HIPK2 catalytic activity and subcellular localization are regulated by activation-loop Y354 autophosphorylation

Abstract

HIPK2 (homeodomain-interacting protein kinase-2) binds to and phosphorylates, at Ser and Thr residues, a large number of targets involved in cell division and cell fate decision in response to different physiological or stress stimuli. Inactivation of HIPK2 has been observed in human and mouse cancers supporting its role as a tumor suppressor. Despite the biological relevance of this kinase, very little is known on how HIPK2 becomes catalytically active. Based on sequence homologies, HIPK2 has been taxonomically classified as a subfamily member of the dual-specificity tyrosine-regulated kinases (DYRKs) and the activation-loop Y354 of HIPK2 has been found phosphorylated in different cells; however, the relevance of this Y phosphorylation is presently unknown. Here, we show that HIPK2, which is extensively phosphorylated at S/T sites throughout its functional domains, becomes catalytically active by autophosphorylation at the activation-loop Y354. In particular, we found that, in analogy to DYRKs, HIPK2-Y354 phosphorylation is an autocatalytic event and its prevention, through Y354 substitution with non-phosphorylatable amino acids or by using the kinase inhibitor purvalanol A, induces a strong reduction of the HIPK2 S/T-kinase activity on different substrates. Interestingly, at variance from DYRKs, inhibition of HIPK2-Y354 phosphorylation induces a strong out-of-target Y-kinase activity in cis and a strong cytoplasmic relocalization of the kinase. Together, these results demonstrate that the catalytic activity, substrate specificity, and subcellular localization of HIPK2 are regulated by autophosphorylation of its activation-loop Y354.

Highlights

► HIPK2 is catalytically activated by activation-loop Y354 autophosphorylation. ► Inhibition of HIPK2-Y354 phosphorylation results in out-of-target Y kinase activity. ► Out-of-target Y kinase activity delocalizes HIPK2 into cytoplasmic aggresomes. ► HIPK2 cytoplasmic aggresomes are reminiscent of similar observations in human cancers.

1. Introduction

HIPK2 is considered a nuclear S/T kinase that binds and phosphorylates a still enlarging body of targets involved in the regulation of cell survival and proliferation during development and in response to genotoxic damage or hypoxia [1–4]. Among these HIPK2 targets, there are several transcription and cotranscription factors, [5–8], signal transducers and scaffold proteins [9–14], oncogenes and tumor suppressors [15–17], and E3 components of SUMO and ubiquitin ligases [18–22]. The relevance of these multiple HIPK2 interactions is highlighted by the finding that HIPK2 can work as a haploinsufficient tumor suppressor in mice [23] and a few mechanisms of HIPK2 inactivation have been identified in human cancers, such as HIPK2 forced cytoplasmic relocalization [15,24], HIPK2 mutations [25], and allele-specific loss of heterozygosity [26].

HIPK2 is a member of the evolutionarily conserved family of homeodomain-interacting protein kinases (HIPKs) that includes three highly homologous S/T kinases originally identified for their capacity to interact with homeodomain transcription factors (i.e., HIPK1, HIPK2, and HIPK3) [27], and a fourth member (HIPK4), that has been discovered through in silico analysis of the human kinome [28]. Based on the homology among catalytic domains, HIPKs have been taxonomically classified as a subfamily of the dual-specificity tyrosine-regulated kinase (DYRK) family that also includes two other subfamilies, the DYRK kinases and the pre-mRNA processing 4 kinases [28]. Alignment of HIPKs' catalytic domains with several DYRKs allowed the identification of evolutionarily conserved consensus motives, including the catalytic loop with a Lys residue that contacts the primary phosphate (K221 in HIPK2), experimentally shown to be required for HIPK2 kinase activity [27], and a Tyr residue (Y354 in HIPK2) located in the context of the activation-loop [29,30].

Phosphorylation of the activation-loop by upstream kinases, as in the MAPK signaling cascade, or by autophosphorylation, as in the DYRK subfamily [31], is one of the key regulatory mechanisms for catalytic activation of protein kinases [32,33]. Worth mentioning, DYRKs are defined as dual-specificity protein kinases because they become catalytically active by autophosphorylation of Tyr residues in their own activation-loop, while they phosphorylate their substrates only on Ser and Thr residues [34].

Although different posttranslational modifications, such as sumoylation, ubiquitylation, acetylation, and caspase cleavage, and interactions with scaffold proteins have been shown to regulate the cellular activities of HIPK2 [35,36], the mechanism through which HIPK2 becomes catalytically active is poorly understood. What is known is that HIPK2 can autophosphorylate in vitro while its kinase-defective (KD) derivative, obtained by substitution of the phosphate-contact Lys at position 221 with an Arg (HIPK2^K221R), has a strongly reduced kinase activity on both exogenous targets and HIPK2 itself [5,6,27]. Starting from these observations, here we provide strong evidence that autophosphorylation at the activation-loop Y354 promotes the catalytic activation of HIPK2 and regulates its substrate specificity and subcellular localization. Indeed, inhibition of Y354 phosphorylation promotes a strong out-of-target Y autophosphorylation in cis that results in accumulation of HIPK2 into cytoplasmic aggresomes, reminiscent of similar observations in human cancers [15,24].

