RUNX family: Regulation and diversification of roles through interacting proteins


  • Linda Shyue Huey Chuang,

    1. Cancer Science Institute of Singapore, National University of Singapore, Singapore
    Search for more papers by this author
  • Kosei Ito,

    1. Nagasaki University, Graduate School of Biomedical Sciences, Nagasaki, Japan
    Search for more papers by this author
  • Yoshiaki Ito

    Corresponding author
    1. Cancer Science Institute of Singapore, National University of Singapore, Singapore
    2. Institute of Molecular and Cell Biology, Proteos, Singapore
    • Cancer Science Institute of Singapore, National University of Singapore, 14 Medical Drive #12-01, Singapore 117599, Singapore
    Search for more papers by this author
    • Tel.: +65-6516-2242, Fax: +65-6873-9664


The Runt-related transcription factors (RUNX) belong to an ancient family of metazoan genes involved in developmental processes. Through multiple protein-interacting partners, RUNX proteins have been implicated in diverse signaling pathways and cellular processes. The frequent inactivation of RUNX genes in cancer indicates crucial roles for RUNX in tumor suppression. This review discusses the abilities of RUNX proteins, in particular RUNX3, to integrate oncogenic signals or environmental cues and to initiate appropriate tumor suppressive responses.

RUNX proteins are crucial transcription factors that regulate a wide range of biological processes to orchestrate proper cell fate determination. Traditionally described as a DNA-binding protein that binds CBFβ (also known as PEBP2β) to form the heterodimeric core-binding factor (CBF) or polyomavirus enhancer-binding protein 2 (PEBP2) transcription complex, RUNX is now viewed as a multifaceted protein that associates with diverse proteins to direct biological outcomes in a context-dependent manner.1 This review discusses how combinatorial interactions—dynamically regulated in a spatiotemporal manner—have endowed RUNX proteins with a plethora of functions, some seemingly opposite. Because of space constraints, emphasis will be placed on the tumor suppressive properties of RUNX3.

RUNX proteins have been identified in many metazoans: most, if not all, play requisite roles in developmental processes. While primitive metazoans such as sea urchin and C. elegans appear to possess a single RUNX family member, multiple RUNX proteins have been identified in mammals, Drosophila and Fugu,2 indicating (i) evolutionary conserved roles in metazoans and (ii) the necessity for elaborate and tight regulation of RUNX activity when specifying complex developmental events in higher organisms. In mammals, there are three RUNX family members: RUNX1, RUNX2 and RUNX3. Mouse models have shown that disruption of individual Runx genes resulted in distinct phenotypes, indicating nonredundant, tissue-specific roles: Runx1 knockout is associated with lack of definitive hematopoiesis; Runx2 knockout resulted in the absence of bone formation and Runx3 knockout led to mice with defects in cytotoxic T-cell development, as well as gastrointestinal (GI) and neural disorders.3–5 During T-lymphocyte development, Runx3 involvement in the establishment of epigenetic silencing of CD4 is critical for lineage specification and homeostasis, and the absence of Runx3 is associated with defective cytotoxic T cells6; in the intestine, heterozygous inactivation of Runx3 induced colon adenoma; in the gastric epithelia, Runx3 deficiency is associated with a precancerous state, distinctly characterized by loss of chief cells7 and in the lung, Runx3 plays requisite roles during bronchiolar epithelial cell differentiation and tumor suppression.8, 9 All offer compelling evidence for a causal link between loss of Runx3, deregulated differentiation and preneoplasia. Together with the fact that Runx3-deficient mice are tumor prone, with spontaneous tumor development in the intestine, lung and breast (Fig. 1),10 it is reasonable to suggest a tumor suppressor role for RUNX3 during the early stages of solid tumor formation.

Figure 1.

Tumor formation in Runx3+/− mice with BALB/c background at 65 weeks. Tissues were fixed with 4% paraformadehyde, sectioned and stained by hematoxylin and eosin (H&E). (a) A small intestinal adenomatous polyp stained with methylene blue; (b) histological analysis (H&E staining) of adenomatous polyp formed in small intestine; (c) the boxed region in (b) is enlarged [images (b) and (c) are reprinted from Ref.3, with permission from ©Cell Press]; (d) whole lung (arrow indicates tumor growth); (e) histological analysis (H&E staining) of tumor in (d); (f) tumor formation in mouse breast (arrows indicate tumors) and (g) histological analysis (H&E staining) of tumor in (f).

