|Home | About | Journals | Submit | Contact Us | Français|
The epicardium contributes to the majority of nonmyocardial cells in the adult heart. Recent studies have reported that the epicardium is derived from Nkx2.5-positive progenitors and can differentiate into cardiomyocytes. Not much is known about the relation between the myocardial and epicardial lineage during development, whereas insights into these embryonic mechanisms could facilitate the design of future regenerative strategies.
Acquiring insight into the signaling pathways involved in the lineage separation leading to the differentiation of myocardial and (pro)epicardial cells at the inflow of the developing heart.
We made 3D reconstructions of Tbx18 gene expression patterns to give insight into the developing epicardium in relation to the developing myocardium. Next, using DiI tracing, we show that the (pro)epicardium separates from the same precursor pool as the inflow myocardium. In vitro, we show that this lineage separation is regulated by a crosstalk between bone morphogenetic protein (BMP) signaling and fibroblast growth factor (FGF) signaling. BMP signaling via Smad drives differentiation toward the myocardial lineage, which is inhibited by FGF signaling via mitogen-activated protein kinase kinase (Mek)1/2. Embryos exposed to recombinant FGF2 in vivo show enhanced epicardium formation, whereas a misbalance between FGF and BMP by Mek1/2 inhibition and BMP stimulation causes a developmental arrest of the epicardium and enhances myocardium formation at the inflow of the heart.
Our data show that FGF signaling via Mek1/2 is dominant over BMP signaling via Smad and is required to separate the epicardial lineage from precardiac mesoderm. Consequently, myocardial differentiation requires BMP signaling via Smad and inhibition of FGF signaling at the level of Mek1/2. These findings are of clinical interest for the development of regeneration-based therapies for heart disease.
In contrast to the adult heart, the embryonic heart tube is devoid of nonmyocardial cells and an epicardium, consisting of an outer myocardial and an inner endocardial layer separated by cardiac jelly. The heart tube expands by recruitment of progenitor cells from the splanchnic mesoderm at both poles of the heart.1,2 The formation of the majority of nonmyocardial cells starts with the development of the proepicardium from splanchnic mesodermal cells at the inflow of the heart. Its villous outgrowths extend into the pericardial cavity, attach to the atrioventricular canal and gradually envelop the entire “naked” heart tube, deriving the epicardium. A subset of epicardial cells undergoes epithelial-to-mesenchymal transformation. The formed subepicardial mesenchymal cells contribute to the nonmyocardial component of the heart ie, the coronary vessels and the cardiac fibroblasts. In the adult heart, the nonmyocardial component occupies approximately 25% of the myocardial volume but comprises 60% to 70% of the cells.3,4
Recently, it was reported that (pro)epicardial cells are derived from Nkx2.5-expressing progenitors and contribute to a small proportion of the cardiomyocytes.5–8 This finding is underscored by the observation that explanted proepicardial cells spontaneously differentiate into contractile cardiomyocytes.9,10 These findings suggest that the proepicardium is derived from precardiac mesoderm, rather than from the septum transversum, ie, splanchnic mesoderm adjacent to the precardiac mesoderm.11 The molecular mechanism that regulates the separation of precardiac mesoderm into (pro)epicardial and myocardial cells is not known. Understanding this mechanism might help the development of regenerative approaches to induce differentiation of epicardial or epicardial-derived interstitial cells into cardiomyocytes.
In this study, we show that proepicardial cells and inflow myocardial cells are derived from a common precursor pool of cells, using 1,1′-dioctadecyl-3,3,3′,3′-tetramethylin-docarbocyanine perchlorate (DiI) labeling analysis. This precursor pool is directed into the myocardial lineage by bone morphogenetic protein (BMP) signaling via Smad or into the (pro)epicardial lineage by fibroblast growth factor (FGF) signaling via mitogen-activated protein kinase kinase (Mek)1/2. Mek1/2-mediated FGF signaling is dominant over BMP signaling. In embryos treated with BMP2 and the Mek1/2 inhibitor (U0126), epicardial development was blocked and inflow myocardium formation enhanced. Treatment with FGF2 revealed a reciprocal phenotype, showing stimulated epicardium and inhibited inflow myocardium formation. Taken together, we show that BMP-mediated myocardial differentiation is inhibited by FGF signaling via Mek1/2 and extracellular signal-regulated kinase (Erk)1/2, which is required to separate the epicardial lineage from a common progenitor pool.
