Roles of STAT3 Defined by Tissue-Specific Gene Targeting

Shizuo Akira*,1

1Department of Host Defense, Research Institute for Microbial Diseases, Osaka University; CREST of Japan Science and Technology Corporation, 3-1 Yamada-oka, Suita, Osaka 565-0871, Japan

The physiological role of each individual STAT protein and cardiotrophin-1, utilize combinations of a com- is now being examined through the study of `knockout' mon signal-transducing subunit, gp130, and various (KO) mice, harboring a null allele for the particular ligand-binding subunits (Kishimoto et al., 1995). gene. In contrast to other STATs de®cient mice that are Propagation of these cytokine signals requires gp130, born alive, STAT3-de®cient mice die during early which activates STAT3 and the Ras/MAP kinase embryogenesis. However, the role of STAT3 in adult pathway (Akira, 1997) (Figure 1). The binding of tissues can be assessed by utilizing the Cre-loxP ligands to the ligand-binding subunit induces the recombination system to ablate the gene in later life. homodimerization of gp130 and subsequently activates Analyses of tissue-speci®c STAT3-de®cient mice indicate gp130-associated JAKs by transphosphorylation. that STAT3 plays a crucial role in a variety of biological JAKs then phosphorylate tyrosine residues in the functions including cell growth, suppression and induction cytoplasmic portion of gp130 as well as JAKs of apoptosis, and cell motility. Oncogene (2000) 19, themselves. Phosphotyrosines on the receptor recruit 2607 ± 2611. src homology (SH2) domain-containing signaling molecules. Human gp130 has six tyrosine residues in Keywords: STAT3; knockout; conditional gene targeting the cytoplasmic domain. Among them, the membrane proximal second tyrosine residue (Y759) is responsible for activation of the MAP kinase cascade through Introduction SHP2 (Fukada et al., 1996), and each of other four tyrosine residues [the third(Y767), fourth(Y814), The Janus kinase-signal transducer and activator of ®fth(Y905) and sixth(Y915)] containing the YxxQ transcription (JAK-STAT) signaling pathway is acti- motif is required for STAT3 activation (Stahl et al., vated in response to a larger number of cytokines, 1995). hormones and growth factors (Darnell, 1997). The JAK Ras activation leads to activation of serine/threonine family of protein tyrosine kinases, which phosphorylate kinase Raf-1, followed by sequential phosphorylation and activate STAT proteins, consists of Jak1, Jak3 and of MEK and MAP kinases. Activated MAP kinases Tyk2. These JAK kinases are characterized by the translocate to the nucleus, and phosphorylate tran- possession of a kinase-like domain and a bona ®de scription factors, which results in modulation of gene kinase domain in its C-terminal (Ihle, 1996). STAT expression. Ras activation via cytokine receptors proteins are a family of latent cytoplasmic transcription require several adaptor proteins. The adaptor protein factors that contain the DNA binding domain in the Grb2 associates with the Ras GTP-GDP exchange middle, as well as an SH2 domain, a tyrosine factor Sos, and induces the conversion of inactive phosphorylation site and a transcriptional activation GDP-bound Ras to the active GTP-bound Ras. SHP2 domain in the C-terminal portion. (also called Syp, SHPTP-2, SHPTP-3, PTP2C, and STAT3 was initially identi®ed as APRF (acute phase PTP1D) is a protein tyrosine phosphatase containing response factor), an inducible DNA binding protein two SH2 domains, and functions as an adaptor protein that binds to the IL-6 responsive element within the in recruitment of the Grb2/Sos complex to gp130. promoters of hepatic acute phase protein genes SHP2 is shown to be inducibly associated with gp130 (Wegenka et al., 1993). APRF protein was puri®ed (Fuhrer et al., 1995). from a pool of mouse liver nuclear extracts using DNA STAT3 proteins are initially present in inactive anity chromatography, cDNA encoding mouse forms in the cytoplasm. Upon ligand binding, they APRF was cloned based on the partial amino acid become associated with gp130 via recognition of the sequences (Akira et al., 1994). Mouse APRF cDNA receptor phosphotyrosines by the STAT SH2 domains. encodes an open reading frame of 770 amino acids The activated JAK kinases then phosphorylate STAT predicting a protein of 88 kD. A DNA data base proteins at their tyrosine residues. Thereafter, the search revealed that mouse APRF has a high degree of phosphorylated STAT proteins detach from the homology to STAT1 (52.5% homology). STAT3 was receptor, become homodimerized or heterodimerized, also independently cloned by Zhong et al. (1994). and translocate to the nucleus to activate transcription by interaction with speci®c DNA sequences. In addition to the tyrosine phosphorylation required for Signal transductions that transmit for gp130 both dimerization and translocation to the nucleus, The receptors for the interleukin-6 (IL-6)-type cyto- STAT proteins also require serine phosphorylation for kines, consisting of IL-6, IL-11, LIF, OSM, CNTF transcriptional activation (Wen et al., 1995; Zhang et al., 1995). The carboxy-terminal regions of STAT1, STAT3, STAT4, and STAT5 contain the MAPK *Correspondence: S Akira consensus sequence, although there is no evidence Tissue-specific knockout of STAT3 S Akira 2608 dependent for undierentiated growth. These data show that self-renewal of pluripotent embryonic stem cells is mediated via activation of STAT3, and that STAT3 has a speci®c and nonredundant function in ES cells. It is also shown that SHP2 and MAP kinase activation through gp130 is dispensable for the self- renewal of ES cells. STAT3 activation has further been shown to mediate IL-6- or LIF-induced astrocytic dierentiation of primary cortical neuroepithelial cells (Bonni et al., 1998). It has also been shown that STAT3 is activated by hepatocyte growth factor and mediates epithelial tubulogenesis (Boccaccio et al., 1998).