RUNX in cancer

Mutations of RUNX genes have been reported in human diseases. Genetic aberrations of RUNX1 include somatic point mutations in myelodysplasia and chromosomal translocations in acute myeloid leukemia (AML) and childhood B-cell acute lymphocytic leukemia.11–13 Moreover, familial platelet disorder (FPD) with propensity to myeloid malignancy (FPD/AML) is attributed to inherited germline RUNX1 mutations.14, 15 RUNX1 was also reported to serve as tumor suppressor in TLX1- and TLX3-mediated T-ALL.16 The frequent monoallelic RUNX1 mutations firmly implicate RUNX1 haploinsufficiency in hematopoietic disorders. Recurrent mutations of RUNX1 and CBFB were also discovered in breast cancer, suggesting that aberrations in RUNX-mediated epithelial cell differentiation may drive breast cancer development.17, 18 Furthermore, as CBFB is closely linked to the essential functions of all three RUNX proteins, loss-of-function mutations in CBFB would, arguably, inactivate all RUNX proteins, and the consequence of which reflected in the ensuing cancer phenotype. With the recent finding of recurrent RUNX1 mutations in GI adenocarcinomas of the esophagus,19 it is likely that RUNX1 also plays important tumor suppressive roles in solid tumors. Monoallelic RUNX2 mutations are associated with cleidocranial dysplasia, an autosomal-dominant disorder characterized by skeletal anomalies. RUNX3 inactivation in cancer cells has been found to occur mainly through aberrations in DNA methylation, histone modification,20–22 hemizygous deletion or cytoplasmic sequestration. Compared to RUNX1 and RUNX2, point mutations in the RUNX3 gene are less frequently identified. Nevertheless, inactivating mutations of RUNX3 have been observed in patients with cancer (Table 1); in particular, the R122C mutation isolated from a gastric cancer patient exhibited strong oncogenic properties.4 Moreover, RUNX3 is located at chromosomal region 1p36, which is frequently deleted in various cancer types.24–26 Importantly, multiple studies have shown that hypermethylation and subsequent silencing of RUNX3 gene expression (up to 80%) is prevalent in solid tumors of breast, colon, lung, bladder and gastric origins,4, 27, 28 indicating that aberrant methylation of RUNX3 may be useful for cancer detection or prognosis.

Table 1. Examples of RUNX3 mutations in human diseases
inline image

It is becoming increasingly evident that cancer may arise from both genetic as well as epigenetic aberrations. Silencing of tumor suppressor genes by hypermethylation is a common occurrence in cancer cells (e.g., HIC1, APC, Tp53, Tp73, PTEN, and BRCA1). Aberrant methylation patterns are believed to be modulated by environmental and dietary factors (e.g., low folate intake, alcohol consumption and smoking),29 and there is now strong evidence to link cancer risk factors and hypermethylation of RUNX3. Oxidative damage is correlated with elevated occupancy of DNA methyltransferase DNMT1 at the RUNX3 promoter, hypermethylation and reduced RUNX3 expression30; RUNX3 methylation increases with age and smoking in bladder cancer and duration of H. pylori infection in gastric cancer.31, 32 It is likely that heritable loss of RUNX3 expression, possibly due to lifestyle choices (e.g., smoking), resulted in reduced cellular capacity for tumor suppression and is a causative rather than correlative factor in cancer.

Perhaps the scarcity of RUNX3 mutations in cancer tissues diverts attention from the fact that RUNX3 inactivation is a very frequent occurrence in cancer. It is important to note that frequency of RUNX3 inactivation is comparable to that of p53 in various tumors. There is considerable variation in the frequencies of p53 mutation among tumor types (e.g., 10% in hematopoietic malignancies and 50–70% in colorectal, ovarian and head and neck cancers).33 p53 is often inactivated by a single monoallelic missense mutation at its DNA-binding domain, resulting in a stable dominant-negative mutant p53 protein.33 In contrast, somatic mutations of RUNX3 are rare: RUNX3 inactivation in cancer cells are more commonly elicited by aberrations in DNA methylation, histone modification, hemizygous deletion and cytoplasmic sequestration. Indeed, inactivation of RUNX3 by hypermethylation or cytoplasmic sequestration was observed in about 80% of gastric cancer tissues.34 Importantly, RUNX3 hypermethylation is more common in bladder tumors of higher grade31, 35 while RUNX3 levels were reduced in 93% of preneoplastic lung lesions,9 suggesting that dysfunctional RUNX3 constitutes an early event in bladder cancer and lung adenocarcinoma development, and detection of which may be used for timely treatment.

The functional domains and protein-interacting domains of RUNX protein

The mammalian RUNX protein contains several conserved regions (Fig. 2). At the N-terminus is the highly conserved Runt domain, which is shared by all members of the RUNX family. The Runt domain is necessary for binding to CBFβ as well as DNA containing the consensus sequence PyGPyGGTPy.46 The Runt domain of all three human RUNX members exhibit close to 90% homology, indicating that similar mechanisms underlie Runt domain functionality. The C-terminus comprises the activation domain, the inhibitory domain (ID), PPxY (or PY) motif, the VWRPY motif and nuclear matrix-targeting signal (NMTS). The PY motif is a proline-rich peptide that interacts with proteins harboring the WW domain, whereas the VWRPY motif interacts with WD domain of Groucho/TLE transcription corepressors. The multiple protein-interacting sites in RUNX suggest that different outcomes can be generated through combinatorial protein binding. Moreover, comparison of RUNX family members revealed relatively less homology downstream of the Runt domain, and this may account for functional differences between RUNX proteins. As evident from knockout models and human diseases, members of RUNX family, despite close similarities in the Runt domain, cannot completely compensate for other members in the event of inappropriate inactivation.

Figure 2.

Functional domains of RUNX proteins and examples of their interacting proteins. The human RUNX3, RUNX1 and RUNX2 proteins refer to sequences with accession numbers NP_004341, NP_001001890 and NP_004339, respectively. Some RUNX interactants were identified using mouse Runx proteins, which are highly homologous to the human counterparts; some interactions were reported with other RUNX isoforms, the domain of which showed close homology to the isoform used here. The numbers indicate amino acid positions. AD, ID, C-term and NMTS refer to activation domain, inhibitory domain, C-terminus and nuclear matrix-targeting sequence, respectively. RUNX1 interactants not shown here include CEBF/α, ELF2, ELF4, HIPK1, HIPK2, SMRT,36 TAZ,37 MyoD,38 SWI-SNF, CHIP,39 TAL1 and TRAF640; RUNX2 interactants not shown here include androgen receptor,41 cyclin B1, MSX2,42 CHIP,43 CoAA,44 UBF1, SL1 and various SMADs and RUNX3 interactants not shown here include RBPJK, Notch145 and various HDACs.

