To investigate the mechanisms whereby class IIa HDACs regulate the growth of cardiac myocytes we first identified HDAC4 target genes. HDAC4 was mildly overexpressed in primary cultures of neonatal rat ventricular cardiomyocytes (NRVMs) using recombinant adenoviruses expressing HDAC4 or control β-galactosidase, after which RNA was collected for transcript profiling. The array expression data showed that 1,130 genes were changed with HDAC4 overexpression, of which 815 genes were down-regulated and 315 up-regulated (Fig. S1
). Gene-grouping analysis using DAVID (Dennis et al., 2003
; Huang et al., 2009
) showed that HDAC4-repressed genes were enriched for cardiac transcripts, such as myofilament genes (12.3-fold enrichment, benjamini P < 2.2E-14), and heart contraction genes (11.4-fold enrichment, benjamini P < 1.8E-7). Interestingly, we also observed a significant enrichment for Ca2+
ion homeostasis genes (5.8-fold enrichment, benjamini P < 3.3E-6), a cardiac gene group not previously shown to be regulated by class IIa HDACs. These data suggest that HDAC4 represses specific gene groups associated with cardiac function and hypertrophy.
To build off the HDAC4 array screening approach and further refine the list of potential direct HDAC4 target genes in cardiomyocytes we performed a genome-wide occupancy mapping analysis in vivo using the DamID technique (Vogel et al., 2007
). The approach consisted of expressing an HDAC4-adenine methyltransferase (Dam, from Escherichia coli
) fusion protein to tag occupied DNA sequences by methylation in vivo. Methylated DNA fragments were then amplified, labeled, and hybridized to genomic arrays. We performed the DamID screen with both quiescent and hypertrophic cardiomyocytes using lentiviral vectors and a weak promoter to produce low levels of expression to guard against nonspecific methylation (Guelen et al., 2008
). The screen identified 6,434 genomic targets under quiescent conditions, 4,961 targets under hypertrophic conditions, and 2,751 targets present under both conditions (Fig. S1). Thus, more loci were identified under quiescent/repressed conditions, and for loci that were present under both conditions; the peak ratio (HDAC4-Dam/Dam) was significantly higher under quiescent conditions than under hypertrophic stimulation (1.16 ± 0.001-fold increase; P < 0.0001). Corroborating these data, the number of loci with multiple peaks was significantly higher under quiescent conditions (33.8 vs. 29%; P < 0.0001). These data support the paradigm that class IIa HDACs are mostly repressive under quiescent conditions and that hypertrophic stimuli relieve this inhibition (Chang et al., 2004
We next crossed the expression data with the genomic occupancy data to identify dynamically regulated HDAC4 targets. Of the 1,130 genes whose expression was significantly changed by HDAC4, 372 genes were identified as direct targets only under quiescence, 321 under hypertrophic conditions, and 221 under both conditions with the DamID screen (Fig. S1; data deposited in public database, see Materials and methods). These direct targets from the combined dataset displayed even higher enrichment scores for the contraction and the Ca2+ ion homeostasis gene groups (Fig. S1, chi square; P < 0.02), indicating that HDAC4 specifically controls select functional gene groups in the heart. We also performed a search for evolutionary conserved motifs in the promoters (−2,000 to +500 bp) of the HDAC4 target genes. However, multiple analyses did not reveal a conserved DNA sequence motif, suggesting that HDAC4 can function more broadly and is not restricted to conserved transcription factor binding sites.
We next experimentally validated the DamID and expression analysis results for the selected gene groups. Quantitative RT-PCR across 11 selected genes showed that the HDAC4 targets were either significantly repressed or activated, consistent with the expression arrays following expression of HDAC4 (). We also analyzed several of the direct HDAC4 target genes by chromatin immunoprecipitation (ChIP) for their promoter regions, which validated the DamID screen and showed occupancy of HDAC4 within 10 selected gene regions corresponding to their promoters (, band sizes given in Materials and methods). Changes in gene expression through HDAC4 also correlated with changes in protein levels of key Ca2+-handling genes (). We observed a reduction in expression of the Na+/Ca2+ exchanger (NCX1), phospholamban (PLN), and the sarcoplasmic reticulum Ca2+ ATPase (Serca2), a profile that should alter Ca2+ handling in myocytes. Indeed, we directly measured intracellular Ca2+ concentration and handling in stimulated cardiomyocytes using a fluorescent Ca2+ indicator (). Expression of HDAC4 resulted in a significant increase in baseline Ca2+ levels and a significant decrease in the transient relaxation time constant. Thus, HDAC4 acts in an orchestrated manner to alter the expression of several Ca2+-handling genes, validating the expression and promoter occupancy screen with a functional correlate.