Figure 1 JAK-STAT and Ras-MAP kinase pathways in gp130- Early embryonic lethality of STAT3-deficient mice dependent signaling STAT3 activity is detected during early postimplantation development in the mouse, suggesting that STAT3 plays a role during early embryogenesis (Duncan et al., 1997). showing that MAPKs are actually involved in the In fact, STAT3-de®cient mice die early in embryogenesis, serine phosphorylation of STATs in vivo. prior to gastrulation (Takeda et al., 1997). By 7.5 days STAT3 is also activated in response to G-CSF and post-coitum STAT3 mRNA is expressed in the extra leptin, the receptors for which are both homologous to embryonic visceral endoderm, which is the principal site gp130. Furthermore, STAT3 is activated in response to of nutrient exchange between the maternal and embryo- stimulation of several receptor tyrosine kinases (epi- nic environments. The timing of the degeneration of dermal growth factor, CSF-1, and PDGF), and by STAT37/7 embryos coincides with the onset of STAT3 members of the interferon (IL-10, IFNg, and IFNa) expression in visceral endoderm in wild-type mice, and IL-2 (IL-2, IL-7, and IL-15) families. These suggesting that STAT37/7 lethality may be due to a receptor molecules harbor a common STAT3 docking defect in visceral endoderm function, such as nutritional motif (YxxQ) in their cytoplasmic domain (Stahl et al., insuciency. The ligand that activates STAT3 in visceral 1995). endoderm remains unknown.

