The function of many protein kinases is controlled by the phosphorylation of a critical tyrosine residue in the activation loop. Dual specificity tyrosine-phosphorylation-regulated kinases (DYRKs) autophosphorylate on this tyrosine residue but phosphorylate substrates on aliphatic amino acids. This study addresses the mechanism of dual specificity kinase activity in DYRK1A and related kinases. Tyrosine autophosphorylation of DYRK1A occurred rapidly during in vitro translation and did not depend on the non-catalytic domains or other proteins. Expression in bacteria as well as in mammalian cells revealed that tyrosine kinase activity of DYRK1A is not restricted to the co-translational autophosphorylation in the activation loop. Moreover, mature DYRK1A was still capable of tyrosine autophosphorylation. Point mutants of DYRK1A and DYRK2 lacking the activation loop tyrosine showed enhanced tyrosine kinase activity. A series of structurally diverse DYRK1A inhibitors was used to pharmacologically distinguish different conformational states of the catalytic domain that are hypothesized to account for the dual specificity kinase activity. All tested compounds inhibited substrate phosphorylation with higher potency than autophosphorylation but none of the tested inhibitors differentially inhibited threonine and tyrosine kinase activity. Finally, the related cyclin-dependent kinase-like kinases (CLKs), which lack the activation loop tyrosine, autophosphorylated on tyrosine both in vitro and in living cells. We propose a model of DYRK autoactivation in which tyrosine autophosphorylation in the activation loop stabilizes a conformation of the catalytic domain with enhanced serine/threonine kinase activity without disabling tyrosine phosphorylation. The mechanism of dual specificity kinase activity probably applies to related serine/threonine kinases that depend on tyrosine autophosphorylation for maturation.
Structured digital abstract
- CMGC group
CDKs, MAPKs, GSK3 and CLKs
dual specificity tyrosine-phosphorylation-regulated kinase
green fluorescent protein
glycogen synthase kinase 3
homeodomain-interacting protein kinase 2
mitogen-activated protein kinase
N-terminal autophosphorylation accessory
Few protein kinases can transfer the γ-phosphate group from ATP to either aliphatic (serine or threonine) or aromatic (tyrosine) hydroxyl groups of amino acid side chains of their substrate proteins . Best known among these so-called dual specificity kinases are the upstream kinases that activate members of the mitogen-activated protein kinase (MAPK) family by phosphorylating a conserved TXY motif in the activation loop [2, 3]. These MAPK kinases (also called MAP2K) are unique in their absolute substrate specificity for the cognate MAPKs they activate . Several members of the CMGC group, which comprises the cyclin-dependent kinases (CDKs), MAPKs, glycogen synthase kinases 3 (GSK3) and CDK-like kinases (CLKs), have also become known as dual specificity kinases [4-6]. These kinases phosphorylate only serine and threonine side chains in their substrates but autophosphorylate on tyrosine.
Many CMGC group kinases depend on the phosphorylation of a conserved tyrosine residue in the activation loop to attain full catalytic activity. Phosphorylation of this tyrosine stabilizes the active conformation by allowing electrostatic interactions of the phosphate with two conserved arginine residues . Only the members of the MAPK family are regulated by upstream kinases (MAP2K), whereas DYRKs, homeodomain-interacting protein kinase 2 (HIPK2), GSK3 and intestinal cell kinase autophosphorylate on the corresponding tyrosine [8-11]. On the other hand, CLKs lack the conserved tyrosine residue in the activation loop but are also capable of tyrosine autophosphorylation [4, 6, 12]. It is an open question whether other CMGC kinases can also target tyrosine residues other than that in the activation loop.
Kinases of the DYRK family have been the model system for the paradigm study of the biochemical mechanism of tyrosine autophosphorylation in the activation loop. The Drosophila DYRKs, minibrain (MNB) and dDyrk2, autophosphorylate on tyrosine intramolecularly while the polypeptide is still bound to ribosomes . After translation, the mature kinase appears to lose tyrosine kinase activity. The authors hypothesize that a conformational change after translation and tyrosine autophosphorylation results in this change of residue specificity . This model is supported by their observation that tyrosine autophosphorylation by the translational intermediate and the phosphorylation of exogenous substrates on serine or threonine by the mature kinase exhibit differential inhibitor sensitivity .
A similar but not identical mechanism was found for GSK3β, which autophosphorylates in a heat shock protein 90 dependent manner shortly after translation . In contrast to MAPK, the critical tyrosines in DYRK and GSK3 appear to be constitutively phosphorylated and serve no regulatory function . Collectively, these findings support the hypothesis that this mechanism of autoactivation also applies to other CMGC kinases that are not regulated by upstream kinases [15, 16]. However, it must be noted that even the role of tyrosine phosphorylation in the activation of DYRK1A is not uncontested .
In dDyrk2 from Drosophila and TbDYRK2 from Trypanosoma brucei, an N-terminal autophosphorylation accessory (NAPA) region has been defined that is required for tyrosine autophosphorylation . This sequence motif is conserved in a subgroup of the DYRK family called class 2 DYRKs (including mammalian DYRK2, DYRK3 and DYRK4) but not in class 1 DYRKs (including DYRK1A and DYRK1B) [18, 19]. The NAPA region is proposed to provide a chaperone-like function that is required for tyrosine autophosphorylation activity of class 2 DYRKs.
