CaMKII specifically induces cytosolic accumulation of HDAC4.
To determine whether different class IIa HDACs might display specific responses to upstream kinases, we began by expressing the 4 class IIa HDACs — 4, 5, 7, and MEF2-interacting transcription repressor (MITR; a splice variant of HDAC9) — in COS cells with kinases implicated in HDAC phosphorylation, including PKD and different isoforms of CaMK. HDAC4, -5 and -7 were exported from the nucleus to the cytoplasm, and MITR, which lacks a nuclear export signal, was redistributed in the nucleus from a punctate to a homogeneous pattern in response to PKD, CaMKI, and CaMKIV (Figure , A and B, and data not shown). Remarkably, however, constitutively active forms of CaMKII carrying point mutations (T287D) that mimic autoactivation caused the complete translocation of HDAC4 from the nucleus to punctate dots in the cytoplasm, but the other class IIa HDACs did not change their predominant nuclear localization in response to activated CaMKIIδB and -δC, -γ, -β, or -α (Figure , A and B, and data not shown). However, HDAC7 appeared to display partial responsiveness to CaMKII.
Selective response of HDAC4 to CaMKII.
To confirm that cytosolic accumulation of HDAC4 reflected its phosphorylation, we assayed for the association of HDAC4 with 14-3-3, which correlates with phosphorylation of class IIa HDACs (3
). FLAG-HDAC4 immunoprecipitates were analyzed by Western blot analysis using an antibody against endogenous 14-3-3 protein. Consistent with our immunocytochemical results, we observed that CaMKIIδB induced 14-3-3 binding to HDAC4 but not HDAC5 whereas PKD1 and CaMKI acted on both class IIa HDACs, suggesting that the latter enzymes act as general class IIa HDAC kinases (Figure C).
HDAC4 colocalizes with activated CaMKII.
The CaMKIIδ splicing variant B contains a nuclear localization signal (NLS) and thus likely targets nuclear proteins (23
). Unexpectedly, activated CaMKIIδB, which was generated by a T287D mutation (CaMKIIδB-T287D), localized predominantly to the cytosol, in contrast to the WT (inactive) form, which was mainly localized to the nucleus (Figure A). Phosphorylation of S332 next to the NLS (328
) of CaMKIIδB has been shown to induce nucleocytoplasmic shuttling of CaMKIIδB (23
). Consistent with these findings, the additional S332A point mutation (CaMKIIδB-T287D/S332A) caused CaMKIIδB to remain in the nucleus (Figure A).
Regulation of CaMKII subcellular localization.
The observation that CaMKIIδB-T287D is predominantly cytosolic while HDAC5 is exclusively nuclear in unstimulated cells raised the possibility that the insensitivity of HDAC5 to CaMKII might reflect its sequestration in a different subcellular compartment than CaMKII. We therefore coexpressed the constitutively active and nuclear CaMKIIδB-T287D/S332A mutant with HDAC4 and HDAC5. Like CaMKII-T287D, this mutant enzyme also induced cytosolic accumulation of HDAC4 but not HDAC5 (Figure B). Although CaMKIIδB-T287D/S332A cannot be phosphorylated next to its NLS and is clearly localized to the nucleus when expressed in the absence of HDAC4, it colocalized with HDAC4 to the cytosol (see Figure B), suggesting a possible physical interaction between HDAC4 and CaMKII that results in coshuttling of the 2 proteins to the cytosol.
CaMKII induces nuclear export and blocks nuclear import of HDAC4.
The above findings raised the question of how the predominantly cytosolic CaMKIIδB-T287 mutant causes accumulation of HDAC4 to the cytosol. The same phenomenon was observed for the CaMKIIδ splice variants A and C, CaMKIIαA and CaMKIIβ'e, which do not contain an NLS but which exerted the same effect on HDAC4 localization as CaMKIIδB (data not shown). In addition, a CaMKIIδB mutant in which the NLS was destroyed (T287D/K328,329N) (23
) and that localized exclusively to the cytosol, caused HDAC4 to colocalize and accumulate to the cytosol (Figure C).
