SF3b1 is a high affinity in vitro substrate of DYRK1A
We have recently identified SF3b1 as an in vitro substrate of DYRK1A by screening of a cDNA expression library from human fetal brain . In order to further characterise SF3b1 as a substrate of DYRK1A, we performed a kinetic analysis of the phosphorylation of His6-SF3b1304–493, the fusion protein produced from the library clone, by GST-DYRK1A-ΔC. The C-terminally deleted mutant of GST-DYRK1A was used for in vitro-kinase assays since this construct exhibits the same substrate specificity but is more active than wild type GST-DYRK1A [15,16]. The Km value obtained for total phosphate incorporation into the substrate was 2.16 +/- 1.72 μM (mean of three independent experiments +/- S.E.M.), characterising SF3b1 as a high affinity substrate of DYRK1A. A representative experiment is shown below in Fig. 1A. Notably, His6-SF3b1304–493 contains 14 Thr-Pro dipeptide motifs (Fig. 1B) which are potential target sites for DYRK1A.
Phosphorylation of SF3b1 by DYRK1A in COS-7 cells
In order to assess whether DYRK1A phosphorylates SF3b1 in vivo we co-transfected COS-7 cells with expression plasmids for GFP-SF3b1-NT and GFP-DYRK1A. Assuming that DYRK1A may phosphorylate one ore more of the Thr-Pro dipeptides, we took advantage of a commercially available antibody which recognises phosphothreonine C-terminally flanked by proline (pTP) to detect phosphorylation of SF3b1. This antibody detected two bands in the immunoprecipitates from cells overexpressing GFP-SF3b1-NT, of which the lower one (apparent molecular weight of about 95 kD) also reacted with the GFP-specific antibody (Fig. 2A). The difference from the calculated molecular weight (82.4 kD) is possibly due to post-translational modifications. As shown in Fig. 2C, both bands were eliminated by treatment with alkaline phosphatase. Furthermore, the upper band was also found after immunoprecipitation with an SF3b1-specific antibody, but not in untransfected cells (Fig. 2D). Thus, this band most likely represents a highly phosphorylated form of GFP-SF3b1-NT which is present in too low amounts to be detected by the GFP-specific antibody (see also below, Fig. 7B). Co-transfection of DYRK1A caused a very pronounced and dose-dependent increase in the phosphorylation of SF3b1-NT (95kD-band) (Fig. 2A), strongly suggesting that DYRK1A phosphorylates SF3b1 in COS-7 cells. This effect required low amounts of GFP-DYRK1A compared with its substrate GFP-SF3b1-NT, as evidenced by the direct comparison of GFP-immunoreactivity (second lane in Fig. 2A).
To test the specificity of this reaction, we compared the effects of GFP-CLK3 and GFP-DYRK1A on the phosphorylation of SF3b1-NT. Protein kinases of the CLK family are related with the DYRK family and also phosphorylate splicing factors . As a further control, we used GFP-DYRK1A-K188R which carries a point mutation in the ATP binding site and exhibits greatly reduced catalytic activity (1–3% of residual activity [16,18]). As shown in Fig. 2B, co-expression of GFP-CLK3 failed to induce phosphorylation of SF3b1 as compared to GFP alone. As shown by immunodetection with the GFP-specific antibody, GFP-CLK3 was expressed at similar levels as wild type GFP-DYRK1A (see also Fig. 7B). Immunocomplex kinase assays with myelin basic protein as substrate confirmed that GFP-CLK3 was an active protein kinase when expressed in COS-7 cells (data not shown). Unexpectedly, co-expression of DYRK1A-K188R significantly enhanced phosphorylation of SF3b1-NT, although the effect was much weaker than that of the wild type kinase (note also that in the experiment shown DYRK1A-K188R was expressed at a higher level than wild type DYRK1A). The result that a mutant of DYRK1A with reduced activity (K188R), but not the related kinase CLK3, enhanced threonine phosphorylation of SF3b1-NT is evidence of the specificity of this reaction.
