Abstract
The multitude of mechanisms regulating the activity of protein kinases includes phosphorylation of amino acids contained in the activation loop. Here we show that the serine/threonine kinase HIPK2 (homeodomain-interacting protein kinase 2) is heavily modified by autophosphorylation, which occurs by cis-autophosphorylation at the activation loop and by trans-autophosphorylation at other phosphorylation sites. Cis-autophosphorylation of HIPK2 at Y354 and S357 in the activation loop is essential for its kinase function and the binding to substrates and the interaction partner Pin1. HIPK2 activation loop phosphorylation is also required for its biological activity as a regulator of gene expression and cell proliferation. Phosphorylation of HIPK2 at Y354 alone is not sufficient for full HIPK2 activity, which is in marked contrast to some dual-specificity tyrosine-phosphorylated and regulated kinases where tyrosine phosphorylation is absolutely essential. This study shows that differential phosphorylation of HIPK2 provides a mechanism for controlling and specifying the signal output from this kinase.
Introduction
Protein kinases are important regulators of virtually all aspects of cell life and share the basic architecture of their catalytic domain, which consists of a smaller N-terminal lobe (N-lobe) and a larger C-terminal lobe (C-lobe) (Hanks et al., 1988; Hanks and Hunter, 1995). A cleft between these two lobes serves as a docking site for adenosine triphosphate (ATP). The N-lobe forms a hydrophobic pocket that contacts the adenine ring of ATP, while a conserved lysine binds α and β phosphates (Lochhead, 2009). The C-lobe contains the activation loop (also known as the T-loop), which contains phosphorylatable amino acids in most but not all kinases (Nolen et al., 2004). The activation loop is defined as the region between the tripeptide motifs DFG and APE. The name of the activation loop reflects its key role for the control of kinase activity as phosphorylation events allow for conformational changes that lead to the correct disposition of substrate binding and catalytic groups (Johnson et al., 1996). Aspartate within the DFG motif chelates a magnesium ion that positions the phosphates for phosphotransfer. The APE sequence is part of the so-called P1 loop and is of critical importance for substrate binding (Nolen et al., 2004). The catalytic activity of all kinases depends on an aspartate molecule that mediates the correct orientation of the hydroxyl acceptor group in the peptide substrate. Kinases controlled by the primary phosphorylation site in the activation loop are characterized by a conserved arginine immediately preceding the conserved catalytic aspartate and were termed ‘RD’ kinases (Taylor and Radzio-Andzelm, 1994; Kornev et al., 2006). Phosphorylation of amino acids in the activation loop is important for the proper alignment of residues that mediate binding of substrate, ATP, and the transfer of the phosphate group, thus converting a kinase from the inactive to the active conformation (Huse and Kuriyan, 2002). There are three main routes that can lead to activation loop phosphorylation: (i) by trans-phosphorylation from an upstream kinase as part of a multi-step signaling cascade, as exemplified by classical MAP (mitogen-activated protein) kinase activation (Dhillon et al., 2007), (ii) by trans-autophosphorylation in which an active kinase trans-phosphorylates an inactive kinase, as seen, for example, for Src (Cooper and MacAuley, 1988), or (iii) by cis-autophosphorylation that can occur during the generation of the kinase at the ribosome or with the help of binding partners (Lochhead et al., 2005, 2006). This type of activation has thus far only been described for CMGC kinases (containing cyclin-dependent kinases, mitogen-activated protein kinase (MAPK), glycogen synthase kinase (GSK), and CDK-like kinases families) such as DYRK (dual-specificity tyrosine-phosphorylated and regulated kinase), p38α, or GSK3 (Lochhead, 2009). The mechanism of autophosphorylation of the activation loop in cis is not well understood. Tyrosine cis-autophosphorylation of DYRK2 is absolutely essential for its kinase activity and occurs during the production of the nascent kinase at the ribosome. This shorter transitory intermediate form of DYRK2 allows for autophosphorylation at a tyrosine, while all other substrates are phosphorylated at serine or threonine residues (Lochhead et al., 2005). Autophosphorylation of a transitional intermediate of GSK3 proceeds by a different mechanism that depends on chaperone activity (Lochhead et al., 2006). Within the mammalian CMGC group of kinases, DYRK kinases show the closest relationship to homeodomain-interacting protein kinases (HIPKs) (Manning et al., 2002). This family consists of the highly homologous kinases such as HIPK1, HIPK2, and HIPK3, and the more distantly related protein HIPK4 (Rinaldo et al., 2008). The best studied member of the HIPK family is HIPK2, which localizes mainly in subnuclear speckles and is important for the response to morphogenic signals or to DNA damage (Calzado et al., 2007). HIPK2 can stimulate gene expression upon phosphorylation of transcription factors including insulin promoter factor-1/pancreatic duodenal homeobox-1 (IPF1/PDX1) and p53 (D'Orazi et al., 2002; Hofmann et al., 2002; Boucher et al., 2009). Additionally, HIPK2 can also repress gene transcription upon recruitment of general regulators of gene expression such as the methyl-DNA-binding proteins ZBTB4 (Yamada et al., 2009) or MeCP2 (Bracaglia et al., 2009). All of the more than 40 known HIPK2 phosphorylation substrates are modified at a serine or threonine that is directly flanked by one or even two proline residues. HIPK2 serves as a regulator of cell proliferation, and its functional role in cancer was revealed in HIPK2-deficient mice, which develop more skin tumors after treatment with carcinogens (Wei et al., 2007). Loss of heterozygosity for the Hipk2 gene occurs in radiation-induced tumors in mouse cells (Mao et al., 2012) and in human papillary thyroid carcinomas (Lavra et al., 2011), suggesting that the activity of HIPK2 needs to be tightly controlled. Accordingly, HIPK2 levels are restricted by degradative ubiquitination employing at least four different ubiquitin E3 ligases including Siah (seven-in-absentia) proteins (Gresko et al., 2006; Rinaldo et al., 2007; Choi et al., 2008; Shima et al., 2008; Calzado et al., 2009). HIPK2 activity is also regulated by caspase-mediated removal of an autoinhibitory domain in the C-terminus (Gresko et al., 2006). A variety of experimental approaches has implicated HIPK2 in a bewilderingly large number of signaling pathways; these include p53 activation, Wingless and Hedgehog signaling, regulation of the hypoxic response by a hypoxia-regulated interaction with Siah2 and the DNA damage response (D'Orazi et al., 2002; Hofmann et al., 2002; Kanei-Ishii et al., 2004; Calzado et al., 2009; Lee et al., 2009). However, it remains completely unclear how a given kinase can participate in the regulation of so many distinct signaling pathways. Thus far, HIPK2 has never been identified as a member of a classical kinase cascade, where one kinase phosphorylates the activation loop of a downstream kinase and thus allows directional signal transmission and amplification.