Because HDAC4 was located in the cytosol in approximately 20% of cells under basal conditions (Figure B), we hypothesized that it might cycle between the nucleus and the cytosol and that cytosolic forms of CaMKII might block nuclear import of HDAC4. Therefore, we examined the effect of leptomycin B, an inhibitor of CRM1-dependent nuclear export, on the intracellular redistribution of HDAC4 in response to different active CaMKIIδB mutants (T287D, which is predominantly cytosolic; T287D/S332A, which is exclusively nuclear; and T278D/K328,329N, which is exclusively cytosolic) (Figure , C and D). All these mutants are constitutively active and all induced cytosolic accumulation of HDAC4 in the absence of leptomycin B. After short-term treatment with leptomycin B (4 hours, 1 nM), HDAC4 accumulated in the nucleus only after stimulation with nuclear CaMKIIδB-T287D/S332A, suggesting that this mutant induced nuclear export of HDAC4. In contrast, in the presence of cytosolic CaMKIIδB-T287D and -T287D/K328,329N, leptomycin B failed to induce nuclear accumulation of HDAC4, indicating that these mutants block nuclear import. Thus, even when CaMKII is in the cytosol, it can phosphorylate HDAC4, favoring derepression of HDAC target genes.
Analysis of HDAC4 phosphorylation sites by 14-3-3 binding in response to CaMKI and CaMKII.
To assess whether CaMKII uses different phosphorylation sites on HDAC4 than CaMKI, we performed additional coimmunoprecipitation experiments with endogenous 14-3-3 protein in COS cells. Consistent with previous work by others (5
), the replacement of S246, S467, and S632 with alanine (mutant S246,467,632A) abolished 14-3-3 binding at baseline as well as in the presence of constitutively active CaMKI or CaMKII (Figure A). Whenever S246 was mutated (alone or in combination with S467 or S632), baseline 14-3-3 binding and CaMKI-induced 14-3-3 binding were impaired (Figure , A and B). Conversely, the S467,632A mutant was still phosphorylated at baseline and sensitive to CaMKI, identifying S246 as the major phosphorylation site for the endogenous COS cell kinase activity and CaMKI. In contrast, whenever S467 was mutated, CaMKII-induced 14-3-3 binding was impaired, but the S246,632A mutant was still sensitive to CaMKII, identifying S467 as the major phosphorylation site for this kinase (Figure , A and B). CaMKI could also phosphorylate S467 to a modest degree, but its phosphorylation of S632 was almost undetectable. In contrast, CaMKII efficiently phosphorylated S632. Although 14-3-3 binding to the S467,632A mutant was readily detectable, there was no increase above baseline in response to CaMKII, indicating that S246 is not a CaMKII site.
Detection of 14-3-3 binding sites of HDAC4 in response to CaMKI and CaMKII.
These data were confirmed using a mammalian 2-hybrid assay that detects the specific association of 14-3-3 fused to VP16 (transcriptional activation domain) with the N terminal half of HDAC4 (amino acids 1–740) fused to GAL4 (DNA-binding domain) (Figure C). A similar system with HDAC5 was successfully used to detect HDAC kinases in a high-throughput expression screen (25
). Phosphorylation of HDAC4 by CaMKIIδB-T287D created docking sites for VP16–14-3-3, resulting in activation of the GAL4-dependent reporter (Figure C). In contrast, CaMKIIδB-T287D failed to activate the reporter in the presence of GAL4-HDAC5 and VP16–14-3-3, confirming the selective responsiveness of HDAC4 to CaMKII and establishing this system as a reliable method for detecting 14-3-3 binding to HDAC4. In this assay, we replaced each of the 3 signal-responsive serines (S246, S467, and S632) with alanines. Disruption of S246 did not affect the interaction of 14-3-3 with HDAC4 in response to CaMKIIδB-T287D (Figure D). Consistent with our 14-3-3 coimmunoprecipitation data, disruption of S467 or S632 reduced 14-3-3 binding. Taken together, the results of the direct and indirect 14-3-3–binding assays demonstrate that CaMKI preferentially phosphorylates S246 and S467, whereas CaMKII phosphorylates S467 and S632 on HDAC4 to induce 14-3-3 binding, as schematized in Figure E.
Mapping a CaMKII-responsive region of HDAC4.