Comparison of the phosphorylation of SF3b1 by DYRK1A and cyclin E/CDK2
SF3b1 is phosphorylated concomitant with or just after catalytic step one of the splicing reaction . The kinase responsible for this phosphorylation during splicing catalysis has not been characterised to date, but Seghezzi et al.  have identified SF3b1 as a potential target of cyclin E/CDK2 complexes. In order to compare the phosphorylation of SF3b1-NT-His6 by DYRK1A and cyclin E/CDK2, we performed a preliminary kinetic analysis of both reactions by measuring the velocities of phosphate incorporation at two different substrate concentrations (0.7 and 7 μM). The approximate Km values calculated from the results shown in Fig. 3 are very similar for both kinases (2.75 μM for DYRK1A and 3.51 μM for cyclin E/CDK2) and indicate that both kinases have a high affinity for SF3b1.
To answer the question whether both kinases target the same phosphorylation site(s) in SF3b1, we generated phosphopeptide fingerprints. SF3b1-NT-His6 was phosphorylated by either GST-DYRK1A-ΔC or cyclin E/CDK2 in vitro, and tryptic peptides of SF3b1-NT were analysed by two-dimensional peptide mapping. The pattern of phosphopeptides derived from DYRK1A-labelled SF3b1 (Fig. 4A) differed completely from the pattern obtained by cyclin E/CDK2 (Fig. 4B), and mixing of the peptides from both experiments revealed no detectable comigration of phosphopeptides (A+B). This result indicates that both kinases phosphorylate different sites in SF3b1.
DYRK1A phosphorylates SF3b1 in vitro on physiologically relevant sites
Next we asked whether the phosphorylation pattern of SF3b1 in vivo better matches the in vitro-pattern obtained with DYRK1A or with cyclin E/CDK2. COS-7 cells were transfected with pEGFP-SF3b1-NT and metabolically labelled by incubation with 32P-orthophosphate. Phosphopeptide mapping of the immunoprecipitated GFP-SF3b1-NT fusion protein showed that the in vivo-phosphorylation pattern (Fig. 4C) strikingly resembled the phosphorylation pattern obtained by in vitro-phosphorylation with DYRK1A (Fig. 4A). Six of the spots on the in vivo-map matched phosphopeptides generated by DYRK1A in vitro and comigrated in the map of a mixed sample (Fig. 4A+C), strongly suggesting that the phosphopeptides generated in vitro by DYRK1A are identical with those generated in vivo. Unlike in vitro, however, spot 1 was much more intense than spot 2. A possible explanation for this difference is a superposition of signals derived from spot 1 and a comigrating phosphopeptide (spot X) that is phosphorylated in vivo by a kinase other than DYRK1A (see below, Fig. 5B). No match was detectable between the cyclin E/CDK2 phosphopeptide map and the in vivo map (Fig. 4B+C). This result provides evidence that the major part of the phosphorylation within the Thr-Pro-rich domain of SF3b1 is catalysed by DYRK1A or a related kinase with similar substrate specificity. However, it cannot be excluded that relevant CDK2 sites escaped detection because phosphopeptides were lost during purification or were poorly soluble in the running buffers.
