Cardiac myocyte p38α kinase regulates angiogenesis via myocyte-endothelial cell cross-talk during stress-induced remodeling in the heart

Stress-induced p38 mitogen-activated protein kinase (MAPK) activity is implicated in pathological remodeling in the heart. For example, constitutive p38 MAPK activation in cardiomyocytes induces pathological features, including myocyte hypertrophy, apoptosis, contractile dysfunction, and fetal gene expression. However, the physiological function of cardiomyocyte p38 MAPK activity in beneficial compensatory vascular remodeling is unclear. This report investigated the functional role and the underlying mechanisms of cardiomyocyte p38 MAPK activity in cardiac remodeling induced by chronic stress. Using both in vitro and in vivo model systems, we found that p38 MAPK activity is required for hypoxia-induced pro-angiogenic activity from cardiomyocytes and that p38 MAPK activation in cardiomyocyte is sufficient to promote paracrine signaling-mediated, pro-angiogenic activity. We further demonstrate that VEGF is a paracrine factor responsible for the p38 MAPK-mediated pro-angiogenic activity from cardiomyocytes and that p38 MAPK pathway activation is sufficient for inducing VEGF secretion from cardiomyocytes in an Sp1-dependent manner. More significantly, cardiomyocyte-specific inactivation of p38α in mouse heart impaired compensatory angiogenesis after pressure overload and promoted early onset of heart failure. In summary, p38αMAPK has a critical role in the cross-talk between cardiomyocytes and vasculature by regulating stress-induced VEGF expression and secretion in cardiomyocytes. We conclude that as part of a stress-induced signaling pathway, p38 MAPK activity significantly contributes to both pathological and compensatory remodeling in the heart.

In response to mechanical stress or loss of myocytes due to pathological injury, a network of stress-signaling pathways is activated in the myocardium, including stress-activated MAP kinase p38. In previous studies from both in vitro and in vivo, activated p38 MAPK activity in cardiomyocytes was shown to induce many aspects of pathological remodeling in heart, including hypertrophy, myocyte death, loss of contractility, interstitial fibrosis, fetal genes, and proinflammatory gene induction. Constitutive activation of the p38 pathway in mouse heart was associated with lethal cardiomyopathy, whereas inhibiting p38 MAPK activity attenuated myocardial injury to ischemia/reperfusion injury or ameliorated the onset of heart failure in experimental animals. These studies laid down the mechanistic foundation for a human clinic trial for ischemic heart failure by targeted inhibition of p38 kinase. However, p38 MAPK activity was also reported to be reduced in end-stage heart failure, and genetic inactivation of p38␣ in mouse heart actually accelerated the progression of heart failure under chronic stress. These observations seemingly contradict to the previous notion that the functional consequence of p38 MAPK activation in heart is largely pathological without a significant beneficial effect.
It is known that both adaptive and maladaptive remodeling occur in stressed heart. In addition to interstitial fibrosis, fetal gene induction, and myocyte hypertrophy, changes in the coronary and capillary network of the heart is also a major part of the remodeling process. In the heart, adequate perfusion is required to maintain cardiac function. Therefore, cardiac vascular function is constantly regulated by cross-communication between myocytes and endothelium. As in other organs, the cardiac vascular function can be regulated in an acute fashion by vasoconstriction or vasodilation (1) or more gradually through the growth of new vessels and angiogenesis under chronic conditions (2). Recently a number of studies demonstrate that myocyte hypertrophy is accompanied by a corre-sponding increase in capillary density (3)(4)(5)(6). Furthermore, they have also shown that a reduction in capillary density plays a critical role in the transition from a compensated hypertrophy to a decompensated state of heart failure. Indeed, therapeutic angiogenesis in the heart has been the focus of a number of investigations with significant potential (2).
Angiogenesis is a complex process that requires coordinated action from multiple cytokines secreted from both the surrounding tissues and the vascular cells themselves (7,8). These cytokines act in either a paracrine or autocrine manner upon endothelial cells, vascular smooth muscle cells, and pericytes. Although the cytokines and their effect on vascular cells are well studied, the signaling pathways that promote angiogenic cytokine production for paracrine signaling, particularly in stressed cardiomyocytes, are not well understood.
In this report we employed both in vitro and in vivo model systems to interrogate the function of p38␣, the dominant isoform in mouse heart, in the context of stress response in heart. We found that p38 MAPK activation in cardiomyocytes was sufficient and necessary for the induction of a pro-angiogenic activity secreted from the stressed cardiomyocytes. We further demonstrated that vascular endothelial growth factor (VEGF) was responsible for the pro-angiogenic paracrine activity induced by p38 MAPK in cardiomyocyte. VEGF was induced in stressed myocytes or intact hearts in a p38-dependent manner, and VEGF mRNA levels and its secretion was induced by p38 activation via Sp1 transcription factor in cardiomyocytes. Most relevantly, we found that the cardiomyocyte-specific inactivation of p38␣ led to a blunted angiogenesis after mechanical overload in intact heart associated with an accelerated progression to heart failure. Therefore, we have uncovered an unexpected beneficial role for the cardiomyocyte p38␣ activity in promoting cardiac angiogenesis by regulating VEGF induction in response to stress. This finding offers a mechanistic basis for the contradicting functions of stress signaling during cardiac remodeling and demonstrates a complex cell-cell communication network regulated by p38 MAPK pathway that has both beneficial and pathological consequences in a stressed heart.

