Calcium-mediated histone modifications regulate alternative splicing in cardiomyocytes

Significance

Calcium is an important intracellular second messenger that regulates many biological processes. Many extracellular environmental cues lead to cellular calcium-level changes, which impact on the output of gene expression. In cardiomyocytes, calcium is known to control gene expression at the level of transcription, whereas its role in regulating alternative splicing has not been explored. Our studies demonstrate that in these cells a network of alternatively spliced exons exists, which responds to the altered calcium levels by changing their splicing patterns. Our studies further elucidate an epigenetic regulatory mechanism, triggered by calcium signaling pathways, that leads to histone hyperacetylation along gene bodies, which increases the transcriptional elongation rate of RNA polymerase II and impacts alternative splicing.

Abstract

In cardiomyocytes, calcium is known to control gene expression at the level of transcription, whereas its role in regulating alternative splicing has not been explored. Here we report that, in mouse primary or embryonic stem cell-derived cardiomyocytes, increased calcium levels induce robust and reversible skipping of several alternative exons from endogenously expressed genes. Interestingly, we demonstrate a calcium-mediated splicing regulatory mechanism that depends on changes of histone modifications. Specifically, the regulation occurs through changes in calcium-responsive kinase activities that lead to alterations in histone modifications and subsequent changes in the transcriptional elongation rate and exon skipping. We demonstrate that increased intracellular calcium levels lead to histone hyperacetylation along the body of the genes containing calcium-responsive alternative exons by disrupting the histone deacetylase-to-histone acetyltransferase balance in the nucleus. Consequently, the RNA polymerase II elongation rate increases significantly on those genes, resulting in skipping of the alternative exons. These studies reveal a mechanism by which calcium-level changes in cardiomyocytes impact on the output of gene expression through altering alternative pre-mRNA splicing patterns.

Alternative splicing is a robust mechanism that regulates the functional output of a genome. More than 90% of human protein-coding genes undergo alternative splicing, which significantly increases proteomic complexity of the human genome (1, 2). The precise spatial–temporal regulation of alternative splicing plays a crucial role in controlling gene expression. Decades of studies have yielded a wealth of knowledge on how tissue- and developmental stage-specific alternative splicing is regulated (3). However, signaling-controlled alternative splicing in response to environmental cues has attracted far less attention, and regulatory mechanistic insights have only begun to emerge in recent years (4).

Calcium is an important intracellular second messenger that regulates many biological processes. Although most studies have focused on how calcium regulates gene expression at the transcriptional level (5, 6), a small number of studies have also explored how changes in intracellular calcium levels can lead to alternative splicing pattern alterations (7–12). For example, in neuronal cells, a splicing-sensitive exon array identified more than 5,000 genes that change their splicing pattern in response to elevated calcium levels (13). However, because only a handful of studies have investigated the underlying mechanisms, it remains largely unknown how calcium-induced alternative splicing changes are regulated. These studies have revealed two distinct mechanisms by which calcium-responsive alternative exons can be regulated: one RNA-binding protein (RBP)–dependent and the other RBP-independent (14, 15).

Several studies using depolarized neurons have revealed RBP-dependent mechanisms that influence alternative exon inclusion, which involve calcium-induced interactions between short RNA sequences and RBP splicing regulators (7, 12). Changes of these RNA–RBP interactions are caused by calcium-induced phosphorylation and/or cellular localization changes of RBPs. For example, in response to higher calcium levels, the splicing regulators heterogeneous ribonucleoprotein A1 (hnRNPA1) and RNA binding protein, fox-1 homolog (Rbfox1) were shown to have increased nuclear localization, whereas transformer-2 protein homolog beta 1 (Tra2β1) was shown to translocate to the cytoplasm (7, 16, 17). The altered nuclear levels of these proteins lead to changes in alternative splicing of the downstream target genes that are regulated by these splicing factors.

By contrast, a recent study of neural cell adhesion molecule (NCAM) alternative splicing implicated an RBP-independent, epigenetic regulatory mechanism that is induced by calcium (11). This study indicated that, in neurons, depolarization-induced skipping of NCAM exon 18 is correlated with a localized H3K9 hyperacetylation and H3K36 trimethylation (11). These results, in combination with pre-mRNA accumulation data, imply that the local transcriptional elongation rate of RNA polymerase II (RNAPII) on NCAM near exon 18 was increased in depolarized cells (11). As such, this study supports a signaling-responsive mechanism that is consistent with the kinetic coupling model of transcription and splicing, where an increased transcriptional elongation rate leads to skipping of alternative exons because they are generally surrounded by suboptimal splicing signals and thus need more time to be recognized (18). Although the proposed mechanism is very intriguing, it is based on a single example, and it is not clear whether the proposed mechanism can be generalized to other genes and cell types.

Calcium and calcium-dependent signaling also play important roles in regulating gene expression in cardiomyocytes, and abnormal calcium levels can affect gene expression at the level of transcription (19). Specifically, a number of studies have demonstrated that elevated intracellular calcium levels activate the calcium/calmodulin-dependent protein kinase (CaMK) or the protein kinase C/D (PKC/PKD) (20–25). Consequently, the activated kinases lead to increased transcription of a number of cardiac-specific transcription factors through a cascade of events including changes in histone modifications (20–25). No studies to date have explored whether calcium signaling can also affect pre-mRNA processing in the cardiovascular system.

Here we establish that in cardiomyocytes there is a calcium-controlled network of alternative splicing events. Using cardiomyocytes isolated from neonatal mice, we provide a number of examples where elevated intracellular calcium levels lead to a significant yet reversible increase in skipping of alternative exons. Interestingly, our analysis suggests that calcium-mediated splicing regulation likely occurs through an RBP-independent mechanism. This mechanism links calcium-responsive kinase activities and alternative splicing through controlling histone modifications and transcriptional elongation rate.

