Runx2 Trans-Activation Mediated by the Msx2-Interacting Nuclear Target Requires Homeodomain Interacting Protein Kinase-3

Abstract

Runt-related transcription factor 2 (Runx2) and muscle segment homeobox homolog 2-interacting nuclear target (MINT) (Spen homolog) are transcriptional regulators critical for mammalian development. MINT enhances Runx2 activation of osteocalcin (OC) fibroblast growth factor (FGF) response element in an FGF2-dependent fashion in C3H10T1/2 cells. Although the MINT N-terminal RNA recognition motif domain contributes, the muscle segment homeobox homolog 2-interacting domain is sufficient for Runx2 activation. Intriguingly, Runx1 cannot replace Runx2 in this assay. To better understand this Runx2 signaling cascade, we performed structure-function analysis of the Runx2-MINT trans-activation relationship. Systematic truncation and domain swapping in Runx1:Runx2 chimeras identified that the unique Runx2 activation domain 3 (AD3), encompassed by residues 316–421, conveys MINT+FGF2 trans-activation in transfection assays. Ala mutagenesis of Runx2 Ser/Thr residues identified that S301 and T326 in AD3 are necessary for full MINT+FGF2 trans-activation. Conversely, phosphomimetic Asp substitution of these AD3 Ser/Thr residues enhanced activation by MINT. Adjacent Pro residues implicated regulation by a proline-directed protein kinase (PDPK). Systematic screening with PDPK inhibitors identified that the casein kinase-2/homeodomain-interacting protein kinase (HIPK)/dual specificity tyrosine phosphorylation regulated kinase inhibitor 2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole (DMAT), but not ERK, c-Jun N-terminal kinase, p38MAPK, or other casein kinase-2 inhibitors, abrogated Runx2-, MINT-, and FGF2-activation. Systematic small interfering RNA-mediated silencing of DMAT-inhibited PDPKs revealed that HIPK3 depletion reduced MINT+FGF2-dependent activation of Runx2. HIPK3 and Runx2 coprecipitate after in vitro transcription-translation, and recombinant HIPK3 recognizes Runx2 AD3 as kinase substrate. Furthermore, DMAT treatment and HIPK3 RNAi inhibited MINT+FGF2 activation of Runx2 AD3, and nuclear HIPK3 colocalized with MINT. HIPK3 antisense oligodeoxynucleotide selectively reduced Runx2 protein accumulation and OC gene expression in C3H10T1/2 cells. Thus, HIPK3 participates in MINT+FGF2 regulation of Runx2 AD3 activity and controls Runx2-dependent OC expression.

Fibroblast growth factor (FGF)2 signaling plays a critical role in stage-specific control of osteoblast proliferation, gene expression, and bone mass accrual ( 1–5). Hurley and co-workers ( 6–9) were the first to establish the contributions of FGF2 to bone formation in vivo. The skeletal response is biphasic ( 10), but FGF2 is one paracrine mediator of anabolism elicited by the prototypic and clinically useful bone anabolic hormone, PTH ( 7, 9). In our studies of osteoblast gene transcription, we identified that FGF2/FGF receptor signaling and muscle segment homeobox homolog 2 (Msx2)-dependent transcriptional suppression converge at the osteocalcin (OC) FGF response element (OCFRE) ( 11, 12). This element, encompassed by nucleotides −154 to −113 relative to the rat OC transcriptional start site, binds a multiprotein complex containing runt-related transcription factor 2 (Runx2) and the nuclear matrix proteins Ku antigen and Msx2-interacting nuclear target (MINT) ( 13–15). Msx2 binds to MINT and Runx2, thus preventing stable association of Runx2 with the OC gene in native chromatin context ( 14). Systematic analyses revealed that it was the Runx2 component of this complex that possessed the FGF2 responsive trans-activation function (TAF) and that this FGF2-regulated TAF is augmented by MINT ( 14, 15).

We wished to better understand the structural features of the Runx2 TAFs ( 16) that confer regulation by MINT+FGF2( 14). We identify that Runx2 activation domain (AD)3 ( 16) encompasses the TAF targeted for activation by the MINT Msx2-interacting domain (MID) domain in conjunction with FGF2 signaling ( 13, 14). Intact Ser/Thr-Pro motifs in the N-terminal part of Runx2 AD3 are required for MINT+FGF2 activation. Signaling via homeodomain-interacting protein kinase (HIPK)3, a proline-directed protein kinase (PDPK) ( 17, 18), is required for robust OCFRE activation by MINT+FGF2, and HIPK3 regulates Runx2 protein accumulation and OC gene expression in C3H10T1/2 multipotent osteoprogenitors ( 19).

Results

MINT augments FGF2-stimulated Runx2 trans-activation via the MID domain

Our early studies of MINT/Spen demonstrated that the N-terminal RNA recognition motif (RRM) domain could bind a G/T-rich element in the OC promoter ( 13); expression of the N-terminal 90-kDa fragment encompassing the three RRM domains suppressed OCFRE activity ( 13). However, subsequent studies with full-length MINT(1-3576) demonstrated enhancement of OCFRE-dependent transcription, with MINT activation dependent upon the Runx2 osteoblast-specific element 2 cognate present in the OCFRE ( 14). As with the concatemerized OCFRE, in the presence of FGF2, full-length MINT stimulates transcription driven by the OCFRE in the native context of the 0.2-kb OC promoter (the 0.2-kb OC promoter fragment encompasses the Runx2/OC-specific element 2 and the G/T-rich RRM binding motif) (Fig. 1A) ( 14). Activation by MINT was promoter specific, because receptor activator of nuclear factor (NF)-κB ligand (RANKL) and nuclear factor of activated T cells (NFAT)-promoter-luciferase reporter gene (LUC) constructs were not stimulated (Fig. 1A). We wished to further refine the domains of MINT necessary for stimulation of OC gene expression. Therefore, we generated a systematic series of MINT truncation mutants and tested transcriptional activation of the OCFRE using C3H10T1/2 cells, a murine osteoprogenitor line ( 19). For these studies, we took advantage of OCFRE7-Rous sarcoma virus (RSV)LUC, a luciferase reporter construct possessing seven copies of the OCFRE cloned upstream of the RSV minimal promoter. Deletion of the MINT Spen paralog and ortholog C-terminal (SPOC) domain ( 20) altered the dose response but did not eliminate activation of the OCFRE (Fig. 1B). Further C-terminal truncation that removes the Notch-regulating recombination signal-binding protein suppressor of hairless (RBP-J) binding domain (RAM7) domain of MINT ( 21, 22), as in MINT(1-2640), had little additional impact (Fig. 1B). However, MINT(1-2180), lacking an intact MID at residues 2070-2394, completely lost the capacity to trans-activate the OCFRE (Fig. 1B). Western blot analyses for hemagglutinin (HA)-tagged MINT fusion proteins demonstrated expression of all HA-MINT variants (Supplemental Fig. 1 published on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). Of note, all MINT variants studied retained at least two nuclear localization signals and localized to the nucleus (data not shown). Thus, OCFRE activation by MINT requires an intact motif present in residues 2180-2400, separate from the RAM7 and C-terminal SPOC domains (Fig. 1C).

