Atrogin-1 associates with α-actinin-2 and calcineurin A.
As a first step to investigate the functions of atrogin-1, we searched for partner proteins that interact directly with atrogin-1 via a yeast 2-hybrid assay. A systematic search for functional domains within atrogin-1 indicated the presence of potential protein-protein interaction domains, including sequences adjacent to the F-box that might be required for substrate recognition, in the C-terminus of atrogin-1 (12
). We therefore expressed the C-terminus of atrogin-1 as a fusion protein with the GAL4 DNA-binding domain in pGBKT7 vector as bait to screen an adult human heart cDNA library, which was cloned in pACT2. Among the potential binding partners found in this screen, the Z-disc–localizing actin-binding protein α-actinin-2 exhibited strong and specific interaction (Figure A). The interaction between atrogin-1 and α-actinin-2 was confirmed by in vitro glutathione-S
-transferase (GST) pull-down assays (Figure B). The interaction of atrogin-1 with α-actinin-2 was further tested by coimmunoprecipitation of epitope-tagged proteins in COS-7 cells. Xpress-tagged atrogin-1 efficiently coprecipitated hemagglutinin-tagged (HA-tagged) α-actinin-2 (Figure C). Atrogin-1 overexpression did not affect levels of α-actinin-2 protein in cardiomyocytes, which suggested to us that α-actinin-2 is not a ubiquitination target of atrogin-1 and might instead participate in multiprotein complexes with atrogin-1 and other proteins that are themselves targets for the activity of atrogin-1.
Figure 1 Molecular interaction of atrogin-1 with α-actinin-2 and calcineurin A (CnA). (A) Yeast 2-hybrid analysis of atrogin-1 interaction with α-actinin-2 (left) and calcineurin A (right). pGBKT7 and pACT2 are empty plasmids. pGAD-T7 vector expresses (more ...)
To explore the relevance of this association, we tested whether endogenous atrogin-1 can interact with α-actinin-2 under physiologic conditions. We performed immunoprecipitation and immunoblotting experiments using extracts prepared from rat neonatal cardiomyocytes and found that α-actinin-2 efficiently coimmunoprecipitated with endogenous atrogin-1 (Figure D). Immunostaining indicated that endogenous atrogin-1 expression is concentrated at the Z-disc in a pattern that overlaps with that of α-actinin-2 (Figure E). Among known interaction partners of α-actinin-2 in striated muscle, calcineurin A stands out because of its role in coordinating myocyte gene expression programs that determine cell size (5
). We therefore compared the immunolocalization of atrogin-1 and calcineurin A and found that these proteins were also similarly distributed and concentrated at the Z-disc. This expression pattern suggested that atrogin-1 and calcineurin A may also directly interact. The association of atrogin-1 and calcineurin A was confirmed in yeast 2-hybrid and GST-binding assays (Figure , A and B) and coimmunoprecipitation experiments (Figure , C and D). To test whether these interactions are direct, we performed in vitro binding reactions using recombinant proteins. Both α-actinin-2 and calcineurin A bound efficiently to a GST–atrogin-1 fusion protein in these assays (Figure F). Taken together, these experiments indicate that atrogin-1 is a bona fide interaction partner of both α-actinin-2 and calcineurin A within the cardiac myofibril, most likely participating in a ternary multiprotein complex.
Interaction domains of atrogin-1, calcineurin A, and α-actinin-2.
To define the regions in atrogin-1 that are responsible for binding with calcineurin A, we tested serially deleted atrogin-1 mutants expressed in bacteria as GST fusion proteins for their ability to bind calcineurin A (Figure A, top). GST pull-down assays revealed that the F-box domain of atrogin-1 is dispensable for protein-protein interactions; amino acids 272–283 of atrogin-1 are necessary for the interaction of atrogin-1 with α-actinin-2 (Figure A, middle), whereas the adjacent amino acids 284–345 of atrogin-1 are required for binding to calcineurin A (Figure A, bottom). To further confirm the region of atrogin-1 required for association with calcineurin A, a full-length calcineurin A GST fusion was tested in GST pull-down assays with a series of atrogin-1 deletion mutants expressed in vivo in COS-7 cells. Amino acids 284–345 of atrogin-1 were also required for calcineurin A binding in this assay (Figure B). These data, summarized in Figure C, indicate that topologically adjacent domains of atrogin-1 are required for α-actinin-2 and calcineurin A binding.
