PREVIEWS

37–42. and Ratcliffe, P. (1997). Proc. Natl. Acad. Sci. Semenza, G. (1999). Annu. Rev. Cell Dev. Biol. USA 94, 8104–8109. 15, 551–578. Kondo, K., and Kaelin, W.G., Jr. (2001). Exp. Cell Res. 264, 117–125. Nakayama, K., Kanzaki, A., Hata, K., Katabuchi, Wiesener, M.S., Munchenhagen, P.M., Berger, I., H., Okamura, H., Miyazaki, K., Fukumoto, M., Morgan, N.V., Roigas, J., Schwiertz, A., Kondo, K., Klco, J., Nakamura, E., Lechpammer, and Takebayashi, Y. (2002). Cancer Lett. 176, Jurgensen, J.S., Gruber, G., Maxwell, P.H., M., and Kaelin, W.G., Jr. (2002). Cancer Cell 1, 215–223. Loning, S.A., et al. (2001). Cancer Res. 61, this issue, 237–246. 5215–5222. Ryan, H.E., Lo, J., and Johnson, R.S. (1998). Maranchie, J.K., Vasselli, J.R., Riss, J., EMBO J. 17, 3005–3015. Zhong, H., De Marzo, A.M., Laughner, E., Lim, Bonifacino, J.S., Linehan, W.M., and Klausner, M., Hilton, D.A., Zagzag, D., Buechler, P., Isaacs, R.D. (2002). Cancer Cell 1, this issue, 247–255. Ryan, H., Poloni, M., McNulty, W., Elson, D., W.B., Semenza, G.L., and Simons, J.W. (1999). Gassmann, M., Arbeit, J., and Jonhson, R. Cancer Res. 59, 5830–5835. Maxwell, P., Dachs, G., Gleadle, J., Nicholls, L., (2000). Cancer Res. 60, 4010–4015. Harris, A., Stratford, I., Hankinson, O., Pugh, C.,

RUNX: A trilogy of cancer genes

The RUNX family of transcription factors plays pivotal roles during normal development and in neoplasias. Recent data involve RUNX3 as an important tumor suppressor in gastric cancers and pose interesting questions about how perturbed levels and interspecific competition among RUNX family members may contribute to tumorigenesis.

On a worldwide basis, gastric cancer is a The work of Li and coworkers ly metastatic tumors. Upon examination major cancer-related killer. In Japan and (2002) completes the mouse knockout of 119 tumor samples, Li et al. were able certain other Asian countries gastric can- analyses of the Runx family and under- to identify only a single nucleotide transi- cer tops the list of cancer-induced deaths. scores the role of Runt family members tion causing an arginine-to-cytosine con- While clear links have been discovered as cancer-related genes. As previously version within the conserved Runt between environmental factors, such as found for Runx1 (Okuda et al., 1996), domain, which did not strengthen the Helicobacter pylori infection, dietary com- Runx2 (Otto et al., 1997), and Cfbβ tumor suppressor argument much. ponents, and gastric cancer frequencies, (Wang et al., 1996), genetic ablation of However, the high CpG nucleotide con- the genetic basis for gastric cancer devel- Runx3 has profound effects. While tent triggered analysis of DNA methyla- opment is still largely unclear. In the April Runx3 knockout mice are born in tion status, and interestingly RUNX3 5 issue of Cell, Li and coworkers (2002) Mendelian ratios, they die soon after hypermethylation in a large number of describe the uncovering