After 48 hours of incubation, eggs were windowed, and stage 11 embryos were injected with BMP2 (50 ng/mL), FGF2 (50 ng/mL), and/or U0126 (10 µmol/L) into the yolk sac, taking into account the diluent volume of the egg. Control embryos were injected with growth factor solvent. After 24 hours, the embryos (stage16) were isolated or reinjected and incubated for another 24 hours (stage19–20). At least 3 embryos per group were used for in situ hybridization and 3D reconstruction.12
Whole mount in situ hybridization showed that the transcription factor Tbx18 is expressed upstream of the linear heart tube and subsequently becomes confined to the proepicardium in mouse and chicken.16 Functional disruption of Tbx18 results in normal development of the epicardium but aberrant muscularization of the cardiac inflow.17 Moreover, lineage analysis showed that the Tbx18 population contributes to both the myocardial and epicardial lineages.5,17 To explore the relation of the Tbx18-expressing cells of the forming proepicardium and inflow myocardium, 3D reconstructions were prepared of serial sections stained for Tbx18 and a myocardial marker (ventricular myosin heavy chain or cardiac troponin I [cTnI]). At stage 11, Tbx18 expression can be seen (Figure 1a and 1b) in mesodermal cells covering the vitelline vein. In this region all cells express Tbx18 though cardiac sarcomeric proteins and morphological characteristics of proepicardial development are absent. At stage 13, proepicardial villi start to develop within the Tbx18 positive area. Proepicardial villi at the left vitelline vein disappear whereas those at the right side expand and reach the dorsal aspect of the atrioventricular canal at stage 17.16 At this stage, Tbx18 is observed in the flanking inflow myocardium (Figure 1c through 1f), being more extensive at the right side. At stage 21, the entire inflow myocardium is Tbx18-postive and the epicardium now covers the entire dorsal aspect of the heart and has started to envelop the ventral side. This expression analysis suggests that the Tbx18-positive mesodermal cells covering the vitelline veins contribute to both epicardial and inflow myocardial cells.
To explore this hypothesis, we placed a DiI label within the Tbx18-positive population covering the right vitelline vein at stage 11 (Figure 2). At stage 16, the labeled cells were found in both the proepicardium and inflow myocardium (Figure 2b through 2d). When cells of the left vitelline vein were labeled, the label was only found in the myocardium (Figure 2f). These experiments, together with the expression analysis and lineage analysis in mice,6,7 suggest that the proepicardium and inflow myocardium develop from a common progenitor pool.
Explant assays have shown that proepicardial cells can differentiate into myocardium and that BMP2 stimulates, and FGF2 inhibits myocardium formation,9,10 suggesting that these factors are regulators of the separation of the myocardial and epicardial lineages. Immunofluorescently marking the canonical BMP- and FGF-signaling pathways (Figure 3b) in vivo reveals the nuclear localization of P-Smad in the proepicardial and flanking myocardial cells (Figure 3a and 3b). In the distal part of the proepicardium P-Smad can also be observed in the cytoplasm (Figure 3b′), In these cells, P-Erk could also be seen (Figure 3c and 3c′). The cytoplasmic localization of P-Smad in the presence of P-Erk and the nuclear localization of P-Smad in the absence of P-Erk are indicative for an intracellular interaction between these pathways. Such an interaction has been shown in Xenopus, in which P-Erk phosphorylates P-Smad, resulting in abrogation of BMP signaling attributable to cytoplasmic P-Smad accumulation and degradation (Figure 3d).18,19 A comparable interaction may also be operational in the separation of the progenitor population into the proepicardial and myocardial lineages.
To evaluate whether FGF signaling via Mek and Erk effects Smad phosphorylation, rat cardiomyocyte-like cells (H10 cells) were treated with U0126 in the absence or presence of BMP2 and/or FGF2. Western blot analysis revealed that P-Erk could be detected in the absence or presence of FGF2 (Figure 3e), indicating that FGF signaling is endogenously active in H10 cells. Phosphorylation of Erk is strongly inhibited by U0126 even in the presence of FGF2. Stimulation of H10 cells with BMP2 induced low levels of P-Smad after 15 minutes, becoming more prominent after 30 minutes (Figure 3f). H10 cells treated with U0126+BMP2 reveal high levels of P-Smad after 15 minutes (Figure 3f), showing an inhibiting effect of FGF signaling via Mek on Smad-mediated BMP signaling. To further substantiate the inhibition of FGF signaling on Smad phosphorylation, H10 cells were preincu-bated with FGF2 and subsequently stimulated with BMP2. Preincubation with FGF2 reduced the level of P-Smad in BMP2- and in BMP2+FGF2-treated cells compared to cells that were not pretreated with FGF2 (Figure 3g).
