Oncogene (2000) 19, 2638 ± 2644 ã 2000 Macmillan Publishers Ltd All rights reserved 0950 ± 9232/00 $15.00 www.nature.com/onc Modulation of STAT signaling by STAT-interacting

K Shuai*,1

1Departments of Medicine and Biological Chemistry, University of California, Los Angeles, California, CA 90095, USA

STATs (signal transducer and of ) play important roles in numerous cellular processes Interaction with non-STAT transcription factors including immune responses, and di€erentia- tion, cell survival and , and oncogenesis. In Studies on the promoters of a number of IFN-a- contrast to many other cellular signaling cascades, the induced identi®ed a conserved DNA sequence STAT pathway is direct: STATs bind to receptors at the named ISRE (-a stimulated ) cell surface and translocate into the nucleus where they that mediates IFN-a response (Darnell, 1997; Darnell function as transcription factors to trigger activa- et al., 1994). Stat1 and Stat2, the ®rst known members tion. However, STATs do not act alone. A number of of the STAT family, were identi®ed in the transcription proteins are found to be associated with STATs. These complex ISGF-3 (interferon-stimulated gene factor 3) STAT-interacting proteins function to modulate STAT that binds to ISRE (Fu et al., 1990, 1992; Schindler et signaling at various steps and mediate the crosstalk of al., 1992). ISGF-3 consists of a Stat1:Stat2 heterodimer STATs with other cellular signaling pathways. This and a non-STAT named p48, a member of the article reviews the roles of STAT-interacting proteins in IRF (interferon regulated factor) family (Levy, 1997). the regulation of STAT signaling. Oncogene (2000) 19, p48 alone can bind to ISRE weakly. Although Stat1 2638 ± 2644. and Stat2 can not bind to ISRE without p48, the ISGF3 complex has a high binding anity toward Keywords: STAT-interacting proteins; signal transduc- ISRE. The ISGF3 complex established a paradigm in tion; transcription which STAT proteins may a€ect transcription by modulating other non-STAT transcription factors. In fact, since STAT proteins were ®rst identi®ed through Introduction the characterization of ISGF3, it had been questioned whether STATs alone can bind to DNA. Characteriza- In unstimulated cells, a STAT protein exists in the tion of IFN-g-induced gene activation demonstrated as a monomer. Upon activation by that Stat1 is a sequence speci®c DNA binding protein phosphorylation in response to that drives transcription through GAS (g-activation stimulation, STAT forms a dimer through SH2- site) (Decker et al., 1991; Shuai et al., 1992). It now phosphotyrosyl interactions (Becker et al., 1998; Chen appears that a STAT protein may participate in et al., 1998; Shuai et al., 1994). The STAT dimer transcriptional activation through four distinct me- then translocates into the nucleus to activate chanisms: (1) A STAT protein may bind to its own transcription. It is believed that the STAT dimer is DNA target site to directly drive transcription; (2) A dephosphorylated by an unidenti®ed protein tyrosine STAT protein may form a transcriptional complex (PTPase) to form non-phosphorylated with a non-STAT to trigger STAT monomers in the nucleus (Darnell, 1997; transcription through a STAT; or (3) a non-STAT Shuai, 1999). The inactivated STAT monomer is DNA binding element; (4) A STAT and a non-STAT relocated back to the cytoplasm where it can be transcription factor may cooperate to activate tran- reactivated, completing an activation-inactivation scription through binding to clustered independent cycle (Figure 1). The hypothesis that a STAT can DNA binding sites (Figure 2). shuttle between cytoplasm and nucleus was initially derived from results of biochemical fractionation and p48 pulse-chase experiments (Haspel and Darnell, 1999; Haspel et al., 1996), and is clearly supported by The molecular basis of ISGF3 complex formation has recent studies using GFP (green ¯uorescence protein) been extensively studied. Although Stat1, Stat2 and p48 fusion proteins to visualize the tracking of Stat1 in can form a stable complex in the presence of DNA, initial living cells (Keoster and Hauser, 1999). A number of attempts to detect association of Stat1 or Stat2 with p48 proteins have been identi®ed that interact with in solution had failed. Subsequently, in vivo coimmuno- STATs and modulate the activity of STATs at precipitation studies using speci®c antibodies against p48 various steps of the activation-inactivation cycle or Stat2 demonstrated that Stat2 can interact with p48 in (Figure 1). The focus of this article is to review the the absence of DNA (Martinez-Moczygemba et al., roles of these STAT-interacting proteins in STAT 1997). The Stat2-p48 interaction was detectable in cells signaling. with or without IFN-a treatment. But similar coimmu- noprecipitation analysis failed to detect an interaction between Stat1 and p48. These results suggest that either Stat1 does not interact with p48 in solution in vivo,or *Correspondence: K Shuai, Division of Hematology-Oncology, 11- alternatively, the interaction of p48 and Stat1 is weak 934 Factor Bldg, 10833 LeConte Avenue, Los Angeles, California and transient in the absence of the DNA target. 90095-1678. Consistent with the latter hypothesis, the Stat1-p48 STAT-interacting proteins K Shuai 2639

