Cardiomyocyte ATP release through pannexin 1 aids in early fibroblast activation

Abstract

Fibrosis following myocardial infarction is associated with increases in arrhythmias and sudden cardiac death. Initial steps in the development of fibrosis are not clear; however, it is likely that cardiac fibroblasts play an important role. In immune cells, ATP release from pannexin 1 (Panx1) channels acts as a paracrine signal initiating activation of innate immunity. ATP has been shown in noncardiac systems to initiate fibroblast activation. Therefore, we propose that ATP release through Panx1 channels and subsequent fibroblast activation in the heart drives the development of fibrosis in the heart following myocardial infarction. We identified for the first time that Panx1 is localized within sarcolemmal membranes of canine cardiac myocytes where it directly interacts with the postsynaptic density 95/Drosophila disk large/zonula occludens-1-containing scaffolding protein synapse-associated protein 97 via its carboxyl terminal domain (amino acids 300–357). Induced ischemia rapidly increased glycosylation of Panx1, resulting in increased trafficking to the plasma membrane as well as increased interaction with synapse-associated protein 97. Cellular stress enhanced ATP release from myocyte Panx1 channels, which, in turn, causes fibroblast transformation to the activated myofibroblast phenotype via activation of the MAPK and p53 pathways, both of which are involved in the development of cardiac fibrosis. ATP release through Panx1 channels in cardiac myocytes during ischemia may be an early paracrine event leading to profibrotic responses to ischemic cardiac injury.

in cardiac injury, fibroblasts are activated, causing them to transdifferentiate into myofibroblasts, leading to profibrotic responses such as production of extracellular matrix proteins, collagen, and cytokines (8). With time, expansion of the extracellular matrix leads to separation of surviving myocyte bundles, forming a heterogeneous substrate that leads to slowed conduction which promotes life-threatening arrhythmias. In regions distant to the site of injury, fibroblasts activate and migrate to sites of injury and are highly responsive to cytokines and chemokines (64). While activated fibroblasts are known to produce and secrete many of the signals for myofibroblast formation and migration, thus causing a positive feedback loop of activation, little is known about the initial stimulus within the injured region that begins the activation process. It has been assumed that fibroblasts sense the environment and respond accordingly, but how and what they are sensing during very early stages of cardiac injury are unknown. Our studies suggest that ATP released from cardiac myocyte pannexin 1 (Panx1) channels initiates signaling in fibroblasts to begin the fibrotic process.

Panx1 is ubiquitously expressed and is the best understood of the pannexin family (71). Panx1 pannexons (hexameric single membrane channels) exhibit permeability to moderately sized molecules (5, 22, 25, 43) such as second messengers Ca²⁺, inositol 1,4,5-trisphosphate, cAMP, and metabolites (glucose, lactate, glutamate, and ATP) and are glycosylated when expressed at the plasma membrane (12). Panx1 channels have single-channel conductances on the order of 500 pS (4). They are opened by membrane depolarization (4, 13, 26, 74) as well as mechanical stress (4). In immune cells, Panx1 channels are ATP release pores that activate under cellular stress (71). ATP then acts in an autocrine or paracrine manner to activate purinergic receptors on their own plasma membrane and on plasma membranes of neighboring cells (35, 49). The recent discovery of pannexon inhibitor 4-(dipropylsulfamoyl)benzoic acid (probenecid, a drug previously used for gout) allows for more precise functional study of this channel than carbenoxolone and flufenemates that inhibit both pannexin and connexin channels (69).

A widely studied role for Panx1 channels is in immune cells where it participates in the activation of the inflammasome following injury (37, 68). Interestingly, Panx1 transcripts have been found in human heart (6, 10, 60), leading us to hypothesize that this particular pannexin may play a role in inflammatory responses in injured myocardium. Our data now show a novel Panx1 paracrine pathway involved in early stages of cardiac fibrosis that may be amenable to pharmacological intervention to limit subsequent cardiac pathologies such as sudden cardiac death and heart failure.

METHODS

Canine myocardial infarction studies.

All animal studies were done in accordance with Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee approval (129-2008). Myocardial infarction (MI) was induced in six mongrel dogs weighing 25–30 kg as previously described (78). Five normal dogs were used as controls.

HL-1 cells.

HL-1 cells were cultured in 0.2% gelatin-fibronectin-precoated flasks at 37.2°C and 5% CO₂ atmosphere in Claycomb medium (JRH Biosciences, Lenexa, KS), supplemented with 10% fetal bovine serum (JRH Biosciences), 100 U/ml penicillin, 100 mg/ml penicillin-streptomycin (Life Technologies, Carlsbad, CA), and 100 mM norepinephrine (Sigma-Aldrich, St. Louis, MO).

Mouse embryonic fibroblasts.

Mouse embryos taken at day 13 (pc) had brain and red organs (heart and liver) excised and then minced and resuspended in 0.25% trypsin-EDTA. Tissue was allowed to settle in media (high-glucose DMEM, containing l-glutamine, 10% fetal bovine serum, and 100 mg/ml penicillin-streptomycin), and then the supernatant was removed and centrifuged at 3,000 rpm for 15 min. Cells were preplated and then resuspended in media containing 100 μM 2′-3′-O-(4-benzoylbenzoyl)ATP (BzATP; Sigma-Aldrich) for 6 h. Dosage was chosen based on the published literature (56).