We next examined the impact of N-terminal truncation on MINT activation of the OCFRE ( 13, 23). Deletion of the RRM domain ( 13, 23) encoded by MINT residues 269-522 significantly reduced but did not completely ablate induction (Fig. 1B). This suggested that activation might be independent of intrinsic MINT nucleic acid binding activity. To directly test this notion and to further map this MINT activation function, we constructed a series of enhanced yellow fluorescence protein (EYFP)-MINT-nuclear localization sequence (NLS) fusion proteins that provide a uniform NLS in addition to an EYFP tag (Fig. 1C, lower panel). As shown in Fig. 1D, in the presence of Runx2, the expression of the MID domain fusion protein EYFP-MINT(2030-2640)-NLS enhanced OCFRE7-RSVLUC activity 3-fold as compared with the EFYP-NLS control. By contrast, EFYP-NLS fusion proteins containing the RAM7 and SPOC domains, either alone or in combination, failed to stimulate OCFRE-dependent transcription (Fig. 1D). Western blot analysis for the EYFP tag confirmed expression of all EFYP-MINT fusion proteins (Supplemental Fig. 2). Thus, MINT MID domain (2030-2640) confers OCFRE activation.

The results obtained using EYFP-MINT(2030-2640)-NLS indicated that OCFRE activation could be mediated independent of the intrinsic MINT nucleic acid binding activity but dependent upon Runx2 ( 13, 23). To provide additional evidence to confirm this notion, we implemented the one-hybrid assay. In this assay, transcription is driven by Runx2 TAF present in chimeric Gal4 DNA binding domain (G4)-Runx2 fusion protein. However, the TAF is recruited to a heterologous pFR-LUC via the G4 DNA recognition cognate multimerized upstream of the LUC transcription initiation site ( 14). As shown in Fig. 1E, full-length MINT(1-3576) augmented transcription driven by G4-Runx2 in FGF2-treated cultures. Activation was transcription factor specific, because MINT did not augment activation of G4-upstream stimulatory factor (USF)1 in one hybrid assay (Fig. 1E). MINT(1141-3576) also significantly enhanced FGF2-stimulated G4-Runx2 transcription and was dependent upon Runx2 recruitment via the Gal4DBD (Supplemental Fig. 3). Thus, MINT augments Runx2-dependent transcription in concert with FGF2 signaling. Intrinsic MINT nucleic acid binding activity is not required for stimulating Runx2 TAF. MINT core activation functions mapped between residues 2030–2400, encompassing the MINT MID.

Runx2 residues 316–421, encoding AD3, mediate MINT-dependent trans-activation

We wished to better understand the structural features of the Runx2 TAFs ( 16) regulated by MINT. To map domains that respond to MINT+FGF2, we generated Runx2 variants lacking known regulatory domains (Fig. 2, upper panel) and then tested trans-activation of OCFRE7-RSVLUC in transient transfection assays. Although much less active than full-length Runx2(1-528), Runx2(1-392) retained significant responsiveness to full-length MINT+FGF2 in C3H10T1/2 cells [Runx2(1-392) retains nuclear localization cues but lacks the nuclear matrix targeting signal (NMTS)] (Fig. 3A) ( 24, 25). By contrast, Runx2(1-312) was completely inactive, revealing that the DNA binding Runt domain and N-terminal AD1 and AD2 domains were insufficient to convey MINT+FGF2 responsiveness. Moreover, Runx2(45-528) and Runx2(105-528), lacking Runx2 AD1 and AD1+AD2, respectively, were still capable of conveying significant responsiveness to MINT in the presence of FGF2 stimulation (Fig. 3A). Thus, AD1, AD2, and the NMTS were not required for Runx2 responses to MINT+FGF2.

We next generated a series of internal deletion constructs to remove AD3 and its adjacent segments in the context of Runx2(105-528). Runx2(105-528; Δ339-374) and Runx2(105-528; Δ339-396) conveyed MINT responsiveness to an extent similar to that of Runx2(105-528) (Fig. 3A). However, Runx2(105-528; Δ302-396), lacking residues 302-339 in AD3 ( 16), exhibited significantly reduced activity vs. Runx2(105-528) (for Runx2(105-528; Δ302-396), see Fig. 3A, asterisk; residues 302-339 underlined in Fig. 3B). We wished to further refine the features of Runx2 AD3 that convey MINT+FGF2 responsiveness. Therefore, we took advantage of our observation that the related family member, Runx1(1-451), conveys at best only modest induction of the OCFRE7-RSVLUC, with little if any response to full-length MINT (see below). We generated and functionally evaluated a series of Runx1:Runx2 chimeric proteins, including the placement of Runx2 AD3 and surrounding elements in the equivalent region of Runx1 (see Materials and Methods and Fig. 3B). Runx1 residues Ser²⁶⁴ through Ala³⁴² were replaced in contiguous open reading frame by Runx2 residues Gln³¹⁶ through Ala⁴²¹, generating the chimeric protein Runx1(Met¹-Ser²⁶³):Runx2(Gln³¹⁶-Ala⁴²¹):Runx1 (Met³⁴³-Tyr⁴⁵¹). For brevity, this chimera is denoted as Runx1:Runx2(316-421). As shown in Fig. 3C, Runx2 residues 316-421 conveyed MINT+FGF2 responsiveness onto Runx1 [compare Runx1:Runx2(316-421) with Runx1(1-451)]. However, the chimera Runx1:Runx2(316-360) that lacked the C terminus of Runx2 AD3( 16) was inactive (Fig. 3C). Thus, unique structural motifs present in Runx2(316-421), a 106 amino acid sequence encompassing the AD3 core ( 16), encodes information necessary and sufficient to convey MINT responsiveness onto the unresponsive heterologous family member, Runx1.