Roles of STAT3 revealed in cell culture systems Tissue-specific targeting of STAT3 The role of STAT3 has been investigated using the In an attempt to assess the role of STAT3 in adult dominant-negative STAT3 mutant (STAT3DN) or a tissues, we utilized the Cre-loxP recombination system, conditionally active form of STAT3. In the mouse in which a speci®c region of DNA ¯anked by loxP sites myeloid leukemic M1 cells, overexpression of can be deleted by expression of the Cre protein (Figure STAT3DN abolished the dierentiation response to 2). We ®rst generated ¯oxed-STAT3 mice, in which IL-6 or LIF, indicating that STAT3 activation is two loxP sites were introduced 5' and 3' of the exon essential for IL-6 or LIF-mediated growth arrest and encoding the tyrosine residue critical for STAT dierentiation of M1 cells (Minami et al., 1996; activation. Floxed-STAT3 mice were mated with Nakajima et al., 1996). In the mouse pro-B cell line transgenic mice expressing Cre protein in speci®c BAF-B03, overexpression of STAT3DN did not tissues. For T cell-speci®c deletion, we used transgenic proliferate and underwent cell death accompanied by mice expressing Cre protein speci®cally in T cells under DNA fragmentation, indicating that STAT3 transmits the control of the Lck promoter. STAT3-de®cient T an anti-apoptotic signal (Fukada et al., 1996). cells displayed a severely impaired proliferative re- ES cells are nontransformed stem cells that can be sponse to IL-6 due to a defect in IL-6 mediated continuously propagated in vitro in the presence of suppression of apoptosis, demonstrating the anti- LIF. ES clones constitutively expressing STAT3DN apoptotic function of STAT3 (Takeda et al., 1998) showed an increased tendency to dierentiate (Boeuf et (Figure 3). A similar result is shown in the mouse pro- al., 1997). Expression of STAT3DN using an inducible B cell line BAF-B03. Dierent from the result with the promoter in ES cells growing in the presence of LIF pro-B cells in which STAT3 is involved in the speci®cally abrogated self-renewal and promoted expression of bcl-2, an anti-apoptotic gene, STAT3- dierentiation (Niwa et al., 1998). Recently, STAT3 mediated anti-apoptosis in T cells is not mediated by activation is demonstrated to be sucient for the self- bcl-2 induction. The anti-apoptotic gene regulated by renewal of ES cells by using a conditionally active form STAT3 in T cells remain to be unknown. The of STAT3, that is, a fusion protein between STAT3 mechanisms of STAT3-mediated anti-apoptosis may and ER ligand binding domain in which STAT3 is be distinct, depending on the cell type. activated in response to the synthetic ligand 4- STAT3-de®cient T cells also show a partial defect in hydroxytamoxifen (4HT) (Matsuda et al., 1999). ES IL-2-induced proliferation. IL-2 receptors are com- cells expressing STAT5aER or STAT6ER did not posed of the combination of three distinct subunits, IL- maintain the undierentiated state of ES cells in 2Ra, b and g. IL-2Ra is required to convert response to 4HT. Although STAT1 can be activated intermediate-anity receptors (containing IL-2Rb and in response to LIF in ES cells, STAT17/7 ES cells g) into high-anity receptors (containing all three retained responsiveness to LIF and remained LIF chains), which confers a 100-fold increase in binding

Oncogene Tissue-specific knockout of STAT3 S Akira 2609 anity for IL-2 as well as ecient cellular responsive- proliferation. Thus, STAT3 is indirectly involved in T ness to the low concentrations of IL-2. The partial cell proliferation by upregulating the expression of IL- defect in IL-2-induced proliferation of STAT3-de®cient 2Ra and forming high anity receptors. The similar T cells are found to be due to a defect in IL-2-induced result is also demonstrated in STAT5a-de®cient T cells expression of IL-2Ra (Akaishi et al., 1998). Interest- (Nakajima et al., 1997). Therefore, both STAT3 and ingly, high concentrations of IL-2 relieved the defect of STAT5a are independently responsible for the IL-2Ra expression since the absence of one is not compensated by the other (Figure 3). We have also generated mice in which STAT3 is de®cient speci®cally in macrophages and neutrophils (Takeda et al., 1999). These mutant mice were highly susceptible to endotoxin shock and demonstrated increased production of in¯ammatory cytokines such as TNFa, IL-1, and IFNg. Production of in¯ammatory cytokines from STAT3-de®cient macrophages were dramatically augmented in response to LPS. The mice also showed a polarized immune response of the Th1 type as shown by increased secretion of IFNg from splenic T cells. Aging mutant mice developed chronic enterocolitis. These phenotypes are quite similar to those seen in mice lacking IL-10, a cytokine with pleitorophic bioactivities, and is relatively unique in its ability to potentially inhibit production of proin¯ammatory cytokines (Kuhn et al., 1993). Indeed, the response to IL-10 was completely abolished in macrophages and neutrophils. These results indicate that STAT3 functions in vivo in macrophages and neutrophils to signal anti-in¯ammatory responses mediated by IL-10, and that IL-10-mediated anti- in¯ammatory response by macrophages and neutrophils plays a critical role in prevention of excessive Th1 response and chronic in¯ammation (Figure 4). IL-10 receptor system is composed of two subunits, IL-10Ra and IL-10Rb. Binding of IL-10 to the extracellular Figure 2 Cre-loxP system for tissue-speci®c gene targeting. The domain of IL-10Ra activates STAT3. Murine IL-10Ra Cre enzyme recognizes a sequence motif of 34 bp, called loxP. If harbors two redundant STAT3 recruitment sites the target gene is ¯anked by two loxP sites in the same orientation, Cre protein excises the intervening target gene. Tissue-speci®c deletion of the target gene is generated by crossing the mutant mice harboring the target gene ¯anked by two loxP sites to various strains expressing Cre protein in tissue-speci®c manner

