A dynamic pathway for calcium-independent activation of CaMKII by methionine oxidation

Oxidation directly activates CaMKII

CaMKII is activated by Ca²⁺/CaM, but autophosphorylation at T287 sustains catalytic activity after dissociation of Ca²⁺/CaM (Fig. 1A) because the negatively charged phosphate prevents reassociation of the catalytic domain and autoinhibitory region (Hudmon and Schulman, 2002). CaMKII activity may also be enhanced by pro-oxidant conditions (Zhu et al., 2007); we therefore hypothesized that oxidation of the regulatory domain in the vicinity of T287 could sustain CaMKII catalytic activity by an analogous mechanism. Exposure of purified CaMKII to H₂O₂ in the absence of any pre-treatment yielded no discernable CaMKII activity (Fig. 1B). However, exposure to H₂O₂ after pretreatment with Ca²⁺/CaM yielded persistent CaMKII activation even in the presence of EGTA. These data suggest that Ca²⁺/CaM binding exposed a key segment of CaMKII for oxidation, and that oxidation interfered with the interaction of the autoinhibitory and catalytic domains. Activation of wild type (WT) CaMKII by H₂O₂ was dose-dependent (Fig. 1C). The concentration of EGTA used was sufficient to block CaMKII activity without the addition of H₂O₂ (Fig. 1B), suggesting that activity observed in the pro-oxidant condition was independent of sustained Ca²⁺/CaM binding.

Pretreatment with Ca²⁺/CaM was also necessary for autophosphorylation-dependent CaMKII activation, indicating that autophosphorylation and oxidation of CaMKII occur by parallel mechanisms. CaMKII bearing a T287A substitution underwent normal Ca²⁺/CaM-dependent activation but did not maintain persistent Ca²⁺-independent activity in the presence of ATP (Fig. 1D). However, the T287A mutant was activated by H₂O₂ (Fig. 1C), and the extent of this activation was statistically indistinguishable at all but the highest concentration of H₂O₂ tested (1μM). We interpret these observations as evidence that activation of CaMKII by ROS and autophosphorylation occur by a similar mechanism, but by independent modifications to nearby sites. Activation of the kinase by either mechanism requires the enzyme to be initially ‘opened’ by Ca²⁺/CaM to allow access to the autoinhibitory domain for oxidation or autophosphorylation (Fig. 1A, E). Either of these modifications can prevent subsequent interaction of the autoinhibitory region with the catalytic domain, providing for sustained Ca²⁺-independent activation of CaMKII. Consistent with these ideas, direct measurements of intrinsic fluorescence revealed that autophosphorylation and oxidation of CaMKII independently induce similar conformational changes in CaMKII (Supplemental Fig. 1).

Proteomic analysis of the synthetic peptide that contains the 281/282 methionine residues was used to probe for oxidative modification upon treatment with H₂O₂. We observed a clear decrease in the unoxidized form coupled with an increase in the various oxidized forms of this peptide based on the chromatographic traces and on the change in the number of observed spectra (Supplemental Fig. 2). In addition to the synthetic peptide, we analyzed the peptide containing the 281/282 methionine residues after treatment of the whole protein with H₂O₂ followed by trypsin cleavage. We were able to determine the relative change in oxidation of this peptide upon hydrogen peroxide treatment (Supplemental Table 1,2). The MS/MS spectra of the oxidized forms of the peptide were identical to those from the synthetic peptide, verifying that the oxidized peptide was correctly identified.