Proline-directed kinase motifs in Runx2(316-421) are important for MINT-dependent stimulation

We wished to further define the region of Runx2(316-421) conveying MINT+FGF2 responsiveness. Runx2 residues 316-421 encompass three Ser and Thr residues immediately upstream and adjacent to Pro residues (Fig. 3B). PDPKs recognize such motifs as substrates, and the FGF2/FGF receptor utilizes a number of cellular PDPKs to mediate signal transduction ( 26). To identify whether the putative PDPK motifs in Runx2(316-421) participate in activation by MINT+FGF2, we first performed Ala mutagenesis in the Runx1:Runx2 chimera (for a schematic representation, see Fig. 4A); we replaced codons for Thr³¹⁸-Ser³¹⁹, Thr³²⁵-Thr³²⁶, and Thr⁴⁰⁷-Thr⁴⁰⁸ with Ala-Ala codons to disrupt PDPK cognates in the N-terminal aspect of the Runx2 AD3 domain (for convenience, only the amino acid immediately preceding the Pro residue is denoted). As shown in Fig. 4B, Runx1:Runx2(316-421:319A;326A) and Runx1:Runx2(316-421:326A) exhibited significantly reduced responsiveness to MINT as compared with Runx1:Runx2 (316-421). By contrast, Runx1:Runx2(316-421:408A) was unaffected. Mutation of the three central PDPK sites (Fig. 4B), as in Runx1:Runx2(316-421:319A,326A,408A), did not further diminish the response to MINT but retained the reduced activity level of Runx1:Runx2(316-421:326A) (Fig. 4B, asterisks). Thus, all mutants pointed to the functional importance of the PDPK cognate at Thr³²⁶. Of note, protein kinase C and casein kinase-2 (CK2) cognates as Ser330, Ser364, Ser347, and Ser388 were also mutated to Ala, but none of these individually contributed to MINT/FGF2-dependent Runx2 trans-activation (data not shown).

At the N-terminal junction of the Runx1-Runx2 chimera, we noted a 4th putative PDPK cognate, a Ser-Pro motif conserved in both Runx1 and Runx2 (Fig. 3B); Runx1 Ser-249 is a PDPK cognate equivalent to Runx2 Ser-301 (Fig. 3B, bold underline). Furthermore, we realized that our previous Runx2 AD3 internal deletions had left Ser-301 intact with an adjacent Pro residue (Fig. 3, A and B). Thus, this 4th PDPK cognate was mutated in the context of the Runx1:Runx2(316-421) chimera (Fig. 4A, upper panel). As shown in Fig. 4C, mutation of that single cognate, as in Runx1(249A):Runx2(316-421), significantly reduced MINT+FGF2 stimulation. Mutation of all four of these AD3 PDPK cognates in the chimera had a profound effect; unlike Runx1:Runx2(316-421), Runx1(249A):Runx2(316-421:319A,326A,408A) chimera was completely unresponsive to MINT+FGF2 (Fig. 4C). Finally, deletion of Runx2 residues 300-397 in the native Runx2(105-528) context, thus removing all of AD3, including Ser³⁰¹, abrogated MINT+FGF2 activation with little effect on basal activity [see Runx2(105-528;Δ300-394) in Fig. 4D].

To confirm these observations, we generated a series of Ala mutations and Asp “phosphomimetic” alterations in Runx2 AD3 within the context of full-length native Runx2 (for a schematic representation, see Fig. 4A, lower panel). As shown in Fig. 4E, mutation of the PDPK cognates in residues Ser-301, Ser-319, and Thr-326 significantly reduced MINT activation of Runx2 without altering basal transcription. By contrast, introduction of the negatively-charged, phosphomimetic Asp side chain at residues 301, 319, and 326 in Runx2 AD3 significantly enhanced activation by MINT (Fig. 4F). As in the Runx1:Runx2 chimera, Ser-301 was critical for MINT regulation in the context of full-length Runx2 (Fig. 4F). Of note, Western blot analyses confirmed equivalent levels of these Runx2 protein variants (Supplemental Fig. 4). Thus, PDPK cognates at Ser³⁰¹ and Thr³²⁶ in Runx2 AD3 contribute to Runx2 trans-activation by MINT+FGF2.

A 2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole (DMAT)-sensitive kinase participates in AD3 activation by MINT+FGF2