The current model of the mechanism of tyrosine autophosphorylation in the DYRK family is based on the landmark study of the Drosophila DYRKs (dDYRK2 and MNB). Of the mammalian DYRKs, DYRK1A has attracted particular interest as a regulator of neurogenesis and appears to play an important role in altered brain development in Down syndrome [20, 21]. We have recently found that tyrosine autophosphorylation of DYRK1A is an intrinsic capacity of the catalytic domain and does not depend on the presence of chaperones .
The present study aims to scrutinize the mechanism of dual specificity kinase activity in mammalian DYRKs. Specifically, we address the question whether tyrosine phosphorylation is a strictly co-translational event or can still take place in mature DYRKs. Furthermore, we examine whether the regulatory tyrosine in the activation loop is the only target site of DYRK tyrosine kinase activity and whether inhibitors can differentially target the tyrosine kinase form and the mature serine/threonine-specific conformation of DYRK1A.
Tyrosine autophosphorylation of CMGC kinases
We selected several kinases of the CMGC group with known phosphotyrosine sites in the activation loop to assess their capacity for tyrosine autophosphorylation when expressed in Escherichia coli as single domain proteins (lacking the non-catalytic regions). CLK1 was included as a closely related kinase that lacks the conserved tyrosine residue in the activation loop (for a sequence alignment of the activation loop see Table S1). An additional construct for DYRK2 was generated to assess the requirement of the NAPA region for tyrosine autophosphorylation .
Following expression in E. coli, phosphotyrosine was detected by immunoblot analysis for DYRK1A, DYRK2, HIPK2 and CLK1 (Fig. 1A). DYRK1B did not exceed the background level defined by the kinase dead mutant DYRK1A-K188R but gave a very weak phosphotyrosine signal when expressed at room temperature (not shown). Unexpectedly, tyrosine autophosphorylation of human DYRK2 did not depend on the NAPA region (Fig. 1A,B). The same result was obtained with two other phosphotyrosine-specific monoclonal antibodies (PY20, PY100; data not shown).
Next we examined tyrosine autophosphorylation in an E. coli-based coupled in vitro transcription–translation system. The PURExpress system is reconstituted from recombinant proteins and purified ribosomes, which excludes effects of unknown proteins such as chaperones. DYRK1A, HIPK2 and CLK1 were autophosphorylated on tyrosine residues when expressed in vitro, indicating that tyrosine kinase activity of these kinases is an intrinsic property of the catalytic domain and does not require co-factors (Fig. 1B). All attempts to produce tyrosine phosphorylated DYRK2 or DYRK2-NAPA by in vitro translation failed. We decided to focus on DYRK1A as a model system to study the mechanism of tyrosine autophosphorylation.
Tyrosine autophosphorylation of DYRK1A is not limited to Y321 in the activation loop
To assess whether DYRK1A and DYRK2 can also autophosphorylate on tyrosines other than those in the activation loop, we expressed tyrosine→phenylalanine point mutants in bacteria. Strikingly, the level of tyrosine phosphorylation was not altered in DYRK1A-Y321F, whereas no phosphotyrosine was detectable in DYRK2-Y382F (Fig. 2A). Mutation of another tyrosine in DYRK1A (Y145) that has been found phosphorylated in phosphoproteomic studies  (phosphositeplus, www.phosphosite.org) also did not reduce immunodetectable phosphotyrosine. This result was reproduced in the in vitro translation system (Fig. 2B).
Tyrosine autophosphorylation of DYRK1A is not strictly coupled to translation
When monitoring in vitro translation of DYRK1A in time-course experiments, tyrosine phosphorylation was already detectable before the Strep-tag (Fig. 3A), indicating that tyrosine autophosphorylation of DYRK1A is a very rapid event. Next we asked whether tyrosine autophosphorylation of DYRK1A can only take place during translation. When translation of DYRK1A was stopped after 2 h and incubation continued for 2 h, relative tyrosine phosphorylation increased further by a factor of 2. However, this effect did not reach statistical significance (P = 0.067; Fig. S5). Therefore, DYRK1A was expressed in the presence of the unspecific kinase inhibitor staurosporine to produce a hypophosphorylated protein. Staurosporine effectively reduced DYRK1A autophosphorylation at a concentration of 30 μm without affecting in vitro translation (Fig. 3B). After translation, the sample was diluted 100-fold to obtain a final concentration of 0.3 μm at which staurosporine did not inhibit autophosphorylation activity (Fig. 3B). This approach takes advantage of the fact that tyrosine autophosphorylation of DYRKs is an intramolecular reaction  and thus independent of enzyme concentration. This experiment showed that tyrosine autophosphorylation of DYRK1A is possible after termination of translation (Fig. 3C). It must be noted that the phosphotyrosine-specific antibody does not discriminate between the phosphorylation of Y321 in the activation loop and other tyrosines (see Fig. 2).
