LY2228820 Dimesylate, a Selective Inhibitor of p38 Mitogen-activated Protein Kinase, Reduces Angiogenic Endothelial Cord Formation in Vitro and in Vivo

Background: Angiogenesis is a critical process for tumor growth and survival.

Results: LY2228820 dimesylate, a p38 MAPK-specific inhibitor, or shRNA knockdown of p38α, MK2, or HSP27 reduced angiogenic cord formation.

Conclusion: p38 MAPK modulated soluble factors released from stromal and tumor cells and reduced their downstream signaling in endothelial cells.

Significance: Antiangiogenic activity of LY2228820 dimesylate may lead to anti-tumor growth effects.

Abstract

LY2228820 dimesylate is a highly selective small molecule inhibitor of p38α and p38β mitogen-activated protein kinases (MAPKs) that is currently under clinical investigation for human malignancies. p38 MAPK is implicated in a wide range of biological processes, in particular those that support tumorigenesis. One such process, angiogenesis, is required for tumor growth and metastasis, and many new cancer therapies are therefore directed against the tumor vasculature. Using an in vitro co-culture endothelial cord formation assay, a surrogate of angiogenesis, we investigated the role of p38 MAPK in growth factor- and tumor-driven angiogenesis using LY2228820 dimesylate treatment and by shRNA gene knockdown. p38 MAPK was activated in endothelial cells upon growth factor stimulation, with inhibition by LY2228820 dimesylate treatment causing a significant decrease in VEGF-, bFGF-, EGF-, and IL-6-induced endothelial cord formation and an even more dramatic decrease in tumor-driven cord formation. In addition to involvement in downstream cytokine signaling, p38 MAPK was important for VEGF, bFGF, EGF, IL-6, and other proangiogenic cytokine secretion in stromal and tumor cells. LY2228820 dimesylate results were substantiated using p38α MAPK-specific shRNA and shRNA against the downstream p38 MAPK effectors MAPKAPK-2 and HSP27. Using in vivo models of functional neoangiogenesis, LY2228820 dimesylate treatment reduced hemoglobin content in a plug assay and decreased VEGF-A-stimulated vascularization in a mouse ear model. Thus, p38α MAPK is implicated in tumor angiogenesis through direct tumoral effects and through reduction of proangiogenic cytokine secretion via the microenvironment.

Introduction

The p38 mitogen-activated protein kinases (MAPKs) are strongly activated by stress and inflammatory cytokines, leading to modulation of many cellular functions, including proliferation, differentiation, and survival (1). Four different p38 MAPK isoforms have been identified, p38α, -β, -γ, and -δ, which may have both overlapping and specific functions (1, 2). p38α MAPK (p38α) and p38β MAPK (p38β) are ubiquitously expressed, whereas p38γ MAPK and p38δ MAPK demonstrate specific tissue expression. p38α, the most abundant isoform, is present in most cells and is exclusively critical for mouse development (3–5). Upstream p38 MAPK kinases (MKKs),² such as MKK3 and MKK6, can differentially regulate p38 isoforms, as evidenced by the inability of MKK3 to effectively activate p38β (6). A major substrate for p38 MAPK is MAPK-activated protein kinase-2 (MAPKAPK-2; MK2), a serine threonine kinase that directly phosphorylates the ubiquitously expressed heat shock protein 27 (HSP27). HSP27 is activated in response to osmotic stress, reactive oxygen species, and inflammatory cytokines and may play a role in cell migration, apoptosis, and actin cytoskeleton organization (7).

Previous reports indicate a role for p38 MAPK in a wide range of biological processes, in particular tumor cell proliferation in vitro and in vivo (8, 9). Angiogenesis is required for tumor growth and metastasis; therefore, many new potential cancer therapies are directed against the tumor vasculature. Angiogenesis is the formation of vascular tubes composed of an inner lining of endothelial cells, and, as they mature, vessels acquire a coating of perivascular cells (referred to as pericytes, smooth muscle cells, or mural cells) that envelop the surface of the vascular tube and are critical for the development and maintenance of the vasculature (10–11). Angiogenesis is stimulated by a variety of soluble factors, including vascular endothelial growth factor (VEGF), basic fibroblast growth factor (bFGF), endothelial growth factor (EGF), and interleukin 6 (IL-6) (12, 13). Endothelial cells and pericytes communicate via cytokine signaling, and pericytes play a role in maintaining the integrity of endothelial cells by serving as support structures (14). In addition to vascular stabilization, pericytes are important for modulation of endothelial cell migration, proliferation, and survival (11, 15). Previous findings suggest a role for p38 MAPK in modulating tumor angiogenesis in tumor cells and/or host endothelial cells (7, 16–20), but this potential role is not well defined.

We investigated the role of p38 MAPK in individual cytokine and tumor-driven angiogenesis through pharmacological inhibition of p38 MAPK using LY2228820 dimesylate treatment and by shRNA gene knockdown. LY2228820 dimesylate (Fig. 1A) is a potent small molecule ATP-competitive inhibitor of p38 MAPK that is highly selective for the p38α and p38β isoforms and is currently under clinical investigation for human malignancies. We report that the p38 MAPK pathway is activated in endothelial cells in response to VEGF, bFGF, EGF, and IL-6 stimulation and is involved in individual cytokine-driven and tumor-driven cord formation. Inhibition of p38 MAPK in tumor cells also led to decreased secretion of VEGF, bFGF, EGF, IL-6, IL-8, and other proangiogenic factors. Small molecule inhibitor results were substantiated by shRNA knockdown of p38α and downstream p38 MAPK effectors MK2 and HSP27 but not p38β. Consequently, p38α plays a role in endothelial cell angiogenesis along with stromal and tumor cell cytokine secretion. LY2228820 dimesylate treatment yielded antiangiogenic effects in vivo via decreased hemoglobin content in a Matrigel^TM plug assay, a measure of functional neoangiogenesis, and decreased VEGF-A-stimulated vascularization in a mouse ear model. p38α and its downstream effectors, MK2 and HSP27, are therefore implicated in tumor angiogenesis, and p38α plays an integral role in key proangiogenic cytokine secretion.