Because our structure-function studies indicated that a member of the PDPK family was necessary for MINT+FGF2 activation of Runx2, we screened a series of pharmacological inhibitors of known PDPKs. Additionally, we included analysis of CK2, a kinase with cognates in Runx2(316-421) that did not contribute to MINT+FGF2 activation (data not shown). C3H10T1/2 cells were transiently transfected with OCFRE7-RSVLUC reporter, full-length Runx2, and/or MINT expression plasmids. Twenty-four hours later, transfected cultures were treated with vehicle (dimethylsulfoxide 0.1%) or inhibitors at doses equal or higher than their IC₅₀ for 1 h before and during an 18-hr treatment with FGF2 or its vehicle (PBS/BSA). As shown in Fig. 5A, the ERK1/ERK2 inhibitor PD98059 did not significantly diminish MINT+FGF2 transcriptional activation, whereas SB203580 (p38MAPK antagonist) had no effect. The c-Jun N-terminal kinase inhibitor SP600125 was also ineffective (10 μm) (data not shown). By contrast, and to our initial surprise, DMAT, a unique benzimidazole that inhibits dual specificity tyrosine phosphorylation regulated kinase (Dyrk)/HIPK proline-directed kinases, as well as CK2, profoundly reduced MINT+FGF2 activation of OCFRE7-RSVLUC (Fig. 5A). However, the action of DMAT was not mediated via inhibition of CK2, because other specific CK2 inhibitors, such as emodin and ellagic acid, did not antagonize MINT+FGF2 activation (Supplemental Fig. 5). Inhibition by DMAT was dose responsive, exhibiting an IC₅₀ value less than 5 μm (Fig. 5, B and C). Moreover, DMAT inhibitory effects were promoter specific, because NF-κB-LUC and TOPFLASH reporter activities were not reduced by DMAT in transfected cells (Fig. 5C). To confirm that a DMAT-sensitive activity was necessary for functional Runx2 trans-activation, we tested the impact of DMAT in the one-hybrid system described above. As shown in Fig. 5D, 10 μm DMAT significantly reduced MINT(1141-3576) trans-activation of G4-Runx2. The alternative CK2 inhibitors emodin and ellagic acid failed to inhibit MINT+FGF2 activation of G4-Runx2 (activation actually enhanced by 20 μm emodin) (Fig. 5D). Thus, a DMAT-sensitive kinase that is not CK2 is required for MINT+FGF2 activation of Runx2 AD3.

HIPK3, a DMAT-sensitive PDPK, supports MINT+FGF2 activation of Runx2

DMAT has been shown to potently inhibit four types of kinases: CK2, proviral integration site for murine leukemia virus kinases, PKD1, and the Dyrk/HIPK family of PDPKs ( 27). Because other well-characterized CK2 inhibitors had no effect, we discounted CK2 as a mediator. Furthermore, emodin, a potent CK2 inhibitor that also inhibits proviral integration site for murine leukemia virus kinase ( 27), did not inhibit Runx2 activation by MINT+FGF2 but actually enhanced responsiveness. Although PKD1 does participate in Runx2 regulation ( 28), it is not a PDPK, and PKD1 regulation maps to the Runt domain ( 28). However, Dyrk/HIPK family members are DMAT-sensitive PDPKs ( 17, 29, 30). Therefore, we studied the impact of depleting each Dyrk/HIPK family member upon MINT+FGF2 activation of the OCFRE. The eight family members evaluated (Dyrk1a, Dyrk1b, Dyrk2, Dyrk3, Dyrk4, HIPK1, HIPK2, and HIPK3) were all targeted by small interfering RNA (siRNA) screening, with CK2 subunit targeting included as one type of negative control (Fig. 6A) (data not shown). A scrambled siRNA served as a second negative control, and anti-LUC siRNA duplex served as a positive control for RNA interference within this transient transfection/LUC reporter assay system. Of all the Dyrk/HIPKs targeted by RNAi, only siRNA directed toward HIPK3 significantly inhibited MINT+FGF2 activation of the OCFRE7-RSVLUC reporter (Fig. 6A) (data not shown). One siRNA directed toward HIPK1 also consistently resulted in a nonsignificant trend for reducing promoter activation (Fig. 5A). However, RT-quantitative PCR (qPCR) analysis demonstrated that this HIPK1 siRNA partially down-regulated the related HIPK3 mRNA in addition to HIPK1, whereas HIPK3 RNAi was specific for HIPK3 message (Fig. 6B). Moreover, the HIPK3 siRNA efficiently down-regulated enhanced green fluorescent protein (EGFP)-HIPK3 protein accumulation (Supplemental Fig. 6). Consistent with our pharmacologic data, despite a reduction of 80% in endogenous CK2α levels (data not shown), no changes were observed for MINT+FGF2 activation of OCFRE reporter (Fig. 6A). To confirm these data, the impact of HIPK3 RNAi was tested on activation via Runx2 AD3 using the Runx1:Runx2(316-421) chimera. Once again, HIPK3 RNAi significantly reduced MINT+FGF2 activation as compared with the scrambled control siRNA (Fig. 6C). Thus, HIPK3 participates in MINT+FGF2 activation of Runx2 AD3 in C3H10T1/2 mesenchymal cells.

HIPK3 binds Runx2 and MINT and colocalizes with MINT in the nuclear matrix

The functional requirement of HIPK3 for MINT- and FGF2-dependent activation of the Runx2-responsive OC promoter prompted us to look for evidence of physical interaction and intracellular proximity. Thus, coprecipitation studies were performed, initially using unlabeled V5-epitope-tagged HIPK3 “bait” protein to interrogate potential prey proteins radioactively labeled with ³⁵S-methionine (all proteins generated by coupled in vitro transcription-translation). Osteopontin (OPN) was chosen as a conveniently available negative control. Unlabeled V5-tagged HIPK3 coprecipitates radiolabeled Runx2 but not radiolabeled OPN (Fig. 7A). Conversely, unlabeled Runx2 coprecipitates radiolabeled HIPK3 but not OPN (Supplemental Fig. 7). We next examined the subcellular localization of HIPK3 with respect to the nuclear matrix protein MINT ( 13, 14). Although immunovisualization of endogenous HIPK3 with a polyclonal antibody exhibited some cytosolic staining, the strongest signal was nuclear (Fig. 7B, upper panel), colocalizing with MINT in the nuclear matrix ( 13). To confirm nuclear colocalization, we assembled a construct coding for HIPK3 fused to the C terminus of the GFP (EGFP-HIPK3). As observed with endogenous HIPK3, EGFP-HIPK3 also colocalized with MINT in C3H10T1/2 cells (Fig. 7B, lower panel). Moreover, in coprecipitation assays, HIPK3 interacts with the MINT MID domain but not with OPN (Fig. 7C). Thus, HIPK3 participates in protein-protein interactions with both Runx2 and MINT and colocalizes with MINT in the nuclear matrix ( 13).