Given these observations and the recognized susceptibility of methionine residues to oxidation (Hoshi and Heinemann, 2001), we made methionine to valine mutations for the paired residues (M281/282V) and for another methionine (M308V) in the CaM-binding region. These mutants were exposed to H₂O₂ and assayed for activity in the presence of EGTA (Fig. 1C). The H₂O₂-dependent activation of CaMKII was preserved in the M308V mutant. However, oxidation-dependent CaMKII activity was completely abolished in the M281/282V and M281/282/308V mutants. Our data, obtained in cell free assay conditions, point to direct oxidation of the M281/282 pair as the primary H₂O₂-dependent activation pathway for CaMKII. Importantly, all the methionine to valine mutants showed a normal activity response to autophosphorylation (Fig. 1D), further supporting the concept that Ca²⁺ autonomous CaMKII activation by ROS or T287 autophosphorylation are independent events. While the paired methionine motif is conserved in the β, γ, and δ isoforms of CaMKII, the neuronal α isoform substitutes a cysteine residue for the first methionine of the pair (position 280 in CaMKIIα). The side chain of cysteine is also susceptible to oxidation. We generated a M281C mutant of CaMKIIδ to mimic the substitution in CaMKIIα. Additionally, we generated and purified CaMKIIα. Both the M281C CaMKIIδ mutant and the purified CaMKIIα were activated by H₂O₂, indicating that the cysteine substitution seen in CaMKIIα also supports ROS-dependent activation (Fig. 1C). To further elucidate the role of M281 and M282 in ROS-dependent activation, these sites were individually mutated (Supplemental Fig. 3). The M282V mutation completely ablated oxidation-dependent activation, while the M281V mutation partially reduced activation by 65%, indicating that a single oxidation event within the regulatory domain is insufficient to activate CaMKII.

Autophosphorylation at T287 dramatically increases the binding affinity of CaMKII for CaM, a phenomenon known as “CaM trapping” (Meyer et al., 1992). In the absence of ATP the Ca²⁺/CaM/CaMKII complex was very rapidly dissociated following addition of EGTA, independent of the redox state, as measured by fluorescence anisotropy of dansylated CaM (Supplemental Fig. 4A). CaMKII exposure to H₂O₂ for 10 min induced Ca²⁺/CaM-independent activity (as in Fig 1B), but also failed to induce CaM trapping (not shown). These observations indicate that under normal experimental conditions, oxidation of CaMKII is not sufficient to induce CaM trapping. Dissociation of Ca²⁺/CaM from autophosporylated CaMKII and CaM was significantly slower than from nonphosphorylated enzyme, consistent with CaM trapping. However, pretreatment with H₂O₂ prior to EGTA had no significant effect on the dissociation kinetics. Thus, oxidation of CaM or CaMKII does not prevent or enhance CaM trapping by autophosphorylated CaMKII. CaM trapping is reduced by phosphorylation of T306/307 (Colbran 1993), so we investigated if oxidation of M308 might prevent CaM trapping by a parallel mechanism. We did observe a significant slowing of dissociation of the CaM/CaMKII complex after H₂O₂ treatment of the M308 mutant (Supplemental Fig. 4B,C). These data suggest that the absence of CaM trapping during oxidation is partly due to M308.

It seemed possible that conditions capable of oxidizing methionine residues would also oxidize unprotected cysteine residues. Although mutation of methionine residues at 281 and 282 was sufficient to completely ablate ROS-dependent activation of CaMKII, we created a C290V mutant to determine whether this cysteine residue within the CaMKII regulatory domain could also play a role. Both the Ca²⁺/CaM-dependent and ROS-dependent activity of the C290V mutant were indistinguishable from WT CaMKII (Supplemental Fig. 4D). Our finding that oxidation of paired amino acids (M281/282 in CaMKIIδ) were required for activation by H₂O₂ support a view that oxidation of a lone residue is insufficient to confer Ca²⁺/CaM autonomous CaMKII activity. In order to comprehensively test the potential role of all accessible cysteines in contributing to oxidation-dependent CaMKII activity we measured CaMKII activity responses to H₂O₂ in the presence of iodoacetic acid, a reagent that blocks oxidation of unprotected cysteine residues (Zangerle et al., 1992). Cysteine protected CaMKIIδ showed equivalent H₂O₂ activity responses compared to CaMKIIδ without iodoacetic acid (Supplemental Fig. 2D). We used an established colorometric assay to quantify the available cysteine residues and verify that cysteine protection by iodoacetic acid was effective. Our results confirmed that most or all of the 11 cysteines in CaMKIIδ were accessible to the Ellman’s reagent after Ca²⁺/CaM binding, while treatment with iodoacetic acid blocked the accessibility of cysteine residues to biochemical modification (data not shown). Taken together, these findings demonstrate that oxidative activation of CaMKIIδ is independent of cysteines.