Figure 1 The activation and inactivation cycle of STAT. STAT shuttles between cytoplasm and nucleus. STAT-interacting proteins a€ect various steps of STAT signaling. See text for detailed description

Figure 2 Mechanisms of transcriptional activation by STATs. STATs alone or in cooperation with non-STAT transcription factors activate transcription through a STAT site or a non-STAT site in gene promoters. See text for details interaction was observed when analysed by the yeast al., 1997) (Figure 3). The coiled-coil domain of Stat2 is two-hybrid assay, a highly sensitive method to detect essential for Stat2-p48 interaction. A single protein-protein interactions. In addition, in vitro GST (K161A) in the p48 contact region of Stat1 inhibited pull-down experiments suggested that both Stat1 and the Stat1-p48 interaction as well as the IFN-a-triggered Stat2 can interact with p48, although the Stat2-p48 ISG15 gene induction, whereas this mutation had no interaction is several fold stronger than the Stat1-p48 e€ect on either the Stat1-Stat2 interaction or the IFN-g association (Horvath et al., 1996). response which is independent of p48 (Horvath et al., In vitro reconstitution experiments identi®ed the 1996). These results strongly support the importance of COOH-terminal region of p48 (aa 217 ± 377) is required the Stat1-p48 interaction in IFN-a signaling. for ISGF3 formation (Veals et al., 1992). Detailed domain mapping studies suggest that the NH2-terminal c-Jun and Sp1 region of Stat1 (aa 152 ± 239) or Stat2 (aa 1 ± 324 in Stat2) contacts with the COOH-terminal portion of It has been well established that the transcriptional p48 (Horvath et al., 1996; Martinez-Moczygemba et activation of mammalian genes usually requires the

Oncogene STAT-interacting proteins K Shuai 2640

Figure 3 Regions of STATs involved in binding to non-STAT proteins. The domain structure of STAT is shown. The contact regions of STAT-interacting proteins are underlined. ND: the N domain; CCD: domain; DBD: DNA binding domain; LD: Linker domain; SH2: Src homology 2 domain; TAD: Transcription activation domain

