Pim-1 kinase phosphorylates RUNX family transcription factors and enhances their activity

Background The pim family genes encode oncogenic serine/threonine kinases which in hematopoietic cells have been implicated in cytokine-dependent signaling as well as in lymphomagenesis, especially in cooperation with other oncogenes such as myc, bcl-2 or Runx family genes. The Runx genes encode α-subunits of heterodimeric transcription factors which regulate cell proliferation and differentiation in various tissues during development and which can become leukemogenic upon aberrant expression. Results Here we have identified novel protein-protein interactions between the Pim-1 kinase and the RUNX family transcription factors. Using the yeast two-hybrid system, we were able to show that the C-terminal part of human RUNX3 associates with Pim-1. This result was confirmed in cell culture, where full-length murine Runx1 and Runx3 both coprecipitated and colocalized with Pim-1. Furthermore, catalytically active Pim-1 kinase was able to phosphorylate Runx1 and Runx3 proteins and enhance the transactivation activity of Runx1 in a dose-dependent fashion. Conclusion Altogether, our results suggest that mammalian RUNX family transcription factors are novel binding partners and substrates for the Pim-1 kinase, which may be able to regulate their activities during normal hematopoiesis as well as in leukemogenesis.


Background
The pim-1 proto-oncogene was first identified as a common proviral insertion site associated with murine leukemiavirus-induced lymphomagenesis, and its oncogenic activity was verified with transgenic mice overexpressing pim-1 in the lymphoid compartment [1]. These mice show a low incidence of spontaneous T-cell lymphomas, the development of which can be accelerated by activation of cooperating oncogenes, such as myc family genes, bcl-2 or Runx2 [1][2][3]. Two additional, functionally redundant pim family members have been identified with partially overlapping expression patterns. The murine pim-1 gene encodes 44 and 34 kD isoforms of a serine/threonine-specific kinase [4], whose expression in hematopoietic cells can be induced by a variety of cytokines, such as interleukins 2, 3, 6 and interferon-α [5][6][7]. We and others have shown that Pim-1 is involved in cytokine-dependent signaling via its ability to regulate activities of the NFATc [8] and c-Myb [9] transcription factors, the Epstein-Barr virus nuclear antigen-2 [10] and the SOCS family suppressors of cytokine signaling [11,12]. Pim kinases also enhance hematopoietic cell survival and participate in regulation of the cell cycle [13].
RUNX family proteins (also known as AML, PEBP2α or CBFα) [14] are DNA-binding α-subunits of heterodimeric transcription factors that are essential for both cell proliferation and differentiation during development [15]. Homozygous disruption of murine Runx2 results in complete lack of bone formation, Runx1 knockout mice are embryonally lethal due to failure of definitive hematopoiesis, and Runx3-deficient mice display abnormal development of gastric epithelium and dorsal root ganglion as well as defects in thymopoiesis. In addition, strict spatiotemporal expression of all Runx family genes is critical for normal hematopoiesis [16]. The RUNX proteins contain an evolutionary conserved region, the Runt domain, which has been named after their structural homologue in Drosophila [17]. This region is required for DNA-binding as well as for dimerization with the β-subunit. While three mammalian genes encode α-subunits: RUNX1 (PEBP2αB), RUNX2 (PEBP2αA) and RUNX3 (PEBP2αC), only one gene has been identified for the β-subunit (PEBP2β/CBFβ). The β-subunit can enhance DNA-binding by the Runt domain but does not contact DNA itself [15,18]. There is less sequence similarity between RUNX family members outside the Runt domain, except for the highly conserved five amino acid C-terminus (VWRPY) known to bind transcriptional repressors, but the C-terminal regions are rich in proline, threonine and serine (PTS) and contain domains involved in transcriptional activation or inhibition [19]. RUNX activity has recently been shown to be regulated by several extracellular signaling pathways resulting in post-translational modifications, such as phosphorylation, acetylation and ubiquitination. [20].
The involvement of RUNX genes in cancer was first discovered as chromosomal translocations associated with acute myeloid leukemia [21]. These translocations had resulted in fusion proteins lacking the C-terminal transactivation domains of RUNX1. Evidence for Runx1 function as a tumor suppressor gene was obtained from knock-in mice where a single Runx1-eto fusion allele caused a similar phenotype as observed for the Runx1 null mice [22,23]. Human RUNX3 has also been shown to act as a tumor suppressor in gastric carcinomas [24]. However, recent retroviral tagging studies have indicated that any of the three murine Runx genes can also operate as dominant oncogenes that can co-operate with myc and pim family genes in lymphomagenesis [3,25]. Human RUNX genes have also been observed to be amplified in childhood leukemias [26,27].
Here we show that the Pim-1 kinase can physically interact with RUNX family transcription factors, colocalize with them within nuclei and phosphorylate them in vitro. Furthermore, the transactivation ability of Runx1 is potentiated by Pim-1, suggesting a mechanism via which Pim-1 may regulate the activity of RUNX family transcription factors during hematopoiesis as well as in leukemogenesis.