Quantitative PCR.

Epicardial tissue was homogenized in 1 ml of TRIzol reagent (Life Technologies) as per the manufacturer's instructions, and a reverse transcription reaction was performed using RT-PCR kit (Applied Biosystems, Foster City, CA). Conditions for reverse transcription were 25°C for 10 min, 37°C for 60 min, and 85°C for 5 min. Quantitative PCR (qPCR) was performed using SYBR green (Applied Biosystems) on 7300 Applied Biosystems qPCR machine (Applied Biosystems). Conditions for qPCR were 50°C for 2 min, 95°C for 10 min, 95°C for 15 s, and 60°C for 1 min for 40 cycles.

Western blot analysis.

Samples were run for Western blot analysis as previously described (40, 52) and blotted with rabbit Panx1 polyclonal 57 antibody or mouse monoclonal Panx1 503 antibody, with rabbit anti-synapse-associated protein 97 (SAP97; Abcam, Cambridge, MA) and with ERK1/2 and phosphorylation of ERK1/2 (Cell Signaling). The Panx1 57 antibody (epitope in the NH₂-terminal region of the cytoplasmic loop) and the 503 antibody (epitope in the NH₂-terminus) were generated against synthetic peptides and validated by the Sosinsky Laboratory (11, 18). Antibodies preincubated with blocking peptide were used as a negative control. All blots were reprobed for GAPDH as a loading control.

N-glycosidase F treatment.

PNGase treatment was performed using N-glycosidase F (PNGase; New England BioLabs) according to the manufacturer's instructions. Normal epicardial tissue and epicardial border zone tissue after 3 h of coronary occlusion (CO) were lysed as for Western blot analysis. For each reaction, 9 μl of lysate (with protein concentration of 40 μg/ul) were used. After 1 μl of 10× glycoprotein denaturing buffer was added, the mixture was boiled at 100°C for 10 min. The reaction was cooled to room temperature. Then, 2 μl of 10× G7 reaction buffer, 2 μl of 10% NP-40, 2 μl of PNGase F, and 2 μl of water were added and the mixture was incubated at 37°C for 1 h. The resulting lysate was mixed with 2× Laemmli sample buffer containing 1% bromophenol blue, 1 mM dithiothreitol, 10% SDS, and 1 M Tris·HCL at pH 6.8 and run on SDS-PAGE gel side by side with samples that underwent the same temperature treatments with PNGase F samples, but contained H₂O rather than PNGase F. Longer incubation times in PNGase F of 3 h and overnight did not affect banding patterns.

Coimmunoprecipitation experiments.

We performed coimmunoprecipitation (39) using rabbit 57 anti-Panx1 antibody or rabbit anti-SAP97 antibody for precipitation of proteins as previously described (40). After SDS-gel electrophoresis and transfer to nitrocellulose membrane, Western blot analysis was performed for Panx1 (using mouse 503 anti-Panx1 antibodies) and SAP97 as described above.

Immunofluorescence imaging.

Rapidly frozen heart samples from the epicardial border zone were sectioned to 15-μm thickness using a Leica 3050S cryostat. Murine embryonic fibroblasts (MEFs) were plated on coverslips following BzATP treatment as described above. Immunostaining was performed as previously described (40).

Yeast two-hybrid assay.

Yeast two-hybrid studies were done as previously described (23).

Dye uptake assays.

Fluorescent dyes were prepared in Dulbecco's PBS (DPBS) consisting of 1× PBS plus 0.1 g/l CaCl₂ anhydrous and 0.1 g/l MgCl₂·6H₂O (pH 7.4) at 5 mg/ml for both Lucifer yellow (457 mol wt, Life Technologies) and Texas red dextran (10,000 mol wt, Life Technologies). Cultures were subjected to 24 h of hypoxia with or without probenecid (Sigma-Aldrich) in 0–1% O₂-5% CO₂-95% N₂ followed by osmotic stimulation with 10× DPBS for 1 min, rinsed, and then incubated on ice for 10 min. Dishes were examined under a Leica DM5000 B epifluorescence microscope for intensity of Lucifer yellow (LY) fluorescence.

ATP assays.

Cells were subject to hypoxia for 24 h (0–1% O₂-5% CO₂-95% N₂) and osmotically stimulated with 0.1× DPBS, containing 1 μM ARL 67156 (Toscis, Bioscience, Ellsville, MO) with or without 1 mM probenecid (Sigma-Aldrich) for 15 min. Intracellular and extracellular ATP concentrations were assessed using ATP Bioluminescence Assay Kit HS II (Roche, Switzerland) according to the manufacturer's instructions.

Transwell cocultures.

HL-1 cells were seeded from homogeneous aliquots onto the lower well surface of the transwell system and allowed to settle for 24 h in normoxic conditions. Dishes were either maintained under normoxic conditions (controls) or transferred to hypoxic (1% O₂) conditions overnight (16 h) with or without either 1 μM probenecid or 25 μM carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone (z-vad-fmk). After 16 h, MEFs were dissociated as described above and plated on the upper well of the transwell surface in the media from the overnight HL-1 incubation. Cells were incubated in the HL-1 media for 4 h in normoxic conditions and then processed for immunostaining as described above.

Gene array experiments.