2. Materials and methods

2.1. Expression vectors and transfection

The following plasmids were employed: pGEX-HIPK2 [5]; pEGFP-HIPK2, pEGFP-K221R [19]; pcDNA3GST-HIPK2 and pcDNA3GST-K221R [37]. pEGFP-Y354F, pEGFP-Y354A, and pCDNA3GST-Y354F were obtained by QuickChange site-directed mutagenesis (Stratagene). pcDNA3GST-Y354F(Δ1050–1189) was obtained by SalI digestion of the pcDNA3GST-Y354F vector to remove the 1050–1189 amino acid fragment. To obtain pcDNA3GST-HIPK2(784–1189), DNA fragment 784–1189 of murine Hipk2 cDNA was amplified from pBKS-clone#46 [5] by PCR and cloned into pCDNA3GST vector as previously described [37]. To obtain pGEX-HIPK2(158–557), pGEX-K221R(158–557) and pGEX-Y354F(158–557), the 158–557 fragments were amplified from the corresponding pEGFP vectors by PCR and cloned into EcoRI/XhoI digested pGEX6p-2rbs (kindly provided by A. Musacchio). His-HIPK2^(158–557) was cloned from SalI-HindIII into the pQE30 expression vector (Qiagen) and purified using Ni-NTA metal-affinity chromatography (The Qiaexpressionist, Qiagen). All constructs were controlled by sequencing. pEGFP-DYRK1A and pEGFP-Y321F were kindly provided by W. Becker and pEGFP-NF-YA by G. Piaggio. The expression vectors were transfected by Lipofectamine LTX and Plus reagent (Invitrogen).

2.2. Cells, culture conditions, and recombinant vaccinia virus

Human H1299, HEK293, U2OS, RKO and HEL cells, hTert-immortalized human fibroblasts [17], and p53-null mouse embryo fibroblasts were cultured in DMEM or RPMI supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen). Purvalanol A, TBB (4,5,6,7-tetrabromo-2-azabenzimidazole), and myelin basic protein (MBP) were obtained from Sigma-Aldrich. The His-tagged kinase domain of HIPK2 expressed in Sf21 insect cells, His-HIPK2^(158–557) was purchased from Millipore, the Mito Tracker Red from Invitrogen, and MG132 from Calbiochem. Recombinant vaccinia virus vTF7-3 carrying the bacteriophage T7 gene 1, kindly provided by Dr. B. Moss, was propagated in HEK293 cells and used as described [38].

2.3. His-purification, GST-pulldown, immunoprecipitation, λPPase treatment and kinase assay

Affinity adsorption to Ni-NTA agarose (Qiagen) and Glutathione-Sepharose 4 Fast Flow (GE Healthcare) were performed accordingly to the manufacturer's instructions. For immunoprecipitation, total cell extracts (TCEs) were prepared in lysis buffer [50 mM Tris–HCl pH 7.5, 5 mM EDTA pH 8, 300 mM NaCl, 1% NP40, 50 mM NaF, and 2 mM NaOV₄]. Dephosphorylation was performed by λPPase (New England Biolabs). After λPPase treatment, the phosphatase was washed out, phosphatase inhibitors (Pierce) were added, and kinase assays performed as reported [17].

2.4. Mass spectrometry (MS) of HIPK2 phosphorylated sites

Purified HIPK2 recombinant proteins were resolved by SDS-PAGE, detected by Coomassie Colloidal Blue staining (Invitrogen) and the appropriate bands were excised and subjected to in-gel digestions by sequencing grade modified porcine trypsin (Promega) or chymotrypsin (Sigma-Aldrich) as described [39]. To recognize phosphorylated peptides, digestion mixtures were co-crystallized with the matrix, α-cyano-4-hydroxycinnamic acid (Sigma-Aldrich) on the MALDI-TOF plate. Analyses were performed by a Voyager-DE STR MALDI-TOF (Applied Bio-systems) operating in reflectron mode. Spectra were internally calibrated and processed by Data Explorer software.

To identify phosphorylated sites, peptides were analyzed by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Peptide mixtures were separated using an Ultimate 3000 HPLC (DIONEX) connected on line with a linear ion trap (LTQ, Thermo Electron). Peptides were separated in a reverse phase, 10-cm capillary column. A data-dependent strategy was used to fragment the five more intense ions present in each full MS scan by collision-induced dissociation. Tandem mass spectra were interpreted through the SEQUEST algorithm [40], taking into account the potential for phosphorylation on Ser, Thr, or Tyr residues. A MS/MS was considered legitimately matched with cross-correlation scores of 1.8, 2.5, and 3 respectively for one, two, and three charged peptides and a probability cut-off for randomized identification of p < 0.001.