Here we show that HIPK2 autophosphorylates its activation loop residues Y354 and S357. Each of these two residues controls kinase activity, substrate affinity and localization of the kinase, thus suggesting a means of how activation loop phosphorylation can shape signal transmission.
Results
HIPK2 undergoes pronounced autophosphorylation
Early studies on HIPK2 function had already shown that the gel electrophoretic mobility of wild-type (wt) HIPK2 was significantly slower in comparison to a kinase inactive point mutant (D'Orazi et al., 2002; Hofmann et al., 2002). In order to address the question of whether this upshift is due to phosphorylation events, cells were transfected to express HIPK2 or a kinase inactive point mutant (HIPK2-K221A), followed by isolation of the kinases by immunoprecipitation (IP) and treatment with λ phosphatase. Analysis of the gel electrophoretic behavior showed that phosphatase treatment converted the mobility of the wt kinase to that of the kinase-inactive version (Figure 1A). These results suggest that the slower migration of HIPK2-wt is due to phosphorylation and raise the question for the occurrence of HIPK2 autophosphorylation. To address this issue, HIPK2 and HIPK2-K221A were produced by in vitro translation in the absence or presence of the previously described HIPK2 inhibitor D-115893 (de la Vega et al., 2011). Subsequent immunoblotting showed (i) that the differential electrophoretic behavior also occurred for kinases that were produced in vitro and (ii) that differential migration on reducing sodium dodecyl sulfate (SDS) gels was largely lost in the presence of the HIPK2 inhibitor (Figure 1B). In order to identify HIPK2 phosphorylation sites in an unbiased approach, overexpressed HIPK2 was isolated by IP and phosphorylation sites were identified by mass spectrometry. Phosphorylated amino acids were detected distributed throughout the kinase and also at Y354 and S357 in the activation loop (Figure 1C). Mass spectrometric analysis of HIPK2-K221A showed remaining phosphorylation at T831 and S820, but absent modification of the other phosphorylation sites including Y354 and S357 in the activation loop. This suggests that modification of Y354 and S357 occurs by autophosphorylation, while T831 and S820 can be modified by transphosphorylation. A sequence comparison with the activation loops of other CMGC group kinases showed the strict conservation of the modified tyrosine and serine (Figure 1D).
Autophosphorylation of HIPK2. (A) 293T cells transfected to express Flag-HIPK2 wild-type (Flag-HIPK2) or Flag-HIPK2-K221A were lysed. An aliquot of the lysates was used for the input control (lower), while the remaining lysate was used for IP of HIPK2 and treatment with λ phosphatase as indicated. Equal amounts of protein were separated by SDS–PAGE and analyzed by immunoblotting with the specified antibodies. (B) Flag-HIPK2 or HIPK2-K221A was translated in vitro in the presence or absence of the HIPK2 inhibitor D-115893 (500 nM). The reaction was stopped by adding SDS sample buffer, and in vitro translated Flag-HIPK2 was subjected to western blotting. (C) Schematic representation of the various HIPK2 domains and the phosphorylated amino acids. Y354 and S357 in the activation loop are shown underlined. The amino acids that were also phosphorylated in the kinase-inactive HIPK2-K221A mutant are highlighted with an asterisk. (D) Architecture of the activation loops of HIPK2 and related kinases in the human genome. The anchoring sequences are underlined, the phosphorylated tyrosines and serines are given in bold, and the degree of sequence homology is shown in the lower part.
Trans-autophosphorylation of HIPK2
Autophosphorylation of HIPK2 may either occur once during translation at the level of transitional intermediates or alternatively autophosphorylation could be permanently performed by the full-length kinase. Since continuously ongoing phosphorylation can be conveniently detected upon inhibition of the antagonizing phosphatases, it was interesting to test the effect of the serine/threonine phosphatase inhibitor calyculin A on HIPK2 phosphorylation. Cells expressing wt or kinase-inactive HIPK2 were treated for various periods with calyculin A. Western blotting showed that phosphatase inhibition caused a further strong upshift of HIPK2-wt, while the electrophoretic migration of HIPK2-K221A was slowed only very slightly (Figure 2A). These data suggest that kinase activity of HIPK2 is required for the continuous phosphorylation of the kinase. It was then interesting to compare HIPK2 phosphorylation between the in vitro translated kinase and HIPK2 that was expressed in intact cells. The electrophoretic mobility of HIPK2 expressed in cells was slower than the in vitro translated material (Figure 2B), showing that only a part of the phosphorylation sites can be recapitulated in this in vitro system. Is HIPK2 autophosphorylation proceeding via cis- or trans-autophosphorylation? Since (at least transient) mutual binding is a prerequisite for trans-autophosphorylation, the occurrence of homotypic interactions was investigated. Differentially tagged HIPK2 proteins showed mutual binding in coimmunoprecipitation experiments (Figure 2C). The occurrence of trans-autophosphorylation would imply that active HIPK2 has the ability to phosphorylate a kinase-inactive version of HIPK2 that is defective in autophosphorylation. To test this experimentally, kinase-inactive GFP-HIPK2-K221A was coexpressed with various Flag-tagged HIPK2 forms that were either wt or mutated in the phosphorylated residues in the HIPK2 activation loop (Y354 and S357). This experiment showed that active HIPK2 was able to trans-phosphorylate kinase-inactive HIPK2, as revealed by the appearance of the upshifted phosphorylated form (Figure 2D). This approach also revealed that mutation of activation loop Y354 and/or S357 failed to induce this upshift, suggesting that these amino acids are essential for the ability to trans-autophosphorylate HIPK2.
