Subunit 3 of the COP9 Signalosome Is Poised to Facilitate Communication between the Extracellular Matrix and the Nucleus through the Muscle-Specific β1D Integrin

Abstract

Yeast two-hybrid analysis (Fields and Song, , Nature, 340:245–246) was used to screen a human heart library to isolate proteins interacting with the adult muscle-specific β1D integrin but not with β1A integrin. In addition to previously identified interactions (RACK 1(Liliental and Chang, , Journal of Biological Chemistry, 273:2379–2383) and α-actinin (Otey et al., , Journal of Cell Biology, 111:721–729), the authors isolated several novel candidates. These include subunit 3 (CSN3/Sgn3) of the COP9 signalosome complex, cyclins D1, D2, and D3, RanBPM, and a recently identified protein COG8/DOR1. These protein interactions were specific for β1D integrin, as no binding to β1A integrin cytoplasmic domain was measurable by two-hybrid analysis. This paper presents the initial characterization of the interaction of CSN3 with β1D integrin, the localization of CSN3 and the other COP9 signalosome subunits in embryonic and adult cardiac myocytes and their response to muscle cell differentiation.

• β1D integrin
• COP9 signalosome
• CSN3
• heart

INTRODUCTION

The extracellular matrix (ECM) is an important determinant of muscle mechanics. The ECM maintains the alignment of myocytes and vessels and prevents myocyte slippage during muscle contraction. Effectively, the ECM functions as a force coupler between adjacent myocytes. Integrin receptors are interposed between the ECM connective tissue skeleton outside cells and the force-generating actin-myosin system within the cell. Integrin-mediated adhesion profoundly influences diverse cellular processes, including cell shape, migration, gene expression, and differentiation (Hynes 1992). In cardiac and skeletal muscle, integrins are localized to the Z-discs and myotendenous junctions or intercalated discs where the integrin-mediated ECM-cytoskeletal linkage contributes to the structural integrity of the muscle (Bao et al. 1993; Gullberg et al. 1998). The ECM-cytoskeletal linkage allows molecular and mechanical signals to be transduced through the ECM across the cell membrane to the cytoplasm, impacting upon the organization of the cytoskeleton and activating signaling pathways that alter muscle gene expression. In fact, integrins function as mechanotransducers in noncardiac cells, activating intracellular signaling in response to mechanical stress (Ingber 1991; MacKenna et al. 1998; Wang et al. 1993). Mechanical stress is a critical component of pressure or volume overload in the heart, leading to altered phenotypes of cardiac myocytes (Sadoshima and Izumo 1997). Therefore, integrins are ideal candidates for the role of mechanotransducers in cardiac myocytes. Significant evidence supports an important role for integrin-ECM interactions in sensing mechanical strain in the heart, leading to alterations in extracellur signal-regulated kinase (ERK), p38 mitogen-activated protein kinase (MAPK), and c-Jun N-terminal kinase (JNK) signaling pathways (Liang et al. 2000). However, the specific pathways activated by integrins to regulate cardiac myocyte structure and functions remain largely undefined.

Integrins are heterodimeric α/β glycoprotein receptors. The ligand specificity of integrins is dictated by their α/β ?pairing (Sonnenberg 1993). When cells bind to ECM (e.g., fibronectin, laminin, collagen, etc.) through integrins, they cluster into cell surface complexes essential for the activation of intracellular responses. The predominant integrins expressed on cardiac myocytes belong to the β1 family. Four splice variants of the cytoplasmic domain of β1 integrin (β1A to β1D) have been described in humans (Baudoin et al. 1996). β1A and β1D are identical with the exception of 13 amino acids in their cytoplasmic domains. β1A, the most prevalent isoform, is ubiquitously expressed except in mature striated muscle. In contrast, the β1D splice variant is uniquely expressed in adult cardiac and skeletal muscle, where it may strengthen cell-ECM adhesion (Belkin et al. 1997). β1B and β1C are rare human isoforms that are absent in the mouse.

The expression of β1A and β1D isoforms is developmentally regulated in cardiac cells. β1A is expressed during embryogenesis whereas β1D expression begins around birth and becomes the only isoform expressed on adult cardiomyocytes. In cardiac muscle, β1D is predominantly paired with α7B to form a laminin receptor (van der Flier et al. 1995). The unique carboxyl-termini of these splice variants suggest distinct functions and unique molecular interactions reside within this region. To probe the function of these splice variants in vivo, Sonnenberg and colleagues created mice expressing exclusively β1A or β1D (Cachaco et al. 2003; Gimond et al. 2000). β1A-expressing mice were viable and normal, whereas mice expressing only β1D died during development (Gimond et al. 2000). Thus, β1A is functionally distinct from β1D. Furthermore, the restricted temporal and spatial expression of β1D in cardiomyocytes suggests that it may have a unique function in the heart. In fact, β1D-mediated signaling, but not β1A signaling, inhibited cell cycle progression in normal myoblasts and fibroblasts (Belkin and Retta 1998).