Pim-1 interacts with RUNX family proteins
To search for putative Pim-1-interacting partners, we used the yeast two-hybrid system as previously described [9,28]. A kinase-deficient K67M mutant of Pim-1 fused to the LexA DNA-binding domain was used as a bait to screen a library of cDNA clones that had been isolated from Epstein-Barr virus-transformed human lymphocytes and fused to the VP16 activation domain. Out of the approximately 6 × 10 6 yeast transformants tested, 220 clones were recovered that were able to activate two separate reporter genes in a strictly Pim-1-dependent fashion.
Sequence analysis revealed that one of the strongly interacting cDNA clones, designated B19, encoded amino acids 264-404 of the human RUNX3 protein, including its C-terminal transactivation and inhibition domains, but lacking the most C-terminal end (see Figure 3B). To test whether this fragment interacted also with the wildtype Pim-1 protein, a mating assay was carried out using a modified two-hybrid assay [29] with baits fused to the GAL4 DNA-binding domain. Results from this assay indicated that the VP16-B19 fusion protein interacts equally well with both the wild-type and mutant GAL4-Pim-1 fusion proteins, but not with any control proteins tested, such as lamin or the Src kinase, as judged by the ability of the diploid yeast strains to grow on selective medium lacking histidine ( Figure 1A and data not shown). The bait proteins did not activate the reporter genes on their own in the absence of the B19 fragment, while all strains were able to grow on non-selective medium containing histidine ( Figure 1A and data not shown). The physical interaction observed in vivo in yeast cells was also biochemically confirmed in a GST pull-down assay, where bacterially produced GST-B19 fusion protein was specifically able to associate with in vitro translated 35 S-labeled Pim-1 protein (data not shown).
To further examine the interaction between Pim-1 and full-length RUNX transcription factors within living cells, COS-7 cells were transiently transfected with vectors expressing Pim-1 and either MYC-tagged Runx1, Runx3 or FLAG-tagged Runx1. Two days later, cells were collected and lysed, after which the cell lysates were subjected to immunoprecipitation with anti-MYC or FLAG antibodies followed by Western blotting with anti-Pim-1 antibody. This analysis revealed that Pim-1 can be coprecipitated together with both Runx1 and Runx3 full-length proteins ( Figure 1B and data not shown).

Pim-1 colocalizes with Runx1 and Runx3 proteins in nucleus
To be able to investigate the intracellular distribution and possible colocalization of Pim-1 and RUNX proteins, COS-7 cells were transiently transfected with vectors expressing Pim-1 fused to the enhanced cyan fluorescent protein (ECFP) and either Runx1 or Runx3 fused to the enhanced yellow fluorescent protein (EYFP). As expected from previous studies with untagged proteins [30][31][32], Pim-1 protein was found both in the nucleus and the cytoplasm of interphase cells, while Runx1 and Runx3 proteins predominantly localized to the nuclei in a granular expression pattern (Figure 2). More intriguingly, both Runx1 and Runx3 colocalized with Pim-1 in the nuclei of double-positive cells, as demonstrated by scattergram analysis of merged fluorescent images from fixed as well as from living cells (Figures 2B, 2C and data not shown). Altogether, our immunoprecipitation and imaging results indicated that Pim-1 can colocalize and physically interact with Runx family proteins within the nuclear compartment.