of RUNX3/ birth probably due to starvation. Runx3 primary tumor samples was found to cor- AML2/CBFA3/PEBP2αC as a candidate is strongly expressed in gastrointestinal relate with gene silencing, which indi- tumor suppressor in gastric cancer devel- organs in the developing embryo and cates an unusual strong prevalence for opment. RUNX3 belongs to the Runt throughout adult life of the mouse and epigenetic gene silencing. To further domain family of transcription factors, the gastric epithelium in Runx3 knock- establish a causal link between RUNX3 which consists of 3 DNA binding α sub- outs displays hyperplasia and a reduced expression and oncogenesis, Li et al. units, RUNX1, RUNX2, and RUNX3 (see apoptotic rate. Interestingly, when ana- injected gastric cancer cells engineered Table 1 for alternative names), each of lyzed in primary cultures Runx3−/− gastric to overexpress either wild-type or Runt which is capable of forming heterodimers epithelial cells are less sensitive to TGF- domain mutated RUNX3 into nude mice. with the common β subunit CBFβ. RUNX β-mediated growth inhibition, based on a While RUNX3 mediated a significant heterodimers are relatively weakly acting marked failure to enter apoptosis when reduction in tumorigenicity, the Runt transcriptional regulators, the potency of treated with TGF-β. To investigate the domain mutant aggravated tumor forma- which can be induced by associations potential connection between RUNX3 tion. Moreover, gastric epithelial cells with transcriptional (co)activators, such and gastric cancers in humans, Li et al. immortalized by loss of p53 only were as MYB, ETS, and p300/CBP, or co- (2002) analyze a series of gastric cancer able to form tumors in nude mice when repressors such as TLE1 and mSin3A cell lines and primary human gastric Runx3 was also deleted. Together, these (Perry et al., 2002). The main family fea- tumors. Out of 46 primary human results underline the role of RUNX3 as a ture, the 128 amino acid runt domain tumors, 30% displayed hemizygosity of bona fide tumor suppressor. named for its high homology to the RUNX3 with a significant correlation It is now clear that all three Runx Drosophila pair-rule protein runt, facili- between RUNX3 loss and gastric cancer family members play important roles in tates dimerization and DNA binding. Like progression stage. Furthermore, RUNX3 normal developmental processes as their counterparts in D. melanogaster expression analysis revealed that on well as in cancers (Table 1). RUNX1, and C. elegans, mammalian RUNX fami- average 60% of the analyzed primary perhaps better known as AML1, plays a ly transcription factors play important human gastric tumors exhibited reduced critical role in hematopoietic develop- roles in cell fate determination during RUNX3 levels rising to nearly 90% ment, and genetic ablation of either development. among the late stage, representing high- Runx1 or Cbfβ results in embryonic