Figure 4 Development of chronic colitis in mice devoid of STAT3 in macrophage and neutrophil-speci®c manner. Gut macrophages are expected to be continuously activated by foreign substances such as bacteria and their products present in the mucosa, and secrete in¯ammatory cytokines and mediators including TNFa, IL-1, and NO, which may result in tissue damage of the intestinal wall. Activated macrophages also secrete Figure 3 Involvement of STAT3 in IL-2- and IL-6-mediated T IL-12 and IL-18, which induce development of Th1 cell to cell proliferation. STAT3-de®cient T cells show a defect in IL-6- produce IFNg, which, in turn, activates macrophages. In normal and IL-2-induced proliferations. STAT3 is involved in T cell mice, IL-10 is simultaneously secreted from activated macro- proliferation by distinct mechanisms. STAT3 activation is phages and suppresses their activation to maintain the ®nely responsible for anti-apoptosis in IL-6-induced T cell proliferation reglated homeostasis in vivo. However, in STAT3-de®cient whereas in the case of IL-2-induced proliferation STAT3 is macrophages and neutrophils, IL-10-induced suppression does indirectly involved in T cell proliferation by upregulating the not occur, and both macrophages and neutrophils are constitu- expression of IL-2Ra tively activated, resulting in progression to chronic in¯ammation

Oncogene Tissue-specific knockout of STAT3 S Akira 2610 (427YQKQ430 and 477YLKQ480) (Weber-Nordt et in vitro migration of STAT3-de®cient epidermal cells al., 1996). Structure-function analysis of the intracel- was severely impaired, although proliferation was lular domain of the IL-10Ra chain shows that two unaected. This suggests that the defect in wound redundant STAT3 recruitment sites are required for all healing is due to the poor motility of epidermal cells. IL-10 dependent eects, whereas IL-10-dependent anti- Furthermore, the mutant mice had sparse hair and in¯ammatory function requires the presence of a developed ulcers spontaneously with age. Additionally carboxy-terminal sequence on the intracellular domain the mutant mice expressed aberrant hair follicles, and of the IL-10Ra (Riley et al., 1999). This result indicates marked hyperplasia of the epidermis (acanthosis) with the IL-10-induced inhibition of TNFa production hyperkeratosis and sclae-crusts. There was pronounced requires two distinct regions of the IL-10Ra intracel- in¯ammatory in®ltration and ®brosis throughout the lular domain and thereby establish a distinctive dermis. The phenotype in aged STAT3-de®cient mice molecular basis for the proliferative versus the anti- appears to be consequence to the impaired wound in¯ammatory action IL-10. healing and disorganized hair cycling. The functional role of STAT3 in skin was assessed Mammary gland involution is characterized by by crossing the ¯oxed-STAT3 mice with mice expres- extensive apoptosis of the epithelial cells. STAT5 is sing the Cre-transgene from the keratin 5 promoter activated during pregnancy and lactation but is rapidly (Sano et al., 1999). The mutant mice were viable and down regulated during involution. By gene targeting, displayed no developmental alterations in the epidermis STAT5 has been shown to be essential for normal and hair follicles by postnatal day 11 (PD11), showing mammopoiesis and lactogenesis (Liu et al., 1997; that STAT3 in keratinocytes is not involved in the Teglund et al., 1998). Conversely, STAT3 is speci®cally morphogenesis of skin and hair follicles. However, the activated at the start of involution. The reciprocal second hair cycle was impaired in STAT3-de®cient activation of STAT3 and 5 at the onset of apoptosis mice. Follicular morphogenesis starts at 14.5 days post suggests opposing roles for these STATs in the coitus through morphogenic mesenchymal-epithelial regulation of apoptosis in the mammary gland. The interactions. Hair follicular rudiments grow down- role of STAT3 in the mammary gland was addressed wards and dierentiate to develop the complex hair by crossing the ¯oxed-STAT3 mice with mice expres- structure until around PD17. Then, follicles undergo sing the Cre-transgene form the milk protein gene b- cellular quiescent process (catagen) and ®nally, com- lactoglobulin (BLG) promoter (Chapman et al., 1999). plete rest phase (telogen). Around PD21, the second Following weaning, a decrease in apoptosis and a anagen is initiated in response to remodeling mesench- dramatic delay of involution were observed in STAT3- ymal signal. In STAT3-de®cient mice, the second de®cient mammary tissue. No marked dierences were anagen was not observed and remained in a telogen seen in the regulation of Bcl-xL or Bax between the stage. These results suggest that STAT3 is essential to normal and STAT3-de®cient mammary glands. Involu- the second and subsequent hair cycles (skin remodel- tion is normally associated with a signi®cantly ing) although it is dispensable for the ®rst hair cycle increased level of insulin-like growth factor-binding (morphogenesis). Skin wound healing was also severely protein-5 (IGFBP-5), which has been suggested to impaired in STAT3-de®cient mice when the mice were induce apoptosis by sequestering insulin-like growth wounded with a biopsy punch and the process of factor-1 (IGF-1) to casein micelles, thereby inhibiting healing was monitored. The fact that no dierence was its survival function. The increase in IGFBP-5 levels shown between STAT3-de®cient mice and control mice was strongly suppressed in STAT3-de®cient mice, in the dermal responses to wound such as granulation, showing that IGFBP-5 is one target for STAT3. in¯ammation and neovascularization indicates that the However, it remains unclear whether IGFBP-5 is retarded wound healing is due not to an impairment in directly dependent on STAT3 binding to the IGFBP- secondary development of dermal components but 5 promoter, although the human IGFBP-5 promoter rather to a fault in epidermal regeneration. Since cell contains a consensus STAT-binding element. migration and proliferation are critical events in re- Taken together, these analyses of tissue-speci®c epithelialization of wounds, the motility and growth of STAT3-de®cient mice demonstrate that STAT3 plays keratinocyte was examined. Growth factor-dependent a crucial role in a variety of biological functions