2.5. RNA extraction and quantitative real-time RT-PCR

RNA extraction and real-time RT-PCR were performed as described [17].

2.6. Protein extraction and Western blotting

TCEs were prepared in lysis buffer [50 mM Tris–HCl (pH 8), 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% NP40, 1 mM EDTA, and 2 mM Na₃VO₄] supplemented with protease-inhibitor mix (Roche). Nuclear/cytoplasmic differential extractions were performed using NE-PER™ Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific) according to the manufacturer's instructions. Soluble and insoluble fractions were obtained as described [41]. Proteins were resolved by SDS-PAGE using NuPAGE® Novex Bis-Tris Gels (Invitrogen), transferred onto nitrocellulose membranes (Bio-Rad), and analyzed with the indicated antibodies (Abs). The following Abs were employed: α-pTyr PY99 monoclonal Ab (moAb) (Santa Cruz Biotechnologies), α-pTyr 4G10 moAb (Upstate), rabbit α-HIPK2 Ab (kindly provided by M.L. Schmitz), α-alpha-tubulin moAb (Immunological Sciences), α-GFP moAb (Santa Cruz Biotechnologies), rabbit α-HDAC-1 Ab (Sigma-Aldrich), α-GST moAb (kindly provided by M. Fanciulli), α-His moAb (Millipore), rabbit α-mouse IgG, HRP-conjugated goat α-mouse and α-rabbit IgG (Cappel). Immuno-reactivity was detected by the ECL-chemi-luminescence reaction (Amersham Corp).

2.7. In vitro translation

GST-HIPK2 was expressed in Reticulocyte Lysate TNT system (Promega) using T7 RNA polymerase in a 2 h-reaction and purified by affinity adsorption to Glutathione-Sepharose (GE Healthcare). TBB and purvalanol A were added prior to the initiation of transcription at the indicated concentrations.

2.8. Microscopy

For immunofluorescence (IF) experiments, cells were seeded onto poly-l-lysine coated coverslips, fixed and permeabilized as empirically predetermined for each Ab [4]. The following Abs were used: rabbit α-calreticulin and rabbit α-LAMP2 (Affinity BioReagents), α-GM130 moAb (BD Bioscences), rabbit α-γ-tubulin Ab (Sigma), α-ubiquitin moAb (Santa Cruz), and α-HIPK2 moAb [17]. Appropriate secondary FITC- or TRITC-conjugated Abs (Jackson Immuno Research Lab) were employed. Fluorescence signals were recorded by Leica DMIRE2 microscope equipped with a Leica DFC 350FX camera and elaborated by Leica FW4000 software (Leica).

3. Results and discussion

Although activation-loop Y residues of HIPKs were found phosphorylated in different conditions [42–44], it is actually unknown whether they are phosphorylated by upstream kinase(s) or by autocatalytic events and whether their phosphorylation is relevant for catalytic activation. To address the possible contribution of the activation-loop Y354 phosphorylation on HIPK2 activity, we first confirmed its phosphorylation status together with the phosphorylation of other HIPK2 sites. Recombinant His- and GST-tagged HIPK2 proteins [e.g., the kinase domain, the wild-type full-length protein or its kinase defective (KD), HIPK2^K221R derivative — Supplementary Fig. S1] were expressed in prokaryotes (Escherichia coli) and eukaryotes (insect Sf21 cells and human H1299 cells), purified, and analyzed by MS and LC–MS/MS. The activation-loop Y354 was phosphorylated in all the recombinant proteins we analyzed with the exception of the KD HIPK2^K221R mutant (Fig. 1A-C and data not shown), suggesting that it might be an autophosphorylation site. In addition, we found 22 phosphorylated S/T sites distributed throughout the different HIPK2 domains, some of which were maintained in the recombinant protein expressed in prokaryotes (Fig. 1D, E and Supplementary Table 1), indicating that HIPK2 autophosphorylates also on S/T sites.

Fig. 1.

HIPK2 phosphorylated sites. A–B) Full-length wild-type GST-HIPK2 produced in E. coli (A) and human H1299 cells (B) were analyzed by MS (left panels) and LC–MS/MS (right panels). Spectra of the peptides containing the activation-loop Y354 are reported. Squares in panels A and B indicate the peak corresponding to the phosphorylated peptide containing Y354. C) KD GST-HIPK2^K221R mutant was produced in human H1299 cells and analyzed by MS as in (B). The dotted square indicates the peak corresponding to the unphosphorylated peptide containing Y354. D) Schematic representation of the distribution of phosphorylated sites on full-length wild-type GST-HIPK2 (eukaryotic GST-HIPK2) and KD GST-HIPK2^K221R purified from human cells (eukaryotic GST-K221R) and on full-length GST-HIPK2 purified from E. coli (prokaryotic GST-HIPK2). E) Schematic representation of HIPK2 structural domains with the 23 phosphoresidues identified by MS in human cells. The sites found phosphorylated also in the recombinant HIPK2 produced in E. coli are indicated with asterisks. N-ter: N-terminal domain; HID: Homeobox Interactive Domain; PEST: PEST sequence; AID: Auto-Inhibitory Domain; YH: Y and H rich domain.

