The protein kinase DYRK1A is distributed throughout the nucleoplasm, accumulating in speckle-like regions. We have found that this punctuated nuclear distribution is determined by the contribution of several elements. Although the nuclear import is mediated by two distinct nuclear localization signals, one at the N-terminus and the other located in the linker region, between subdomains X and XI of the catalytic domain, the accumulation in speckles that are SC35 positive depends on a sequence motif that is located C-terminal to the kinase domain and comprises a histidine tail. A similar sequence is also responsible for the targeting of cyclin T1. Therefore the histidine-rich region represents a novel splicing speckle targeting signal. Moreover, overexpression of DYRK1A induces speckle disassembly. Such disassembly is DYRK1A activity specific, since the overexpression of a DYRK1A kinase inactive mutant, the paralogous DYRK1B or a chimeric protein DYRK1B that has been directed to the speckles via the DYRK1A targeting signal, leaves the SC35 speckle pattern untouched. Thus DYRK1A protein kinase may play a role in regulating the biogenesis of the splicing speckle compartment.
The mammalian cell nucleus is an organelle with a high degree of morphological and functional complexity. The interphase nucleus contains numerous subcompartments, collectively referred to as foci or subnuclear compartments, that have been identified over recent years and are, in general, related to RNA biogenesis, including transcription, processing or transport (reviewed in Matera, 1999). One of the most-studied subnuclear compartments is the nuclear speckle (also known as splicing factor compartment or SFC): 20-50 irregularly shaped dots in which pre-mRNA splicing factors concentrate. Electron microscopy has allowed two distinct structures to be identified within the speckle: a central region termed interchromatin granule clusters (IGCs), which are recruitment domains for splicing factors, and the peripheric regions or perichromatin fibrils (PFs), where it is believed the RNA processing occurs. Speckles are enriched in splicing factors primarily belonging to the serine/arginine-rich (SR) protein family, but transcription factors, snRNAs, ribosomal proteins, SR kinases and phosphatases have also been found in this compartment (reviewed in Matera, 1999). Speckles appear to act as the storage/assembly sites for splicing machinery components and as such may play a role in alternative splicing by modulating the relative concentration of different splicing factors at the processing site. They are highly dynamic structures, and it has been suggested that reversible phosphorylation is the major mechanism of control that regulates distribution of splicing factors within the speckle, either by modulating SR protein shuttling activity or by altering the substrate association specificity of different components of the splicing machinery (reviewed in Misteli, 1999). In fact the phosphorylation of SR splicing factors by different kinases, such as CLKs or SRPKs, causes them to be released from the speckle (Gui et al., 1994; Colwill et al., 1996; Wang et al., 1998; Koizumi et al., 1999).
DYRK1A (Guimera et al., 1996) is a protein kinase, the exact cellular functions of which are still unknown, although may participate in different cellular processes and signal transduction pathways (Yang et al., 2001; Hammerle et al., 2002; Mao et al., 2002), including central nervous system development (Fotaki et al., 2002). DYRK1A belongs to a highly conserved family of protein kinases called DYRKs (dual-specificity tyrosine-regulated kinases or dual-specificity yak-related kinases), with members from yeast to humans (Becker et al., 1998). The lower eukaryotic members of this family, such as Yak1p in Saccharomyces cerevisiae (Garrett et al., 1991), Pom1p in Schizosaccharomyces pombe (Bahler and Pringle, 1998) and YakA in Dictyostelium discoideum (Souza et al., 1998), have been associated with growth control and development. The DYRK1A homologue in Drosophila, minibrain, seems to be involved in postembryonic neurogenesis (Tejedor et al., 1995). The human DYRK1A gene maps to chromosome 21, and it is ubiquitously expressed in adult and fetal tissues, with a high level of expression in the brain (Guimera et al., 1996; Guimera et al., 1999). Its expression pattern in the adult central nervous system (Marti et al., 2003) and in neurogenesis (Hammerle et al., 2002), its overexpression in Down's syndrome fetal brains (Guimera et al., 1999), and the phenotype of DYRK1A transgenic and knock-out mice (Altafaj et al., 2001; Fotaki et al., 2002) make it likely that DYRK1A contributes to some of the neuropathological traits observed in Down's syndrome.