Figure 2 Mapping the interaction domains of atrogin-1, α-actinin-2, and calcineurin A. (A) The residues of atrogin-1 required for binding to α-actinin-2 and calcineurin A were determined with GST pull-down assays. GST–atrogin-1 fusion proteins (more ...)
To identify the sequences in calcineurin A required for interaction with atrogin-1, a series of truncations of calcineurin A expressed as GST fusions (Figure D, top) was used in pull-down assays with Myc–atrogin-1 expressed in COS-7 cells. The region of calcineurin A containing amino acids 287–337 was required for binding to atrogin-1 (Figure D, bottom). These data (summarized in Figure E) indicate that the atrogin-1–interacting domain of calcineurin A resides within the C-terminal portion of its catalytic domain and is distinct from the binding site for other known calcineurin-interacting proteins.
Atrogin-1 reduces endogenous calcineurin A protein levels and decreases calcineurin A activity in neonatal cardiomyocytes.
Because atrogin-1 is a putative ubiquitin ligase component, we tested the ability of atrogin-1 to regulate the level of endogenous calcineurin A protein in cardiomyocytes infected with adenoviruses Ad-GFP and Ad-atrogin-1-GFP after FBS stimulation. There are 3 known genes that encode catalytic calcineurin A proteins, calcineurin Aα, calcineurin Aβ, and calcineurin Aγ. Infection with Ad-atrogin-1-GFP caused a dose-dependent reduction in the level of total endogenous calcineurin A, including both calcineurin Aα and calcineurin Aβ (Figure A), demonstrating that atrogin-1 reduces calcineurin A protein levels in cardiomyocytes. To establish the cellular consequences of the effects of atrogin-1 on endogenous calcineurin A protein, we measured calcineurin phosphatase activity in cardiomyocytes infected with either Ad-GFP or Ad-atrogin-1-GFP after FBS stimulation. Calcineurin phosphatase activity was significantly decreased in a dose-dependent manner in Ad-atrogin-1-GFP–infected compared with Ad-GFP–infected cardiomyocytes (Figure B; P < 0.05 or 0.001). To determine whether atrogin-1 might influence endogenous proteins in other signaling pathways, cardiomyocytes were infected with adenoviruses Ad-GFP and Ad-atrogin-1-GFP after FBS stimulation. Immunoblotting indicated that atrogin-1 does not appear to have a major role in regulating the levels of other endogenous proteins, including α-actinin-2, ERK, JNK, PKCδ, p38 MAPK, and Akt (Figure C); this demonstrates at least some degree of specificity for the effect of atrogin-1 on calcineurin A.
Figure 3 Atrogin-1 regulates endogenous calcineurin A levels and activity. (A) Cardiomyocytes were infected with increasing multiplicities of infection (MOI) of Ad-atrogin-1-GFP and Ad-GFP with FBS stimulation for 36 hours. The levels of expressed atrogin-1 and (more ...)
In addition, we tested whether endogenous atrogin-1 is critical in regulating the level of calcineurin A in cardiomyocytes using an adenoviral small interfering RNA (siRNA) strategy. siRNA targeted to unique sequences in atrogin-1 effectively reduced atrogin-1 protein levels by 80% in FBS-treated cardiomyocytes compared with an adenovirus expressing a control siRNA, and suppression of atrogin-1 protein was accompanied by increased calcineurin A expression (Figure D). Similarly, downregulation of atrogin-1 resulted in a significant increase in calcineurin phosphatase activity, whereas the control siRNA had no effect (Figure E). Together, these data demonstrate that atrogin-1 inhibits endogenous calcineurin A activity by reducing calcineurin A levels in FBS-stimulated cardiomyocytes.
Atrogin-1 represses calcineurin A–dependent transcriptional responses in neonatal cardiomyocytes.