To further examine the molecular basis for the selective responsiveness of HDAC4 to CaMKII, we generated mutant constructs encoding chimeric HDAC4/HDAC5 proteins. As shown in Figure A, only those chimeric proteins containing residues 529–657 of HDAC4 were responsive to CaMKIIδB-T287D. Because this region contains S632, we asked whether this phosphorylation site determines the selective responsiveness of HDAC4 to CaMKIIδB-T287D. The amino acid sequence surrounding S632 of HDAC4 differs at 4 positions from the corresponding region in HDAC5, which surrounds S661. We mutated 3 of these residues in addition to 2 differing amino acids in the consensus sequence around S498 of HDAC5 to those of HDAC4 (HDAC5 S494G/S499A/G657S/T659A/A665S). Despite these changes, this HDAC5 mutant was still nonresponsive to CaMKII (Figure B). These results suggest that differences in the CaMKII phosphorylation sites of HDAC5 were insufficient to account for its insensitivity to CaMKII.
Mapping the CaMKII-responsive region of HDAC4.
Activated CaMKII directly interacts with a unique domain of HDAC4.
Based on the observation that CaMKII-T287D colocalized with HDAC4, we performed immunoprecipitation experiments to determine if the proteins interacted. As shown in Figure C, we found that CaMKII-T287D bound strongly to HDAC4 but showed almost no detectable interaction with HDAC5. Whereas weak binding of CaMKII to HDAC5 was detectable under low stringency conditions in the immunoprecipitation buffer, it was not detectable under high stringency conditions. However, CaMKII still bound strongly to HDAC4 under high stringency conditions, suggesting that HDAC4 possesses a unique domain that mediates a strong interaction with the kinase. Interestingly, we observed that activated CaMKII-T287D but not inactive WT CaMKII strongly bound to HDAC4 (Figure D). This finding suggests that autophosphorylation induces a conformational change in CaMKII, which allows it to bind to HDAC4.
Coimmunoprecipitation experiments using deletion mutants of HDAC4 delineated the CaMKII-binding domain to amino acids 585–608 of HDAC4 (Figure , E and F). Although HDAC4 shares extensive amino acid homology with other class IIa HDACs throughout its length, this CaMKII-binding region is not homologous to other class IIa HDACs (Figure A).
R601 of HDAC4 is required for full responsiveness to CaMKIIδB-T287D.
To pinpoint the residue(s) required for interaction of HDAC4 with CaMKII, we systematically mutated the residues in the minimal CaMKII-binding domain to alanines and tested the mutants for their ability to bind CaMKII by coimmunoprecipitation. As shown in Figure , A and B, substitution of R601 by alanine or phenylalanine markedly disrupted the physical interaction between HDAC4 and CaMKII.
To test whether R601 of HDAC4 was required for CaMKII responsiveness, we examined the subcellular localization of 2 HDAC4 mutants (R601A and R601F) in the presence of CaMKIIδB-T287D. In contrast to WT HDAC4, these mutants only partially accumulated in the cytosol in response to CaMKIIδB-T287D (Figure , C and D). Moreover, these HDAC4 mutants did not colocalize with CaMKIIδB-T287D. In contrast, HDAC4-R601A and R601F mutants were still responsive to CaMKI (Figure D), indicating that R601 is specifically required for CaMKII sensitivity.
GAL4-HDAC4 mutants containing alanine, phenylalanine, lysine or leucine in place of R601 were markedly impaired in their ability to bind VP16–14-3-3 in response to CaMKIIδB-T287D (Figure E) whereas the basal 14-3-3–binding activity of these HDAC4 mutants was comparable to that of WT HDAC4 (not shown). While HDAC4-R601A and R601K showed a slight CaMKII-induced increase in 14-3-3 binding (below 2-fold), mutations of R601 to nonpolar hydrophobic amino acids (phenylalanine or leucine) prevented 14-3-3 binding in response to CaMKII completely. Again, these CaMKII nonresponsive mutants were not affected in their ability to bind 14-3-3 in response to CaMKI (not shown).
To further test whether phosphorylation of HDAC4 by CaMKII is dependent on a direct interaction between these 2 proteins, we purified a bacterially expressed HDAC4 protein fragment (amino acids 419–670), which contains the CaMKII phosphorylation sites S467 and S632 and the CaMKII docking site as a glutathione-S-transferase (GST) fusion protein (GST-HDAC4-WT). When incubated together with purified His-CaMKIIδ, Ca2+/calmodulin, and [γ-32P]-ATP, direct phosphorylation was observed (Figure F). In contrast, phosphorylation of GST-HDAC4-R601F was 7-fold weaker, indicating that full responsiveness of HDAC4 to CaMKII depends on stable docking of the kinase on HDAC4. In this assay, active WT CaMKII enzyme was used, and HDAC4 phosphorylation was abolished by Ca2+ depletion with EGTA, confirming that phosphorylation of HDAC4 by CaMKII is not artificially caused by the T287D point mutation but that it also occurs with endogenously activated CaMKII.