Five members of the DYRK kinase subfamily exist in mammals (DYRK1A, DYRK1B, DYRK2, DYRK3 and DYRK4) that share a high degree of conservation in the catalytic domain, but are very divergent in their N- and C-terminal domains (Becker et al., 1998). Features of DYRK1A are a bipartite nuclear localization signal (NLS) at the N-terminus and a PEST domain, a histidine tail and a region rich in serines and threonines at the C-terminus. The last two domains, both of unknown function, are exclusively present in DYRK1A. It has been defined as a dual specificity kinase because of its ability to phosphorylate serine/threonine and tyrosine residues (Kentrup et al., 1996; Himpel et al., 2001). Substrate specificity studies have identified a consensus phosphorylation sequence (RPX(S/T)P), which defines DYRK1A as a proline-directed kinase (Himpel et al., 2000). A few DYRK1A substrates have been reported over recent years, including cytosolic proteins, such as the ϵ subunit of eukaryotic initiation factor 2B (eIF2Bϵ), the microtubule-associated protein tau (Woods et al., 2001a) and dynamin (Chen-Hwang et al., 2002), and several transcription factors, such as FKHR (Woods et al., 2001b), CREB (Yang et al., 2001) and Gli1 (Mao et al., 2002).
In this study we have analyzed the subcellular localization of DYRK1A, characterizing a second NLS at the C-terminal end of the catalytic domain. We show that DYRK1A accumulates in nuclear speckles or SFCs targeted by the histidine-rich segment present in its C-terminal region, and thus we identify a specific role for this domain of the kinase and define it as a novel speckle-targeting signal that is different from those previously described. Supporting this, we have also identified a signal with similar characteristics in cyclin T1. Furthermore, we show that the overexpression of DYRK1A, but not that of a kinase inactive mutant or its close relative DYRK1B, induces the redistribution of SC35 splicing factor from speckles to the nucleoplasm. Together, these results suggest a new potential role for DYRK1A in RNA synthesis or processing.
Materials and Methods
Monoclonal antibody (HA.11) to the hemagglutinin (HA) epitope tag was purchased from BAbCO (Berkeley, CA). Monoclonal antibody to the SC35 splicing factor was obtained from BD Pharmingen (San Diego, CA), and rabbit polyclonal antibody to PML was from Santa Cruz Biotechnology (Santa Cruz, CA). The Y12 antibody (anti-Sm proteins) was kindly provided by J. Valcárcel (Barcelona, Spain). Horseradish peroxidase (HRP)-conjugated rabbit anti-mouse antibodies were purchased from DAKO (Glastrup, Denmark). Fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse (Southern Biotechnology Associates, Birmingham, AL), Texas Red-labelled sheep anti-mouse (Amersham Biosciences, Uppsala, Sweden) and Cy3-conjugated goat anti-rabbit (Amersham Biosciences) were used as secondary antibodies in immunofluorescence.
The cloning of DYRK1A full-length open reading frame (the 754 amino acids alternative spliced form) as an N-terminal HA-tagged fusion protein has already been described (Marti et al., 2003). HA-DYRK1A deletion mutants were generated either by removing different fragments with restriction enzymes and subsequent ligation or by PCR amplification using specific primers with appropriate restriction sites for subcloning. DYRK1A cDNA was also cloned into a pEGFP-C1 vector (BD Clontech, Palo Alto, CA) to generate an N-terminal GFP fusion protein. All the GFP constructs containing different fragments of DYRK1A were prepared by subcloning PCR products into pEGFP-C1 or by creating stop codons using the Quick-Change kit of Stratagene (La Jolla, CA). The DYRK1A inactive mutant K179R was generated by site-directed mutagenesis.