Nuclear translocation and activation of target genes by NFAT depends on calcineurin signaling in cardiomyocytes (4
), which led us to test the effects of atrogin-1 on calcineurin-NFAT function. Cardiomyocytes were cotransfected with reporter construct pIL2-Luc (containing an NFAT-binding site from the IL-2 promoter upstream of the luciferase gene) and plasmids expressing activated calcineurin A alone or in combination with atrogin-1. Atrogin-1 potently repressed calcineurin-dependent transactivation (Figure A). In contrast, transfection of a plasmid expressing atrogin-1 siRNA enhanced calcineurin-dependent transactivation by more than 45% (Figure B). Interestingly, atrogin-1 had no effect on pIL2-Luc transactivation by constitutively activated NFAT (ΔNFAT, Figure C), which suggests that atrogin-1 does not impair NFAT-dependent transactivation directly and instead exerts its effects at the level of calcineurin A activity.
Figure 4 Atrogin-1 blocks calcineurin-dependent transcriptional responses and nuclear translocation of NFATc4 but does not inhibit a constitutively active form of NFAT in cardiomyocytes. (A) The calcineurin-dependent transcriptional response was measured in cardiomyocytes (more ...)
To extend this observation, we tested whether atrogin-1 alters nuclear translocation of NFAT, which occurs rapidly after dephosphorylation by calcineurin and is required for its activity. Neonatal cardiomyocytes were immunostained 36 hours after plasmid transfection with an antibody against NFATc4 (Figure D, red), and nuclei were counterstained with DAPI (Figure D, blue). In cells transfected with vector alone, NFATc4 was localized to the cytoplasm. Following stimulation with activated calcineurin A, NFATc4 translocated to the nucleus, as has been described previously (15
). Cotransfection with activated calcineurin A and atrogin-1 together completely blocked nuclear translocation of NFATc4, which was similar to the effects of the calcineurin A inhibitor cyclosporin A on NFAT mobilization. We measured transfected NFATc4 translocation quantitatively in cardiomyocytes by Western blot analysis to confirm that atrogin-1 selectively blocked calcineurin-dependent NFAT nuclear translocation (Figure E). These experiments indicate that atrogin-1 can inhibit calcineurin-dependent nuclear translocation of NFAT in cardiomyocytes, without having direct effects on NFAT or on other pathways that affect NFAT activity.
Atrogin-1 inhibits calcineurin-dependent cardiomyocyte hypertrophy.
Previous studies have shown that calcineurin is activated by hypertrophic agonists in cultured cardiomyocytes (4
). To determine directly whether atrogin-1 regulates the agonist-induced calcineurin signaling pathway that mediates cardiac hypertrophy, cardiomyocytes were infected with Ad-GFP or Ad-atrogin-1-GFP and subsequently treated with phenylephrine (PE) or FBS for 36 hours. Calcineurin phosphatase activity was increased approximately 5-fold by stimulation with PE and FBS in cardiomyocytes infected with Ad-GFP (Figure A). In comparison, Ad-atrogin-1-GFP infection completely blocked the increases in calcineurin activity. In addition, Western blotting demonstrated that Ad-atrogin-1-GFP infection reduced basal calcineurin A protein levels and repressed agonist-induced induction of calcineurin A protein in response to PE and FBS stimulation (Figure B).
Figure 5 Ad-atrogin-1-GFP infection blocks agonist-induced cardiomyocyte hypertrophy. (A) Cardiomyocytes were infected with Ad-GFP or Ad-atrogin-1-GFP and were stimulated with PE (100 μM) or FBS for 36 hours. Calcineurin phosphatase activity was measured (more ...)
To investigate the effects of atrogin-1 on the morphologic changes induced by hypertrophic agents, cardiomyocytes were infected with Ad-GFP, Ad-atrogin-1-GFP, or no virus and stimulated with PE or FBS for 36 hours, fixed, and stained with anti–sarcomeric α-actinin antibody. Whereas either medium alone (Figure C, top) or infection with Ad-GFP (Figure C, middle) had no effect on agonist-induced hypertrophy, cardiomyocytes infected with Ad-atrogin-1-GFP repressed PE- or FBS-induced hypertrophy (Figure C, bottom). Measurement of cell surface area indicated that cells infected with Ad-atrogin-1-GFP did not increase in size after PE or FBS stimulation (Figure D). Furthermore, infection of cardiomyocytes with Ad-siRNA-atrogin-1 enhanced PE-induced cardiomyocyte hypertrophy (Figure E). Cell surface area was increased by 1.4-fold in Ad-siRNA-atrogin-1–infected cardiomyocytes compared with cells infected with Ad-siRNA-control (Figure F).