One in three Americans will be affected by heart disease (HD) during their lifetime. At the cellular level, the manifestations of HD include uncontrolled cell growth (hypertrophy) and/or fibrosis. In order to devise more effective treatment strategies for HD, it is necessary to understand (at the molecular level) how cardiac myocytes hypertrophy and how the ECM remodels in response to high blood pressure or myocardial infarction. Induction of the hypertrophic response pathway in the cardiac myocyte is complex. Each of the three MAPK pathways contributes to the development of hypertrophy. Increased ECM deposition and altered integrin expression characterize in vivo models of hypertrophy (Kuppuswamy et al. 1997; Sun et al. 2003). In addition, integrin signaling is important in regulating the cytoskeletal reorganization. Moreover, in vitro, engagement of β1 integrins with ECM ligand is required for the hypertrophic response of cardiac myocytes to α-adrenergic stimuli (Ross et al. 1998). Therefore, integrin activation/ligation and the resultant signaling are essential components of cardiac hypertrophy and fibrosis. Despite these provocative observations, surprisingly little is known about the molecular linkages coupling integrins with gene expression and cell growth in cardiac myocytes. This study identified a specific molecular interaction between the muscle-specific β1D integrin and the CSN3 subunit of the COP9 signalosome, identifying a new potential signaling pathway in cardiac and skeletal myocytes.

MATERIALS AND METHODS

Two-Hybrid Constructs

To create binding domain (BD) constructs, cDNAs corresponding to amino acid residues 752 to 801 of integrin β1D and 752 to 798 of integrin β1A cytoplasmic domain (β-tails) were cloned into pAS2-1. cDNAs were a generous gift of Dr. Joseph Loftus.

Library Screening

The β-tail–BD constructs, along with a human heart cDNA activation domain library (pACT2-human heart cDNA; Clontech, catalog number HL4042AH) were cotransformed into the reporter yeast strain Y190 using standard LiAc transformation methods (Clontech, Yeast Protocol Handbook). Y190 has both a HIS3 and lacZ gene under control of a GAL4 responsive element. Transformants were plated onto SD medium lacking histidine. His+ colonies were screened by the colony-lift filter assay for β-galactosidase activity (Clontech, Yeast Protocol Handbook).

Isolation of Plasmid DNA from Yeast

Pellets from a 3.0-ml stationary-phase culture were resuspended in a 0.2-ml solution of 2% Triton X-100, 1% sodium dodecyl sulfate [SDS], 100 mM NaCl, 10 mM Tris-HCl, pH 8.0, and 1 mM EDTA. Then 0.2 ml of phenol/chloroform (1:1) and 0.3 g of glass beads were added. The samples were vortexed for 2 min, centrifuged at 13000 rpm, for 5 min, and the aqueous supernatant removed. This was followed by ethanol precipitation and washing with 70% ethanol. The DNA was resuspended in 30 µl sterile water and 4 µl were electroporated into Esherichia coli.

Sequencing

All sequencing was performed by the Mayo Clinic Sequencing Core.

Pull-Down Assays

Cobalt or glutathione-agarose beads were incubated with combinations of affinity-purified glutathione S-transferase (GST) fusion protein containing the 1D cytoplasmic domain (GST-1D), affinity-purified His-tagged CSN3 fusion protein, and/or GST protein for 1 h at 25°C or overnight at 4°C. Beads were washed with profound lysis buffer 5 times (Pierce). Bound proteins were eluted by boiling 5 min in 2× SDS sample buffer. Immunoblotting with anti-β1D antibody (2B1; generous gift of A. Sonnenberg) detected GST-1D that coprecipitated with the His-CSN3 protein.