MEFs (2 controls, 4 ATP treated) were lysed in 1 ml of TRIzol reagent (Invitrogen) as per manufacturer's instructions. Samples were pooled to limit experimental differences. RNA extraction was performed as described above (see qPCR analysis). Samples were processed at Microarray Core of Dana Farber Cancer Institute using GeneChip Mouse Gene 1.0 ST Array. Results were analyzed using dChip software (2). Pathway analysis was performed with Genecodis 2.0 (15, 58).

Statistical analysis.

A Student's t-test analysis was used to compare two experimental groups. For comparison of more than two groups, one-way ANOVA analysis was performed (with Tukey's honestly significant difference correction). The difference was considered significant at P < 0.05.

RESULTS

Panx1 expression increases in early ischemia.

Using semiquantitative and qPCR, we found that Panx1 was expressed at low levels in normal heart but was significantly increased (7-fold) after 3 h of ischemia (Fig. 1, A and B). By Western blot analysis, Madin-Darby canine kidney cells overexpressing Panx1 typically show three Panx1 bands (nonglycosylated GLY0, partially gylcosylated GLY1, and fully gylcosylated GLY2). Heart lysates showed a significant increase in the GLY2 membrane associated form of Panx1 (Fig. 1, C and D). To confirm that the 50-kDa band represented the GLY2 form of Panx1, tissue lysates were treated with PNGase and run on a Western blot. The data showed an increase in mobility of the 50-kDa band in both normal (Fig. 1E, left) and ischemic (Fig. 1E, right) tissue, indicating the presence of glycosylation of Panx1, with an increase in this glycosylation after 3 h CO. The shift in mobility did not completely collapse the GLY2 band, suggesting that there are other potential modifications of the protein. Being as glycosylation of Panx1 is associated with plasma membrane expression (10), the increase in glycosylation of the 50-kDa GLY2 band suggests that after 3 h CO, significant amounts of Panx1 channels are now relocalized to the cardiac sarcolemmal membranes.

Fig. 1.

Identification and characterization of pannexin 1 (Panx1) in canine epicardial border zone myocytes in normal and ischemic heart. A: quantitative PCR showed low levels of Panx1 expression in normal heart that increased significantly after 3 h of coronary occlusion (CO; B). C: Western blot analysis showed both 40 kDa (nonglycosylated GLY0) and 50 kDa (fully gylcosylated GLY2) Panx1 isotypes in overexpressing Madin-Darby canine kidney (MDCK) cells. A partially gylcosylated form, Gly1, is found between GLY0 and GLY2. These bands were absent in the HEK293 cells which do not express Panx1. Mouse brain showed only 2 bands (40 kDa and 43 kDa, GLY0 and GLY1). Following 3 h coronary occlusion, there was a significant increase in the 50-kDa band in heart (D). AU, arbitrary units. To determine whether the 50-kDa band represented the membrane localized fully glycosylated form (GLY2) lysates of normal and ischemic heart were subjected to treatment with N-glycosidase F (PNGase; E). The increase in the 50-kDa band seen in ischemia was lost with PNGase treatment in both normal (left) and 3 h CO (right). This shift with PNGase indicates that the upper band (GLY2) is the glycosylated plasma membrane form of Panx1. *P < 0.05.

Panx1 colocalizes with the scaffolding protein SAP97.

To determine whether increased GLY2 Panx1 caused an increase in membrane localization of the channel, we immunostained 14-μm sections of canine left ventricular epicardium and costained the sections with a pan-cadherin antibody (Fig. 2, A–C, green, or blue, D–F) to mark the intercalated disk. Panx1 was found in low levels in plasma membranes of cardiac myocytes where it did not colocalize with the intercalated disk (Fig. 2A). Following 3 h ischemia, there was an increase in total Panx1 levels at the cell membrane that exhibited a highly organized pattern reminiscent of T-tubule patterns (Fig. 2B). Use of a blocking peptide indicated specificity of our Panx1 antibody (Fig. 2C). The chicken 4515 polyclonal Panx1 antibody raised against an intracellular epitope gave the same staining pattern as our rabbit polyclonal antibody (49). Because of the fact that connexin protein family members are known to interact with postsynaptic density 95/Drosophila disk large/zonula occludens-1 (PDZ) -containing scaffolding proteins such as zonula occludens-1 (ZO-1) and SAP97 and based on the pattern similarity of Panx1 with the cardiac PDZ-containing scaffolding protein SAP97, we examined the interaction between these two proteins. ZO-1, a known connexin binding partner, localizes to the intercalated disk (39) where we did not see Panx1 to be localized. In contrast to ZO-1, SAP97 localized to cardiac sarcolemmal membranes in a pattern very similar to Panx1 (Fig. 2D). This localization did not change after 3 h ischemia (Fig. 2E). A no primary antibody control is shown in Fig. 2F. Costaining of Panx1 with SAP97 after 3 h ischemia (Fig. 2G, Panx1, red; SAP97, green; cadherin, blue) showed almost complete overlap (Fig. 2G, bottom right). These data suggest that Panx1 and SAP97 may form a complex at cardiac sarcolemmal membranes. Quantification of Panx1 expression in cardiac myocyte plasma membranes showed significant increase in membrane localized Panx1 following 3 h of CO (Fig. 2H).

Fig. 2.