HIPK2 is autophosphorylated in trans. (A) Cells were transfected to express Flag-HIPK2 or HIPK2-K221A and treated for increasing time periods with calyculin A (50 nM) as shown. Western blotting was used to investigate HIPK2 phosphorylation, as revealed by the occurrence of the slower migrating form. Phosphorylation of the NF-κB p65 subunit was detected as a positive control for calyculin A function. (B) The electrophoretic migration of HIPK2 produced by in vitro translation was compared with that of HIPK2 or HIPK2-K221A proteins that were derived from expression in 293T cells. HIPK2 was revealed with an anti-Flag antibody as shown. (C) Flag-HIPK2 or HA-HIPK2 was either expressed alone or together in 293T cells. Lysates were analyzed for adequate protein expression (input) or subjected to coimmunoprecipitation using the anti-Flag antibody or control mouse IgG, followed by immunoblotting. (D) GFP-HIPK2-K221A was coexpressed with Flag-HIPK2-wt or different Flag-HIPK2 point mutants as shown. Cells were lysed and analyzed by western blotting for HIPK2 phosphorylation with the indicated antibodies. The position of phosphorylated GFP-HIPK2-K221A is indicated.
Cis-autophosphorylation of Y354 in the activation loop of HIPK2
The occurrence of HIPK2 tyrosine phosphorylation was further investigated upon analysis of wt HIPK2 and mutants where the tyrosine residues contained in the activation loop were changed to phenylalanine. These HIPK2 forms were expressed in 293T cells, followed by IP of the epitope-tagged kinases and detection of tyrosine phosphorylation by immunoblotting (Figure 3A). These experiments showed that only mutation of Y354 largely impaired HIPK2 tyrosine phosphorylation. This could either mean that Y354 is the main phosphorylated tyrosine or alternatively that Y354 phosphorylation is required for tyrosine phosphorylation at other sites. Kinase-inactive HIPK2-K221A was not phosphorylated at Y354, showing the importance of HIPK2 kinase function for this modification. Tyrosine phosphorylation occurred also for endogenous HIPK2 in unstimulated cells (Figure 3B). Tyrosine phosphorylation was also detected for in vitro translated HIPK2 and the kinetics of this phosphorylation exactly paralleled that of HIPK2 production without any lag time (Figure 3C), thus corroborating the observations suggesting that HIPK2 Y354 is modified by autophosphorylation. The potential occurrence of trans-autophosphorylation in the activation loop was investigated upon measuring the ability of GFP-tagged HIPK2 to trigger phosphorylation of Y354 in kinase-inactive Flag-HIPK2. These experiments showed no trans-autophosphorylation of Y354 (Figure 3D), thus providing evidence that this modification is mediated by autophosphorylation in cis.
HIPK2 Y354 autophosphorylation in cis. (A) 293T cells were transfected to express different HIPK2 constructs that were either changed in the phosphorylated amino acids or in all tyrosines (3Y/F) contained in the activation loop. One part of the lysates was further analyzed by immunoprecipitation and western blotting with antibodies specific for phosphorylated tyrosine residues (upper). The remaining material was used for the input controls (lower). (B) 293T cells were lysed in NP-40 buffer and endogenous HIPK2 was immunoprecipitated using anti-HIPK2 antibodies. As a control a small fraction of the total cell lysate was incubated with rabbit IgG. Immunoprecipitates were separated by SDS–PAGE and immunoblotted with anti-HIPK2 and anti-phospho-tyrosine antibodies. (C) Flag-HIPK2 was translated in vitro for the indicated times. After immunoprecipitation with the anti-Flag antibody, precipitates were subjected to western blotting and analyzed with antibodies detecting the Flag-Tag and phosphorylated tyrosine as shown. (D) Flag-HIPK2-K221A was coexpressed with GFP-HIPK2 and GFP-HIPK2-K221A. After immunoprecipitation of Flag-tagged HIPK2, phosphorylation of Y354 was analyzed with the anti-phospho-tyrosine antibodies (upper). The lower part shows the input material.
Short forms of HIPK2 cis-autophosphorylate the activation loop
Since autophosphorylation of DYRK kinases is mediated by the nascent kinase (Lochhead et al., 2005), it was interesting to investigate whether C-terminally truncated forms of HIPK2 corresponding to incompletely translated proteins can also undergo autophosphorylation. To identify the minimal length of catalytically active HIPK2, mutants lacking various portions of the C-terminus were tested for their ability to trans-phosphorylate the Siah2 substrate protein. These experiments showed intact HIPK2 kinase activity for the version encompassing amino acids 1–520, while further deletion of C-terminal amino acids prohibited kinase activity (Figure 4A) and HIPK2 tyrosine phosphorylation (Supplementary Figure S1). To investigate whether the short active form of HIPK2 also undergoes tyrosine autophosphorylation, HIPK2 1–520 and longer forms were expressed in cells and isolated by IP. The material remained untreated or was incubated with λ phosphatase in order to analyze the accessibility of the modified Y354. Immunoblotting showed Y354 phosphorylation already for HIPK2 1–520 (Figure 4B), showing that tyrosine cis-autophosphorylation occurs for the C-terminally truncated HIPK2 form that mimics the incompletely synthesized protein. This experiment also revealed that λ phosphatase efficiently dephosphorylated Y354 contained in HIPK2 1–520 and HIPK2 1–768, while it failed to remove the phosphate group in HIPK2 1–916 where the kinase has presumably already adopted a different conformation that precludes access by the phosphatase.