p38 MAPK activity regulated pro-angiogenic paracrine activity in cardiomyocyte
One of the most potent inducers of angiogenesis among different cell types is hypoxia. In neonatal rat ventricular myocyte (NRVMs) 5 cultures (with Ͼ90% pure for cardiomyocytes; see Fig. 1, 4 h of hypoxia robustly induced p38 MAPK activation, as demonstrated by phospho-p38/total p38 ( Fig. 2A). The conditioned media from the hypoxia-treated NRVMs promoted vascular tube formation compared to untreated cells (Fig. 2B). Furthermore, pretreatment with 2 M SB202190, a p38 specific inhibitor in NRVMs, significantly reduced this hypoxia-induced pro-angiogenic activity (Fig. 2B). To test whether the p38 MAPK activity induced during hypoxia is sufficient to promote pro-angiogenic paracrine activity from myocytes, NRVMs were transfected with Adv-MKK3bE, an upstream kinase with specific activity for p38 MAPK without a significant impact on ERK or Akt activities (Fig. 3A) (9). The conditioned media from the p38-activated NRVMs were applied to porcine aortic endothelial cells to measure the pro-angiogenic ability as described (10). The conditioned media from the MKK3bE-expressing NRVMs had a potent pro-angiogenic activity (2.1-fold over control) similar to that observed from the hypoxia-treated NRVMs (Fig.  3, B and C). Co-expressing p38␣ dominant negative (DN) mutant in NRVMs significantly reduced MKK3b(E)-induced paracrine activity for endothelial cell growth (Fig. 3, B and C). These results suggest that cardiomyocyte p38␣ activity is both necessary and sufficient for stress-induced pro-angiogenic paracrine activity from cardiomyocytes.  Phospho-p38 and total p38 were measured by immunoblotting as described. *, p Ͻ 0.05. B, p38 MAPK activity is necessary for hypoxiainduced paracrine pro-angiogenic activity. Conditioned media from NRVM untreated (Control), treated with p38 inhibitor (SB202190, 2 M), exposed to 4 h of hypoxia, and treated with p38 inhibitor in combination with hypoxia exposure were applied to an in vitro Matrigel-based angiogenesis assay for 6 -8 h. Vascular tube formation was quantified as described under "Experimental procedures." *, p Ͻ 0.05.