Results

Robust and Dynamic Calcium-Induced Changes of the Alternative Splicing of Nf1 Exon 23a in Mouse Cardiomyocytes.

The Nf1 gene plays an essential role in heart development and function (26, 27). Nf1 exon 23a is alternatively spliced, and its inclusion inserts 21 amino acids into the GTPase-activating protein (GAP)–related domain of the Nf1 gene product neurofibromin, leading to a significant reduction of its Ras-GAP activity in multiple cell types (28–30). In mouse cardiomyocytes, the alternative exon Nf1 exon 23a is included in 70% of transcripts (Fig. 1). To investigate whether the splicing pattern of this exon changes in response to environmental stimuli, we exposed cardiomyocytes prepared from postnatal day 1–3 mouse hearts to increasing concentrations of extracellular potassium chloride (KCl), which elevates intracellular calcium levels by depolarizing cells. Sodium chloride (NaCl) was used as a negative control. Semiquantitative reverse transcriptase–PCR (RT-PCR) performed with primers annealing to the exons surrounding Nf1 exon 23a using total RNA extracted from these cells indicates that exon inclusion decreased drastically from 69 to 15% with increasing amounts of KCl (Fig. 1 A and B). In a time course experiment, we observed that inclusion of Nf1 exon 23a decreases from 72 to 9% within 24 h of treatment with 100 mM KCl (Fig. 1C). We observed the splicing pattern change starting after 6 h of the KCl treatment, which is expected because the Nf1 mRNA half-life is 6 h (Fig. S1A). Remarkably, when we replaced KCl-containing culture media with regular media, the splicing pattern change was completely reversed (Fig. 1D), indicating that the splicing pattern of Nf1 exon 23a is highly dynamic in responding to environmental stimuli. Note that the cells appeared normal throughout the duration of the experiments. In fact, the rhythmic beating of the primary cardiomyocytes, which ceased during KCl treatment, was restored after KCl was removed. Also, the level of neurofibromin, the protein product of Nf1, remained constant in untreated and treated cardiomyocytes (Fig. S1B).

Fig. 1.

KCl induces a change in the alternative splicing pattern of Nf1 exon 23a in primary mouse cardiomyocytes. (A) Schematic diagram indicating the alternative splicing pattern of Nf1 exon 23a. The arrows show the locations of RT-PCR primers. (B–D) Semiquantitative RT-PCR analysis of the alternative splicing of Nf1 exon 23a in primary mouse neonatal cardiomyocytes treated with the indicated concentrations of KCl or control NaCl. The numbers indicate the % inclusion of Nf1 exon 23a relative to total Nf1 transcripts. (B) Dose–response of KCl treatment. Cells were treated with increasing concentrations of KCl, as indicated, for 24 h. n = 5. (C) Time course analysis of the KCl treatment. Cells were treated with 100 mM KCl and collected for RNA isolation at the indicated times. n = 3. (D) Recovery time course analysis. Cells were treated with 100 mM KCl for 24 h followed by recovery without KCl for the indicated times. n = 3. Error bars represent standard deviations.

To determine the cell-type specificity of this phenomenon, we conducted similar experiments in several cell types. Whereas cardiomyocytes differentiated from mouse ES cells showed similar changes in the splicing pattern of Nf1 exon 23a (Fig. S2 A–C), no changes were observed in undifferentiated mouse ES cells (Fig. S2D). These results suggest that the KCl-induced splicing changes of Nf1 exon 23a are specific to cardiomyocytes.

The consistently observed stronger RT-PCR signal in KCl-treated samples prompted us to investigate whether the Nf1 mRNA transcript level changed in these cells. Real-time RT-PCR analyses indicate that KCl treatment leads to up to a fourfold increase of Nf1 transcript level in both primary and ES cell-derived cardiomyocytes but not in ES cells (Fig. S1C). The Nf1 transcript level is significantly reduced in cells upon recovery from KCl treatment (Fig. S1C). These results indicate that the KCl-induced alternative splicing change of Nf1 exon 23a is accompanied by increased transcript level.

We hypothesized that the altered levels of intracellular calcium are responsible for the KCl-induced splicing changes of Nf1 exon 23a, because elevated levels of extracellular KCl are known to lead to calcium influx via L-type voltage-gated calcium channels (31–33). We used two established pharmacological approaches to examine the role of calcium in splicing regulation. We found that nifedipine, an L-type calcium channel blocker, significantly reduced the effect of KCl on alternative splicing of Nf1 exon 23a (Fig. S3A). A similar result was observed with the use of AP5, an NMDA receptor antagonist, in conjunction with BAPTA-AM, a calcium chelator (Fig. S3B). To further examine the role of intracellular calcium in stimulus-induced regulation of Nf1 alternative splicing, we treated primary or ES cell-derived cardiomyocytes with increasing concentrations of caffeine, which is known to increase intracellular calcium levels by acting as a rynodine receptor agonist, promoting release of calcium from the sarcoplasmic reticulum to the cytoplasm (34). Interestingly, caffeine treatment led to a reduction of Nf1 exon 23a inclusion from 68% of the transcripts in the control cells to 3% of the transcripts in cardiomyocytes in a dose-dependent manner (Fig. S3 C and D). These results establish that the alternative splicing pattern of Nf1 exon 23a changes in response to altered intracellular calcium levels.

Calcium-Regulated Changes in the Alternative Splicing Pattern of Additional Genes.