Absence of detection in MS does not necessarily indicate absence of phosphorylation, thus, we verified that HIPK2 autophosphorylates at Y residue by performing WB analysis on the full-length wild-type HIPK2 and its KD mutant, HIPK2^K221R, with two specific anti-phospho-tyrosine (α-pTyr) Abs (Fig. 2A). To confirm that the autophosphorylated Y is the activation-loop Y354, we assessed the ability of an in vitro dephosphorylated HIPK2 to rephosphorylate Y354 in the presence of ATP. Since efficient dephosphorylation of DYRKs can be obtained only with truncated versions of the kinases [45], HIPK2^(158–557) deletion mutant purified from insect cells (Fig. 2B), or from E. coli (Fig. 2C) were employed. α-pTyr WB showed that λPPase treatment efficiently dephosphorylates His-HIPK2^(158–557) at Y residue, whereas λPPase removal and subsequent kinase reaction rescues α-pTyr immunoreactivity (Fig. 2B and C) as well as catalytic activity on an exogenous substrate, i.e., MBP (Supplementary Fig. S2). Next, similarly dephosphorylated/rephosphorylated samples were analyzed by MS. A peak matching a singly phosphorylated ACVSTYLQSR peptide (m/z: 1264.5 vs the expected 1184.5 of the nonphosphorylated peptide) was present in the untreated sample (Fig. 2D), disappeared in the λPPase treated sample (Fig. 2E), and reappeared upon rephosphorylation (Fig. 2F). LC–MS/MS fragmentation demonstrated that Y354 is the phosphorylated residue (Fig. 2G and H), together with S434 (data not shown) while no other phosphorylated Ys were detected in any of these conditions. Taken together, these data show that HIPK2-Y354 phosphorylation is an autocatalytic event.

Fig. 2.

HIPK2-Y354 phosphorylation is an autocatalytic event. A) The indicated proteins expressed and GST-purified from human H1299 cells were analyzed by WB with α-pTyr PY99 and 4G10 Abs. Normalization was performed by WB with α-HIPK2 Ab. B) His-tagged HIPK2^(158–557) produced and purified from insect cells were untreated (first lane), treated with ATP (second lane), treated with phosphatase (third lane), or treated with phosphatase, washed to remove phosphatase, and treated with ATP (fourth lane). WB were performed with PY99 α-pTyr Ab. Normalization was performed with α-His Ab. C) His-tagged HIPK2^(158–557) samples derived from E. coli were treated and analyzed as in B. D–H) His-tagged HIPK2^(158–557) produced and purified from insect cells was treated as in (B) and analyzed by MS (D, E, and F) and LC–MS/MS (G and H). Spectra of the peptides containing the activation-loop Y354 are reported. Squares in D and F indicate the peak corresponding to the phosphorylated peptide containing Y354. Dotted square in E indicates the absence of the peak corresponding to the phosphorylated peptide containing Y354.

In the DYRK family, phosphorylation of the activation-loop Y takes place in cis on the nascent, ribosome-bound protein. In addition, it has been proposed that after release from the ribosome, the mature kinase loses the Y autophosphorylation activity and becomes a S/T kinase toward exogenous substrates [46]. Thus, we assessed whether HIPK2-Y354 behaves as a DYRK activation-loop Y and its phosphorylation is required for HIPK2 catalytic activity on itself and its substrates. To this aim, we first evaluated HIPK2 sensitivity to purvalanol A and TBB, two small-molecules that differentially inhibit DYRKs' activity. In particular, purvalanol A, but not TBB, inhibits the activation-loop Y autophosphorylation in rabbit reticulocyte lysates, while both molecules inhibit trans-phosphorylation on exogenous substrates [46]. α-pTyr WB of in vitro translated HIPK2 showed that Y autophosphorylation is inhibited by purvalanol A but not TBB (Fig. 3A). In contrast, both small molecules inhibited HIPK2 trans-phosphorylation on MBP, as assessed by in vitro kinase assays (Fig. 3B). These results show that the catalytic structure of HIPK2 able to autophosphorylate on Y residue is different from the catalytic structure that phosphorylate exogenous substrates on S/T residues and, in analogy with DYRKs, support the hypothesis that activation-loop Y354 autophosphorylation might take place on nascent HIPK2 protein.