HIPK2 tyrosine autophosphorylation in cis. (A) Flag-HIPK2 mutants lacking various parts of the C-terminus were coexpressed with the substrate protein Flag-Siah2-Rm and immunoblotted after cell lysis. Siah2 phosphorylation was revealed by the phosphorylation-dependent upshift of Siah2 or with a Siah2 phospho-specific antibody as shown. (B) Three Flag-HIPK2 truncation mutants were expressed and further analyzed by immunoprecipitation and treated with λ phosphatase as shown. Western blotting was used to reveal HIPK2 precipitation and tyrosine phosphorylation as shown. The lower part shows the input material. One representative experiment from n = 4 experiments is shown.
HIPK2 activation loop phosphorylation controls kinase activity and substrate-binding affinity
To investigate the relevance of activation loop phosphorylation on its function as a protein kinase, we compared HIPK2 and activation loop mutants thereof for their ability to phosphorylate the Siah2 protein, which is phosphorylated by HIPK2 at many residues which results in multiple upshifted versions (Calzado et al., 2009). In order to reveal potential differences in individual phosphorylation sites, it was important to use phospho-specific antibodies recognizing individual phosphorylated amino acids. Antibodies for phosphorylated Siah2-T26 and S68 were generated and their specificity was ensured in control experiments (Supplementary Figure S2). To test the importance of HIPK2 activation loop phosphorylation for kinase activity, cells were transfected to express the various HIPK2 mutants along with Siah2, followed by analysis of Siah2 phosphorylation by immunoblotting (Figure 5A). Expression of the wt kinase triggered massive Siah2 phosphorylation at many sites, as revealed by the almost quantitative upshift of the Siah2 band and also by phospho-specific antibodies recognizing Siah2-T26, S28, and S68. While HIPK2-K221A lacked any kinase activity, the tyrosine mutant HIPK2-Y354F showed an intermediate kinase activity. Similarly, HIPK2-S357A showed a reduced but significant activity as a protein kinase. Mutation of both residues resulted in an HIPK2 variant that had completely lost its ability to phosphorylate Siah2 at T26 and S28, while S68 was still phosphorylated to some extent. Similarly, HIPK2-Y354F/S357A expression still caused the occurrence of upshifted Siah2 bands, suggesting that phosphorylation of HIPK2-Y354 and S357 does not function as an ‘on-off switch’ but rather controls kinase activity and specificity. The distinct activities of the HIPK2 activation loop mutants were also recapitulated in an in vitro system where the bacterially expressed and purified glutathione S-transferase (GST)-Siah2 was differentially phosphorylated by HIPK2 and its activation loop mutants irrespective of whether the kinases were obtained from immunoprecipitated (Figure 5B) or from in vitro translated (Figure 5C) material. Comparison of kinase activities between HIPK2 variants using Siah2-S68A as a substrate revealed an impaired but significant Siah2-T26 and S28 phosphorylating activity of HIPK2-Y354F and HIPK2-S357A (Supplementary Figure S3). These data show that phosphorylation of each site alone is not required for HIPK2 activity per se, while mutation of both sites largely abolished kinase activity.
Activation loop phosphorylation determines strength of substrate phosphorylation. (A) An expression plasmid for Flag-Siah2 was transfected into 293T cells along with indicated Flag-HIPK2 constructs. During the last 12 h before lysis, the cells were grown in the presence of the proteasome inhibitor MG-132 (5 µM) in order to inhibit Siah2-mediated HIPK2 ubiquitination/degradation. HIPK2-mediated Siah2 phosphorylation was detected with the different phospho-specific antibodies. (B) Flag-tagged HIPK2 variants were expressed in 293T cells and isolated by immunoprecipitation. Kinase assays (KAs) were performed upon incubation of the immunoprecipitated material with recombinant and purified GST-Siah2 substrate protein and ATP. HIPK2 substrate phosphorylation efficiency was revealed by the anti-phospho-Siah2-S28 antibody after western blotting. The lower part shows the recombinant substrate protein stained by Coomassie brilliant blue (CBB) and the positions of molecular weight markers are indicated. (C) The experiment was done as in B with the exception that HIPK2 was produced by in vitro translation.
HIPK2 activation loop phosphorylation controls substrate binding
In order to obtain insight into the structural parameters underlying phosphorylation-dependent substrate specificity, the structure of HIPK2 was modeled according to the crystal structures of the related kinases DYRK1a (Ogawa et al., 2010) and DYRK2 (PDB 3KVW) (Figure 6A). Structural comparison and sequence alignment showed the conservation of the residues forming the activation loop and of the ATP-binding pocket. The phosphorylated S357 of HIPK2 is in close proximity to K319, which is sandwiched by the two acidic amino acids D317 (proton acceptor) and E321. Phosphorylated Y354 is closer to the surface and surrounded by R358, R361, and E399, which interacts with K423. Given that activation loop phosphorylation can affect substrate binding, we then performed an in silico analysis that compared the electrostatic charge in a surface region adjacent to the ATP-binding pocket between HIPK2 versions representing the unphosphorylated and the phosphorylated activation loop (Figure 6B). The models visualize the predicted consequences of activation loop phosphorylation on the surface charge of HIPK2, which in turn can affect the ability to bind the unmodified substrate and to release the phosphorylated substrate at the end of the catalytic cycle (Johnson et al., 1996; Nolen et al., 2004). To compare the influence of HIPK2 kinase activity and its autophosphorylation for their substrate-binding affinities, coimmunoprecipitation experiments were performed. Cells were transfected to express HIPK2, HIPK2-Y354F/S357A, or HIPK2-K221A along with Siah2. The results from the coimmunoprecipitation experiments showed an inverse correlation between the kinase activity and the ability to bind Siah2 (Figure 6C). Enhanced interaction between Siah2 and HIPK2-Y354F/S357A could be either due to absent HIPK2 activation loop phosphorylation and resulting conformation changes in the kinase or alternatively to the impaired phosphorylation of Siah2. To distinguish between these two possibilities, we determined the interaction between the HIPK2 variants and a phosphomimetic Siah2 mutant where the three known (Calzado et al., 2009) and two newly identified phosphorylation sites (M. Lienhard Schmitz, unpublished data) were changed to aspartic acid (Siah2-5D). Coimmunoprecipitation experiments revealed an inverse correlation between HIPK2 kinase activity and the ability to bind to the phosphomimetic Siah2 substrate (Figure 6D). These data suggest that HIPK2 activation loop phosphorylation has an impact on the kinase activity, which then in turn controls the affinity to the substrate.