FIGURE 1.

LY2228820 dimesylate treatment reduced VEGF-, bFGF-, EGF-, and IL-6-driven cord formation. A, chemical structure of LY2228820 dimesylate (LY). B, whole cell protein extracts were isolated from ECFCs or ADSCs following pretreatment with DMSO (−) or 1 μm LY2228820 dimesylate (+) for 30 min prior to the addition of 10 ng/ml VEGF, bFGF, EGF, or 100 ng/ml IL-6, and then the extracts were subjected to Western blot analysis using antisera directed against p-p38, p38α, p38β, p-MK2, total MK2, p-HSP27, total HSP27, and β-actin as a loading control. C, whole cell protein extracts were isolated from ECFCs following 15-min PBS (basal), 10 ng/ml VEGF, bFGF, EGF, or 100 ng/ml IL-6 treatment, and then extracts were subjected to p-p38 and p-MK2 analysis by a phosphoprotein immunoassay. Graphs represent means ± S.E. (error bars) from three independent experiments, and asterisks denote statistically significant (*, p < 0.05) differences compared with basal controls. D, the ADSC/ECFC co-cultures were treated with DMSO or 1 μm LY2228820 dimesylate simultaneously with PBS (basal) or 10 ng/ml VEGF, bFGF, EGF or 100 ng/ml IL-6 for 96 h prior to immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (×5 magnification) are shown; graphs represent means ± S.E. after basal cord formation data were subtracted from VEGF, bFGF, EGF, and IL-6 data from three independent experiments; and asterisks denote statistically significant (*, p < 0.05) differences compared with DMSO controls. E, conditioned medium was collected from the ADSC/ECFC co-cultures or ADSCs alone treated with DMSO or 1 μm LY2228820 dimesylate for 72 h and subjected to ELISA analysis for VEGF, bFGF, EGF, and IL-6. Graphs represent means ± S.E. from three independent experiments, and asterisks denote statistically significant (*, p < 0.05) differences compared with DMSO controls.

EXPERIMENTAL PROCEDURES

Cell Culture

U-87-MG, MDA-MB-231, SK-OV-3, A-2780, NCI-H1650, and PC-3 cells were grown according to the American Type Culture Collection (ATCC, Manassas, VA) guidelines. LXFA-629 non-small cell lung adenocarcinoma cells (Oncotest, Freiburg, Germany) were maintained in RPMI 1640 medium supplemented with 10% heat-inactivated FBS and 1% glutamine (all from Invitrogen). All cells were grown and treated in uncoated tissue culture-treated flasks in a humidified atmosphere at 37 °C and 5% CO₂.

shRNA Knockdown

U-87-MG and MDA-MB-231 cells were transduced (multiplicity of infection 9) with MISSION® shRNA lentiviral transduction particles (Sigma-Aldrich) (non-target control, SCH202V; p38α, NM_001315; p38β, NM_002751), selected with 5 μg/ml puromycin, and screened for protein knockdown by Western blot analysis as described below. Adipose-derived stem cell (ADSC)/endothelial colony-forming cell (ECFC) co-cultures were transduced following ECFC plating in cord formation as described below with 30 μl of MISSION® shRNA lentiviral transduction particles (Sigma-Aldrich) (non-target control, SCH202V; p38α, NM_001315; p38β, NM_002751; MK2, NM_032960; HSP27, NM_001540) for 72 h prior to analysis for Western blot, cord formation, cytokine secretion, or phosphoprotein immunoassay as described below.

In Vitro Cord Formation Assay

ADSCs (Zen-Bio, Research Triangle Park, NC) were plated at 75,000 cells/well into 96-well HTS Transwell® (Corning Inc.) receiver plates (tumor-driven) or 50,000 cells/well (growth factor-driven) into 96-well black poly-d-lysine-coated plates, and tumor cells were plated at 25,000 cells/well in 96-well HTS Transwell® (Corning Inc.) plates in co-culture medium (MCDB-131 medium (Invitrogen) supplemented with l-ascorbic acid 2-phosphate, dexamethasone, tobramycin, insulin (all from Sigma-Aldrich), and CellPrime rTransferrin AF (Millipore, Ballerica, MA)) for 24 h. ADSC medium was removed, and 6,000 (tumor-driven) or 5,000 (growth factor-driven) human ECFCs (Lonza, Basel, Switzerland) per well were overseeded. Treatment with 10 ng/ml VEGF, bFGF, or EGF or 100 ng/ml IL-6 (all from Invitrogen) and DMSO or 1 μm LY2228820 dimesylate treatment occurred 4 h following ECFC plating and continued for 96 h. Cells were directly fixed for 10 min with 3.7% formaldehyde (Sigma-Aldrich) followed by ice-cold 70% ethanol for 20 min at 25 °C. Cells were rinsed once with PBS, blocked for 30 min with 1% BSA, and immunostained for 1 h with antiserum directed against cluster of differentiation 31 (CD31) (R&D Systems, Minneapolis, MN) diluted to 1 μg/ml in 1% BSA. Cells were washed three times with PBS and incubated for 1 h with 5 μg/ml donkey α-sheep-Alexa-488 (Invitrogen), α-smooth muscle actin Cy3 conjugate (1:200; Sigma-Aldrich), and 200 ng/ml Hoechst 33342 (Invitrogen) in 1% BSA, washed with PBS, and then imaged using the cord formation algorithm on the Cellomics® ArrayScan® VTI at an image magnification of ×5 (Thermo Fisher Scientific). For assessment of proliferation and apoptosis, cells were plated, treated, and fixed for cord formation as mentioned above and then immunostained with Ki67 (1:100; Millipore), 5 μg/ml goat α-rabbit-Alexa-647 (Invitrogen), and 200 ng/ml Hoechst 33342 (Invitrogen) or the In Situ Cell Death Detection Kit (Roche Applied Science) according to the manufacturer's recommendations and then imaged using the target activation algorithm on the Cellomics® ArrayScan® VTI at an image magnification of ×20 (Thermo Fisher Scientific). Cell motility was analyzed using the Cellomics® cell motility kit (Thermo Fisher Scientific) by plating 500 ADSC or ECFC cells on prepared blue fluorescent microsphere plates according to the manufacturer's recommendations. Treatment with DMSO or 1 μm LY2228820 dimesylate and 10 ng/ml VEGF, bFGF, or EGF or 100 ng/ml IL-6 occurred 24 h following cell plating and continued for 18 h. Cells were then fixed, stained, and imaged using the cell motility algorithm on the Cellomics® ArrayScan® VTI at an image magnification of ×20 (Thermo Fisher Scientific) according to the manufacturer's recommendations.