Fig. 3.

Effects of purvalanol A and TBB on HIPK2 kinase activity. A) Full-length GST-HIPK2 was in vitro translated with rabbit reticulocyte lysate in the presence of purvalanol A (upper panel) or TBB (lower panel) at the indicated concentrations. In vitro translated proteins were purified by GST-pulldown, resolved by SDS-PAGE, and immunoblotted with PY99 α-pTyr and α-HIPK2 Abs. B) In vitro kinase assays of full-length GST-HIPK2 purified from human H1299 cells on MBP substrate in the presence of increasing amounts of TBB and purvalanol A. MBP phosphorylation was measured evaluating [³²P]-incorporation upon SDS-PAGE and autoradiography. Data are shown as mean with standard deviation (SD) of three independent experiments.

To further evaluate whether Y354 phosphorylation is required for the S/T kinase activity of HIPK2, we compared the full-length wild-type HIPK2 and its KD HIPK2^K221R mutant with the HIPK2^Y354F derivative in which the activation-loop Y354 is non-phosphorylatable. In vitro kinase assays showed that substitution of the activation-loop Y with F almost abolished the ability of HIPK2 to phosphorylate the non-specific substrate MBP (Fig. 4A) and strongly inhibited phosphorylation of two specific substrates, p53 and p63 (Fig. 4B and C). These data support the idea that phosphorylation of the activation-loop Y354 is responsible for the catalytic activation of HIPK2. However, when we assessed the three GST-purified proteins from human cells for their autophosphorylation, HIPK2^Y354F exhibited an intermediate activity between HIPK2 and the KD HIPK2^K221R mutant (Fig. 4D), indicating that inhibition of Y354 phosphorylation does not abolish per se the catalytic activity of the kinase. Of relevance, comparable results were obtained by assessing the kinase activity of full-length wild-type HIPK2 translated in vitro (IVT-HIPK2) in the presence of purvalanol A to inhibit Y354 phosphorylation on the nascent protein, as confirmed by α-pTyr WB (Fig. 4E and F). Surprisingly, when the same proteins were analyzed by anti-pTyr WB, an increased immunoreactivity was detected in the GST-purified HIPK2^Y354F compared with wild-type HIPK2 and its KD mutant (Fig. 5A) and in the wild-type IVT-HIPK2 translated in the presence of purvalanol A, washed and incubated in the presence of ATP (Fig. 5B). Together with the data reported above, these results suggest that the residual autophosphorylation is due, at least in part, to Y-kinase activity. Indeed, a direct analysis of the phosphorylation pattern of HIPK2^Y354F by MS showed the appearance of a novel phosphorylated Y residue at position 1186 (Fig. 5C and Supplementary Fig. S3A). Since we never detected Y1186 phosphorylation in wild-type HIPK2, even upon incubation with ATP to maximize autophosphorylation (data not shown), these results suggest that the activation-loop Y354 autophosphorylation prevents aberrant Y phosphorylation activity.

Fig. 4.

Impairment of HIPK2-Y345 phosphorylation affects HIPK2 kinase ability. A–C) MBP, p53, and p63 phosphorylation by the indicated full-length HIPK2 proteins, GST-purified from human H1299 cells were assessed upon in vitro kinase assays in the presence of [γ-³²P]-ATP. D) Autophosphorylation activity of the same recombinant proteins employed above. (A–D) Indicative autoradiographies with the relative Coomassie blue stainings are reported while the upper graphics represent the mean with SD of kinase activities of at least three independent experiments. E) Full-length wild-type HIPK2 was in vitro translated (IVT) with rabbit reticulocyte lysate in the presence or absence of 150 μM of purvalanol A. Samples translated in the presence of purvalanol A were washed to remove the drug and employed for in vitro kinase assays with [γ-³²P]-ATP in the presence or absence of MBP. Autoradiographies and densitometric analysis and Coomassie staining are shown (upper panels). Full-length wild-type HIPK2 was in vitro translated with rabbit reticulocyte lysate in the presence or absence of 150 μM of purvalanol A, as in E. Samples were washed to remove the drug and subjected to a kinase assay in the presence of ATP. The level of pY of each sample was analyzed by WB with PY99 α-pTyr Ab. Normalization was performed by WB with α-HIPK2 Ab (lower panels). F) Graphic representation of the densitometric values of the autophosphorylation and MBP phosphorylation shown in E.

Fig. 5.