The pEGFP-DYRK1A (1-176) (Becker et al., 1998) and pEGFP-DYRK1B (Leder et al., 1999) plasmids were kindly provided by W. Becker (Aachen, Germany). To generate the DYRK1B/A chimeric protein, a PstI fragment (amino acids 386-589) in pEGFP-DYRK1B was replaced by a PstI fragment containing the C-terminal region of DYRK1A (amino acids 465-754).
The SUMO-1 open reading frame was amplified from a human brain cDNA library with specific primers and inserted in-frame into the BamHI and XhoI sites of pcDNA-HA1 to generate an N-terminal HA-tagged fusion protein. UBC-9 cDNA was amplified from mouse brain RNA with specific primers and cloned into pRcCMV (Invitrogen, Carlsbad, CA). Human cyclin T1 cDNA fragments (nucleotides 1621-1920 and 1836-1920 in GenBank Acc. No. NM_001240) were amplified from a human brain cDNA library with specific primers and cloned in-frame into the BglII and SalI sites of pEGFP-C1.
All constructs made by PCR, as well as all the in-frame fusions, were verified by DNA sequencing.
Cell culture and transfection
COS-7 and HeLa cells were purchased from the European Collection of Cell Cultures (ECACC) and maintained at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS) and antibiotics. Treatment of COS-7 cells with α-amanitin (50 μg/ml; Sigma, St Louis, MI) was carried out for 5 hours at 37°C. Transient transfections were performed using the calcium phosphate precipitation method.
COS-7 cells were plated in six-well dishes (1×105 cells/well), transfected with 3 μg of plasmid DNA and harvested in ice-cold phosphate-buffered saline (PBS) 48 hours post transfection. Cells were lysed in Laemmli's SDS buffer (250 mM Tris-HCl pH 6.8, 4.6% SDS, 10% 2-mercaptoethanol, 20% glycerol and 20 μM bromophenol blue). Samples were subjected to 10% SDS-PAGE, transferred onto a nitrocellulose membrane (Hybond C, Amersham Biosciences) and blocked with 10% skimmed milk in PBS. The membrane was incubated for 1 hour with the anti-HA antibody (1:2000 in PBS with 0.1% Tween-20 and 5% skimmed milk) and then with an HRP-conjugated rabbit anti-mouse antibody (1:2000 in PBS with 0.1% Tween-20 and 5% skimmed milk) for 45 minutes. Washes were done in PBS-0.1% Tween-20. Detection was performed by enhanced chemiluminiscence (SuperSignal West Pico, Pierce, Rockford, IL).
COS-7 cells growing on coverslips in six-well dishes were transfected with different expression constructs. 48 hours after transfection the cells were washed in PBS, fixed in 4% paraformaldehyde in PBS for 15 minutes and permeabilized in 0.1% Triton X-100 in PBS for 10 minutes. In general the cells were blocked in 10% FCS in PBS, except for the detection of Sm proteins, in which case a blocking solution containing 50% FBS, 6% skimmed milk, 3% BSA and 0.2% Tween-20 was used. Cells were incubated with primary antibodies for 1 to 2 hours and washed extensively with PBS before and after incubation with secondary antibodies for 45 minutes. All incubations were done at room temperature. Anti-HA, anti-PML, anti-SC35 and Y12 antibodies were used at 1:800, 1:100, 1:100 and 1:100 dilutions in PBS-1% FCS, respectively. FITC-conjugated goat anti-mouse, Texas-Red-conjugated sheep anti-mouse and Cy3-conjugated goat anti-rabbit antibodies were used at 1:400, 1:50 and 1:1000 dilutions in PBS-10% FCS, respectively. Coverslips were mounted onto slides using Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA) with 0.2 μg/ml 4,6-diamidino-2-phenylindole (DAPI) and analyzed with an Olympus BX60 microscope with the appropriate filters. Images were captured using a digital camera (Spot RT Colour, Diagnostic Instruments) with SPOT Advanced version 3.2.4. (Diagnostic Instruments) software and processed for presentation with Adobe Photoshop. When indicated in the figure legend, images were acquired in an inverted Leica SP2 Confocal Microscope, using an HCX PL APO 63× 1,32 Oil Ph3 CS objective, and the double images were all taken sequentially. GFP was excited with the 488 nm line of the Argon laser and IgG Texas Red was excited with a 543 nm HeNe laser.