To establish whether this interaction is operational in proepicardia, freshly isolated proepicardia were stimulated with BMP2, resulting in phosphorylation of Smad (Figure 3h and 3i). Reduced Smad phosphorylation was found when proepicardia were simultaneously stimulated with FGF2. Stimulating with FGF2 alone did not effect Smad phosphorylation. The inhibiting effect of FGF signaling on Smad phosphorylation was abolished when U0126 was added, indicating that also in proepicardia FGF signaling via Mek and Erk mediates an inhibitory effect on BMP-induced Smad phosphorylation.
Taken together, these results suggest that Smad-mediated BMP signaling recruits progenitor cells into the myocardial lineage and Erk-mediated FGF signaling inhibits Smad-mediated BMP signaling, preventing differentiation of the proepicardium into the myocardial lineage.
To evaluate the interaction between BMP signaling and FGF signaling on the differentiation of proepicardial cells, the myocardial area and the number of myocardial and nonmyo-cardial cells were determined in cultured proepicardia (Figure 4a through 4d, Online Figure I, and Online Table I). In control, BMP2-, or FGF2-treated explants, the total number of cells was not significantly different after 5 days of culture. In BMP2+FGF2-treated cultures the total number of cells was significantly (P<0.05) larger from day 2 onward. Cardiomyocytes were virtually absent from FGF2- and BMP2+FGF2-treated cultures. In controls and BMP2-treated cultures cardiomyocytes were observed from 2 days onwards. At 5 days of culture, the myocardial area was 2.1-fold larger in BMP2-treated cultures compared to controls. Calculating the average cardiomyocyte size showed a similar size in all conditions (279±80 µm2), indicating that the changes in myocardial area are the result of de novo differentiation rather than of hypertrophy. These observations suggest that the inhibition of myocardial differentiation by FGF2 is dominant over the stimulatory effect of BMP2. To further substantiate this hypothesis, proepicardia were cultured in the presence of U0126 (Figure 4a′ through 4d′). In control and BMP2-treated cultures, U0126 caused a significant reduction of 64% and 40%, respectively, in the total number of cells without significantly altering the myocardial area, number or size of cardiomyocytes. However, in FGF2-treated cultures, U0126 addition resulted in a 23% increase in total cell number with myocardial formation returning to control levels. In BMP2+FGF2-treated cultures U0126 did not significantly influence the total cell number but like FGF-treated cultures, myocardial area returned to control levels (Figure 4e and 4f and Online Table II).
The relative contribution of myocardial and nonmyocardial cells under the different conditions was calculated and plotted (Figure 4f), showing that in control and BMP2-treated explants U0126 reduces the number of noncardiomyocytes. FGF2 stimulation results in an increase in the noncardiomyocyte population. On FGF2 or BMP2+FGF2 treatment, virtually no cardiomyocytes are formed. Under these conditions cotreatment with U0126 reverses the inhibitory effect on cardiomyocyte formation to control levels, without inhibiting the stimulatory effect on noncardiomyocyte formation.
Taken together, these data show that BMP-mediated myocardial differentiation is inhibited by FGF signaling via Mek1/2 and Erk1/2. Shifting the balance in favor of BMP2 signaling by adding BMP2 and simultaneously inhibiting Mek1/2-mediated FGF signaling results in an almost complete differentiation of proepicardial cells into cardiomyocytes.
The expression levels of genes characteristic for the myocardial and epicardial lineages were determined using quantitative RT-PCR and compared between HH16 proepicardium, sinus venosus, atrium, ventricle, and HH24 epicardium (Figure 4g). In control and U0126+BMP2-treated cultures, the expression levels of the myocardial genes AMHC, VMHC, and BNF closely resembled the expression observed in atria. On FGF2 or BMP2+FGF2 treatment, the expression levels of myocardial genes were lower. In control explants, the expression levels of Cx43, a marker for working myocardium, was approximately 2-fold higher than in atrial samples and more than 3-fold higher than in BMP2+FGF2-treated explants. In FGF2- or BMP2+FGF2-treated explants, the expression was at a similar low level as in proepicardium. In situ hybridization analysis showed that Tbx18 expression tapers off and Nkx2.5 expression gradually increases in the sinus venosus myocardium (Figure 4h). Taken together, these analyses indicate that proepicardial explants differentiate into working myocardium during the culture and BMP2+U0126 further promotes this differentiation.