Western Blot

Whole cell protein extracts were isolated by cell lysis with 1% SDS and brief sonication, and protein concentration was quantified using the Bradford method. Thirty micrograms of protein were subjected to electrophoresis on 4–20% precast Tris-glycine gradient gels (Invitrogen), transferred to nitrocellulose (Invitrogen), blocked with 5% blotting grade blocker (Bio-Rad) in Tris-buffered saline containing 0.1% Tween (TBST), probed with primary antiserum, washed with TBST, and incubated with an appropriate horseradish peroxidase-labeled secondary antibody. Membranes were washed with TBST, and signal was detected by ECL (Thermo Fisher Scientific). Antisera directed against p38α, p38β, and total HSP27 (all from Cell Signaling Technology, Danvers, MA); phospho-p38 MAPK (Thr-180/Tyr-182), phospho-HSP27 (Ser-15), phospho-MAPKAPK2 (Thr-334), and total MAPKAPK2 (all from Epitomics, Burlingame, CA); and β-actin (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were diluted with 5% blotting grade blocker (Bio-Rad) in TBST. Densitometry was performed using ImageJ analysis software (National Institutes of Health) as per the request of the ImageJ developers.

Phosphoprotein Immunoassays

ADSCs/ECFCs were plated following the cord formation protocol above, or ECFCs were plated in normal growth medium (2 × 10⁴) in 96-well tissue culture dishes for 24 h prior to replacement of medium with co-culture medium for 2 h. ADSCs/ECFCs or ECFCs were then stimulated for 15 min with 10 ng/ml VEGF, bFGF, EGF, 100 ng/ml IL-6, or conditioned medium from U-87-MG or MDA-MB-231 cells (described below) and analyzed with the phospho-p38 (Thr-180, Tyr-182) or phospho-MAPKAPK2 (Thr-334) whole cell kit according to the manufacturer's recommendations (Meso Scale Discovery, Rockville, MD).

Cytokine Analysis

ADSCs or ADSCs/ECFCs were plated in co-culture medium following the cord formation protocol above. Tumor cells (2 × 10⁵) were plated in co-culture medium in 6-well tissue culture dishes. Four hours after ECFC plating or 24 h after tumor cell plating, medium was replaced, and treatments were added for 72 h prior to medium collection and cell number counts. Cell debris was removed from conditioned medium by centrifugation, and samples were analyzed fresh or were frozen at −20 °C until analysis. Samples were analyzed with Quantikine® Colorimetric Sandwich ELISAs (R&D Systems) according to the manufacturer's recommendations.

In Vivo Matrigel^TM Plug Angiogenesis Assay

ADSCs (0.5 × 10⁶) and ECFCs (2 × 10⁶) were mixed with 200 μl of growth factor-reduced Matrigel^TM (BD Biosciences) on ice and subcutaneously injected into the flanks of athymic nude female mice (Harlan, Indianapolis, IN), one implant per animal. Mice were dosed orally three times daily with LY2228820 dimesylate (20 and 40 mg/kg) or twice daily with sunitinib (25 mg/kg), which were prepared internally, beginning 4 h prior to cell implantation. After 5 days of dosing, implants were removed and flash-frozen in liquid nitrogen, and hemoglobin was quantified using the QuantiChrom^TM hemoglobin assay kit (Bioassay, Hayward, CA) as described previously (21).

Angiogenesis Ear Assay

Animal protocols were approved by the Imclone System Inc. and Mispro Biotech Services Corporation Animal Care and Use Committee. A nonreplicating adenoviral vector engineered to express the predominant (164-amino acid) murine isoform of VEGF-A (Ad-VEGF-A¹⁶⁴; 1 × 10⁸ plaque-forming units), as described previously (22), was injected intradermally into the dorsal ears of 8-week-old athymic nu/nu mice (Charles River, Wilmington, MA) as described previously (23). Mice (n = 5) were dosed orally with sunitinib at 40 mg/kg daily, LY2228820 dimesylate at 30 mg/kg twice a day, or vehicle (HEC-Tween) twice a day starting 1 day before injection of adenovirus VEGF-A and harvested at day 5 after adenovirus injection. Ears were mounted flat under a glass slide with immersion oil and photographed using a Leica M80 photomicroscope as described previously (23). The images of ear vasculature were quantified by Image-Pro Analyzer version 7.0 (MediaCybernetics, Bethesda, MD).

Statistical Analysis

Statistical significance of data were assessed by a two-tailed Student's t test with equal variance compared with data obtained for DMSO or non-target shRNA controls (in vitro) or vehicle controls (in vivo). Statistical significance was assigned to p values of <0.05.