COS-7 cells were plated in 10-cm dishes at 7×105 cells/plate, transfected with 14 μg of the different HA-fusion constructs and harvested in ice-cold PBS 48 hours after transfection.
Cells were lysed in a buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl, 2 mM EDTA, 25 mM sodium fluoride, 30 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 1% Nonidet P-40 (NP-40) and a protease inhibitor cocktail (Roche, Manheim, Germany). Cell extracts were clarified by centrifugation (10,000 g, 15 minutes) and incubated overnight at 4°C with protein-G Sepharose beads (Amersham Biosciences) pre-bound with anti-HA antibody. The beads were washed four times with a buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl and 2 mM EDTA, adding 0.1% NP-40 for the initial washes.
The immunocomplexes were subjected to an in vitro kinase assay using the peptide DYRKtide as exogenous substrate, as described previously (Himpel et al., 2000). Briefly, the immunocomplexes were incubated for 20 minutes at 30°C in 20μ l of phosphorylation buffer containing 200 μM DYRKtide (a kind gift of W. Becker), in the presence of 10 μM [32P]-ATP (5×10-3 μCi/pmol). Phosphorylation of DYRKtide was assayed with the phosphocellulose method and samples were resolved in 10% SDS-PAGE and visualized by autoradiography of the dried gels to detect DYRK1A autophosphorylation. The presence of HA-DYRK1A proteins in the immunocomplexes was detected by western blot.
DYRK1A accumulates in speckle-like regions of the nucleus
DYRK1A protein kinase localizes in the nucleus, probably targeted by a bipartite nuclear localization signal (NLS) present in its N-terminal region, and concentrates in subnuclear dots of unidentified origin (Becker et al., 1998). In order to analyze DYRK1A nuclear and subnuclear accumulation in greater detail, we have generated a series of HA-tagged deletion mutants covering the full-length protein sequence (Fig. 1A). All the mutant proteins were expressed by transient transfections and showed their expected molecular weight, as demonstrated by western blot analysis of the transfected cell extracts (Fig. 1B). When analyzed by immunofluorescence, cells expressing DYRK1A wildtype (1-754) displayed, in addition to a nucleoplasmic diffuse staining with exclusion from the nucleolus, a nuclear speckle-like pattern (Fig. 1C). Deletion of the C-terminal third of the protein, up to the end of the kinase domain, did not affect DYRK1A nuclear accumulation. However, the truncated form of DYRK1A 1-588 displayed a nuclear diffuse staining with no sign of speckle enrichment, whereas a mutant with a shorter deletion (1-616) still localized to discrete foci, indicating that the DYRK1A speckle-like pattern was dependent on a region comprising amino acids 588-616. Surprisingly, a mutant lacking the N-terminal region (DYRK1A 147-754) that harbours the previously described NLS presented a staining pattern indistinguishable from the wild-type protein. Since the size of the truncated protein is above the upper limit for passive transport to the nucleus, an additional NLS should be present in DYRK1A.