Structural model of the HIPK2 kinase domain. (A) The published DYRK1a and DYRK2 structures were used as templates in order to calculate the structure of HIPK2 using the Swiss-Model server. The activation loop is colored yellow, with important residues shown as ball and stick and phosphorylations are shown in red. The ATP-binding site is marked by an indirubin ligand (blue), taken from the superimposed DYRK2 structure. The positions of amino acids in the vicinity of Y354 and S357 are indicated. (B) The electrostatic potential of the molecular surfaces of unphosphorylated and phosphorylated HIPK2 are color coded. (C) The indicated Flag-HIPK2 constructs were coexpressed with Flag-Siah2 in 293T cells. One day after transfection cells were grown in the presence of the proteasome inhibitor MG-132 in order to prevent Siah2-mediated HIPK2 degradation. Cells were lysed and HIPK2 was immunoprecipitated with anti-HIPK2 antibodies, followed by detection of proteins with the indicated antibodies. (D) The experiment was done as in C with the exception that binding to expressed Siah2-5D was tested.
Activation loop phosphorylation is important for multiple biological functions of HIPK2
Since the kinase activity of HIPK2 influences its intracellular localization (Moller et al., 2003), it was interesting to investigate the consequences of missing activation loop phosphorylation on the intracellular distribution of the kinase. Cells were transfected to express moderate amounts of the wt kinase, HIPK2-K221A, or the activation loop mutant HIPK2-Y354F/S357A, followed by immunofluorescence analysis (Figure 7A). HIPK2-wt was mainly found in nuclear speckles and HIPK2-K221A localized to the nucleoplasm, which is in accordance with the published data (Moller et al., 2003). In contrast, a large portion of the activation loop mutant HIPK2-Y354F/S357A showed a cytosolic distribution that resembles the localization of HIPK2 in the presence of its kinase inhibitor D-115893 (de la Vega et al., 2011) or the cytoplasmic delocalization of HIPK2 occurring in human breast carcinomas (Pierantoni et al., 2007). Since HIPK2 is an important regulator of gene expression, it was then interesting to investigate the functional consequences of activation loop phosphorylation on the transcription of HIPK2-dependent genes. HIPK2-deficient mouse embryonic fibroblasts (MEFs) were stably reconstituted in order to express physiological levels of HIPK2-wt or HIPK2-Y354F/S357A and analyzed by quantitative polymerase chain reaction (qPCR) for the expression of endogenous genes known to be regulated by HIPK2 (Hattangadi et al., 2010; Puca et al., 2010). While the wt kinase suppressed transcription of the early growth response protein 1 (Egr1) and peroxisomal acyl-coenzyme A oxidase 1 (Acox1) genes, the activation loop mutant had completely (Acox1) or partially (Egr1) lost the gene-repressing ability (Figure 7B). The differentially reconstituted MEFs were then compared for potential differences in cell proliferation (Figure 7C). Consistent with the published antiproliferative effect of HIPK2 (Wei et al., 2007), cells expressing this kinase grew significantly slower when compared with HIPK2−/− cells. The antiproliferative effect of HIPK2 was lost in HIPK2-Y354F/S357A expressing cells, indicating that HIPK2 activation loop phosphorylation is required for its function in cell cycle regulation. In order to reveal the importance of activation loop phosphorylation on inducible gene expression, HIPK2 and its mutant derivatives were compared for their ability to trigger IPF1/PDX1-induced transcription (Boucher et al., 2009). Cells were transfected with an IPF1/PDX1-dependent reporter construct along with an expression vector for IPF1/PDX1 and various HIPK2 expression constructs (Figure 7D). Analysis of luciferase activity showed that HIPK2-mediated enhancement of IPF1/PDX1 activity was impaired upon mutation of the individual phosphorylation sites in the activation loop. Mutation of both residues largely abrogated the stimulatory effect of HIPK2 and only allowed some residual activity. The functions of the HIPK2 phosphorylation sites outside of the activation loop are not known, but phosphorylation events can allow docking of further proteins such as 14-3-3 family members or the prolyl isomerase Pin1 (Verdecia et al., 2000; Gardino and Yaffe, 2011). Since Pin1 is important for HIPK2-mediated phosphorylation of p53 at S46 (Grison et al., 2011), we were tempted to test binding of HIPK2 and its mutants to Pin1 in GST pull-down experiments. These experiments showed robust binding of wt HIPK2 to Pin1, while HIPK2-K221A and HIPK2-Y354F/S357A showed no or strongly reduced interaction with this prolyl isomerase (Figure 7E).
