Stat2 is a transcriptional activator that requires sequence-specific contacts provided by stat1 and p48 for stable interaction with DNA.

Transcriptional responses to interferon (IFN) are mediated by tyrosine phosphorylation and nuclear translocation of transcription factors of the signal transducer and activator of transcription (Stat) family. The Stat1 protein is required for all transcriptional responses to IFN (both type I and type II). Responses to type I IFN (α and β) also require Stat2 and the IFN regulatory factor family protein p48, which form a heterotrimeric transcription complex with Stat1 termed ISGF3. Stat1 homodimers formed in response to IFN-γ treatment can also interact with p48 and function as transcriptional activators. We now show that Stat2 is capable of forming a stable homodimer that interacts with p48, can be recruited to DNA, and can activate transcription, raising a question of why Stat1 is required. Analysis of the transcriptional competence, affinity, and specificity of Stat2-p48 complexes compared with other Stat protein-containing transcription factor complexes suggests distinct roles for each component. Although Stat2 is a potent transactivator, it does not interact stably with DNA in complex with p48 alone. Adding Stat1 increases the affinity and alters the sequence selectivity of p48-DNA interactions by contacting a half-site of its palindromic recognition motif adjacent to a p48 interaction sequence. Thus, ISGF3 assembly involves p48 functioning as an adaptor protein to recruit Stat1 and Stat2 to an IFN-α-stimulated response element, Stat2 contributes a potent transactivation domain but is unable to directly contact DNA, while Stat1 stabilizes the heteromeric complex by contacting DNA directly.


Transcriptional responses to interferon (IFN) are mediated by tyrosine phosphorylation and nuclear translocation of transcription factors of the signal transducer and activator of transcription (Stat) family. The Stat1
protein is required for all transcriptional responses to IFN (both type I and type II). Responses to type I IFN (␣ and ␤) also require Stat2 and the IFN regulatory factor family protein p48, which form a heterotrimeric transcription complex with Stat1 termed ISGF3. Stat1 homodimers formed in response to IFN-␥ treatment can also interact with p48 and function as transcriptional activators. We now show that Stat2 is capable of forming a stable homodimer that interacts with p48, can be recruited to DNA, and can activate transcription, raising a question of why Stat1 is required. Analysis of the transcriptional competence, affinity, and specificity of Stat2-p48 complexes compared with other Stat proteincontaining transcription factor complexes suggests distinct roles for each component. Although Stat2 is a potent transactivator, it does not interact stably with DNA in complex with p48 alone. Adding Stat1 increases the affinity and alters the sequence selectivity of p48-DNA interactions by contacting a half-site of its palindromic recognition motif adjacent to a p48 interaction sequence. Thus, ISGF3 assembly involves p48 functioning as an adaptor protein to recruit Stat1 and Stat2 to an IFN-␣-stimulated response element, Stat2 contributes a potent transactivation domain but is unable to directly contact DNA, while Stat1 stabilizes the heteromeric complex by contacting DNA directly.
Interferon-␣ (IFN-␣) 1 binding to its receptor leads to the activation of a multiprotein DNA-binding complex called IFNstimulated gene factor 3 (ISGF3). Purification of its component proteins led to the identification (1, 2) and cloning of the first two members of a novel family of signal transducers and activators of transcription (Stats), Stat1 (3) and Stat2 (4), and the DNA-binding protein ISGF3␥ p48 (5). The Stat1 and Stat2 proteins are present in the cytoplasm of untreated cells; upon stimulation with IFN-␣, they become rapidly activated by tyrosine phosphorylation at a single site (6) catalyzed by receptor associated Jak (Janus) kinases (7)(8)(9)(10). This activation subsequently allows them to assemble into stable homo-and heteromeric complexes through specific SH2 domain-phosphotyrosyl interactions (11) and translocate to the nucleus where they bind specific enhancer elements in the promoters of IFN-inducible genes. Interestingly, Stat1 (but not Stat2) is also activated by tyrosine phosphorylation in response to a number of other cytokines (12)(13)(14), including IFN-␥ (15; for review, see Refs. 16 and 17).
