp2 and Bmp4 was a reduction of pSmad1/5/8 in the Shox2 mutant AV junction (Fig. 7, I and J), indicating disrupted BMP signaling.

Shox2 and DMP Development
Disruption of BMP Signaling in the Developing DMP Recapitulates Shox2 Mutant Phenotype-Although it has been reported that BMP signaling plays a critical role in DMP development (5), we wondered if disrupted BMP signaling attributes to the fate change of the developing DMP in Shox2 mutants. We ablated Bmp4 from Shox2 expression domains using a Shox2-Cre allele and floxed Bmp4 mice (16,29). Histological analyses showed that mice carrying Shox2 Cre/ϩ ;Bmp4 F/F compound alleles manifest a hypoplastic DMP at E12.5 (Fig. 8, A and B). Immunohistochemistry further revealed a significantly reduced domain and level of Hcn4 expression in the developing DMP of Shox2-Cre;Bmp4 F/F mice (Fig. 8, C and D). Although the phenotypes appeared less compromised as compared with that in Shox2 mutants, the results nevertheless demonstrate an essential role of Bmp4 in the maintenance of the proper fate of the DMP.
Because Bmp2 is expressed in the adjacent AV canal at E11.5 and become localized to the developing DMP at E12.5, a functional compensation from Bmp2 may account for the relatively milder DMP defects in Shox2-Cre;Bmp4 F/F mice. To further establish a role for BMP signaling in DMP development and its fate maintenance, we overexpressed Noggin, a potent BMP antagonist preferentially binding to BMP2, BMP4, and BMP7 in Shox2-expressing cells by crossing Shox2-Cre mice to mice carrying a conditional Noggin transgenic allele (pMes-Nog). Taking the advantage of the co-expressed EGFP along with transgenic Noggin by the pMes-Nog allele (28), we were able to define the DMP domain where Shox2-Cre is expressed and trace molecular changes in Shox2 Cre/ϩ ;pMes-Nog transgenic animals. Not surprisingly, the phosphorylation of Smad1/5/8 was jeopar-dized severely in the AV junction (Fig. 8, E-H). As we expected, immunohistochemical analyses showed that Hcn4 expression was almost completely abolished accompanied by ectopic expression of Cx40 and aberrant activation of Caspase3 within the hypoplastic DMP region of Shox2-Cre;pMes-Nog mice (Fig.  8, I-X), resembling that observed in Shox2 mutants.
Smad-mediated BMP Signaling Regulates Hcn4 Expression-Although extensive studies have pinpointed the critical function of Hcn4 in CCS, the upstream regulators of Hcn4 expression have yet to be identified. Because phosphorylation of Smad1/5/8 is compromised along with the down-regulation of Hcn4 in the DMP of Shox2 mutants or mice carrying disrupted BMP signaling, we wondered if Smad-mediated BMP signaling acts as a positive upstream regulator of Hcn4 expression in the developing DMP. We conducted a loss-of-function study by creating Shox2 Cre/ϩ ;Smad4 F/F mice. In Smad4 conditional knock-out animals, again we found consistently deprived Hcn4 expression in the DMP at E12.5 and E14.5, establishing a Smad-Hcn4 regulatory cascade (Fig. 9, A and B; and data not shown). To test if Smad complexes regulate Hcn4 expression directly, we took a bioinformatic approach to search for SBEs in about 17 kb covering the Hcn4 locus among placental mammals (43,44). Although four potential SBEs were identified, one was highly conserved among the placental mammals, which is located within the first intron (Fig. 9D). Because the first intron of Hcn4  has been demonstrated to harbor regulatory elements that are required for Hcn4 expression in cardiac myocytes (45,46) as well as for its faithful recapitulation of Hcn4 expression in the AV conduction axis of transgenic animals (47), we therefore focused on this conserved Smad binding within the first intron. Subsequent in vivo ChIP assay using E13.5 heart extract and anti-Smad1/5/8 antibodies confirmed binding of the Smad1/ 5/8 complex to the conserved sequence (Fig. 9C), suggesting that Hcn4 is a direct target of Smad-dependent BMP signaling. In support with this notion, we found that Smad1/5/8 is activated in the AVN, coinciding with the abolishment of pSmad1/ 5/8 in the Shox2 ϩ lineage cells at E18.5 at which time Hcn4 expression became activated in the AVN but absent in the Shox2 ϩ lineage domain (Fig. 9, E and F).
