In this study, we identified PSF as a Fox-3 interacting protein and demonstrated that this interaction is essential for Fox-3 to activate neural cell-specific alternative splicing of the N30 exon of NMHC II-B. We also compared the expression pattern of Fox-3 to those of Fox-1 and 2 at the cellular level in the mouse central nervous system and found a correlation between Fox-3 expression and N30 splicing.
Two laboratories have reported the tissue distribution and subcellular localization of Fox-1 in the mammalian nervous system (24
). Although they agreed that Fox-1 is expressed in neuronal cells, there is a discrepancy as to whether it localizes to the nucleus or cytoplasm. Our observations largely agree with that of Black and colleagues that Fox-1 localizes predominantly to nuclei. However, we also detected Fox-1 in the somas and nuclei of motor neurons in the spinal cord. Discrepancies in histological observations might be due in part to differences in the age and species of the animals, differences in methods for tissue fixation and antigen retrieval and differences in the epitopes recognized by the Abs. Of interest, it has been reported that the depolarization-induced change in the Fox-1 isoform results in a change in the nucleo-cytoplasmic ratio of Fox-1 (32
). We have also previously reported that different isoforms of Fox-1 show different subcellular localizations when they are exogenously expressed in cultured cells (23
). Therefore, it may be necessary to determine which isoform of Fox-1 is dominant in each particular system.
Our co-immunostaining analysis and FACS followed by immunoblotting demonstrated that some neuronal cells expressing Fox-1 or 2 do not express Fox-3. Of note is that the expression levels of Fox-3, but not Fox-1 or 2, correlate with the N30 splicing patterns in the cerebellum, brain stem and spinal cord. Most of the previous studies on brain specific alternative splicing have used whole brain or anatomically dissectible parts of the brain. We made use of FACS to separate cells according to the expression levels of Fox-3. A combination of cell sorting and biochemical analysis led us to find different splicing patterns between the Fox-3 positive and negative cell populations. In spite of our previous observation that exogenous expression of Fox-1 or 2 enhances N30 inclusion in cultured cells (23
), brain cells which express Fox-1 and/or Fox-2 but not Fox-3 do not activate N30 splicing. This could be explained in part by the following. Since the average expression level of Fox-1 or 2 is similar between Fox-3 positive and negative cells in the cerebellum, brain stem and spinal cord, additional expression of Fox-3 simply increases the total amounts of all three Fox proteins. Another reason might be related to the isoforms of Fox-1 and 2. Fox-1 and 2 genes generate multiple alternatively spliced isoforms. Differences in amino acid sequences at the C-terminal region result in significant differences in the splicing activity toward N30 (23
). Currently our study does not distinguish differences in the C-terminal amino acids, since anti-Fox-1 was generated against the N-terminal region common to Fox-1 isoforms and the same is true for anti-Fox-2. Therefore, cells negative for Fox-3 might express Fox-1 and/or Fox-2 isoforms with lower splicing activities. A third reason might be a difference in affinity for interacting protein(s). A number of proteins including ataxin-1 and 2, atrophin-1, quaking, Fyn tyrosine kinase and estrogen receptor-α have been reported to interact with Fox-1 or Fox-2 in mammals (29
). However, most of these proteins have not been functionally characterized in the context of pre-mRNA splicing. Of particular interest is a report showing the interaction of an U1 snRNP component, U1C protein, with Fox-1 and 2 in yeast two-hybrid screening (47
), although the functional outcome of this interaction has not been studied. In this study, we identified PSF as an essential interacting protein of Fox-3 for activation of N30 splicing. Even though we compared the isoforms of Fox-1 and 2 which show the highest sequence similarity to Fox-3 and the highest activity for N30 splicing among the known isoforms, PSF binds to Fox-3 more efficiently compared to Fox-1 and 2. Therefore this could explain why Fox-3 is more active in N30 inclusion. This affinity difference also suggests that Fox-3 might play a role in the determination of neural specificity of alternative splicing. However differences in affinity of PSF for Fox proteins still need to be studied in greater detail biochemically and evaluated more precisely in the cellular context.
Although we found a good correlation between the level of Fox-3 expression and the extent of N30 splicing, this study does not exclude a possible contribution of Fox-1 and 2 to N30 splicing. Fox-1 and 2 have been reported to interact with each other using a yeast two-hybrid system (46
). Our lab also detected the interaction of Fox-2 with Fox-3 using the yeast two-hybrid system as well as the interaction of Fox-3 with Fox-1 and with Fox-2 by co-immunoprecipitation (unpublished observations). In fact, we rarely observed neuronal cells which express only Fox-3. It was unexpected to find that Fox-1 localized predominantly to nuclei in intact brain, since all Fox-1 isoforms which we have analyzed were diffusely distributed in both the nuclei and cytoplasm or predominantly localized to the cytoplasm when they were exogenously expressed in cultured cells (23
). In contrast, exogenously expressed Fox-3 isoforms localize almost exclusively to nuclei in cultured cells (Supplementary Figure S1D
). The Fox-2 isoforms which we have analyzed localize predominantly to nuclei (23
). These observations raise the possibility that Fox-1 might dimerize with Fox-3 or Fox-2 and localize to nuclei in vivo
. The Fox proteins might function as heterodimers or heteromultimers, especially when the pre-mRNA contains multiple UGCAUG elements. Whether there is a difference in splicing activity among heterodimers and homodimers still needs to be determined.