Oxidation of CaMKII occurs in vivo

We developed a new immune serum against oxidized M281/282 to detect ROS effects on CaMKII in vivo. We validated the fidelity of the antiserum using purified CaMKII protein by immunoblotting against WT CaMKII and the M281/282V mutant in control conditions and after treatment with H₂O₂ or Ca²⁺/CaM/ATP. Blots were also assayed with a phospho- and site-specific antibody against T287 (p-287). WT CaMKII exposed to H₂O₂ after pretreatment with Ca²⁺/CaM showed significant reactivity to our oxidized M281/282 antiserum, but untreated and T287-phosphorylated CaMKII samples were not recognized by our antiserum (Fig. 2A). The M281/282V mutant had minimal reactivity to our antiserum among the three treatments. These findings demonstrated that phospho-T287 and oxidized M281/282 were immunologically distinct sites. We performed additional immunoblots in which oxidized CaMKII was probed with the antiserum along with increasing concentrations of the peptide antigen (Fig. 2B). Band intensity decreased with increasing peptide concentration, indicating that the immune serum was specific for oxidized CaMKII.

To determine the role of CaMKII oxidation in apoptosis, mice were treated with saline, AngII, or isoproterenol (Iso) for one week, and transverse heart sections from these mice were probed for the production of oxidized CaMKII in vivo. WT mice treated with AngII produced more oxidized CaMKII than those treated with saline or Iso (Fig. 2C). Total CaMKII immunoreactivity remained constant regardless of treatment. Conversely, mice lacking a critical subunit of NADPH oxidase (p47^−/−) did not show increased levels of oxidized CaMKII in response to AngII. The p47^−/− mice do not assemble the ROS-producing complex NADPH oxidase (Munzel and Keaney, 2001), the main source of ROS due to AngII stimulation in many cell types (Lyle and Griendling, 2006). Heart sections from WT mice showed increased staining for T287-phosphorylated CaMKII after AngII treatment, while p47^−/− mice were unaffected (Supplemental Fig 5). Other studies have suggested that protein phosphatase activity is decreased by pro-oxidant conditions (Howe et al., 2004), indicating the possibility of coordinate activation of CaMKII both by direct oxidation at the Met281/282 sites and by phosphatase inactivation leading to increased phosphorylation at the T287 site.

We also homogenized hearts from mice treated with saline, AngII, or Iso, and whole heart lysates were analyzed by immunoblot for oxidized CaMKII. While total CaMKII was not significantly different among the three treatment groups, heart lysates from mice treated with AngII showed significantly increased oxidized CaMKII levels (Fig. 2D). Taken together, these findings demonstrate that oxidation of CaMKII occurs in vivo, and that elevated levels of AngII increase CaMKII oxidation at M281/282 compared to saline or Iso.

AngII triggers ROS production and CaMKII-dependent apoptosis in cardiomyocytes

Given our results, we hypothesized that cells deficient in ROS production or CaMKII activity would be resistant to AngII mediated apoptosis. We treated cardiomyocytes from mice that express an inhibitory peptide against CaMKII (AC3-I, Zhang et al., 2005) with 100nM AngII for 24 hours in parallel with isolated cardiomyocytes from WT and p47^−/− mice. AngII caused a significant increase in the percent of TUNEL positive nuclei in WT cells but had no significant effect in p47^−/− or AC3-I cardiomyocytes (Fig. 3A). Activity assays for caspase-3, a downstream target enzyme in the CaMKII apoptotic signaling pathway in heart, recapitulated the results from the TUNEL assay (Fig. 3B). Importantly, direct addition of ROS in the form of H₂O₂ caused significant apoptosis in p47−/− cells, demonstrating that their resistance to AngII-induced apoptosis is a result of impaired ability to produce ROS rather than a lack of sensitivity to ROS. The apoptotic effect of H₂O₂ in AC3-I cells was blunted by more than half compared to WT or p47^−/− cells (Fig. 3A), indicating the critical importance of CaMKII activation to ROS and Iso-dependent apoptosis.