Ectopic Bmp4 Expression Rescues DMP Defect in Shox2 Mutants-The evidence that BMP signaling is compromised in the developing DMP of Shox2 mutants and disruption of BMP signaling produces DMP defects similar to that in Shox2 mutants strongly suggests that BMP signaling is a major player to mediate Shox2 function during DMP development. To test this hypothesis, we conducted genetic rescue experiments using a conditional transgenic Bmp4 allele (pMes-Bmp4) compounded with Shox2-Cre and Shox2-LacZ alleles (Shox2 Cre/LacZ ). In such compounded mice (pMes-Bmp4; Shox2 Cre/LacZ ), transgenic Bmp4 was activated in Shox2-expressing cells including the DMP under the Shox2 mutant background. In the developing DMP of pMes-Bmp4;Shox2 Cre/LacZ mice, the pSmad1/5/8 level was resumed, and its expression domain was expanded as compared with wild-type controls and mutants (Fig. 10, A-C). Similarly, Hcn4 expression within the DMP region was also resumed (Fig. 10, D-F). Interestingly, although the majority of pMes-Bmp4;Shox2 Cre/LacZ mice died perinatally, we were able to harvest one survived such mutant mouse at the birth. Furthermore, the addition of exogenous BMP4 to isolated Shox2 LacZ/LacZ DMP cells in cell culture for 18 h rescued the phenotype of electrophysiological properties, particularly the maximum rate of rise of the AP upstroke ( Fig.  10, G-I; Table 1). These observations indicate that the Shox2 function in the regulation of DMP cell fate decision and development is mediated, at least partially, by BMP signaling.

Shox2-expressing DMP May Function as a Temporary
Pacemaker during Embryogenesis-We and others have shown previously that Shox2 is expressed in the SAN to promote cell proliferation and to function as a repressor of working myocardium genes and an inducer of CCS markers (10,48). In our current study we revisited the expression pattern of Shox2 in the developing heart using a Shox2-LacZ mouse line and revealed a novel Shox2 expression domain in the developing AV junction. By lineage tracing, we were able to pinpoint that the Shox2-expressing population in the AV junction gives rise to the vestibular spine, indicating its DMP origin, and does not contribute to the adjacent AVN. Although the mature AVN does not consist of Shox2-Cre ϩ lineage cells, a gene expression assay demonstrated that the Shox2-expressing DMP has the strongest combined expression of pacemaker maker Hcn4 and Tbx3 as well as nodal-like electrophysiological properties in the AV junction before the emergence of an identifiable AVN at late gestation stage. Thus, it appears that the embryonic DMP possesses the pacemaker feature and probably acts as a temporary pacemaker before the emergence of a mature AVN to ensure the proper sequence of conducting cardiac electrical pulses in the AV junction. The developing DMP appears to lose its pacemaking function, which may be gradually taken over by the maturing AVN at late gestation and early post-natal stages.
Shox2 Is Essential for DMP Development-Similar to the SAN defects observed in Shox2 null mutants, the Shox2 mutant DMP also displayed hypoplasia and defective differentiation. However, unlike the Shox2 Ϫ/Ϫ SAN in which reduced cell proliferation rate was thought to contribute to the hypoplastic phenotype, the developing DMP in Shox2 mutants exhibited a slightly reduced cell proliferation rate but a significantly elevated cell apoptosis, suggesting that Shox2 regulates the development of SAN and DMP by controlling different cellular event. It was reported previously that some adjacent AV canal cells undergo apoptosis to facilitate the formation of AV structures in the four chambered heart (49). It appears that Shox2 functions to shield DMP cells from apoptotic fate, and the DMP cells in Shox2 mutants adopt similar cellular properties of neighboring working myocardium. The aberrant electrophysiological character and the loss of the expression of CCS-specific gene Hcn4 and ectopic activation of the working myocardial marker gene Cx40 within the DMP domain of Shox2 mutants, similar to the phenotype observed in the SAN, indicate a deviation of DMP cell fate and precocious adoption of working myocardial fate. These observations suggest an essential role for Shox2 not only in the regulation of tissue size but also in fate determination of both the SAN and DMP. It was reported previously that the defective SAN in Shox2 mutant embryos is the causative of bradycardia, which possibly contributes to embryonic lethality (10,11,48). However, the sinus bradycardia or even loss of sinus function normally is not a direct cause of death within a short period of time, because the AV junction is able to compensate the conducting firing rate (50). Thus, the severely slowed heart rate in Shox2 mutant animals is likely due to the disrupted functions in both the SAN and AV junction.
