p22phox is upregulated in the failing human heart and in cardiomyocytes in response to pressure overload
The level of p22 was evaluated in heart samples obtained from patients with dilated or ischemic cardiomyopathy and from donors with normal hearts. The level of p22 was significantly higher in the patients' hearts than in the control hearts (Fig. 1a,b and Supplementary Table 1).
To assess the effects of pressure overload, we subjected mice to transverse aortic constriction (TAC) or sham surgery. p22 expression in the heart was significantly elevated at both mRNA and protein levels after 1 week and 4 weeks of TAC (Fig. 1c-e). p22 was also upregulated at the mRNA (Fig. 1f) and protein (Fig. 1g,h) levels in cardiomyocytes isolated from wild-type (WT) mice after 1 week of TAC, suggesting that p22 is produced and upregulated in cardiomyocytes in response to pressure overload.
To investigate the role of endogenous p22 in the heart, we generated cardiomyocyte-specific p22 knockout (p22 cKO) mice (Supplementary Fig. 1a,b). As p22 is also expressed in non-myocytes, residual protein was still detectable in whole hearts (Supplementary Fig. 1c,d). However, the p22 mRNA was significantly reduced in isolated cardiomyocytes from p22 cKO mice (Supplementary Fig. 1e), and TAC-induced upregulation of p22 observed in control mice was absent in p22 cKO cardiomyocytes (Supplementary Fig. 1f), confirming successful gene deletion. At baseline, p22 cKO mice showed normal cardiac structure and function (Supplementary Tables 2 and 3). However, following 1 week or 4 weeks of TAC, p22 cKO mice developed more severe left ventricle (LV) dilation (increased LV end-diastolic diameter (LVEDD) and LV end-systolic diameter (LVESD)) and greater LV dysfunction (reduced LV ejection fraction (LVEF)) than control mice (Fig. 2a-e). In addition, lung congestion and LV end-diastolic pressure (LVEDP) were significantly increased after 4 weeks of TAC in p22 cKO mice (Fig. 2f,g), and aortic pressure gradients were reduced (Fig. 2h), consistent with impaired contractility. These mice also showed a blunted inotropic response to dobutamine (Supplementary Fig. 2a-d), indicating reduced cardiac reserve. Histological analyses show that the cardiomyocyte cross-sectional area (CSA) was comparable between groups after TAC (Fig. 2i,j), but fibrosis was significantly exacerbated in p22 cKO hearts (Fig. 2k,l). Importantly, survival after TAC was markedly lower in p22 cKO mice (Fig. 2m). Collectively, these findings show that although p22 is dispensable at baseline, its absence impairs the heart's ability to compensate under pressure overload, leading to worsened LV dysfunction, enhanced fibrosis and increased mortality.
To uncover how p22 loss contributes to cardiac dysfunction under pressure overload, we searched for p22-binding proteins. Lysates from adult mouse cardiomyocytes were incubated with or without Flag-tagged p22, followed by anti-Flag immunoprecipitation. Interacting proteins were identified using liquid chromatography with tandem mass spectrometry (LC-MS/MS) with spectral counting. We identified 555 proteins by LC-MS/MS, of which 231 proteins bound well to Flag-p22 (ratio > 1.5) with respect to the control group (Supplementary Table 4). The LC-MS/MS analysis showed that SERCA2a was the protein that bound most abundantly to Flag-p22 in the mouse heart.
To further validate the interaction between p22 and SERCA2a, co-immunoprecipitation revealed that SERCA2a was specifically detected in anti-p22 immunoprecipitates, but not when a nonspecific control antibody was used (Fig. 3a). Conversely, p22 was detected in the immunoprecipitate with anti-SERCA2a antibody (Fig. 3b). These results suggest that p22 and SERCA2a physically interact with one another. A direct interaction between SERCA2a and p22 was confirmed by in vitro binding assays using recombinant proteins (Fig. 3c). This interaction was further validated in cells using a proximity ligation assay (PLA), which revealed robust signals in the perinuclear region of cardiomyocytes when both anti-p22 and anti-SERCA2a antibodies were applied, consistent with the complex formation in the sarcoplasmic reticulum (Fig. 3d). To map the interaction domain, a series of SERCA2a deletion mutants were tested. p22 bound to full-length SERCA2a, as well as SERCA2a 1-666, 334-666 and 334-998 constructs, but not to SERCA2a 1-333, identifying the 334-666 amino acid region as necessary for binding (Fig. 3e).