Immunolocalization of Panx1 and synapse-associated protein 97 (SAP97) in canine cardiac myocytes. A: low levels of Panx1 (red) were seen in sarcolemmal membranes but not associated with intercalated disks, stained with a pan-cadherin antibody (green). B: following 3 h of coronary occlusion, membrane-localized Panx1 increased. C: blocking peptide to Panx1 shows specificity of the antibody. D: immunolocalization of SAP97 (green) in cardiac myocytes showed sarcolemmal staining that did not colocalize with cadherin (blue). SAP97 did not change after coronary occlusion (E). F: no primary control. G: colocalization of Panx1 and SAP97 in epicardial border zone myocytes after 3 h of coronary occlusion showed a close interaction of these 2 proteins in the sarcolemmal membranes that did not occur at the intercalated disk (cadherin, blue). H: quantification of Panx1 staining in epicardial border zone myocytes showed a significant increase in Panx1 in the sarcolemmal membranes after coronary occlusion.

Panx1 and SAP97 bind directly to each other.

While intriguing, spatial colocalization within a cell at light microscopic resolution is not sufficient proof of a direct binding between proteins. To determine whether these two proteins directly interact, we performed Yeast-2-Hybrid studies using domains of SAP97 as bait and Panx1 carboxyl-terminal domain (CT) as prey. The Panx1 CT was chosen based on data showing that the interaction of many proteins with PDZ-containing scaffolding proteins is via CT interactions (67). We tested four domains of SAP97 (Fig. 3A) and full-length Panx1 CT (Fig. 3B, Full CT), a truncated Panx1 CT (CT-A), and two distal CT domains (CT-B and CT-C) (Fig. 3B). We used pV3 + SV40 as a positive control. Test plates (−2 plate, lacking leucine and tryptophan) showed the efficacy of our system (Fig. 3C). Removal of histidine (−3 plate) was allowed for yeast growth only where the two proteins interacted. We found that Panx1 and SAP97 interact directly, and the interaction domain of SAP97 was the SH3-Hook-guanylate kinase (GUK) domain (amino acid 569–764) rather than a PDZ-domain interaction (Fig. 3D). This is not without precedent as a similarly folded gap junction protein, connexin 32, also binds to SAP97 via this domain (23). Interestingly, we found that all Panx1 CT fragments bound directly to SAP97 suggesting two binding sites (Fig. 3D). To determine which binding site had the higher affinity, we used a higher stringency plate (−Ade, −His) for screening. We found that increasing stringency decreased the binding of the full CT and the distal ends of the CT to SAP97, whereas CT-A (amino acids 300–357) had the highest affinity binding. These data suggest that the primary binding site is within the proximal region, but amino acids 378 to 426 (the distal region of the CT) mask the 300–357 binding site (Fig. 3E).

Fig. 3.

Characterization of the SAP97 and Panx1 COOH-terminus (CT) interaction. A: schematic of the SAP97 domains. B: summary of the Panx1 CT constructs used in the yeast 2-hybrid assay. C: Saccharomyces cerevisiae yeast strain AH109 was cotransformed with GAL4ad fusion constructs GAL4ad-NT [SAP97(1–233)], GAL4ad-PDZ1–2 [SAP97(200–463)], GAL4ad-PDZ3 [SAP97(451–587)], GAL4ad-SH3-Hook-guanylate kinase (GUK) and GAL4ad-SV40 Large T antigen [control (Ctr)], with GALbd fusion constructs GAL4bd-CTFull [Panx1(300–426)], GAL4bd-CTA[Panx1(300–357)], GAL4bd-CTB [Panx1 (379–426)], GAL4bd-CTC [Panx1 (406–426)] and GAL4bd-pVA3 murine p53 (control). Cotransformants were assayed for growth on low-stringency (B, +HIS), medium-stringency (C, −HIS), and high-stringency (D, −HIS/−ADE) media (ad, activation domain; bd, DNA-binding domain). E: higher stringency yeast two hybrid showed that the highest interaction was between SH3-Hook domain of SAP97 and the proximal carboxyl-terminal domain of Panx1. While all Panx1 domains interacted with the SH3-Hook-GUK domain of SAP97 in the medium stringency plates, only the CT-A (proximal region of the Panx1 CT) interacted with SAP97 in the high-stringency plate (F). Coimmunoprecipitation (co-IP) study from normal and 3 h CO canine heart. Panx1 antibody (10) was used for IP and then precipitated proteins were blotted for SAP97. These co-IP experiments revealed a low level of interaction between these 2 proteins in normal heart, but an increase in the interaction after 3 h CO. Mouse brain lysates (no IP) were used as a positive control for SAP97. G: IP with SAP97 antibodies followed by Western blot analysis of the supernatant with the Panx1 503 antibodies showed that the 50-kDa band was no longer present. The absence of this band demonstrates that the GLY2 form of Panx1 interacts with SAP97, whereas the GLY0 cytoplasmic form remained following removal of the SAP97 complex.