In order to validate the connection between AngII and ROS in our experimental model, we treated isolated cardiomyocytes from WT and p47^−/− mice with 100nM AngII and monitored production of ROS by imaging DHE, a fluorescent reporter for superoxide and hydrogen peroxide (Fig. 3C). We also incubated WT cardiomyocytes with fura-2 AM, a cell-permeant calcium indicator to observe changes in intracellular Ca²⁺ ([Ca²⁺]_i). Treatment with AngII caused a significant increase in ROS production in WT but not in p47^−/− cardiomyocytes (Fig. 3D). On the other hand, increases in [Ca²⁺]_i were significantly less after AngII compared to Iso treatment (Fig. 3E, F) for cardiomyocytes from both WT and p47^−/− mice. These data show AngII signaling predominantly increases ROS, while Iso predominantly increases [Ca²⁺]_i under our experimental conditions.

CaMKII knockdown prevents AngII- and Iso-induced apoptosis

In order to further test the role of CaMKII and specifically define the effects of M281/282 on myocardial apoptosis, we used a knock down and replacement strategy in cultured neonatal cardiomyocytes. Rat cardiomyocytes were cultured and treated with shRNA encoding lentivirus against rat CaMKIIδ. After 48 hours CaMKII expression was significantly reduced, as measured by immunoblot and activity assays (Fig. 4A). Cells were then transduced with lentivirus encoding shRNA-resistant WT or M281/282V mutant CaMKII. Control cells were transduced with GFP encoding lentivirus. After 48 hours, cells transduced with CaMKII rescue constructs showed significant recovery of CaMKII expression compared to control cells. Addition of Ca²⁺/CaM to lysates from cells transduced with either CaMKII encoding virus had similar total activity as native cells. However, H₂O₂-induced activity was only rescued in cells expressing the WT CaMKII construct. These cellular studies support our earlier finding with molecular CaMKII (Fig. 1) by showing that oxidation of M281/282 is critical for ROS triggered CaMKII activity. In addition, this strategy created cardiomyocytes expressing ROS-resistant CaMKII, providing a novel system for investigating ROS and CaMKII-dependent apoptosis.

Cardiomyocytes treated with shRNAs/CaMKIIδ encoding lentivirus were exposed to saline, AngII or Iso as above (Fig. 4B). The apoptotic response to AngII and Iso was significantly attenuated in CaMKII knockdown cells compared to myocytes without shRNA. Moreover, expression of shRNA-resistant WT CaMKII fully rescued apoptotic responses to both agonists (Fig. 4C). In contrast, expression of the ROS-resistant M281/282V CaMKII mutant rescued the apoptotic response to Iso but, importantly, failed to rescue the apoptotic response to AngII after 24 hours (Fig. 4C). Cells expressing the M281/282V CaMKII remain susceptible to Iso-induced apoptosis (Fig. 4B, C), indicating that elimination of these residues does not affect activation of CaMKII by catecholamine stimulation. These cellular studies are performed in a complex biological environment compared to studies with isolated CaMKII, but nevertheless support a concept that direct oxidation of CaMKII by AngII is sufficient to confer enhanced CaMKII activity and trigger apoptosis.

ROS production and CaMKII activity are critical for AngII-mediated cardiac apoptosis in vivo

Intracellular ROS levels increase dramatically in models of structural heart disease (Hare, 2001), particularly those initiated by AngII (Tojo et al., 2002). Stimulation by AngII leads to activation of the NADPH oxidase complex, increasing intracellular superoxide and hydrogen peroxide levels. To establish an in vivo context for our previous findings and to test the role of CaMKII in AngII-stimulated cardiac apoptosis, p47^−/−, AC3-I, and WT mice were treated with saline, AngII or Iso for one week. Transverse heart sections from these mice were stained for evidence of apoptosis. After one week WT mice treated with either AngII or Iso showed significant cardiac apoptosis, as determined by TUNEL staining of heart sections (Fig. 5). The p47^−/− mice had no significant increase in cardiac apoptosis after treatment with AngII, most likely because these mice were unable to produce ROS in response to AngII stimulation (Fig. 3D). However, the p47^−/− mice showed a preserved apoptotic response to Iso, suggesting that Iso-induced apoptosis occurs independently of oxidative stress generated by NADPH oxidase in vivo under these conditions. In contrast, the AC3-I mice with CaMKII inhibition were resistant to apoptosis induced by either AngII or Iso, indicating that CaMKII is a necessary signal element for apoptosis initiated by AngII or Iso in vivo.