Recombinant HIPK3 recognizes Runx2 AD3 as a kinase substrate in vitro

Because 1) DMAT treatment and HIPK3 RNAi inhibited MINT+FGF2 activation of Runx2 AD3; 2) intact Ser/Thr-Pro PDPK cognates in this region were required for AD3 activation by MINT+FGF2; and 3) phosphomimetic Asp substitutions at these Runx2 AD3 Ser/Thr-Pro motifs enhanced activation by MINT, we hypothesized that the PDPK activity of HIPK3 would recognize Runx2 AD3 as a substrate. To directly test this notion, we examined whether recombinant purified HIPK3 could catalyze radiolabeled phosphate transfer from γ-[³²P]-ATP to a Runx2 AD3 synthetic peptide corresponding to residues Tyr³¹³ to Arg³³², thus encompassing the proline-directed kinase cognates specific to Runx2 AD3 and not shared with Runx1(Fig. 8A). Two additional Arg residues were added to the C terminus to facilitate interaction of the peptide to phosphocellulose in this filter binding assay. As shown in Fig. 8A, HIPK3 increases [³²P]-Runx2(313-332) formation in a time- and concentration-dependent fashion. Moreover, replacement of all hydroxyl amino acids immediately upstream of the Pro residues in this AD3 segment significantly reduced radiolabeled phosphate incorporation (Fig. 8A). Heating HIPK3 for 5 min at 95 C before addition of peptide substrate inactivated the enzyme and reduced radiolabeled phosphate incorporation to background (data not shown). Moreover, DMAT dose dependently inhibited HIPK3 phosphorylation of the Runx2 AD3 domain, with an IC₅₀ approximately 100-fold lower than that of the potent CK2 inhibitor emodin (Fig. 8B). Thus, the PDPK activity of HIPK3 phosphorylates Runx2 AD3 residues (313-332) in vitro, and phosphorylation is potently inhibited by DMAT.

HIPK3 supports Runx2 protein accumulation and OC gene expression

The steady-state accumulation of Runx2 protein is an important aspect of Runx2-dependent gene regulation ( 31–33). FGF2 treatment increases Runx2 protein levels in C3H10T1/2 cells, and this increase is inhibited by DMAT treatment, suggesting a HIPK3-dependent mechanism (Supplemental Fig. 8). To test this notion, we examined the effect of depleting HIPK3 in this same assay. Transfection of cells with HIPK3 antisense oligodeoxynucleotide phosphorothioate (ASO) recapitulated the effects of DMAT treatment, inhibiting basal and FGF2-stimulated Runx2 protein accumulation (Fig. 9A). Moreover, HIPK3 ASO resulted in a 46% decrease in endogenous OC mRNA expression (P < 0.001) compared with cells transfected with the control sense-strand oligonucleotide (Fig. 9A). Col1A1 mRNA accumulation was unaffected by HIPK3 ASO (Fig. 9B). Thus, depletion of HIPK3 selectively inhibits OC gene expression and FGF2-regulated Runx2 protein accumulation in C3H10T1/2 mesenchymal cells.

Discussion

The large size of MINT has made detailed structure-function analyses of this nuclear matrix transcription factor very difficult ( 13). However, based upon protein-protein interactions and protein-nucleic acid interactions, five domains [the C-terminal SPOC domain (3410-3576; murine sequence), the N-terminal RRM-containing domain (269-522), the central RBP-J-interacting domain RAM7 (2638-2777), the adjacent MID (2070-2394), and the overlapping nuclear receptor-interacting domain (RID) (2148-2637)] have been identified as playing critical roles in MINT-regulated gene transcription (Fig. 1C). Honjo and co-workers ( 22) first identified the important role of MINT as an inhibitor of Notch1 signaling; via its RAM7 domain, MINT sequesters and prevents RBP-J DNA binding, thus precluding trans-activation directed by the RBP-J coadapter, Notch1 intracellular domain ( 22). As such, MINT fulfills a function of Drosophilan Hairless, a negative regulator of RBP-J that is not present in the vertebrate genome. The C-terminal SPOC domain participates in transcriptional suppression via type I histone deacetylase recruitment ( 23, 34), and we have confirmed the results of Honjo and co-workers ( 22) in our own mammalian one-hybrid assays (our unpublished observations). The MINT SPOC domain may also be involved in Notch inhibition ( 34). However, our data demonstrate that the MINT SPOC domain is not required for Runx2 regulation of the OCFRE. The N-terminal RRM domain of human MINT was shown by Evans and co-workers ( 23) to recognize steroid receptor RNA activator, an RNA-based nuclear receptor coactivator. Their studies also identified a nuclear receptor binding motif called RID located in MINT residues 2148-2637 (numbered per murine sequence for coregistration) ( 23). The RID segment of MINT partially overlaps the MID that we identified upon the interaction cloning of MINT ( 14). We now identify that the MINT MID segment also encompasses the core activity for MINT+FGF2 activation of the Runx2 AD3. Although contributions are certainly made by the N-terminal RRM domain, the MID domain encompasses the core function of MINT-mediated Runx2 activation. Consistent with this, MINT+FGF2 activation of OCFRE7-RSVLUC is exquisitely sensitive to even very low levels of Msx2 ( 14), inhibited by concentrations of Msx2 expression plasmid 2 log units lower than those required for Runx2 or MINT responses (our unpublished observations).

Our structure-function studies also newly indentify HIPK3 in the signaling pathway by which FGF2 and MINT stimulate Runx2 trans-activation. HIPK3 is required for OCFRE activation, binds Runx2 and MINT, phosphorylates Runx2 AD3, controls AD3 activation by MINT+FGF2, and is necessary for FGF2-mediated enhancement of Runx2 protein accumulation. Thus, we conclude that this functionally and physically interacting Runx2 kinase is a key mediator of MINT+FGF2-regulated OC gene expression (Supplemental Fig. 9). Whether the MINT MID domain is also phosphorylated by HIPK3 is being investigated. Based upon the data of Franceschi and co-workers ( 36) and Franceschi et al. (4, 37), we anticipate that osteoblast HIPK3 and ERK1/2 signaling will sequentially participate in the stage-specific control of OC gene expression. Because DMAT is much more effective than PD98059 in inhibiting Runx2 actions in C3H10T1/2 osteoprogenitors, HIPK3 is like to play a critical role upstream of ERK1/2 during the sequential elaboration of the osteoblast phenotype. However, this notion remains to be tested in vivo.