As SERCA2a transports cytosolic Ca into the SR, Ca transients were measured in isolated cardiomyocytes. At baseline, p22 cKO cardiomyocytes showed a significantly reduced peak amplitude compared with the control (Fig. 4a,b). The peak amplitude of the Ca transient after caffeine administration, indicating the total SR Ca content, was then evaluated. At baseline, a rapid caffeine spike emptying the intracellular Ca store in single cardiomyocytes revealed a significantly reduced SR Ca content in the p22 cKO mouse cardiomyocytes compared with in the control mouse cardiomyocytes (Fig. 4c). There was no significant difference in the percentage of fractional Ca release between control and p22 cKO cardiomyocytes at baseline (Fig. 4d). The decay constant of the Ca transient (tau) was significantly greater in cardiomyocytes isolated from p22 cKO mice than in those from control mice at baseline (Fig. 4e). The Ca transients were also assessed in isolated cardiomyocytes after 1 week of pressure overload (TAC). The peak amplitude and the total SR Ca content were significantly lower in the p22 cKO mouse cardiomyocytes than in the control mouse cardiomyocytes after 1 week of TAC (Fig. 4a-c). There was no significant difference in the percentage of fractional Ca release between control and p22 cKO cardiomyocytes after 1 week of TAC (Fig. 4d). The tau was significantly greater in cardiomyocytes isolated from p22 cKO mice than in those from control mice after TAC (Fig. 4e). These results are consistent with the notion that the activity of SERCA2a is reduced in p22 cKO cardiomyocytes both at baseline and after TAC. On the other hand, the tau for the decay of Ca transient after caffeine administration, reflecting the activity of the Na/Ca exchanger (NCX), was not significant between control and p22 cKO mice at baseline and after TAC (Fig. 4f), although it was significantly smaller after TAC compared with the sham operation in p22 cKO mice, suggesting compensatory activation of NCX.
Given altered calcium handling in p22 cKO mice, we examined how endogenous p22 regulates key cardiac calcium handling genes at baseline and 1 week after pressure overload. The relative mRNA levels of SERCA2a, PLN, NCX and ryanodine receptor (RyR) were not significantly different between control and p22 cKO mice at baseline (Fig. 5a,b and Supplementary Fig. 2e,f). Although the levels of SERCA2a, PLN and RyR mRNA expression were significantly decreased in both p22 cKO and control mice after 1 week of pressure overload, there was no significant difference between p22 cKO and control mice (Fig. 5a and Supplementary Fig. 2f). As p22 interacts with SERCA2a, we examined SERCA2a protein levels. Despite there being no change in the mRNA, the SERCA2a protein was significantly reduced in p22 cKO mice compared with controls, both at baseline and after pressure overload (Fig. 5c,d). Similarly, in cultured neonatal rat cardiomyocytes with small interfering RNA-mediated p22 knockdown, SERCA2a mRNA levels were unaltered but the protein level of SERCA2a was significantly smaller after p22 knockdown (Fig. 5e-g). These results suggest that p22 may control the SERCA2a protein level through posttranscriptional mechanisms. PLN inhibits SERCA2a, but this inhibition is relieved upon phosphorylation by protein kinase A (ref. ). In p22 cKO mice, phosphorylated phospholamban protein levels were significantly higher than in controls, suggesting a compensatory response to reduced SERCA2a (Fig. 5h,i). Relative ATPase activity in isolated SR fractions from control and p22 cKO mice showed no significant difference when normalized to SERCA2a levels (Fig. 5j).