The biological activities of RUNX proteins are arguably the sum of its interactions with other proteins. As described below, many of these protein partners have been implicated in tumorigenesis. Studying the molecular interactions of RUNX is likely to yield insights on the pleiotropic effects of RUNX in tumor suppression and may even lead to novel therapeutic interventions for patients with cancer.

Involvement of RUNX in signaling pathways implicated in tumorigenesis

TGF-β signaling

The TGF-β pathway plays prominent roles in metazoan development. It maintains tissue homeostasis by regulating numerous processes including proliferation, differentiation, apoptosis and adhesion.47 Although it can exert strong tumor suppressive effects, TGF-β signaling is, paradoxically, responsible for highly malignant tumor processes such as epithelial–mesenchymal transition (EMT), evasion from immune surveillance and cell invasion. Inappropriate TGF-β signaling is therefore a critical factor in malignant progression. Multiple lines of evidence support cooperation between RUNX and the two major branches of TGF-β superfamily, namely, TGF-β and bone morphogenetic proteins (BMPs); all RUNX members interact with the key effectors of the TGF-β pathway—the SMAD transcription factors.48, 49 Interaction of RUNX3 with TGF-β receptor-regulated SMADs (e.g., SMAD2 and SMAD3) synergistically induced transcription from the immunoglobulin germline Cα promoter.50, 51 RUNX2 interacts with BMP-specific SMADs (e.g., SMAD1 and SMAD5) to perform critical roles in bone formation. Conversely, the interaction of SMAD3 with RUNX2 results in the repression of RUNX2 transactivation ability on osteoblast-specific genes, reflecting the inhibitory action of TGF-β in osteoblast differentiation.52 Differential cooperation of RUNX members with TGF-β, therefore, enables cell context variations of RUNX activity.

Importantly, the ability of RUNX3 to augment the TGF-β pathway is essential for tumor suppression of gastric cancer4: Runx3 knockout mice develop gastric hyperplasia due, in part, to defective TGF-β-mediated apoptosis; RUNX3 upregulates cell cycle inhibitor p21WAF1, proapoptotic gene Bim and tight junction gene Claudin-1 through the TGF-β pathway in gastric cells.53–55 Furthermore, ablation of Runx3 in p53−/−-immortalized gastric epithelial cells resulted in spontaneous EMT, generating mesenchymal-like cells with stem-like and tumorigenic properties.56 As the loss of Runx3 is associated with increased cellular susceptibility to TGF-β-induced EMT rather than apoptosis, it is likely that RUNX3 protects against EMT by influencing the output of TGF-β response during malignant progression. Taken together, RUNX3 appears to possess the ability to block tumorigenesis at distinct stages, namely, initiation, progression and metastasis. On the other hand, RUNX2 was reported to promote EMT through TGF-β-dependent induction of SNAI2,57 suggesting complex interplay of RUNX proteins with the TGF-β pathway.

Wnt signaling

Abnormal Wnt signaling has been implicated in cancer, maintenance of stemness, osteoporosis and aging.58 Wnt-mediated transcription requires the stabilization of β-catenin and its interaction with T-cell factor/lymphoid enhancer factor (TCF/LEF) family of DNA-binding transcription factors. LEF1 was reported to bind RUNX2 via the Runt domain, resulting in the repression of RUNX2 transactivation of the osteocalcin promoter.59 This repression, which is enhanced by β-catenin, indicates a role for Wnt in regulating RUNX2 activity during osteoblast development. Moreover, as the Runt domain is highly conserved among RUNX members, it is likely that LEF1 interacts with RUNX1 and RUNX3 as well and that additional studies need to be done about the possible interaction between RUNX1/RUNX3 and LEF1.

A clear role for RUNX3 in the suppression of oncogenic Wnt signaling has been recently reported.3 RUNX3 forms a ternary complex with key Wnt effectors TCF4-β-catenin, resulting in reduced affinity of TCF4-β-catenin for DNA and attenuation of Wnt signaling. Runx3 knockout mice exhibit increased Wnt signaling and oncogenicity in intestinal epithelial cells, as indicated by intestinal adenoma formation in Runx3−/− mice. Therefore, RUNX3 performs a parallel, but independent, role to that of APC in suppressing colorectal tumorigenesis. Enhanced Wnt activity was also observed in the gastric epithelia of Runx3−/− mice. Increased levels of Wnt target Cdx2 correlated with induction of an intestinal phenotype in the stomach.7 Intestinal metaplasia (IM) is a histological feature in human gastric cancer. Although the formation of goblet cells is a hallmark of IM in humans, goblet cells do not normally develop in mouse stomach even when Cdx2 is expressed. Intestinalization of mouse stomach epithelium by increased levels of Cdx2, therefore, recapitulates part of the gastric subregion observed in human gastric cancer. Additionally, a new type of metaplasia, called spasmolytic polypeptide-expressing metaplasia (SPEM), is observed in both premalignant mouse and human gastric epithelium.60 SPEM, generally induced when acid-producing parietal cells lose their function, is characterized by the loss of chief cells and parietal cells. In Runx3−/− epithelium, SPEM is induced even though parietal cell function appears normal. It may be that Runx3 regulates pathways downstream of parietal cell loss to induce SPEM.7 This robust ability to attenuate oncogenic Wnt signaling strongly indicates a gatekeeper role—via balancing differentiation and proliferation—for RUNX3 in the early stages of GI cancer development.