Of the HIPK family members described to date, HIPK1 and HIPK2 have been the most extensively studied. Ito recently demonstrated that Runx1 is phosphorylated by HIPK2, a crucial step in sequential Runx1-p300 modifications that direct subnuclear localization of transcriptional complexes ( 38). HIPK1 and HIPK2 knockout mice are grossly normal, but HIPK1^−/−;HIPK2^−/− mouse embryos die in utero due to severe defects in vasculogenesis and angiogenesis ( 39). This embryonic-lethal phenotype likely reflects impaired regulation of Runx1-dependent transcriptional programs ( 39). By comparison, HIPK3 has been only superficially studied. Known functions of HIPK3 include augmentation of androgen receptor ( 40) and c-Jun (41) transcriptional responses. Via an independent line of biochemical investigation, we have identified that HIPK3 also participates in Runx2 transcriptional regulation by MINT+FGF2. This adds to emerging data indicating an important role for the HIPK family in Runx-dependent gene transcription ( 38). To date, the phenotype of HIPK3-null mice has not been reported. Given the inability of the intact HIPK3 gene to support early development in HIPK1^−/−;HIPK2^−/− mice ( 39), and the unique contributions of HIPK3 to Runx2-dependent trans-activation, we anticipate that HIPK3 will serve unique physiological roles during later vertebrate development, including skeletogenesis.

Of note, FGF2 ( 6, 42), Runx2 (43), and Msx2 ( 44–47) signals significantly contribute to diseases of heterotopic calcification ( 48, 49), as well as to orthotopic bone formation ( 50, 51). All of these signals, along with osteoregulatory Notch signaling ( 48), functionally converge upon MINT. Thus, targeting the HIPK3-MINT interaction with novel antagonists, or DMAT-like kinase inhibitors, may offer new therapeutic strategies for limiting aberrant osteogenic mineralization as occurs in calcific aortic stenosis ( 52, 53). Our future studies will examine the potential impact of HIPK3 inhibition on valvular and vascular calcification, using in vitro and in vivo models of disease.

Materials and Methods

Chemicals, reagents, and antibodies

Protein kinase inhibitors emodin, apigenin, 1,3-dihydro-3-[(2,4,6-trimethoxyphenyl) methylene]-2H-indol-2-one, DMAT, and PD98059 were obtained from Sigma Chemical Co. (St. Louis, MO). The following antibodies were used: mouse anti-Runx2 8G5 was from MBL International, Inc. (Naka-ku Nagoya, Japan). Mouse anti-V5 epitope tag Clone V5–10, mouse M-2 anti-Flag, and rabbit anti-Runx1 antibody (N terminal; R-0406) were obtained from Sigma-Aldrich (St. Louis, MO). Rabbit anti-His tag (sc-803), goat polyclonal anti-HIPK3 (sc-46360), and goat polyclonal anti-CK2 (sc-6480) were obtained from Santa Cruz Biotechnology, Inc. (San Diego, CA). All other biochemical reagents were purchased from Fisher Scientific (Pittsburgh, PA), Invitrogen Corp. (Carlsbad, CA), Millipore Corp. (Billerica, MA), QIAGEN (Valencia, CA), or Sigma Chemical Co. as indicated.

Reporter and expression plasmids

As previously detailed ( 11, 15), the reporter for OCFRE activity is a seven-copy concatemer of rat OC promoter nucleotides −154 to −113 that places cognates for Runx2 and Ku antigen upstream of the minimal RSV promoter and LUC cDNA. The cytomegalovirus (CMV)-LacZ control vector ( 14, 15) and the OC-LUC ( 54) and RANKL-LUC (55) promoter-reporter constructs been previously described. The latter kindly provided by Xu Feng. NFAT-LUC and NF-κB-LUC reporters were purchased from Stratagene (La Jolla, CA). Full-length MINT (amino acids 1-3576; also known as murine SHARP or SPEN) corresponds to RefSeq NM_019763 as initially reported and cloned by us ( 13, 14). The cDNAs for wild-type Runx2 (Osf2/Cbfa1; amino acids 1-528; Expasy Q08775-5 or National Center for Biotechnology Information NP_001139510) ( 16, 56) and Runx1 (PEBP2αB1, murine AML1b; amino acids 1-451; Expasy isoform 1 Q03347-1, National Center for Biotechnology Information isoform 3 NM_001111023.1) ( 57, 58) have been previously reported. The pFR-LUC reporter and G4-Runx2 and G4-USF1 expression plasmids used in one hybrid analyses have been previously described ( 14, 59).

To prepare N- or C-terminal truncation variants of MINT, full-length MINT cDNA originally cloned from MC3T3-E1 calvarial osteoblasts ( 14) was used as a template in high-fidelity PCRs, and fragments were ligated into pCMV-HA (cytomegalovirus promoter-HA tag) vector as previously detailed (Fig. 1D) ( 14). Likewise, to study the contribution of short stretches of Runx2 within the context of Runx1 and their interaction with MINT, the distinct regions of Runx2 indicated in Fig. 2 were PCR amplified and inserted in-frame into unique endonuclease restriction sites found along Runx1 isoform 1 cDNA (BamH1, SacI, SalI, and SphI). Murine HIPK3 ( 17) was PCR amplified using IMAGE clone no. 40110751 as a template, then cloned into pcDNA3.1-V5-His vector. HIPK3 was also cloned in open reading frame onto the C terminus of the EGFP plasmid (CLONTECH, Mountain View, CA) and used in transient transfection to image the intracellular distribution of EGFP-HIPK3. Vectors and methods for one-hybrid analysis of Runx2 have been previously described ( 14). All plasmid inserts were sequenced to verify fidelity, and protein expression validated by coupled transcription/translation (Promega, Inc., Madison, WI), Western blot analysis, and/or fluorescence imaging.