p38␣ MAP kinase regulates angiogenesis in the stressed heart Cardiomyocyte p38 MAPK activity regulated VEGF expression and secretion
To investigate the molecular basis of p38 MAPK-mediated pro-angiogenic paracrine activity in cardiomyocytes, we investigated the impact of p38 MAPK activation on pro-angiogenic factor VEGF in cardiomyocytes. As shown in Fig. 4A, activation of p38 MAPK led to a significant increase (3.9-fold) in VEGF mRNA levels in myocytes. The induction level was significantly reduced with the co-expression of p38␣ DN (Fig. 4A) or pretreatment with p38 inhibitor SB202190 (Fig. 4B). ERK and AKT are both implicated in VEFG regulation (11,12). However, coexpression of p38␣ DN with MKK3b(E) in NRVMs actually increased the basal phospho-Akt levels modestly (Fig. 3A) but reduced VEGF mRNA induction. Similarly, inhibiting ERK activity with U0126 did not significantly affect MKK3b(E)-induced VEGF expression (Fig. 4C). Therefore, specific activation of p38␣ is sufficient to induce VEGF mRNA levels in myocytes. Reflecting what was seen at the mRNA level, myocytes expressing MKK3b(E) showed an increase in total intracellular VEGF protein levels (Fig. 4D), whereas co-expressing with p38␣ DN reduced it. Finally, the conditioned media collected over a 24-h incubation period from the NRVMs treated with Adv-MKK3b(E) had significantly higher concentrations of VEGF than those from cells treated with Adv-LacZ (2.3-fold, 3.7-fold, and 4.8-fold at 48, 72, and 96 h post-infection, respectively) or Adv-p38␣-DN (2.3-fold, 3.6-fold, and 2.5-fold at 48, 72, and 96 h post-infection, respectively) (Fig. 4E). Furthermore, co-expressing p38␣ DN and MKK3b(E) resulted in a significant decrease in VEGF induction in the conditioned media (23.2% and 23.9% reduction at 48 and 72 h post infection, respectively), and the same trend remained at the 96-h time point where the decrease (31.0%) had p ϭ 0.084 (Fig. 4E). Because p38 activation in cardiomyocytes can lead to cardiomyocyte hypertrophy (13,14), we examined the effect of pro-hypertrophic treatment of phenylephrine in NRVMs but observed no significant impact on the VEGF expression (Fig. 4F). In contrast, phenylephrine (PE)-treated NRVMs showed significant hypertrophy in size and induction of ANF gene expression (Fig. 4, G-I) associated with sustained changes in MAP kinases and AKT signaling (Fig. 4, J and K). These results indicate that p38 MAPK Figure 3. p38 MAPK activation is sufficient to promote paracrine pro-angiogenic activity from cardiomyocytes. A, selective p38 activation by MKK3bE in NRVM. Western blots from NRVM total protein treated with adenovirus vectors expressing LacZ, p38␣ DN (p38DN) and MKK3bE as indicated. B and C, p38 MAPK activity is sufficient to induce paracrine pro-angiogenic activity from cardiomyocytes. Representative images (B) and quantification (C) of vascular endothelial cell growth after treated with conditioned media from NRVM-expressing LacZ, p38DN, MKK3bE, and MKK3bE in combination with p38␣ DN. * p Ͻ 0.05 compared with other groups.

p38␣ MAP kinase regulates angiogenesis in the stressed heart
activity in cardiomyocytes is both sufficient and necessary for VEGF expression and secretion. VEGF induction is not a direct consequence of cardiomyocyte hypertrophy but, rather, a specific downstream event of p38-mediated stress response.

VEGF induction was essential to p38 MAPK-mediated pro-angiogenic paracrine activity from cardiomyocytes
We further examined the role VEGF induction in p38 MAPK-mediated pro-angiogenic paracrine activity from myocytes. PAE-VEGFR2 cells treated with media from MKK3b(E)-expressing myocytes showed a significant level of phospho-VEGFR2 compared with the other treatment groups (Fig. 5A). Furthermore, treatment of the NRVMs media with an anti-VEGF antibody before application in the in vitro angiogenesis assay blocked the pro-angiogenic effect significantly (decreased 62.4%, Fig. 5, B and C). This result indicates that VEGF induction is essential to p38 MAPK-mediated pro-angiogenic paracrine activity from cardiomyocytes.