Hippo/MST2 pathway

The Hippo/MST2 signaling pathway is a major regulator of tissue growth and organ size and has been linked to carcinogenesis. Hippo pathway component YAP1 (Yes-associated protein) interacts with the PY motif of RUNX proteins via its WW domain.61 RUNX proteins recruit YAP1 to RUNX target promoters (e.g., Osteocalcin and Itch) to enhance transcription.61, 62 Interestingly, during deleterious DNA damage, tyrosine-phosphorylated YAP1 discriminated against RUNX proteins, showing instead stronger affinity for p73.63 This led to preferential recruitment of YAP1 to p73-target genes over RUNX-target genes, resulting in selective coactivation of p73-target genes. RUNX–YAP collaboration is, therefore, modulated by the cellular environment. Recently, RUNX3 was shown to interact with two other components of the Hippo/MST2 pathway, scaffold protein SAV1 and tumor suppressor LATS2 kinase.64 Unlike YAP–RUNX interaction, the RUNX3–SAV1 interaction requires the Runt domain and is enhanced following phosphorylation of RUNX3 by MST2. Ablation of RUNX3 abolishes MST2-mediated cell death,64 suggesting a central role for RUNX3 in this aspect of the MST2 pathway. As SAV1 and MST2 are involved in centrosome disjunction65 and LATS2 localizes to the centrosome to regulate γ-tubulin accumulation and mitotic spindle formation,66 it is tempting to speculate that RUNX3 collaborates with Hippo/MST2 components at the centrosome.67 Moreover, the sharing of RUNX proteins by Hippo, Wnt and TGF-β signaling pathways suggests a role for RUNX in integrating signals and providing crosstalk between these pathways for development and tissue homeostasis.

RUNX in cell cycle progression

RUNX protein levels, due to changes in transcription and protein stability, oscillate throughout the cell cycle.68 RUNX1 level increases during G1 to S transition and remain elevated at G2/M phase,69 whereas RUNX2 protein levels are high during early G1 phase and low from early S phase to mitosis in osteoblastic cells.70 This suggests that variations in RUNX activity influence cell cycle progression. Indeed, RUNX1 stimulates G1 to S progression in hematopoietic cells,71–73 in part through induction of cyclin D3 transcription. In contrast, overexpression of RUNX2 in MC3T3-E1 osteoblastic cells delays progression from G1 to S phase.70 Furthermore, RUNX1 and RUNX2 have been reported binding to Cdk6 (at Intron 2) and cyclin B2 (at the promoter), respectively, suggesting that RUNX regulates transcription of cell cycle regulators and thus the cell cycle at multiple stages.74–76 It is important to note that the effects of RUNX on the cell cycle are highly context dependent and dose sensitive. For example, in hematopoietic stem and progenitor cells, the relative expression of RUNX1 splice isoforms AML1a and AML1b affects cell fate determination: enforced AML1b expression promoted differentiation, whereas elevated AML1a is associated with self-renewal. The antagonistic effects of RUNX1 isoforms suggest self-regulation through variations in RNA splicing.77

Moreover, RUNX activity is modulated in various ways by cell cycle regulators such as cyclins and cyclin-dependent kinases (Cdks). For example, RUNX proteins are ubiquitinated and degraded following phosphorylation by Cdk–cyclin complexes such as Cdk4/cyclin D1, Cdk2/cyclin A and Cdk1/cyclin B.78–80 Moreover, the properties of RUNX proteins are altered by phosphorylation: Cdk1/cyclin B can increase RUNX2 affinity for DNA during mitosis81, 82 or stimulate RUNX1 transactivation ability by reducing its affinity for histone deacetylase (HDAC1/3).83 Interestingly, Cdks and cyclins also affect RUNX function by kinase-independent mechanisms. Cdk6 disrupts Runx1 interaction with DNA and C/EBPα84 by binding to the Runt domain; cyclin D3 inhibits RUNX1 activity by physically displacing Cbfβ from the Runt domain and reducing its affinity for DNA85 and cyclin D1 prevents RUNX3 from binding to transcription coactivator p300.86 From a clinical perspective, cyclin D-mediated inhibition of RUNX transactivation ability (and presumably its tumor suppressive properties) is important because of the frequent overexpression of D-type cyclins in cancer.

The retinoblastoma protein (pRb) is a master regulator of cell cycle progression from G1 to S phase. Although best known for promoting cell cycle exit, pRb has also been implicated in cell fate decisions and lineage commitment.87, 88 RUNX2 interacts preferentially with hypophosphorylated pRb via the NMTS.89 The ensuing potentiation of RUNX2 transcriptional activity increased the expression of p27KIP1 and other osteogenic genes, resulting in exit from cell cycle and subsequent induction of terminal osteoblast differentiation.87, 90 This has implications in tumorigenesis because disruption of this process is linked to osteosarcoma, perhaps a consequence of dedifferentiation to multipotent cancer stem cells.