Impairment of HIPK2-Y345 phosphorylation results in a strong and out-of-target Y-kinase activity in cis. A) WB with PY99 α-pTyr Ab of the indicated proteins expressed and GST-purified from human H1299 cells and treated with ATP. Normalization was performed with α-HIPK2 Ab. B) Full-length wild-type HIPK2 was in vitro translated with rabbit reticulocyte lysate. Samples untreated (first lane), treated with 150 μM of purvalanol (second lane), or treated with 150 μM of purvalanol, washed to remove the drug, and treated with ATP (third lane) were analyzed by WB with PY99 α-pTyr Ab. Normalization was performed by WB with α-HIPK2 Ab (lower panels). C) LC–MS/MS profiles of the extract ions relative to the phosphorylated peptide with sequence TGYPLSPAKVNQYPYI acquired for GST-HIPK2^Y354F (upper panel) and GST-HIPK2 (lower panel) expressed and purified from H1299 cells and subjected to an in vitro kinase assay. A peak at retention time 28.2 min was detected only in the GST-HIPK2^Y354F sample, while a peak at retention time 30.9 min, containing a Ser-phosphorylated residue was present in both recombinant proteins. D) WB with PY99 α-pTyr Ab was performed on the indicated full-length or Δ1050-1189 GST-purified proteins from human cells, kept untreated or subjected to in vitro kinase assays in the presence of cold ATP. Normalization was performed with α-GST Ab. E) WB with PY99 α-pTyr Ab of the indicated proteins GST-purified from E. coli. Normalization was performed with α-GST Ab. F) The indicated recombinant proteins were mixed with HIPK2^(784–1189), expressed and purified from E. coli, and subjected to in vitro kinase assays in the presence of [γ-³²P]-ATP or cold ATP. Autoradiography of the substrate is shown in the upper panel. The histogram reports the quantification of ³²P-incorporation into the substrate. Substrate phosphorylation at Y residues was evaluated by WB with PY99 α-pTyr Ab. Substrate normalizations were obtained by Coomassie blue staining and by probing with α-GST Ab, respectively.

To further characterize this behavior, we assessed the Y kinase activity of HIPK2^Y354F upon deletion of its C-terminal region (aa1050–1189) that contains the newly phosphorylated Y1186 residue. Similar deletion mutants of wild-type HIPK2 and its KD mutant were analyzed as control, and showed no significant difference before and after incubation in the presence of ATP, as assessed by α-pTyr WB (Fig. 5D, lower panel) and MS (data not shown). α-pTyr WB showed no reduction in the ability of HIPK2^{Y354F(Δ1050–1189)} to autophosphorylate Y residues (Fig. 5D, upper panel), indicating the maintenance of a strong and out-of-target Y-kinase activity. Indeed, comparable results were obtained by employing the sole HIPK2 kinase domain, i.e., GST-HIPK2^(158–557), in the wild-type form or with the K221R or the Y354F mutations purified from bacteria, analyzed by α-pTyr WB in the absence of incubation with ATP (Supplementary Fig. S1 and Fig. 5E), showing that different Y residues can be the target of autophosphorylation in the absence of activation-loop Y354 phosphorylation.

Finally, we asked whether the persistent, out-of-target Y autophosphorylation of HIPK2^Y354F takes place in cis or in trans. To this aim, the HIPK2 C-terminal region containing the Y1186 phosphorylation site, HIPK2^(784–1189) (Supplementary Fig. S1) purified from E. coli, was used as substrate for HIPK2, HIPK2^Y354F, and HIPK2^K221R in in vitro kinase assays in the presence of [γ-³²P]-ATP or cold ATP. In the first case, autoradiography showed that HIPK2 and HIPK2^Y354F are able to trans-phosphorylate HIPK2^(784–1189), though HIPK2^Y354F was less efficient than wild-type HIPK2 (Fig. 5F, graphic and upper panels). In the second case, α-pTyr WB showed no immunoreaction with any of the samples (Fig. 5F, lower panels), indicating that Y1186 autophosphorylation by HIPK2^Y354F occurs in cis, while the trans-phosphorylation detected by autoradiography applies to S/T residues. Similar results were obtained by employing the HIPK2^{K221R(158–557)} as substrate (Supplementary Fig. S3B and S3C).

Altogether, these results indicate that, similarly to DYRKs, phosphorylation of the activation-loop Y354 of HIPK2 is an autocatalytic event and regulates the S/T kinase activity. However, at variance with DYRKs, in which inhibition of the activation-loop Y autophosphorylation results in a complete loss of Y-kinase activity [45,46], impairment of HIPK2-Y354 phosphorylation results in the acquisition of a strong out-of-target Y-kinase activity.