To show the role of SERCA2a downregulation in cardiac dysfunction in p22 cKO mice under pressure overload, we performed rescue experiments using cardiomyocyte-specific SERCA2a expression via AAV9-cTnT-SERCA2a. Injection of AAV9-cTnT-SERCA2a increased cardiac SERCA2a expression in both control and p22 cKO mice in both the sham and TAC groups (Extended Data Fig. 1a-d). No significant difference in LV tissue weight (LVW)/tibia length (TL) was observed in the presence of SERCA2a overexpression either after the sham operation or 4 weeks after TAC compared with in an eGFP overexpression group (control) (Extended Data Fig. 1e). The increase in lung weight observed after 4 weeks of TAC in p22 cKO mice was attenuated in p22 cKO mice injected with AAV9-cTnT-SERCA2a (Extended Data Fig. 1f). Moreover, LVEF after 4 weeks of TAC was significantly reduced in p22 cKO mice, but not in p22 cKO mice injected with AAV9-cTnT-SERCA2a (Extended Data Fig. 1g). These data suggest that supplementation of SERCA2a levels in cardiomyocytes rescues cardiac dysfunction associated with a lack of p22. Thus, SERCA2a downregulation has an important role in mediating cardiac dysfunction in p22 cKO mice.
We investigated how p22 regulates SERCA2a protein expression, given that SERCA2a is sensitive to oxidative modification, which can promote its degradation or reduce its activity in a context-dependent manner. As p22 interacts with SERCA2a, and its loss reduces SERCA2a protein levels, we hypothesized that p22 modulates SERCA2a oxidation.
To assess oxidative stress, we measured total dityrosinated proteins and tissue-released HO in p22 cKO mice. While both markers increased after pressure overload in control and p22 cKO mouse hearts, their levels were significantly lower in p22 cKO mice (Fig. 6a-c). However, ER-localized HyPer fluorescence revealed higher HO signals in p22 knockdown cardiomyocytes compared with controls (Fig. 6d,e and Extended Data Fig. 2a,b), indicating that p22 loss increases local oxidative stress in the ER, despite lower total cardiac oxidative stress.
As multiple cysteine residues are subjected to oxidative modifications, the extent of cysteine oxidation in SERCA2a was evaluated by blocking free reactive cysteine thiols with biotinylated iodoacetamide (BIAM) labeling and pull-down. There was significantly less BIAM-labeled SERCA2a in the heart in p22 cKO mice than in control mice (Fig. 6f,g), indicating that the SERCA2a cysteine residues are more oxidized in p22 cKO mice than in control mice.
We then evaluated redox-sensitive cysteines in p22 cKO and control mouse hearts. Reduced free cysteines in the heart at the time of collection were irreversibly labeled with BIAM, whereas oxidized cysteines were first reduced with dithiothreitol (DTT) and then labeled with biotin-free IAM. LC-MS/MS showed that BIAM labeling of 7 cysteines (Cys344/349, Cys447, Cys471, Cys498, Cys560 and Cys669) was reduced by 30% or more in p22 cKO mouse hearts compared with in control hearts, excluding low-label-efficiency (<1%) cysteines. Although several cysteines in SERCA2a were more oxidized in p22 cKO mice than in control mice, Cys498 was oxidized to the greatest extent (Fig. 6h). To evaluate the functional significance of Cys498 oxidation, we generated adenoviruses harboring SERCA2a-C498S (Ad-Flag-SERCA2a-C498S) and studied how the Cys498 oxidation resistant mutation affects the overall oxidation status of SERCA2a. Cardiomyocytes transduced with Ad-Flag-SERCA2a-C498S showed significantly less BIAM labeling of Flag-SERCA2a than those transduced with Ad-Flag-SERCA2a WT (Fig. 6i,j). Furthermore, cardiomyocytes transduced with Ad-Flag-SERCA2a-C498S showed no further reduction in the amount of BIAM-labeled Flag-SERCA2a in the absence of p22 compared with in the presence of p22. These results indicate that Cys498 is a major site of cysteine oxidation in SERCA2a in the absence of p22. In addition, the decrease in the SERCA2a level in the absence of p22 was significantly alleviated in the presence of SERCA2a-C498S in cardiomyocytes (Fig. 6k-m). These results suggest that the absence of p22 results in the oxidation of SERCA2a at Cys498 and negatively affects the protein level of SERCA2a.