RESULTS

VEGF, bFGF, EGF, and IL-6 Activate p38 MAPK Signaling

LY2228820 dimesylate (Fig. 1A) is a highly selective ATP-competitive inhibitor of p38α and p38β that does not alter p38 MAPK activation but reduces downstream p38 MAPK signaling (24). The human kinome map of LY2228820 dimesylate activity indicates the specificity of the inhibitor for p38 MAPK compared with that of sunitinib, a multitargeted receptor tyrosine kinase inhibitor with an antiangiogenic mechanism of action (25) (supplemental Fig. 1). Smooth muscle cells and endothelial cells are exposed to many growth factors and proinflammatory cytokines that contribute to angiogenesis, and p38 MAPK is implicated in downstream cytokine signaling (18, 26); therefore, p38 MAPK signaling was analyzed in ECFCs, a subtype of umbilical cord blood-derived endothelial cells that can form intrinsic in vivo vessels upon transplantation into immunodeficient mice (27), and ADSCs, which are similar to mesenchymal stem cells, which can give rise to cells with pericytic properties that can stabilize vascular assembly in vitro (28). In ECFCs, there was a robust increase in VEGF-dependent phosphorylated p38 MAPK (p-p38) expression (p < 0.05), whereas bFGF, EGF, and IL-6 showed a modest (∼25%) increase in p-p38 by a highly sensitive immunoassay (Fig. 1C). Importantly, ligand-induced up-regulation of p-p38 in ECFCs resulted in increased expression of the downstream effectors p-MK2 and p-HSP27 by Western blot (Fig. 1B) and a significant increase (p < 0.05) in p-MK2 by immunoassay (Fig. 1C). Activation of these effectors was significantly impaired following inhibition of p38 MAPK by LY2228820 dimesylate treatment (Fig. 1B). Furthermore, LY2228820 dimesylate treatment abrogated basal p38 pathway activity in ECFCs and ADSCs (Fig. 1B).

LY2228820 Dimesylate Treatment Reduced VEGF-, bFGF-, EGF-, and IL-6-driven Cord Formation

Because p38 pathway activation occurs following treatment with proangiogenic factors, a surrogate cord formation assay to model key morphogenic features of blood vessel formation (29) was used to analyze the effect of LY2228820 dimesylate treatment on in vitro endothelial cord formation. The ECFC and ADSC co-culture can establish cord networks in vitro and functional blood vessels in vivo (21). ADSCs, which serve as the feeder layer for the ECFCs, can differentiate into pericyte-like cells that express α-smooth muscle actin, a contractile filament expressed in pericytes. Formation of vascular networks by ECFCs can be visualized by CD31 immunostaining (30), accompanied by ADSC migration and increased density near the cords along with increased α-smooth muscle actin expression. LY2228820 dimesylate treatment significantly reduced (p < 0.05) basal and VEGF-, bFGF-, EGF-, and IL-6-driven cord formation along with α-smooth muscle actin expression (Fig. 1D). In addition, cords that were established without stimulation (basal) or with VEGF, bFGF, EGF, and IL-6 for 4 days prior to LY2228820 dimesylate treatment showed significant regression (p < 0.05) in both cord formation and α-smooth muscle actin expression (supplemental Fig. 2). Cell viability of the ADSC/ECFC co-culture was not altered by LY2228820 dimesylate treatment, which ensured that the observed effects on cord formation were not due to cell toxicity (data not shown).

Endothelial cells require cytokine stimulation for survival, proliferation, and migration, all of which are essential for angiogenesis, and the paracrine signaling between endothelial cells and pericytes is important for cord formation. In addition, endothelial cells secrete factors, partly to attract pericytes to envelop the vessel wall and promote vessel maturation (31); therefore, a role for p38 MAPK on ADSC/ECFC co-culture, ADSC, and ECFC cytokine secretion was analyzed. The ADSC/ECFC co-culture along with ADSCs alone showed a significant reduction (p < 0.05) in VEGF, bFGF, EGF, and IL-6 secretion upon LY2228820 dimesylate treatment (Fig. 1E). ECFCs secreted undetectable to minimal amounts (<8 pg/ml) of VEGF, bFGF, EGF, and IL-6 (data not shown). However, the amount of VEGF, bFGF, EGF, and IL-6 was slightly enhanced (∼10–30%) in the ADSC/ECFC co-cultures compared with ADSCs alone, indicating that cell contact between ADSCs and ECFCs may be important for cytokine secretion. These data indicate that p38 MAPK is involved both in downstream cytokine signaling and cytokine release from stromal cells. Statistically significant decreases (p < 0.05) in ECFC proliferation, assessed by expression of the proliferation marker Ki67, ECFC motility, or ECFC apoptosis, assessed by TUNEL analysis, were not observed with LY2228820 dimesylate treatment upon VEGF, bFGF, EGF, or IL-6 stimulation (supplemental Fig. 3).

Knockdown of p38α, MK2, or HSP27 Reduced VEGF-, bFGF-, EGF-, and IL-6-driven Cord Formation

The use of a p38α/β-selective inhibitor, such as LY2228820 dimesylate, is critical in establishing a role for p38 MAPK function in angiogenesis but does not distinguish between the functions of p38α and p38β. To further investigate a role for the p38 MAPK pathway in angiogenesis, gene knockdown using shRNA targeted against p38α and p38β in ADSC/ECFC co-cultures were generated. Knockdown of p38α but not p38β in ADSCs/ECFCs was effective in blocking p38 MAPK signaling, as evidenced by reduced levels of p-p38, p-MK2, and p-HSP27 by Western blot (Fig. 2B) and by immunoassay (Fig. 2D), which led to a significant reduction (p < 0.05) in VEGF-, bFGF-, EGF-, and IL-6-driven (knockdown of p38β also significantly reduced IL-6-driven cord formation but to a lesser extent than p38α) cord formation (Fig. 2A). These results are similar to those of LY2228820 dimesylate treatment, which strengthens the notion that compound effects are specific to inhibition of p38α. To further support a role for p38 MAPK and the p38 MAPK pathway in cord formation, knockdown of downstream signaling effectors MK2 and HSP27 in the ADSC/ECFC co-culture also significantly reduced (p < 0.05) VEGF-, bFGF-, EGF-, and IL-6-driven cord formation (Fig. 2A). Knockdown of p38α, MK2, or HSP27 but not p38β also significantly reduced (p < 0.05) cytokine secretion of VEGF, bFGF, and IL-6 from ADSC/ECFC co-cultures (Fig. 2C). Secretion of EGF from ADSC/ECFC co-cultures was significantly reduced (p < 0.05) with p38α knockdown and was only slightly reduced with MK2 or HSP27 knockdown (Fig. 2C). Knockdown of p38α significantly inhibited (p < 0.05) activation of p-p38 and p-MK2 following VEGF stimulation (Fig. 2D). Knockdown of MK2 also significantly reduced (p < 0.05) basal and VEGF-induced p-MK2 expression (Fig. 2D). In contrast, knockdown of p38β, MK2, or HSP27 did not alter VEGF-induced activation of p-p38 by immunoassay analysis (Fig. 2D). This indicates that p38α and not p38β is the main mediator of VEGF, EGF, bFGF, and IL-6 cytokine secretion and downstream signaling through MK2 and HSP27 in our proangiogenic co-culture system. As observed with LY2228820 dimesylate treatment, cell viability of ADSC/ECFC co-cultures was not altered upon shRNA treatment; therefore, the anti-cord forming effects did not stem from a cytotoxic event (data not shown). Similar results were obtained with two additional shRNA clones targeting different mRNA regions of p38α, p38β, MK2, and HSP27, indicating that the effects observed are probably due to reduced expression of the intended target genes (data not shown).