cooperative e€ect of many transcription factors. (Figure 3). Point in the interaction regions Studies on a number of genes have shown that of Stat3 abolished Stat3-c-Jun interaction as well as independent but closely spaced DNA binding sites for their cooperation in IL-6-induced transcriptional STAT and other transcription factors are required for activation of the a2-macroglobulin gene. Further- maximal transcriptional activation. Two well studied more, these mutations did not a€ect the ability of examples involving cooperative interactions between Stat3 to drive noncooperative IL-6-induced transcrip- STAT and Sp1 or STAT and c-Jun are discussed in tion. It is unclear why di€erent regions of c-Jun detail here. were observed to interact with Stat3 by di€erent The intercellular adhesion molecule-1 (ICAM-1) investigators. Nevertheless, these results demonstrate gene is induced by IFN-g. In the of ICAM- the importance of the Stat3-c-Jun interaction in gene 1 gene, contiguous but independent DNA binding sites activation. for Stat1 and Sp1 are present and are shown to be required for the full transcriptional activation of Glucocortical ICAM-1 gene in response to IFN-g. (Look et al., 1995). Similarly, both a Stat3 binding sequence and an The ®nding that receptor (GR) is adjacent Sp1 binding site present in the promoter of associated with Stat5 came from studies on the b- the C/EBPd (CCAAT/enhancer binding protein d) gene casein gene. Upon interaction with its ligand, are shown to be required for the induction of C/EBPd GR binds to speci®c DNA sequence, the glucocorti- by IL-6 (Cantwell et al., 1998). In addition, Stat1 and coid response element (GRE) to activate transcrip- Sp1 association in solution has been observed. These tion (Beato et al., 1995). The optimal expression of studies demonstrate that Stat1 or Stat3 can cooperate b-casein gene is achieved when costimulated with with Sp1 to activate genes. and prolactin (PRL), a growth The involvement of Stat3 and c-Jun cooperation hormone that activates Stat5 (Schmitt-Ney et al., in transcriptional activation has been well documen- 1991). A functional Stat5 binding site and half- ted for a number of genes, including the induction palindromic GREs are present in the promoter of b- of a2-macroglobulin gene by IL-6 (Ito et al., 1989), casein gene. Mutation of the Stat5 binding site in the induction of c-fos by multiple ligands (Robertson the b-casein gene promoter abolishes both glucocor- et al., 1995), the induction of matrix metalloprotei- ticoid- and PRL-induced transcription. In vivo nase 1 gene by (Korzus et al., 1997), coimmunoprecipitation studies detected a physical and the induction of the vasoactive intestinal peptide association between Stat5 with GR (Cella et al., gene by ciliary neurotrophic factor (Lewis et al., 1998; Steocklin et al., 1996). In addition, GR is 1994). The initial identi®cation that c-Jun interacts present in PRL-induced Stat5 DNA-binding complex with Stat3 came from a yeast two-hybrid screen and can enhance Stat5-mediated transcription. It has using the NH2-terminal segment of c-Jun as the bait been suggested that GR may act as a transcriptional (Schaefer et al., 1995, 1997). It was found that co-activator for Stat5 (Steocklin et al., 1996). Stat3b, a variant of Stat3, cooperates with c-Jun to Interestingly, a recent report suggests that cotransfec- activate the transcription of the a2-macroglobulin tion of GR and Stat5 resulted in the enhancement gene. Detailed Stat3-c-Jun interaction studies have of the DNA binding activity and the tyrosine recently been reported (Zhang et al., 1999). The phosphorylation of Stat5 (Wyszomierski et al., COOH-terminal region of c-Jun (aa 105 ± 334) is 1999). It will be of great interest to understand shown to interact with two regions in the NH2- the molecular basis of the observed e€ect of GR on terminal portion of Stat3 (aa 130 ± 358). One region Stat5 activation and to examine if the enhanced is within the coiled-coid domain of Stat3, whereas of Stat5 by GR is solely the other region is located in the DNA binding responsible for the transcriptional synergy exhibited domain but distant from DNA contact sites of Stat3 by Stat5 and GR.