To identify the source of HO in the ER in the absence of p22, we evaluated whether the increased ER-localized oxidative stress is mediated by Nox4, a partner of p22 in the ER. We have shown previously that a lack of anti-apoptotic HS-1-associated protein X-1 (HAX-1), a regulator of ER-mediated cell survival, leads to increased oxidative stress in the ER through a Nox4-dependent mechanism. The ER-localized HyPer fluorescence ratio was not affected by Nox4 downregulation in either the presence or absence of p22 knockdown (Extended Data Fig. 2a,b). Although the knockdown of p22 decreased the protein levels of both Nox4 and SERCA2a, the knockdown of Nox4 did not decrease SERCA2a levels (Extended Data Fig. 2c-e). p22 downregulation also decreased Nox2 protein levels (Extended Data Fig. 3a,b). These data suggest that upregulation of HO and downregulation of SERCA2a in response to p22 downregulation are independent of Nox4 or Nox2. Furthermore, knockdown of p22 did not alter the mitochondria-localized Mito-HyPer fluorescence ratio compared with the control (Extended Data Fig. 3e,f), excluding mitochondria-derived oxidative stress in the p22-deficient condition. Interestingly, the ER-localized HyPer fluorescence ratio was high when SERCA2a was downregulated (Extended Data Fig. 3e,f). Thus, increased oxidative stress in the ER in response to downregulation of p22 is associated with SERCA2a downregulation.
As the protein level of SERCA2a was downregulated in the absence of p22, the possibility of SERCA2a ubiquitination and degradation by the proteasome was investigated. Flag-tagged WT SERCA2a was immunoprecipitated from neonatal rat cardiomyocytes transduced with adenoviruses harboring LacZ or p22 short hairpin RNA and administered epoxomicin, an irreversible selective proteasome inhibitor. SERCA2a was more highly ubiquitinated (K48-linked ubiquitination) in the presence of p22 downregulation (Supplementary Fig. 3a). Cycloheximide chase assays were conducted in neonatal rat cardiomyocytes expressing either Flag-tagged wild-type SERCA2a or C498S SERCA2a in the presence of either LacZ or p22 shRNA and epoxomicin. Although WT SERCA2a was susceptible to proteasomal degradation in the absence of p22, C498S SERCA2a was more stable than WT SERCA2a even in the absence of p22 (Extended Data Fig. 4a-d). Moreover, siRNA-mediated knockdown of proteasome subunit alpha type-3 (PSMA3), a subunit of the 20S proteasome core and a part of the 26S proteasome, rescued the SERCA2a levels downregulated in the presence of p22 knockdown (Extended Data Fig. 4e,f), suggesting that SERCA2a undergoes proteasomal degradation in the absence of p22.
We next examined how p22-SERCA2a interaction influences SERCA2a stability. We expressed a small fragment of SERCA2a encompassing the p22-SERCA2a interaction domain as a minigene in cardiomyocytes to block endogenous p22-SERCA2a interaction in vitro. Adenoviruses harboring SERCA2a (334-666) from rat SERCA2a with an N-terminal HA tag was transduced into rat neonatal cardiomyocytes, and the level of SERCA2a was determined by western blotting. Expression of the SERCA2a (334-666)-HA protein resulted in downregulation of endogenous SERCA2a levels in a dose-dependent manner (Extended Data Fig. 5a,b). As observed earlier in this study, immunoprecipitation of the SERCA2a (334-666)-HA protein co-immunoprecipitated endogenous p22 (Extended Data Fig. 5c), and, as expected, the interaction between endogenous full-length SERCA2a and p22 was diminished in the presence of SERCA2a (334-666)-HA (Extended Data Fig. 5d). Proteasome inhibition with MG132 partially rescued the lowered SERCA2a protein level in the presence of SERCA2a (334-666)-HA (Extended Data Fig. 5). To evaluate how the presence of the minigene affects the oxidation status of SERCA2a, BIAM pull-down assays were conducted with cardiomyocytes transduced with either Ad-LacZ or Ad-SERCA2a (334-666)-HA in the presence of MG132. Cardiomyocytes treated with HO served as positive control. BIAM-SERCA2a pull-down was lower in the presence of SERCA2a (334-666)-HA than in the presence of LacZ (Extended Data Fig. 5g,h). Furthermore, WT mice injected with AAV9-cTnT-mSERCA2a (333-666)-Flag (minigene) showed lower endogenous SERCA2a levels than those injected with AAV9-cTnT-eGFP (Extended Data Fig. 5i,j). These data suggest that SERCA2a (334-666) binds and sequesters endogenous p22, thereby leading to increased oxidation and degradation of endogenous SERCA2a protein.