We performed coimmunoprecipitation assays of Panx1 and SAP97 to determine complex formation. We used Panx1 antibodies to pull down the complex and blotted for SAP97 (Fig. 3F). Whole lysates of mouse brain were used as a positive control for the presence of SAP97. We found that a low level of interaction between Panx1 and SAP97 was present in normal heart but that this interaction increased following 3 h CO. In contrast, Panx1 did not coimmunoprecipitate with antibodies against another PDZ-containing protein typically found in heart, MINT-1 (44). These data suggest that increased localization of Panx1 at the cardiac sarcolemmal membrane was associated with an increase in formation of the Panx1/SAP97 complex. In order for the complex to be increased at cell membranes, the SAP97 interaction needs to be with the GLY2 form of Panx1. To determine which form interacted, we did an “upside down” immunoprecipitation where we added antibodies directed against SAP97, pulled down the complex, and ran the remaining lysates on a Western blot for the identification of which species of Panx1 was lost. We did this to avoid artifactual identification of binding based on presence of a band that would be expected to run at the identical size to the heavy chain of the pulldown antibody. Examination of lysates remaining after the upside-down immunoprecipitation showed that the Panx1 band remaining was the 40-kDa Gly0 band both in the normal (control) and the 3-h CO samples. Thus these data demonstrate that the interaction of SAP97 is with the GLY2 membrane associated form of Panx1.

Panx1 channels in cardiac myocytes release ATP under stress.

We examined whether cardiac myocyte Panx1 opens in response to stress, as was found in immune cells (16), using a cell culture system of HL-1 cells. HL-1 cells are derived from murine atrial myocytes and retain much of the cardiac myocyte phenotype. Immunostaining of HL-1 cells showed Panx1 primarily intracellularly in control cells but increased at cell membranes in hypoxia. We cultured cells in normoxic or hypoxic conditions (Fig. 4A) and tested their ability to take up the Panx1 permeant dye LY following osmotic stimulation. Normal HL-1 cells take up small amounts of dye, an event that is not blocked by the Panx1 channel blocker probenecid (Fig. 4A, top). While probenecid inhibits the organic anion transporters of the kidney and the human bitter taste receptor (TASR16) (14, 30), these receptors are not found in heart and therefore the effects of probenecid on cardiac myocytes are most likely its closure of Panx1 channels. Here, low level of dye uptake is most likely due to pinocytosis or endocytosis. In contrast, cells subjected to hypoxic stimuli take up significantly higher amounts of LY (Fig. 4A, bottom left) that is blocked by probenecid (Fig. 4A, bottom right). Quantification of dye uptake studies show that fluorescence is increased by almost fourfold in the stressed cells, whereas probenecid completely blocks this increase (Fig. 4B). Thus cellular stress in HL-1 cells causes Panx1 channels to open. Using a luciferase assay to test for ATP released from HL-1 cells in the conditions described above, we found that cellular stress significantly increased ATP levels in the extracellular media that was blocked by probenecid (Fig. 3C). Together, these data suggest that cardiac myocytes undergoing cellular stress open Panx1 channels and release ATP.

Fig. 4.

Cellular stress increases opening of Panx1 channels and results in ATP release from the cells. A: dye uptake assay of HL-1 cells. Control cells do not take up Lucifer yellow to any great extent (A, top, left). Probenecid treatment shows no significant differences in dye uptake compared with the control (A, top, right). In contrast, under conditions of cellular stress, HL-1 cells showed significant increase in Lucifer yellow uptake (A, bottom, left) compared with control cells, which was completely abolished by 1 mM probenecid that selectively blocks Panx1 channel (A, bottom, right). Therefore, this increase in dye uptake is a Panx1 channel-specific event. B: quantification of fluorescence intensity shows significant increase of Lucifer yellow uptake under conditions of cellular stress that is blocked by probenecid application. C: luciferase ATP assay showed that cellular stress causes a significant increase in ATP release from HL-1 cells (relative to control), which is also abolished by addition of 1 mM probenecid, indicating that enhanced Panx1 function results in ATP release to the extracellular media. *P < 0.05.

ATP plays a role in initiation of fibroblast transformation to the myofibroblast phenotype.

Cardiac myocytes are in intimate contact with neighboring fibroblasts (8) that are known to have ATP (purinergic) receptors and are responsive to ATP (48, 76). To determine whether ATP released from cardiac myocytes activates fibroblasts, we treated freshly isolated MEFs with 100 μM of nonhydrolyzable ATP (BzATP). MEFs were used to limit culture-induced transformation to the myofibroblast phenotype. Treated MEFs were plated in BzATP containing media for 6 h. Western blot analysis showed that smooth muscle actin (SMA), a marker of myofibroblasts, was significantly upregulated compared with nontreated control MEFs (Fig. 5A). To determine whether stress fibers of SMA were induced by ATP stimulation, we immunostained MEFs for SMA. We found that control MEFs had low levels of SMA, possibly because of low levels of transformation in culture. In contrast, MEFs treated with ATP had extensive stress fibers of SMA and increased SMA at sites of membrane extensions, reminiscent of focal adhesion sites (Fig. 5B). Analysis of activated cells/total number of cells showed that BzATP stimulation of MEFs more than doubled the number of SMA positive cells compared with controls (63.9 vs. 31.1%).

Fig. 5.

Activation of fibroblasts by ATP. A: murine embryonic fibroblasts (MEFs) were stimulated with nonhydrolyzable 2′–3′-O-(4-benzoylbenzoyl)ATP (BzATP), plated for 6 h, and then lysed for Western blot analysis of smooth muscle actin (SMA) levels. Stimulation with BzATP significantly increases SMA in MEFs, indicating activation of the fibroblasts to the myofibroblast phenotype. *P < 0.05. B: immunolocalization of SMA in control MEFs showed low levels of SMA (B, a) with only a few areas of focal SMA indicating transformation (B, b), an event that occurs at low levels in vitro. In contrast, BzATP stimulated cells have increased levels of SMA and increased numbers of cells, which are SMA positive (B, c). Cells show extensive focal SMA staining indicating a high level of activation (B, d). C: quantification of the percentage of activated cells shows a significant increase in activated fibroblasts in the ATP-treated cells (SMA-to-nuclei ratio).