Differential effects of kinase inhibitors on autophosphorylation and substrate phosphorylation
Following the approach of Lochhead et al. , we aimed to pharmacologically distinguish the different conformational states of DYRK1A that may account for its dual specificity kinase activity. For a broad coverage of chemical scaffolds, we compiled a series of compounds comprising known DYRK1A inhibitors (harmine , TG003 [24, 25], KH-CB19 , 4,5,6,7-tetrabromobenzotriazole (TBB) [27, 28]) and unspecific kinase inhibitors (staurosporin, curcumin). TBB was analysed because this compound had been used to establish the differential inhibitor sensitivity of the translational intermediate and mature form of dDYRK2 . Unfortunately, TBB inhibited DYRK1A only weakly. In addition, we included 5-iodotubercidin, a compound previously reported to inhibit DYRK2 , and NIH54, a newly developed derivative of KH-CB19. Structures of the inhibitors are provided in Fig. S3. Inhibitors were used at concentrations that resulted in the near maximal inhibition of substrate phosphorylation (Fig. 4, lower panels). We used recombinant splicing factor 3B1 (SF3B1) as a substrate, which is phosphorylated by DYRK1A at T434 . Only the three most potent inhibitors, namely 5-iodotubercidin, harmine and staurosporine, reduced tyrosine autophosphorylation of DYRK1A by more than 50%. This result shows that the DYRK1A autophosphorylation assay is less sensitive to the known DYRK inhibitors than the substrate phosphorylation assay.
Given that tyrosine autophosphorylation is considered necessary for maturation of DYRKs, DYRK1A-Y321F should be incapable of achieving the mature conformation and thus be resistant to compounds that selectively target the mature conformation of DYRK1A. DYRK1A-Y321F is sufficiently active to be tested in in vitro kinase assays . The effects of the inhibitors on DYRK1A-Y321F and the wild-type kinase showed a very similar pattern (Fig. 4B). This result does not support the hypothesis that the tested compounds selectively inhibit the mature form of DYRK1A. The differential inhibition by TBB (30 μm) of DYRK1A-Y321F and wild-type kinase was not reproduced in assays with GST-DYRK1A fusion proteins isolated from E. coli and remains an unexplained peculiarity of this in vitro expression system.
Inhibitor sensitivity of tyrosine versus threonine autophosphorylation
We reasoned that the tyrosine autophosphorylation assay was more resistant to pharmacological inhibition because a single phosphorylation event in 2 h suffices for maximal signal intensity, whereas the substrate phosphorylation assay had been optimized to maximize substrate turnover. To compare the effects of the inhibitors on tyrosine and threonine phosphorylation under the same reaction conditions, the sequence segment of SF3B1 comprising T434 was fused to the C-terminus of DYRK1Acat (Fig. 5A) in a vector suitable for bacterial in vitro translation. In contrast to the assay shown in Fig. 4A, tyrosine autophosphorylation was rather more sensitive to the tested inhibitors than threonine autophosphorylation (Fig. 5B,C). Hence, the differential inhibition of tyrosine versus threonine phosphorylation in Fig. 4 is a consequence of the different reaction conditions (turnover rate and enzyme/substrate ratio) in these assays.
Tyrosine phosphorylation of DYRK1A expressed in HeLa cells
Next we expressed GFP-DYRK1A in HeLa cells to examine whether mature DYRK1A isolated from mammalian cells retains tyrosine kinase activity. As a control, we generated the D287N point mutant in which the catalytic aspartate was substituted by asparagine. We preferred this mutant as a kinase dead control because DYRK1A-K188R retains very low but detectable catalytic activity when expressed in mammalian cells . As shown in Fig. 6B,C, levels of phosphotyrosine were increased 2-fold after incubation of wild-type GFP-DYRK1A with ATP, providing evidence that mature DYRK1A autophosphorylates on tyrosine. Very weak phosphotyrosine bands were observed for GFP-DYRK1A-D287N, indicating that most of the phosphotyrosine in the wild-type protein is due to autophosphorylation and not other co-purified protein kinase activity. GFP-DYRK1A-Y321F contained minimal phosphotyrosine after immunoprecipitation (‘–ATP’ sample) but was strongly tyrosine phosphorylated when incubated with ATP. In contrast, autophosphorylation of DYRK1A-ΔNΔC did not increase, suggesting that the relevant tyrosines are located outside the catalytic domain. This construct corresponds to the bacterially expressed protein and retains full catalytic activity (not shown).
To identify the tyrosine residue(s) autophosphorylated by DYRK1A, we analysed the immunoprecipitated proteins by LC-MS/MS. Consistent with previous observations, Y321 was only found in the phosphorylated state (Fig. 6D). In addition, wild-type DYRK1A was partially phosphorylated on Y104 and Y111 in the N-terminal region and was autophosphorylated on Y720 after incubation with ATP. Consistent with the immunochemical detection, Y104 and Y111 phosphotyrosine-containing peptides were detected in DYRK1A-Y321F only after incubation with ATP.