RUNX during stress response

The Ku70/Ku80 heterodimer is a key component of the nonhomologous end-joining pathway (NHEJ). It detects and binds DNA double-strand breaks (DSBs) and recruits other NHEJ components for repair.91 RUNX3 was found to associate with Ku70, which enhances RUNX3-mediated transactivation of p21.92 Similarly, Ku70 interacts with Runx2, leading to synergistic activation of the osteocalcin promoter.93 However, whether the Ku70–RUNX complex plays a role in response to DSB remains unknown. Nevertheless, there are clear indications of RUNX involvement in DNA damage response. Following adriamycin insult, RUNX3 binds to p53 and enhances p53-mediated transcription activation, and RUNX3 can also associate with phosphorylated ATM.94 It is extremely significant that the RUNX homolog in C. elegans RNT-1 is stabilized in the intestines in response to oxidative stress via the p38 mitogen-activated protein kinase (MAP) pathway.95 This suggests an evolutionarily conserved role for RUNX proteins in stress response. Moreover, given the ability of RUNX proteins to induce cell cycle inhibitors (e.g., p21WAF1 and p27KIP1) and proapoptotic genes (e.g., Bax and Bim),53, 54, 96 RUNX proteins are well poised to play a dominant role in sensing and responding to mutagenic insults.

Compelling evidence indicate a major role for RUNX proteins in oncogene-induced senescence (OIS), a well-established mode of tumor suppression that counteracts inappropriate proliferative signals elicited by deregulated oncogenes. Following oncogenic Ras activation, all RUNX proteins cooperate with p19alternative reading frame (ARF)/p53 pathway to induce a senescence-like growth arrest in primary mouse embryonic fibroblasts (MEFs); for example, Runx2−/− MEF cells characterized by enhanced growth advantage and susceptibility to spontaneous immortalization acquired resistance to Ras-induced senescence. These properties are attributed to increased G2/M cyclin–Cdk and p38MAPK activity in Runx2−/− cells. Moreover, it is interesting that the key contributors of Ras-induced senescence include p38MAPK, Raf-MEK-ERK signaling, DNA damage, SWItch/Sucrose NonFermentable (SWI/SNF) complexes and tumor suppressors such as p53 and pRb, many of which have associations with RUNX proteins. Aside from OIS, ectopic expression of RUNX1 in human fibroblasts also induces senescence-like growth arrest, suggesting involvement of RUNX in the induction of different senescence phenotypes.97

Importantly, there is frequent co-occurrence of RAS activation with RUNX1 mutations in human leukemia, and knockout of Runx1 in the mouse bone marrow is associated with increased stem/progenitor cell fraction and impairment of RAS-induced senescence.98 Therefore, RUNX1 is involved in a fail-safe mechanism to prevent malignant transformation in hematopoietic stem/progenitor cells.

Another kind of stress is viral infection. DNA tumor viruses exploit cellular machinery to promote aberrant proliferation and suppress apoptosis—the resultant, a formula for tumorigenesis. Adenovirus oncoprotein E1A binds to RUNX3 and inhibits its transcriptional activity on p21WAF1.99 This suggests that the adenovirus inhibits RUNX3, in addition to tumor suppressors pRb and p53, to create a conducive environment for viral multiplication.

Modulating RUNX activity by post-translational modification

RUNX activity is dynamically regulated by a growing list of post-translational modifications such as phosphorylation, ubiquitination, acetylation and methylation.100 As described below, post-translational modification of RUNX exerts important functional consequence and may explain how cells fine tune RUNX activity to regulate cell cycle progression or respond to external stimuli.


RUNX proteins are substrates of diverse kinases such as Pim-1, ERK and cyclin–Cdks. Proto-oncogene Pim-1 phosphorylates RUNX proteins, leading to increased RUNX1 transactivation ability,101 increased RUNX3 protein stability and its cytoplasmic localization.102 Aberrant expression of Runx2 and Pim-1 led to synergistic T-cell lymphoma development, suggesting functional cooperation between Runx2 and Pim-1.103 RUNX1 is phosphorylated by the serine/threonine kinase ERK2, resulting in increased transactivation ability but reduced protein stability due to dissociation from corepressor Sin3A.104, 105 Owing to the involvement of Pim-1 and ERK2 in cancer, their respective abilities to affect RUNX transcriptional activity have important implications in tumor progression. Moreover, homodomain-interacting protein kinase HIPK2 phosphorylates RUNX1 to promote cooperation between RUNX1, MOZ and p300 for transcriptional activation, suggesting crosstalk between acetyltransferase and kinase in regulating RUNX function.106, 107


All mammalian RUNX proteins have been shown to be acetylated by p300 and which in turn modulate RUNX activity. Acetylation of RUNX1 by p300 promotes its DNA-binding ability108; BMP-2 induces acetylation of RUNX2, resulting in enhanced transactivation properties109 and acetylated RUNX3 shows increased protein stability.110 p300 is frequently mutated in tumors,111 and it remains to be seen if p300 deficiency affects RUNX function during tumor development.