ISGF3␥ p48 (p48) is unrelated to Stat proteins but is a member of the IFN regulatory factor family (18) of DNA-binding proteins that recognize the positive regulatory domain I (PRDI) of the IFN-␤ gene (19,20) and the IFN-␣-stimulated response element (ISRE) of IFN-␣-stimulated genes (5,21). p48 binds the ISRE more efficiently than it binds PRDI (22), and it is required for many transcriptional responses of IFN-␣ target genes containing ISRE enhancer sequences (23,24). By itself p48 has a low affinity for the ISRE sequence and no transcriptional activity, but complexed with the Stat1-Stat2 heterodimer that is formed upon IFN-␣ treatment, it forms a significantly more stable protein-DNA complex (5,21,22) capable of transactivating gene expression. Sequence specific recognition by ISGF3 has been shown to involve multiple protein-DNA contacts by both the Stat proteins and p48 (22,25).
Phosphorylated Stat1 homodimers are formed following treatment with either IFN-␣ or IFN-␥ (26). As a dimer (known as gamma-activated factor, or GAF) Stat1 binds directly to an enhancer element called gamma-activated site, or GAS (27) and activates transcription of target genes containing this site in their promoters. A major difference between transcription factor binding at the GAS and the ISRE sequences is the presence of a p48 recognition element in the ISRE which allows p48 binding and thus Stat protein recruitment to this distinct site. As might be predicted from the sequence similarity between Stat1 and Stat2, it was found that Stat1 homodimers, in addition to Stat1-Stat2 heterodimers, also interact with p48 and bind the ISRE in response to IFN-␥ treatment (28). This Stat1-p48 complex binds the ISRE with a recognition specificity similar to ISGF3. Together the three distinct multimeric complexes activated by IFN-␣ and IFN-␥ (ISGF3, GAF, and Stat1-p48) contribute to the characteristic patterns of gene transcription induced by these cytokines (18).
Mutagenesis of Stat2 suggested that this protein contributes the transactivation function of ISGF3 (29), prompting us to question the role for Stat1 in ISGF3. We found that Stat2, like * This work was supported by Grant AI28900 from the National Institutes of Health and by grants from the Pew Scholars Program and the Arthritis Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
other Stat proteins, is capable of forming a stable homodimer when phosphorylated in response to IFN-␣. In conjunction with p48, Stat2 homodimers are capable of activating transcription from ISRE-containing genes in the absence of Stat1. However, Stat2-p48-DNA complexes are very unstable, only forming under conditions where these proteins are abundant. The increased affinity of ISGF3 over Stat2-p48 appears to be caused by additional sequence-specific DNA contacts contributed by Stat1 adjacent to the primary p48 binding sequence. Stat2, in contrast, is incapable of directly contacting the ISRE sequence.
Transfection Procedure and Luciferase Assay-U3A, U3A-Stat2, U3A-p48, or U3A-Stat2-p48 cells were seeded at 5 ϫ 10 5 cells/6-cm dish. The next day, cells were transfected with the ISG54-LUC reporter gene construct (1 g) alone or together with expression plasmids directing the synthesis of Stat1, Stat2, or p48, using the calcium phosphate precipitation method (42). After overnight incubation with the precipitate, incubation was continued in the absence or presence of human IFN-␣-2a (500 units/ml) for an additional 6 h. Luciferase activity in cell extracts was determined according to the Luciferase Assay System (Promega).
Electrophoretic Mobility Shift Assays (EMSAs)-Cytoplasmic and nuclear protein extracts were prepared from transiently transfected 293T cells, or transiently or stably transfected U3A cells, and assayed for ISRE binding activity using the high affinity ISG15 ISRE sequence, as described previously (22). Cells in 6-or 10-cm dishes were transfected with expression plasmids directing the production of p48, Stat1, Stat2, Stat2-HA, ⌬Stat2, or Stat2(F) (2.5 g). Ligand-independent activation of the different Stat2 proteins was achieved by cotransfection with 0.4 g of an expression plasmid for the Jak1 tyrosine kinase.