Fox proteins can function as activators or repressors depending on their binding location on pre-mRNAs relative to the regulated exons. Recent genome-wide studies together with earlier studies using model systems have proposed a general rule for Fox proteins to influence the choice of exons (8
). When Fox proteins bind to the intron downstream of the alternative exon, exon inclusion occurs. On the other hand, when Fox proteins bind to the intron upstream of the alternative exon, exon skipping occurs. Recently a few reports have begun to address the mechanism by which Fox proteins repress or activate the usage of alternative exons. Using F1γ pre-mRNA, Fox-1 which is recruited to the upstream intron has been shown to block formation of the early pre-spliceosome complex on the intron downstream of the regulated exon (48
). In the case of calcitonin/CGRP pre-mRNA, Fox-2 which is bound to the upstream intron inhibits the recruitment of SF1 at the branch site, and further, Fox-2 which is bound to the alternative exon inhibits the recruitment of U2AF at the 3′ splice site (27
). The interaction of Fox proteins (FOX-1 and ASD-1) with another sequence-specific RNA-binding protein SUP-12, which is expressed specifically in muscle, has been reported in C. elegans
. The FOX-1 (or ASD-1) and SUP-12 interaction enhances their binding to their adjacent target RNA elements on the egl-15 pre-mRNA and leads to inclusion of a muscle specific mutually exclusive exon. This report provides a mechanism for the strict tissue specificity of Fox-regulated alternative splicing (49
In this study we identified PSF and the PSF–NonO complex as proteins interacting with Fox-3. The human ortholog of NonO is often called p54nrb. PSF and NonO are structurally related proteins that contain two RRMs and can bind to RNA as well as DNA (40
). PSF and NonO interact with each other to form a heterodimer and the participation of PSF or the PSF–NonO complex has been reported in many aspects of nuclear function including transcription, transcriptional termination, pre-mRNA splicing, 3′ end processing of mRNA and RNA retention (40
). PSF and NonO are expressed in various tissues and cell types. We demonstrated that the C-terminal region of Fox-3 binds directly to the N-terminal region of PSF, and Fox-3 and NonO interact indirectly via PSF. PSF enhances the binding of Fox-3 to the target UGCAUG element in an in vitro
crosslinking assay. Moreover the presence of PSF enhances the recruitment of Fox-3 to the IDDE of the NMHC II-B transcript, which contains the UGCAUG elements, in intact cells. The effect of PSF on Fox-3 binding to the target RNA element in intact cells shows an apparently greater degree of enhancement (>30-fold) than in vitro
(2- to 3-fold). This difference could be due to the different efficiencies of the Fox-3 and PSF complex formation between in vitro
and intact cells. Another possibility is that PSF and Fox-3 might be co-transcriptionally recruited to the Fox-3 target sites of pre-mRNAs. PSF and NonO have been reported to be associated with phosphorylated RNA polymerase II in a large complex containing transcriptional elongation factors (52
). If Fox-3 was included in that complex via PSF, Fox-3 could be more efficiently recruited to target sites of pre-mRNAs during transcriptional elongation, compared to simple diffusion in the nucleus.
Although we showed that Fox-3 binding to the target RNA element was enhanced by PSF, the role of PSF in Fox-3-dependent activation of alternative splicing may not be limited to this effect. PSF was originally found as a spliceosome associated protein (54
). A number of studies have demonstrated the presence of PSF in the pre-spliceosome and in the spliceosome at different stages and in complexes containing snRNPs (55
). PSF has also been reported to interact directly with U5 snRNA and with the 5′ splice site under splicing conditions (42
). Therefore PSF may function as a mediator between Fox-3-bound pre-mRNA and the splicing machinery. This notion is supported by our observation that PSF does not bind directly to the IDDE or to the pre-mRNA from the NMHC II-B minigene in the absence of other nuclear factors. Additionally, both Fox-3 (or other Fox proteins) and PSF are essential for the UGCAUG-dependent activation of N30 splicing. In the presence of endogenous PSF and Fox-2, exogenous Fox-3 activates N30 splicing to some extent as does exogenous PSF. These effects of exogenous Fox-3 and PSF are apparently additive. When endogenous Fox-2 or PSF is eliminated, however, neither exogenous Fox-3 nor PSF can activate N30 splicing. The enhancing effect of PSF on N30 splicing is dependent on the UGCAUG element, although it does not bind to this element, and is absolutely dependent on the presence of Fox-3 (or other Fox proteins). The enhancing effect of Fox-3 on N30 splicing is also absolutely dependent on the presence of PSF. Thus, the effects of the two proteins are actually not additive, but rather cooperative. PSF seems to function as a coactivator or mediator of Fox-3 during the splicing process. The simplest model to accommodate all the data is that PSF or the PSF-containing complex bridges Fox-3 and the splicing machinery. Although the detailed molecular mechanism of N30 exon recognition following the Fox-3 and PSF interaction remains to be elucidated, this interaction has now been shown to be an integral part of the mechanism responsible for Fox protein regulated activation of alternative exon inclusion via a downstream intronic enhancer.