The preceding experiments suggest that inhibition of calcineurin activity by atrogin-1 is sufficient to inhibit agonist-induced hypertrophy in cardiomyocytes. To confirm that the effects of atrogin-1 on cell size reflect inhibition of the hypertrophic fetal gene expression profile in neonatal cardiomyocytes, cells were infected with either Ad-GFP or Ad-atrogin-1-GFP, stimulated by agonists, and immunostained with antibody against atrial natriuretic factor (ANF). Cardiomyocytes infected with Ad-atrogin-1-GFP did not appreciably increase ANF protein in response to stimulation by PE and FBS (Figure A, bottom), whereas ANF was abundantly expressed in cells conditioned with media alone or Ad-GFP (Figure A, top and middle). Quantitative analysis indicated that the number of ANF-positive cells was not increased above base line after PE and FBS treatment in cells treated with Ad-atrogin-1-GFP (Figure B). Conversely, cardiomyocytes infected with Ad-siRNA-atrogin-1 demonstrated a modest but statistically significant increase in ANF positivity in response to stimulation by PE (Figure , C and D). In addition, RT-PCR analysis demonstrated that atrogin-1 prevented induction of the mRNA for hypertrophy markers ANF, β–myosin heavy chain (β-MHC), and skeletal α-actin (Figure E). Taken together, these data indicate that atrogin-1 blocks calcineurin A–dependent, agonist-induced hypertrophy and prevents hypertrophic gene induction in vitro, which is consistent with its ability to inhibit calcineurin A activity and decrease calcineurin A protein levels.
Figure 6 Ad-atrogin-1-GFP inhibits fetal gene expression in cardiomyocytes. (A) Cardiomyocytes were infected and were stimulated as indicated. Cells were fixed and stained with antibody against ANF (perinuclear red signal). A representative field is shown for (more ...) The SCFatrogin-1 complex has an E3 ubiquitin ligase activity and promotes ubiquitination of calcineurin A in vitro.
Atrogin-1 contains an F-box domain (amino acids 228–267), which is characteristic of proteins that are components of SCF ubiquitin ligase complexes (11
). F-box proteins are adaptors that associate with Skp1, Cul1, and Roc1 of the SCF complex through the F-box and simultaneously directly bind and recruit proteins to be ubiquitinated (17
). F-box proteins therefore provide substrate specificity to SCF complexes, and they are also the rate-limiting components for substrate ubiquitination. To determine whether atrogin-1 associates with other components of the SCF complex, we purified GST–atrogin-1 fusion proteins in bacteria (WT, lacking the F-box, or lacking the calcineurin-binding domain; Figure A) and tested formation of the Skp1-Cul1-Roc1 complex by coimmunoprecipitation of epitope-tagged proteins. Under these conditions, Myc-tagged Cul1 immunoprecipitated T7-tagged Skp1 and HA-tagged Roc1 (Figure B). GST pull-down assays demonstrated that Skp1, Cul1, and Roc1 proteins present in cell extracts bound efficiently to immobilized GST–atrogin-1 but not GST alone, and only weakly to a fusion protein lacking the F-box (Figure C). These observations extend previous reports (11
) and confirm that atrogin-1 incorporates into SCF complexes in an F-box–dependent fashion.
Figure 7 Atrogin-1 participates in an SCFatrogin-1 complex that ubiquitinates calcineurin A in vitro. (A) GST–atrogin-1 (WT), atrogin-1 ΔF-box, and atrogin-1 1–284 were purified from bacteria. The purity of each purified protein was verified (more ...)
Ubiquitin ligases interact functionally with specific E2 ubiquitin-conjugating enzymes to facilitate substrate recognition and ubiquitin transfer (18
). We tested the ability of the SCFatrogin-1
complex to assemble polyubiquitin chains in the presence of a panel of E2 ubiquitin-conjugating enzymes. We found that UbcH3/CDC34 was able to interact efficiently with the SCFatrogin-1
complex. However, other ubiquitin-conjugating enzymes — including UbcH5, UbcH4, E2-20, and E2-25 — were ineffective in facilitating SCFatrogin-1
complex–mediated polyubiquitin chain formation (data not shown). To determine whether atrogin-1–containing SCFatrogin-1
complexes have intrinsic ubiquitin ligase activity, we performed an in vitro ubiquitination assay with the immunoprecipitated Skp1-Cul1-Roc1 complex, purified GST–atrogin-1 or GST alone, recombinant E1, E2 (UbcH3/CDC34), and ubiquitin. SCFatrogin-1
complex–dependent multiubiquitin chain assembly (an established marker of ubiquitin ligase activity [ref. 18
]) was readily detected and dependent on the presence of atrogin-1 and E1 and E2 enzymes (Figure D). Thus, atrogin-1 assembles a functional ubiquitin ligase complex in vitro.