Identification of a nuclear localization signal in the catalytic domain of DYRK1A
PSORT II (http://psort.ims.u-tokyo.ac.jp/) identified, apart from the N-terminal NLS (NLS1), a consensus monopartite type NLS, PKRAKFF, at amino acid residues 380-386 (NLS2). As previously described (Becker et al., 1998), a GFP fusion protein comprising the N-terminal region of DYRK1A (NLS1) was found exclusively in the nucleus (Fig. 2, 1-176). When a similar approach was taken for the predicted additional NLS2, the fusion failed to completely accumulate GFP in the nucleus (Fig. 2, 378-386). A more detailed analysis of the primary sequence around the region revealed the presence of another stretch of basic residues between amino acids 397-407 (KKTKDGKREYK), suggesting the possibility that this sequence element was also necessary for nuclear targeting, as a bipartite-like NLS. A GFP fusion protein including both polybasic sequences was also unable to fully accumulate GFP in the nucleus (Fig. 2, 378-407). However, a fusion protein with an extended fragment of the region, defined above, localized exclusively in the nucleus (Fig. 2, 378-466), indicating that the flanking residues of the predicted NLS2 are also determinant for the nuclear enrichment. To confirm that the predicted NLS2 was nevertheless necessary for nuclear targeting, amino acids 380-386 were removed from the last construct, which induced the accumulation of the fusion protein both in the nucleus and the cytosol (Fig. 2, 390-466). These results, together with the deletion analysis, indicate that DYRK1A contains two independent NLSs, each of which individually is sufficient for targeting DYRK1A to the nucleus.
DYRK1A has a novel nuclear speckle-targeting signal
Having defined a region between amino acids 588-616 (Fig. 1) that seems to be required for the targeting of DYRK1A to the nuclear speckles, we wanted to investigate whether this particular domain of DYRK1A was sufficient for localizing a heterologous protein to the speckles. We first generated a GFP fusion protein with an extended fragment of the minimal region defined above (378-616) to include the second NLS identified in case nuclear import was necessary for targeting to the subnuclear compartment. This fusion protein showed the same staining pattern as that seen for the full-length DYRK1A fused to GFP (Fig. 3, compared wildtype and 378-616 images). When the amino acid sequence containing the histidine tail was deleted, accumulation in speckles was no longer detected, although nuclear localization was maintained (Fig. 3, 378-588). A fusion protein, consisting of the histidine repeat alone, clearly accumulated in the speckle compartment (Fig. 3, 590-616), indicating that the presence of this sequence is sufficient for nuclear speckle localization. Furthermore, since this GFP fusion protein does not harbour an NLS, and accordingly is not restricted to the nucleus, the results allow us to suggest that the speckle targeting promoted by this sequence does not depend on receptor-mediated nuclear import. Therefore, a novel signal, both necessary and sufficient for targeting to nuclear speckles, is present in DYRK1A.
Colocalization of DYRK1A with splicing factors
Next, we attempted to determine whether DYRK1A associates with any of the known subnuclear structures by means of colocalization studies of GFP-DYRK1A using reported markers for different subnuclear compartments. On the basis of the type and number of DYRK1A nuclear dots, we concentrated on the splicing factor compartment, PODs, and SUMO nuclear bodies, the last two based on the fact that the protein kinase HIPK2, a close relative of the DYRK family of proteins, localizes to nuclear dots when modified by SUMO (Kim et al., 1999) and accumulates within PML nuclear bodies (Trost et al., 2000).
Immunostaining with an antibody against the promyelocytic leukemia protein (PML), a major component of PML bodies, did not show colocalization (Fig. 4A). Nor were we able to detect any colocalization of DYRK1A with SUMO when exogenously expressed, neither in the absence or presence of its E2 enzyme UBC-9 (Fig. 4B and data not shown). However, a marked colocalization was observed when we used an antibody to the SC35 protein, a non-snRNP splicing factor of the SR family of proteins that is commonly used to define splicing nuclear speckles (Fig. 4C, upper panel). To determine whether DYRK1A associates with components of the spliceosome in addition to SC35, we tested an antibody to Sm proteins (Y12 antibody) that recognizes an epitope found on a number of snRNPs, and colocalization was also detected (Fig. 4C, lower panel). Results with the fusion GFP-DYRK1A 590-616 confirmed that the discrete foci for the novel targeting signal were also SC35 positive (Fig. 4D).
