Integration of stress signals by homeodomain interacting protein kinases

Evolutionary conservation of the HIPK family

The members of the evolutionary conserved family of HIPKs have a similar basic architecture, with an N-terminal kinase domain and multiple functional domains in the C-terminus. In contrast to these canonical HIPKs (HIPK1-3), the family member HIPK4 only shares the kinase domain with the other family members and thus is very likely to display distinct biological roles. The HIPK family of protein kinases belongs to the CMGC (containing CDK, MAPK, GSK, CLK families) group of protein kinases (Manning et al., 2002). Within this group, HIPKs are most closely related to the family of dual-specifity tyrosine phosphorylation regulated kinases (DYRK), as schematically shown in Figure 1. Most of the information discussed here has been gathered for HIPK2, which is the most-studied member of this family of protein kinases. HIPK2 is detected much more frequently than its sister kinases when unbiased screens such as functional experiments or yeast two-hybrid screens are employed. The nature of this bias is not known and the preferred detection of HIPK2 may suggest its special biological relevance. The phylogenetic ancestor of HIPKs and DYRKs is the yeast Yak1 protein, which already shows most of the principle functional features displayed by the canonical HIPK family members.

Figure 1

The CMG group of kinases in the phylogenetic tree of human kinases (Manning et al., 2002).

Yak1, an ancestral kinase linking cell stress to transcriptional regulation

The Yak1 protein negatively regulates cell proliferation and its kinase activity is upregulated during S phase arrest of yeast cells (Garrett et al., 1991). Also the late mitotic regulatory network controlling cyclin destruction is controlled by Yak1, as revealed by genetic evidence. The growth defect of several late mitotic mutants leading to cell cycle arrest in anaphase could be rescued by overexpression of Yak1 (Jaspersen et al., 1998), but the molecular mechanisms mediating the rescue remain to be established.

Besides its involvement in cell cycle regulation, many studies have identified Yak1 as a stress sensor that mediates gene regulation in response to heat shock, acid stress and metabolic stress. Glucose availability is reflected at the levels of the two kinases PKA (protein kinase A) and TOR (target of rapamycin), which both regulate the localization and activity of Yak1, as schematically shown in Figure 2. Initial evidence for the regulation of Yak1 by glucose came from a study describing that deletion of the YAK1 gene allows growth of yeast cells lacking the catalytic subunit of the cAMP-dependent protein kinase PKA (Garrett and Broach, 1989). Yak1 is not only negatively regulated by PKA, but also by TOR, the second relevant nutrient-sensing kinase pathway (Martin et al., 2004; Schmelzle et al., 2004). In unstressed cells, the Yak1 protein localizes throughout the entire cell. Induction of stress, by removal of glucose or rapamycin-induced inhibition of TOR, results in nuclear accumulation of Yak1 (Moriya et al., 2001). Cytosolic localization of Yak1 requires PKA-dependent phosphorylation of Yak1 on Ser295 (Lee et al., 2011). Autophosphorylation at Thr335 enables binding of a yeast 14-3-3 protein, which decreases the catalytic activity of Yak1 and also contributes to the cytoplasmic retention of inactive Yak1 under high glucose conditions (Lee et al., 2011). The group of Yak1-regulated transcription factors includes the stress responsive transcription factors heat shock factor protein 1 (Hsf1) and Msn2/Msn4. They are direct phosphorylation targets of Yak1 under conditions where PKA activity is lowered by glucose depletion (Lee et al., 2008). While it is known that Yak1-mediated phosphorylation of Hsf1 increases the DNA-binding activity of the transcription factor, the functional relevance of Msn2 phosphorylation remains to be established (Lee et al., 2008). Genetic evidence shows that Msn2/Msn4 trigger the expression of enzymes mediating synthesis of the storage carbohydrate glycogen. In addition, these transcription factors augment Yak1 expression and thus establish an autoregulatory loop (Smith et al., 1998). Limitation of glucose also leads to Yak1-dependent phosphorylation of Pop2/Caf1, a homolog of the human cNOT7 protein which is contained in the Ccr4-Not (carbon catabolite repressor 4 negative on TATA-less) complex (Moriya et al., 2001). This multi-subunit complex has been implicated in many different aspects of the mRNA life cycle (Miller and Reese, 2012) and interestingly Yak1 seems also to associate with other components of this complex, such as cell division cycle (Cdc)39/Not1 (http://www.yeastgenome.org/). These findings raise the possibility that Yak1 also contributes to post-transcriptional gene regulation. A further study proposed the transcription factor Haa1 as a Yak1 target under conditions of acidic stress (Malcher et al., 2011). The same study also implied the transcription factors Sok2 and Phd1 as downstream targets of Yak1, showing that this kinase can regulate a multitude of transcription factors (Malcher et al., 2011).