Oncogene STAT-interacting proteins K Shuai 2641 Interaction with p300/CBP and Nmi luciferase reporter assays. An antisense Nmi expression vector can inhibit IL-2-induced Stat5 luciferase p300/CBP reporter in YT cells, while a sense Nmi expression reporter can enhance Stat5-mediated gene activation in STATs, like many signal-responsive transcriptional an IL-2 responsive NIH3T3-derived cell line. Interest- factors, are found to be regulated by coactivators ingly, Nmi was found to augment the association of p300 and CBP (Bhattacharya et al., 1996; Horvai et al., Stat5 with CBP. These results strongly suggest that 1997; Zhang et al., 1996), the histone acetyltransferases Nmi may function to enhance the transcriptional involved in chromatin remodeling (Kadonaga, 1998). activity of Stat5 by augmenting the recruitment of Stat2 was identi®ed as a candidate protein that CBP to Stat5. Furthermore, in vitro GST pull-down interacts with the cysteine-histidine-rich domain 1 analysis indicates that all STATs except Stat2 can (CH1) of p300 in an expression screen. The p300/ interact with Nmi. Consistently, Nmi can enhance CBP interactive region of Stat2 is located at the IFN-g-dependent Stat1-mediated transcription. COOH-terminal portion of Stat2 that functions as a In contrast to the proposed nuclear function of Nmi, transcriptional activation domain (Figure 3). Stat1 has several studies indicate that Nmi is a cytoplasmic protein also been shown to interact with p300/CBP through (Bannasch et al., 1999; Lebrun et al., 1998; Lee et al., two separate regions: the NH2-terminal region of Stat1 1999). The endogenous Nmi or ectopically expressed interacts with the CREB-binding domain of p300/CBP full-length Nmi was found to be localized in the and the COOH-terminal region of Stat1 interacts with cytoplasm, in punctate speckles, in a number of cell the CH3 domain of p300/CBP that is also known to lines. Although these studies can not exclude the bind to adenovirus E1A protein (Figure 3). Because possibility that a small fraction of Nmi is present in the p300/CBP is required for E1A to activate transcription nucleus, it seems likely that Nmi may function mainly in during viral replication and cellular transformation by the cytoplasm. Alternatively, Nmi may shuttle between adenovirus, these results suggest that E1A may cytoplasm and nucleus and may reside in the nucleus interfere with the anti-viral activity of IFNs by transiently due to fast nuclear export. Further experi- competing for binding to p300/CBP. Indeed, wild type ments are needed to solve this discrepancy. E1A, but not mutant E1A defective in binding to p300/ CBP, is able to inhibit IFN-induced STAT-dependent transcription. Thus, p300 and CBP potentiate STAT- mediated transcription by linking STAT to the basal Protein inhibitors of activated STAT transcriptional machinery. The interruption of STAT- p300/CBP interactions by E1A contributes to adeno- Isolation of the PIAS family of proteins virus survival in its natural host. In a yeast two-hybrid screen aimed at the identi®cation of potential regulators of Stat1, a partial cDNA clone Nmi encoding the COOH-terminal region of an unknown protein, which later named as PIAS1 (protein inhibitor Nmi (N- interactor) was ®rst identi®ed through a of activated Stat1), was isolated based on its ability to yeast two-hybrid screen using the COOH-terminus of interact with Stat1b (Liu et al., 1998; Shuai, 1999). In N-Myc that includes the bHLH-Zip (basic helix ± the process of characterizing PIAS1 in Stat1 signaling, loop ± helix/) region as the bait (Bao and a protein named GBP (Gu binding protein) that is Zervos, 1996). In addition to N-Myc, Nmi was shown highly related to PIAS1 was reported to bind to Gu to be able to interact with c-myc and Max (contain a RNA helicase (Valdez et al., 1997). The major bHLH-Zip domain), daughterless (contains a bHLH di€erence between PIAS1 and GBP is that GBP lacks domain), and c-fos (contains a bZip motif) in yeast the NH2-terminal 9 residues present in two-hybrid assays. Interaction of Nmi with N-Myc or PIAS1 (if the ®rst methione residue present in GBP c-myc was also detected when overexpressed in 293 protein sequence is counted as the ®rst amino acid of cells. The human Nmi is a protein of 307 amino acids. GBP). Subsequent studies suggest that the NH2- Sequence analysis suggests that the NH2-terminus of terminal 9 amino acid residues of PIAS1 is essential Nmi contains a coiled-coil heptad repeat, whereas its for PIAS1 function (Liu and Shuai, unpublished data). COOH-terminal region shows homology to an inter- It is possible that GBP represents an incomplete feron-induced leucine zipper protein, IP35 (Bao and version of PIAS1. Zervos, 1996). Subsequent studies indicate that Nmi Four additional mammalian proteins related to expression is also interferon-inducible (Lebrun et al., PIAS1 were identi®ed through database search and 1998). cDNA library screen (Chung et al., 1997; Liu et al., In an independent yeast two-hybrid screen using the 1998): PIAS3, PIASy, PIASxa, and PIASxb. Sequence coiled-coil region (aa 182 ± 337) of Stat5b as the bait, analysis suggests that PIAS3 and PIASy are novel Nmi was identi®ed as a Stat5-interacting protein (Zhu proteins. PIASxa and PIASxb are identical, except they et al., 1999). The expression of Nmi is inducible by IL- have distinct COOH-terminal regions, and are prob- 2 in peripheral blood lymphocytes (PBL). Nmi and ably the products of di€erentially spliced messages Stat5 association was detected in PBL cells with or from the same gene. An incomplete portion of PIASxb without IL-2 treatment by in vivo coimmunoprecipita- (aa 134 ± 621), named as Miz1, was previously shown tion analysis. Detailed mapping studies indicate that to interact with a DNA binding protein two regions of Nmi (aa 57 ± 99 and aa 143 ± 202) are Msx2 (Wu et al., 1997). able to interact with the amino acid 232 ± 321 region of EST clones encoding proteins that have sequence Stat5b (Figure 3). The importance of Nmi in IL-2- homology with PIAS are also found in Drosophila, C. activated Stat5 signaling was suggested by transient elegans, yeast, and plants. Zimp, the Drosophila PIAS