In the absence of p22, SERCA2a undergoes ubiquitination and proteasomal degradation. Therefore, the E3 ubiquitin ligase targeting SERCA2a under oxidative stress was investigated. Smad ubiquitination regulatory factor 1 (Smurf1), a member of the HECT family of E3 ligases, is involved in activin type II receptor-induced degradation of SERCA2a during aging and heart failure. Co-immunoprecipitation with SERCA2a antibody showed that SERCA2a interacts with Smurf1 at baseline, and this interaction is enhanced under HO-induced oxidative stress (Supplementary Fig. 4a). Interestingly, immunoprecipitation of HA-tagged ubiquitin also showed enhanced Smurf1 association with SERCA2a under oxidative stress (Supplementary Fig. 4a). The results also show that Smurf1 is more highly ubiquitinated under oxidative stress (Supplementary Fig. 4a), probably owing to autoubiquitination, which is known to take place in HECT type E3 ubiquitin ligases. Knockdown of Smurf1 using siRNA partially rescued the HO-mediated downregulation of SERCA2a (Supplementary Fig. 4b,c).
SERCA2 undergoes ER-associated degradation (ERAD). Therefore, the involvement of HMG-CoA reductase degradation protein 1 (Hrd1), an ERAD-specific RING family E3 ubiquitin ligase, in the degradation of SERCA2a was also investigated in this study. Co-immunoprecipitation experiments showed that SERCA2a interacts with Hrd1 and that this interaction is also enhanced in the presence of HO (Supplementary Fig. 4d). Knockdown of Hrd1 alleviated HO-induced downregulation of SERCA2a (Supplementary Fig. 4e,f).
Co-immunoprecipitation assays revealed that SERCA2a interacts with both Smurf1 and Hrd1, with these interactions markedly enhanced in p22 knockdown cells treated with epoxomicin (Fig. 7a). Knockdown of Smurf1 or Hrd1 independently led to a partial rescue of SERCA2a levels in p22-deficient cardiomyocytes (Fig. 7b-e). Immunofluorescence showed increased colocalization of SERCA2a and Hrd1 under p22 knockdown conditions (Fig. 7f,g). In addition, colocalization of SERCA2a, Hrd1 and the 20S proteasome subunit PSMA3 was elevated in the absence of p22 (Fig. 7h,i). These findings indicate that Smurf1 and Hrd1 mediate proteasomal degradation of SERCA2a in the absence of p22.
