Abstract
Alternative splicing represents a mechanism by which a single gene can be used to create proteins with different functions. Neurons use alternative splicing to produce channels with different sequences and biophysical or regulatory properties. O'Donovan and Darnell discuss a mechanism by which neurons can alter channel splicing in response to neuronal activity through a signal generated by calcium and calcium/calmodulin-dependent kinase activity.
Alternative splicing of pre-messenger RNA (pre-mRNA) molecules endows neurons with the ability to generate a staggering variety of mature messenger RNA (mRNA) transcripts from a single pre-mRNA. Over the past 20 years, researchers have found that almost all neurotransmitter receptor and ion channel pre-mRNAs undergo extensive alternative splicing to generate multiple isoforms. Although it is believed that the inclusion or exclusion of a particular exon can regulate important variations in protein function, there has been little insight into how neurons regulate this process. Recent studies have demonstrated that the neuron-specific RNA binding protein Nova is able to regulate alternative splicing of the inhibitory glycine and γ-aminobutyric acid (GABA)A receptors in vivo (1). Xie and Black have now taken the field to new ground (2), correlating depolarization and calcium/calmodulin-dependent kinase (CaMK) activity with regulation of alternative splicing, a link they make through an inhibitory RNA element termed the CaMKIV-responsive RNA element (CaRRE).
In a number of pre-mRNAs, predominantly those encoding cellular receptors and channels, specific functions have been clearly ascribed to the protein products of alternative splicing. The N-methyl-D-aspartate (NMDA)-R1 (NR1) receptor pre-mRNA undergoes alternative splicing to generate specific isoforms that affect subcellular localization (3), protein-protein interactions (4), and zinc-sensitivity (5). The vertebrate Slo (also referred to as Slowpoke or BK) genes, which encode calcium- and voltage-gated potassium channels that are expressed widely in the nervous system (6-8), also undergo extensive alternative splicing. These large conductance potassium channels set the resting membrane potential of neurons and endocrine cells by regulating both repolarization and after-hyperpolarization following an action potential (9, 10). Changes in the concentrations of stress hormones regulate alternative splicing of the so-called STREX exons (for stress axis-regulated exon) in the rat Slo pre-mRNA in vivo (11). When expressed in Xenopus oocytes, STREX-containing Slo channel variants exhibit enhanced channel activation, as well as slowed deactivation. Phosphorylation of the Slo channel by adenosine 3′,5′-monophosphate (cAMP)-dependent protein kinase (PKA), which physically interacts with the Drosophila version of the channel (12) and inhibits current flow through STREX-containing channels, whereas it activates Slo channels lacking the STREX exon (13). However, up to the present, nothing was known about the mechanisms underlying the regulation of STREX exon splicing other than the observation that gross changes in hormone levels could modulate STREX exon inclusion.
Xie and Black took advantage of a pituitary tumor-derived cell line, GH3 cells, which expresses Slo channels with and without the STREX exon cassette. To test whether perturbations in cellular signaling pathways might modulate alternative splicing of the STREX exon, the authors used a standard method of cellular depolarization with high (25 mM) extracellular potassium. Six hours after depolarization of the GH3 cells, they observed a 50% decrease in endogenous STREX exon inclusion. This effect was abrogated by pretreatment of the cells with KN93, a nonspecific CaMK inhibitor, and partially abrogated by pretreatment of the cells with an inactive analogue, KN92. Thus, these data provide evidence for dynamic regulation, in part mediated through a CaMK pathway, of Slo pre-mRNA splicing in a pituitary cell line. Most of the rest of the paper is devoted to characterizing the RNA elements required for the depolarization-induced changes.
A hybrid "minigene" was constructed containing the STREX exon and some of its flanking intronic sequences cloned between two constitutively spliced β-globin exons. When this construct was transfected into the transformed human embryonic kidney cell line HEK-293, overexpression of CaMKIV partially suppressed STREX exon inclusion. In contrast, overexpression of CaMKI, CaMKII, or a catalytically inactive CaMKIV mutant had no effect on STREX exon splicing. These experiments clearly demonstrate that CaMKIV mediates repression of STREX exon inclusion on the hybrid minigene in HEK-293 cells. This conclusion cannot be drawn for the endogenous STREX exon in GH3 cells, because CaMKI, CaMKII, and CaMKIV are all expressed in GH3 cells, activated by depolarization, and inhibited by KN93. It will be of interest to know whether overexpression of CaMKIV in GH3 cells can similarly decrease endogenous STREX exon inclusion. Notably, CaMKIV is predominantly localized to the nucleus, suggesting the possibility that it may directly phosphorylate proteins involved in STREX splicing.
The authors also determined which exonic and intronic sequences mediate the CaMKIV repression of STREX exon inclusion. The STREX exon itself contains what appear to be two purine-rich enhancer elements, as well as a longer pyrimidine-rich repressor element. Deletion of these sequences from the minigene construct described above had only a slight effect on splicing in untreated HEK-293 cells, whereas CaMKIV-mediated repression was completely abolished upon deletion of the pyrimidine-rich exonic sequence. However, further analysis revealed that intronic sequences, including the upstream 3′ splice site of the STREX exon, also contribute to STREX splicing regulation. Thus, as is often the case, multiple intronic and exonic elements contribute to the overall regulation of splicing. For the minigene experiments, the authors include a number of important controls for relative differences in transcription. Nonetheless, in addition to its newfound effects on splicing, CaMKIV phosphorylates and activates the cAMP response-element-binding protein (CREB) and serum-response factor (SRF) transcription factors [reviewed in (14)]. Thus, CaMKIV could both activate the transcription and regulate the splicing of a target pre-mRNA.