FIGURE 2.

Knockdown of p38α, MK2, or HSP27 reduced VEGF-, bFGF-, EGF-, and IL-6-driven cord formation. A, the ADSC/ECFC co-cultures were treated with non-targeting (control), p38α, p38β, MK2, or HSP27 shRNA for 72 h prior to induction of cord formation without (PBS, basal) or with 10 ng/ml VEGF, bFGF, EGF, or 100 ng/ml IL-6 for 96 h before immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (×5 magnification) are shown; graphs represent means ± S.E. (error bars) from three independent experiments after basal cord formation data were subtracted from VEGF, bFGF, EGF, and IL-6 data; and asterisks denote statistically significant differences (*, p < 0.05) compared with non-targeting shRNA controls. B, whole cell protein extracts were isolated from the ADSC/ECFC co-cultures following 72-h shRNA treatment for the indicated gene and subjected to Western blot analysis using antisera directed against p38α, p38β, p-p38, p-MK2, total MK2, p-HSP27, total HSP27, and β-actin as a loading control, and protein quantification was determined with densitometry. C, conditioned medium was collected from the ADSC/ECFC co-cultures following 72 h shRNA treatment for the indicated gene and subjected to ELISA analysis for VEGF, bFGF, EGF, and IL-6. Graphs represent means ± S.E. from three independent experiments, and asterisks denote statistically significant (*, p < 0.05) differences compared with DMSO controls. D, whole cell protein extracts were isolated from ECFCs following 72-h shRNA treatment for the indicated gene following by 15 min of PBS (basal) or 10 ng/ml VEGF treatment, and then extracts were subjected to p-p38 and p-MK2 analysis by a phosphoprotein immunoassay. Graphs represent means ± S.E. from three independent experiments, and asterisks denote statistically significant (* and #, p < 0.05) differences compared with respective shRNA control PBS-treated (*) or shRNA control VEGF-treated (#) samples.

LY2228820 Dimesylate Treatment Reduced Tumor-driven Cord Formation

To more closely represent tumor angiogenesis, instead of individual cytokines stimulating cord formation, LY2228820 dimesylate effects on tumor-conditioned medium-driven and tumor cell-driven cord formation were analyzed. Conditioned media from commonly used, well characterized U-87-MG glioblastoma and MDA-MB-231 breast cancer cells increased protein expression of p-p38, p-MK2, and p-HSP27 by Western blot and significantly increased (p < 0.05) p-p38 and p-MK2 by immunoassay (supplemental Fig. 4, A and B). This indicates that cytokines secreted from U-87-MG and MDA-MB-231 tumor cells activate p38 MAPK signaling. LY2228820 dimesylate treatment significantly reduced (p < 0.05) U-87-MG and MDA-MB-231 tumor-conditioned medium-driven cord formation, indicating a function for p38 MAPK in stromal cells downstream of tumor-secreted cytokines (supplemental Fig. 4C). LY2228820 dimesylate also significantly reduced (p < 0.05) tumor-driven cord formation and smooth muscle actin expression from a range of tumor histologies, including U-87-MG, MDA-MB-231, ovarian (A-2780 and SK-OV-3), lung (LXFA-629 and NCI-H1650), and prostate (PC-3) (Fig. 3A). LY2228820 dimesylate treatment inhibited downstream p38 MAPK signaling (p-MK2 and p-HSP27) in U-87-MG and MDA-MB-231 cells (Fig. 3B) along with each tumor cell line analyzed (data not shown). In tumor-driven cord formation, LY2228820 dimesylate effects on the tumor cells and downstream cytokine signaling could not be separated because all three cell lines (tumor, ADSC, and ECFC) were concurrently treated with LY2228820 dimesylate, necessitating pretreatment of U-87-MG or MDA-MB-231 tumor cells with LY2228820 dimesylate prior to cell plating. Pretreatment of tumor cells with LY2228820 dimesylate significantly reduced cord formation (supplemental Fig. 5) and secretion (p < 0.05) of VEGF, bFGF, EGF, and IL-6 from U-87-MG (Fig. 3C) along with A-2780, SK-OV-3, and PC-3 tumor cell lines (data not shown) and VEGF, bFGF, and IL-6 from MDA-MB-231 tumor cells (Fig. 3C; basal EGF is below the limit of detection in MDA-MB-231 cells). LY2228820 dimesylate treatment also reduced secretion of IL-8 and other proangiogenic cytokines (angiogenin, HGF, PlGF, PDGF-AA) secreted from U-87-MG, MDA-MB-231, SK-OV-3, and A-2780 tumor cells (data not shown). Pretreatment of tumor cells with LY2228820 dimesylate and the addition of compound into the cord formation assay led to the greatest inhibition of cord formation (supplemental Fig. 5), further supporting the notion that LY2228820 dimesylate treatment has a direct effect on tumor cell cytokine secretion and cytokine signaling in ADSC/ECFC cells, especially because cell viability of tumor cells and ADSC/ECFC co-cultures were unchanged (data not shown).