To evaluate the functional implications of impairing HIPK2-Y354 autophosphorylation, we transfected human U2OS cells with EGFP-tagged wild-type HIPK2 (EGFP-HIPK2) or its Y354F derivative (EGFP-HIPK2^Y354F). As expected [47], EGFP-HIPK2 showed a prevalent nuclear, dotted distribution and, to a lower extent, a nuclear/cytoplasmic, dotted localization (Fig. 6A and B). Surprisingly, EGFP-HIPK2^Y354F showed an opposite distribution with a prevalent nuclear/cytoplasmic or solely cytoplasmic localization (Fig. 6A and B), the latter one being represented by dot-like structures and larger aggregations (Fig. 6A and C). These different subcellular distributions were confirmed by WB analyses performed on nuclear and cytoplasmic fractions (Fig. 6D). In addition, comparable results were obtained i) employing an EGFP-HIPK2^Y345A derivative, in which the non-phosphorylatable F residue was substituted with an A, to exclude amino acid-based artifacts (Fig. 6B and C); ii) using a lower amount of plasmid DNA (Supplementary Fig. S3A-D); iii) transfecting cells from a different line (i.e., human RKO cells) (Supplementary Fig. S3C and D); and iv) inhibiting Y autophosphorylation of wild-type HIPK2 with purvalanol A. In the last experiment, we assessed the subcellular distribution of EGFP-HIPK2 expressed in U2OS cells (Fig. 6E and F) or of the endogenous HIPK2 in human fibroblasts (Fig. 6G and H), in the presence or absence of purvalanol A or TBB. Treatment with purvalanol A, but not TBB, induced a strong cytoplasmic delocalization of wild-type HIPK2 (Fig. 6E–H, and data not shown) but not of the transcription factor NF-YA employed as control (Supplementary Fig. S3E). Furthermore, we observed that the cytoplasmic delocalization is characteristic of HIPK2, since the homologous DYRK1A^Y321F derivative maintains a nuclear localization (Supplementary Fig. S3F). Of note, despite the absence of Y354 phosphorylation described above, the KD HIPK2^K221R mutant retains a nuclear distribution (Fig. 6I and L) [47]. This behavior suggests that the cytoplasmic relocalization observed upon inhibition of the sole Y354 phosphorylation might be due to the out-of-target Y phosphorylation, rather than to a generalized inhibition of HIPK2 kinase activity. Indeed, this hypothesis was strongly supported by the observation that a double HIPK2^K221R/Y354F mutant exhibits a nuclear localization comparable with the single HIPK2^K221R mutant (Fig. 6I and L). This result might explain why DYRK1A^Y321F derivative, that at variance from HIPK2 loses its Y-kinase activity, maintains a nuclear localization.

Fig. 6.

Inhibition of activation loop Y354 autophosphorylation induces HIPK2 cytoplasm delocalization. A–C) U2OS cells were transduced with the indicated proteins. After 24 h from transfection cells were fixed, stained with Hoechst to visualize DNA, and analyzed for EGFP autofluorescence. Representative images are shown in (A). Percentages of cells with nuclear, nuclear/cytoplasmic, and cytoplasmic localizations (B) or with cytoplasmic dotted or aggregated distribution (C) are reported for each indicated protein. At least 300 cells per samples were analyzed here and in the following experiments and data are shown as mean ± SD of three independent experiments. D) U2OS cells were transfected as above and subjected to subcellular fractionation. Fractions were resolved by SDS-PAGE and immunoblotted with α-HIPK2 Ab. α-tubulin and α-HDAC1 Abs were used as markers of cytoplasmic and nuclear extracts. E, F) U2OS cells were transfected with an EGFP-HIPK2-carrying vector in the presence or absence of 100 μm purvalanol A. After 16 h from transfection, subcellular localization of the tagged protein was evaluated by EGFP autofluorescence. Representative images are shown in (E) and percentages of cells with different subcellular localization are reported in (F). G, H) Human fibroblasts were cultured in the presence or absence of 100 μm purvalanol A for 72 h, fixed and immunostained for the endogenous HIPK2 with α-HIPK2 Ab as described [4]. Representative images are shown (G) and percentages of cells with different subcellular localization are reported in (H). I, L) U2OS cells were transduced with the indicated proteins and analyzed as in (A). Representative images are shown in (I) and percentages of cells with different subcellular localization are reported in (L).

To further characterize the HIPK2 subcellular distribution observed upon impairment of Y354 phosphorylation, colocalization analyses were performed with organelle-specific Abs. We found that EGFP-HIPK2^Y354F does not colocalize with calreticulin, GM130, and Mitotracker, respectively employed as markers of the endoplasmic reticulum, Golgi apparatus, and mitochondria (Supplementary Fig. S4). In contrast, colocalization was observed with the lysosomal marker LAMP2 and with ubiquitin, while γ-tubulin immunostaining showed proximity of the EGFP-HIPK2^Y354F aggregations with microtubule organizing centers (Fig. 7A). This distribution is compatible with that of the intracellular “storage bins” known as aggresomes [48], usually involved in the clearance of misfolded or aggregated proteins [49]. Since aggresome-stored proteins are partially insoluble [41], we compared the solubility of HIPK2 and its HIPK2^Y354F derivative. To this aim, TCEs were obtained from U2OS cells transfected to express comparable amounts of EGFP-HIPK2 and EGFP-HIPK2^Y354F mRNAs (Fig. 7B), and both soluble and insoluble fractions were analyzed by WB for the presence of the kinases. As shown in Fig. 7C, whereas wild-type HIPK2 was almost completely present in the soluble fraction, about 50% of HIPK2^Y354F was found in the insoluble one. This behavior is not a mutation-linked artifact since we could induce cytoplasmic delocalization and aggresome storage of wild-type EGFP-HIPK2 by cell treatment with MG132 to block the proteasome. As shown by WB for soluble and insoluble fractions (Fig. 7D) and by immunofluorescence with an anti-ubiquitin Ab (Fig. 7E and F), a strong increase in the presence of HIPK2 in the insoluble fraction and a shift from nuclear, ubiquitin-negative dots to cytoplasmic, ubiquitin-positive aggregates were observed in the MG132-treated cells.