Multiple sites in RUNX1 are methylated by arginine methyltransferase PRMT1. PRMT1-mediated methylation at C-terminus of RUNX1 abrogates its interaction with Sin3A, leading to increased transactivation potential.112 One of the methylation sites, located within the Runt domain and conserved in all three RUNX members, has been found mutated in RUNX1 in breast cancer (Fig. 3).18 In close proximity are two other RUNX1 mutations, also isolated in breast cancer, thus hinting at the importance of this region for RUNX1-mediated tumor suppression. This region is highly enriched with other post-translational modification sites, namely, Pim-1 phosphorylation sites, AKT consensus motif (A. Lim and Y. Ito, unpublished data) and p300 acetylation site. The overlap of these sites with nucleotide binding and AKT1 phosphorylation consensus motifs suggest functional crosstalk between methylation and phosphorylation similar to that reported in FOXO113 to exert tight regulation on ATP/nucleotide binding. Further studies on this region should facilitate understanding of RUNX tumor suppressor activity.

Figure 3.

Highly conserved peptide within RUNX enriched with protein motifs and modification sites. Peptides range from amino acids 141 to 160 in RUNX3 (NP_004341). This region is 100% conserved in RUNX1, RUNX2 and RUNX3. Me, Ac and P indicate methylation, acetylation and phosphorylation. Text in red indicates sites mutated in RUNX1 in breast cancer. Walker consensus motif A (for nucleotide binding) is GXXXXGKS/T; AKT1 phosphorylation motif is RXKXXS/T and T151 (in blue) is a putative AKT phosphorylation site.

Ubiquitin-mediated degradation

RUNX protein levels are controlled by ubiquitin-mediated proteosome degradation. Perhaps reflecting the necessity for rapid fine tuning of RUNX activity in response to different stimuli, multiple E3 ubiquitin ligases such as MDM2, anaphase-promoting complex and SMAD-specific SMURF1 and SMURF2 promote degradation of RUNX proteins, whereas interaction of RUNX with diverse proteins such as CBFβ, SIN3A and mixed-lineage leukemia H3K4 methyltransferase protein prevents proteasomal degradation.110, 114–116 Moreover, RUNX protein stability can also be altered by other protein modifications. In the presence of TGF-β, p300-mediated acetylation of RUNX3 protects against SMURF-mediated degradation,110 whereas phosphorylation of RUNX proteins by various Cdk–cyclins and ERK2 led to RUNX degradation.80 Recently, prolyl isomerase Pin1 was reported to bind phosphorylated RUNX3, leading to its ubiquitination and degradation.117 This has important implications in breast cancer where frequent upregulation of Pin1 may nullify RUNX3 tumor suppressor properties. Moreover, pathogen and gastric cancer risk factor H. pylori expressed oncogenic protein CagA, which interacted with RUNX3 at its PY motif, resulting in its degradation by proteasome.118 It is thus likely that H. pylori infection promotes reduction of RUNX3 expression—be it by degradation or promoter methylation—to suppress apoptosis for effective colonization and that inactivation of RUNX3 may well be one of the major mechanisms of H. pylori-associated gastric cancer.

Augmentation of RUNX transcriptional activity by chromatin-modifying proteins

There is a growing list of chromatin-modifying proteins that interact with RUNX proteins. How RUNX proteins switch between diverging roles such as transcription activation and repression may be, in part, attributed to its interaction with chromatin-modifying proteins.

RUNX and transcription activation

Acetyltransferases MOZ (MYST3) and MORF (MYST4) physically interact with RUNX1 and RUNX2, strongly stimulating RUNX transactivation activity.119, 120 There is, however, no report of MOZ/MORF interaction with RUNX3. Importantly, mutations of MOZ and MORF, which may affect RUNX tumor suppressive ability, have been reported in patients with gastric cancer and leukemia.121 Another acetyltransferase that interacts with RUNX proteins is the coactivator p300,110, 122 which augments transcription activation and protein stability.

Chromatin organization

RUNX has been shown to collaborate with the SWI/SNF chromatin modeling complex for transcription activation. Interaction of RUNX1 with key components of the SWI/SNF complex, such as BRG1 and INI1, and their co-occupancy of RUNX1 target gene promoters correlates with active chromatin, whereas depletion of RUNX1 reduces co-occupancy of BRG1 and INI1, suggesting that RUNX1 recruits the SWI/SNF complex to specific gene targets to regulate hematopoietic functions.123 Interestingly, RUNX2 also engages SWI/SNF, albeit through C/EBPβ, to support osteoblast-specific gene expression.124 This indicates conserved functional interaction between RUNX members and the SWI/SNF complex. The SWI/SNF complex has been heavily implicated in tumorigenesis, and recently, a component ARID1A was described as one of the most frequently mutated genes in gastric and other cancers.121, 125–127 It is becoming increasingly apparent that mutations in epigenetic regulators/machinery and consequent epigenetic aberrations constitute major drivers of cancer progression. Whether inactivation of ARID1A, which is located at chromosomal region 1p36 and in close proximity with RUNX3, affects RUNX function during tumorigenesis would, therefore, be of great interest.