Identification of SF3b1 phosphorylation sites
His6-SF3b1304–493 was phosphorylated with GST-DYRK1A in vitro and tryptic peptides were analysed for phosphorylation by tandem mass spectrometry (MS2). Two phosphorylated peptides were identified: (1) VLPPPAGYVPIRTPAR, containing Thr426 (underlined) as the phosphoamino acid; the phosphorylated residue completely inhibited tryptic cleavage at the preceding Arg whereas in an unphosphorylated sample, this cleavage occurred freely. (2) KLTATPTPLGGMTGFHMQTEDR (with both Met residues in the sulphoxide form – a side reaction of preparation). MS2 and MS3 analysis of this peptide indicated that the phosphorylation was confined to either of the first two threonines of the peptide (Thr432 or Thr434) but from the data the labelled residue could not be distinguished. This was because the predicted fragment ions needed to resolve this question laid beyond the dynamic range of the ion-trap instrument. Attempts to examine smaller MS2 ions further by MS3 to gain access to this region were unproductive, as were secondary digest attempts with chymotrypsin. We considered Thr434 the more likely target because DYRK1A is a proline-directed kinase. Therefore we prepared alanine mutants of Thr426 and Thr434 by site directed mutagenesis of SF3b1-NT. In addition, Thr273 and Thr303 were mutated because the surrounding sequences of both threonines (Thr273: GRGDT273P; Thr303: TERDT303P) matched known target sequences of DYRK1A (RXXS/TP; RXS/TP).
The mutant proteins were phosphorylated with GST-DYRK1A-ΔC in vitro and analysed by peptide mapping. Mutation of Thr434 resulted in the loss of the two most prominent spots (spots 1 and 2; right panels of Fig 5A), indicating that Thr434 is the major phosphorylation site for DYRK1A. The existence of two different phosphopeptides containing Thr434 can be explained by incomplete tryptic cleavage as the MS analysis showed this peptide to exist with and without the lysine at the N-terminus. Such ragged N- or C-termini can be expected when an XRKX sequence is cleaved by trypsin. The absence of one spot in the phosphopeptide map of the T273A mutant identified Thr273 as one of the minor in-vitro phosphorylation sites of DYRK1A (Fig. 5A, arrow in the left panel). The mutants T426A and T303A yielded the same pattern of spots as the wild type protein (data not shown). The failure to detect the VLPPPAGYVPIRTPAR phosphopeptide containing Thr426, which was identified as a phosphorylated residue by MS, may be due to poor solubility of this peptide under the conditions applied.
Next we asked whether Thr273 and Thr434 are in vivo phosphorylation sites of SF3b1. The respective point mutants of GFP-SF3b1-NT were metabolically labelled in COS-7 cells and subjected to phosphopeptide mapping. Analysis of SF3b1-NT-T273A did not reveal differences between the wild type and the mutated protein (data not shown). In contrast, one of the major phosphopeptides (spot 2) was absent in the map of GFP-SF3b1-NT-T434A as compared to the wild type protein (Fig. 5B). This result confirms our conclusion that spot 2 represents the same phosphopeptide in the in vitro and the in vivo-maps (Fig. 4). We assume that the absence of the other peptide (spot 1) is masked by a comigrating phosphopeptide (spot X). Spot X is lacking in SF3b1-NT-T434A after phosphorylation by DYRK1A in vitro, hence this phosphopeptide appears to harbour the only major phosphorylation site not recognised by DYRK1A. These data indicate that Thr434 in SF3b1 is phosphorylated by endogenous kinases in COS-7 cells.
Overexpression of DYRK1A increases phosphorylation of SF3b1 at in vivo-phosphorylation sites
As shown in Fig. 2A, overexpression of DYRK1A increases the phosphorylation of SF3b1 in COS-7 cells. To investigate whether DYRK1A targets the same sites that are already phosphorylated in vivo, we compared the phosphopeptide map of GFP-SF3b1-NT phosphorylated by endogenous kinases in COS-7 cells with the phosphopeptides obtained after cotransfection of GFP-DYRK1A. As shown in Fig. 6, intensities of at least five peptides (spots 2–6) increased upon coexpression of GFP-DYRK1A relative to spot X/1. It should be noted that the comparison with spot X/1, which includes the DYRK1A-phosphorylated spot 1, underestimates the degree of the increase caused by cotransfection of DYRK1A. In addition, three new spots appeared that were not detectable when SF3b1 was labelled without coexpression of DYRK1A (arrows). This result demonstrates that DYRK1A can phosphorylate other residues in addition to Thr434 that are endogenous phosphorylation sites.