Panx1 function in HL-1 cells is sufficient to stimulate activation of neighboring MEFs.

To determine whether the channel opening of Panx1 in a cardiac myocyte-like cell was sufficient to initiate fibroblast to myofibroblast transformation, we used a coculture system where HL-1 cells were plated in the lower chamber and MEFs were cultured in the upper chamber. HL-1 cells were stimulated with hypoxia either with or without probenecid to determine whether inhibition of Panx1 channels was capable of inhibiting MEF transformation (Fig. 6). We found that MEFs placed in media from control cells had very few cells that were stimulated to produce SMA (Fig. 6A). In contrast, MEFs incubated in transwells with media from HL-1 cells that were hypoxic, a significant increase in the number of SMA positive cells was seen (Fig. 6B). This increase did not occur if the HL-1 cells were incubated with probenecid during the hypoxic period (Fig. 6C), indicating that blockade of the Panx1 channel inhibited the neighboring MEFs from becoming activated. Additionally, to determine whether the activation of Panx1 was caspase dependent as had previously been shown in immune cells (16), we incubated HL-1 cells with the pan-caspase inhibitor z-vad-fmk during the hypoxic period. Inhibition of caspases in the HL-1 cells decreased the SMA upregulation in the neighboring fibroblasts similar to the effect of blockade of the Panx1 channel (Fig. 6D). These data indicate that HL-1 cells release enough ATP during hypoxia to stimulate neighboring fibroblasts to become activated to the myofibroblast phenotype and that this ATP release is likely arises through open Panx1 channels as the blockade of this channel inhibited transformation. Thus inhibition of caspases also decreased activation of the neighboring fibroblasts suggests that caspase activation is a required step in the activation of ATP release through Panx1 channels.

Fig. 6.

MEFs cocultured in a transwell system with HL-1 cells and stained for SMA. A: MEFs cocultured with control HL-1 cells that had been maintained in normoxic conditions show few cells with SMA staining. B: MEFs cocultured with hypoxic HL-1 cells showed a significant increase in the number of SMA positive cells. C: blockade of the Panx1 channel in the HL-1 cells during the hypoxic treatment led to a significant decrease in the number of activated MEFs in the adjacent transwell inserts. D: additionally, inhibition of caspase activity in HL-1 cells using the pan-caspase inhibitor carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone (z-vad-fmk) led to a significant decrease numbers of activated MEFs in adjacent transwell inserts. (P < 0.05, n = 6). *P < 0.05.

ATP signaling in MEFs leads to activation of profibrotic signaling pathways.

Gene array analysis of BzATP-treated MEFs showed changes in over 600 genes compared with control MEFs. The majority of changes were increases (∼400 genes), whereas the rest were downregulated genes. A list of 20 genes whose expression levels were the most altered is shown in Table 1. When broken down into categories, we found changes in metabolism, cell growth, and signal transduction pathways (Fig. 7A). We examined only upregulated genes (Fig. 7B) for signal transduction pathways and found that the two pathways upregulated were the profibrotic MAPK and p53 pathways (Table 2), suggesting that effects on cell growth and survival are influenced by Panx1 channel regulation. Upregulation of signaling pathways at the level of the gene is not always indicative of activation of these pathways. Levels of protein phosphorylation within the pathway are more reliable indicators of activation. To determine whether activation of the MAPK kinase pathway was induced, we examined the phosphorylation status of one of the primary proteins in the path, the ERK. MEFs were stimulated with ATP and Western blots for phospho-ERK were run. ATP caused a dramatic increase in phosphorylation of ERK1 (Fig. 7C, top band) with no change in total levels of ERK protein.

Table 1.

Genes showing greatest changes in expression following BzATP stimulation

Upregulated Genes	Downregulated Genes
Hspa1a: heat shock protein 1A	Cyp2d10: cytochrome P450, family 2, subfamily d, polypeptide 10
Chac1: ChaC, cation transport regulator-like 1	Olfr1085: olfactory receptor 1085
Trib3: tribbles homolog 3 (Drosophila)	D430041D05Rik: RIKEN cDNA D430041D05 gene
Hspa1b: heat shock protein 1B	Sfrs3: splicing factor, arginine/serine-rich 3 (SRp20)
Il1b: interleukin 1 beta	Irgm1: immunity-related GTPase family M member 1
4930486G11Rik: RIKEN cDNA 4930486G11 gene	Nppb: natriuretic peptide precursor type B
Adm2: adrenomedullin 2	Tslp: thymic stromal lymphopoietin
Vldlr: very low density lipoprotein receptor	Usp18: ubiquitin specific peptidase 18
Sat2: spermidine/spermine N1-acetyl transferase 2	LOC664934: similar to spermiogenesis specific transcript on the Y 1
Stc2: stanniocalcin 2	Nes: nestin
Gpt2: glutamic pyruvate transaminase (alanine aminotransferase) 2	Pi15: peptidase inhibitor 15
Zfp455: zinc finger protein 455	Bhlhb8: basic helix-loop-helix domain containing, class B, 8
Sesn2: sestrin 2	Cd209c: CD209c antigen
Cth: cystathionase (cystathionine gamma-lyase)	C130073F10Rik: RIKEN cDNA C130073F10 gene
Ddit3: DNA-damage inducible transcript 3	Cd180: CD180 antigen
Cd14: CD14 antigen	Ly6d: lymphocyte antigen 6 complex, locus D
Pck2: phosphoenolpyruvate carboxykinase 2 (mitochondrial)	Txnip: thioredoxin interacting protein
EG665858: predicted gene, EG665858	Ms4a6c: membrane-spanning 4-domains, subfamily A, member 6C
Aldh1l2: aldehyde dehydrogenase 1 family, member L2	Rassf10: Ras association (RalGDS/AF-6) domain family (N-terminal) member 10
Sprr2 g: small proline-rich protein 2G	Klhdc8a: kelch domain containing 8A