Oncogene STAT-interacting proteins K Shuai 2642 homologue, is found to be essential in Drosophila 1997). Phosphorylation at Ser727 in the Stat1 TAD development (Mohr and Boswell, 1999). domain is required for the maximal transcriptional activity of Stat1 (Wen et al., 1995). Using GST-TAD in protein-trapping assays, a group of nuclear proteins The structure of PIAS speci®cally associated with TAD were identi®ed (Zhang Sequence analysis reveals several interesting structure et al., 1998). One of these nuclear proteins has been features of the PIAS family of proteins (Figure 4). A identi®ed as MCM5, a member of the minichromo- conserved putative LXXLL signature motif is present some maintenance (MCM) family of proteins known to in the NH2-terminal region of all PIAS proteins. The be involved in DNA replication (Dutta and Bell, 1997). LXXLL motif, also known as NR () Interestingly, the interaction of MCM5 with the Stat1 box, is found to be present in a number of nuclear TAD domain is Ser727-dependent. The substitution of receptor coregulators (Heery et al., 1997; Torchia et al., Ser727 with Ala inhibited the association. Most 1997, 1998). The LXXLL signature motif mediates importantly, MCM5 can enhance Stat1-mediated ligand-dependent coactivator-nuclear receptor interac- reporter gene assays. These results strongly suggest tions or coactivator-coactivator interactions. A puta- the importance of MCM5-Stat1 interaction in regulat- tive zinc binding motif is present near the middle of ing the transcriptional activity of Stat1. MCM proteins PIAS proteins. The COOH-terminal regions of PIAS have helicase activity and are known to associate with proteins are least conserved and consist of a highly initiation and elongation complexes during DNA acidic region and a serine/threonine rich region. replication (Dutta and Bell, 1997). Thus, it seems Interestingly, unlike other PIAS proteins, PIASy lacks likely that MCM5 may play a dual role in DNA the COOH-terminal serine/threonine rich domain. The replication and Stat1-mediated gene transcription. roles of these domains in PIAS function remain to be determined. Interaction with NPI-1 PIAS1 and PIAS3 as inhibitors of STAT signaling One important feature of STAT activation is the rapid In vivo coimmunoprecipitation studies using speci®c nuclear translocation of STAT upon ligand stimula- antibodies against PIAS1 and PIAS3 indicated that tion. The formation of the nuclear pore-targeting PIAS1 and PIAS3 interacts with Stat1 and Stat3, complex (PTAC) in the cytoplasm is required for the respectively (Chung et al., 1997; Liu et al., 1998). Most translocation of nuclear localization signal (NSL)- interestingly, the PIAS-STAT interaction is dependent containing proteins (Imamoto et al., 1995). The PTAC on stimulation. In vitro DNA binding analysis complex consists of a karyophilic protein and two suggests that PIAS can block the DNA binding activity essential a and b subunit families (Sekimoto and of STAT. Consistently, PIAS can inhibit STAT- Yoneda, 1998). Studies on the nuclear translocation mediated gene activation. Thus, these results suggest of Stat1 have identi®ed a physical association of Stat1 that PIAS1 and PIAS3 are inhibitors of Stat1 and with NPI-1, a member of the a subunit family Stat3. (Sekimoto et al., 1997). The binding of Stat1 to NPI- The observation that PIAS is associated with STAT 1 allows an indirect association of Stat1 with the b only in cells stimulated with suggests that the subunit. Interestingly, in vitro binding assays suggest PIAS-STAT interaction may be regulated. Deletional that NPI-1 can trap Stat1 only when extracts from cells analysis indicates that a region immediately after the stimulated with IFN-g were used, suggesting the zinc binding motif of PIAS1 interacts with the NH2- interaction of NPI-1 and Stat1 may be regulated. Stat1 terminal portion (aa 1 ± 191) of Stat1 (Figure 3). A contacts the COOH-terminal portion of NPI-1, a mutant PIAS1 lacking the Stat1 interactive domain is region distinct from the SV40 large T antigen NLS- defective in suppressing Stat1-mediated gene activation binding domain. The NPI-1 interacting region of Stat1 (Liao et al., 2000, submitted). has not been identi®ed. Microinjection of antibodies against NPI-1 or b subunit e€ectively blocked the nuclear translocation of Stat1. These results demon- Interaction with MCM5 strate the importance of NPI-1 and Stat1 interaction in the regulation of Stat1 nuclear translocation. The transcriptional activation domain (TAD) of Stat1 is localized at the COOH-terminal region (Darnell, Interaction with ERKs