Functional consequences of activation loop phosphorylation. (A) HIPK2-deficient MEFs were transiently transfected to express the indicated HIPK2 forms. Upper: The intracellular localization of HIPK2 was analyzed by indirect immunofluorescence. Nuclear DNA was visualized with Hoechst (blue), and representative localization for HIPK2 forms is shown. Lower: quantification of intracellular HIPK2 distribution, as revealed by microscopical analysis of 200 cells for each construct. (B) HIPK2−/− cells were lentivirally transduced to express comparable levels of HIPK2 wt and HIPK2-Y354F/S357A as shown (lower). The cells were analyzed for the expression of the HIPK2 target genes by real-time PCR as shown. In order to facilitate comparison, gene transcription in the knockout cells was arbitrarily set as 100%. Experiments were performed in triplicates, error bars display standard deviations. (C) Equal numbers of the differentially reconstituted MEFs were seeded into 6-well dishes and cell numbers were counted after several time points in a flow cytometer. The graph shows the increase of cell numbers over time, and error bars show standard errors of the mean of two independent experiments performed in triplicate. (D) HEK293T cells were transfected with a reporter gene under the control of five binding sites for IPF/PDX1 together with expression vectors for IPF/PDX and the indicated HIPK2 variants. Cells were lysed in NP-40 buffer and luciferase activity was measured in a luminometer. Basal luciferase activity without coexpression of IFP1/PDX1 and HIPK2 was set as 1, bars indicate standard deviation from two experiments measured in duplicates. (E) HIPK2 binds Pin1 in vitro. Total cell extracts from cells transfected to express Flag-HIPK2 and mutants thereof were tested for protein expression (input) and for binding to bacterially produced GST-Pin1 by pull-down experiments as shown. The upper part shows a western blot displaying eluted HIPK2 and the CBB-stained GST fusion proteins used for this experiment.
Discussion
The activities of protein kinases must be tightly controlled, as unrestrained kinase activities as they occur for example in the oncogenic mutant versions of Src or Raf-1 kinases can facilitate tumor development (Mansour et al., 1994). Given the relevance of HIPK2 for central signaling pathways, its activity is regulated by several mechanisms. These include binding to the scaffold protein Han11 to control the threshold, amplitude, and kinetics of HIPK2-triggered signaling events (Ritterhoff et al., 2010). Also the intracellular localization of HIPK2 can be regulated, as aberrant expression of the high-mobility group A1 protein, which can occur in human cancers, promotes relocalization of HIPK2 to the cytoplasm (Pierantoni et al., 2007). Additionally, HIPK2 is prominently governed at the level of protein stability by the ubiquitin/proteasome system, since four different ubiquitin E3 ligases (MDM2, Siah, WSB-1, and Fbx3-formed SCF) ensure the tight limitation of HIPK2 amounts (Gresko et al., 2006; Rinaldo et al., 2007; Choi et al., 2008; Shima et al., 2008; Calzado et al., 2009). The need to control HIPK2 abundance is corroborated by this study, which shows that production of HIPK2 per se is sufficient to generate a kinase with phosphorylated residues in the activation loop and thus competent for kinase activity.
Our data show HIPK2 phosphorylation throughout the entire kinase and a comparison with data retrieved from the public PhosphoSitePlus© database confirmed the identified phosphorylation sites at Y354, S357, S819/320, and T831. The database also shows phosphorylation at T789, but intriguingly also documents phosphorylation at the activation loop residues S352, T353, and Y359/360 (Supplementary Figure S4). We also observed phosphorylation at T353 in one experiment, raising the possibility that more residues in the activation loop can be phosphorylated and thus allow further control of HIPK2 activity. While our data show no relevance of Y359/360 for HIPK2 activity (see Figure 5A), it will be interesting in future experiments to quantify the possible occurrence and regulation of S352 and T353 phosphorylation using stable isotope labeling by amino acids in cell culture experiments. While our experiments detected basal phosphorylation of HIPK2 in unstimulated cells, it will also be relevant to map the phosphorylation pattern of this kinase under conditions (such as DNA damage or hypoxia) known to regulate its activity (Rinaldo et al., 2008; Calzado et al., 2009). Our results do not exclude the possibility of further activation loop regulation by additional kinases or phosphatases. Such an example for the occurrence of two independent mechanisms mediating activation loop phosphorylation is provided by the MAPK p38α. This kinase can be activated through the classical MAPK pathway by its activation loop phosphorylation mediated by the upstream kinases MKK3 or MKK6. On the other hand, activation of p38α can also be induced by its interaction with TAB1 (transforming growth factor-β-activated protein kinase 1-binding protein 1), which promotes p38α activation loop autophosphorylation (Ge et al., 2002). However, once phosphorylated, activation loops can also be principally subjected to phosphatase-mediated dephosphorylation (Sangwan et al., 2008; Chabot et al., 2009), so that future work can address the question of whether additional enzymes may act on the HIPK2 activation loop. Both cis-autophosphorylated residues in the activation loop do not correspond to the consensus phosphorylation site of the mature full-length kinase. Of note, the observation that activation loop autophosphorylation differs from substrate phosphorylation has also been described for other kinases (Pike et al., 2008). The exact mechanism by which HIPK2 autophosphorylates its activation loop in cis is not clear, but it is likely that Y354 phosphorylation occurs by a mechanism that has been observed for DYRKs (Lochhead et al., 2005). While GSK3 autophosphorylation depends on the chaperone activity of heat shock protein 90 kDa (Hsp90) (Lochhead et al., 2006), HIPK2 autophosphorylation still occurs in the presence of Hsp90 inhibitors (Supplementary Figure S5). HIPK2 phosphorylation in regions outside of the activation loop can proceed via trans-autophosphorylation (see Figure 2D) but probably also by cis-autophosphorylation. The reduced phosphorylation of in vitro translated HIPK2 is compatible with the possibility that other kinases also contribute to the modification of HIPK2. Candidates are known interactors such as DYRK kinases and also the kinase TAK1 and its adapter protein TAB1 (Kanei-Ishii et al., 2004; Ritterhoff et al., 2010), but in preliminary experiments their expression did not result in the occurrence of upshifted and phosphorylated HIPK2 (Supplementary Figures S6 and S7). Phosphorylation of S357 is not a prerequisite for Y354 phosphorylation, as revealed by efficient tyrosine phosphorylation of in vitro translated HIPK2-S357A (Supplementary Figure S8). However, phosphorylation of the activation loop is critical for phosphorylations occurring in the remaining part of the kinase, as also seen by the increased electrophoretic mobility of HIPK2 activation loop mutants (see also Figure 5A).