To determine whether the splicing change regulated by intracellular calcium levels in cardiomyocytes occurs with other alternative exons, we performed semiquantitative RT-PCR on nine alternative exons known to be differentially spliced during heart development (35). Interestingly, over 50% of these exons (five of nine) showed changes in their splicing patterns after exposure of the primary cardiomyocytes to KCl. These exons include Ktn1 exon 36, Ank2 exon 21, Enah exon 5, and Mef 2A exon 11 (Fig. 2). In all cases, we observed increased skipping of the alternative exons after KCl treatment (Fig. 2). Importantly, for all of these exons, the splicing pattern was reverted back to the baseline level in untreated cells when KCl was removed (Fig. 2). Among the alternative exons that did not show a change in splicing is Capzb exon 8 (Fig. 2), which provided an important negative control in the subsequent mechanistic studies. These results demonstrate that calcium regulates alternative splicing of several genes in cardiomyocytes. Interestingly, we noted an apparent correlation between the splicing pattern change and transcript levels for these genes too (Fig. 2).

Fig. 2.

Calcium-mediated dynamic changes in alternative splicing of other alternative exons. Semiquantitative RT-PCR analysis using total RNA isolated from primary cardiomyocytes treated with 100 mM NaCl (Na), KCl (K), or KCl followed by recovery (R). Primers annealing to exons surrounding the alternative exons were used. Products representing inclusion (open circles) or skipping (solid circles) of an alternative exon are indicated. n = 3. Error bars represent standard deviations.

Calcium-Regulated Alternative Splicing Is Mediated by the CaMKIIδ or PKD1 Signaling Pathways.

Abnormal calcium signaling has been shown to cause cardiac remodeling through several calcium signaling molecules, including PKC/PKD and CaMKII (21, 25). To investigate the signaling pathways involved in the calcium-mediated changes in splicing patterns, we treated primary cardiomyocytes with inhibitors blocking the function of these kinases, Gö6976 targeting PKC/PKD and KN93 targeting CaMK, in conjunction with the KCl treatment. Interestingly, as shown in Fig. 3A, the KCl-induced splicing pattern change of Nf1 exon 23a was significantly reduced only when the CaMK and PKC/PKD pathways were inhibited simultaneously, but not when the pathways were inhibited individually. These results suggest that the CaMK and PKC/PKD signaling pathways have redundant functions in regulating calcium-induced splicing changes of Nf1 exon 23a.

Fig. 3.

Calcium-mediated alternative splicing is controlled by the CaMKIIδ and PKD1 signaling pathways. Semiquantitative RT-PCR analysis was carried out to examine the Nf1 exon 23a splicing pattern in primary cardiomyocytes treated with 100 mM control NaCl or KCl for 24 h in the absence or presence of inhibitors (A) or siRNA (B). (A) In addition to NaCl or KCl, cells were treated with the PKC/PKD inhibitor Gö6976 (10 μM) and the CaMK inhibitor KN93 (1 μM) individually or in combination. (B) In addition to NaCl or KCl, cells were treated with 200 pmol of siRNA targeting PKD1 or CaMKIIδ individually or in combination. A scrambled nontargeting siRNA was used as a control. n = 3. Error bars represent standard deviations.

We next carried out siRNA knockdown experiments to further investigate the role of these two signaling pathways in splicing regulation. Both the CaMK and PKC/PKD protein kinase families contain multiple members. We chose to knock down CaMKIIδ and PKD1, because their roles in regulating gene expression in cardiomyocytes have been well-established (21, 23, 25). We found that siRNA-mediated depletion of both CaMKIIδ and PKD1 completely abolished the effect of KCl on Nf1 exon 23a alternative splicing, whereas individual knockdown had no effect (Fig. 3B), corroborating the inhibitor study. The siRNA knockdown efficiency is shown in Fig. S4 A and B. siRNA knockdown of CaMKIIδ and PKD1, either individually or in combination, had no effect on the alternative splicing of Nf1 exon 23a in cardiomyocytes treated with the NaCl control (Fig. 3B). These results indicate the important role of CaMKIIδ and PKD1 in calcium-regulated alternative splicing.

We further tested whether depletion of these kinases affected alternative splicing of the other calcium-responsive exons shown in Fig. 2. We found that KCl-induced skipping of Ktn1 exon 36 and Ank2 exon 21 was also abolished when the cells were treated with siRNAs targeting both CaMKIIδ and PKD1, whereas the splicing pattern of Capzb exon 8 was not affected (Fig. S4C). These results support the hypothesis that the CaMKIIδ and PKD1 pathways play an important role in regulating multiple alternative splicing events in response to changes in calcium levels.

Histone Hyperacetylation in Depolarized Cardiomyocytes.

Up-regulation of the CaMKIIδ and PKD1 pathways has been shown to increase cytoplasmic localization of class II histone deacetylases HDACs 4, 5, 7, 9) under hypertrophic stress in neonatal rat cardiomyocytes (20, 22–24). To determine whether the depolarization-induced calcium-level increase leads to export of class II HDACs from the nucleus to the cytoplasm, we performed an immunofluorescence staining assay to examine the cellular localization of HDACs in NaCl- or KCl-treated cells and in KCl-treated cells followed by recovery. Using an antibody that recognizes HDACs 4, 5, and 7, we found that KCl treatment significantly depleted these class II HDACs from the nucleus, which was reversed after recovery, whereas the level of these HDAC proteins remained constant (Fig. S5A). The cellular localization change is likely specific to class II HDACs, because we did not observe any change with HDAC2, a class I HDAC (Fig. S5B). Furthermore, Western blot analysis using nuclear and cytoplasmic protein fractions indicated that the levels of the class II HDACs are reduced in the nucleus and increased in the cytoplasm (Fig. S5C).