Inactivation of p38␣ activity in cardiomyocyte-impaired vascular remodeling in response to pressure overload
To investigate the role p38␣ MAPK plays in myocyte-endothelial cell cross-talk in vivo, we examined whether or not the loss of p38␣ would affect the angiogenic response in the intact heart both at baseline and after pressure overload induced by transverse aortic constriction (TAC). Cardiac-specific knockout (CKO) of p38␣ was accomplished by cardiac-specific cre/ loxP-mediated recombination (15,16). The resulting p38␣ CKO animals showed a significant loss of p38␣ protein in the myocardium (Fig. 7A) (15). Similar to what was reported earlier (16), the p38␣ CKO mice were phenotypically normal at baseline, showing no gross anatomical abnormalities in heart (Fig.  7B) and exhibited no baseline functional differences in heart rate (HR), ejection fraction (EF), or fractional shortening (FS) (Fig. 7C).
In agreement with earlier reports, we observed that p38␣ CKO hearts developed an accelerated progression to heart failure after TAC without a significant impact on hypertrophy (Fig.  8, A-C). As shown in Fig. 9, cross-sectional areas measured from WGA-stained sections showed no significant differences between p38␣ CKO and wild-type (WT) hearts 28 days after TAC (Fig. 9, A and B). We also measured the fibrotic area and found no significant difference between p38␣ CKO and WT hearts 28 days after TAC (Fig. 9, C and D). The capillary density in the myocardium showed no significant difference between p38␣ CKO and WT hearts at baseline (Fig. 10, A and B), and the observed ratio of capillary/myocyte in the ventricle was in good agreement with other studies (17,18). After 28 days of TAC, capillary density in p38␣ WT controls was significantly increased by ϳ2.5-fold. However, this increase was significantly blunted in the p38␣ CKO hearts (Fig. 10, A and B). Quantitative mRNA analysis showed vascular marker genes such as vascular endothelial cadherin (VE-CAD) and CD31 were significant induced the WT hearts, and this induction was also blunted in the p38␣ CKO hearts at 28 days post-TAC (Fig. 10C). However, at 28 days post-TAC, VEGF expression was also not induced in the p38␣ CKO heart after TAC (Fig. 10D). All these results suggest that loss of myocyte-specific p38␣ MAPK signaling blunts the chronic stress-induced angiogenic response in vivo. These data suggest that cardiomyocyte p38␣ activity has a beneficial role in compensatory vascular remodeling in heart.

Discussion
In this study we demonstrated that p38 MAPK activation is both necessary and sufficient to a stress-induced pro-angiogenic paracrine activity from cardiomyocyte. VEGF expression is induced through p38 MAPK/SP1 pathway in cardiomyocytes and is shown to be responsible for this pro-angiogenic paracrine activity. Loss of p38␣ activity from adult cardiomyocytes impairs stress-induced angiogenesis and vascular remodeling in intact heart, leading to accelerated heart failure after pressure overload. Therefore, all these data provide the first in vitro and in vivo evidence that a pathologically related stress-re-sponse pathway, p38 MAPK, also has an important beneficial role in compensatory vascular remodeling by mediating the cross-talk between cardiomyocytes and endothelial cells via a VEGF-dependent paracrine signaling mechanism.
Previous studies in a number of cell types, including myocytes, have shown the importance of p38 MAPK signaling in inflammatory cytokine production. Our study provides direct evidence that the p38 MAPK pathway also plays a critical role in angiogenesis by inducing a potent paracrine factor VEGF from cardiomyocytes. This observation is consistent with similar observations in other cell types, including breast cancer cells (19), Lewis lung carcinoma cells (20), osteoblasts (21), skeletal muscle (22), and vascular smooth muscle (21,(23)(24)(25). One previous study showed that IL-1␤ treatment of cardiomyocytes increased VEGF in a p38-dependent manner (26), suggesting a potential synergy between these two downstream events of p38 MAPK signaling: inflammatory response and angiogenesis.
Our study established the underlying mechanism of VEGF induction via the p38 MAPK/SP1 pathway as well as the functional link between p38 MAPK activity and VEGF expression in cardiomyocyte. Previous work showed that a cardiomyocyteenriched transcription factor GATA4 is a downstream target of p38, MAPK and its transcriptional activity is induced by activated p38 MAPK (27). GATA4 is also reported to regulate VEGF expression in the heart (3). Additionally, MEF2c as another transcription factor downstream of p38 and has been reported to be required for myocardial VEGF expression and paracrine signaling during development (28). Sp1 is also a transcription factor that is activated by p38 MAPK and induces angiogenesis via VEGF induction (29,30). PGC1␣ is reported to regulate angiogenesis via VEGF in skeletal muscle and is a downstream target of p38 MAPK (31,32). COX2 is a well known downstream target of p38 MAPK in heart, and it is reported to regulate VEGF expression (33,34). HIF1␣ is one of the most well known activators of VEGF expression that is also reported to be regulated by p38 MAPK activity (35)(36)(37)(38)(39)(40). HIF1␣ expression in hearts was also reduced when treated with a p38 inhibitor in vivo (41). We found that SP1 knockdown resulted in a 20.4% decrease in p38-induced VEGF expression. However, the loss of function studies reported here indicated that GATA4, MEF2c, PGC1␣, and COX2 were not required for p38mediated VEGF induction, and loss of HIF1␣ during p38 activation only resulted in a modest but insignificant reduction of VEGF mRNA. Therefore, p38 MAPK-mediated VEGF regulation in stressed myocardium is most likely carried out in a SP1dependent mechanism but in GATA4, MEF2C, PGC1␣, COX2, and HIF1␣ independent mechanisms. Because inflammatory cytokines, including IL-1␤ and TNF␣ (42,43), are regulated by p38 MAPK and they can function as a potent inducer of VEGF expression, p38 MAPK may function through proinflammatory cytokines and induce VEGF expression indirectly via autocrine effects of other cytokines. It is intriguing that SP1 appears to be specifically responsible for p38␣ MAPK-mediated VEGF induction in heart, The specific molecular basis for such a specificity remains enigmatic, and molecular mechanisms involved p38 MAPK-mediated VEGF regulation in heart should be further investigated.