Dissociation Rate Determination-Dissociation rates of protein-DNA complexes were determined essentially as described (1), using cytoplasmic extracts containing DNA-binding complexes composed of p48, Stat2-p48, or Stat2-Stat1-p48 (ISGF3). The approximate half-life of each complex was determined by quantitative PhosphorImager analysis.

RESULTS
A Stat2-p48 Protein Complex Binds the ISRE-In response to IFN-␣, tyrosine phosphorylated Stat1 and Stat2 multimerize to form either Stat1 homodimers that bind GAS sequences through an intrinsic DNA binding domain (46) or Stat1-Stat2 heterodimers that bind ISRE sequences through the DNA binding domain of p48 (22). To explore the possibility that Stat2 may homodimerize in a manner analogous to phosphorylated Stat1, we first examined the ability of Stat2 to bind ISRE sequences in the presence of p48 (Fig. 1A). Tyrosine phosphorylated Stat2 and p48 in extracts from transfected 293T cells were assayed for protein-DNA interaction by gel mobility shift using an ISRE probe. Cell extracts containing activated Stat2 (lane 1) or p48 (lane 2) displayed no ISGF3-like ISRE binding activity. However, mixing protein extracts containing activated Stat2 with extracts containing p48 produced a slowly migrating complex (lane 3). This complex bound antibodies specific for p48 (lane 4) or Stat2 (lane 5), whereas antibodies against Stat1 (lane 6) or Stat3 (lane 7) had no effect, showing that endogenous Stat1 protein did not participate in formation of this complex. The sequence recognition specificity of the complex was tested by competition assay. A 20-fold molar excess of unlabeled ISRE abrogated all binding to the probe (lane 8). In contrast, unlabeled PRDI (an ISRE-related sequence oligonucleotide; lane 9) or GAS (lane 10) competed poorly or failed to compete. These results indicated that Stat2 binds DNA in the presence of p48 with a specificity similar to ISGF3.
Phosphorylated Stat2 Can Form Homodimers-Formation of an ISGF3-like complex composed of Stat2 and p48 suggested that Stat2 might homodimerize, analogous to Stat1. This possibility was tested by coexpressing full-length and truncated, recombinant Stat2 (Fig. 1B). Full-length Stat2 tagged with an HA epitope (Stat2-HA) and a shorter, untagged form (⌬Stat2) that lacks the carboxyl-terminal 100 amino acids were expressed in 293T cells. Extracts containing Stat2-HA (lane 1), ⌬Stat2 (lane 2), or Stat2-HA and ⌬Stat2 together (lane 3) showed equal expression levels for the different proteins. A smaller protein, apparently a proteolytic product of recombinant Stat2-HA lacking the carboxyl terminus and epitope tag, was also detected (lanes 1 and 3). Endogenous Stat2, which is expressed in 293T cells at much lower levels than recombinant protein, was not detected at this exposure. When these extracts were precipitated with a monoclonal antibody against the HA epitope followed by Western blotting using the Stat2 antibody (lanes 4 -6), untagged ⌬Stat2 was recovered from extracts of cells where it was coexpressed with Stat2-HA (lane 6), but not from extracts containing Stat2-HA (lane 4) or ⌬Stat2 (lane 5) alone. Efficient interaction between Stat2-HA and ⌬Stat2 required tyrosine phosphorylation, as indicated by a need for Jak1 kinase and by a lack of interaction with a Stat2 mutant in which tyrosine 690 was replaced by phenylalanine (data not shown).