Immunoprecipitation Assay

Adult mouse hearts were isolated, minced on ice, and homogenized in IP Buffer (50 mM Tris-HCl, pH 8.0, 120 mM NaCl, 1.0mM EDTA, pH 8.0, 0.1% Tween 20, 1.0 mM dithiothreitol [DTT], 1× protease inhibitor cocktail, and 1× phosphatase inhibitor cocktail). Lysates were centrifuged for 10 min at 13000 rpm and the pellets discarded. A 500-µg sample of this protein lysate was incubated with anti-CSN3 antibody (1:200) at 4°C overnight. After 1 h incubation with protein A/G–linked sepharose beads, the sample was centrifuged for 1 min at 13000 rpm. The beads were than washed 5× in IP Buffer. The resultant precipitate was eluted with 2× SDS sample buffer and immunoblot analysis with antibodies 2B1 (β1D integrin–specific antibody) and anti-Jab1 (CSN5) was performed.

Culturing and Fractionation of C2C12 Cells

Mouse C2C12 myoblasts were cultured at 37°C/5% CO₂ in Dulbecco's modified Eagle's medium (DMEM), supplemented with 4.5 g/L glucose, 4 mM l-glutamine, and 10% fetal bovine serum (FBS). To differentiate cells, 80% confluent C2C12 cells were fed with DMEM + 10% horse serum for 11 to 14 days. Ten plates of C2C12 cells were rinsed with phosphate-buffered saline (PBS), scraped into 50-ml conical tube, and centrifuged for 5 min at 1000 rpm. The pellet was weighed and 9 volumes of ice-cold homogenization buffer (HB; 0.25 M sucrose, 50 mM HEPES, pH 7.4, 1× protease inhibitor cocktail, and 1× phosphatase inhibitor cocktail) added. Cells were dounce homogenized (40 passes) and lysis was verified by light microscopy. Lysate was centrifuged at 600×g for 10 min at 4°C. The supernatant was transferred into a new tube (crude cytosol). The pellet (crude nuclei) was washed three times with ice-cold HB, centrifuged (2500 rpm, 10 min, 4°C), then 250 µl of HB was added. To pellet the mitochondrial fraction, the crude cytosol was centrifuged (5000×g, 15 min., 4°C). The supernatant (cytosol + microsomal) was transferred into a new tube and the pellet (mitochondrial fraction) was washed three times with HB and resuspended in 200 µl. The microsomal fraction was pelleted (100,000×g, 1 h, 4°C). The final supernatant (cytosolic fraction) was transferred into a new tube and the pellet (microsomal fraction) was resuspended in 100 µl HB.

Immunoblot Analysis

The BCA Protein Assay (Pierce, Rockford, IL) was used to quantitate the cytosolic fraction. The loading of each fraction was proportional to the amount of starting material. Each cytosolic fraction contained 40 µg protein. Fractions were separated upon 4% to 12% Bis-Tris SDS–polyscrylamide gel electrophoresis (PAGE) (Invitrogen, Carlsbad, CA). The proteins were transferred onto Hybond-C membrane (GE Healthcare, UK), blocked in 5% milk/PBS, and incubated with indicated antibodies. Horseradish peroxidase (HRP)-conjugated secondary antibodies were detected using the SuperSignal West Pico or Femto Maximum Sensitivity Substrate (Pierce).

Antibodies

Antibodies to Signalosome subunits CSN1, CSN4, CSN7, and CSN8 were purchased from Biomol International (UK). Antibody against CSN3 subunit was obtained from Bethyl Labs (Montgomery, TX) and antibody against CSN5 subunit was from GeneTex (San Antonio, TX). Antibody against β1A integrin (MC229) was generated in house. Antibody against β1D integrin (2B1) was generously provided by Dr. A. Sonnenberg. Antibody against β1 integrin (9EG7) was generously provided by Dr. D.V. Weber. Antibody against α-actinin (EA-53) was purchased from Sigma-Aldrich (St. Louis, MO).

RESULTS

Two-Hybrid Screen Identifies Molecules that Interact with the Unique β1D Integrin Cytoplasmic Domain

The Matchmaker System (Clontech) was used to identify novel protein/protein interactions. Specifically, the yeast transcriptional activator Gal4p contains two domains, the activation domain (AD) and the DNA binding domain (BD). The two domains, when physically juxtaposed in vivo, activate transcription of GAL4-dependent reporter genes, GAL1-lacZ and GAL1-HIS3. In this system, the AD and BD are made as individual proteins that have been fused in frame to the proteins of choice. Interaction between the fusion portions of the two proteins allows DNA binding and transcriptional activation. Specifically, plasmid pAS 2.1 was used to fuse the AD to the β1D integrin cytoplasmic domain (residues 752 to 801) to create the “bait.” The BD library screened was a human heart Matchmaker cDNA library (Clontech, catalog number HL4042AH).