RUNX and transcription repression

RUNX can also associate with histone-modifying enzymes, such as histone acetyltransferase HDAC, to elicit transcription repression. Interestingly, phosphorylation of RUNX1 by Cdk–cyclins resulted in reduced interaction of RUNX1 with HDAC1 or HDAC3 and subsequent stimulation of RUNX1 transactivation activity.83 Cyclin–Cdks, as well as cell cycle phases, therefore influence whether RUNX functions as transcription repressor or activator. Moreover, RUNX1 and RUNX3, unlike RUNX2, bind to histone-lysine N-methyltransferase SUV39H1,128 indicating shared roles in transcription repression, heterochromatin organization and chromosome segregation. Recently, RUNX1 was shown to form a multiprotein complex with CBFβ and polycomb-repressive complex Bmi1 to regulate genes involved in hematopoiesis.129 The fact that Bmi1 interacted with the highly conserved Runt domain strongly suggests a similar interaction with RUNX2 and RUNX3 and would be an informative area of research. The repertoire of proteins that cooperates with RUNX in transcription repression, including SIN3A and Groucho/Transducin-like enhancer of split (TLE) corepressor, is continually increasing, and this reflects the broad range of RUNX activity and the necessity of fine tuning RUNX activity according to cell context.

Binding of RUNX proteins to transcription factors and transcription machinery

RUNX proteins are, by themselves, not strong transcription regulators. They interact with diverse transcription factors, such as ETS, AP-1, GATA-1, SOX9, C/EBPα, STATs, HES-1 and ALY, to exert strong transcriptional activation or repression.84, 130–134 Many of these transcription factors, including PAX5,135 SOX9, STAT5 and AP-1, interact with RUNX via the Runt domain, suggesting that they may compete with each other for binding to RUNX. These different combinations of transcription factors greatly diversify RUNX transcriptional roles.

Interaction of RUNX proteins with transcription factors

Because of space constraints, a few examples are highlighted. RUNX1 and RUNX2 bind to c-fos and Jun, which constitutes the AP-1 transcription factor, through the Runt domain to activate the collagenase-3 promoter.136 RUNX1 interacts with Ets family members such as Ets1 and MEF to promote transactivation.137, 138 Interaction with other ETS family members results in different effects: interaction of RUNX1 with NERF-2 and NERF-1 results in B-cell-specific blk gene activation and repression, respectively.139 The interaction of TAL1 with ETS1 and RUNX1/3 tethers TAL1 to genes involved in T-cell differentiation.76 Interestingly, RUNX3 can destabilize the estrogen receptor-α via proteasome-mediated degradation; this ability is likely to be important in suppression of breast tumorigenesis.10

Transcription factors that inhibit RUNX activity

The transactivation activity of RUNX proteins can be inhibited in multiple ways by transcription factors: for example, the binding of Foxp3 to RUNX1 at its ID results in the suppression of RUNX1 in regulatory T cells140; RUNX1-mediated repression of CD4 transcription is likely due, in part, to interaction with cycT1 at the ID and resultant inhibition of transcription elongation.141 Similarly, RUNX2 can associate with RNA Pol I transcription factors UBF1 and SL1 at the nucleoli to repress ribosomal RNA transcription.142 Interaction of RUNX2 and SOX9 during osteochondroprogenitor cell fate determination resulted in the repression of RUNX2 via decreased DNA-binding ability and increased proteasome-mediated degradation,143 and Gli3 repressor inhibits Runx2 by competitively binding to DNA.144

Subcellular distribution of RUNX proteins

Besides temporal control by post-translational modification, RUNX activity is also modulated by subcellular distribution. Although predominantly nuclear, RUNX proteins have been detected at the cytoplasm, centrosome, mitotic spindle and midbody. Nucleocytoplasmic shuttling of RUNX proteins has been attributed to RUNX association with microtubules via α-tubulin.145 Within the nucleus, RUNX proteins are concentrated at microenvironments—transcription sites and nuclear matrix—via the NMTS146 or at the nucleolus for regulation of ribosomal RNA genes. This allows for discrete and concentrated output of RUNX activity for efficient transcription regulation. Moreover, RUNX proteins bind to mitotic chromatin for retention of parental phenotype in progeny cells.75, 142, 147 Recently, RUNX proteins were found to interact with centrosomal proteins such as rootletin and γ-tubulin and localize at the centrosome.67 Considering that the centrosome is important in microtubule organization, mitotic division and as a scaffold for regulatory proteins involved in transitions between various cell cycle phases,148 it would be interesting to ascertain whether centrosomal RUNX performs any of these roles. In addition, RUNX3 was also detected at the mitotic spindle and midbody.67 The collective presence of RUNX proteins at the centrosome, spindle and midbody—all nontranscriptional sites—indicates the involvement of RUNX in the mechanics of the mitotic apparatus. These observations, together with differential RUNX phosphorylation status during mitosis, indicate necessary but varied roles for RUNX during mitosis.

The importance of spatial distribution for proper RUNX function is underlined by the observation of cytoplasmic localization of RUNX3 in various cancer cells, such as breast, colorectal and gastric cancer.3, 34, 149 Inappropriate localization of RUNX3 in the cytoplasm would inactivate RUNX3-mediated transcriptional regulation and presumably its tumor suppressive properties. Interestingly, it has been shown that oncoproteins such as MDM2 E3 ligase, Pim1 kinase, Jab1 kinase and Src kinase promote cytoplasmic localization of RUNX3 through ubiquitination and phosphorylation, respectively, suggesting that cytoplasmic sequestration is a mechanism for oncogenicity.102, 114, 150, 151 Cytoplasmic localization has also been observed for RUNX2: STAT1 inhibits nuclear localization of RUNX2.131 It is thus likely that cytoplasmic sequestration of RUNX proteins is a conserved mechanism for cellular/developmental processes, and indeed, high amounts of RUNX3 have been detected in the cytoplasm of chief cells.34