p38␣ MAP kinase regulates angiogenesis in the stressed heart
In the myocyte-specific p38 MAPK knock-out mouse hearts (CKO), VEGF expression and vascular density, as measured from immunohistochemistry and mRNA levels of vascular markers (CD31, VEGFR2, and VE-CAD), were not significantly different from the wild-type controls at the basal state. Therefore, p38 MAPK activity is not essential for normal vascular homeostasis in heart. However, myocyte-specific loss of p38 kinase activity significantly blunted angiogenic growth after chronic pressure overload as supported by histological and molecular evidences. Therefore, p38␣ activity in myocytes appears to play a specific and essential role in stress-induced compensatory vascular remodeling. The lack of compensatory vascular growth after TAC may be a potential mechanism for the accelerated heart failure observed in the p38␣ CKO mice. It raises an important issue that a stress-induced signaling pathway associated with cardiac pathological remodeling and deterioration of heart function also has a beneficial role in the heart by promoting cardiac angiogenesis. This insight reveals an important mechanism in the intricate regulatory network between cardiomyocytes and cardiac vasculature. Because p38 MAPK is induced by a variety of stressors, from mechanical to ischemia/reperfusion injury, its activity can serve as an important signal to translate cardiomyocyte stress into compensatory response from the surrounding endothelial cells. The p38 MAPK-mediated VEGF expression from cardiomyocytes and the paracrine effect on the surrounding cardiac vasculature offer an effective and elegant feedback loop to match blood supply with long-term metabolic demand in heart (Fig. 10E). The p38 MAPK pathway is a target of clinic trials for inflamma-tory disease and ischemic injury in heart. Our data implicate a potential complication for the outcome of such an approach as we now need to balance the effect of its pathological versus beneficial role in chronic remodeling in heart.

NRVM purity by flow cytometry
Twenty-four hours after culture, NRVMs were stained with an antibody directed against myosin heavy chain, MF-20, directly conjugate to FITC, or an isotype-matched FITC-conjugated control antibody, to assess the purity of cardiomyocytes in culture. Purity of immunostained NRVMs was determined by flow cytometry (6).