Cyclin T1 has been shown to localize to SFCs, and a region (amino acids 433-533) containing a histidine-rich segment has been defined by deletion analysis as important for its localization in this compartment (Herrmann and Mancini, 2001). Not only the fusion GFP-cyclin T1 433-533 but also a shorter version including only the histidine segment (HKEKHKTHPSN(H)5NHHSHKHSHSQ) clearly colocalized with SC35 staining in the transfected cells (Fig. 4E).
Although the results from above strongly suggest that kinase activity is not required for DYRK1A nuclear speckle localization, we wanted to confirm it by studying the behaviour of a kinase-negative mutant. The mutant (K179R), generated by replacing the conserved lysine in the ATP-binding site with an arginine, was unable to autophosphorylate and failed to phosphorylate a peptide substrate (DYRKtide) (Fig. 5A), a result similar to that obtained with the equivalent mutant (DYRK1A K188R) in the 763 amino acids rat DYRK1A form (Kentrup et al., 1996). This inactive kinase mutant accumulated in nuclear speckles and colocalized with the SC35 splicing factor (Fig. 5B). These results, all together, indicate that DYRK1A concentrates in nuclear speckles with splicing factors and that it does not require its kinase activity for such localization.
Effect of transcriptional inhibition on DYRK1A localization
When cells are treated with transcription inhibitors, splicing activity is reduced and foci associated with RNA processing undergo dynamic changes, in particular, speckles labelled with SC35 become fewer, increase in size and round up (O'Keefe et al., 1994). To determine whether the localization of DYRK1A is dependent on the transcriptional activity of the cell, transfected cells were treated with α-amanitin, an inhibitor of transcription mediated by RNA polymerase II. As shown in Fig. 6A, DYRK1A localized to larger speckles with a complete overlap with SC35. The same effect was observed for the kinase-negative mutant (DYRK1A K179R) and for DYRK1A 590-616 (Fig. 6B,C), although in this case a few GFP-positive foci were not co-stained with SC35. Therefore, DYRK1A accumulation is sensitive to the transcriptional state of the cell, responding as other members of the splicing machinery with regard to such behaviour.
DYRK1A disassembles nuclear speckles
In DYRK1A-expressing cells we repeatedly detected a marked population of cells in which DYRK1A staining appeared to be diffuse. In these cells the endogenous SC35 staining was not concentrated in speckles, but nucleoplasmic staining was still detected (Fig. 7A). Quantification of this phenotype (Fig. 7B) indicates that, in around 40% of the transfected cells, overexpression of DYRK1A wildtype, but not of the kinase-negative mutant, was able to induce the disassembly of nuclear speckles. DYRK1A 590-616 did not affect disassembly, supporting the release of splicing factors from speckles is dependent on DYRK1A kinase activity. Surprisingly, overexpression of a truncated form, lacking the C-terminal region of DYRK1A (DYRK1A 1-522, ΔC) and unable to localize to speckles, also induced the redistribution of endogenous splicing factors at a similar level to DYRK1A wildtype (Fig. 7B). Treatment with α-amanitin had no effect on the percentages of speckle disassembly (Fig. 7B). Therefore, DYRK1A, similarly to CLKs or SRPKs (Gui et al., 1994; Colwill et al., 1996; Duncan et al., 1997; Wang et al., 1998), is able to induce the redistribution of SR proteins from speckles when the active kinase is overexpressed.