To examine the effect of HDAC translocation on histone acetylation, we performed Western blot analysis using acid-extracted histones from primary cardiomyocytes and antibodies against acetylated histone H3 and H4. We observed a 2.5-fold increase of histone acetylation after depolarization for both histones H3 and H4 (Fig. 4A). The total protein levels of histone H3 and H4 remained unchanged (Fig. 4A). Histone acetylation was completely restored to basal levels 24 h after KCl removal (Fig. 4A). Similar results were obtained using ES cell-derived cardiomyocytes, whereas no change was observed with undifferentiated ES cells (Fig. S6A). Histone hyperacetylation induced by KCl was blocked when inhibitors of PKC/PKD and CaMK were used simultaneously but not individually (Fig. S6B). These results demonstrate that depolarization of cardiomyocytes leads to an increase in global acetylation of histone H3 and H4.

Fig. 4.

Histone hyperacetylation is induced by elevated intracellular calcium levels. (A) Primary cardiomyocytes were treated as described in Fig. 2. Total histones were acid-extracted from the treated cells. Western blot analysis was carried out using antibodies specific for total or pan-acetylated histone H3 or H4. The graphs show the fold differences between KCl- and NaCl-treated cells. n = 3. Error bars represent standard deviations. (B) Western blot analysis using histones prepared as described in A and specific antibodies as indicated. Arrows indicate hyperacetylation at specific lysines. (C and D) ChIP analysis using primary cardiomyocytes treated with NaCl (light gray bars) or KCl (black bars). IP was performed using antibodies against pan-acetylated H3 or total H3 and real-time PCR assay using primers targeting different regions of Nf1 (C) or Capzb (D) as indicated. The ratio of acetylated to total histone H3 was plotted in the graphs with the value for each region in NaCl-treated cells set at 1. n = 3. The P values of the Student t test for each region were calculated, and the asterisk represents a statistically significant P value (<8.9 × 10⁻⁸).

To study acetylation of specific lysines on histone H3, we examined H3 acetylated at lysine 9, 14, and 27 and H4 acetylated at lysine 16, and found similar change patterns for all of the examined lysines (Fig. 4B). We also examined the histone methylation status of H3K4 and H3K36 and found no changes (Fig. 4B). These results suggest that depolarization of cardiomyocytes induces histone hyperacetylation at multiple lysines.

The seemingly global hyperacetylation result predicts that hyperacetylation occurs on broad instead of localized chromosomal regions. Thus, we hypothesize that for any given gene, hyperacetylation occurs along the gene body instead of on a limited region such as the promoter region. To test this hypothesis, we performed a chromatin immunoprecipitation (ChIP) assay comparing Nf1 and Ktn1, the two genes that contain calcium-responsive alternative exons, with Capzb, the gene that contains a nonresponsive alternative exon (Fig. 2). Our results showed a positive correlation between histone hyperacetylation throughout the gene body and calcium responsiveness, with approximately fourfold increase of histone H3 acetylation along the Nf1 and Ktn1 genes and no change along the Capzb gene after depolarization (Fig. 4 C and D and Fig. S6C). Note that throughout the manuscript, each ChIP analysis was carried out in three independent experiments (i.e., n = 3) and each real-time PCR assay was carried out in duplicate (two PCR assays with each experiment). For ChIP analysis, the IP values were normalized to input.

The Role of Histone Acetyltransferases in Calcium-Induced Alternative Splicing Changes.

We reasoned that the calcium-induced translocation of class II HDACs would provide more access for histone acetyltransferases (HATs) to acetylate histones on genes that show a change in alternative splicing pattern. To test this hypothesis, we focused on two HATs, CBP and p300, which are known to play important roles in heart functions (36–38). We first examined their association with the three genes discussed above, Nf1, Ktn1, and Capzb. Our ChIP analysis using antibodies against CBP or p300 showed that, in mouse primary cardiomyocytes treated with KCl, association of CBP and p300 along the Nf1 and Ktn1 genes increased moderately by up to 1.5-fold (Fig. 5A and Fig. S7A), whereas their association with the Capzb gene did not change (Fig. 5B). Statistical analysis indicated that the moderate increase of CBP and p300 along Nf1 and Ktn1 is statistically significant (Fig. 5A and Fig. S7A).

Fig. 5.

Role of the HATs p300 and CBP in a calcium-induced alternative splicing pattern change in primary cardiomyocytes. (A and B) ChIP analysis to examine the association of the HATs CBP and p300 with the Nf1 (A) or Capzb (B) genes in NaCl- or KCl-treated cells. Cells were treated with 100 mM NaCl or 100 mM KCl for 24 h. ChIP analysis was carried out using anti-CBP or anti-p300 antibodies and real-time PCR using primers targeting the indicated regions of Nf1 or Capzb. The coimmunoprecipitated gene fragments were normalized to the input value. Fold differences of the HAT association with Nf1 or Capzb in NaCl- or KCl-treated cells are shown in the graphs. n = 3. The P values of the Student t test for each region were calculated, and the asterisk represents a statistically significant P value (<8.9 × 10⁻⁵). (C) Cardiomyocytes were treated with siRNAs targeting CBP or p300. A nonspecific siRNA was used as a control. All cells were then treated with 100 mM KCl for 24 h or 100 mM NaCl, followed by semiquantitative RT-PCR to examine the Nf1 exon 23a splicing pattern. Expression of CBP and p300 protein was detected by Western blot analysis using anti-CBP or anti-p300 antibodies. γ-Tubulin was used as a loading control. n = 3. Error bars represent standard deviations.