The yeast reporter strain Y190 was cotransformed with the bait and library plasmids. A total of 1×10¹¹ independent TRP + /LEU+ yeast colonies were obtained. To screen for colonies that could activate the GAL1-HIS3 reporter, the colonies were replica plated to SD-Trp-Leu-His + 3AT. To add specificity, the resultant HIS+ isolates were screened for the ability to activate GAL1-lacZ, with the X-gal filter assay (Clontech, User Manual PT3024-1). Approximately 100 independent colonies were identified that grew in the absence of histidine and contained β-galactosidase activity.

To identify the unknown proteins responsible for the interaction, plasmids from colonies were isolated and individually retransformed into yeast Y190 with the β1D-bait vector. To determine if the protein/protein interaction was dependent upon unique β1D sequence, the positive plasmids were also cotransformed with an empty bait plasmid and a bait construct expressing the β1A integrin cytoplasmic domain (residues 752 to 798). The fresh colonies were reexamined for activation of both GAL reporter genes. Thirty-five of the isolated plasmids transmitted the activation of GAL reporter genes in the presence of β1D-bait and not with the empty bait vector. All 35 positive plasmids were sequenced and compared to sequences deposited in GenBank. Table 1 shows the proteins encoded by these plasmids (when identifiable). At least three of the proteins (encoded by the plasmids) were not specific to β1D, as shown by cross-reaction with the β1A-bait.

Table 1.  Two-hybrid screen identifies human proteins that interact with the unique β1D integrin cytoplasmic domain

		Bait in β-gal assay
Candidate protein/cDNA	No. of isolates (independent)	β1D	β1A
Cyclin D1/BCL1/PRAD1	10	+ + +	−
Cyclin D2	4	+ +	−
Cyclin D3	3	+ + +	−
Mitofilin	2	+ + +	−
α-Actinin/ACTN2	2	+ + +	+ + +
RanBPM	2	+ + +	+
MAP1A	2	+ +	−
Inhibitor of CDK	2	+	−
Rack1	1	+ + +	+ + +
Subunit 3 of COP9/CSN3	1	+ + +	−
COG8	1	+ + +	−
DEAF1	1	+	−
Macropain (α-subunit of 20S proteasome)	1	+	−
Meis3	1	+	−
Unidentified protein (Chromosome 20)	1	+	−

+ + +, strong blue; ++, blue; +, faint blue; −, white.

Confirming the specificity of the screen, we isolated two proteins known to interact with β1 integrins, RACK 1 (Liliental and Chang 1998) and α-actinin (Otey et al. 1990) (Table 1). In addition to these previously identified interactions, we isolated a number of novel candidates. The strongest specific interactions were with subunit 3 (CSN3/Sgn3) of the COP9 signalosome complex, cyclins D1and D3, mitofilin, and a recently identified protein COG8/DOR1. The remaining candidates failed to meet the β1D integrin specificity requirements or showed only weak interaction.

One protein identified, subunit 3 of the COP9 signalosome, caught our attention. Interestingly, subunit 5 (JAB1/CSN5) of the same complex was previously shown to interact with LFA1 (a β2 integrin expressed in lymphocytes) (Bianchi et al. 2000). Lending further support, we isolated RanBPM. RanBPM also associates with LFA1 and is postulated to be a molecular scaffold that couples integrins with intracellular signaling pathways (Denti et al. 2004). Therefore, our initial focus was to pursue the putative interaction between CSN3 and β1D integrin.

Pull-Down Assays Confirm the Interaction of CSN3 and β1D Integrin

Traditional pull-down assays (coprecipitation) were used to see if purified glutathione S-transferase–β1D integrin fusion protein could bind to His/V5-tagged CSN3 protein. More specifically, the β1D cytoplasmic domain was cloned into a glutathione S-transferase (GST) fusion protein vector (pGEX). The GST-β1D fusion protein was expressed in bacteria and purified over glutathione-linked agarose beads. Similarly, CSN3 was cloned into a His/V5-tagged expression vector (pBADThio) and His-CSN3 fusion protein was purified on nickel beads. Purified proteins were incubated with cobalt- or glutathione-linked beads in the presence and absence of the interacting protein. After incubation, the beads were extensively washed and the bound proteins eluted with SDS sample buffer. After separation on a 10% polyacrylamide gel, the proteins were transferred to Hybond-C membrane and probed with antibody to CSN3 or β1D integrin. The resultant immunoblot after precipitation with cobalt beads is shown in Figure 1A. GST-β1D fusion protein was coprecipitated only in the presence of the His-CSN3 protein. Conversely, immunoblot detection of His-CSN3 after precipitation with glutathione-linked agarose demonstrated the coprecipitation of His-CSN3 only in the presence of GST-β1D (Figure 1B, lane 3). Although compelling, these results did not eliminate the possibility that the interaction of the two fusion proteins occurred on the protein tags.