Figure 2

Glucose responsiveness of the yeast Yak1 protein. Under conditions of high glucose, Yak1 is retained in the cytosol by TOR and PKA-dependent signals. Low glucose levels allow phosphorylation and activation of Hsf1 and Msn2/4, which in turn lead to induced synthesis of Yak1 and of enzymes catalyzing glycogen synthesis.

Yak1 also acts on transcriptional coregulators such as the corepressor Crf1 which participates in the negative regulation of ribosomal protein genes and thus of ribosome biogenesis. TOR/PKA-mediated signals negatively regulate Yak1 and therefore maintain the inhibitory protein Crf1 in the cytoplasm. TOR inactivation leads to Yak1 induction, which in turn phosphorylates and activates Crf1 (Martin et al., 2004). The phosphorylated corepressor accumulates in the nucleus and inhibits transcription of genes encoding ribosomal proteins, thus providing a signaling mechanism connecting environmental stress signaling to ribosome biogenesis. Another Yak1 substrate protein involved in the control of ribosomal protein synthesis is Ifh1. While the direct phosphorylation can be recapitulated using in vitro kinase assays (Kim et al., 2011), the functional consequences of this phosphorylation remain to be studied. A genetic interaction screen revealed the necessity of Yak1 for the mutual binding of the chromatin assembly factors Msi1 and Cac1 (Pratt et al., 2007). Further transcriptional regulators found in a proteomic screen include the dual function helix-loop-helix protein Cbf1, subunits of the RNA polymerase II mediator complex and a subunit of the chromatin-modifying COMPASS (Complex Proteins Associated with Set1) complex (http://www.yeastgenome.org/). These data show that the phosphorylation targets of Yak1 fall roughly into the same categories as the phosphorylation targets of HIPKs, which act mainly on transcription factors and transcriptional coregulators.

How can the Yak1 kinase activity be activated? Biochemical experiments show that Yak1 can autophosphorylate at Tyr530 in the activation loop. Removal of this phosphorylation by treatment with protein tyrosine phosphatase 1B drastically reduces Yak1 kinase activity (Kassis et al., 2000). This regulatory principle has been maintained throughout evolution and accordingly also tyrosine autophosphorylation of DYRK2 and HIPK2 is important for its kinase activity (Lochhead et al., 2005; Saul et al., 2013).