CANCER CELL : APRIL 2002 213 PREVIEWS

Table 1. Functions and oncogenesis of Runt-related transcription factorsa

Gene Alternative Proposed essential Mouse (−/−) phenotype Tumorigenesis gene names function

RUNX1 AML1/CBFA2/ Definitive Embryonic lethal. Absence Hemizygocity in humans predisposes to acute PEBP2αB hematopoiesis of fetal liver hematopoiesis. myeloid leukemia. Frequent translocation breakpoint. Common insertion site in retrovirus- induced mouse leukemia. RUNX2 AML3/CBFA1/ Bone ossification Dies at birth from Transgenic Runx2 overexpression predisposes to PEBP2αA respiratory failure. T-cell lymphomas. Common insertion site in retrovirus-induced mouse leukemia. RUNX3 AML2/CBFA3/ Development of the Dies soon after birth. Frequently inactivated in human gastric cancers. PEBP2αC gastrointestinal tract Hyperplastic gastric epithelium. Common insertion site in retrovirus-induced mouse leukemia. aSee text for references lethality and a complete lack of fetal liver apparently opposing effects of RUNX 2000). So where does RUNX3 fit into the hematopoiesis (Okuda et al., 1996; family members as, on the one hand picture? On the one hand, Runx3 loss Wang et al., 1996). RUNX1 has long dominant-acting oncogenes, while on may be an early event, as is suggested been recognized as an important the other hand being important tumor by the hyperplasia in the Runx knockout translocation breakpoint in human suppressors? One possible explanation mice; on the other hand, the finding of Li leukemias, with the TEL-AML1 t(12;21) could be that since all three RUNX het- and coworkers (2002) that RUNX3 loss fusion accounting for 20% of acute lym- erodimers are known to recognize the increases with cancer progression sug- phoblastic leukemia (ALL) cases and same consensus DNA core motif gests a role in tumor progression. This the AML1-ETO t(8;21) fusion accounting (TGt/cGGT), erroneous expression of dual action of Runx loss may be for 12% of acute myeloid leukemias RUNX members outside of their normal explained by its effects on inducing both (AML) (Look, 1997). Importantly, a num- tissues might exert oncogenic effects cell proliferation (inducing hyperplasia) ber of the translocation breakpoint through competitive interference with the and on preventing apoptosis (which products involving RUNX1 have been RUNX heterodimer species normally could be selected for in tumor progres- shown to have trans-dominant effects present within that particular tissue. Also, sion). Indeed, a role for RUNX3 in and block the normal functions of the RUNX members differ in their associa- proapoptotic pathways is sustained by RUNX1/CBFB heterodimer (reviewed in tions with coacting factors; RUNX3 for the fact that Runx3−/− gastric epithelial Perry et al., 2002).The hypothetical func- instance entirely lacks a domain present cells display apoptotic defects in tion of RUNX1 as a tumor suppressor for in RUNX1 and RUNX2 responsible for response to TGF-β. In support of a role AML was further strengthened by the interaction with Ets family members (Bae for RUNX family members in TGF-β and finding that familial mutations with et al., 1995). Along the same lines, the bone morphogenic protein signaling RUNX1 cause thrombocytopenia and “wiring” of RUNX members, in terms of pathways are the findings that all three predispose to acute myeloid leukemia activating and repressing signaling path- RUNX members have been found to bind (Song et al., 1999). The Runx2 gene ways, may differ in different cells types. It R-Smads (Hanai et al., 1999) and that product is pivotal for osteoblast differenti- will be of obvious importance to use thymocytes from Runx2 transgenic mice ation and bone ossification (Otto et al., expression profiling and other tech- are hypersensitive to TGF-β (Vaillant et 1997). Furthermore, RUNX2 is haploin- niques to find the relevant downstream al., 1999). sufficient, and heterozygous RUNX2 targets for each of the Runx het- Interestingly, Runx3 loss appears loss of function gives rise to cleidocranial erodimeric transcription factors. specifically to predispose to gastric can- dysplasia in both humans and mice, a The recognition of Helicobacter pylori cer but not to colon cancer, despite disorder characterized by multiple skele- as a major causative agent in gastric can- equally high expression of Runx3 in both tal abnormalities (Otto et al., 1997; cer development has led to the classifica- gastric and colonic epithelial cells. Mundlos et al., 1997). The oncogenic tion of H. pylori as a Class I carcinogen by Among several other speculative expla- properties of Runx2 were demonstrated the World Health Organization. H. pylori nations, one is that this could relate to in transgenic mice where Runx2 overex- infection causes release of reactive oxy- the unique milieu and cofactors in the pression pertubates T cell development gen species, mitogenic stimulation, and stomach, such as low pH, dietary factors and synergizes strongly with c-myc in constitutive inflammatory response in and H. pylori. Given the capability of both lymphomagenesis (Vaillant et al., 1999). gastric epithelium (Peek and Blaser, H. pylori and loss of Runx3 to induce ini- Interestingly, all three Runx family mem- 2002; Karin et al., 2002). Early steps in tial lesions and hyperplasia, an intriguing bers have been identified as putative gastric neoplasias involve transition of possibility could be that Runx3 is one of proto-oncogenes in mouse insertional normal epithelial mucosa to gastritis, a the downstream “targets” of H. pylori mutagenesis leukemia models (Li et al., condition promoted by H. pylori, implicat- action. Having the Runx3 knockout mice 1999; Stewart et al., 1997; A.H.L. and ing H. pylori in tumor initiation and pro- in hand, this and several other important M.v.L., submitted), hence further indicat- motion (Peek and Blaser, 2002). Later questions can now be investigated ing that absolute levels of the Runx het- stages involve many of the “common directly.While the early death of Runx3−/− erodimers are important and need to be themes” in tumorigenesis: p53 loss of mice precludes tumorigenesis experi- tightly controlled. function, K-ras activation, and E-cad- ments, careful monitoring of Runx3+/− How can we seek to understand the herin loss or mutation (Kuniyasu et al., mice, perhaps in combination with other