To examine directly whether the SCFatrogin-1
complex can promote calcineurin A ubiquitination in vitro, we reconstituted a reaction using the immunoprecipitated Skp1-Cul1-Roc1 complex, GST–atrogin-1, and recombinant E1 and UbcH3. Reactions were performed with calcineurin A expressed as a GFP fusion in COS-7 cells. Ubiquitination of calcineurin A was evaluated by immunoblotting. A high–molecular weight species indicative of polyubiquitin chain assembly was detected in the presence of the SCFatrogin-1
complex, ubiquitin, E1, and UbcH3; Roc1 (the RING finger component of the reaction) dramatically enhanced the ubiquitination of calcineurin A (Figure E). Omitting ubiquitin, E1, UbcH3, or GST–atrogin-1 abolished formation of the polyubiquitin chain, and GFP alone was not efficiently ubiquitinated under these conditions (data not shown), which is consistent with previous observations that unmodified GFP is not a substrate for the ubiquitination machinery (19
). More importantly, omission of the SCF complex also prevented calcineurin A polyubiquitination. Thus, the observed product in our assays represents the polyubiquitinated form of calcineurin A. These data suggest that an SCF complex containing atrogin-1 as an F-box component catalyzes polyubiquitination of calcineurin A.
Since efficient transfer of ubiquitin to calcineurin A requires the direct interaction of calcineurin A with the UbcH3/SCFatrogin-1 machinery, atrogin-1 mutants lacking the F-box or the calcineurin-binding site should be resistant to ubiquitination. To test this hypothesis, we engineered 2 mutant forms of atrogin-1, containing either a deletion of the entire F-box (atrogin-1 ΔF-box) or atrogin-1 1–284, which lacks the calcineurin A–binding site. In the presence of the SCFatrogin-1 complex, atrogin-1 dramatically enhanced the ubiquitination of calcineurin A in a time-dependent manner (Figure F). Moreover, deletion of the F-box or the calcineurin A–binding site markedly attenuated atrogin-1–mediated enhancement of calcineurin A polyubiquitination (Figure G, compare second and third lanes with first lane). Collectively, these data indicate that atrogin-1 functions as a limiting component of an F-box–dependent E3 ubiquitin ligase and that the F-box domain and calcineurin A–binding site (amino acids 284–345) of atrogin-1 are both required for calcineurin A ubiquitination.
Atrogin-1 transgenic mice are resistant to pressure overload–induced cardiac hypertrophy.
To determine whether atrogin-1 is also capable of repressing calcineurin A activity and cardiac hypertrophy in vivo, we created transgenic mice that overexpress atrogin-1 in the heart under the control of the α-myosin heavy chain (α-MHC) promoter (data not shown) (20
). Two independently derived atrogin-1 founders (Tg5 and Tg14) contained approximately 3 and 15 copies, respectively, of the transgene as assessed by Southern blot analysis (data not shown). Northern blot analysis demonstrated that expression of the atrogin-1 transgene was restricted to the heart (data not shown), and expression of atrogin-1 in Tg5 and Tg14 was increased 4.4- and 12.2-fold, respectively, compared with that in nontransgenic mice (Figure A). (Because the transgenic lines showed a similar phenotype, we will only describe data from line Tg14.) Transgenic mice were phenotypically indistinguishable from their WT littermates, and none of the transgenic lines was characterized by premature lethality. There was no significant difference in heart weight/body weight ratios between WT and atrogin-1 transgenic mice, and echocardiography failed to reveal any abnormalities in cardiac function in transgenic mice up to 8 months of age (data not shown). This is not surprising, since calcineurin inhibitors such as cyclosporin A do not have significant effects on base-line cardiac function (1
Figure 8 Effects of overexpressed atrogin-1 on calcineurin A activity, protein level, and cardiac function. (A) Northern blot analysis of transgene expression. Ten micrograms of total RNA from heart of Tg5 and Tg14 (second and third lanes) and WT (first lane) (more ...)