The increase in phosphotyrosine after incubation with ATP implies that DYRK1A is not fully phosphorylated on tyrosine residues when expressed in HeLa cells. To exclude the possibility that tyrosine autophosphorylation was only possible under in vitro conditions, we tested whether tyrosine phosphorylation of DYRK1A could be increased by treating the cells with the tyrosine phosphatase inhibitor sodium vanadate. As shown in Fig. 6E,F, vanadate treatment resulted in a small but statistically significant increase of phosphotyrosine immunoreactivity of wild-type DYRK1A. Although it cannot be formally ruled out that vanadate directly influences residue specificity of DYRK1A, the in vitro assays in Fig. 6B suggest that tyrosine autophosphorylation activity is a regular feature of DYRK1A but is counteracted in living cells by the action of tyrosine phosphatases. Consistent with the in vitro assay, vanadate treatment strongly increased tyrosine phosphorylation of DYRK1A-Y321F but not DYRK1A-ΔNΔC. The majority of phosphotyrosine is likely to result from autophosphorylation, since vanadate treatment induced much weaker tyrosine phosphorylation of DYRK1A-D287N. However, these results do not exclude the possibility that DYRK1A is phosphorylated by another tyrosine kinase whose activity depends on catalytically active DYRK1A.
Tyrosine autophosphorylation in the DYRK family and related kinases
Next we tested other members of the DYRK family along with HIPK2 and CLK3 for their capacity to autophosphorylate on tyrosine after immunoprecipitation from HeLa cells. As expected, phosphotyrosine was readily detectable in all kinases (Fig. 7A) except for CLK3, which lacks the conserved tyrosine in the activation loop. Only very weak basal tyrosine phosphorylation of CLK3 was observed in some experiments (see Fig. 7B). Tyrosine phosphorylation appeared to be moderately increased in some of the DYRKs (DYRK1B, DYRK2), but the effects did not reach statistical significance due to the variable degree of the basal phosphorylation. Strikingly, CLK3 contained high levels of phosphotyrosine after the autophosphorylation reaction.
Vanadate treatment of HeLa cells caused moderate but non-significant increases of phosphotyrosine in DYRK1B and DYRK3 (Fig. 7B) and again strongly increased tyrosine phosphorylation of CLK3. These results indicate that CLK3 autophosphorylates on tyrosine residues but contains little phosphotyrosine under basal conditions due to the action of tyrosine phosphatases.
Tyrosine kinase activity of DYRK1A-Y321F and DYRK2-Y382F
We noticed that tyrosine autophosphorylation was particularly well detectable in those kinases that lack the activation loop tyrosine (DYRK1A-Y321F and CLK3). Therefore we analysed whether the corresponding mutant of the activation loop tyrosine (Y382) in DYRK2 also had tyrosine kinase activity. Similar to DYRK1A, DYRK2-Y382F contained no detectable phosphotyrosine when isolated from HeLa cells but autophosphorylated on tyrosine when incubated with ATP (Fig. 8A). Furthermore, two bands co-purified with GFP-DYRK2-Y382F were tyrosine phosphorylated in the in vitro assay (Fig. 8A, open arrowheads). This reaction was sensitive to inhibition by 5-iodotubercidin, strongly suggesting that it was catalysed by DYRK2-Y382F and not by a co-purified tyrosine kinase. The same pair of co-purified bands was phosphorylated by DYRK1-Y321F in a harmine-inhibitable manner (Fig. 8B). These bands were not recognized by antibodies directed against green fluorescent protein (GFP) or N- and C-terminal epitopes in DYRK1A, excluding the possibility that they represent proteolytic fragments of DYRK1A. DYRK1A-Y321F was also able to trans-phosphorylate tyrosine(s) in the N-terminal domain of DYRK1A (GFP-DYRK1A(80–156), Fig. 6A) when expressed as an independent protein (Fig. 8C). Wild-type DYRK1A did not produce higher levels of phosphotyrosine than the background obtained with the catalytically inactive mutant D287N. These results show that DYRK1A-Y321F and DYRK2-Y282F can phosphorylate tyrosines in exogenous substrates and have higher tyrosine kinase activity than the wild-type kinases.
Dual specificity kinase activity is a salient feature of the DYRK family of protein kinases and is intrinsically linked to the distinctive mechanism of autoactivation of the DYRKs. The current model of DYRK activation by co-translational, one-time tyrosine autophosphorylation in the activation segment is based on the analysis of the Drosophila dDYRK2 and MNB but has not been scrutinized for general validity within the DYRK family. Here we provide evidence that DYRK1A retains tyrosine kinase activity after translation is finished and that tyrosine kinase activity of DYRK1A is not limited to the intramolecular, activating autophosphorylation.