Several STATs can be serine phosphorylated through ERK-dependent and -independent mechanisms (see review by T Decker in this issue). Growth hormone (GH) can induce the activation of Stat5 and ERK pathways (Pircher et al., 1997; Winston and Bertics, 1992). It has been shown that Stat5a which contains a putative ERK phosphorylation site (RLSP) is physi- cally associated with ERK1/2 (Pircher et al., 1999). The ERK interaction region of Stat5a is mapped at the Figure 4 The domain structure of PIAS proteins. ZBD: zinc binding domain; AD: acidic domain; S/T: serine-threonine rich COOH-terminal portion of Stat5a (Figure 3). It is region. S/T domain is not present in PIASy. Only the conserved proposed that the association of Stat5a with ERK LXXLL signature motif is indicated results in the phosphorylation of Stat5a at Ser780 by

Oncogene STAT-interacting proteins K Shuai 2643 ERK. Physical interactions of STATs with STAT Third, crosstalk between STAT and other cellular serine are clearly important for the modulation signaling pathways can be achieved by the action of of STAT activity. certain STAT-associated proteins. For example, an Phosphorylation of STATs by non-JAK tyrosine interaction between ERK and STAT may connect the kinases have also been well documented. This topic is MAPK and STAT signaling pathways. Members of reviewed in this issue by R Jove. the PIAS family have also been suggested to function in androgen signaling (Moilanen et al., 1999; Tan et al., 2000). Conclusion The NH2- and the COOH-terminal regions of STATs are largely responsible for mediating protein- A growing number of STAT-interacting proteins have protein interactions (Figure 3). Since the binding sites been described. These proteins have distinct functions for many STAT-interacting proteins are overlap, it in the modulation of STAT signaling. What is the seems likely that multiple STAT complexes may exist signi®cance of STAT-interacting proteins in STAT in vivo. Recent studies have suggested the presence of signaling? First, STAT-interacting proteins can func- high molecular mass Stat3-containing complexes in the tion as positive or negative factors to regulate various (Ndubuisi et al., 1999). The chaperone GRP58/ basic aspects of STAT signaling. For example, ER-60/ERp57 was identi®ed as a component of the STAT-interacting kinases/ directly a€ect large complex. The involvement of sca€old proteins in the activation/inactivation process of STATs; p300/ the formation of large signaling complexes, the so- CBP links STATs to the basal transcriptional called `signalsome', has been suggested in several machinery; the relative level of STATs and PIAS signaling pathways (Cohen et al., 1998; Whitmarsh et proteins for a given cell type may a€ect the overall al., 1998). strength of STAT signaling. Second, since STATs The physiological importance of many STAT- play important roles in numerous fundamental interacting proteins as well as the molecular basis of cellular processes, the involvement of other non- their interactions with STATs remain to be determined. STAT transcription factors certainly contributes to Since the dysregulation of STAT activity appears to the speci®city and complexity of STAT-mediated gene contribute to certain human diseases and , future activation. Indeed, many of the STAT-interacting studies on STAT-interacting proteins may enhance our proteins identi®ed so far are directly involved in the ability to design rational therapeutic strategies. regulation of the transcriptional activity of STATs.