Cell culture and transient transfection assays

Our methods of cell culture and transiently transfect ion have been previously described ( 14, 15). Briefly, mouse C3H10T1/2 cells were passaged in 10% fetal bovine serum and basal medium eagle and then plated at 70% of confluence in 24-well plates (1.9-cm² area/well) for transfection studies. Next day, cells were transiently transfected with 600 ng of total plasmid DNA/well with 2.5 μl of Superfect dendrimer reagent (QIAGEN) in 30 μl of OptiMEM (Invitrogen Corp.). After 4 h of transfection, cultures were allowed to recover overnight in complete growth medium, then treated with 25 ng/ml of human FGF2 (R&D Systems, Minneapolis, MN) for another 18 h. Cultures were harvested for luciferase reporter activity (routine β-galactosidase normalization) as previously detailed ( 14, 15).

Small interfering RNA- and antisense-mediated knock-down methods

To systematically screen the DMAT-sensitive Dyrk family protein kinases that might participate in MINT+FGF2 activation of the OCFRE, siRNAs were transiently cotransfected along with the indicated Runx2 expression and LUC-reporter plasmids. Anti-LUC siRNA (Dharmacon, ThermoFisher Scientific, Inc., Lafayette, CO) served as positive control for RNA interference and was taken to define maximum silencing. The AllStars control siRNA (QIAGEN) was used as the negative control for initial screening, with subsequent revalidation using scrambled siRNA control sequence for HIPK3 (see below). Briefly, siRNAs were dissolved in serum-free medium containing Mirrus Transit-TKO reagent (Mirrus Co., Madison, WI), incubated, and pooled with plasmid-liposome mixture prepared separately, using diluted Superfect dendrimer transfection reagent and plasmid dissolved on OptiMEM. Complete growth medium was added to bring siRNA concentration to 100 nm, gently mixed, and then added to adherent C³H10T1/2 cells. After a 6-h incubation, cultures were turned to growth media and then replaced the next morning with fresh medium supplemented with either vehicle or FGF2 as previously detailed. Cell cultures were harvested for either LUC reporter activity or RT-qPCR using methods previously described (n = 3–6 per treatment condition, repeated at least twice) ( 14, 15). Gene silencing was confirmed by RT-qPCR or by Western blot analysis. For assessing knock-down of the EGFP-HIPK3 fusion protein, cultures cotransfected with CMV-EGFP-HIPK3 and either control or anti-Hipk3 siRNA as above, then fixed with 2% paraformaldehyde and evaluated by fluorescence microscopy in lieu of extract preparation. The siRNA duplexes were directed against the following murine DNA sequence targets: CK-2α, CAGAAGATTTATATGACTATA; CDK2, GTTGTGGCGCTTAAGAAGATC; Hipk1, CTGATTCATGCTGACCTTAAA; Hipk2, CGGGAGTTCATTGACCTGTTA; and Hipk3, CAGGAGTAATTCATTGCAGAA. Control siRNAs were either the QIAGEN AllStars RNAi control (CAGGGTATCGACGATTACAAA) or a scrambled Hipk3 duplex sequence (ATACCAGTGGCATAAGTAGAT). All other siRNA duplex sequences were obtained commercially via QIAGEN GenomeWide Design. These sequences are available upon request. For quantifying Hipk family members, specific TaqMan reagents purchased from Applied Biosystems (Foster City, CA) were Mm00501689_m1 (Hipk1), Mm00439329_m1 (Hipk2), and Mm00468880_m1 (Hipk3). Other primer set sequences are available upon request.

Gene silencing using antisense oligonucleotides

Phosphorothioate antisense oligonucleotide (GGTAGACCAAGACTTGTGAG) directed against HIPK3 or the sense-strand control (CTCACAAGTCTTGGTCTACC) were transfected similarly to siRNA methods described above, but the concentration used was 40 nm. pCMV-Runx2 was cotransfected at 0.1 μg/cm²·well. Forty-eight hours after transfection, cells were treated with FGF2 (25 ng/ml) and then harvested 24 h later. Whole-cell lysates harvested in 2×Laemmli sodium dodecyl sulfate sample buffer were processed for immunodetection of Runx2 and the housekeeping gene eIF2α as indicated.

Confocal microscopy

C³H10T1/2 cells were plated and transfected with CMV-EGFP-HIPK3, pCMV-HIPK3-V5, pCMV-Runx2, or p-CMV-HA-MINT-FLAG plasmids and incubated in normal growth conditions for 24 h. Cells were washed with ice-cold PBS and fixed with 4% paraformaldehyde in PBS for 10 min. After permeabilization in 0.1% Triton X-100/PBS, cells were washed and incubated in blocking buffer (2% BSA in PBS + 0.1% Triton X-100), and incubated with primary antibodies with gentle shaking at 4 C overnight. For Runx2, mouse monoclonal anti-Runx2 clone 8G5 (MBL International Corp., Woburn, MA) was used at 1:1000. Mouse monoclonal M2 anti-Flag antibody (1:500) was used to detect MINT (epitope tag added to the C terminus). For endogenous HIPK3, goat polyclonal anti-HIPK3 (sc-46360 N-13; Santa Cruz Biotechnology, Inc.) used at 1:100–200. Donkey anti-IgG secondary antibodies conjugated with fluorochromes Alexa 488 or Alexa 647 were added at 1:400 for 1.0 h and washed. Slides were mounted with Vecta-Mount aqueous media, sealed with nail polish, and imaged in a Bio-Rad Radiance 2000-Nikon TE300 inverted microscope. Confocal microscopy images were acquired by λ strobe sequential scanning at four-six zoom in 1024 × 1024 pixel files, and one-two Airy discs of pinhole. Grayscale picture files were deconvoluted with a constrained iteration deblurring algorithm set at 1.5 of brightness and two iterations (Microtome Deconvolution, VayTek, Fairfield, IA).