Figure 1.  (A) Coprecipitation of purified His-CSN3 and GST-β1D on cobalt beads. Lane 1, the eluate from the incubation containing purified GST-β1D fusion protein. Lane 2, the eluate from His-CSN3 fusion protein. Lane 3, the eluate when the incubation contained both His-CSN3 and GST-β1D. The blot was probed with β1D-specific antibody 2B1 (1/200). GST-β1D was only detected (lane 3) when His-CSN3 was present in the incubation. (B) Coprecipitation of purified His-CSN3 and GST-β1D on glutathione-S-transferase–linked beads. Lane 1, the eluate from the incubation containing purified GST-β1D fusion protein. Lane 2, the eluate from His-CSN3 fusion protein. Lane 3, the eluate when the incubation contained both His-CSN3 and GST-β1D. The blot was probed with V5-specific antibody. His/V5-CSN3 was only detected (lane 3) when GST-β1D was present in the incubation. (C) Immunoprecipitation of both β1D integrin and CSN5- with CSN3-specific antibody (1/2000; Bethyl Labs, BL564). Lane 1, total cell lysate. Lane 2, the IP eluate. The blots were probed with β1D (2B1) and CSN5 (1/1000; AB495, ABcam) antibodies. (D) The immunobolt shown in C probed with an antibody specific to β1A integrin (MC229).

CSN3 Precipitates Both β1D Integrin and CSN5/JAB1 from Adult Heart Lysates But Not β1A Integrin

To demonstrate that the interaction of CSN3 and β1 integrin was independent of the protein tags, specific to β1D integrin, and occurred with the native proteins, a traditional immunoprecipitation (IP) was performed. Adult mouse heart lysates were prepared and incubated overnight with CSN3-specific antibodies bound to protein A sepharose beads. After five successive washes, the beads were eluted. The eluate was analyzed by immunoblot analysis with antibodies specific for β1D integrin, β1A integrin, and CSN5 (a signalosome subunit previously shown to interact with CSN3). In an adult heart extract, only β1D integrin is expressed by the myocytes and β1A integrin is expressed by the fibroblasts. As shown in Figure 1C, IP with anti-CSN3 antibodies coprecipitated both β1D integrin and CSN5. However, as shown in Figure 1D, β1A integrin is not precipitated. This result confirms specificity of the interaction of β1D integrin with the CSN3 subunit of the signalosome complex.

CSN3 Can Localize to Z-bands in Adult Cardiac Myocytes But Not Focal Adhesions in Proliferating Myoblasts

To determine the subcellular localization of CSN3, and the signalosome complex subunits in adult cardiac myocytes, isolated myocytes were incubated with antibodies against various COP9 subunits and visualized by confocal immunofluorescence microscopy. Given the high degree of similarity between the mouse, human, and plant subunits, we postulated that antibodies to one species may cross-react with another species. (In fact, antibodies to subunits 1 and 8 were known to cross-react with mouse.) Antibodies generated against the human subunits 1 2, 3, 6, 7, and 8 along with the Arabidopsis subunit 4 and the mouse subunit 5 (Jab1) were used in subsequent experiments. Cardiac myocytes were isolated from a 12-week-old FVB/N mouse heart and double-stained with antibodies to each signalosome subunit and α-actinin (a marker for the Z-band). As shown, CSN3 (Figure 2A) stained in a punctate pattern. A fraction of CSN3 staining colocalized with α-actinin at the Z-band. The Z-band localization was consistent with the well-characterized β1D integrin localization in adult mouse cardiac myocytes (Figure 2A, inset). A similar punctate distribution was seen with CSN1, CSN2, CSN4, CSN5, CSN6, and CSN8 (data not shown). These results demonstrate that CSN3 and β1D integrin can both localize at Z-bands in adult cardiac myocytes (β1A integrin is not expressed in these cells). Staining of a cell line that specifically expressed β1A integrin (proliferating C2C12 myoblasts) demonstrated that CSN3 localized to both the cytoplasm and the nucleus in a punctate pattern (Figure 2B). However, CSN3 was not found to localize with the β1A integrin at focal adhesions. This result reinforces the specificity of the β1D integrin–CSN3 interaction.