References

Bannasch D, Weis I and Schwab M. (1999). Oncogene, 18, Heery DM, Kalkhoven E, Hoare S and Parker MG. (1997). 6810 ± 6817. Nature, 387, 733 ± 736. Bao J and Zervos AS. (1996). Oncogene, 12, 2171 ± 2176. Horvai AE, Xu L, Korzus E, Brard G, Kalafus D, Mullen Beato M, Herrlich P and Scheutz G. (1995). Cell, 83, 851 ± TM, Rose DW, Rosenfeld MG and Glass CK. (1997). 857. Proc. Natl. Acad. Sci. USA, 94, 1074 ± 1079. Becker S. Groner B and Muller CW. (1998). Nature, 394, Horvath CM, Stark GR, Kerr IM and Darnell Jr JE. (1996). 145 ± 151. Mol. Cell. Biol., 16, 6957 ± 6964. Bhattacharya S, Eckner R, Grossman S, Oldread E, Arany Imamoto N, Shimamoto T, Kose S, Takao T, Tachibana T, Z, D'Andrea A and Livingston DM. (1996). Nature, 383, Matsubae M, Sekimoto T, Shimonishi Y and Yoneda Y. 344 ± 347. (1995). FEBS Lett., 368, 415 ± 419. Cantwell CA, Sterneck E and Johnson PF. (1998). Mol. Cell. Ito T, Tanahashi H, Misumi Y and Sakaki Y. (1989). Nucleic Biol., 18, 2108 ± 2117. Acids Res, 17, 9425 ± 9435. Cella N, Groner B and Hynes NE. (1998). Mol. Cell. Biol., Kadonaga JT. (1998). Cell, 92, 307 ± 313. 18, 1783 ± 1792. Keoster M and Hauser H. (1999). Eur. J. Biochem., 260, Chen X, Vinkemeier U, Zhao Y, Jeruzalmi D, Darnell Jr JE 137 ± 144. and Kuriyan J. (1998). Cell, 93, 827 ± 839. Korzus E, Nagase H, Rydell R and Travis J. (1997). J. Biol. Chung CD, Liao J, Liu B, Rao X, Jay P, Berta P and Shuai Chem., 272, 1188 ± 1196. K. (1997). Science, 278, 1803 ± 1805. Lebrun SJ, Shpall RL and Naumovski L. (1998). J. Inter. Cohen L, Henzel WJ and Baeuerle PA. (1998). Nature, 395, Cytokine Res., 18, 767 ± 771. 292 ± 296. Lee ND, Chen J, Shpall RL and Naumovski L. (1999). J. Darnell Jr JE. (1997). Science, 277, 1630 ± 1635. Inter. Cytokine Res., 19, 1245 ± 1252. Darnell Jr JE, Kerr IM and Stark GR. (1994). Science, 264, Levy DE. (1997). Cytokine Rev., 8, 81 ± 90. 1415 ± 1421. Lewis SE, Rao MS, Symes AJ, Dauer WT, Fink JS, Landis Decker T, Lew DJ, Mirkovitch J and Darnell Jr JE. (1991). SC and Hyman SE. (1994). J. Neurochem., 63, 429 ± 438. Embo J., 10, 927 ± 932. LiuB,LiaoJ,RaoX,KushnerSA,ChungCD,ChangDD Dutta A and Bell SP. (1997). Ann. Rev. Cell Dev. Biol., 13, and Shuai K. (1998). Proc. Natl. Acad. Sci. USA, 95, 293 ± 332. 10626 ± 10631. Fu XY, Kessler DS, Veals SA, Levy DE and Darnell Jr JE. Look DC, Pelletier MR, Tidwell RM, Roswit WT and (1990). Proc. Natl. Acad. Sci. USA, 87, 8555 ± 8559. Holtzman MJ. (1995). J. Biol. Chem., 270, 30264 ± 30267. Fu XY, Schindler C, Improta T, Aebersold R and Darnell Jr Martinez-Moczygemba M, Gutch MJ, French DL and Reich JE. (1992). Proc. Natl. Acad. Sci. USA, 89, 7840 ± 7843. NC. (1997). J. Biol. Chem., 272, 20070 ± 20076. Haspel RL and Darnell Jr JE. (1999). Proc. Natl. Acad. Sci. Mohr SE and Boswell RE. (1999). Gene, 229, 109 ± 116. USA, 96, 10188 ± 10193. Moilanen AM, Karvonen U, Poukka H, Yan W, Toppari J, Haspel RL, Salditt-Georgie€ M and Darnell Jr JE. (1996). Jeanne OA and Palvimo JJ. (1999). J. Biol. Chem., 274, EMBO J., 15, 6262 ± 6268. 3700 ± 3704.