Pim-1 can phosphorylate Runx1 and Runx3 proteins in vitro
To find out whether human or murine RUNX proteins act as substrates for the Pim-1 kinase, in vitro kinase assays were carried out with bacterially expressed proteins fused to the glutathione S-transferase (GST) protein. Wild-type GST-Pim-1, but not the corresponding kinase-deficient K67M mutant was able to phosphorylate itself, the C-terminal interacting fragment of human RUNX3 as well as the full-length murine Runx3 protein, but not the GST moiety ( Figure 3A and data not shown). Pim-1 phosphorylated also several C-terminal fragments of murine Runx1 ( Figure 3A). Since not all the Runx1 and Runx3 fragments overlapped with each other ( Figure 3B), this suggests that there are multiple target sites for Pim-1 within the RUNX proteins.

Pim-1 potentiates transcriptional activity of the RUNX1 transcription factor
Phosphorylation by extracellular signal-regulated kinase (ERK) has previously been shown to potentiate the transactivation ability of RUNX1 [33]. To investigate whether phosphorylation of Runx1 by Pim-1 had similar consequences, transient transactivation experiments were carried out in Jurkat T cells using a previously established luciferase reporter assay with which the functional domains of Runx1 had been determined [19]. There the luciferase gene is driven by the macrophage-colony stimulating factor receptor (M-CSF-R) promoter containing binding sites for RUNX, PU.1 and C/EBP transcription factors. In Jurkat T cells, the reporter is inactive in the absence of ectopic expression of any RUNX family member and their heterodimeric binding partner CBFβ, while Runx1 alone only slightly activates it. Also in our assays, ectopic expression of Pim-1 was unable to stimulate luciferase activity in the absence of Runx1 (data not shown). As shown in Figure 4A, increasing amounts of wild-type Pim-1 were able to enhance Runx1/Cbfβ-dependent transactivation of the luciferase reporter in a statistically significant and dose-dependent fashion, as also confirmed by analysis of the steady-state levels of Pim-1 protein by Western blotting. By contrast, the kinase-deficient K67M mutant of Pim-1 did not have any major effects on Runx1 activity, while the more extensive NT81 mutant lacking the N-terminal 80 amino acids of Pim-1 even slightly inhibited it ( Figure 4B). We have previously shown that this mutant can act in a dominant negative fashion to downregulate the effects of the endogenously expressed wild-type Pim-1 protein e.g. on NFATc activity [8].
To examine whether the effects of Pim-1 were mediated via the activation domain of Runx1 that Pim-1 was able to phosphorylate, additional assays were carried out with a GAL4-dependent luciferase reporter coexpressed with a fusion protein where the yeast GAL4 DNA-binding domain had been fused with the activation domain of Runx1 (amino acids 262-371) containing two major transactivation elements, TE1 and TE2 [ [19], see Figure  3B]. Indeed, wild-type Pim-1 was able to increase luciferase activity when coexpressed with the GAL4-Runx1 fusion protein, but not with the GAL4 DNA-binding domain alone ( Figure 4C). Since the kinase-deficient mutants remained inactive in this assay (data not shown), our results suggest that the effects of Pim-1 are dependent on the presence of its phosphorylation target sites within the activation domain of RUNX proteins.
Ser 249 and Ser 266 of RUNX1 have been shown to be targeted by the extracellular signal-regulated kinase (ERK) [33]. More recent studies have indicated that phosphorylation by ERK affects not only activity, but also localization and stability of RUNX1 [34]. Unphosphorylated RUNX1 interacts with the transcriptional repressor mSin3A and is associated with nuclear matrix. Phosphorylation of RUNX1 by the ERK-dependent pathway releases RUNX1 from mSin3A and nuclear matrix, and this is accompanied with enhanced transcriptional activity. However, since binding to mSin3A protects RUNX1 from proteosome-mediated degradation, corepressor release from RUNX1 may regulate its transcriptional activity in a time-dependent fashion, and thereby prevent prolonged RUNX1 activation in response to cytokines or growth factors. Since none of the amino acid sequences surrounding ERK-phosphorylated or other C-terminal serine or threo-Pim-1 colocalizes with Runx-1 and Runx-3 nine residues in RUNX1 show obvious homology to the reported Lys/Arg-rich Pim-1 consensus phosphorylation site [35], the Pim-1 target sites as well as the putative effects of the Pim-1 kinase on stability of RUNX proteins remain to be identified.