214 CANCER CELL : APRIL 2002 PREVIEWS

genetic lesions such as p53 ablation, is Selected reading Otto, F., Thornell, A.P., Crompton, T., Denzel, A., expected to reveal gastric cancer pre- Gilmour, K.C., Rosewell, I.R., Stamp, G.W., disposition and could provide more Bae, S.C., Takahashi, E., Zhang, Y.W., Ogawa, Beddington, R.S., Mundlos, S., Olsen, B.R., et al. (1997). Cell 89, 765–771. detailed insight into the roles of Runx3 in E., Shigesada, K., Namba, Y., Satake, M., and Ito, Y.(1995). Gene 159, 245–248. tumor initiation and progression and Peek, R.M., and Blaser, M.J. (2002). Nat. Rev. its possible interactions with cofactors. Hanai, J., Chen, L.F., Kanno, T., Ohtani-Fujita, N., Cancer 2, 28–37. Alternatively, a conditional Runx3 knock- Kim, W.Y., Guo, W.H., Imamura, T., Ishidou, Y., Perry, C., Eldor, A., and Soreq, H. (2002). Leuk. out mouse may serve this purpose. In Fukuchi, M., Shi, M.J., et al. (1999). J. Biol. Res. 26, 221–228. Chem. 274, 31577–31582. addition, such mouse models will allow Song, W.J., Sullivan, M.G., Legare, R.D., investigation of the extent to which loss Karin, M., Cao, Y., Greten, F.R., and Li, Z.-W. Hutchings, S., Tan, X., Kufrin, D., Ratajczak, J., of RUNX3 complements or substitutes (2002). Nat. Rev. Cancer 2, 301–310. Resende, I.C., Haworth, C., Hock, R., et al. for other lesions known from gastric can- Kuniyasu, H., Yasui, W., Yokozaki, H., and Tahara, (1999). Nat. Genet. 23, 166–175. cers, such as E-cadherin loss, RAS E. (2000). Langenbecks Arch. Surg. 385, 69–74. Stewart, M., Terry, A., Hu, M., O’Hara, M., Blyth, mutations, and loss of DCC. It can be Li, J., Shen, H., Himmel, K.L., Dupuy, A.J., K., Baxter, E., Cameron, E., Onions, D.E., and expected that this mouse model for gas- Largaespada, D.A., Nakamura, T., Shaughnessy, Neil, J.C. (1997). Proc. Natl. Acad. Sci. USA 94, 8646–8651. tric cancer will greatly assist unraveling J.D., Jr., Jenkins, N.A., and Copeland, N.G. of the underlying molecular causes for (1999). Nat. Genet. 23, 348–353. Vaillant, F., Blyth, K., Terry, A., Bell, M., Cameron, this important widespread disease in the Li, Q.-L., Ito, K., Sakakura, C., Fukamachi, H., E.R., and Stewart, M. (1999). Oncogene 18, 7124–7134. coming years. Inoue, K.-I., Chi, X.-Z., Lee, K.-Y., Nomura, S., Lee, C.-W., Han, S.-B., et al. (2002). Cell 109, Wang, Q., Stacy, T., Miller, J.D., Lewis, A.F., Gu, 113–124. T.L., Huang, X., Bushweller, J.H., Bories, J.C., Anders H. Lund and Look, A.T. (1997). Science 278, 1059–1064. Alt, F.W., Ryan, G., et al. (1996). Cell 87, Maarten van Lohuizen1 697–708. Mundlos, S., Otto, F., Mundlos, C., Mulliken, J.B., Division of Molecular Genetics Aylsworth, A.S., Albright, S., Lindhout, D., Cole, The Netherlands Cancer Institute W.G., Henn, W., Knoll, J.H., et al. (1997). Cell 89, Antoni van Leeuwenhoek ziekenhuis 773–779. Plesmanlaan 121, 1066CX, Amsterdam Okuda, T., van Deursen, J., Hiebert, S.W., The Netherlands Grosveld, G., and Downing, J.R. (1996). Cell 84, 1 E-mail: [email protected] 321–330.

The secrets of selective estrogen receptor modulation: Cell-specific coregulation

A specific increase in the level of a single coactivator appears to enhance estrogen action with tamoxifen at some gene targets in uterine cells but not breast cells.

The discovery and development of antie- estimated that 400,000 women are alive action in the breast. The actions of strogens as treatments for estrogen today because of the success of tamox- tamoxifen in the endometrium are impor- receptor (ER) positive breast cancer ifen treatment. However, tamoxifen is a tant because they illustrate the concept (Lerner and Jordan, 1990) introduced a pioneering medicine over and above its of selective ER modulation. When both new approach for targeted therapy with ability to save lives. breast and endometrial tumors are few side effects compared to traditional The recognition of selective estrogen implanted into immune deficient mice cytotoxic chemotherapy. The novel so- receptor modulation in the laboratory (Gottardis et al., 1988), tamoxifen called nonsteroidal antiestrogens, initial- during the 1980s (Jordan, 2001) has had enhances the growth of endometrial can- ly investigated during the 1960s, were all important implications not only for the cer but prevents the growth of breast classified as partial estrogen agonists in evaluation of the side effects associated cancer. The tamoxifen ER complex is the rat uterus but with a predominantly with tamoxifen, but also has established perceived as an estrogen in endometrial antiestrogen action. This pharmacologic the rationale for a new class of drugs, the cancer cells but as an antiestrogen in activity in the laboratory extrapolated to selective estrogen receptor modulators breast cancer. This concept translated to antitumor action by blocking estrogen- (SERMs). It is now clear that SERMs the clinic with a predicted modest rise in stimulated breast tumor growth at the have potential as multifunctional medi- the incidence of endometrial cancer in ER. Tamoxifen was introduced clinically cines. The SERMs express estrogen-like postmenopausal women during tamox- during the 1970s, and the drug has had a actions in bone, lower circulating choles- ifen therapy. The question is, how is the profound effect on patient survival. It is terol, but produce an antiestrogenic pharmacology of tamoxifen reversed in

CANCER CELL : APRIL 2002 215