FIGURE 3.

LY2228820 dimesylate treatment reduced tumor-driven cord formation. A, ADSC/ECFC co-cultures with permeable transwells containing medium (no cells) or the indicated tumor cells (U-87-MG, MDA-MB-231, A-2780, SK-OV-3, LXFA-629, NCI-H1650, and PC-3) were treated with DMSO or 1 μm LY2228820 dimesylate (LY) for 96 h prior to immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (×5 magnification) are shown; graphs represent means ± S.E. (error bars) from three independent experiments after no cells (basal) data were subtracted; and asterisks denote statistically significant (*, p < 0.05) differences compared with DMSO controls. B, whole cell protein extracts were isolated from the indicated tumor cells following treatment with DMSO (−) or 1 μm LY2228820 dimesylate (+) for 4 h and subjected to Western blot analysis using antisera directed against p-p38, p38α, p38β, p-MK2, total MK2, p-HSP27, total HSP27, and β-actin as a loading control. C, conditioned medium was collected from U-87-MG or MDA-MB-231 tumor cells treated with DMSO or 1 μm LY2228820 dimesylate for 72 h and subjected to ELISA analysis for VEGF, bFGF, EGF, and IL-6. Graphs represent means ± S.E. from three independent experiments, and asterisks denote statistically significant (*, p < 0.05) differences compared with DMSO controls.

Knockdown of p38α in Tumor Cells Reduced Tumor-driven Cord Formation

To further investigate the role of p38α and p38β in tumor cell cytokine secretion and tumor-induced cord formation, stable knockdown of p38α or p38β was assessed in U-87-MG and MDA-MB-231 cells. It is important to note that complete protein knockdown was not achieved for either isoform, which may have affected results on cord formation. Knockdown of p38α but not p38β in both cell lines reduced expression of p-p38, p-MK2, and p-HSP27 (Fig. 4B and supplemental Fig. 6B), and significantly reduced (p < 0.05) tumor-driven cord formation (Fig. 4A and supplemental Fig. 6A) along with VEGF, bFGF, EGF (in U-87-MG, undetectable in MDA-MB-231), and IL-6 secretion (Fig. 4C and supplemental Fig. 6C). Similar results were obtained with additional U-87-MG and MDA-MB-231 stable shRNA knockdown lines, which confirmed the importance of p38α in controlling tumor cell cytokine secretion and cord formation.

FIGURE 4.

Knockdown of p38α MAPK in U-87-MG tumor cells reduced tumor-driven cord formation. A, stable U-87-MG shRNA cell lines for a non-targeting shRNA (control), p38α, p38β, or medium (no cells) were plated in permeable transwells with ADSC/ECFC co-cultures for 96 h and then subjected to immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (×5 magnification) are shown, graphs represent means ± S.E. (error bars) from three independent experiments after no cells data (basal cord formation) were subtracted, and asterisks denote statistically significant (*, p < 0.05) differences compared with the non-targeting shRNA control. B, whole cell protein extracts were isolated from the indicated cell lines and subjected to Western blot analysis using antisera directed against p38α, p38β, p-p38, p-MK2, total MK2, and β-actin as a loading control. C, conditioned medium was collected from stable U-87-MG shRNA cell lines and subjected to ELISA analysis for VEGF, bFGF, EGF, and IL-6. Graphs represent means ± S.E. from three independent experiments, and asterisks denote statistically significant (*, p < 0.05) differences compared with non-targeting shRNA controls.

Cord Formation Rescue with the Addition of VEGF, bFGF, EGF, and IL-6 in Conditioned Media from Tumor Cells with Stable p38α Knockdown

To determine if the reduction in VEGF, bFGF, EGF (U-87-MG), and IL-6 is contributing to the reduction in cord formation observed with stable knockdown of p38α in U-87-MG and MDA-MB-231 tumor cells, we performed an add back experiment. Conditioned media from control shRNA cells or p38α stable knockdown cells for U-87-MG and MDA-MB-231 cells were collected and analyzed for VEGF, bFGF, EGF (U-87-MG cells only; MDA-MB-231 cells had undetectable amounts of EGF), and IL-6 cytokine levels. The addition of 1× or 2× amounts (compared with the control shRNA conditioned medium) of the individual deficient cytokines or a mixture of the deficient cytokines to the p38α stable knockdown conditioned medium was used to assess cord formation. Importantly, the addition of a 1× or 2× mixture of VEGF, bFGF, EGF (U-87-MG only), and IL-6 significantly increased (p < 0.05) cord formation compared with the U-87-MG or MDA-MB-231 p38α knockdown tumor conditioned medium (supplemental Fig. 7). This indicates that reduction in VEGF, bFGF, and, to a lesser extent, EGF and IL-6 upon p38α knockdown in tumor cells contributes to the reduction in cord formation observed.

LY2228820 Dimesylate Treatment Reduced Hemoglobin Content and Ear Angiogenesis in Vivo

To extend a role for the p38 MAPK pathway in angiogenesis in vivo, we tested LY2228820 dimesylate in a neoangiogenesis Matrigel^TM plug model consisting of ADSCs/ECFCs that form blood vessels following co-implantation into the flank of a nude mouse (28). Five days after implantation, ADSC/ECFC cells formed extensive networks of blood vessels whose functionality was assessed by measuring hemoglobin content. LY2228820 dimesylate and sunitinib, an approved angiogenesis inhibitor used as a positive control, were given at clinically relevant doses, and both caused a significant reduction (p < 0.05) in hemoglobin content (Fig. 5A).

FIGURE 5.