To determine whether the activities of CBP and p300 are responsible for the calcium-induced histone hyperacetylation on Nf1 and Ktn1, we carried out an siRNA knockdown assay to reduce the protein level of CBP or p300 (Fig. 5C, Right). The RT-PCR splicing analyses indicated that the KCl-induced splicing changes for Nf1 exon 23a and Ktn1 exon 36 were at least partially reversed when the CBP or p300 level was reduced (Fig. 5C, Left, and Fig. S7B). Knockdown of both CBP and p300 in KCl-treated cells led to a moderately stronger reversal of the splicing phenotype of Nf1 exon 23a (Fig. S7D). As expected, no change was observed in the splicing of Capzb exon 8 (Fig. S7C). The NaCl-treated cells did not show any changes in splicing of these exons when CBP or p300 was knocked down (Fig. 5C and Fig. S7 B and C). Furthermore, interestingly, the increased association of CBP and p300 on Nf1 and Ktn1 was abolished when kinase inhibitors were used (Fig. S8). These results suggest that increased activity of HATs, such as CBP and p300, along gene bodies is responsible for histone hyperacetylation, which leads to increased skipping of alternative exons when cells are stimulated by calcium.

The Role of Histone Deacetylases in the Specificity of Calcium-Induced Alternative Splicing Changes.

An important question in calcium-induced alternative splicing change is how gene specificity is achieved. In other words, why is the splicing pattern for Capzb exon 8 not changed after KCl treatment? One hypothesis is that class II HDACs are differentially recruited to specific genes and, thus, only those genes that are associated with class II HDACs in untreated or NaCl-treated cells will show splicing pattern change after the KCl treatment, which induces the cytoplasmic translocation of these HDACs. Our ChIP analyses using antibodies specific for the class II HDACs HDAC4 or HDAC5 or the class I HDAC HDAC2 indicate that the baseline levels of the HDAC association are similar with the three genes (Fig. S9). Interestingly, however, when cardiomyocytes were treated with KCl, the association of class II HDACs was reduced along Nf1 and Ktn1, but not Capzb, and the reduction was blocked by use of the two kinase inhibitors, whereas the association of HDAC2 does not change on any of the three genes (Fig. S10). These results suggest that the differential dissociation of class II HDACs provides the specificity of the KCl-induced splicing change.

Elevated Intracellular Calcium Levels Correlate with Increased Transcriptional Elongation Rate in Cardiomyocytes.

These results, combined with our observation that altered calcium levels did not affect either the total protein level or cellular localization pattern of the splicing regulators known to affect Nf1 exon 23a alternative splicing (Fig. S11), prompted us to test the role of the kinetic coupling model in the calcium-induced skipping of alternative exons. The kinetic coupling model posits that the RNAPII elongation rate affects the splicing outcome of an alternative exon by changing the time allowed for an alternative exon to be recognized by the splicing machinery assembled on a transcribing, nascent pre-mRNA (18, 39, 40). The higher elongation rate leads to increased skipping of alternative exons, which are typically associated with suboptimal splicing signals, due to a shorter window of time for the exons to be defined. We hypothesized that calcium-induced hyperacetylation led to a more relaxed chromatin configuration, allowing for faster RNAPII elongation and, thus, increased exon skipping.

To test this hypothesis, we examined the RNAPII occupancy on the three genes discussed above by ChIP analysis using the H5 antibody specific to Ser2 phospho-C-terminal domain (CTD) of the large subunit of RNAPII, a hallmark of elongating RNAPII. We observed a significant reduction of RNAPII occupancy on the Nf1 and Ktn1 genes when cells were depolarized (Fig. 6A and Fig. S7E), but no changes on Capzb (Fig. 6A). These observations indicate an inverse correlation between RNAPII occupancy and exon skipping in depolarized primary cardiomyocytes, and are consistent with our hypothesis that RNAPII moves faster on Nf1 and Ktn1 in depolarized cells.

Fig. 6.

Calcium-induced RNAPII elongation rate increase. (A) RNAPII occupancy examined by ChIP analysis using H5 antibody. The coimmunoprecipitated gene fragments were normalized to the input value. The graphs indicate the relative level of RNAPII accumulation at the indicated regions along the Nf1 or Capzb genes in primary cardiomyocytes treated with 100 mM NaCl (light gray) or KCl (black). n = 3. (B) The RNAPII transcriptional elongation rate along three genes in ES cell-derived cardiomyocytes treated with 100 mM NaCl or KCl. The elongation rate was measured as described in Materials and Methods. n = 3. The asterisk represents a statistically significant P value (<2.5 × 10⁻⁴). (C) Semiquantitative RT-PCR to examine the Nf1 exon 23a splicing pattern in primary cardiomyocytes that were treated with 3.5 μM CPT in addition to the NaCl or KCl treatment. n = 4. Error bars represent standard deviations.

To provide definitive evidence to support our model, we measured the transcriptional elongation rate of RNAPII on the three genes in ES cell-derived cardiomyocytes in the absence or presence of KCl using a procedure initially described by Singh and Padgett and modified by our laboratory (41, 42). In this experiment, ES cell-derived cardiomyocytes were treated with 5,6-dichloro-1-β-d-ribofuranosylbenzi​midazole (DRB) to inhibit transcription. After removal of DRB, cells were collected at 0, 15, 30, 60, 75, and 90 min, and pre-mRNA accumulation of specific regions of a given gene was analyzed by real-time RT-PCR using several sets of primer pairs spanning the entire gene of interest (Figs. S12–S14). The transcriptional elongation rate was calculated using the pre-mRNA accumulation data, as described (41). Interestingly, the transcriptional elongation rate for Nf1 and Ktn1 nearly doubled in KCl-treated cells compared with NaCl-treated cells, from 3.2 kb/min to 6.8 kb/min for Nf1 (Fig. 6B and Fig. S12) and from 2.4 kb/min to 4.14 kb/min for Ktn1 (Fig. 6B and Fig. S13). Most importantly, the transcriptional elongation rate for Capzb did not show a statistically significant change, from 1.93 kb/min to 2.08 kb/min (Fig. 6B and Fig. S14). The increased elongation rate is likely throughout the affected genes, because the calculated rate for Nf1 is the same using data of pre-mRNA accumulation in more than one region of the gene.