Table 1 Phenotypes of tissue-speci®c knockout of STAT3 Strain expressing Cells expressing Cre protein Cre protein Phenotype

Lck-Cre T cell Normal T cell development, impaired IL-6-dependent T cell proliferation due to lack of anti-apoptotic activity, partial defect in IL-2-dependent T cell proliferation due to defective IL-2-induced IL-2Ra expression

LysM-Cre Macrophage, Complete abolishment of suppressive effects of IL-10 on inflammatory cytokines from neutrophil macrophages and neutrophils, augmented production of inflammatory cytokines from macrophages in response to endotoxin, high susceptibility to endotoxin shock, enhanced Th1 response and development of chronic enterocolitis

Keratin5-Cre Keratinocyte Normal development of hair cycle and hair follicles, severely compromized hair cycle and wound healing processes, impaired growth factor-dependent in vitro migration of keratinocytes despite normal proliferative responses

BLG-Cre Mammary gland Decrease in mammary epithelial apoptosis and delay of involution

Oncogene Tissue-specific knockout of STAT3 S Akira 2611 including cell growth, anti-apoptosis, apoptosis and cell the STAT2 and 6 on chromosome 6 genes have arisen motility depending on the cell type and stimulus (Table via a tandem duplication of an ancestral locus 1). (probably chromosome 11 harboring the STAT3 and 5 genes). Classical and tissue-speci®c targeting disruption of Discussion the STAT3 gene clearly showed the critical role of STAT3 in many aspects of biological functions. In the Targeted disruption has disclosed the speci®c function future the identi®cation of target genes regulated by of each STAT protein. It is noteworthy that STAT3 STAT3 will reveal the molecular mechanisms under- and 5 are expressed in many cell types, activated by a lying the STAT3-mediated biological responses in variety of cytokines and growth factors, and play a role various tissues. in various aspects of biological responses whereas other STAT proteins (STAT1, 2, 4, and 6) plays speci®c roles in host defenses. This suggests that the development of host defense mechanisms in mammals may have Acknowledgments required STAT proteins to evolve speci®c roles in the I thank Dr K Takeda for preparing the Figures, and T immune response. In fact, the chromosomal localiza- Aoki for excellent secretarial assistance. This work was in tion of the mouse STAT genes suggests that the part supported by grants from the Ministry of Education STAT1 and 4 genes cosegregating on chromosome 1, of Japan.