Post-translational tyrosine kinase activity of DYRK1A
Tyrosine phosphorylation of in vitro translated DYRK1A parallels the production of the protein (Fig. 3A), which is consistent with a co-translational autophosphorylation mechanism. Very similar kinetics have been reported for HIPK2  and Drosophila dDYRK2 , whereas tyrosine autophosphorylation of GSK3β takes place with some delay after translation . Here we show that DYRK1A retains the capacity of autophosphorylating on tyrosine residues after termination of in vitro translation (Fig. 3C). In this experimental setup, inhibitor treatment might have arrested the catalytic domain in the nascent conformation, allowing delayed autophosphorylation after washout of the inhibitor. Importantly, mature DYRK1A purified from HeLa cells was also able to autophosphorylate on tyrosine (Fig. 6B,C). Hence, tyrosine kinase activity of DYRK1A is not limited to a transient folding state that only exists during translation. Consistently, a DYRK1A protein monophosphorylated on Y321 was able to autophosphorylate several other tyrosine residues in vitro . The DYRK orthologue from yeast, Yak1p, has also been shown to autophosphorylate on tyrosine when immunoprecipitates were incubated with ATP . However, it is important to note that none of these results contradicts the hypothesis that the autophosphorylation in the activation loop is a one-time-only event during translation. Nevertheless, HIPK2 has been shown to be capable of rephosphorylating the activation loop tyrosine after dephosphorylation of the mature enzyme .
Tyrosine kinase activity of DYRKs is not limited to the conserved tyrosine in the activation loop
The capacity of tyrosine phosphorylation in DYRKs and related CMGC kinases such as GSK3 and HIPK2 is essential for the activation of these kinases and has been thought to be strictly limited to cis-autophosphorylation of one specific tyrosine residue in the activation loop [13, 14, 32]. Here we show that DYRK1A-Y321F and DYRK2-Y382F are capable of phosphorylating other tyrosines when expressed in E. coli (Fig. 2) or in HeLa cells (Figs 6 and 8). This result appears to contradict previous studies in mammalian cells or insect Sf9 cells, in which mutation of the activation loop tyrosine eliminated immunodetectable phosphotyrosine in DYRK1A, DYRK1B, DYRK4, dDYRK2 and HIPK2 [31, 32, 35-37]. Furthermore, no phosphorylated tyrosines except those in the activation loop were detected by mass spectrometry analysis of dDYRK2 or MNB from Sf9 cells . In fact, tyrosine autophosphorylation of DYRK1A-Y321F and DYRK2-Y382F was also minimal in our study unless HeLa cells were treated with vanadate. Thus, the constitutive tyrosine kinase activity of mature DYRK1A is masked in eukaryotic cells unless tyrosine phosphatases are inhibited by vanadate, but can be revealed in the absence of phosphatases, such as in in vitro reactions. No vanadate-induced increase of phosphotyrosine was observed in DYRK1A-ΔNΔC, suggesting that the major part of the reversible tyrosine autophosphorylation takes place in regions that do not include the activation loop. This result is consistent with the current view that the activation loop tyrosine in DYRKs is constitutively phosphorylated and has no regulatory function [13, 16].
Phosphoproteomics studies have identified several phosphotyrosines outside the activation segment of DYRK1A, including pY104, Y111, pY145 and pY147 in the aminoterminal domain (www.phosphosite.org). This observation excludes the possibility that tyrosine phosphorylation of mature DYRK1A is only found in in vitro reactions or under unphysiological conditions (vanadate treatment). Interestingly, the phosphorylation status of Y145/Y147 has recently been correlated with the subcellular localization of DYRK1A in brain neurons . Here we identify Y104 and Y111 in the N-terminal region of DYRK1A as sites that are partially (auto-)phosphorylated when DYRK1A is expressed in HeLa cells. We and others have previously identified Y111 as an autophosphorylation site of bacterially expressed DYRK1A [31, 33]. Autophosphorylation of these tyrosines may be an ancestral feature of class 1 DYRKs, because Y104 and Y111 are extremely well conserved even in very distantly related unicellular eukaryotes including slime mould (Dictyostelium) and the flagellate Trypanosoma (see Fig. S7 for a sequence alignment). The repeated identification of Y111, but not the neighbouring Y112, as an autophosphorylation site of DYRK1A provides evidence of target site selectivity in tyrosine phosphorylation by DYRK1A.
Model of dual specificity protein kinase activity of DYRKs
The available evidence supports the hypothesis that DYRKs are activated by the co-translational, intramolecular autophosphorylation of the conserved tyrosine in the activation loop, which appears to be irreversible . However, the current model of a switch between mutually exclusive conformational states conferring either serine/threonine or tyrosine kinase activity is not consistent with our observation that mature DYRK1A retains tyrosine kinase activity as a mature protein. Furthermore, DYRK1A-Y321F cannot attain the autoactivated state but has significant serine/threonine kinase activity (10–20% of the wild-type protein [31, 39]). Here we want to propose a modified version of the model for autoactivation of DYRK family kinases (Fig. 9). We presume that the catalytic domain of DYRK1A exists in a dynamic equilibrium between two conformational states, one capable of tyrosine autophosphorylation and one capable of phosphorylating serine or threonine residues within the consensus sequence (RXS/TP). If Y321 in the activation loop is unphosphorylated, DYRK1A exists predominantly in the tyrosine kinase conformation but can also adopt the serine/threonine kinase conformation. Interestingly, crystal structures of GSK3β revealed that this kinase can indeed exhibit the same active conformation whether or not Y216 in the activation loop is phosphorylated [40-43].