Phosphorylation of Thr434 in endogenous SF3b1
In order to facilitate detection of phosphorylated Thr434, we raised a polyclonal antiserum against a peptide comprising residues 429–439 of SF3b1, phosphorylated at Thr434. The affinity-purified antibody recognised wild type SF3b1-NT-His6 after in vitro-phosphorylation by DYRK1A, but not the unphosphorylated protein or SF3b1-NT-His6-T434A (Fig. 7A). In contrast, the commercial pThrPro-specific antibody also bound to other phosphorylated ThrPro motifs in the T434A mutant of SF3b1. This result shows that the pT434-directed antibody exhibits high specificity for this phosphorylation site in SF3b1.
The anti-pT434 antibody was then used to study the phosphorylation of Thr434 in transfected COS-7 cells. To assess the specificity of the reaction, several nuclear protein kinases were tested in parallel with DYRK1A for their capacity to enhance phosphorylation of Thr434 in GFP-SF3b1-NT. DYRK1B is the kinase most closely related to DYRK1A (85% of identical amino acids in the catalytic domain ). HIPK2 (homeodomain-interacting protein kinase 2) was selected as a more distant member of the DYRK family (42% identity) , and CLK3 is a kinase known to phosphorylate splicing factors (see above). As shown in Fig. 7B, the pThr434-specific antibody detected SF3b1-NT in cells that did not overexpress DYRK1A (leftmost lane). This result is consistent with the labelling of spots 1 and 2 by endogenous kinases in COS-7 cells (Fig. 4C, 5B, and 6A). Signal intensity was dose-dependently enhanced by co-expression of DYRK1A or DYRK1B but not DYRK1A-K188R. SF3b1-T434A (lane 5) was not recognised by the antibody, confirming that the antibody was indeed specific for phosphoThr434. Notably, co-expression of HIPK2, but not CLK3, also resulted in an increased phosphorylation of Thr434 in SF3b1-NT. A second band (marked by an asterisk in Fig. 7B) could also be identified as a form of SF3b1-NT because of its absence in cells transfected with SF3b1-NT-T434A. As noted above (Fig. 2A), it is likely that this band represents a posttranslationally modified form of the protein.
In addition to the recombinant SF3b1-NT protein, the anti-pT434 antibody labelled a band with an apparent molecular mass of 150 kDa that was only detectable in lysates of cells overexpressing catalytically active DYRK1A or DYRK1B. This band co-migrated with the endogenous SF3b1 protein as identified by a commercially available antibody. As shown in Fig. 7C, the 150-kDa band was also detected in a nuclear protein fraction purified from DYRK1A-overexpressing COS-7, further supporting the identification as SF3b1. These data provide evidence that DYRK1A and DYRK1B can phosphorylate the full length, endogenous SF3b1 protein in intact cells. In contrast, overexpression of HIPK2 did not enhance phosphorylation of Thr434 in SF3b1, suggesting that this kinase cannot phosphorylate the endogenous protein in the spliceosome.
Phosphorylation of SF3b1 by endogenous DYRK1A
In order to assess the role of endogenous DYRK1A in the phosphorylation of SF3b1, we constructed two plasmids expressing small hairpin RNA (shRNA) for specific downregulation of human DYRK1A. The target sequences were carefully selected to avoid potential effects on DYRK1B mRNA. As shown in Fig. 8A, transient transfection of either one of the shRNA constructs efficiently reduced the level of GFP-DYRK1A, suggesting that they should also downregulate endogenous DYRK1A which is expressed at much lower levels. Next we determined the effect of the shRNA constructs on the phosphorylation of Thr434 in SF3b1-NT in two different human cell lines (Fig. 8B). Transient transfection of either construct resulted in a marked reduction of Thr434 phosphorylation, indicating that DYRK1A is the major Thr434 kinase in HEK193T cells and in HepG2 cells.