Figure 1. Validation of the expression and occupancy screens. (A) Validation of the transcription array profiling using normalized quantitative RT-PCR in NRVM at baseline or with HDAC4 (HD4) adenoviral-mediated overexpression showing HDAC4 repressed and activated (more ...)
To investigate the transcriptional mechanisms whereby HDAC4 might selectively control gene expression in the heart we performed a modified yeast two-hybrid screen, termed the Ras recruitment system (RRS; Aronheim, 2004
). We used the N-terminal regulatory domain of HDAC4 (aa 3–666) as the bait with a heart cDNA library to identify prey plasmids that specifically interacted. This screen identified the C terminus of the Nup155 (aa 886–1391) as an HDAC4 binding partner. Examination in yeast showed that an HDAC4 N-terminal sequence (aa 3–281) was sufficient to bind Nup155 C terminus (). Coimmunoprecipitation (coIP) in mammalian cells using HA-tagged full-length Nup155 and His-tagged HDAC4 fragments showed that peptides containing amino acid 3–220 and 3–185 of HDAC4 were sufficient for the interaction, whereas a 3–165 fragment was not, suggesting that the interacting domain in HDAC4 is between amino acids 165 and 185 (). This domain in HDAC4 also mediates an interaction with MEF2 and HDAC1 (Chan et al., 2003
). Finally, recombinant GST was fused to the C terminus of Nup155 (aa 886–1391) and used to assess interaction with the recombinant HDAC4 N terminus (aa 3–628) fused to the maltose-binding protein (MBP). This analysis showed that GST-Nup155 was unable to bind MBP alone, although it did bind MBP-HDAC4 recombinant protein (, asterisk). GST alone did not bind MBP-HDAC4.
Figure 2. HDAC4 partners with Nup155 and associates with the NUPs. (A) Yeast two-hybrid growth assay with baits containing either the N terminus (aa 3–281 or 3–666) or C terminus (aa 632–1084) of HDAC4 (HD4) cotransfected with the library (more ...)
Coimmunoprecipitation experiments showed that endogenous HDAC4 interacted with endogenous Nup155 in extracts from primary cultures of NRVMs (). We used the well-validated mAb414 monoclonal antibody raised against the entire rat nuclear pore, which recognizes several NUPs but not Nup155 (Davis and Blobel, 1986
), to show that other NUPs (bands at 62 and 107 kD) may be part of this complex (, and unpublished data). Remarkably, immunofluorescent staining of cardiomyocytes showed that endogenous HDAC4 is enriched at the nuclear periphery, and partially colocalized with the mAb414 antibody (). Interestingly, HDAC5 was not able to bind Nup155 (unpublished data), suggesting greater specificity for HDAC4. Collectively, these data suggest that Nup155 and HDAC4 can interact in conjunction with other NUPs, especially at the nuclear periphery where the NPC is located.
Global HDAC inhibition was previously shown to alter chromatin association with the NUPs for several gene loci (Brown et al., 2008
), and a study in Drosophila
cells showed that mAb414-positive NUPs were found to associate with sites of active transcription (Capelson et al., 2010
). Therefore, we hypothesized that HDAC4 could modify the association of select genomic loci with the NUPs, and that the interaction between HDAC4 and Nup155 was required for this change in association. To test this hypothesis we initially performed ChIP assays using the mAb414 antibody in NRVMs, with or without expression of HDAC4 and TSA treatment. Many of the identified HDAC4 target genes indeed displayed some degree of association with the NUPs. Remarkably, as we hypothesized, overexpression of HDAC4 decreased the association of many target loci with the NUPs, and the addition of TSA restored this association (). Interestingly, several HDAC4 targets showed a reversed association pattern. Loci like Pln
, and Csrp3
displayed low association with NUPs at baseline but overexpression of HDAC4 enhanced the association, which was reversed with TSA (). Although this differential behavior of specific loci has been described in yeast (Casolari et al., 2004
) and mammalian cells (Brown et al., 2008
), the reasons for it are unknown. The reversal of the dissociation by TSA shows that the process is acetylation/deacetylation dependent, although the enzymes involved, such as association with class I HDACs, and the recipients of the acetyl group, are unknown. Thus, our results identify a mechanism whereby HDAC4 can modify chromatin association of select gene loci with NUPs.
Figure 3. HDAC4 binds and regulates the association of specific chromatin loci with NUPs. (A) ChIP assay in NRVM with the NUP (mAb414) antibody showing association of loci with NUPs with or without HDAC4, with or without TSA treatment. Input represents amplification (more ...)