References

AkaishiH,TakedaK,KaishoT,ShinehaR,SatomiS, Nakajima H, Liu X, Wynshaw-Boris A, Rosenthal LA, Takeda J and Akira S. (1998). Int. Immunol., 10, 1747 ± Imada K, Finbloom DS, Henninghausen L and Leonard 1751. WJ. (1997). Immunity, 7, 691 ± 701. Akira S. IL-6-regulated transcription factors. (1997). Int. J. Nakajima K, Yamanaka Y, Nakae K, Kojima H, Ichiba M, Biochem. Cell. Biol., 29, 1401 ± 1418. Kiuchi N, Kitaoka T, Fukada T, Hibi M and Hirano T. Akira S, Nishio Y, Inoue M, Wang X, Wei S, Matsuzaka T, (1996). EMBO J., 15, 3651 ± 3658. YoshidaK,SudoT,NarutoMandKishimotoT.(1994). Niwa H, Burdon T and Smith A. (1998). Genes Dev., 12, Cell, 77, 63 ± 71. 2048 ± 2060. Boccaccio C, Ando M, Tamagnone L, Bardelli A, Michieli P, Riley JK, Takeda K, Akira S and Schreiber RD. (1999). J. Battistini C and Comoglio PM. (1998). Nature, 391, 285 ± Biol. Chem., 274, 16513 ± 16521. 288. Sano S, Itami S, Takeda K, Tarutani M, Yamaguchi Y, Boeuf H, Hauss H, De Graeve F, Baran N and Kedinger N. Miura H, Yoshikawa K, Akira S and Takeda J. (1999). (1997). J. Cell. Biol., 138, 1207 ± 1217. EMBO J., 18, 4657 ± 4668. Bonni A, Sun Y, Nadal-Vicens M, Bhatt A, Frank DA, Stahl N, Farruggella TJ, Boulton TG, Zhong Z, Darnell Jr, Rozovsky I, Stahl N, Yancopoulos GD and Greenberg JE and Yancopoulos GD. (1995). Science, 267, 1349 ± ME. (1998). Science, 276, 477 ± 483. 1353. Chapman RS, Lourenco PC, Tonner E, Flint DJ, Selbert S, Takeda K, Clausen BE, Kaisho T, Tsujimura T, Terada N, Takeda K, Akira S, Clarke AR and Watson CJ. (1999). FoÈ rster I and Akira S. (1999). Immunity, 10, 39 ± 49. Gene Dev., 13, 2604 ± 2616. Takeda K, Kaisho T, Yoshida N, Takeda J, Kishimoto T and Darnell Jr JE. (1997). Science, 277, 1630 ± 1635. Akira S. (1998). J. Immunol., 161, 4652 ± 4660. Duncan SA, Zhong Z, Wen Z, Darnell Jr JE. (1997). Dev. Takeda K, Noguchi K, Shi W, Tanaka T, Matsumoto M, Dyn., 208, 190 ± 198. Yoshida N, Kishimoto T and Akira S. (1997). Proc. Natl. Fuhrer DK, Feng GS and Yang YC. (1995). J. Biol. Chem., Acad.Sci.USA,94, 3801 ± 3804. 270, 24826 ± 24830. Teglund S, McKay C, Schuetz E, van Deursen JM, Fukada T, Hibi M, Yamanaka Y, Takahashi-Tezuka M, Stravopodis D, Wang D, Brown M, Bodner S, Grosveld Fujitani Y, Yamaguchi T, Nakajima and Hirano T. G and Ihle JN. (1998). Cell, 93, 841 ± 850. (1996). Immunity, 5, 449 ± 460. Weber-Nordt RM, Riley JK, Greenlund AC, Moore KW, Ihle JN. (1996). Cell, 84, 331 ± 334. Darnell JE and Schreiber RD. (1996). J. Biol. Chem., 271, Kishimoto T, Akira S, Narazaki M and Taga T. (1995). 27954 ± 27961. Blood, 86, 1243 ± 1254. Wegenka UM, Buschmann J, Lutticken C, Heinrich PC and Kuhn R, Lohler J, Rennick D, Rajewsky K and Muller W. Horn F. (1993). Mol. Cell. Biol., 13, 276 ± 288. (1993). Cell, 75, 263 ± 274. Wen Z, Zhong Z and Darnell Jr JE. (1995). Cell, 82, 241 ± Liu X, Robinson GW, Wagner KU, Garrett L, Wynshaw- 250. Boris A and Henninghausen L. (1997). Gene Dev., 11, ZhangX,BlenisJ,LiHC,SchindlerCandChen-KiangS. 179 ± 186. (1995). Science, 267, 1990 ± 1994. Matsuda T, Nakamura T, Nakao K, Arai T, Katsuki M, Zhong Z, Wen Z and Darnell JE. (1994). Science, 264, 95 ± Heike T and Yokota T. (1999). EMBO J., 18, 4261 ± 4269. 98. Minami M, Inoue M, Wei S, Takeda K, Matsumoto M, Kishimoto T and Akira S. (1996). Proc. Natl. Acad. Sci. USA, 93, 3963 ± 3966.

Oncogene