Figure 2.  Immunofluorescence microscopy of adult cardiac myocytes and proliferating myoblasts. (A) CSN3 (green) was detected with PW 8235 (Biomol), α-actinin (red) with EA-53 (Sigma), and the nucleus with bis-benzamide (blue). Orange/yellow shows colocalization of some CSN3 with α-actinin at the Z-bands. The inset shows a similar pattern of localization for β1D integrin (9EG7). (B) CSN3 (green) localized in a punctate pattern in both the cytoplasm and nucleus of proliferating C2C12 myoblasts. However, it does not localize with β1A integrins (red) at focal adhesions.

CSN3 Expression Increased 3.5-Fold in Response to Myocyte Differentiation

Given the spatial and temporal regulation of β1D integrin, we decided to examine if CSN3 expression is regulated in response to muscle cell differentiation. To accomplish this, C2C12 myoblasts were induced to differentiate into C2C12 skeletal myotubes. The C2C12 cells were examined at three time points: the proliferative stage (myoblasts), mid-way through differentiation (day 10), and when mature myotubes were present (day 14). Immunoblot analysis was used to determine the level of expression of CSN3 and its subcellular localization was determined by fractionation. In proliferating myoblasts, the expression of CSN3 was equally distributed between the cytosol and the nucleus. Surprisingly, there was a transient 3.5-fold increase of CSN3 expression during differentiation (44378 units versus 152153 units; Un-Scan-It), a loss of nuclear localization and small (but significant) redistribution to the membrane fraction (Figure 3). To determine if additional COP9 signalosome subunits redistributed in response to differentiation, the immunoblots were reexamined with different primary antibodies. In proliferating myoblasts, the majority of CSN1 was predominantly localized to the cytoplasm with a small amount in the nucleus (Figure 3). Upon differentiation, a small fraction of CSN1 was found in the membrane fraction. CSN4 was detected in several isoforms around 50 kDa and predominantly localized to the cytoplasm, in proliferating cells. However, after differentiation, a larger isoform CSN4 was found in the mitochondrial fraction and a significant amount was found in the nuclear and membrane fractions. In general, CSN5, CSN7, and CSN8 were found in the cytoplasmic, nuclear, and membrane fractions of both proliferating and differentiating C2C12 cells. Interestingly, a second larger isoform of CSN5 is found in the cytoplasm during differentiation. In contrast to CSN5 and CSN8, CSN7 was significantly decreased in the nuclear and membrane fractions during differentiation (day 10). Of particular note is that CSN5 and CSN8 consistently maintained a significant nuclear fraction. Unfortunately, CSN2 and CSN6 antibodies did not cross-react with the mouse subunits by immunoblot analysis. Overall, the increase and redistribution of CSN3 and CSN1 during differentiation correlated with β1D integrin expression.

Figure 3.  Representative fractionation and immunoblot analysis of the COP9 signalosome subunits in C2C12 skeletal myoblasts/tubes. Proliferating (P) C2C12 myoblast fractions. C2C12 myoblast/tube fractions after 10 days (D₁₀) of differentiation. C2C12 myotube fractions after 14 days (D₁₄) of differentiation. Four fractions were examined: cytoplasm, mitochondrial, nuclear, and membrane. Antibodies specific to various COP9 subunits are indicated. Note: CSN3 was only found with the membrane fraction in differentiated C2C12 cells when β1D integrin was expressed (data not shown).

The Distribution of CSN3 and Other COP9 Signalosome Subunits Are Distinct in Embryonic versus Adult Hearts

To further characterize the localization of CSN3 and the COP9 signalosome subunits relative to β1D integrin expression in vivo, subcellular fractionation studies were performed on embryonic (no β1D integrin present) versus adult (β1D integrin present) mouse hearts. To ensure enough protein would be recovered to do multiple analyses, two pools of embryonic hearts were fractionated and examined in duplicate. These results were compared to three to five individual adult mouse heart lysates. As illustrated in Figure 4A, almost all of CSN3 fractionates with the cytosol in embryonic hearts. However, in adult hearts 18% of CSN3 fractionates with the nucleus. In contrast, CSN5 in the same fractions showed a distinct profile (Figure 4A). In embryonic cardiac myocytes, 50% of CSN5 was cytosolic, 15% nuclear, and 27% microsomal compared to 76% cytosolic, 8% nuclear, and 11% microsomal in adult hearts. Examination of similar fractions with antibodies to CSN1, CSN4, CSN7, and CSN8 show a significant redistribution of CSN1 and CSN4 to the nuclear fraction of adult cardiac myocytes (Figure 4B versus C). These results show that adult hearts (which express β1D integrin) contain a significantly larger nuclear fraction of CSN3, CSN1, and CSN4 relative to embryonic hearts and the independent nature of each COP9 signalosome subunit.