Shox2 Functions through BMP Signaling to Regulate DMP Development-BMP signaling plays vital roles in the regulation of proliferation, apoptosis, and differentiation during cardiogenesis. Deletion of either Bmp2 or Bmp4 resulted in abnormal  (51,52). Although Bmp2 and Bmp4 are expressed in the embryonic AV canal (21,22), their roles in CCS development have not been addressed. The similar AV septal defect found in mice carrying cardiac-specific ablation of either Bmp4, Alk3, or Shox2 (5, 21) (this study) and the fact that Shox2 ablation led to deprived Bmp4 expression in the embryonic SAN (15) suggest the existence of a Shox2-Bmp4 regulatory cascade in the developing DMP. The down-regulation of Bmp4 expression within the Shox2-expressing DMP in the AV junction supports this notion, consistent with the observation found in the Shox2 Ϫ/Ϫ SAN (15). Furthermore, Bmp2 expression was also reduced dramatically in the Shox2 Ϫ/Ϫ DMP. Consistent with the abrogation of BMP ligands, phosphorylation of  Smad1/5/8 was almost abolished. It should be pointed out that at E11.5, Bmp2 expression is largely not overlapped with Shox2 in the AV junction. The strong Bmp2 expression in the surrounding AV canal tissue likely compensates for the loss of Bmp4 expression, which explains why Shox2 null mice did not display obvious DMP defect at this stage.
Although the importance of BMP signaling in DMP development has been illustrated in mice carrying Alk3 deletion in the second heart field (5), in our current study we found that mice bearing Bmp4 deletion (Shox2-Cre;Bmp4 F/F ) or Noggin overexpression (Shox2-Cre;pMes-Nog) in Shox2-expressing cells exhibit not only hypoplastic but also defectively DMP differentiation that is precocious deviation of its pacemaking-like fate to working myocardium. However, the relatively milder DMP phenotype seen in Shox2-Cre;Bmp4 F/F mice as compared with that in Shox2-null and Shox2-Cre;pMes-Nog mice further supports a functional redundant role of Bmp2 and Bmp4 during DMP development. A piece of additional compelling evidence to support the critical role of BMP signaling in DMP development came from the rescue experiments in which activation of a transgenic Bmp4 allele in the Shox2-expressing domain under the Shox2 mutant background resumed Hcn4 expression and rescued, if not fully, the hypoplastic phenotype of DMP. In addition, the AP pattern of Shox2 Ϫ/Ϫ DMP cells could also be rescued by exogenous BMP4. Taken together, our results demonstrate that Shox2 functions through Bmp2/Bmp4 to regulate DMP development and differentiation, and BMP signaling is both necessary and sufficient to induce Hcn4 expression in the developing DMP.
Smad-dependent Pathway Regulates Hcn4 Expression-The hyperpolarization-activated, cyclic nucleotide-gated 4 (Hcn4) is regarded as a molecular marker for CCS and functions to generate I f (funny/pacemaker) current that is essential for pacemaking activity. Indeed, deprivation of Hcn4 results in rhythm disorders and embryonic lethality (24,25). Several transcription factors including Tbx3, MEF2, NRSF, and SP1 have been reported to regulate Hcn4 transcription, but these factors were either shown to bind to the Hcn4 regulatory elements by in vitro assays or characterized as an inducer for ectopic Hcn4 expression in chamber myocardium at a late developmental stage (45,46,53). BMP signaling is known to be mediated by Smad-dependent canonical and Smad-independent non-canonical pathways (18,19). Although it has been reported that Bmp2-mediated Smad signaling functions through Tbx2 as a repressor of chamber-specific genes such as Cx40, Cx43, and ANF in the developing AV conduction axis (23), little is known about the direct role of BMP/Smad in promoting conduction cell differentiation. In our study we present evidence that Smad-mediated signaling is essential for DMP development and for Hcn4 expression. Our in vivo ChIP assay confirmed direct binding of pSmad1/5/8 to the conserved sequence within the first intron of Hcn4 that has been shown to be essential for Hcn4 expression in cell cultures and for faithful recapitulation of Hcn4 expression in transgenic animals (45)(46)(47), suggesting a direct regulation of Hcn4 transcription by Smad complex. Nevertheless, our studies reveal an essential function for Shox2 in DMP development and differentiation and establish a Shox2-BMP/Smad signaling axis in the regulation of Hcn4 expression in the developing DMP, which likely functions as a temporary pacemaking tissue in the AV junction during cardiogenesis.