Fig. 7.

Cytoplasmic HIPK2 accumulates into aggresomes. A) U2OS cells were transiently transfected with EGFP-Y354F-expressing vector. After 24 h from transfection, cells were immunostained with the indicated Abs. Hoechst was used to visualize DNA. Colocalization with EGFP-Y354F was evaluated through EGFP autofluorescence. Representative images of at least 200 cells scored are shown. B, C) U2OS cells were transfected with EGFP-HIPK2, EGFP-Y354F or EGFP-empty vector (e.v.). After 24 h from transfection, cells were harvested and divided into aliquots to obtain total mRNA, TCEs, soluble and insoluble fractions. HIPK2 and Y354F mRNA levels were analyzed by real-time RT-PCR (B). TCEs, soluble, and insoluble fractions were resolved by SDS-PAGE and immunoblotted with α-HIPK2 Ab. Normalization was obtained with α-tubulin Ab (C). D–F) U2OS cells were transiently transfected with EGFP-HIPK2-expressing vector, treated for 16 h with 5 μM MG132 and analyzed by WB and IF as follows: harvested to obtain soluble and insoluble fractions that were resolved by SDS-PAGE and analyzed by WB with the indicated Abs (D); and fixed and immunostained with α-ubiquitin Ab and Hoechst to visualize DNA. Colocalization with EGFP-HIPK2 was evaluated by EGFP autofluorescence (E) and percentages of cells with different subcellular localization are reported in (F). At least 300 cells per sample were analyzed and percentages of cytoplasmic dots and aggresomes are reported. G) Human HEL leukemia cells were immunostained with α-HIPK2 (red) and α-ubiquitin (green) Abs. Representative images of at least 200 cells scored are shown.

The HIPK2 aggresome accumulation induced by inhibition of Y354 phosphorylation is reminiscent of HIPK2 delocalization observed in human breast carcinomas and leukemias [15,24]. Interestingly, IF analyses of endogenous HIPK2 in tumor cells with high cytoplasmic HIPK2 expression, such as HEL leukemia cells, showed a significant colocalization of HIPK2 and ubiquitin in aggresome-like structures (Fig. 7G), suggesting that aggresome formation, such as that induced by inhibition of HIPK2-Y354 phosphorylation, might contribute to inactivate the tumor-suppressive function(s) of HIPK2 in human cancers.

4. Conclusion

The phosphorylation of one or two amino acids in the kinase activation-loop, a sequence of 12-37 residues conserved among members of the same family, is one of the mechanisms through which protein kinases regulate their own activity [32]. Activation-loop phosphorylation might be catalyzed by upstream kinases or occurs by autophosphorylation in cis or in trans. In this study, we identify the activation-loop Y354 of HIPK2 as a critical autophosphorylation site to acquire kinase activation, substrate specificity, and subcellular localization. Indeed, differently from DYRKs, impairment of HIPK2-Y354 autophosphorylation results in a strong out-of-target Y-kinase activity in cis that cause a significant cytoplasmic relocalization with aggresome formation. Further studies will be needed to determine whether this activity pertains to a dysfunctional state of HIPK2, leading to inactivation of its tumor-suppressing functions, for example through mutations in the activation-loop or overexpression/iperactivation of tyrosine phosphatases and pseudo-phosphatases. Alternatively, it might belong to a physiological state of HIPK2, involved in regulating its stability and/or activity. The dephosphorylation/rephosphorylation data obtained with the isolated kinase domain suggest that HIPK2 can fold properly and its dephosphorylation does not impair the Y354 kinase activity. In contrast, impairment of Y354 autophosphorylation during translation, as mimicked by the Y354F mutant or purvalanol A treatment, might inhibit a proper folding of HIPK2, or induce a different folding, resulting in out-of target Y auto-phosphorylation. The resemblance of HIPK2 accumulation upon inhibition of Y354 phosphorylation with the cytoplasmic delocalization of HIPK2 in human cancers suggests that neoplasias might be used as experimental systems to discriminate between these possibilities.

Appendix A. Supplementary data

Supplementary Table

Supplementary Figures