Architecture of HIPKs

The various domains found in HIPKs are schematically displayed in Figure 3, which also visualizes the close relationship between HIPK1 and HIPK2. HIPK1-3 contain an N-terminal domain of unknown function. The actual N-terminal end of HIPK2 has not been experimentally determined, as the putative translational start site is preceded by seven amino acids encoded in a separate exon. This also explains the coexistence of two amino acid numbering systems in HIPK2. The N-terminal part of all HIPK family members also contains the kinase domain, which is followed by a protein/protein interaction domain that mediates binding to homeodomain transcription factors and further regulatory proteins. Interestingly, the protein/protein interaction domain also occurs in a truncated form where splicing creates a variant that is 27 amino acids shorter (Kurokawa et al., 2013). This has important biological consequences, as the shorter version fails to bind the ubiquitin E3 ligase seven in absentia (SIAH) 1 and thus escapes from the constant ongoing ubiquitination and proteasomal degradation occurring for the full-length kinase (Kurokawa et al., 2013). It will be interesting to see whether the splicing event leading to the generation of a SIAH1-resistant HIPK2 form can be regulated under (patho)physiological conditions. The protein/protein interaction domain is followed by a PEST domain and a C-terminus rich in short repeats of serines, glutamines and alanines. The function of these C-terminally iterated amino acids is not known, but the repeated occurrence of amino acids may be indicative for an unstructured region. Disordered regions allow conformational flexibility and increase intramolecular flexibility to expand the repertoire of possible protein/protein interactions (Dunker and Gough, 2011). A regulated conformation of the HIPK proteins would be compatible with a recent finding that describes the autophosphorylation-dependent interaction of HIPK2 with the prolyl isomerase Pin1 (Bitomsky et al., 2013; Saul et al., 2013). These enzymes catalyze the cis/trans conversion between phosphorylated Ser/Thr-Pro motifs in proteins and thus regulate the protein shape (Liou et al., 2011). Autophosphorylation of HIPK2 at Thr880/Ser882 creates a binding signal for Pin1, which in turn stabilizes the kinase by inhibiting HIPK2 polyubiquitination and thus triggering cell death (Bitomsky et al., 2013). While HIPK1-3 is found in all vertebrates, HIPK4 only occurs in mammals. The N-terminal lysine 25 contained in HIPK2 can be attached to SUMO1 (small ubiquitin-like modifier), which decreases the protein stability of the modified kinase (Gresko et al., 2005) and impairs its ability to activate c-Jun N-terminal kinase (JNK) (Hofmann et al., 2005). Experimental evidence shows SUMOylation of HIPK3 (Gresko et al., 2005) and HIPK1 (Li et al., 2005). The SUMOylated lysine 25 in HIPK2 also occurs in HIPK1 and HIPK3, but it remains to be experimentally tested whether SUMOylation really takes place at these homologous positions. The C-terminus of HIPK2 contains a region that is important for localization of the kinase to nuclear speckles (Kim et al., 1999). The ability to reside in subnuclear speckles critically relies on the presence of a functional SUMO-interaction motif (SIM), which allows non-covalent binding of the SUMO isoforms SUMO1-3. This domain allows the formation of intra- and intermolecular protein/protein interactions and thus functions as a glue that enhances specific interactions. The SIM motif is perfectly conserved in HIPK1 and HIPK3 and our data show that also the HIPK1 SIM is functional in SUMO binding (M.L.S., unpublished observation). The nuclear localization of HIPK2 can be also controlled by two nuclear localization sequences (NLS1 and NLS2). Two lysines contained in NLS1 (Lys796 and Lys798) can be modified by acetylation in cells experiencing high levels of reactive oxygen species (ROS) and prohibits localization of HIPK2 to nuclear speckles (de la Vega et al., 2012). Only NLS1 is maintained in HIPK1, while none of the two NLS sequences are preserved in HIPK3 or HIPK4. It will thus be interesting to identify the sequences mediating nuclear localization of HIPK3. Functional experiments showed the occurrence of an autoinhibitory domain between 935 and 1050 in HIPK2 (Rui et al., 2004). The sequence encompassing the autoinhibitory region is well conserved in HIPK1 and thus it will be relevant to study whether this sequence also functions to restrict the catalytic activity of HIPK1.

Figure 3

Domain structure of HIPK family members. The various domains including the inhibitory (inhibit.) domain are indicated. The positions of the activation loop (AL) and SUMO (S) attachment sites are displayed. For further details, see text.

Control of HIPK activity and expression levels

Similar to the related Yak1 and DYRK kinases, HIPK2 also has the ability to autophosphorylate critical residues in the activation loop. Biochemical experiments revealed the relevance of HIPK2 cis-autophosphorylation at Tyr354 and Ser357 in the activation loop for its kinase function (Saul et al., 2013; Siepi et al., 2013). In marked contrast to some DYRK kinases where tyrosine phosphorylation is absolutely essential, Tyr354 phosphorylation of HIPK2 alone is not sufficient for full HIPK2 activity. This raises the possibility that the activation loop of HIPK2 is regulated by additional kinases or phosphatases, but these putative upstream regulators have not yet been identified. A candidate kinase is TGFβ-activated kinase 1 (TAK1), which may be responsible for the increased Tyr354 phosphorylation observed after transforming growth factor (TGF) β stimulation of cells (Shang et al., 2013). Although activation loop phosphorylation has thus far only been studied for HIPK2, the high degree of sequence conservation in this region suggests that also the other HIPK family members are regulated in a similar way.