Cell culture
NRVMs were isolated from P1-P3 day old Sprague-Dawley rat pups of mixed gender as described previously with modifications (9). Briefly, hearts were digested with collagenase, and the resulting cell slurry was fractionated on a Percoll gradient. The myocyte-rich fraction was isolated and plated in plating media. After resting overnight in plating medium, the NRVMs were then changed to serum-free Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented with 1% insulin-transferrin-selenium (BD Bioscience) and 1% penicillin/streptomycin. Cells were incubated at 37°C in 5% CO 2 . Recombinant adenoviruses expressing constitutively activated MKK3 (MKK3b(E)) and p38␣ DN were generated and used as previously reported (9). p38 inhibitor SB202190 and MEK inhibitor U0126 was added 6 h after Adv infection.

p38␣ MAP kinase regulates angiogenesis in the stressed heart
NRVMs were infected with adenovirus at a multiplicity of infection of 25-50 particles/cell and allowed to incubate for 24 -96 h before biochemical analysis. Hypertrophy studies were conducted by treating NRVMs with phenylephrine (10 M, Sigma) for 48 h. Small interfering (si)RNA experiments were conducted as described previously (44). Briefly, cells were transfected with 150 pmol of GATA4, HIF1␣, COX2, MEF2c (Qiagen), PGC1a (Invitrogen), or SP1 (Ambion) siRNA or nonspecific scrambled siRNA (Qiagen) using Lipofectamine RNAiMAX (Invitrogen) according to the manufacture's protocol. 24 h after transfection, NRVMs were infected with adenovirus as described above and incubated for an additional 48 h. In hypoxia experiments, cells were exposed to hypoxia (1% O 2 , 5% CO 2 ) with or without p38 inhibitor for 4 h. Hypoxia incubator (from Thermo Fisher; with variable O 2 and CO 2 regulator) was used to maintain the hypoxic conditions throughout the experiment. Each experiment using cultured cell was repeated more than three times.

Quantitative RT-PCR
Total RNA was extracted from cultured cells or total left ventricular tissue using the TRIzol (Invitrogen) reagent and following the manufacturer's protocol. 0.5 g of the resultant total RNA was reverse-transcribed into cDNA using the SuperScript II RT system (Invitrogen) with oligo dT primers according to the manufacturer's instructions. mRNA levels of selected genes were determined by quantitative PCR from the generated cDNA. 50-l reactions were used with iQ SYBR Green Supermix (Bio-Rad), 250 nM concentrations of each primer, and 1 l of cDNA. The reactions were run on a MyiQ Single Color Real-Time PCR Detection System (Bio-Rad), and data were collected using iQ5 software (Bio-Rad). The cycler program used was an initial denaturation at 95°C for 5 min, 40 cycles of 45 s each of 95°C, 60°C, and 72°C, another denaturation at 95°C for 5 min, and a final product melting curve. Data were normalized to the GAPDH mRNA levels. The primer sequences are listed in Table 1.

VEGF ELISA
To determine the amount of VEGF released from NRVMs, a standard ELISA technique was used. NRVMs were treated with adenovirus as outline above, and media were collected every 24 h for a total of 96 h. Upon collection, media were treated with protease inhibitors (aprotinin 2 mg/ml, leupeptin 10 mg/ml, PMSF 100 mM), centrifuged to remove any cell debris, snapfrozen, and stored at Ϫ80°C until use. Media were then subjected to VEGF ELISA (R&D Systems) according to the man-

p38␣ MAP kinase regulates angiogenesis in the stressed heart
ufacturer's protocol. The resulting absorbance was read at 540 nm in a plate reader (BioTek PowerWave XS).

In vitro angiogenesis assay
Two independent methods were used for in vitro angiogenesis assays. First, is to use porcine aortic endothelial (PAE) cells stably expressing human VEGFR2 (PAE-VEGFR2) cultured as previously described (10). Briefly, PAE-VEGFR2 cells were cultured in F-12 medium (Cellgro) supplemented with 10% bovine growth serum and 1% penicillin/streptomycin at 37°C in 5% CO 2 . Cytodex beads (700,000 beads/ml; Sigma) were incubated with PAE-VEGFR2 cells for 4 h and were then embedded into fibrinogen/fibronectin gel (2.5 mg/ml; Sigma) containing 1 unit/l thrombin (Sigma). 2 ml of medium from adenovirustreated NRVMs was transferred directly from myocytes to gelembedded PAE-VEGFR2-coated beads. NRVMs were infected for 24 h before the first transfer, and media were transferred every 24 h thereafter for 96 h. VEGF-blocking experiments were done by removing the media from the NRVMs adding either anti-VEGF, IgG control (both R&D systems, 2 g/ml), or nothing and incubating at 37°C for 1 h before adding to the PAE culture. Second, is to use human microvascular p38␣ MAP kinase regulates angiogenesis in the stressed heart endothelial cells (HUVECs, Clonetics, Walkersville, MD) to perform endothelial cell tube formation assay as adapted from previous studies (6). Briefly, the formation of tube-like structures was assayed on tumor-derived basement membrane matrix (Millipore). HUVECs were seeded on Matrigel-coated 96-well plate in endothelial cell growth medium (5000 HUVECs/well). After 2-3 h of culture, endothelial cell growth medium was replaced with 150 l of NRVM-conditioned medium (from hypoxia or normoxia) and then incubated for 6 -8 h. After incubation, 6 -8 random microscopic (Åϳ40 magnified) view fields were pictured for each well. ImageJ software was used to calculate tube length.