We wondered whether these mice might be sensitized to hypertrophic stimuli based on our observations in vitro, as would be predicted if atrogin-1 is a suppressor of calcineurin A activity. Transgenic mice and age- and sex-matched WT littermates were subjected to aortic banding to induce pressure overload (or sham surgery) for 14 days. In response to aortic banding, calcineurin A activity was elevated about 3-fold in WT mice, but cardiac calcineurin A activity was only minimally increased in hearts from atrogin-1 transgenic mice (Figure B). Similarly, calcineurin A protein levels were increased by pressure overload in WT mice but were reduced by 70% in atrogin-1 transgenic mice (Figure C). These data confirm our in vitro observations and indicate that atrogin-1 inhibits calcineurin A activity and reduces its protein levels in vivo.
To assess cardiac function accurately, transthoracic echocardiography was performed. In the aortic-banded group, atrogin-1 transgenic mice developed left ventricular dilation, characterized by increased left ventricular end-systolic dimensions (Figure D; Table ). In contrast, WT mice had reduced left ventricular dimensions and increased septal thickness 14 days after banding. Furthermore, atrogin-1 transgenic mice had decreased interventricular septal thickness in systole compared with nontransgenic mice and had a dramatic reduction in ejection fraction, as estimated by fractional shortening (37% versus 59% in WT mice) after aortic banding. No differences in echocardiographic parameters were detected in sham-treated mice in either group. These measurements indicate that atrogin-1 transgenic mice display left ventricular dilatation and reduced cardiac function rather than the expected ventricular hypertrophy after aortic banding.
Echocardiography of dimensions and function in WT and atrogin-1 transgenic mice after 2-week aortic banding
On gross postmortem examination, atrogin-1 transgenic mice did not develop cardiac hypertrophy after banding (Figure A, top) and had a 25% reduction in heart weight/body weight ratios compared with age-matched aortic-banded WT mice (Figure A, bottom). There were no significant differences in heart size or heart weight/body weight ratios between transgenic mice and nontransgenic mice after sham surgery. Histologic analysis of aortic-banded transgenic hearts revealed that the walls of the septum and left ventricular chambers were much thinner than in nontransgenic hearts (Figure B, top). Masson’s trichrome staining of heart sections from transgenic mice indicated reduced interstitial cell fibrosis (Figure B, middle), which is reminiscent of the reduction in collagen deposition observed in MCIP1-overexpressing hearts (21
). Histologic sections were also stained with wheat germ agglutinin–TRITC to determine cross-sectional areas of cardiac myocytes (Figure B, bottom). Myocyte cross-sectional area was not significantly different between transgenic and nontransgenic mice in sham groups; however, the cross-sectional area of cardiac myocytes in nontransgenic hearts was significantly increased, whereas myocytes in transgenic hearts increased very little compared with those in sham-treated mice, with the result that the cross-sectional area of cardiac myocytes in aortic-banded transgenic hearts was 41% less than in controls (Figure C).
Figure 9 Atrogin-1 blunts the hypertrophic response to pressure overload. (A) Representative heart sizes and heart weight/body weight ratios of atrogin-1 transgenic mice were compared with those of nontransgenic mice. Eight-week-old mice were subjected to thoracic (more ...)
We also examined changes in fetal gene expression to determine whether atrogin-1 transgenic hearts have a stress-associated expression pattern. RT-PCR analysis was performed to quantify mRNA levels of hypertrophic genes from hearts of transgenic and nontransgenic mice after banding for 2 weeks (Figure D). Levels of cardiac ANF, β-MHC, and skeletal α-actin mRNA, which are typically induced during hypertrophy, were decreased by more than 55% in hearts from atrogin-1 transgenic mice compared with nontransgenic hearts. Collectively, these data indicate that atrogin-1 acts as a suppressor of the molecular program for cardiac hypertrophy in vivo, at least in part by suppressing calcineurin A protein expression and thus inhibiting calcineurin A activity. In response to pressure overload, transgenic hearts expressing elevated levels of atrogin-1 fail to increase wall thickness, thereby increasing wall stress and causing cardiac chamber enlargement.