To evaluate the role of SERCA2a Cys498 oxidation in vivo, we generated SERCA2a-C498S knock-in (KI) mice (Supplementary Fig. 5). Calcium transients in cardiomyocytes from WT and SERCA2a-C498S KI mice (heterozygous and homozygous) revealed no significant differences in amplitude (F/F), SR calcium content or tau, indicating preserved calcium handling at baseline (Supplementary Fig. 6). Echocardiography confirmed that SERCA2a-C498S KI (homozygous) mice show normal baseline cardiac function and intact cardiac reserve in response to dobutamine (Supplementary Fig. 7a). Following 3 weeks of TAC or sham surgery, both heterozygous and homozygous KI mice showed preserved ejection fraction (Supplementary Fig. 7b,c), reduced hypertrophy (LVW/TL) (Supplementary Fig. 7d), less lung congestion (lung tissue weight (LungW)/TL) (Supplementary Fig. 7e), smaller cardiomyocyte size (Supplementary Fig. 7f,g) and less interstitial fibrosis (Supplementary Fig. 7h,i) compared with WT mice. TAC reduced RyR2, NCX1 and SERCA2a mRNA levels across all groups (Supplementary Fig. 7j-l,o), but SERCA2a-C498S KI mice tended to show lower NCX1 and SERCA2a expression than WT mice (Supplementary Fig. 7k,o). TAC-induced increases in the P-PLN/total PLN (T-PLN) ratio were blunted in SERCA2a-C498S KI mice (Supplementary Fig. 7m,n). Notably, SERCA2a protein levels were preserved in both heterozygous and homozygous SERCA2a-C498S KI mice after TAC, suggesting increased protein stability (Supplementary Fig. 7p,q). Co-immunoprecipitation assays showed reduced interaction between SERCA2a and Smurf1/Hrd1 in SERCA2a-C498S KI mice (Supplementary Fig. 8a). In BIAM pull-down assays, TAC reduced BIAM labeling of SERCA2a in WT mice, consistent with oxidation of cysteine thiols, while labeling remained unchanged in SERCA2a-C498S KI mice, indicating protection from Cys498 oxidation (Supplementary Fig. 8b,c). Together, these findings show that oxidation of SERCA2a at Cys498 during pressure overload promotes its degradation, contributing to cardiac dysfunction, and that the C498S mutation confers protective effects by maintaining SERCA2a levels and cardiac function.
To assess the functional importance of SERCA2a Cys498 oxidation, we crossed SERCA2a-C498S KI mice with p22 cKO mice, followed by 3 weeks of TAC or sham surgery. Consistent with the aforementioned findings, the cross with SERCA2a-C498S KI mice rescued cardiac dysfunction and preserved SERCA2a levels in p22 cKO mice during pressure overload (Fig. 8 and Supplementary Fig. 9). The cross with heterozygous SERCA2a-C498S KI mice restored LVEF, reduced the LVW/TL and lung W/TL ratios (Fig. 8a-d) and mitigated fibrosis and cardiomyocyte hypertrophy (Supplementary Fig. 9a-d) in p22 cKO mice. SERCA2a protein levels were significantly higher in p22 cKO plus SERCA2a-C498S KI mice compared with p22 cKO alone mice in both sham and TAC conditions (Fig. 8e,f). Importantly, the interaction of SERCA2a with Smurf1 and Hrd1, elevated in p22 cKO mice, was suppressed in the presence of the C498S KI mutation (Supplementary Fig. 9e). Cardiomyocyte contractility was not significantly altered by SERCA2a-C498S KI compared with the control. However, the reduced contractility in p22 cKO mice was fully restored by their cross with SERCA2a-C498S KI mice both at baseline and during isoproterenol stimulation (Fig. 8g,h). Similarly, calcium transient defects (reduced amplitude and prolonged tau), seen in p22 cKO cardiomyocytes, were normalized in the presence of SERCA2a-C498S KI, achieving values comparable to those of the wild type (Extended Data Fig. 6a-c). Despite these functional improvements, SERCA2a ATPase activity remained unchanged in WT, p22 cKO, SERCA2a-C498S KI and double-mutant groups (Supplementary Fig. 9f), suggesting that the protein level, not enzymatic function, is the critical determinant of dysfunction in p22-deficient hearts. Collectively, these findings identify Cys498 oxidation as a key regulatory site mediating SERCA2a degradation and cardiac dysfunction in the absence of p22.
The molecules associated with the degradation of SERCA2a identified in this study were evaluated in human LV samples from donor (healthy) and recipient (dilated cardiomyopathy (DCM) or ischemic cardiomyopathy (ICM)) samples. Immunostaining of human heart samples showed that the level of SERCA2a in the myocardium was significantly lower in failing hearts than in non-failing hearts (Extended Data Fig. 7a,b). Immunoblotting also showed that SERCA2a protein levels were downregulated in cardiomyopathic hearts compared with healthy hearts (Extended Data Fig. 7c,d). In addition, the levels of Smurf1 and Hrd1 were upregulated in cardiomyopathic hearts compared with control hearts (Extended Data Fig. 7e,f).