The fact that HIPKs are generated in a fully (or at least partially) activated status raises the need to control their activity by alternative mechanisms. The intracellular localization of HIPK2 can be regulated by acetylation of lysines contained in NLS1 (de la Vega et al., 2012) or by cytoplasmic relocalization of HIPK2 by increased expression of high-mobility group A1 protein (Pierantoni et al., 2007). The intracellular localization of HIPK1 is regulated by its SUMOylation status. In response to stimulation of cells with tumor necrosis factor, HIPK1 is transported to the cytosol where it participates in the activation of JNK kinases. The cytosolic translocation depends on the SUMO-specific protease SENP1 (sentrin-specific protease), which deconjugates SUMO from HIPK1 and thus allows its relocalization to the cytosol (Li et al., 2008). In addition, HIPK2 expression levels are tightly regulated by several mechanisms, thus ensuring that the protein amounts of this important regulator of cell proliferation and survival are kept in check. HIPK2 mRNA expression levels are controlled by at least two different microRNAs (miR-27a and miR-181a) (Li et al., 2010; Huang et al., 2011) and similarly microRNA-mediated regulation of transcript abundance was also reported for HIPK1 (Wu et al., 2013) and HIPK3 (Mosakhani et al., 2013). At the level of protein stability, the amount of HIPK2 protein is tightly controlled by degradative ubiquitination which is exerted by at least four different ubiquitin E3 ligases (MDM2, SIAH, WSB-1 and Fbx3-formed SCF) (Rinaldo et al., 2007; Choi et al., 2008; Shima et al., 2008; Calzado et al., 2009a). Dependent on the physiological situation, these different E3 ligases can be activated to degrade HIPK2. Hypoxia leads to the induced expression of the ubiquitin E3 ligase SIAH2, which associates with HIPK2 and causes its efficient polyubiquitination and proteasomal degradation. As HIPK2 contributes to gene repression of multiple genes upon association with histone deacetylases (HDACs) (Choi et al., 1999; de la Vega et al., 2013) and further repressive proteins, the hypoxia-induced proteasomal elimination of the kinase allows full induction of hypoxia-triggered genes (Calzado et al., 2009b). Continuous HIPK2 ubiquitination in unstressed cells is also mediated by SIAH1 and WSB-1, thus limiting HIPK2 steady state expression levels.

Integration of stress signals by homeodomain interacting protein kinases

Members of the canonical HIPK family have been implicated in a number of conditions representing precarious situations for the cell. HIPKs reportedly show differences in function, posttranslational modification or intracellular localization in response to many distinct adverse stimuli. For example, traumatic brain injury enhances the interaction between HIPK2 and CtBP2 (C-terminal binding protein 2) (Zou et al., 2013). High levels of ROS, as they occur in many pathological situations ranging from inflammation to viral infection and necrosis (Wiseman and Halliwell, 1996), allow the inducible acetylation of HIPK2, which ensures cell survival even under conditions of high oxidative stress (de la Vega et al., 2012). Also the limited supply of oxygen (hypoxia) is a stressful condition regulating HIPK2-dependent signaling. Hypoxia results in a markedly increased binding between HIPK2 and SIAH2, which leads to almost complete degradation of HIPK2 (Calzado et al., 2009a). HIPK2 also interacts with a number of viral proteins. The nucleocapsid proteins from Seoul virus and Hantaan virus bind to HIPK2, as revealed by a yeast two-hybrid screening (Lee et al., 2003). Also the US11 protein of herpes simplex virus binds to the PEST domain of HIPK2. The US11 protein is a sensor of environmental stress and regulates cell survival. Binding of this viral protein to HIPK2 drastically changes the subcellular distribution of the kinase by unknown mechanisms (Giraud et al., 2004). Also the E6 protein encoded by human papillomaviruses directly interacts with HIPK2 both in vitro and in vivo. UV radiation results in colocalization between both proteins in nuclear bodies. The association between E6 and HIPK2 enforces dissociation of the HIPK2/p53 complex and consequently impairs HIPK2-mediated p53 Ser46 phosphorylation (Muschik et al., 2011). It will be interesting to see whether this serves as a viral strategy to favor survival of virally infected keratinocytes experiencing DNA damage. In addition, HIPKs are also regulated by DNA damage and metabolic stress, as discussed in more detail below.

The multitude of different stress signals impinging on HIPKs is reflected by the large variety of signaling pathways employing members of the canonical HIPK family. These pathways include the Salvador-Warts-Hippo pathway, bone morphogenetic protein (BMP), Wnt/Wingless, Notch, TGFb, p53 and mitogen-activated protein kinase (MAPK) signaling networks, as described in several recent reviews (Puca et al., 2010a; D’Orazi et al., 2012). It is important to note that HIPKs were never identified as essential and canonical core components in any of those pathways. In all cases HIPKs were found to modulate signal output and to connect different stress signaling pathways. The involvement of HIPKs in a bewildering large number of signaling pathways rather suggests that they function as general signaling hubs that receive input signals from various stress pathways. HIPKs integrate these signals and feed them into various downstream effector pathways that serve to cope with these precarious situations. The function of HIPKs as auxiliary factors for the integration of stress signals derived from various sources is schematically depicted in Figure 4. The involvement of HIPKs in the shaping of so many downstream effector pathways also frequently implies them in proliferative diseases such as cancer or fibrosis (Saul and Schmitz, 2013).

Figure 4

The function of HIPKs as integrators of stress signals. The stress signals and effector pathways are shown.