Figure 4.  Fractionation and immunoblot analysis of the COP9 signalosome subunits in heart. Four adult mouse hearts and two pools of embryonic hearts (E16/17; 20 hearts examined in duplicate) were isolated and fractionated into soluble cytosol, mitochondria (5000×g), nuclear (600×g), and microsomal (membrane 100,000×g) fractions. Fractions were analyzed by immunoblot analysis with subunit-specific antibodies as indicated. Resultant blots were analyzed with Un-Scanit (Silk Scientific). Results show the mean distribution of all experiments ± SEM. (A) Histogram illustrating the amount (percentage of total) of CSN3 and CSN5 in embryonic fractions and the change in distribution in adult hearts. Notice the large shift of CSN3 from the cytosol to the nucleus of adult lysates. (B) Histogram showing the distribution of CSN1, CSN4, CSN7, and CSN8 in embryonic hearts. (C) Histogram showing the distribution of CSN1, CSN4, CSN7, and CSN8 in adult hearts. Note: β1D integrin is predominantly found in the membrane fractions of adult hearts (data not shown).

DISCUSSION

The purpose of this study was to identify potential integrin-related signaling pathways used in adult cardiomyocytes or striated muscle. We were able to identify proteins known to interact with β1D integrin and several new candidates. One candidate, CSN3, became the focus of more in depth study. In plants, CSN3 is part of a multifunctional complex that integrates signal transduction. This complex is called the COP9 signalosome. It is a 450-kDa multimer with eight unique core subunits (CSN1 to CSN8) (Bech-Otschir et al. 2002; Kapelari et al. 2000). Each subunit has one or more copies of a subunit-subunit interaction domain (termed either PCI or MPN sequences), and additional sequences with dissimilar and distinct functions (Tomoda et al. 2002; Tsuge et al. 2001; Wei and Deng 1999). The complex was originally identified in Arabidopsis as a complex that suppressed light-dependent development and it is evolutionarily related to the 26S proteasome (Seeger et al. 1998). COP9 signalosome subunits interact with a variety of nonsignalosome proteins including p53, c-Jun, integrin β2, Cul1, p27kip1, MIF1, Id3, IRF8/OCSBP, Hif1-a, Smad4, Bcl3, VPR, rLHR, SRC-1, thyroid hormone receptor, retinoic acid receptor, Rpn6, and eIF3e (Bech-Otschir et al. 2002; Chamovitz and Glickman 2002). These interactions do not involve the common PCI domain (Bianchi et al. 2000; Tomoda et al. 2002; Tsuge et al. 2001). Thus, the signalosome subunits/complex interact with signal transduction pathways controlling gene expression and the cell cycle (Chamovitz and Segal 2001; Claret et al. 1996; Naumann et al. 1999). For example, CSN1 inhibits JNK1, repressing Jun-dependent promoter activity (Spain et al. 1996). CSN2 competes with ligand for binding to thyroid hormone and retinoic acid receptors and CSN6 regulates the G2/M phase transition (Lee et al. 1995; Mahalingam et al. 1998).

The best characterized member of the COP9 signalosome complex is subunit 5 (CSN5/JAB1 [Jun activation domain-binding protein 1]) (Chamovitz and Segal 2001; Tomoda et al. 2002). CSN5 regulates both c-Jun and p27^kip-1 (p27). Strikingly, CSN5 also binds an integrin cytoplasmic domain, specifically, the β2 integrin (also known as lymphocyte adhesion integrin [LFA]-1) (Bianchi et al. 2000). CSN5 colocalized with LFA-1 at the cell membrane and when LFA-1 was activated, a portion of CSN5 relocated to the nucleus. Additionally, the resultant increase in the nuclear pool of CSN5 enhanced the binding of c-Jun–containing AP-1 complexes. Thus, signaling through LFA-1 may increase c-Jun–driven transcription by regulating the nuclear localization of CSN5. More recently, CSN5 has also been shown to bind to α_v integrin and regulate gene expression in a similar manner. (Levinson et al. 2004)