CaMKII binds directly to HDAC4 because the same GST-HDAC4 fusion protein used in the in vitro kinase assay pulled down about 30% of active His-CaMKIIδ that was coincubated in the same tube (Figure G). The GST-HDAC4-R601F mutant did not pull down His-CaMKIIδ. The same result was obtained with in vitro–translated [35S]-CaMKIIδB-T287D (not shown).
These findings raise the question of why CaMKIIδB-T287D still induces some cytosolic accumulation of HDAC4-R601F (to about 50%) (Figure D). We imagine 3 explanations for this phenomenon: (a) CaMKII activates an endogenous kinase in COS cells that induces cytosolic accumulation of HDAC4 independent of CaMKII docking; (b) there may be residual binding of CaMKII to the HDAC4-R601F mutant, which is indetectable in the assay but sufficient to induce weak phosphorylation of HDAC4; (c) stable binding may not be absolutely essential for HDAC4 phosphorylation by CaMKII if CaMKII can act inefficiently through a “kiss-and-run” type of mechanism. Evidence for the latter 2 possibilities is provided by the in vitro kinase assay (Figure F) in which CaMKII induced slight phosphorylation of HDAC4-R601F (7-fold less than to the WT fragment) although no direct binding was detected (Figure G).
Another question one could ask is whether other kinases, which act as general class IIa HDAC kinases, such as CaMKI, also require docking to their substrates. Indeed, coimmunoprecipitation of HDAC4 and HDAC5 with CaMKI confirmed that these 2 proteins interact (Figure H). The HDAC4-R601F mutation did not affect CaMKI binding to HDAC4, suggesting that CaMKI uses a different docking site than CaMKII. Further studies are needed to characterize the binding domains of other class IIa HDAC kinases, such as CaMKI, CaMKIV, and PKD.
α-Adrenergic agonists signal to HDAC4 via CaMKII.
A variety of neurohumoral signals have been shown to induce HDAC5 to translocate from the nucleus to the cytoplasm of cardiomyocytes as a consequence of PKD-dependent phosphorylation of 14-3-3–binding sites in its N terminal regulatory domain (16
). To determine whether HDAC4 was regulated in a similar manner, we examined its subcellular distribution in neonatal rat ventricular myocytes (NRVMs) exposed to phenylephrine (PE), a potent hypertrophic agonist that acts through the α-adrenergic receptor. As shown in Figure , A and B, PE caused a time-dependent cytosolic accumulation of HDAC4.
Cytosolic accumulation of HDAC4 in cardiomyocytes.
Cytosolic accumulation of HDAC4 in response to PE was markedly reduced by the general serine/threonine kinase inhibitor staurosporine as well as by the CaMKII inhibitors KN93, KN62, and autocamtide inhibitory peptide II-2 (AIPII-2). In contrast, bisindolylmaleimide I (Bis), Gö6976, and H89, which inhibit PKC, PKD, and protein kinase A, respectively, did not affect HDAC4 localization (Figure , A and C). Consistent with previous findings that the PKC-PKD axis regulates HDAC5 (16
), PE-induced nucleocytoplasmic shuttling of HDAC5 was not affected by KN93 (Figure A). The effectiveness of Bis was confirmed by the observation that it was able to prevent phosphorylation of the PKC phosphorylation sites on endogenous PKD (S744 and S748) in response to stimulation with PMA (Figure D), confirming that, despite the marked inhibition of PKD, PE-induced cytosolic accumulation of HDAC4 was caused by a PKC/PKD-independent signaling pathway, which we conclude from these results is the CaMKII pathway.
Adenoviral overexpression of WT HDAC4 did not prevent features of PE-induced cardiomyocyte hypertrophy, as assessed by the sarcomeric organization of NRVMs (Figure E), [3H]-leucine incorporation (Figure F), and atrial natriuretic peptide (ANP) immunostaining (Figure G), because it is signal responsive and accumulated in the cytosol. In contrast, an HDAC4 mutant, in which the signal-responsive serines were replaced with alanines (S246,467,632A), stayed in the nucleus and blocked the hypertrophic response to PE, suggesting that cytosolic accumulation of HDAC4 is essential for cardiomyocyte hypertrophy in response to CaMKII signaling (Figure , E–G).