Next we wanted to verify that the association with nucleoporins is required for HDAC4 action on gene expression. Unfortunately, knockdown of Nup155 resulted in cardiomyocyte death (unpublished data). To circumvent this problem we constructed a truncated Nup155 mutant, referred to as Nup155ΔC (aa 1–886), which lacks the HDAC4-interacting C-terminal domain. Analysis in cardiomyocytes showed that this truncation mutant is able to localize in a similar manner to full-length Nup155 (, arrowheads). Importantly, overexpression of Nup155ΔC did not seem to result in overt nuclear pore transport defects or otherwise compromise the health of cultured cells (Fig. S2
). Overexpression of Nup155ΔC by adenoviral-mediated gene transfer in NRVMs also did not result in major alterations in other NUPs as assessed with m414 or Nup153 antibody (unpublished data).
Remarkably, adenoviral overexpression of Nup155ΔC replaced endogenous full-length Nup155 and now no longer permitted endogenous HDAC4 to associate with nucleoporins ( and Fig. S3
). Careful analysis and quantification showed that ~30% of the HDAC4 colocalized with NUPs (mAb414) at baseline, and overexpression of Nup155ΔC reduced this association to only 3% (Fig. S3). More importantly, ChIP assays using immunoprecipitated NUPs showed that although HDAC4 is able to modify the association of selected loci with NUPs, coexpression of Nup155ΔC abolished the effect and restored NUPs association (). Similarly, coexpression of activated calmodulin-dependent kinase II (CaMKII), which can shuttle HDAC4 out of the nucleus (Haberland et al., 2009
), abolished the effect of HDAC4 and restored association with nucleoporins (). Thus, recruitment of HDAC4 by Nup155 binding is required for HDAC4’s ability to displace chromatin from the NUPs.
To further verify dynamic association between NUPs and HDAC4 on select target genes in NRVMs we performed fluorescent in-situ hybridization (FISH) combined with confocal microscopy and quantified the results in both a blinded and unblinded manner, producing similar results (). This analysis showed predominant peripheral nuclear localization of the Nppb, Acta1, Cacna1c loci during control conditions, a shift to a more central nuclear position under HDAC4 (HD4) overexpression, and reversal of the HDAC4 effect by expression of Nup155ΔC (). In contrast, the Pln locus demonstrates a reverse pattern. The Nlrx1 locus was used as a control, which showed no significant changes in intranuclear location (not depicted). These results corroborate the ChIP experiments and support the dynamic movement of specific loci to the nuclear periphery, the ability of HDAC4 to modify this balance, and the requirement of HDAC4–Nup155 interaction for this mechanism.
Figure 4. HDAC4 regulates spatial chromatin organization. (A) Confocal FISH of selected loci (green dots, arrows) with nuclear Topro3 counterstain (blue) in NRVMs showing predominant peripheral nuclear localization of the Nppb, Acta1, Cacna1c during control conditions, (more ...)
Consistent with the changes described above, detailed mRNA expression analysis using qRT-PCR showed that overexpression of Nup155ΔC reversed the expression pattern of HDAC4 for many of its target genes (). Functionally, these changes in gene expression correlated with alterations in cardiomyocyte hypertrophy. Indeed, although overexpression of HDAC4 reduced sarcomeric organization at baseline in quiescent NRVMs, expression of the Nup155ΔC mutant induced organization of sarcomeres and induced an increase in cell size, likely by relieving HDAC-dependent repression of select genes involved in myocyte growth ().
Figure 5. Non-HDAC4 binding Nup155ΔC reverses HDAC4 gene expression patterns and alters the growth response of NRVMs. (A) Quantitative qRT-PCR in NRVMs (normalized to Gapdh) shows that combined expression of Nup155ΔC reverses the HDAC4 expression (more ...)
Collectively, we identified clusters of sarcomeric genes and Ca2+
-handling genes that are directly repressed by HDAC4 and showed that HDAC4 exists in a complex with Nup155 and mAb414-positive NUPs, and that HDAC4 modified the association of these chromatin loci with the NUPs to control gene expression and hypertrophy. The association of these loci with the NUPs may not be locus- or sequence-specific, but rather the activation of functional gene groups may induce their association with the NUPs. HDAC4 appears to be affixed to the NUPs at rest in quiescent cardiomyocytes, which might function as a general inhibitory mechanism to quench gene expression of selected target loci. Upon agonist or mitogen stimulation, class IIa HDACs are exported from the nucleus, thereby enhancing chromatin–NUP association, gene transcription, and activation of otherwise repressed loci, now poised for rapid expression. Interestingly, NUPs were shown to interact with and regulate genes in the nucleoplasm, independent of the nuclear periphery (Capelson et al., 2010
; Kalverda et al., 2010
), and some NUPs are highly dynamic and rapidly shuttle between the NPC and the nucleoplasm (Hou and Corces, 2010
). Therefore, although we did not observe it directly in our system, it is possible that HDAC4 can modify the association between chromatin and NUPs in the nucleoplasm or on the NPC at the periphery.