The numerous HIPK substrate proteins can be roughly divided into two groups. Similar to the Yak1 substrates, the first group comprises transcription factors and accessory proteins of the transcriptional machinery. The other substrates form a heterogeneous group of signaling proteins that have enzymatic functions or function as scaffolding proteins. Accordingly, the main function of HIPKs seems to reside in their ability to augment or repress gene expression. An important and not yet solved issue is the functional redundancy between the various HIPK family members. At the biochemical level it seems that various HIPKs have distinct and also overlapping binding abilities (Kim et al., 1998). Some substrates only interact with one HIPK family member, while others bind to two or even all three kinases (Kim et al., 1998). The partial redundancy is also seen at the genetic level, as the individual knockout of HIPK1 or HIPK2 results only in mild phenotypes, while combined knockout of both kinases leads to death of embryos between 9.5 and 12.5 days postcoitus (Isono et al., 2006).

HIPKs as regulators of cell proliferation

The ability of Yak1 to regulate cell proliferation is conserved in HIPK2. Overexpression of the kinase in cultured cells induces cell cycle arrest (and eventually cell death). Deletion of the Hipk2 gene in mice leads to death of 40% of the animals in the first 3 days after birth. Mouse embryonic fibroblasts (MEFs) from these animals were reported to have an increased proliferation rate (Wei et al., 2007), while another group reported on decreased cell proliferation (Trapasso et al., 2009). As immortalization of MEFs causes the induction of alternative signaling pathways allowing cell survival, and inducible knockout systems will help to clarify this issue. ShRNA-mediated knockdown of HIPK2 in murine bone marrow cells results in up-regulation of the cyclin-dependent kinase (CDK) inhibitor p21^(Waf1/Cip1) and subsequent cell cycle arrest (Iacovelli et al., 2009), showing its critical role in non-immortalized primary cells. Along this line, also knockdown of HIPK2 in fetal liver cells impairs their proliferation (Hattangadi et al., 2010). HIPK2 also phosphorylates the CDK inhibitor p27^kip1 at Ser10, which results in the stabilization of p27^kip1. However, p27^kip1 has no effect on HIPK2-regulated cell proliferation, so that the functional consequences of p27^kip1 phosphorylation remain unclear (Pierantoni et al., 2011). HIPK2 is also a regulator of mitosis, as it has the ability to phosphorylate histone H2B at the midbody, a transient structure connecting two daughter cells at the end of cytokinesis. As HIPK2-mediated phosphorylation of H2B at Ser14 is required for efficient cell cleavage, this kinase contributes to the control of cytokinesis and thus prevents tetraploidization (Rinaldo et al., 2012).

The information on the regulation of cell proliferation by HIPK1 is scarce. HIPK1 was reported to cause p53 Ser15 phosphorylation, which in turn results in increased expression of p21^(Waf1/Cip1) and reduced cell proliferation (Rey et al., 2013). This mechanism may be of special relevance in tumor cells, as HIPK1^-/- mice show no defects in proliferation with the exception of B cells that exhibit impaired cell division in response to B cell receptor cross-linking (Guerra et al., 2012). However, this effect is fully reversible upon addition of CD40L, suggesting that the observed defects in cell proliferation are caused by soluble factors rather than to intrinsic effects on the cell cycle machinery.

The role of HIPKs in the DNA damage response (DDR)

While research on HIPK-mediated cell cycle regulation is still in its infancy and may require simultaneous depletion of more than one HIPK family member, much more is known on HIPK-triggered cell death in response to genotoxic agents. A genetic screen examining regulators of cell death in the Drosophila wing epithelium allowed the identification of the Drosophila Hipk gene as a critical determinant for cell numbers and a collective form of cell death, which proceeds without a final engulfment step (Link et al., 2007). In vertebrates the cell death-regulating functions of HIPKs proceed via p53 family-dependent and -independent pathways.

Several types of genotoxic damage including adriamycin, cisplatin and etoposide, but also UV and ionizing radiation activate HIPK2, however the molecular mechanisms are not known. It may well be that a kinase enhances the constitutive phosphorylation at Tyr354 in the activation loop, as it was reported to occur in response to TGFb stimulation (Shang et al., 2013). It will be important in future experiments to compare the HIPK2 modification patterns between cells that have experienced DNA damage and adequate controls. Quantitative data on these modifications can be obtained using stable isotope labeling by amino acids in cell culture (SILAC) experiments, thus allowing the identification of the molecular switches regulating the activity of HIPK2 and also the other HIPK family members.