LY2228820 dimesylate treatment reduced hemoglobin content and ear vascularization in vivo. A, an ADSC/ECFC cell mixture was co-implanted subcutaneously into the flanks of athymic nude mice (8 mice/treatment group). Oral dosing of mice began 4 h prior to cell implantation and occurred three times daily with LY2228820 dimesylate (LY) (20 and 40 mg/kg) or twice daily with sunitinib (25 mg/kg). After 5 days of dosing, Matrigel^TM plugs were removed, and hemoglobin was quantified. The graph is representative of three independent experiments and indicates means ± S.E. (error bars) from one experiment. Asterisks denote statistically significant (*, p < 0.05) differences compared with vehicle controls. B, mice were dosed orally with vehicle (HEC-Tween), LY2228820 dimesylate at 30 mg/kg twice a day, or sunitinib at 40 mg/kg daily, starting 1 day before injection of adenovirus VEGF-A (Ad-VEGF-A¹⁶⁴). Ears were harvested 5 days after adenovirus injection and imaged (representative images from two independent experiments are shown; ×8 magnification), and vasculature was quantified. The graph is representative of two independent experiments, indicates means ± S.E., and asterisks denote statistically significant (*, p < 0.05) differences compared with vehicle controls.

To further analyze the relevant effects of p38 MAPK on angiogenesis in vivo, we tested LY2228820 dimesylate in an ear angiogenesis model consisting of intradermal injection of Ad-VEGF-A¹⁶⁴ into nude mouse ears that induces a robust angiogenic response (23). Similar to the Matrigel^TM plug assay, both LY2228820 dimesylate and sunitinib treatment caused a significant reduction (p < 0.05) in ear vascularity (Fig. 5B), which indicates that LY2228820 dimesylate treatment impaired neoangiogenesis in vivo. All in vitro and in vivo experiments were confirmed with a second generation p38α- and p38β-specific ATP-competitive inhibitor, further indicating that the results obtained in these studies are due to specific inhibition of p38 MAPK signaling (data not shown).

DISCUSSION

Our work examined the role of p38 MAPK signaling in cord formation in vitro and neoangiogenesis in vivo, and our data implicate p38 MAPK both in the release of cytokines from tumor and stromal cells and in mediating their downstream effects on endothelial cells. Previous studies using p38 MAPK inhibitors were limited by an inability to distinguish tumor versus stromal cell effects and delineate p38 MAPK isoform-specific effects. Our data support a positive role for p38 MAPK activation in angiogenesis, similar to what was reported by Jackson et al. (33), where small molecule inhibitors of p38 MAPK kinase reduced angiogenesis in an inflammatory angiogenesis model. Others have also reported a role for p38 MAPK in proinflammatory angiogenesis in vivo in a prostate cancer rat model (16), in shear stress-mediated angiogenesis (34), in human coronary artery endothelial cell tube formation (17), and in vitro in basal and tumor conditioned media capillary-like structure formation in a co-culture system (35). Further strengthening a potential role for p38 MAPK in angiogenesis, lack of the p38 MAPK upstream activator, MKK3, causes deficiency in both primary and placental blood vessel development (36).

In this report, VEGF, bFGF, EGF, and IL-6 were shown to activate the p38 MAPK pathway in endothelial cells. LY2228820 dimesylate treatment reduced VEGF-, bFGF-, EGF-, and IL-6-driven cord formation and α-smooth muscle actin expression (pericyte marker) along with the secretion of those soluble factors from ADSCs and a proangiogenic ADSC/ECFC co-culture. Our results are consistent with previous findings that p38 MAPK mediates both signaling downstream of cytokine receptors and cytokine release (33, 34, 37). Thus, p38 MAPK-dependent production of VEGF, bFGF, EGF, and IL-6 may be critical for angiogenesis. VEGF was reported to induce activation of p38 MAPK and HSP27 in endothelial cells through a p38 MAPK-dependent pathway (7, 18, 26). p38 MAPK signaling was also shown to be responsible for activation of HSP27 and mediate smooth muscle cell migration upon PDGF, IL-1β, and TGFβ stimulation (38). In endothelial cells, p38 MAPK was also activated by bFGF, and inhibition of p38 MAPK abrogated bFGF-mediated tube formation of MSS31 stromal cells and endothelial cell migration (39). Therefore, p38 MAPK is activated by many proangiogenic factors and probably mediates endothelial cell function.

p38 MAPK signaling has also been implicated in the regulation of endothelial and mural cell migration (37, 38) and recruitment during angiogenesis (11). Inhibition of p38 MAPK led to reduced HSP27 phosphorylation, actin reorganization, and endothelial cell migration (18), which was dependent on MK2 following VEGF stimulation (26). Furthermore, overexpression of MKK6, an upstream activator of p38 MAPK kinase, enhanced HSP27 phosphorylation and migration in human umbilical vein endothelial cells (40), whereas siRNA-mediated MK2 knockdown in MSS31 spleen endothelial cells inhibited VEGF-induced cell migration (26). It remains to be determined whether p38 MAPK mediates ECFC or ADSC migration in our cord formation system, but we did not observe significant reduction in migration with cytokine-stimulated ECFCs alone or ADSCs alone upon LY2228820 dimesylate treatment.

In addition to reducing individual cytokine-induced cord formation, LY2228820 dimesylate treatment displayed a more pronounced reduction in tumor-conditioned medium-driven and tumor cell-driven cord formation. The fact that p38 MAPK was shown to be activated downstream of many proangiogenic cytokine receptors suggests that LY2228820 dimesylate treatment may have an additive effect when a multitude of cytokines are being released from tumor cells and/or that LY2228820 dimesylate treatment has an effect on tumor cell cytokine secretion and signaling downstream of cytokine receptors in stromal cells. Previous reports also indicate that p38 MAPK can regulate cytokine secretion from tumor cells, including VEGF secretion in malignant gliomas (41) and other tumor cells (19). Pretreatment of tumor cells with LY2228820 dimesylate, which does not affect tumor cell viability, reduced cord formation which suggests that LY2228820 dimesylate treatment alters cytokine secretion from tumor cells. Indeed, LY2228820 dimesylate treatment significantly reduced secretion of VEGF, bFGF, EGF, IL-6, and other proangiogenic (IL-8 and angiogenin) cytokines from a variety of tumor cell lines. Although IL-8 has been shown to play an important role in tumor growth, angiogenesis, and metastasis (42), it was not a potent inducer of cord formation in our in vitro system (data not shown). Pretreatment of tumor cells with LY2228820 dimesylate along with the addition of LY2228820 dimesylate into the cord formation system led to a greater inhibition of cord formation than just pretreatment of tumor cells with LY2228820 dimesylate or the addition of LY2228820 dimesylate into the cord formation assay. This suggests that LY2228820 dimesylate treatment reduced tumor-driven cord formation in part by decreasing cytokine secretion from tumor and/or stromal cells and by affecting the stromal cell response to cytokine stimulation.