Since the phosphorylation sites are found in various domains such as the kinase domain, the autoinhibitory domain, and also in the homeobox interaction domain, many functional consequences are possible. Our data show that HIPK2 phosphorylation allows the binding to Pin1 that recognizes phosphorylated serine or threonine residues followed by proline, thus resulting in cis/trans-isomerization of the bound protein (Verdecia et al., 2000). The identification of the responsible residue(s) allowing docking of phosphorylated HIPK2 to Pin1 is beyond the scope of this manuscript, but it will be interesting to study the mechanisms and consequences of Pin1-mediated HIPK2 isomerization in the future. Of note, a HIPK2 mutant, carrying serine and threonine to alanine mutations of the putative phosphorylation sites at position S111, S114, S287, S434, T443, and T510, did not show any defects in substrate phosphorylation (data not shown), indicating that these sites are at least not important for HIPK2 kinase activity. HIPK2 and DYRKs share the modification of a critical tyrosine in the activation loop by autophosphorylation. However, while DYRK2 is completely inactive without this modification (Lochhead et al., 2005), the impact of Y354 phosphorylation on HIPK2 activity depends on the substrate protein and resembles more the Y321 of DYRK1A that regulates kinase activity without being strictly essential (Himpel et al., 2001; Adayev et al., 2007). Given the growing number of pathways involving HIPK2 activity, there is a need to achieve high specificity of its activity. We hypothesize that complex phosphorylation patterns of HIPK2 contribute to the specification of its substrate recognition and phosphorylation activity.
Materials and methods
Plasmids
The expression vectors for pcDNA3-Flag-HIPK2 (wt and K221A), pEF-HA-HIPK2, pEGFP-HIPK2 (wt and K221A), pCMV-Siah2, pCMV-Siah2 5D (Calzado et al., 2009; Pérez et al., 2012), and GST-Pin1 (Yi et al., 2005) were described elsewhere. HIPK2 point mutants, HIPK2 C-terminal truncation mutants, and Siah2-Rm (H98A/C101A) were generated using the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. The IPF1/PDX1-dependent reporter construct 5× P1-luc and the expression vector for IPF1/PDX1 were described (Boucher et al., 2009). Lentiviral expression of HIPK2 was performed by cloning HIPK2 or its mutants in the vector 290-pHAGE-hEF1a CAR-PGK Puro. The packaging vectors pMDLg/pRRE, pRSV-Rev, and pHCMVG were from Addgene.
Antibodies and reagents
Mouse monoclonal antibodies against Flag (M2) and Tubulin (Tub2.1) were purchased from Sigma. Antibodies recognizing HA (3F10) and GFP (clones 7.1 and 13.1) (Roche Applied Science), anti-phospho-tyrosine (4G10) (Millipore), phospho-p38 T180/Y182 (#9211), and phospho-NF-κB p65 S536 (#3031) (Cell Signaling Technology) were from the indicated suppliers. Rabbit polyclonal antibodies recognizing phosphorylated Siah2 at T26 and S68 were generated by PolyPeptide Laboratories (Strasbourg). The anti-HIPK2 and anti-phospho-Siah2 (S28) antibodies were previously described (Calzado et al., 2009). Secondary horseradish peroxidase (HRP)-coupled antibodies against mouse or rabbit immunoglobulin G (IgG) were obtained from Dianova, the HIPK2 inhibitor D-115893 was previously described (de la Vega et al., 2011) and MG-132 was purchased from Sigma.
Cell culture and transient transfection
HEK293, 293T, Hela cells, and HIPK2-deficient MEFs were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS), 2 mM l-glutamine, and 1% (v/v) penicillin/streptomycin. HEK293 and Hela cells were transfected with Rotifect according to the manufacturer's instructions (Roth). 293T cells were transiently transfected using polyethylenimine (PEI; Sigma) (Ehrhardt et al., 2006). In brief, 293T cells were seeded in 6-cm dishes and were transfected the day after using 2 µl of PEI (1 mg/ml H2O, pH 7.0) per µg plasmid DNA. DNA:PEI complexes were formed in 200 µl serum- and antibiotic-free DMEM during 20 min of incubation at room temperature. Growth medium from cells was removed and replaced by 2 ml of antibiotic-free DMEM with a normal FCS concentration. After adding the transfection mixture, the cells were incubated 3–5 h before the medium was changed.
Production of lentiviruses and reconstitution of HIPK2-deficient MEFs
293T cells, seeded in 10-cm dishes, were transfected with lentiviral vectors of the third generation and p290-Flag-HIPK2 constructs using Lipofectamine 2000 (Invitrogen). Forty-eight hours post-transfection, viruses were collected and filtered through a 0.45 µm filter. HIPK2-deficient MEFs were infected in the presence of 5 µg/ml polybrene (Sigma). One day after transduction, reconstituted MEFs were selected with 3 µg/ml puromycin (Invivogen).
In vitro transcription and translation
Flag-HIPK2 constructs were expressed in vitro using the TNT T7-coupled Reticulocyte Lysate System (Promega) for 90 min or indicated times. In vitro translated Flag-HIPK2 was immunoprecipitated, directly analyzed by western blotting or subjected to in vitro phosphorylation assays.
Cell lysis, IP and western blotting
Transfected 293T cells were lysed in NP-40 buffer as described (Renner et al., 2011). The lysates were cleared by centrifugation and proteins were separated by SDS–polyacrylamide gel electrophoresis (PAGE) and blotted onto polyvinylidene fluoride membrane (Millipore). After incubation with primary and secondary HRP-coupled antibodies, proteins were visualized using the Western Lightning ECL solutions (Perkin Elmer). For IP cleared cell lysates or in vitro translated Flag-HIPK2 were supplemented with 1 µg precipitating antibody and 25 µl protein A/G agarose (Santa Cruz) and incubated for 4 h at 4°C. Agarose beads were then washed four times with 1 ml cold NP-40 buffer and precipitated proteins were eluted by adding 2× SDS sample buffer.