DYRK1A has a paralogous gene in humans named DYRK1B. The two proteins share over 80% amino-acid identity in the catalytic region, with the similarity extending towards the N-terminus but not to the C-terminus. As previously reported (Leder et al., 1999), a fusion protein GFP-DYRK1B localized primarily in the nucleus, with no detectable accumulation in nuclear speckles, as was expected because of a lack of the histidine repeat, which is not conserved. Moreover, all cells overexpressing the fusion protein presented a normal SC35 staining pattern (Fig. 7C). In contrast, a chimeric protein (DYRK1B/A) in which the C-terminal regions of both proteins have been exchanged, regained the ability to localize to speckles, but not to cause the release of SR proteins (Fig. 7C). These results suggest that, if speckle disassembly is mediated by the phosphorylation of certain factors present in the nuclear speckles, some substrate specificity, which may be different in DYRK1A and DYRK1B, should be involved in the process.
Protein kinase DYRK1A is a nuclear protein (Marti et al., 2003), which behaves as such when exogenously expressed in cells (Becker et al., 1998) (this study). It had been thought that an N-terminal motif, resembling a bipartite NLS, was responsible for this subcellular localization (Becker et al., 1998). We have found that another NLS is present in DYRK1A, demonstrated by the way that deletion of the described NLS still leaves a protein able to accumulate in the nucleus. The new NLS (NLS2) is similar to a canonical monopartite NLS and is located in the linker region between subdomains X-XI of the catalytic domain, a region very variable in length that it is assumed to be exposed (Bourne et al., 1996). NLS2 is necessary, but not sufficient in itself, to induce the accumulation of a heterologous protein within the nucleus because an adjacent region is required to achieve complete nuclear localization, either contributing to nuclear translocation by enhancing recognition by import receptors or working as a nuclear retention domain.
We have shown that both NLSs may work independently, because they are both capable of targeting a heterologous protein to the nucleus and because their deletion in DYRK1A protein still leaves a protein competent in nuclear localization. The existence of multiple NLSs in a single protein has been associated with a faster nuclear uptake of the protein (Boulikas, 1993). It is also possible that if each NLS is subject to different regulatory pathways, they may contribute to the fine control of DYRK1A concentration within the nucleus.
The nuclear speckles in which DYRK1A accumulates have been identified as the splicing factor compartment or SFC, as indicated by DYRK1A colocalization with SC35 and Sm proteins. We did not detect colocalization with PML or SUMO-1, as described for the DYRK family relative HIPK2 (Kim et al., 1999; Trost et al., 2000), although we cannot exclude the possibility that DYRK1A may be modified by other SUMO family members not tested here. Speckle localization is not cell type specific because it can be readily observed in several mammalian cell lines, including NIH 3T3, HeLa and SH-SY5Y (data not shown). DYRK1A protein levels are very low in all these cell lines, precluding the possibility of studying the endogenous pattern. However, in in vitro cultured mouse cerebellar neurons, endogenous DYRK1A immunostaining, as shown by confocal microscopy, resembles the pattern found when exogenously expressed (Marti et al., 2003), excluding an overexpression-mediated effect.
Very little is known about the protein signals that direct accumulation in specific subnuclear sites. In the case of SFCs a few protein domains have been described: the RS-domain and the RNA-recognition motif for SR proteins and CrkRS (Caceres et al., 1997; Gama-Carvalho et al., 1997; Ko et al., 2001); the `forkhead-associated' domain in the case of the protein phosphatase-1 regulator NIPP1 (Jagiello et al., 2000); the ankyrin repeats in IkappaBL (Semple et al., 2002), the TP-domain in SF3b155/SAP155 (Eilbracht and Schmidt-Zachmann, 2001), and a newly described region in SRm160 (Wagner et al., 2003). According to this, the speckle-targeting signal described here is completely novel. We consider it a bona fide signal, given that it meets the criteria of both being necessary for DYRK1A localization and being sufficient to mediate the speckle localization of a heterologous protein. The fact that a similar type of signal is responsible for cyclin T1 localization to SC35 foci highlights the importance of this novel domain and leads us to suggest that the histidine-rich region represents a novel class of targeting signals to SFCs. At present we do not know whether the novel histidine-rich targeting signal works as an RNA-binding domain or as a protein interacting surface. Our attempts to detect interactions from SR-protein-enriched fractions using the DYRK1A targeting signal as a bait in GST pull-downs have been unsuccessful (data not shown), suggesting that the target may not be any of the major SR proteins. Taking into account that the signal is shared by cyclin T1, a regulator of transcription, it would be possible that the target is not a member of the splicing machinery but of RNA synthesis complexes.