BzATP, 2′-3′-O-(4-benzoylbenzoyl)ATP.

Fig. 7.

Gene array analysis of BzATP-stimulated MEFs. A: percentage of cell division regulators altered by BzATP stimulation. B: analysis of signal transduction pathways indicate that ATP stimulation of fibroblasts leads to upregulation of the MAPK, p53, and tRNA biosynthesis pathways. Individual upregulated genes are described in Table 2. C: Western blot analysis confirmation of the activation of the MAPK pathway. Examination of the phosphorylation of ERK1/2 (pERK1/2) in ATP-treated MEFs (treated) shows increased pERK, indicating activation over nontreated cells (control). Total ERK is shown as a loading control.

Table 2.

Signaling pathway genes upregulated in MEFs by ATP stimulation

Component	Pathway	Fold Change
MAPKKK2 (MEK)	MAPK	1.65
Dusp10	MAPK	1.63
Gadd45a	MAPK, p53	2.21
Trib3	MAPK, p53	6.62
Gadd45b	p53	1.58
Sesn2	p53	2.94
Sesn3	p53	1.62

MEFs, murine embryonic fibroblasts.

DISCUSSION

Cardiac ischemia causes cells in the infarcted region to die, whereas cells in surrounding regions survive with altered electrical and biochemical properties (39, 62). In these areas, activation of surviving fibroblasts leads to increases in cytokines, chemokines, extracellular matrix deposition, and fibrosis (8), which separates myocyte bundles, isolating surviving cells from the remaining myocyte syncytium. This heterogeneity of substrate leads to slowed cardiac conduction and formation of arrhythmias (20).

Our studies indicate a novel mechanism by which fibrosis is initiated early in ischemic hearts. We showed that Panx1 channels in intact heart directly interact with the scaffolding protein SAP97 through the proximal end of the Panx1 CT in the region spanned by amino acids 300–357 and the SH3-Hook domain and the proximal end of the inactive GUK domain of SAP97 (amino acid 569–764). Our data indicate that ischemia increases this interaction at myocyte cellular membranes very early following cellular injury. Increased membrane channel density following cellular stress and increased Panx1 function via caspase activation led to ATP release into the extracellular space at levels high enough to impact cells over a distance. Additionally, we showed that ATP stimulation of fibroblasts caused rapid transformation to the myofibroblast phenotype and activates many profibrotic signaling pathways. Together, these data suggest that early in ischemia Panx1 channels may provide a paracrine signal which initiates fibroblast activation, leading to fibrosis and formation of a heterogeneous substrate in the post-MI heart. These data may be supported by findings of Panx1 expression increases in a heart failure model (57).

Purinergic (ATP) receptors are found in abundance in the heart (80), including ionotropic P2X receptors that have been linked to signal transduction in cellular injury (41, 70, 80). ATP activation of P2X receptors has been linked with inflammation, IL-1β production, and MAPK pathway activation in other systems (21, 28, 36, 65, 66). Our gene array analysis of ATP-activated fibroblasts suggests that these cellular events are occurring here as well. It is interesting to note that one of the highest fold change in gene expression in ATP-activated fibroblasts is in IL-1β (Table 2), suggesting that ATP signaling is involved in fibroblasts involvement in the immune response. Additionally, ATP can have an effect on cardiomyocytes through autocrine activation of myocyte purinergic receptors that leads to calcium entry into cells. This has an overall positive ionotropic effect on the myocytes, increasing both the force and rate of contraction of the intact heart (80). Breakdown of ATP yields products that are also bioactive and have similar downstream effects as ATP (42). Thus, while our data show that ATP that is nonhydrolyzable can cause activation of fibroblasts, there may be a more robust signal in intact heart due to activation of signaling by breakdown products of ATP. Changes in Ca²⁺ homeostasis may also be one reason that ATP has been reported to have a proarrhythmic action (33). In ischemic heart, it is likely that both of these events occur concurrently in fibroblasts, eventually forming fibrosis and scar tissue, whereas the remaining surviving myocytes contract more vigorously to maintain heart function in the face of injury.