In addition to small molecule inhibition of p38 MAPK, protein knockdown of p38α in stromal or tumor cells reduced cord formation. Protein knockdown of p38β only reduced IL-6-driven cord formation in stromal cells, indicating a role for both p38α and p38β in IL-6-mediated cord formation. Targeted inactivation of the mouse p38α gene results in embryonic death due to a placental defect (3–5), and angiogenesis was abnormal in the yolk sac and the embryo itself, resulting in immature networks of vessels (4). In contrast, single knock-out of p38β, p38γ, or p38δ, or both p38γ and δ result in viable, healthy mice (43, 44). In addition, MK2 activity is abolished in p38α knock-out mice (3), whereas p38β knock-out mice exhibit normal MK2 activity (43), which highlights diverse downstream pathway activation among p38 MAPK isoforms. One study using SB203580, a small molecule that completely inhibits p38α and only partially inhibits p38β, suggests that p38α is the principal isoform controlling proliferation and migration of endothelial cells (40). Similarly, our results indicate that downstream p38 MAPK signaling through MK2 and HSP27 is mediated by p38α and not p38β in stromal and tumor cells, which suggests independent, non-redundant functions between the p38 MAPK isoforms.

In our experiments, knockdown of p38α but not p38β reduced secretion of VEGF, bFGF, EGF, and IL-6 production from stromal and tumor cells, implicating p38α as a critical mediator of proangiogenic cytokine secretion. Importantly, the addition of the reduced VEGF, bFGF, EGF (U-87-MG only), and IL-6 in stable U-87-MG or MDA-MB-231 p38α knockdown tumor cells was able to rescue cord formation, indicating that these cytokines appear to be key players involved in p38 MAPK signaling and cord formation on our co-culture system. Additional studies indicate that p38α is the predominant isoform involved in cytokine production in vivo following lipopolysaccharide (LPS) stimulation (43), and gene transfer of p38α and the upstream activator MKK3 significantly increased expression of bFGF and PDGF-A in the normal heart (45), further supporting the concept of independent, non-redundant functions of p38 MAPK isoforms. In addition to p38α, knockdown of the p38 MAPK pathway protein MK2 or HSP27 in stromal cells also reduced cord formation along with VEGF, bFGF, and IL-6 secretion, a result substantiated in MK2 knock-out mice, where reduced angiogenesis in wound healing is in concert with reduced expression of several cytokines (GM-CSF, VEGF, IFNγ, MCP1, TNF, IL-6, and IL-1β) compared with wild-type mice (46). In addition, targeted deletion of MK2 in macrophages led to decreased production of LPS-induced tumor necrosis factor (TNF), IL-6, and other cytokines (47). Taken together, these results reveal a role for the p38 MAPK pathway components MK2 and HSP27 in cytokine production and angiogenesis.

In addition to endothelial cells, pericytes play an important role in endothelial cell function and vascular formation (10, 11). Pericytes have been associated mainly with stabilization of blood vessels but also can sense angiogenic stimuli, guide sprouting tubes, elicit endothelial survival, and exhibit macrophage-like activities (31). Several molecules regulate pericyte contractile tone and function as paracrine signals that reveal an interaction between endothelial cells and pericytes in the regulation of blood flow (31). Tumors usually contain a small number of functional pericytes important for vessel stability, function, and endothelial cell survival (31). Highlighting the importance of pericyte vessel stability function, glioblastomas frequently contain vessels that are not covered by pericytes, and these vessels are more dependent on VEGF as an endothelial cell survival factor (48). Furthermore, blocking PDGF receptor signaling resulted in detachment of pericytes from tumor vessels and restricted tumor growth (49). Inhibition of p38 MAPK signaling with LY2228820 dimesylate treatment or knockdown of p38α significantly reduced α-smooth muscle actin expression from pericyte-like cells in our cord formation assay. This suggests that p38 MAPK may play an important role in paracrine signaling between endothelial cells and pericytes, but it remains unknown whether p38 MAPK signaling is important for pericyte function or recruitment in vivo. Tumor vessels without pericytes appear more vulnerable and may be more responsive to anti-endothelial cell drugs (31). Multiple human renal tumor models treated with a combination of LY2228820 dimesylate and sunitinib show potentiation of sunitinib activity,³ which may be due in part to decreased numbers of pericytes interacting with the vessels, causing the endothelial cells to be more susceptible to the antiangiogenic therapy.

MKKs are crucial enzymes involved in several biological pathways that control cell differentiation, proliferation, and survival (50). In response to extracellular stimuli, MKKs become activated and phosphorylate MAPKs, including extracellular signal-regulated protein kinase (ERK), c-Jun NH₂-terminal kinase (JNK), and p38 MAPK. Others report important roles for the other MAPK signaling pathways (ERK and JNK) in tumor angiogenesis (13, 32). A variety of small molecule inhibitors that target MEK, ERK, and JNK were observed to be antiangiogenic in our in vitro co-culture system (data not shown), further evidence indicating the importance of MAPK signaling in angiogenesis. Our results indicate a positive role for p38 MAPK signaling, in particular p38α MAPK, MK2, and HSP27 in angiogenesis and p38α MAPK in tumor and stromal cell cytokine release.