While the posttranslational modifications mediating HIPK2 activity in response to DNA damage are not well known, there is ample evidence that genotoxic damage regulates protein amounts of HIPK2 and its associated proteins, as schematically shown in Figure 5. HIPK2 constitutively associates with WIP1 (wild-type p53-induced phosphatase 1), a negative regulator of the ataxia telangiectasia mutated (ATM)-mediated signaling pathway (Choi et al., 2013). In unstressed cells, HIPK2 constitutively phosphorylates WIP1, which is then proteasomally degraded. Ionizing radiation triggers DNA damage and the activity of ATM, which in turn results in AMP-activated protein kinase (AMPK) α2-mediated HIPK2 phosphorylation at Thr112/Ser114 (Choi et al., 2013). This HIPK2 modification results in the dissociation from WIP1, which allows stabilization of this phosphatase and the termination of genotoxic stress signaling upon dephosphorylation of ATM and further proteins such as p53. This signaling pathway provides a negative feed-back loop, which serves to shut down the DDR or which may be instrumental to limit exaggerated DDR signaling. Severe DNA damage leads to stabilization of HIPK2 by a mechanism that prevents its constitutive degradation, which is caused by constitutive binding to SIAH ubiquitin ligases (Winter et al., 2008; Calzado et al., 2009a). DNA damage leads to ATM/ATR-dependent phosphorylation of SIAH1 at Ser19. This phosphorylation causes the disruption of the SIAH1/HIPK2 complex and results in HIPK2 stabilization (Winter et al., 2008). This in turn results in HIPK2-mediated p53 Ser46 phosphorylation and the induction of cell death. However, SIAH1 is not the only ubiquitin E3 ligase regulating HIPK2 abundance in response to DNA damage. Severe DNA damage abrogates WSB-1-mediated HIPK2 ubiquitination and also causes the stabilization of the death promoting kinase (Choi et al., 2008). Under conditions of non-severe DNA damage, activation of p53 allows expression of MDM2, which mediates ubiquitin/proteasome-dependent degradation of HIPK2, thus explaining why under these conditions p53 Ser46 phosphorylation is not seen (Rinaldo et al., 2007). MDM2 also has the ability to modify HIPK1 by degradative ubiquitination. Binding between HIPK1 and its cognate E3 ligase is facilitated by the NORE1 (Novel Ras effector 1, also called RASSF5) protein and accordingly silencing of NORE1, as it occurs in primary adenocarcinomas, leads to stabilization of HIPK1 (Lee et al., 2012).

Figure 5

Mechanisms regulating the stability of HIPK2 and WIP1 in the absence (left) or presence (right) of massive DNA damage.

HIPK2-mediated cell death proceeds via p53 family-dependent and also p53 family-independent pathways. HIPK2-mediated p53-dependent apoptosis critically involves p53 Ser46 phosphorylation (D’Orazi et al., 2002; Hofmann et al., 2002) and it is currently not known whether the other HIPKs can also modify this site. Ser46 phosphorylation allows recruitment of the histone acetylase (HAT) p300, which then acetylates p53 at Lys382 (Hofmann et al., 2002). The p53 Ser46 phosphorylation affects cell cycle modulatory genes only to a minor extent, while it preferentially activates pro-apoptotic p53 target genes. This is also reflected by ChIP-Seq [chromatin immunoprecipitation (ChIP) coupled to massively parallel sequencing] experiments showing the enrichment of Ser46 phosphorylated p53 at chromatin regions controlling the expression of pro-apoptotic genes (Smeenk et al., 2011). HIPK2 and p53 show multiple levels of cross-regulation, thus illustrating the importance of this interaction. The importance of p53-dependent HIPK2 regulation is schematically shown in Figure 6A. On one hand, p53-induced expression of Caspase 6 leads to HIPK2 cleavage and the generation of a truncated and hyperactive form of the kinase (Gresko et al., 2006). However, p53 can also limit HIPK2 activity by proteolytic destruction of the kinase. Non-severe DNA damage allows expression of MDM2, which in turn mediates ubiquitin/proteasome-dependent degradation of HIPK2, thus preventing DNA damage-induced apoptosis (Rinaldo et al., 2007). While these experiments show p53-mediated regulation of HIPK2, conversely HIPK2 can also regulate the p53 protein (Figure 6B). HIPK2 knockdown leads to increased expression of Nox1, a homolog of the catalytic subunit of the superoxide-generating NADPH-oxidase. Nox1 up-regulation in turn induces p53 deacetylation at Lys382, which results in impaired transcriptional activity of p53 (Puca et al., 2010b). Depletion of HIPK2 also results in p53 misfolding, as detected by immunoprecipitation with conformation-specific p53 antibodies (Puca et al., 2008). The p53 misfolding might be attributed to induced expression of metallothionein 2A (MT2A), which is observed after HIPK2 knockdown. As MT2A binds zinc, it might subduct this metal from a zinc-sensitive protein such as p53. Importantly, the supplementation of zinc restores the native p53 conformation in HIPK2-depleted cells (Puca et al., 2011), thus raising the possibility that zinc could be used therapeutically in cells defective for HIPK2 expression.