Analysis of the epicardial marker genes showed that, in line with previous reports,10 Tbx18 and Raldh2 are down-regulated during culture and are hardly effected by the different culture conditions. Expression of γ smooth muscle actin, a marker for coronary smooth muscle cells, increased in control explants, was not effected by BMP2+FGF2 or BMP2+U0126 treatment and was highest when epicardial differentiation is promoted by FGF2. Flk1, a marker for endothelial/endocardial cells, decreased during culture compared to proepicardium. Based on Flk1 and γ smooth muscle actin expression, proepicardial cells differentiate in coronary smooth muscle cells rather than into endothelial cells. When treated with BMP2+FGF2, proepicardial cells showed an almost 2-fold higher expression level of Flk1 compared to 5 days cultured controls. Taken together, the analyses of the epicardial marker genes suggest that proepicardial cells differentiate along the epicardial lineage. When myocardial differentiation is inhibited by FGF2 or BMP2+FGF2, differentiation into the smooth muscle cell lineage or endothelial/endocardial lineage, respectively, is promoted.
To uncouple FGF signaling and BMP signaling in vivo, we injected eggs before proepicardial induction (stage 11)16 with BMP2, FGF2, or growth factor solvent in combination with or without U0126. After 24 hours of reincubation, the embryos (stage 16 to 17) were isolated and none of the embryos showed gross morphological abnormalities. To visualize all myocardium (ventricular myosin heavy chain or cTnI) and the proepicardium and inflow myocardium (Tbx18), in situ hybridization was used. In all (n=10) embryos treated with U0126+BMP2, a small sac-like proepicardium without villous projections had developed and epicardium consistently failed to develop (Figure 5a and 5b). Within the rudimentary proepicardium, myocardial strands were present. The efficacy of this treatment was assessed by immunohistochemical staining for P-Smad and P-Erk (Online Figure II). Within the rudimentary proepicardium, myocardial strands were present. In the FGF2-treated embryos (n=3), epicardium formation was enhanced compared to control embryos (n=3) (Figure 5a and 5c). The epicardium had almost completely enveloped the atrioventricular canal, which is normally observed at stage 18. In BMP2-treated (n=3) or U0126-treated (n=3) embryos, the proepicardium was similar to controls (Online Figure III).
After 48 hours of reincubation, all isolated embryos showed morphological characteristics that qualified them as stage 20 embryos (Online Figure IV). Treatment with either U0126 (n=3) or BMP2 (n=3) revealed no abnormalities (Online Figure III). FGF2- and U0126+BMP2-treated embryos displayed abnormalities in epicardium and inflow myocardium. To identify abnormalities in the Tbx18- and cTnI-expression domains, 3D reconstructions were made. In control embryos, the epicardium covered the dorsal aspect of the ventricle and the entire atrioventricular canal (Figure 6a). In U0126+BMP2-treated embryos (n=8), epicardium failed to develop; the proepicardium remained a small sac-like structure without villous protrusions (Figure 6b). The myocardium of the ventricles was thin, showed no compact layer and hardly any trabeculae. This abnormality is most probably secondary to the absence of an epicardium. Compared to the controls, extensive myocardial sleeves covering the sinus horns were present (Figure 6b and 6e). In FGF2-stimulated embryos (n=2), on the other hand, (pro)epicardial cells had formed a bridge over the entire length of the systemic inflow that was attached to the atrioventricular canal, whereas inflow myocardium formation was diminished (Figure 6c and 6f and Online 3D PDF).
Taken together, these data show that the interaction between FGF signaling and BMP signaling at the level of Mek1/2 is important in the regulation of differentiation of progenitor cells at the inflow of the heart. Simultaneous inhibition of Mek-mediated FGF signaling and stimulation of BMP signaling results in preferential differentiation of progenitor cells into myocardial cells at the expense of epicardial cells, whereas stimulation of FGF signaling leads to a reciprocal phenotype (Figure 7a).