Interplay of RUNX proteins

The fact that haploinsufficiency and inactivation of RUNX play causal roles in human diseases indicates that RUNX activity must be stringently regulated. For example, in neuroblastoma cells, RUNX1 is required for cell proliferation; however, overexpression of RUNX1 or RUNX3 promotes cell death, further emphasizing that cell proliferation is very sensitive to dosage of RUNX activity.152 Intriguingly, inverse correlation between RUNX1 and RUNX3 expression has been reported, suggesting crosstalk between RUNX members.153, 154 RUNX3 binds to conserved RUNX sites at RUNX1 promoter where it inhibits RUNX1 transcription by its VWRPY motif, which presumably recruits TLE corepressor.155, 156 In B cells immortalized by Epstein–Barr virus (EBV), cell proliferation is associated with high RUNX3 expression (induced by EBV transcription factor EBNA2) and corresponding decrease in RUNX1 levels153; knockdown of RUNX3 results in reactivation of RUNX1 and growth repression. Clearly, a specific stoichiometric balance of RUNX proteins is required for cell proliferation. The exact mechanism of how RUNX proteins regulate this interesting biological phenomenon is an interesting area for further research. Aside from cross repression, RUNX proteins cooperate with each other: chondrocyte maturation is highly dependent on the dosage of both Runx2 and Runx3,157 whereas regulation of sternal morphogenesis involves cooperation of Runx2 and Runx1.158

RUNX3 in relation to tumor suppressor p53

RUNX3 is ideally equipped to suppress tumor formation: its close links with differentiation programs allow surveillance and rapid response at potential cancer cell of origin; its downstream transcriptional targets include proapoptotic and cell cycle inhibitor genes, which on TGF-β stimulation, result in apoptosis or cell cycle arrest; it modulates oncogenic Wnt signals through direct inhibition of TCF4-β-catenin activity and it cooperates with p53–ARF pathway to induce a senescence-like growth arrest in response to oncogenic Ras signals. There are also growing indications of RUNX3 involvement during DNA damage. All point toward a gatekeeper role for RUNX3 when cells are confronted with hyperproliferative signals, a premise further supported by the frequent epigenetic inactivation of RUNX3 in human solid tumors, as well as impaired differentiation and precancerous development in Runx3−/− mouse models. Moreover, RUNX3 possesses multiple post-translational modification sites that allow for dynamic regulation of RUNX3 protein in response to environmental cues. In many aspects, RUNX3 is similar to well-known tumor suppressor p53. In the presence of genotoxic or oncogenic stress, p53 induces transcription of genes that arrest the cell cycle, activates the apoptotic program or senescence. As the stimuli that activate RUNX3 may be different (e.g., TGF-β as opposed to DNA damage), both RUNX3 and p53 appear to function in independent pathways to suppress tumorigenesis. Moreover, it has been argued that inactivation of p53 occurs after RAS mutation at the later stages of cancer.159 This may indicate that p53 suppresses tumor progression at advanced stages of cancer; conversely, RUNX3 appears to be a gatekeeper, suppressing tumor formation during cancer initiation.3 This difference in timing of tumor suppression suggests that p53 and RUNX3 contribute differently in their respective tumor suppressor roles, thereby providing additional safeguards against the myriad ways of tumor progression.

Although multiple lines of evidence indicate that RUNX3 exert tumor suppressor activity (see above topics on induction of apoptosis, attenuation of oncogenic Wnt signaling and tumor-prone phenotype of Runx3 KO mice), the role of RUNX3 as tumor suppressor has been disputed by Levanon et al.160 Their premise was primarily based on their inability to detect the presence of Runx3 in normal mouse GI epithelium; not experimental assessment of the biochemical properties of RUNX3 protein and its tumor suppressor activity.

Whether Runx3 is expressed in GI tract (GIT) epithelium remains controversial.161 Although independent groups have described the presence of RUNX3 in human gastric epithelium,162 the level of Runx3 expression in mouse gastric epithelium appears heavily subjected to environmental conditions in different mouse facilities. This is perhaps expected, considering that RUNX3 expression can be increased on cellular stress. The issue of Runx3 expression in mouse GIT epithelium has been described elsewhere161 and will be examined in detail in the near future.


The RUNX field is constantly evolving, with sometimes contradictory theories put forth. As discussed above, the RUNX protein can draw from a large repertoire of interacting partners that fine tunes its activity and endows it with a broad range of roles. RUNX activity may be significantly modified by multiple proteins that compete for binding to RUNX protein. How RUNX selects its partner, or is in turn selected, depend in part on post-translational modifications as dictated by the cellular environment.

The ability of RUNX proteins to modulate various key signaling pathways in the developmental process (e.g., Wnt, Hippo and TGF-β) is a likely reflection of an ancient metazoan protein that has evolved through the ages to provide for increased biological complexity of higher organisms. Its critical roles in multiple aspects of the development are commensurate with the fact that loss of RUNX function results in aberrations in differentiation and severe developmental defects. In long-lived organisms (e.g., mammals), which require defense against malignant transformation, evolution has bestowed on various RUNX family members' strong tumor suppressor activity, such that RUNX deficiency is strongly linked to tumorigenesis. Understanding the molecular activities of RUNX proteins may therefore have therapeutic value in the early detection and treatment of a wide range of cancers.