Oncogene STAT-interacting proteins K Shuai 2644 Ndubuisi MI, Guo GG, Fried VA, Etlinger JD and Sehgal Torchia J, Glass C and Rosenfeld MG. (1998). Curr. Opin. PB. (1999). J. Biol. Chem., 274, 25499 ± 25509. Cell. Biol., 10, 373 ± 383. Pircher TJ, Flores-Morales A, Mui AL, Saltiel AR, Norstedt Torchia J, Rose DW, Inostroza J, Kamei Y, Westin S, Glass G, Gustafsson JA and Haldosen LA. (1997). Mol. Cell CK and Rosenfeld MG. (1997). Nature, 387, 677 ± 684. Endocrinol., 133, 169 ± 176. Valdez BC, Henning D, Perlaky L, Busch RK and Busch H. Pircher TJ, Petersen H, Gustafsson JA and Haldosaen LA. (1997). Biochem. Biophy. Res. Comm., 234, 335 ± 340. (1999). Mol. Endocrinol., 13, 555 ± 565. Veals SA, Schindler C, Leonard D, Fu XY, Aebersold R, Robertson LM, Kerppola TK, Vendrell M, Luk D, Smeyne Darnell Jr JE and Levy DE. (1992). Mol. Cell. Biol., 12, RJ, Bocchiaro C, Morgan JI and Curran T. (1995). 3315 ± 3324. Neuron, 14, 241 ± 252. Wen Z, Zhong Z and Darnell Jr JE. (1995). Cell, 82, 241 ± Schaefer TS, Sanders LK and Nathans D. (1995). Proc. Natl. 250. Acad. Sci. USA, 92, 9097 ± 9101. Whitmarsh AJ, Cavanagh J, Tournier C, Yasuda J and Davis Schaefer TS, Sanders LK, Park OK and Nathans D. (1997). RJ. (1998). Science, 281, 1671 ± 1674. Mol. Cell. Biol., 17, 5307 ± 5316. Winston LA and Bertics PJ. (1992). J. Biol. Chem., 267, Schindler C, Fu XY, Improta T, Aebersold R and Darnell Jr 4747 ± 4751. JE. (1992). Proc. Natl. Acad. Sci. USA, 89, 7836 ± 7839. WuL,WuH,SangiorgiF,WuN,BellJR,LyonsGEand Schmitt-Ney M, Doppler W, Ball RK and Groner B. (1991). Maxson R. (1997). Mech. Dev., 65, 3±17. Mol. Cell. Biol., 11, 3745 ± 3755. Wyszomierski SL, Yeh J and Rosen JM. (1999). Mol. Sekimoto T, Imamoto N, Nakajima K, Hirano T and Endocrinol., 13, 330 ± 343. Yoneda Y. (1997). EMBO J., 16, 7067 ± 7077. Zhang JJ, Vinkemeier U, Gu W, Chakravarti D, Horvath Sekimoto T and Yoneda Y. (1998). Cytokine Growth Factor CM and Darnell Jr JE. (1996). Proc. Natl. Acad. Sci. USA, Rev., 9, 205 ± 211. 93, 15092 ± 15096. Shuai K. (1999). Prog. Biophy. Mol. Biol., 71, 405 ± 422. Zhang JJ, Zhao Y, Chait BT, Lathem WW, Ritzi M, Shuai K, Horvath CM, Huang LH, Qureshi SA, Cowburn D Knippers R and Darnell Jr JE. (1998). EMBO J., 17, and Darnell Jr JE. (1994). Cell, 76, 821 ± 828. 6963 ± 6971. Shuai K, Schindler C, Prezioso VR and Darnell Jr JE. (1992). Zhang X, Wrzeszczynska MH, Horvath CM and Darnell Jr Science, 258, 1808 ± 1812. JE. (1999). Mol. Cell. Biol., 19, 7138 ± 7146. Steocklin E, Wissler M, Gouilleux F and Groner B. (1996). Zhu M, John S, Berg M and Leonard WJ. (1999). Cell, 96, Nature, 383, 726 ± 728. 121 ± 130. Tan J, Hall SH, Hamil KG, Grossman G, Petrusz P, Liao J, Shuai K and French FS. (2000). Mol. Endocrin., 14, 14 ± 26.

Oncogene