These data establish a strong correlation between hyperacetylation, faster transcriptional elongation, and increased skipping of alternative exons. To further strengthen the conclusion on transcriptional elongation, we treated depolarized primary cardiomyocytes with camptothecin (CPT), a drug used to slow down transcriptional elongation (41, 43). Use of CPT in combination with KCl completely restored the splicing pattern of Nf1 exon 23a (Fig. 6C) and Ktn1 exon 36 (Fig. S15A), but had no effect on Capzb exon 8 (Fig. S15B).

Taken together, the results presented here provide strong support for the RBP-independent model that, in response to increased calcium in cardiomyocytes, the RNAPII elongation rate is increased, leading to more skipping of alternative exons. Furthermore, our studies reveal the mechanistic link between increased calcium and RNAPII elongation rate change. As diagrammed in Fig. 7A, the elevated intracellular level of calcium increases the activity of CaMKIIδ or PKD1, which induces the nucleus-to-cytoplasm translocation of class II HDACs. As a consequence, the association of HATs with the genes containing calcium-responsive alternative exons is increased, resulting in hyperacetylation of H3 and H4 along the gene body of these genes. This allows for faster RNAPII elongation rate on these genes (Fig. 7B). Note that this conclusion is based on studies of two genes that contain calcium-responsive exons (Nf1 and Ktn1). We believe that it can be applied to other genes that contain calcium-responsive exons, such as those shown in Fig. 2. Indeed, at least Ank2 appears to be regulated similarly, as use of kinase inhibitors also reversed the KCl-induced splicing change (Fig. S4C). It will be interesting to study whether histone hyperacetylation and increased transcriptional elongation rate occur with Ank2 and other calcium-responsive exon-containing genes.

Fig. 7.

Model illustrating how increased intracellular calcium levels lead to chromatin changes, which result in changes in alternative splicing patterns. (A) Increased calcium levels activate CaMK and PKD, which phosphorylate class II HDACs, causing their translocation from the nucleus to the cytoplasm. (B) The disrupted HAT–HDAC balance leads to histone hyperacetylation, which relaxes chromatin, resulting in an increase in RNAPII transcriptional elongation rate. The dashed arrow indicates that if exon skipping occurs in an RBP, it may change the ability of the RBP to regulate downstream alternative splicing events. Black ovals indicate acetyl marks on histones. txn, transcription.

Discussion

Our analysis of calcium-mediated splicing regulation in cardiomyocytes has revealed a previously unidentified regulatory mechanism that integrates histone modifications, transcription, and splicing. We also demonstrate that in addition to transcription, calcium-mediated epigenetic changes can regulate gene expression at the level of alternative splicing in cardiomyocytes.

RBP-Independent vs. RBP-Dependent Calcium-Responsive Mechanisms.

In cardiomyocytes, we have discovered an alternative splicing mechanism that appears to be splicing regulator-independent. We show that increased calcium levels lead to changes in chromatin modification, which in turn lead to exon skipping. We find no evidence that this mechanism depends on calcium-induced changes of splicing regulator levels, posttranslational modifications, or cellular localization (Fig. S11). We posit that this seemingly RBP-independent mechanism functions beyond, yet in the context of, classical RBP–RNA interactions.

However, it is unlikely that the mechanism we describe here is the only mechanism induced by calcium. Instead, it is likely to work in conjunction with, and/or in parallel to, other RBP-driven splicing regulatory mechanisms. In this context, we can envision several scenarios. First, when CaMKII and PKD are activated, they may phosphorylate RBPs, which would lead to changes in their activity and splicing patterns of their target pre-mRNAs. An excellent example of this is the depolarization-induced skipping of the stress axis-regulated exon (STREX) of the Slo1 gene. Membrane depolarization of cerebellar neurons induces phosphorylation of hnRNP L at Ser513 by CaMKIV, which enhances the interaction of hnRNP L with the CaMKIV-responsive RNA element (CaRRE1) upstream of STREX. As a result, U2AF binding is reduced and STREX is skipped (44). Second, histone hyperacetylation can cause up-regulation of specific splicing regulators at the transcription level, which will lead to either inclusion or skipping of their target alternative exons (Fig. 7B). Third, if epigenetically induced skipping of an exon occurs in a splicing regulator, alternative splicing changes of target pre-mRNAs can then occur as a secondary event (Fig. 7B). A potential example of this is the Rbfox1-mediated splicing pattern change in depolarization of P19 cell-derived neurons (17). In this case, depolarization induces skipping of exon 19 of Rbfox1, which increases nuclear localization of Rbfox1. As a result, several alternative exons change their splicing patterns (17). It is not known how Rbfox1 exon 19 is skipped in depolarized cells. Because this exon is not associated with any CaRRE motifs, it is tantalizing to speculate that the epigenetic mechanism described here is responsible for the initial skipping event in an alternative splicing cascade.

In our study, use of a calcium chelator in conjunction with an NMDA receptor antagonist in KCl-treated cardiomyocytes relieved Nf1 exon skipping without causing an apparent change of transcript level (Fig. S3 A and B). This result appears to imply that additional, possibly RBP-mediated mechanisms exist in calcium-induced splicing control.

Calcium-Induced Histone Modification Changes That Regulate Alternative Splicing.