The model implies that the autophosphorylation of Y321 stabilizes the serine/threonine kinase configuration of the catalytic domain. This assumption is consistent with the known structural role of the activation loop phosphotyrosine in coordinating the P + 1 loop and defining selectivity of CMGC kinases for serine/threonine residues followed by a proline . The lack of the electrostatic contacts mediated by the phosphate moiety in the activation loop is expected to destabilize the P + 1 loop and thereby (a) reduce the affinity and turnover rates of the unphosphorylated kinase for canonical substrates and (b) relax the selectivity in target recognition. Hence, DYRK1A-Y321F has lower serine/threonine kinase activity than wild-type DYRK1A [31, 39] but higher tyrosine kinase activity. Indeed, DYRK1A-Y321F and DYRK2-Y382F, but not the wild-type kinases, catalysed the tyrosine phosphorylation of two unidentified proteins that were co-purified from HeLa cells (Fig. 8). To our knowledge, this is the first observation of an intermolecular tyrosine phosphorylation reaction catalysed by dual specificity kinases of the CMGC group. Drosophila dDYRK2 and MNB did not even phosphorylate peptides that mimicked the sequence of their own activation loop , and neither DYRK1A, DYRK2 nor DYRK4 phosphorylated tyrosine sites in peptide arrays [31, 35]. It must be noted that the capacity of these DYRK mutants to phosphorylate substrates on tyrosines in immunocomplex kinase assays does not reveal a physiological function but rather a biochemical characteristic of the enzymes.
Effects of kinase inhibitors on tyrosine autophosphorylation and substrate phosphorylation
As we have previously found for harmine , all tested compounds inhibited the phosphorylation of T434 in SF3B1 much more potently than tyrosine autophosphorylation of DYRK1A (Fig. 4). However, this difference was not due to the differential inhibition of threonine versus tyrosine kinase conformations, since the preferential inhibition of threonine phosphorylation was lost when threonine and tyrosine kinase activity were determined in the same assay (Fig. 5). It is impossible per se to perform autophosphorylation assays under conditions of substrate saturation. Minimal catalytic activity suffices to allow the one catalytic cycle to proceed that is required for stoichiometric autophosphorylation within the 2-h incubation time. Therefore, even strong inhibition of catalytic activity hardly reduces the amount of autophosphorylated kinase.
As a secondary finding, the present data identify 5-iodotubercidin as a potent inhibitor of DYRK1A. Iodotubercidin was previously characterized as a specific inhibitor of DYRK2 and Haspin with excellent selectivity in a screen against a panel of 98 kinases . However, the use of 5-iodotubercidin as a DYRK inhibitor in cellular assays or animal experiments is compromised by its high activity as an inhibitor of adenylate kinase (IC50 = 9 nm ).
Tyrosine autophosphorylation of other CMGC kinases
We have previously found that an expression construct of the yeast kinase Yak1p consisting only of the catalytic domain and the adjacent DYRK homology box is capable of tyrosine autophosphorylation in E. coli . Here we show that analogous constructs of DYRK2, HIPK2 and CLK1 also autophosphorylate in E. coli. Unexpectedly, the NAPA region was not required for tyrosine autophosphorylation of DYRK2, although the sequence motif is well conserved in the human protein. It remains to be determined whether the function of the NAPA region differs between individual kinases (human DYRK2 versus Drosophila dDYRK2 and Trypanosoma brucei TbDYRK2) [18, 19] or depends on the experimental system used (expression in E. coli versus rabbit reticulocyte lysate). It must also be noted that our result does not provide direct evidence for NAPA-independent phosphorylation of the activation loop, since the antibody might have detected phosphotyrosines other than Y382.
Kinases of the CLK family lack the conserved tyrosine in the activation loop but were shown to exhibit dual specificity kinase activity in vitro or in bacterial expression systems [12, 47]. We found that CLK3 contained minimal phosphotyrosine when expressed in HeLa cells. However, vanadate strongly stimulated tyrosine phosphorylation of CLK3, revealing for the first time that CLKs can autophosphorylate on tyrosine in a homologous expression system. The catalytic activity of CLK1 has been reported to be unaffected by tyrosine phosphorylation , and the regulatory function, if any, of tyrosine autophosphorylation in CLKs remains to determined. However, it is interesting to note that dual specificity kinase activity is a conserved feature even in those CMGC kinases that are not activated by tyrosine phosphorylation. Although kinases of the MAPK family are a paradigm for the regulation through upstream kinases, autophosphorylation of the activation loop tyrosine has been identified as an alternative pathway of activation in p38α  and in a constitutively active splicing variant of JNK2α . Taken together, existing evidence suggests that the peculiar feature of dual specificity kinase activity is an ancestral trait of CMGC kinases.
Materials and methods
Mouse monoclonal antibodies against phosphotyrosine (PY99; Santa Cruz Biotechnology, Santa Cruz, CA, USA; PY100; Cell Signaling Technology, Danvers, MA, USA; PY20; BD Transduction Laboratories, Lexington, KY, USA) were purchased commercially. The PY99 antibody was used throughout except for the experiment shown in Fig. 1 where the three antibodies were compared. A sheep DYRK1A antibody was kindly provided by Sir Philip Cohen (MRC Protein Phosphorylation Unit, Dundee, UK). The rabbit antibody used in the kinase assays for detecting phosphorylated T434 in SF3B1 (pT434) has been described previously .