Scaffolding proteins containing PDZ domains are important for receptor and ion channel clustering (29). Originally described in the brain at neuronal synapses (75), there is increasing evidence that these scaffolds play a similar role in the heart (1, 7, 27, 39, 45, 46, 53, 61, 63, 81). It has been shown that interaction of cardiac ion channels with SAP97 occurs with a number of potassium channels including Kv1.5, Kv4, Kir2.2 and 2.3, and the hyperpolarization-activated, cyclic nucleotide-gated channels (1, 7, 27, 45, 46, 61, 81). These ion channels interact with SAP97 through standard PDZ-domain binding to the terminal most amino acids of the ion channel (1, 27, 45, 46, 53, 61, 81) and appear to play an important role in regulation of channel density at the cardiac myocyte membranes (1, 79). We show here direct interaction between Panx1 and SAP97, which is important for Panx1 channel density in the membrane. In contrast to the binding of ion channels that mostly bind via PDZ domains, SAP97 binds to Panx1 via its SH3-Hook-GUK domain. This is not the first report of a GUK domain interaction of SAP97. Interestingly, one of the connexin proteins, connexin 32, interacts with SAP97 via a non-PDZ domain, also at the SH3-HOOK domain (23). As with Panx1, the domain on connexin 32 that interacts with SAP97 is also a proximal CT region and a secondary site (23), suggesting that this is a common structural association in this region of SAP97 unique to the connexin-like protein families. This arrangement on SAP97 allows for multiple channels to associate at the same time, forming a multiunit macromolecular complex of potassium, sodium, and ATP channels at the cardiac myocyte sarcolemmal membranes. This highly dynamic structure is likely to play a major role in ion homeostasis in cardiac myocytes as well as in paracrine signaling in diseased heart.

Fibroblasts in normal heart turnover only rarely, but following cardiac injury new fibroblasts, which arise from fibrocytes, play a role in wound healing and the immune response. Fibroblasts from outside of the heart can migrate in and play a role in the immune response as well (8). Additionally, cells in the injured heart undergo endothelial mesenchymal transition (83). These cells may also be activated by ATP signaling. While cells of fibrocyte origin may be activated by ATP signaling, activation at the early time post-MI is likely to be in the resident fibroblasts because very few other fibroblast types are unlikely to be present. Fibroblasts, through receptor activation, sense the environment and respond by increasing production of extracellular matrix proteins. Following injury, multiple cytokines and chemokines such as TGF-β, TNF-α, IL-1β, IL-6, and others (82) begin to increase within several hours. TGF-β is a pleiotropic cytokine that promotes myofibroblast differentiation and is a major signaling agent in cardiac fibrosis. Active TGF-β triggers several signaling pathways (38, 55), including the well-characterized Smad pathway and the phosphotidylinositol 3-kinase pathway (3, 19, 32), which causes activation of focal adhesion kinase and p38 MAPK (34, 47, 59). IL-1α and -β belong to the group of initial stimuli that drive acute inflammatory responses after MI (31). Both IL-1α and -β levels are elevated in infarcted heart (73) and have been shown to be potent activators of the p38 MAPK pathway in fibroblasts (54, 73, 77). Additionally, IL-1β contributes to formation of cardiac fibroblasts migratory phenotype through coordinate regulation of the MAPK signaling cascades (54). While each of these activates fibroblasts via unique signaling pathways, a common downstream effector pathway is the p38 MAPK pathway (34, 47, 59), which, in turn, activates Akt, a serine/threonine kinase, involved in multiple cellular processes including cell proliferation, inflammation, survival, and glucose metabolism. Activation of these pathways has been shown to lead to increases in SMA, collagen 3, and subsequent fibrosis (50, 51). Our finding that ATP activates these pathways in fibroblasts suggests that early activation of fibroblasts may be occurring because of ATP released from cardiac myocyte Panx1 channels, although there are likely to be other early events that also play a role in this activation.

In conclusion, we find that the ATP release channel Panx1 is found in cardiac myocytes and sarcolemma expression increased in early ischemia. The ability of cardiac myocytes to release ATP may be one of the molecular mechanisms that allow for the early activation of fibroblasts and lead to fibrosis. Thus blockade of this channel in the heart after ischemia may lead to treatments that decrease fibrosis and limit formation of the heterogeneous substrate. This in turn would decrease the heterogeneous substrate found in the peri-infarct region of the post-MI heart and provides a potential strategy for designing therapeutics.

GRANTS

Financial Support was contributed by National Institutes of Health Grants HL-083205 (to H. S. Duffy), GM-072631 (to P. L. Sorgen), and GM-065937 and GM072881 (to G. E. Sosinsky), as well as National Science Foundation Grant MCB0543934 (to G. E. Sosinsky), American Heart Association Grant 10SDG2610281 (to D. Boassa) and P41 NIH grant for the National Center for Microscopy and Imaging Research (to G. E. Sosinsky) RR-004050 (to M. Ellisman).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

E.D., M.I.K., P.L.S., G.E.S., and H.S.D. conception and design of research; E.D., G.S., D.B., J.R.B., K.K., C.A., and H.S.D. performed experiments; E.D., G.S., D.B., J.R.B., K.K., C.A., G.E.S., and H.S.D. analyzed data; E.D., G.S., D.B., J.R.B., K.K., C.A., M.I.K., P.L.S., G.E.S., and H.S.D. interpreted results of experiments; E.D., P.L.S., G.E.S., and H.S.D. edited and revised manuscript; E.D., G.S., D.B., J.R.B., K.K., M.I.K., P.L.S., G.E.S., and H.S.D. approved final version of manuscript; J.R.B., P.L.S., and H.S.D. prepared figures; G.E.S. and H.S.D. drafted manuscript.