Spliceosome formation is a dynamic process, and there is continuous switching of binding partners, both protein and RNA components, throughout the splicing reaction. Protein phosphorylation seems to significantly contribute to the modulation of both the structural organization of the spliceosomal complex and the catalytic steps of the splicing reaction. In spite of this, not many kinases have been shown to modulate mRNA splicing (Gui et al., 1994; Duncan et al., 1997; Xiao and Manley, 1998; Prasad et al., 1999; Wang et al., 1999). Although some kinases accumulate in SFCs: PITSLREp110 (Loyer et al., 1998), phosphatidylinositol 3-kinase C2α (Didichenko and Thelen, 2001), phosphatidylinositol phosphate kinases (Boronenkov et al., 1998), casein kinase I (Gross et al., 1999), cyclin T-cdk9 (Herrmann and Mancini, 2001), CrkRS (Ko et al., 2001) and hPRP4 (Kojima et al., 2001), only members of the CLK and SRPK families have been described as able to disturb the normal appearance of SFCs when overexpressed (Gui et al., 1994; Colwill et al., 1996; Kuroyanagi et al., 1998; Wang et al., 1998). The results shown here indicate that DYRK1A belongs to the very short list of kinases capable of disassembling nuclear speckles. Disassembly occurs in about 40% of cells overexpressing the protein, and it does not result from the displacement of a speckle component by competition because the GFP fusion containing only the targeting signal does not cause speckle redistribution. Moreover, on the basis of the results of the kinase-defective mutant, it seems that the effect is definitely dependent on having an active kinase that may either phosphorylate spliceosome components or act as a regulatory kinase for another speckle-disassembly kinase. The pattern shown by the DYRK1B protein targeted to the speckles through the C-terminal region of DRYK1A points to a target of phosphorylation being DYRK1A specific. An intriguing finding is that disassembly is not an `all or nothing' effect, because it is detectable only in a population of transfected cells. The percentages found cannot be related to differences in the level of protein expression because we have not been able to correlate high expression level and speckle redistribution. One possibility is that the kinase is active in a defined cell population, for example, during a specific phase of the cell cycle. Preliminary data indicating that DYRK1A-overexpressing cells are arrested in G1 would suggest that speckle reorganization is not related to cells being in the S, G2 or M phases of the cell cycle (S. Aranda and S.L., unpublished). The effect can also be explained if the phosphorylation target is not present in speckles in all cells. Future experiments will be directed towards addressing these questions, identifying DYRK1A partners in the speckles and discovering whether DYRK1A kinase activity is able to regulate splicing by modulating the biogenesis of nuclear speckles.
We are indebted to Walter Becker (Institute für Pharmakologie und Toxikologie, Aachen) for plasmids and for his help with the kinase assays. We are likewise grateful to Juan Valcárcel for critical reading of the manuscript and Daniel Bilbao (Centre de Regulació Genòmica, Barcelona) for stimulating discussions; Arrate Mallabiabarrena (Universitat Pompeu i Fabra, Barcelona) for her help with confocal microscopy; Alicia Raya for her technical support, and all other lab members for their support. The study was supported by grants from the Ministerio de Ciencia y Tecnología (BMC2001-1580) to S.L. and European Comission (QLG1-CT-2002-00816) to X.E. M.A. is a BEFI fellow.
- Accepted April 15, 2003.
- © The Company of Biologists Limited 2003