During normal development, proepicardial differentiation is induced at both vitelline veins. At the left side, the proepicardial villi disappear, whereas the right-sided proepicardial villi expand.16 To establish the role of apoptosis or proliferation in the inhibited proepicardia and increased inflow myocardium formation in U0126+BMP2-treated embryos, we performed TUNEL staining and analyzed 5-bromodeoxyuridine incorporation. In control and U0126+BMP2-treated embryos, no TUNEL-positive cells could be detected in the inflow region, whereas in other regions of the embryo, TUNEL-positive cells were present as expected (Figure 5e and 5g).20 No differences in proliferation index between (pro)epicardial and inflow myocardial cells and between control and U0126+BMP2-treated embryos were found (2-way ANOVA; P=0.116 and P=0.701, respectively). These findings further support the idea that the observed changes in (pro)epicardium and inflow myocardium formation are attributable to a shift in the separation of precursor cells into the myocardial and (pro)epicardial lineages.
The early heart tube increases in length by recruitment of myocardial cells from flanking mesoderm and becomes ensheathed by an epicardial layer. Using gene expression analysis (Figure 1) and DiI labeling (Figure 2), we show that caudal of the inflow of the heart a Tbx18-positive progenitor population contributes to both inflow myocardium and epicardium. Regarding the DiI-labeling experiment we are aware that this experiment cannot exclude the possibility that 2 different progenitor pools are already present at stage 11. Based on the DiI-labeling experiment and gene expression analyses, we conclude that a small homogenously Tbx18-expressing group of cells gives rise to myocardial and proepicardial cells. In situ hybridization analyses showed that the mesoderm covering the vitelline veins as well as the proepicardium and adjacent inflow myocardium stain homogeneously for Tbx18 (Figure 1). No morphological signs of epicardial differentiation are visible, and no signs of myocardial differentiation are evident, suggesting that progenitors have not yet made a lineage split at stage 11. Moreover, the proepicardial cells still possess the capacity to differentiate into cardiomyocytes in vitro (Online Figure I), also suggestive of a common progenitor pool.
Initial proepicardial explant analyses have shown that BMPs and FGFs influenced myocardium and/or epicardium formation. Gene expression analysis showed that BMP2 is expressed in the proepicardium and inflow myocardium and FGF2 in the stroma of the proepicardium.9,21 In this study, we focused on the signaling pathways conveying BMP2 and FGF2 signals and their interaction, BMP2 being the strongest stimulator of myocardial formation and FGF2 preventing spontaneous and BMP-induced myocardial formation in vitro (Online Figure I).
During development, FGF signaling is transduced by 3 major pathways, phosphatidylinositol 3-kinase, phospho-lipase Cγ, and Erk1/2.22 Testing the effect of inhibitors on the various pathways showed that the inhibiting effect on myocardium formation of FGF2 signaling is transduced via Mek and Erk (Figure 4 and data not shown). This finding was further supported by Western blotting and immunofluorescence, showing P-Erk expression in the proepicardium. BMP signaling is transduced by 2 major pathways, Smad and p38.23 Inhibition of the p38 pathway did not effect myocardium formation (data not shown). Smad inhibitors are not commercially available, but Western blotting and immunofluorescence showed different expression of P-Smad in the proepicardium and inflow myocardium. Interestingly, in the region of the proepicardium that expresses P-Erk, P-Smad is located in the cytoplasm. In Xenopus, phosphorylation of the linker in Smad by P-Erk results in a disruption of BMP-mediated Smad signaling.18,19 In line with this finding, we found that stimulating explants with BMP2+FGF2 prevents myocardium formation and inhibiting FGF signaling, using U0126, in the presence of BMP2 strongly stimulated myocardium formation; on average, 75% of the cells in the explant cultures differentiated into cardiomyocytes (Figure 4). Moreover, directly challenging proepicardia with these substances showed the expected changes in P-Erk and P-Smad on Western blots (Figure 3i). This finding probably also underlies the spontaneous differentiation into cardiomyocytes on explanting, the FGF concentration becomes diluted, thus diminishing the FGF inhibition of BMP-mediated myocardial differentiation.