Figure 6

Mutual regulation of HIPK2 and p53. (A) Effects of p53 on the cleavage and stability of HIPK2. (B) Depletion of HIPK2 inactivates p53 via deacetylation and misfolding. The + and - symbols show the activity status. Further details are given in the text.

HIPK-mediated p53-independent cell death pathways proceed on various routes. UV radiation triggers HIPK2-mediated phosphorylation of the anti-apoptotic transcriptional corepressor CtBP at Ser422. This phosphorylation results in the subsequent proteasome-dependent degradation of CtBP, which eliminates its anti-apoptotic function and thus promotes apoptosis (Zhang et al., 2003). HIPK2 also phosphorylates and promotes proteasomal degradation of ΔNp63α, a dominant negative isoform of the p53 family member p63 with pro-survival functions. Cell death induction by different genotoxic agents depends on ΔNp63α phosphorylation at Thr397 and accordingly expression of a HIPK2-resistant ΔNp63α-Δ390 mutant inhibits cell death (Lazzari et al., 2011). Another example for a p53-independent pathway is provided by the cooperative effect between HIPK2 and Daxx for the induction of TGFβ-triggered cell death in human hepatocellular carcinoma cells deficient for p53 (Hofmann et al., 2003).

While these data collectively suggest a predominantly pro-apoptotic role of HIPK2, the family member HIPK3 may also exert anti-apoptotic functions. Triggering of the death receptor CD95 (Fas/APO-1) leads to apoptosis, which relies on a functional CD95 death-inducing signaling complex (DISC) that comprises CD95, Fas-associated protein with death domain (FADD), procaspase-8, procaspase-10 and c-FLIP (cellular FLICE-like inhibitory protein) proteins (Schleich et al., 2013). CD95 activation also results in JNK activation, which is not important for the execution process of CD95-mediated cell killing (Hofmann et al., 2001) and may even confer resistance to CD95-mediated apoptosis in cancer cells (Curtin and Cotter, 2004). In prostate carcinoma cells, the activation of JNK results in elevated HIPK3 expression that inhibits the interaction between FADD and Caspase 8 by an unknown mechanism (Curtin and Cotter, 2004). Downregulation of HIPK3 restores this interaction and allows the transmission of apoptotic signals, thus suggesting a pro-survival role of HIPK3. The question whether HIPK3 generally exerts an anti-apoptotic function needs to be explored in future studies. Depletion of HIPK2 protects cells from cell death induced by a wide variety of agents ranging from H₂O₂ to UV or adriamycin (D’Orazi et al., 2002; Hofmann et al., 2002; Gresko et al., 2006; de la Vega et al., 2012), suggesting a pro-apoptotic role of this kinase. In contrast, γ-irradiation reduces the survival rate of HIPK2 knockdown cells (Choi et al., 2013), which is consistent with an anti-apoptotic role of the kinase. This finding raises the possibility, that the role of HIPK2 for the induction of cell death may be determined by the nature of the inducing stimulus.