Increased CaMKII oxidation, apoptosis, cardiac dysfunction and death in MsrA^−/− mice

Methionine oxidation is specifically reversed by MsrA (Weissbach et al., 2002), so we hypothesized that MsrA^−/− mice would show enhanced vulnerability to AngII-mediated CaMKII oxidation and apoptosis. In order to test this idea we implanted MsrA^−/− and WT control mice with AngII or saline eluting osmotic mini pumps. Hearts from MsrA^−/− mice treated with AngII in vivo showed significantly more CaMKII oxidation (Fig. 6A,B) and increased TUNEL staining (Fig. 6C) compared to saline treated MsrA^−/− mice and to saline or AngII treated control hearts. The increased CaMKII oxidation by AngII in MsrA^−/− hearts showed that CaMKII oxidation is dynamically regulated by MsrA in myocardium in vivo, and suggested that MsrA^−/− mice would be more vulnerable to severe myocardial stress due to increased methionine oxidation. Myocardial infarction is the most common cause of sudden cardiac death and heart failure in patients, and p47^−/− (Doerries et al., 2007) and AC3-I mice (Zhang et al., 2005) are protected from left ventricular dilation and dysfunction after myocardial infarction surgery, suggesting that ROS activation of CaMKII may be important in myocardial infarction. In order to test if CaMKII oxidation and apoptosis were regulated by NADPH oxidase and MsrA in the setting of myocardial infarction, we subjected MsrA^−/−, p47^−/−, and WT mice to myocardial infarction surgery. MsrA^−/− mice showed significantly more CaMKII oxidation (Fig. 7A, B) and myocardial apoptosis (Fig. 7C) compared to WT and p47^−/− mice. These data indicate CaMKII oxidation is dynamically regulated by NADPH oxidase and MsrA in the setting of myocardial infarction. We performed myocardial infarction surgery on a larger cohort of MsrA^−/− and WT control mice to determine if increased CaMKII oxidation and apoptosis in MsrA^−/− mice translated into poorer functional outcomes. MsrA^−/− mice were significantly more likely to die after surgery compared to WT controls (Fig. 7D, p = 0.0015) and exhibited significantly greater left ventricular dilation (Fig. 7E) and impaired systolic function (Fig. 7F) compared to controls, demonstrating that methionine oxidation increases the pathological impact of myocardial infarction in vivo.

Mouse Models

Mice lacking the p47 gene (p47^−/−) were purchased from Jackson Labs. Mice lacking the MsrA (MsrA^−/−) were supplied by NIH (Bethesda, MD). Mice with genetic CaMKII inhibition (AC3-I) were generated by us as previously described (Zhang et al., 2005).

CaMKII activity assays and protein analysis

Mutant CaMKII cDNAs were generated using a QuikChange site-directed mutagenesis kit (Stratagene). CaMKIIδ (GenBank #NP_001020609) was generated using the Bac-to-Bac baculovirus system (Invitrogen) and purified on a calmodulin-agarose column. For CaMKII activity assays, purified CaMKII was pretreated with 200μM CaCl₂ and 1μM CaM on ice for 1 minute. The protein was then exposed to ATP, H₂O₂, or water at the described concentrations for 10 minutes. Samples exposed to ATP or H₂O₂ were then treated with 10mM EGTA for 10 minutes. CaMKII activity was measured as a function of ³²P-ATP incorporation into a synthetic substrate (syntide-2) at 30°C, as previously described (Wu et al., 2002).

Oxidized M281/282 immune serum production and immune staining

Antiserum to oxidized CaMKII regulatory domain peptide was generated by Epitomics, Inc. Commercial antibodies were used for blots and immunostaining for total (Stressgen Biotechnologies) and phosphorylated (Santa Cruz) CaMKII.

Detection of ROS

Changes in ROS levels in cultured primary cardiac myocytes after agonist stimulation were measured using the fluorgenic probe dihydroethidium (DHE, 5μM, Molecular Probes), as previously described (Zimmerman et al., 2004). DHE fluorescent images were acquired using confocal microscopy (Zeiss LSM510).