To evaluate whether this interaction between FGF signaling and BMP signaling is also operational in vivo, embryos were treated in ovo. In line with our in vitro observations, the combination of U0126+BMP2 resulted in enhanced myocardial differentiation at both poles of the heart (Figure 6a and 6b), blocked expansion of the proepicardium, and consequently, the epicardium did not develop. FGF2 alone resulted in reduced inflow myocardium formation and enhanced epicardium formation. A potential pitfall of the in vivo model is that the entire embryo is exposed, which, as a consequence, might influence other organs as well. Approaches for local administration failed to provide a continuously effective concentration of the applied factors, most probably because of dilution in the egg. However, on visual inspection of forming limbs, somites, etc, the development of the treated embryos appeared normal. Liver development has been reported to be inhibited by U0126.24 Only in U0126+BMP2-treated embryos could we observe a mild inhibition of hepatogenesis (data not shown). The difference in effect of U0126 on the liver is probably attributable to a dose difference, 10 versus 50 µmol/L. An indirect effect of U0126+BMP2 via the liver on the progenitor pool is unlikely because of the presence of a developing liver, the presence of Tbx18 (which is induced by signals from the liver25), and the consistency between the in vitro and in vivo results. Finally, in the FGF2-treated embryos, no ectopic induction of Tbx18 was observed, indicating that FGF2 is most likely not the liver-derived inductor of Tbx18 in this progenitor population.25
The finding that inflow myocardial and epicardial cells originate from a common Tbx18-expressing progenitor pool is underscored by several knockout phenotypes showing inflow and/or epicardium defects. Tbx18 knockout mice show delayed formation of the myocardial sleeves covering the sinus horns, although epicardial development appears normal.17 Podoplanin, a coelomic and myocardial marker, knockout mice have both epicardial and inflow myocardial defects.26 When the mammalian homolog of Caenorhabditis elegans polarity proteins, PAR3 (PARD3–Mouse Genome Informatics), is deficient, epicardial development is defective.27
The common origin of epicardial and myocardial cells led to the idea that epicardially derived cells might play a role in cardiac regeneration and serve as a source for cardiomyocytes. Initial observations in zebrafish support this idea. On ventricular amputation, the myocardium regenerates. During regeneration, embryonic genes, such as Tbx18 and Raldh2, are reactivated in the epicardium, and new cardiomyocytes are formed from progenitor cells.28 The origin of these progenitors is unclear, but the epicardium is a potential candidate.
Whether adult mammalian epicardium is able to behave similarly on myocardial injury remains unclear. Nevertheless, a role of the epicardium during regeneration is plausible based on the following findings. (1) c-Kit-positive cells are present in the subepicardial mesenchyme.29,30 (2) WT1 expression is induced in coronary vessels at the border zone of an infarct.31 (3) Subepicardial mesenchyme starts to proliferate after myocardial infarction.32 (4) A subset of adult cardiac stem cells can differentiate into coronary vessels when injected in infarcted hearts.33 In this respect, it is of relevance to note that the coronary arteries have a (pro)epicardial origin.
Altogether, these observations suggest that the (pro)epicardium and epicardial-derived cells contribute to the cardiac stem cells population observed in several studies.34,35 In this respect, the presence of both P-Smad and P-Erk in stage 33 epicardium indicates that at later stages a balance between FGF and BMP signaling might be operational (Figure 7b through 7d). Differentiation of these progenitors into cardiomyocytes may also require a shift in the balance between BMP signaling and FGF signaling in favor of BMP by Mek1/2 uncoupling, offering promising inroads to cardiac regeneration from (pro)epicardial-derived cells.
We thank Mr J. Hagoort for preparing the 3D PDF.
Sources of Funding
This work was supported by Netherlands Heart Foundation grant 1996M002 (to B.v.W., G.v.d.B., M.J.B.v.d.H., and A.F.M.M.); European Community’s Sixth Framework Program contract (‘HeartRepair’) grant LSHM-CT-2005-018630 (to M.J.B.v.d.H, S.v.d.V., A.F.M.M.); the Deutsche Forschungsgemeinschaft (M.S.); Rosalind Franklin Fellowship of the University of Groningen (to M.S.); and the Duke Brumley NPRI Children’s Heart Foundation (to M.L.K.).
Data Supplement (unedited) at: http://circres.ahajournals.org/cgi/content/full/CIRCRESAHA.109.203083/DC1
Reprints: Information about reprints can be found online at http://www.lww.com/reprints