Coimmunoprecipitation reactions

Coprecipitation/pull-down assays were carried out essentially as described ( 14). Runx2, V5-His-tagged HIPK3, T7-tagged-MID MINT, or OPN cloned in T7-promoter driven plasmids were labeled in vitro with ³⁵S-methionine in T7 polymerase-driven transcription-translation assays (Rabbit reticulocyte lysate; Promega, Inc.). Unlabeled proteins were prepared similarly using nonradioactive amino acids. Protein-protein interaction reactions were set up by mixing unlabeled HIPK3-V5 proteins (5 μl of in vitro reaction) with 5 μl of either radiolabeled OPN (as a negative control) or radiolabeled Runx2 and MID-MINT as indicated and incubated for 1.0 h on ice with occasional mixing in coimmunoprecipitation buffer [50 mm Tris-HCl (pH 7.5), 1.5 mm EGTA, 100 mm NaCl, 0.1% Triton X-100, 10 mm β-glycerol phosphate, 1 mm orthovanadate, and freshly added protease/phosphatase inhibitors]. Reciprocally, unlabeled Runx2 was incubated with radiolabeled OPN or HIPK3 and incubated as above. Immunoprecipitation directed toward the unlabeled protein partner was carried out by adding 2 μg of either mouse monoclonal anti-Runx2 antibody to the unlabeled Runx2 reactions or mouse monoclonal anti-V5 epitope tag or rabbit anti-His tag to pull-down unlabeled V5-tagged HIPK3 reactions and incubated for 1.0 h on ice. Subsequently, 50 μl of protein A/G-sepharose (25% bed volume) was added to each 400 μl coimmunoprecipitation reaction and mixed end-to-end for 1.0 h at 4 C. Immune complexes were washed three times in 1.0 ml in radio-immunoprecipitation assay buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate in Tris-buffer normal saline) and centrifuged at 3000 rpm for 1 min. Laemmli sample buffer was added, each sample heated for 5 min at 95 C, proteins resolved by SDS-PAGE, and visualized by autoradiography after fixation and fluoro-enhancement as previously detailed ( 60).

In vitro HIPK3 kinase assay

To determine whether Runx2 AD3 served as a substrate for phosphorylation by HIPK3, a peptide encompassing Runx2 residues 310-331 (YLSQMTSPSIHSTTPLSSTRRR) or a mutated control peptide (YLSQMVAPSIHAVVPLSSTRRR) were evaluated as kinase substrates in an in vitro assay using purified human HIPK3 (amino acids 161-562; Millipore Corp.). General guidelines were followed as described by Hardie ( 61). Briefly, 25 μl reactions were prepared in 8 mm 3[N-morholino]propanesulfonic acid-0.2 mm EDTA buffer (pH 7.0) with 10.7 mU of HIPK3 (10 ng), and either Runx2 AD3 or control peptide at 200 μm; reaction was started by adding 5 mm Mg/0.01 mm ATP. Radiolabeled γ [³²P]-ATP was premixed with Mg-ATP to achieve a final specific activity of 2200 dpms/pmol·reaction. Reactions were incubated for 30 min at 30 C and stopped with 5 μl of 3% ortho-phosphoric acid. Half volume (15 μl) was spotted on a prewetted P81 phosphocellulose Millipore Multiscreen 96-well plates, and unincorporated radioactivity was removed by three 5-min washes of 75 mm ortho-phosphoric acid and a final methanol wash using a vacuum manifold (Millipore Corp.). Radioactivity was determined by Cerenkov counting of the P81 filters after collection with a Millipore Multiscreen well punch. For studies incorporating the kinase inhibitors DMAT and emodin, inhibitors were added to the reaction at the final concentration before addition of the substrate Runx2 AD3 peptide and kinase activity then quantified as described above.

Statistical analysis

All data points are presented as the mean ± se of independent replicates (n ≥ 3), with each experiment carried out at least twice. All statistical analyses performed by one-way ANOVA or Chi-squared analysis as indicated using GraphPad InStat Software (version 3.06 for Windows; GraphPad, La Jolla, CA). When the ANOVA P was significant (P < 0.05), analysis for intergroup differences was then performed using either Student-Newman-Keuls (for all pair-wise comparisons) or Dunnett’s (for comparison vs. control) post hoc tests for multiple comparisons with GraphPad Instat. For Chi-squared analysis, two-sided Fisher’s exact test was implemented to evaluate statistical significance.

Acknowledgments

This work was supported by National Institutes of Health Grants AR43731 and HL81138 (to D.A.T.) and the Barnes-Jewish Hospital Foundation.

Disclosure Summary: D.A.T. serves as a paid consultant to the Department of Medicine, University of California, Los Angeles Geffen School of Medicine and to the Institute of Medicine/Committee to Review Dietary Reference Intakes for Calcium. O.L.S. has nothing to disclose.

Abbreviations:

 
• AD,

  Activation domain;

 
• ASO,

  antisense oligodeoxynucleotide phosphorothioate;

 
• CMV,

  cytomegalovirus;

 
• CK2,

  casein kinase-2;

 
• DMAT,

  2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole;

 
• Dyrk,

  dual specificity tyrosine phosphorylation regulated kinase;

 
• EGFP,

  enhanced green fluorescent protein;

 
• EYFP,

  enhanced yellow fluorescence protein;

 
• FGF,

  fibroblast growth factor;

 
• HA,

  hemagglutinin;

 
• HIPK,

  homeodomain-interacting protein kinase;

 
• LUC,

  luciferase reporter gene;

 
• MID,

  Msx2-interacting domain;

 
• MINT,

  Msx2-interacting nuclear target;

 
• Msx2,

  muscle segment homeobox homolog 2;

 
• NF,

  nuclear factor;

 
• NFAT,

  NF of activated T cells;

 
• NMTS,

  nuclear matrix targeting signal;

 
• NLS,

  nuclear localization sequence;

 
• OC,

  osteocalcin;

 
• OCFRE,

  OC FGF response element;

 
• OPN,

  osteopontin;

 
• PDPK,

  proline-directed protein kinase;

 
• qPCR,

  quantitative PCR;

 
• RAM7,

  RBP-J binding domain;

 
• RANKL,

  receptor activator of NF-κB ligand;

 
• RBP-J,

  recombination signal-binding protein suppressor of hairless;

 
• RID,

  receptor-interacting domain;

 
• RRM,

  RNA recognition motif;

 
• RSV,

  Rous sarcoma virus;

 
• Runx2,

  Runt-related transcription factor 2;

 
• siRNA,

  small interfering RNA;

 
• SPOC,

  Spen paralog and ortholog C-terminal;

 
• TAF,

  trans-activation function;

 
• USF,

  upstream stimulatory factor.

References