Overexpression of CaMKIV may alter the relative stability of mRNAs that contain or that lack the STREX exon. One way to address this issue is to determine whether the upstream 3′ splice site of the STREX exon alone is sufficient to confer CaMKIV responsiveness, because intronic elements might be expected to have their primary effect on pre-mRNA metabolism. The authors cloned a 53-nucleotide (nt) intronic fragment (plus 1 nt of contiguous exonic sequence) of the STREX pre-mRNA upstream of a constitutively spliced β-globin exon. Overexpression of CaMKIV, but not a catalytically inactive mutant, resulted in robust repression of β-globin exon inclusion. Moreover, point mutations within the polypyrimidine upstream of the STREX 3′ splice site abrogated the CaMKIV-mediated repression. Based on these observations, the authors dub this short nucleotide stretch the CaMKIV-responsive RNA element (CaRRE). The fact that such seemingly subtle mutations abolish CaMKIV-mediated repression suggests that sequence-specific binding of some factor, presumably a direct or indirect target of CaMKIV, is involved in this process. The authors note that they have been unable to detect CaMKIV-mediated phosphorylation of the known splicing repressor polypyrimidine-tract-binding protein (PTB) (15, 16), although they do not address whether a recently described brain-enriched PTB protein (17, 18) may play a role here.
Returning to GH3 cells, Xie and Black tested whether the CaRRE can confer depolarization-induced repression of β-globin exon inclusion. Whereas untreated cells exhibit ~57% β-globin exon inclusion, depolarization reduced exon inclusion to ~34%. Making the same intronic polypyrimidine tract mutations that abolished CaMKIV responsiveness, the authors found that repression of exon inclusion was completely abolished in both depolarized and untreated cells. Because the CaRRE confers exon exclusion in untreated GH3 cells (~80% of the STREX exon is excluded in untreated cells and ~43% is excluded in the context of the β-globin minigene), one wonders whether GH3 cells have relatively high levels of CaMK activity. This question could be addressed by examining whether KN93, the nonspecific CaMK inhibitor, represses exon inclusion under basal conditions, in addition to its action following depolarization (Fig. 1). Also, in this set of experiments, GH3 cells were treated with high potassium for 48 hours. Thus, the observed CaRRE-mediated changes in splicing might occur on a more rapid time scale in response to depolarization, as was observed for endogenous STREX exon inclusion in GH3 cells in other experiments.
CaRRE-mediated suppression of exon inclusion. After depolarization, there is an increase in intracellular Ca2+ and activation of CaMK, which can be inhibited by exogenous application of the nonspecific CaMK inhibitor KN93. Acting through the novel CaRRE RNA element, CaMK leads to a reduction in the inclusion of the Slo gene's STREX exon.
To investigate whether overexpression of CaMKIV can regulate the pre-mRNA splicing of other known ion channel alternative exons, the authors tested another alternate Slo gene exon (SloII87), as well as two NR1 glutamate receptor exons, exons 5 and 21. They subcloned each exon along with some of its flanking intronic sequence into the previously employed β-globin minigene construct. Remarkably, in each case, CaMKIV led to a repression of exon inclusion. In addition, the 3′ splice site of the NR1 exon 5 conferred CaMKIV responsiveness to a heterologous exon. Although these data are tantalizing, it is still not known whether the observed CaMKIV-dependent splicing of SloII87, NR1 exon 5, and NR1 exon 21 is mediated by the CaRRE. Along these lines, it will be of interest to know how prevalent the CaRRE element is in the genome and whether CaMK can mediate its effects on splicing via other RNA elements.
These data have important implications for the regulation of neuron-specific alternative splicing. It now appears that there may be dynamic regulation of alternative splicing of ion channel pre-mRNAs in neurons. Such processes acting through CaMK or the CaRRE or both would almost certainly impact upon the firing properties of neurons, as has been shown previously for Slo potassium channels in adrenal chromaffin cells (10, 19). Moreover, CaMKIV specifically suppresses exon inclusion of the NR1 glutamate receptor subunit, which is a critical component of the NMDA receptor complex. The activation of the NR1-containing NMDA receptor is required for long-term synaptic changes at many excitatory synapses. Therefore, CaMKIV, in addition to its known effects on transcription, might regulate glutamate receptor splicing under such conditions and thereby influence processes like long-term potentiation (LTP) and long-term depression (LTD) (Fig. 2). Because some forms of LTP can last from several hours to days, one might hypothesize that if CaMK were to regulate alternative splicing of ion channel pre-mRNAs in an activity-dependent manner in neurons, it may affect the stability or maintenance of LTP, rather than its induction. Interestingly, late phase LTD in cultured cerebellar Purkinje neurons requires both CaMKIV activation and CREB-dependent transcription (20).
After synaptic activity, there is an increase in dendritic Ca2+ and activation of dendritically localized CaMK. The CaMK signal is propagated to the nucleus where, presumably, CaMK exerts its effects on splicing of receptor and channel pre-mRNAs. Over the long term, the constitution of the receptors and channels at the activated (or depressed) synapse will change as a result of signaling through the CaRRE controlling alternative splicing (shown as altered cytoplasmic domains in the proteins).
A remaining and perplexing question is how timing and specificity within a neuron can be achieved if a particular CaMK-regulated and recently spliced mRNA is to be properly targeted to an activated synapse some distance from the nucleus. Although there is evidence for localization of specific mRNAs to activated synapses (21), is it feasible that an activity-induced and spliced mRNA could stochastically traffic out to the correct activated synapse, undergo translation, and affect synaptic function in a timely manner? Is it possible that unspliced pre-mRNA is targeted to synapses to facilitate local regulation of alternative splicing and translation? In either case, the accumulating evidence indicates that alternative splicing represents a key regulatory step in neuronal signaling.