Hyperacetylation of histones leads to more relaxed chromatin configurations, which support more rapid RNAPII transcription (45, 46). Although it is well-known that histone hyperacetylation at the promoter region leads to increased rates of transcription initiation (47), the impact of hyperacetylation on transcription along the gene body has not been extensively studied. Here we measured transcriptional elongation rate on three genes, two containing calcium-responsive alternative exons and one containing a calcium-nonresponsive alternative exon. We uncovered a very strong correlation between histone hyperacetylation, transcriptional elongation rate, and exon skipping. Significantly, use of CPT to slow down transcriptional elongation abolished the effect of increased calcium levels on alternative splicing (Fig. 6).

A number of recent studies have linked the rate of RNAPII transcriptional elongation and the regulation of alternative splicing of the pre-mRNA being transcribed, although transcriptional elongation rate was not measured directly in most of these studies (42, 48–55). Several distinct mechanisms that regulate transcriptional elongation rate to impact alternative splicing have been characterized (39, 56). The RNAPII elongation rate can be regulated by association of heterochromatic proteins induced by siRNA or small RNA (sRNA) (48, 50, 53), interaction of chromatin remodeling complexes with histone marks or methylated DNA (49, 51, 52), or structural changes in RNAPII (54, 55). A recent study from our laboratory demonstrated that alternative splicing of Nf1 exon 23a can also be regulated at the chromatin level in a Hu protein-dependent manner in neurons (42), which is a distinct mechanism from the one presented here. In this case, binding of Hu proteins to its target sequences on pre-mRNA blocks HDAC activity, leading to a local histone hyperacetylation surrounding exon 23a, increasing the RNAPII elongation rate of the subsequent rounds of transcription (42). We showed that neither the expression level nor cellular localization of HuR, the Hu protein family member expressed in cardiomyocytes, is changed in KCl-treated cells (Fig. S11). In addition, in neurons, Nf1 exon 23a is predominantly skipped, and depolarization would not further increase skipping. These observations highlight the distinct regulatory mechanisms on the same alternative exons in different cell types or in responding to different external conditions.

However, we predict that the mechanism we uncovered in cardiomyocytes exists in other cell types such as neurons, although the specific players of the pathways may be different. To date, only one other example, exon 18 of NCAM, was reported to undergo chromatin-level regulation in response to increased calcium levels, albeit in a different cell type, depolarized neurons (11). Comparing the NCAM study with our results, we notice both similarities and differences. Even though transcriptional elongation was not measured directly, the data for pre-mRNA accumulation of regions downstream and upstream of NCAM exon 18 imply that the elongation rate was increased after depolarization, leading to more skipping of this exon (11). However, although the increased transcriptional elongation rate for Nf1 and Ktn1 appears to be global, it was suggested that the elongation rate increase for the NCAM gene is local, because the localized increase of H3K9 acetylation and H3K36 trimethylation after depolarization correlated with the increased elongation rate and skipping of exon 18 (11). Interestingly, we did not observe any differences in H3K36Me3 or H3K4Me3 after depolarization of cardiomyocytes (Fig. 4B).

Another important question is how specificity is achieved. Our results show that genes containing calcium-responsive exons are associated with more HATs, including CBP and p300, whereas a gene containing a calcium-nonresponsive exon does not show enhanced association with these HATs in response to increased calcium levels (Fig. 5 and Fig. S7). What induces enhanced HAT association is an open question that remains to be investigated. Results of our preliminary experiment suggest that KCl induces differential dissociation of class II HDACs, causing reduced class II HDAC association only from those genes containing calcium-responsive exons. It is possible that chromatin remodelers are mediated by calcium, which removes class II HDACs only from a subset of genes. It is also possible that the rate of nucleosome replacement during transcriptional elongation is differentially regulated on those genes that contain calcium-responsive exons (57).

Gene Expression Regulation by the Calcium-CaMK/PKD-HDAC Pathways.

Our results indicate that the CaMKII and PKD signaling pathways play redundant roles in calcium-mediated splicing regulation, as either one of the two is sufficient to promote the downstream effect: translocation of class II HDACs from the nucleus to the cytoplasm, which perturbs HAT–HDAC balance. The increased association of HATs including CBP and p300 with Nf1 and Ktn1 causes histone hyperacetylation of these genes, enhancing the rate of transcriptional elongation and exon skipping (Fig. 5).

In the cardiovascular system, activation of the calcium-CaMKII-HDAC and calcium-PKD-HDAC pathways has been shown to cause cardiac hypertrophy using genetic mouse models (21, 25, 58). Deletion of one of the class II HDAC kinases is sufficient to rescue pathological cardiac remodeling (21). In this study, CaMKIIδ null mice were viable and displayed no obvious cardiac structural or functional abnormalities. These mutant mice showed reduced phosphorylation of a class II HDAC (HDAC4), suggesting a diminished kinase activity. When challenged with pressure overload hypertrophic stress, these CaMKIIδ null mice were protected against hypertrophy and fibrosis compared with their wild-type littermates (21). Furthermore, in vivo studies using a p300 transgenic mouse model demonstrated a significant role of p300 in the pathology of cardiac hypertrophy (59).

At the molecular level, all of the previous studies have focused on up-regulation of gene expression at the transcription initiation level such as for Myocyte enhancer factor 2 (Mef2), Nuclear factor of activated T cells (Nfat), GATA transcription factors, Cardiac homeobox transcription factors (Csx/Nkx2–5), Atrial natriuretic factor (Anf), and so forth (60–63). Our studies reveal a previously unidentified layer of regulation, supporting a model in which activation of calcium-CaMKII/PKD-HDAC pathways also changes gene expression at the alternative splicing level. Interestingly, the developmental remodeling of the heart is facilitated by both transcriptional changes (64–66) and changes in alternative splicing patterns (35, 67). These changes facilitate adaptation of the developing heart to adult functions. It has been shown that in cardiac hypertrophy and heart failure, the gene expression program often reverts from the adult to fetal patterns (68–70). In our study, we noticed a similar trend for the alternative exons in Ktn1 (exon 36), Ank2 (exon 21), Enah (exon 5), and Mef2A (exon 9) when the calcium level was increased (Fig. 2). It is thus tempting to speculate that such changes contribute to the physiology of the stressed heart, although additional studies need to be conducted to examine the differential functions of specific alternatively spliced isoforms.