Pim-1 may regulate hematopoietic cell fate together with RUNX proteins
Enforced expression of Runx2 and gfi-1 transcription factors in murine thymocytes has been shown to result in delayed thymocyte development at the stage of β-selection where cells rearrange their T cell receptor β (TCRβ) locus [36,37]. Interestingly, Pim-1 is able to promote maturation of double negative (DN) thymocytes into double positive (DP) thymocytes in Rag-deficient and TCRβ enhancer-deleted mice, which are deficient in β-selection  Figure 3 Pim-1 phosphorylates Runx proteins in vitro. (A) Bacterially produced GST fusion proteins expressing either fulllength (FL) or fragments of Runx1 or Runx3 were incubated with GST-Pim-1 in in vitro kinase assays. The phosphorylation products were separated on SDS-PAGE and visualized by autoradiography. GST alone (-) was used as a negative control. * indicates protein degradation products. (B) Schematic presentation of the functional domains of Runx1, including the Runt domain, an activation domain (AD) with two major transactivation elements (TE1 and TE2), a minor transactivation element (TE3), an inhibitory domain (ID) and the C-terminal VWRPY sequence. Shown are also the fragments phosphorylated by Pim-1 in Runx1 or Runx3, which lacks the sequences corresponding to TE3 of Runx1.

Pim-1 phosphorylates Runx proteins in vitro
as are also mice overexpressing Gfi-1 [37][38][39]. In addition, intact Runx1 protein is required for cell proliferation during DN-to-DP transition [40]. Thus, strict spatio-temporal expression of all these proteins is important for development of DN thymocytes into more mature DP T cells and further into functional mature single positive effector T cells.
Both pim and Runx family genes can cooperate with myc family genes in tumor formation [1,25], which correlates well with the observation that both Runx1 and Runx3 were found among the genes that could substitute for pim-1 and pim-2 in retroviral tagging experiments [41]. However, pim-1 and Runx2 have also been shown to cooperate with each other [3], suggesting that these genes are not completely redundant in their oncogenic effects and that RUNX family transcription factors may function both in parallel as well as downstream of Pim kinase-modulated pathways.

Conclusion
Our data indicate that the Pim-1 serine/threonine kinase is able to physically interact with the RUNX family transcription factors, colocalize with them within nuclei and phosphorylate them in vitro. Moreover, the transcriptional activity of at least Runx1, but most likely also of other RUNX family members is potentiated by Pim-1. These results have revealed a previously unrecognized signaling cascade involving Pim-1 kinase and the RUNX family of transcription factors that may control differentiation and transformation of hematopoietic cells.

Yeast two-hybrid assays
Yeast two-hybrid assays were carried out essentially as previously described [9,28]. Briefly, the K67M mutant of Pim-1 fused with the LexA DNA-binding domain was used as a bait to screen a library kindly provided by Stephen Elledge (Baylor College of Medicine, Houston, Texas). The library contained cDNA clones isolated from Epstein-Barr virus-transformed human peripheral blood lymphocytes and fused to the VP16 activation domain. The yeast transformants expressing Pim-1-interacting protein fragments were double-selected for their abilities to grow on histidine-deficient plates containing 25 mM 3aminotriazole and to produce β-galactosidase. To further verify double-positive interactions, mating assays were carried out using a modified yeast two-hybrid assay [29] with baits fused with the GAL4 DNA-binding domain. Nucleotide sequences for the positive clones were determined using an Applied Biosystems automated sequencing apparatus.
Cell imaging assays COS-7 cells were transiently transfected with ECFP or EYFP fusion vectors as described above and plated on coverslips. Two days later, cells were fixed with 4% paraformaldehyde, after which confocal images were captured with Zeiss LSM510 META confocal microscope. ECFP and EYFP fusion proteins were excited with 405 nm and 514 nm laser lines and emissions were collected with BP 435-485 and LP 560 filters, respectively. The optical thicknesses of the two channels were equalized prior to image acquisition and colocalization was visualized with a scattergram plot acquired with Zeiss LSM510 3.2 program.

Authors' contributions
TLTA carried out protein interaction assays, transactivation assays and most of in vitro kinase assays, and drafted the manuscript. JS performed imaging assays and KJP yeast mating assays. YI provided reagents and expertise on Runx proteins and critically revised the manuscript. PJK initiated the project by carrying out the yeast two-hybrid screen and some in vitro kinase assays, and supervised the work as well as writing of the manuscript. All authors read and approved the final manuscript.