Cell imaging
All myocyte and endothelial cell cultures were evaluated, and images were collected using the SPOT digital camera system (Diagnostic Instruments). Endothelial cell area was quantified using the SPOT Advanced software (Diagnostic Instruments).

Animals
All animal handling and procedures were carried out in compliance with institutional guidelines and Institutional Animal Care and Use Committees-approved protocols. Loss of p38 activity in myocytes was achieved using a conditional knockout approach as previously described (15). Briefly mice containing a floxed p38␣ allele (Lexicon Genetics, Inc.) were crossed with mice with myocyte-specific Cre recombinase expression under control of the myosin light chain 2a (MLC-2a) promoter. MLC-2a is expressed throughout the heart during the early stages of development, but its expression is down-regulated in the ventricle at later stages of development (46). Use of this promoter has previously been established as a method to successfully knock-out a gene in a cardiomyocyte-specific manner (47). p38␣ loxP/loxP Creϩ was established as a CKO animal, and their CreϪ littermates (p38␣ loxP/loxP CreϪ) were used as WT controls. Only male mice were used for these studies.

Transverse aortic constriction
TAC was performed as previously described with modifications (9). Briefly, mice were anesthetized with ketamine/ xylazine (80 and 20 mg/kg, respectively) by intraperitoneal injection and mechanically ventilated with 95% O 2 (tidal volume 0.5 ml, 130 breaths/min). A left parasternal thoracotomy was performed to access the transverse aorta, which was tied with a 5-0 nylon suture on a 27-gauge needle. After needle removal, the transverse aortic lumen remained constricted by 65-70%. Sham mice underwent a similar procedure except the aorta was not tied. All surgeries were performed on male mice at the age of 15-18 weeks old.

Histology
The mouse hearts were perfused and fixed with 10% formalin before embedding in paraffin. All sections were cross-sections of the heart taken from the midpoint of the ventricle. To obtain capillary density images, mice were injected with 200 l of FITC-conjugated lectin (Vector Laboratories) diluted 1:2 with phosphate-buffered saline. The hearts were fixed as outlined above. Samples were counter-stained with anti-tropomyosin (Sigma) primary antibody and Alexa 568 (Molecular Probes) secondary. To obtain cross-sectional area image, sections were incubated with Alexa 594-conjugated WGA (Life Technologies) and anti-tropomyosin primary antibody/Alexa 488 secondary. Nuclei were counter-stained with Hoechst 33342 (Molecular Probes). Images were collected using confocal microscopy.

Trichrome staining
Heart sections were stained with Trichrome stain kit (Sigma) according to manufacture protocol. Images were collected using microscopy.

p38␣ MAP kinase regulates angiogenesis in the stressed heart Statistical analysis
Comparisons between more than two groups were accomplished using a factorial analysis of variance (ANOVA) with Fisher's protected least significant difference post-hoc test. Comparisons between two groups were accomplished using an unpaired two-tailed t test. Analysis was carried out using the StatView program (Abacus Concepts, Berkeley, CA). In all cases a significant result was defined as p Ͻ 0.05.
Author contributions-B. A. R. and T. Y. contributed to the experimental design, data collection, analysis interpretation, and manuscript preparation. V. C. and A. Y. K. contributed to the hypoxiarelated experiments. S. R. contributed to animal surgery. L. I.-A. and S. M. contributed to manuscript preparation and experimental design. Y. W. contributed to experimental design, data interpretation, and manuscript preparation.