Materials and Methods

Cell Culture, Drug Treatment, and siRNA Transfection.

Mouse ventricular cardiomyocytes were prepared from hearts of 1- to 3-d-old newborn CD1 mice (Charles River Laboratories) and cultured as previously described (71) with minor modifications. Following the procedure, we obtained cardiac muscle cells with 90–95% purity, indicated by immunofluorescence staining using anti-cardiac troponin T (cTNT) antibody (Developmental Studies Hybridoma Bank). The Institutional Animal Care and Use Committee at Case Western Reserve University approved these experiments and confirmed that all experiments conformed to the relevant regulatory standards.

Mouse embryonic stem cells (R1) were obtained from the Case Transgenic and Targeting Facility and were cultured as described (29). After two passages, the cells were differentiated into cardiomyocytes using the hanging drop method as previously described (72). The cells started beating approximately on day 7 of differentiation and were used by day 10.

Drug treatment was carried out for 24 h in all cases. In double treatment experiments, KCl/NaCl and drugs were added at the same time. Gö6976 (Sigma) was used at a concentration of 10 μM, whereas KN93 (Sigma) was used at 1 μM. Nifedipine (Fisher), caffeine (Fisher), AP5 (R&D Systems), and BAPTA-AM (Life Technologies) were all added at the concentrations indicated in the figures. Camptothecin (Fisher) was used at a concentration of 3.5 μM.

Primary cardiomyocytes were transfected with scrambled control siRNA or siRNA targeting a specific gene (200 pmol for 5 × 10⁶ cells) using Lipofectamine 2000 (Life Technologies) as described by the manufacturer. The sequences of siRNAs are shown in Table S1.

Semiquantitative RT-PCR, Western Blot, and Immunofluorescence Analyses.

Total RNA extraction and RT-PCR analysis were performed as described previously (73). Low-cycle number PCR (16–20) was performed to determine the relative abundance of individual isoforms of RNA transcripts. Primer information is included in Table S2. Quantification of exon inclusion was determined using a Typhoon Trio Variable Mode Imager (GE Healthcare). The results shown are representative of results from at least three independent experiments. Total protein extraction, Western blotting (73), immunofluorescence (74), and acid extraction of total histones (75) were performed as previously described. Antibody information is included in Table S3.

ChIP and Real-Time PCR.

The ChIP experiments were performed with a ChIP Assay Kit (Millipore; 17-295) using 6 × 10⁷ cells that were fixed with 1% formaldehyde for 30 min at 37 °C as described (42). Immunoprecipitation was carried out overnight at 4 °C with 15 μg of H5 antibody (Covance) or anti–acetyl-histone H3 (Millipore), -CBP (Santa Cruz Biotechnology), or -p300 (Santa Cruz Biotechnology) antibodies. Nonspecific rabbit IgG (Sigma) was used as a control. Cross-linking of bound DNA fragments was then reversed using 5 M NaCl, and DNA was dissolved in 100 μL Tris (10 mM)/EDTA (1 mM) and used as a template for real-time PCR. Real-time PCR was carried out as previously described (42). Primers used in the PCR analyses are described in Table S2. Each primer set has one primer annealing to an intronic region and the other annealing to an exonic region. For data analysis, each sample was quantified in duplicate and at least three independent ChIP analyses were performed. The ChIP signals were normalized to input. P values were calculated by conducting a Student t test comparing values obtained from NaCl and KCl treatments for each indicated region of a specific gene.

Transcriptional Elongation Analysis.

Inhibition and reinitiation of transcription and bromouridine (BrU) labeling were performed as described (41, 42). For this experiment, ES cell-derived cardiomyocytes were treated with 50 μM DRB (Sigma) for 8 h before 2 mM BrU (Sigma) was added into the medium and the cells were cultured for another hour. For samples that were treated with KCl, it was added at the same time as BrU. Following the BrU incubation, cells were collected at 0, 15, 30, 45, 60, 75, and 90 min and total RNA was isolated using TRIzol (Life Technologies). The reverse-transcription and real-time PCR analyses were described previously (42). Primers targeting exon–intron or intron–exon junctions were used. Accumulation of pre-mRNA containing different regions of a gene at each time point was plotted comparing NaCl-treated with KCl-treated cells. The rate of transcriptional elongation was calculated by dividing a specific length of transcript by the time taken for its transcription. For example, the Nf1 transcriptional elongation rate in KCl-treated cells was calculated using exon 40, because it is expressed after 30 min of transcription, but at the same time point exon 45, which is only 3 kb away, is not expressed. The primers targeting exon 40 are 205 kb away from the promoter, which gives the transcriptional elongation rate of 6.8 kb/min. Primers targeting exon 17 at the 30-min time point were used to measure the elongation rate for Nf1 in NaCl-treated cells. To calculate the transcriptional elongation rate for Ktn1, exon 30 at the 15-min time point and exon 13 at the 15-min time point were used for KCl-treated and NaCl-treated cells, respectively. Last, to calculate the transcriptional elongation rate for Capzb, primers targeting Capzb exon 7 at the 45-min time point and exon 5 at the 45-min time point were used for KCl-treated and NaCl-treated cells, respectively.