Intracellular Calcium Concentration Measurements

Intracellular calcium concentration was assessed by Fura-2 fluorescence ratio imaging using a microscopic digital imaging system (Photon Technology International), as described previously (Sharma et al., 1995).

Fluorescence measurements

Spectra were collected at 30°C using a Fluorolog 3 (Jobin Yvon, Horiba) spectrofluorometer. For intrinsic fluorescence shift experiments, excitation wavelength was 270nm. Emission spectra were generated at 1nm increments from 280nm to 400nm. Background traces were subtracted from CaMKII spectra to eliminate the contribution from intrinsic fluorescence of CaM. For fluorescence anisotropy experiments, baseline traces of 100nM dansylated CaM in 15mM HEPES buffer, pH 7.2 were measured at baseline and after the addition of 200μM CaCl₂ at 60s. At 180s, 100nM purified CaMKII was added to the CaM solution. For some trials, CaMKII became phosphorylated by the addition of 10mM ATP. 100μM H₂O₂ or an equivalent volume of buffer was added at 250s. Finally, addition of 10mM EGTA at 300s was used to remove free calcium from the solution, uncoupling CaM/CaMKII binding.

Cardiomyocyte TUNEL immunostaining

Myocyte isolations from neonatal mouse or rat pups were modified from previously described methods (Mohler et al., 2007). To ensure that pure populations of cardiomyocytes were obtained, cultures were immunolabeled with alpha-actinin Ig (cardiomyocyte-specific marker). Only cultures with >90% cardiomyocytes were used in experiments (see Supplemental Fig. 6). Lentiviral treatments (shRNA, rescue constructs) and apoptosis inducing agents (Iso, AngII) were used for 24 hours. Cells were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 and sodium citrate, and stained using In Situ cell death detection kits, TMR Red (Roche). TUNEL stain assays were interpreted as previously described (Yang et al., 2006). Nuclei were co-stained with DAPI. Investigators were blinded to the genetic identity and treatment of the mice in all studies.

CaMKII cDNA, shRNA, and virus generation

See supplemental text for sequences and methods.

Heart section TUNEL immunostaining

Minipumps containing saline or AngII (3mg/kg/day) were inserted or daily injections of Iso (30mg/kg/day, intraperitoneal) were given for 7 days. Animals were then sacrificed and the heart was removed, fixed, and embedded in paraffin. To confirm specific labeling of cardiomyocytes, heart sections were labeled with alpha-actinin Ig (cardiomyocyte-specific marker). Only sections with >90% cardiomyocytes were used in experiments (see Supplemental Fig. 7). TUNEL staining was performed using In Situ cell death detection kits, TMR Red (Roche). Nuclei were co-stained with DAPI. An investigator blinded to the identity and treatment of the mice counted the number of total and TUNEL positive cells in each image.

Other sections were treated with either a general CaMKII antibody or oxidized CaMKII antiserum. Nuclei were co-stained with DAPI. Identity of cardiomyocytes was confirmed by co-stain with an antibody against α-actinin. Images were quantified for relative staining intensity using Image J (NIH, USA). Both the investigators and technical personnel assigned to immunostaining and quantification were blinded to the genetic identity and treatment of the mice in all studies.

Myocardial Infarction and echocardiography

Mice were anesthetized with ketamine/xylazine (87.5/12.5 mg/kg, respectively), and the left anterior descending (LAD) branch of the coronary artery was ligated using 8-0 ethilon suture (Ethicon) along the anterolateral border of the heart as close to the left atrial appendage as possible. Successful ligation of the artery is confirmed by blanching of the myocardium. Transthoracic echocardiograms were recorded in conscious sedated mice as described previously (Weiss et al., 2006), using a 15 MHz probe connected to a Sonos 5500 imager (Phillips Medical Systems, Bothell, WA). Images were acquired by an operator blinded to mouse genotype.

Statistical analysis

Statistical significance for mortality study was determined by chi-squares test. All other statistical significance was determined by One-Way ANOVA with post hoc Bonferonni tests. A p value of <0.05 was considered statistically significant. All results are presented as mean ± SEM.