The following kinase inhibitors were obtained commercially: staurosporine (Enzo Life Sciences, Lörrach, Germany), TG003 (Sigma Aldrich, Taufkirchen, Germany), harmine (Fluka, Buchs, Switzerland), TBB and 5-iodotubercidin (Tocris Bioscience, Minneapolis, MN, USA). KH-CB19 was synthesized as described previously . NIH54 is a novel dichloroindolyl enaminonitrile related to KH-CB19 with lower affinity for the off-target CLK3 (Fig. S4). Inhibitors were dissolved in dimethylsulfoxide (DMSO) to create a 1 mm stock solution, except that harmine was dissolved in ethanol as described previously . Further dilutions of the inhibitors were prepared in water. DMSO was present at a final concentration of 3% when in vitro translation reactions or kinase assays were performed in the presence of inhibitors.
Expression constructs for T7 polymerase-driven bacterial expression contained the Strep-tag® II sequence (MAWSHPQFEK) N-terminal of the catalytic domain of the kinases of interest in the vector pET28a. DYRK1A cDNA was also cloned into pBEN-SGC (kind gift from O. Gileadi, Structural Genomics Consortium, Oxford, UK) to produce a fusion protein with an N-terminal streptavidin binding peptide (SBP) tag and a C-terminal sequence segment derived from splicing factor 3B1 (SF3B1). A detailed description of the expression constructs is provided in the supplementary data (Figs. S1 and S2, Table S1).
Mammalian expression plasmids for the DYRKs and CLK3 have been described previously [35, 46]. A deletion construct (GFP-DYRK1A-ΔNΔC) was generated that contained only the catalytic domain and the DYRK homology box (amino acids 135–481; numbering in this paper refers to the splicing variants defined as ‘canonical’ in the UniProt database www.uniprot.org). Point mutants of DYRK1A and DYRK2 were created by DpnI-mediated site directed mutagenesis. The GFP-HIPK2 expression plasmid was kindly provided by M. L. Schmitz (Justus-Liebig-University Giessen, Germany).
Expression in E. coli
The kinase constructs were transformed into E. coli BL21(DE3). Test tubes containing 1 mL LB medium with 30 μg·mL−1 kanamycin were inoculated with 50 μL of an overnight culture and incubated at 37 °C, 230 rpm. After 3 h, the cultures were induced with 0.1 mm isopropyl-β-d-thiogalactopyranoside, and 100 μL samples were taken after further incubation as indicated in the figure legends. The cells were harvested by centrifugation, and each pellet was lysed in 10 μL Lämmli sample buffer and directly subjected to western blot analysis.
Expression in vitro
The kinase constructs were expressed in the PURExpress® In Vitro Protein Synthesis Kit (New England Biolabs, Beverley, MA, USA) at 37 °C for the times given in each experiment. We used 10 ng·μL−1 plasmid in a total volume of 5–20 μL. If indicated, translation was stopped with 30 μg·mL−1 kanamycin and 0.89 mg·mL−1 RNAse A.
Kinase activity assays
To measure the effect of kinase inhibitors on in vitro expressed DYRK1A (Fig. 4), PURExpress® reaction mixes with wild-type pET-ST2-DYRK1A or pET-ST2-DYRK1A-Y321F were diluted 1 : 60 with kinase buffer (25 mm Hepes pH 7.4, 0.5 mm dithiothreitol, 5 mm MgCl2) and then incubated for 5 min at 37 °C with 2 μg His6SF3B1-NT and 500 μm ATP in the presence of the respective compounds. His6-tagged SF3B1-NT was expressed in E. coli and prepared as described previously . Phosphorylation of SF3B1 on T434 was determined by immunoblot analysis with the help of a phosphospecific antibody .
Immunocomplex kinase assay and tyrosine phosphorylation in HeLa cells
HeLa were transfected with expression vectors for GFP fusion proteins using FuGENE HD (Promega, Mannheim, Germany). Two days later, the subconfluent cells from one 85-mm plate were lysed under non-denaturing conditions using 1 mL of native lysis buffer (50 mm Tris/Cl pH 7.5, 150 mm NaCl, 2 mm EDTA, 0.5% Igepal CA-630, supplemented with 1 mm Na3VO4, 1 mm phenylmethylsulfonyl fluoride and 10 μg·mL−1 each of aprotinin, pepstatin and leupeptin) on ice. The samples were cleared by centrifugation and incubated with 15 μL GFP trap® M (single domain camelid GFP antibody coupled to paramagnetic particles; ChromoTek, Martinsried, Germany) for 1 h at 4 °C. The beads were washed three times with washing buffer (50 mm Tris/Cl pH 7.5, 150 mm NaCl, 2 mm EDTA, 0.1% Igepal), resuspended in kinase buffer and split into two aliquots. Supernatants were removed and the immobilized proteins were incubated for 1 h at 30 °C in 30 μL kinase buffer supplemented with 1 mm Na3VO4 in the presence or absence of 500 μm ATP as indicated in the figure legends. Assays involving ATP-competitive inhibitors were performed at 100 μm ATP. Tyrosine phosphorylation was detected by western blotting.