The binding of CSN3 to β1D suggests a role similar to the CSN5-integrin interaction. CSN3 is known to down-regulate p27, a cyclin-dependent kinase (Cdk) inhibitor, which is induced by a variety of antimitogenic stimuli (Tomoda et al. 2002). Additionally, the interaction of CSN3 with inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase gamma (IKKγ) negatively regulates tumor necrosis factor (TNF)-induced nuclear factor (NF)-κB activation pathways (Hong et al. 2001). Interestingly, CSN3 is part of the common region hemizygously deleted in patients with Smith-Magenis syndrome (SMS) (Elsea et al. 1999). SMS is a frequently observed human microdeletion syndrome with a complex phenotype, including mental retardation, craniofacial, eye, brain, cardiac, skeletal muscle, renal, forearm, and sleep abnormalities, hearing impairment, and scoliosis. Of patients with SMS, 37% have cardiac and skeletal muscle abnormalities. Therefore investigating the interaction of CSN3 with β1D integrin may give further insights into the SMS muscle disorders.

Our studies provide strong evidence for an interaction between β1D integrin and CSN3. CSN3 antibodies were able to coprecipitate both β1D integrin and CSN5 from adult heart lysates and immunofluorescence demonstrated that some CSN3 localized to the Z-bands. Subcellular fractionation studies indicated that <1% of CSN3 and 15% of CSN5 was in the adult heart membrane fractions. Together, these data indicate that CSN3 loosely associates with β1D integrin and that subcellular fractionation dissociates this interaction, creating a large cytoplasmic pool. Overall, the data presented showed dynamic changes in COP9 subunit expression and localization upon differentiation of C2C12 myoblasts and in embryonic hearts compared to adult hearts. When β1D integrin expression is stimulated by differentiation of C2C12 myoblasts, there is a 3.5-fold increase in CSN3. Additionally, in adult mouse hearts where the majority of β1D integrin is ligated, 18% of CSN3 fractionated with the nucleus. This is up from <1% in embryonic hearts. This, in conjunction with the prior evidence that CSN5 translocated to the nucleus in response to integrin ligation, leads us to hypothesize that CSN3 and possibly the COP9 signalosome complex acts as a signal transducer during cardiac myocyte differentiation and/or hypertrophy in response to β1D integrin activation/ligation.

A proposed model for the CSN3/COP9 signalosome function in cardiac myocytes is presented in Figure 5. In normal hearts, development and/or physiological stress induce cardiac hypertrophy. The engagement of β1 integrins with ECM ligand is required for this hypertrophic response. Pathological stress, such as volume and pressure overload or myocardial infarction, induces a similar response. We propose that the expression of β1D integrin in the perinatal heart and its subsequent ligation may lead to CSN3 translocation to the nucleus, stimulating physiological cardiac hypertrophy after birth. In turn, in the adult heart, a fraction of CSN3 is bound to β1D integrin. Upon activation of the integrin by pathological stress, CSN3 with or without other COP9 signalosome subunits may translocate to the nucleus. Once in the nucleus, a hypertrophic response could be initiated by activating transcription factors and/or by aiding in the export of transcription factors/repressors from the nucleus. In the future, it will be essential to determine if β1D integrin activation/ligation is responsible for the redistribution of CSN3 to the nucleus in cardiac myocytes. To examine this, we will undertake two approaches. First, we will examine the expression and localization of CSN3 in response to pressure or volume overload in cardiac myocytes (hypertrophy inducing conditions). Second, we will begin a series of experiments on isolated cardiac myocytes to determine if β1D integrin ligation and/or activation influence the activity of genes known to be regulated by CSN3 and/or the COP9 signalosome.

Figure 5.  Model of β1D integrin interaction with the COP9 signalosome complex. Cardiac myocytes respond to physiological stress (e.g., ischemia, pressure or volume fluctuations) through integrin activation. The resultant signaling causes translocation of CSN3-containing (sub)complexes to the nucleus, recruitment of more integrin to the cell surface, and strengthening of cell-ECM linkages. The recruitment of CSN3 to the nucleus results in the regulation of various transcripts responsive to the COP9 signalosome (sub)complex and the selective export of proteins from the nucleus for degradation pathways.

Acknowledgements

The authors would like to thank Iva Neveux and Jason Michaels for their technical assistance. The authors would also like to thank Dr. John A. McDonald, who provided both the support and laboratory space that made this project possible. This project was funded by an American Heart Association Beginning Grant-in-Aid, NIH R01 HL070511, NIH-NCRR P20 RR015581, and the University of Nevada Reno.