A contribution of HIPKs for metabolic stress

The importance of Yak1 for glucose metabolism in yeast has been preserved for HIPKs, which also regulate glucose metabolism in mammals. Early evidence for the responsiveness of HIPKs to glucose levels came from a study that showed HIPK1-mediated phosphorylation and cytoplasmic relocalization of Daxx in response to glucose deprivation (Song and Lee, 2003). The knockdown of HIPK2 in RKO colon cancer cells showed an important role of this kinase for the regulation of metabolic pathways influencing glucose levels. While glucose starvation induced cell killing in control cells, the knockdown of HIPK2 largely protected the cells from cell death. Metabolomic experiments showed marked changes in choline-containing metabolites and individual amino acids, but it remains to be elucidated which biochemical pathway(s) is directly affected by HIPK2 (Garufi et al., 2013). Glucose metabolism is not only controlled by glycolysis, but also by cell-type specific pathways such as gluconeogenesis as well as glycogen synthesis and decay (Shaw, 2006). It will thus be relevant to test the contribution of HIPK2 and the other canonical HIPKs for metabolic changes of glucose metabolism in suitable cells (such as liver cells) in vivo. While these studies showed a contribution of HIPKs for intracellular events regulating glucose levels, recent data also revealed the relevance of HIPKs for the hormonal regulation of glucose homeostasis. Knockdown of HIPK1, HIPK2 and HIPK3 strongly interferes with glucose-induced secretion of insulin in isolated mouse islets. ChIP experiments revealed binding of all three HIPK family members to the insulin promoter, where they apparently serve as coactivators, as their knockdown decreases insulin transcript levels. Chronic activation of insulin secretion by a high-fat diet results in increased mRNA levels for all canonical HIPKs (Shojima et al., 2012). Similarly, transcript levels of all HIPKs are elevated in mice homozygous for a mutation leading to the spontaneous development of diabetes. Direct evidence for the importance of HIPK3 for insulin release was provided in a mouse model, as HIPK3-deficient animals have a significantly impaired glucose tolerance and insulin response. The reduced insulin secretion in Hipk3^-/- mice occurs only in response to high glucose levels. Decreased insulin secretion is not only caused by impaired transcription of the insulin gene, but also to reduced concentrations of the intracellular messengers responsible for insulin vesicle secretion. Prolonged high-fat diet results in expanded islets in wild-type mice, while the proliferation of islet cells in HIPK3-deficient animals is strongly reduced. Collectively these data show that HIPK3 participates in the control of many aspects in insulin signaling, ranging from islet cell proliferation to the control of insulin transcription and release. Also expression of the beta cell specific transcription factor pancreatic and duodenal homeobox 1 (PDX1) is significantly decreased in the islets of Hipk3^-/- mice fed with a high-fat diet for 12 weeks (Shojima et al., 2012). Interestingly another study showed the HIPK2-mediated phosphorylation of PDX1 Ser269 (An et al., 2010). Elevated glucose concentrations result in a decreased PDX1 Ser269 phosphorylation, concomitant with abnormal changes in its subnuclear localization. It will be interesting to study the consequences of this phosphorylation for PDX1-mediated gene expression, which regulates pancreatic development and mature beta cell function. But HIPKs may also influence insulin signaling by an alternative pathway. Wnt signaling modulates beta cell function on a pathway that involves prevention of glycogen synthase kinase (GSK)-3β phosphorylation and accumulation and nuclear translocation of β-catenin (Welters and Kulkarni, 2008). Consistent with previous reports implicating HIPKs as non-canonical regulators of Wnt signaling (Kanei-Ishii et al., 2004; Lee et al., 2009), GSK3β phosphorylation and β-catenin abundance are decreased in the islets of Hipk3^-/- mice, indicating that HIPK3 has the ability to influence beta cell function via the Wnt signaling pathway. An independent experimental approach suggested the involvement of HIPK3 in the regulation of nutritional stress. Obese or overweight adolescents undergoing an intervention for weight reduction were grouped according to their differential weight loss responsiveness. The two groups responding well or poorly to the weight reduction intervention were subsequently analyzed for DNA methylation patterns. A comparative analysis of CpG methylation identified five differentially methylated regions, including sequences in the Hipk3 gene (Moleres et al., 2013). It will thus be interesting to study whether these differential methylation patterns result in different transcript and protein levels of HIPK3 and to investigate a causative role of HIPK3 for weight reduction responsiveness.

Outlook

The last decade has witnessed rapid progress in our understanding of HIPK function in a vast variety of stress signaling pathways and diseases. However, we still do not understand many important features of these protein kinases. The identification of splicing variants raises the question on the occurrence, regulation and functional relevance of these splicing events. We also need to understand the molecular mechanisms and enzymes that mediate full activation and possible inactivation of these kinases. The functional redundancy of HIPKs can only be addressed with the availability of suitable experimental tools. Thus the field is in need of highly specific antibodies that discriminate between the different endogenously occurring HIPK family members. These tools will also show whether the various HIPK family members occur in the same protein/protein complexes and/or show mutual interaction. The partial functional redundancy can only be overcome with mouse models, allowing the inducible and simultaneous inactivation of more than one HIPK family member. As HIPK protein amounts are tightly regulated, it will be also instrumental to have mice with hypomorphic alleles for individual HIPKs. The elucidation of HIPK protein structures will help to understand the complex interplay between the catalytic domains and the regions that allow transient docking of the kinase to its numerous interaction partners. Given the function of HIPKs as highly interconnected signaling hubs and their involvement in proliferative diseases, it will be exciting to unravel more of their secrets in the future.