λ phosphatase treatment
Immunoprecipitated Flag-HIPK2 was incubated for 30 min at 30°C in the appropriate reaction buffer supplemented with 1 mM MnCl2 and 400 U λ phosphatase (New England Biolabs). The samples were washed twice with 1 ml NP-40 buffer and boiled in SDS sample buffer.
Purification of GST fusion proteins, GST pull-down experiments, and in vitro phosphorylation
Recombinant GST fusion proteins were expressed in E. coli BL21 and purified by standard protocols using GSH Sepharose (GE Healthcare). Pull-down experiments were done after dialysis of proteins against PBS buffer. HIPK2 and its derivates were produced by expression in transiently transfected HEK293T cells. Cell lysates were incubated with 3 µg of GST or the GST fusion proteins, respectively, followed by extensive washing in PBS buffer and elution with 1× SDS sample buffer. Eluates were further analyzed by immunoblotting, while the recombinant GST proteins were controlled by SDS–PAGE and Coomassie staining. In vitro phosphorylation experiments were done after production of Flag-HIPK2 in cells or by in vitro translation. The kinase reaction was performed by incubating immunoprecipitated Flag-HIPK2 together with GST-Siah2 in kinase reaction buffer (40 µM ATP, 25 mM HEPES/KOH, pH 7.6, 25 mM β-glycerophosphate, 20 mM MgCl2, 2 mM DTT) for 25 min at 30°C. The reaction was stopped by adding SDS sample buffer and subsequent boiling for 5 min. The samples were analyzed by western blotting.
Quantitative real-time PCR
Total RNA was isolated from reconstituted HIPK2−/− MEFs using the RNeasy mini kit according to the manufacturer's instructions (Qiagen). One microgram total RNA was reverse transcribed using Oligo(dT)12–18 primer and SuperScript II reverse transcriptase (Invitrogen). Quantitative real-time PCR was performed using the following primer: Acox1-F, CCGCCACCTTCAATCCAGAG; Acox1-R, CAAGTTCTCGATTTCTCGACGG; Egr1-F, CAGGAGTTGGAGTGTTGTGG; Egr1-R, TATCCCATGGGCAATAGAGC. Experiments were done in triplicate and data were normalized to the housekeeping gene β-actin.
Indirect immunofluorescence staining
Hela cells were seeded on cover slips in a 12-well plate and transfected. For immunofluorescence staining, cells were washed twice with phosphate-buffered saline (PBS) and fixed for 1 min with cold methanol/acetone (1:1). After air drying, the cells were rehydrated with PBS and blocked with 10% (v/v) goat serum (Sigma) in PBS for 60 min at room temperature. The primary and secondary antibodies were incubated in 1% (v/v) goat serum overnight at 4°C or 2 h at room temperature, respectively. After washing the cells with PBS, nuclei were stained with Hoechst 33342 (Invitrogen) and cells were mounted with Kaiser's glycerol gelatine (Merck). Cells were analyzed with a Nikon eclipse TE2000-E microscope.
Luciferase reporter assay
HEK293 cells were seeded on in 6-well plates and transfected with the specified DNA constructs using Rotifect (Roth). Total DNA amounts were kept equal in all transfections by adding empty pcDNA3 vector. Cells were lysed 32 h post-transfection in 200 µl NP-40 buffer and cleared by centrifugation. Ten microliters of the supernatant were mixed with 10 µl of luciferase buffer and bioluminescence was immediately measured for 10 sec in a luminometer (Berthold DuoLumat LB 9501).
Modeling of HIPK2 structures
The structure of HIPK2 was modeled according to the crystal structure of DYRK1a (unpublished data, PDB code: 2vx3) using the Swiss-Model automated comparative protein modeling server (Arnold et al., 2006). Both kinases share a sequence identity of 35.5% in the modeled range (HIPK2: 165–521; DYRK1: 135–481). Phosphorylation and subsequent refinement of Y354 and S357 was done with the interactive graphics program Coot (Emsley and Cowtan, 2004). Superimposing of DYRK1a and DYRK2 structures was done with secondary-structure matching (Krissinel and Henrick, 2004). Molecular graphics images showing the electrostatic potential were produced using the UCSF Chimera package (Pettersen et al., 2004).
Mass spectrometry
Cells were transfected to express HA-HIPK2 or HA-HIPK2 K221A, followed by IP with anti-HA antibodies in a buffer containing 10 mM sodium fluoride and 0.5 mM sodium orthovanadate in order to maintain phosphorylation. HIPK2 gel bands were excised and subjected to in-gel digestion with trypsin (Shevchenko et al., 2006). The resulting tryptic peptides were extracted with acetonitrile and desalted with reversed phase C18 STAGE tips (Rappsilber et al., 2007). Mass spectrometric experiments were performed on a nano-flow high-performance liquid chromatography system (Proxeon) connected to LTQ-orbitrap and Velos LTQ-Orbitrap-XL instruments (Thermo Fisher Scientific) equipped with a nanoelectrospray source (Proxeon).
Supplementary material
Supplementary material is available at Journal of Molecular Cell Biology online.
Funding
This work was supported by grants from the German Research Foundation projects (SCHM 1417/4-2, SCHM 1417/7-1, SCHM 1417/8-1, GRK 1566/1, SFB/TRR 81), the Excellence Cluster Cardio-Pulmonary System (ECCPS), German Academic Exchange Service (A/08/98404) the LOEWE/UGMLC program (MLS). K.B. was supported by the LOEWE Research Focus ‘Insect Biotechnology’.
Conflict of interest: none declared.
Acknowledgements
We are grateful to Daniela Stock (Justus Liebig University, Giessen) for technical assistance. We thank Dr Silvia Soddu (Istituto Nazionale dei Tumori Regina Elena, Rome) for sharing unpublished observations, Dr Issay Kitabayashi (National Cancer Center Research Institute, Tokyo) for providing HIPK2-deficient MEFs and Dr Irene Seippelt (Æterna Zentaris, Frankfurt) for D-115893.