To analyse tyrosine phosphorylation in living HeLa cells, cells were treated with 1 mm Na3VO4 for 1 h before cells were lysed. GFP-containing proteins were immunoprecipitated and analysed by western blotting.
After in vitro expression in the presence of 30 μm staurosporine, 10 μL samples were diluted 100-fold with DYRK kinase buffer (25 mm Hepes pH 7.0, 0.5 mm MgCl2, 0.5 mm dithiothreitol) and incubated with 1 mm ATP. The reaction was stopped at the indicated time points with 10 mm EDTA, 4 volumes of ice-cold acetone were added and the samples were stored at −20 °C for 1 h. After centrifugation, the pellets were dissolved in Lämmli's sample buffer and subjected to western blotting.
Samples were separated by SDS/PAGE and blotted onto nitrocellulose membranes. The membranes were blocked with 3% BSA in 20 mm Tris/Cl (pH 7.6), 137 mm NaCl containing 0.2% Tween-20 and probed with primary antibodies or Strep-Tactin horseradish peroxidase conjugate (IBA BioTAGnology, Göttingen, Germany). Chemiluminescence signals were detected using a LAS-3000 CCD imaging system and band intensities were densitometrically evaluated with the help of the aida image analyzer 3.52 software (Raytest, Straubenhardt, Germany). Anti-phosphotyrosine immunoreactivity was determined as a quantitative estimate of tyrosine phosphorylation. To compare data obtained in separate western blots in independent experiments, relative anti-phosphotyrosine immunoreactivity was calculated by division by the total protein signal of each lane and normalized to the untreated sample included in every blot. The one-sample t test was applied to test significance of treatment effects with the help of the graphpad prism 5.0 program (GraphPad Software, La Jolla, CA, USA). Effects were considered statistically significant at P < 0.05.
For mass spectrometry analysis, the appropriate bands were excised from Coomassie Colloidal Blue stained gels and subjected to in-gel digestions by sequencing grade trypsin and chymotrpysin (Promega). The resulting tryptic peptides were extracted with acetonitrile. To identify phosphorylated sites, peptides were analysed by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Peptide extracts were dissolved in 0.5% trifluoroacetic acid and separated by nano-HPLC (Dionex Ultimate 3000) equipped with a μ-precolumn (C18, 5 μm, 100 Å, 5 × 0.3 mm) and an Acclaim PepMap RSLC nanocolumn (C18, 2 μm, 100 Å, 150 × 0.075 mm) (all Thermo Fisher Scientific, Vienna, Austria). Samples were injected and concentrated on the enrichment column for 2 min at a flow rate of 20 μL·min−1 with 0.5% trifluoroacetic acid as isocratic solvent. Separation was carried out on the nanocolumn at a flow rate of 300 nL·min−1 using the following gradient, where solvent A is 0.05% trifluoroacetic acid in water and solvent B is 0.05% trifluoroacetic acid in 80% acetonitrile: 0–2 min, 4% B; 2–70 min, 4–28% B; 70–94 min, 28–50% B; 94–96 min, 50–95% B; 96–116 min, 95% B; 116–116.1 min, 95–4% B; 116.1–140 min, re-equilibration at 4% B. The sample was ionized in the nanospray source equipped with stainless steel emitters (ES528; Thermo Fisher Scientific, Austria) and analysed in an Orbitrap velos mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) operated in positive ion mode, applying alternating full scan MS (m/z 400–2000) in the ion cyclotron and MS/MS by higher-energy collisional dissociation of the 10 most intense peaks with dynamic exclusion enabled. The LC-MS/MS data were analysed by searching a database containing the protein sequences of the recombinant proteins and known background proteins with proteome discoverer 1.3 (Thermo Fisher Scientific, USA) and mascot 2.3 (MatrixScience, London, UK). Phosphorylation on tyrosine and oxidation on methionine were entered as variable modifications. A maximum false discovery rate of 5% using decoy search was chosen as the identification criterion .
We thank Angela Maurer, Kristina Reck, Charlotte Spitz, Stefan Spoerk and Barbara Darnhofer for competent experimental assistance. We are grateful to M. Lienhardt Schmitz (University of Giessen, Germany) and Opher Gileadi (SGC, Oxford, UK) for providing plasmids and to Sir Philip Cohen for the gift of the sheep DYRK1A antibody. This work was supported by the Deutsche Forschungsgemeinschaft to FB and WB (BE 1967/3-1), the Proteomics Core Facility (CP) of the IZKF Aachen (Interdisciplinary Center for Clinical Research within the Faculty of Medicine at RWTH Aachen University), the Austrian Science Fund (FWF) (DK-MCD W1226) and the Austrian Research Promotion Agency (FFG)/Comet program (funded by FFG, SFG and province of Styria) (ACIB 824186) to RBG.