The same cell extracts were also assayed by gel shift (Fig.  1C). Extracts containing phosphorylated Stat2-HA, mixed with extracts from p48-transfected cells, displayed a slowly migrating complex (lane These results confirmed that the slower mobility complex contained Stat2-HA homodimers, the faster complex contained ⌬Stat2 homodimers, and the intermediate complex consisted of Stat2-HA-⌬Stat2 heterodimers. Overall, these data demonstrated that activated Stat2 formed homodimers that bind the ISRE when complexed with p48. Formation of the Stat2-⌬Stat2 gel shift complexes, like the interactions detected by immunoprecipitation (Fig. 1B), depended on tyrosine phosphorylation of Stat2. Extracts containing a mutant Stat2 in which tyrosine 690 was changed to phenylalanine (lanes 10 and 11) no longer displayed ISRE binding activity when mixed with p48-containing extracts. Similarly, extracts from cells coexpressing Stat2(F) and ⌬Stat2 only displayed the faster migrating complex (lane 11) consisting of ⌬Stat2 homodimers. Formation of all of these complexes required the presence of an active tyrosine kinase (not shown).

FIG. 2. IFN-␣ activates a Stat2-p48 protein complex that binds the ISRE and activates transcription in Stat1-deficient U3A cells.
Panel A, cytoplasmic extracts of mock-, Stat2-, p48-, or Stat2-p48-transfected U3A cells, treated with or without IFN-␣ for 1 h, were analyzed by EMSA using a labeled ISRE probe. Antibodies specific for p48, Stat2, Stat1, or Stat3 were added in lanes 11-14 at a final dilution of 1:200, as indicated. Panel B, U3A cells transfected with ISG54-LUC reporter alone or together with expression plasmids directing the synthesis of p48, Stat2, or ⌬Stat2, as indicated, were analyzed for luciferase activity. Cells were treated with or without IFN-␣ for 6 h, as indicated.
(lane 8). Antibody reactivity identified the composition of this complex. The slow mobility complex as well as the p48 complex were supershifted by antibodies specific for p48 (lane 11). Likewise, the slowly migrating complex, but not the p48 complex, was disrupted by antibodies against Stat2 (lane 13). In contrast, antibodies against Stat1 (lane 12) or Stat3 (lane 14) had no effect on either complex, indicating that a Stat2-p48-DNA complex formed in response to IFN-␣.
The transcriptional potency of the Stat2-p48 complex was tested on an ISRE-reporter construct. U3A cells were transfected with the IFN-␣-responsive reporter construct ISG54-LUC alone or together with expression plasmids directing the synthesis of Stat2 or p48 (Fig. 2B). U3A cells failed to express ISG54-LUC in response to IFN-␣, nor did U3A cells transiently transfected with plasmids for Stat2 or p48 alone show a significant response to IFN-␣. However, cotransfection of all three plasmids (ISG54-LUC, Stat2, and p48) resulted in a 5-fold induction of luciferase activity in response to IFN-␣, a response that depended on the presence of the carboxyl-terminal transactivation domain of Stat2. We conclude that the Stat2-p48 complex can activate transcription in response to IFN-␣ provided that sufficient quantities of Stat2 and p48 proteins are present. The level of transactivation was similar to that observed following transfection of U3A cells with Stat1 cDNA (28).
In contrast, U3A cells transfected with either construct alone failed to form this complex (lanes 3-6). Whereas U3A cells expressed only minute amounts of endogenous Stat2 and p48 and these proteins failed to accumulate in response to IFN-␣ treatment, U3A-Stat2-p48 cells expressed levels of these proteins approximately equal to those observed in U3A-Stat1 cells treated with IFN-␣ for 24 h (Fig. 3B). Expression of the endogenous ISG54 gene was induced in response to IFN-␣ treatment in U3A-Stat2-p48 cells, but the level of expression was approximately 30-fold lower, as judged by PhosphorImager analysis, than in IFN-␣-treated U3A-Stat1 cells (Fig. 3C, compare lanes  2 and 4).
Similarly low levels of induction were observed for the endogenous Stat2 gene (Fig. 3D). Polymerase chain reaction primers were designed which allowed expression of endogenous Stat2 mRNA to be distinguished from expression of the transfected gene. Expression of endogenous Stat2 was 6-fold inducible in U3A-Stat1 cells in response to IFN-␣, whereas it was induced only approximately 2-fold in U3A-Stat2-p48 cells (com- pare lanes 2 and 4). Thus, although Stat2 is a potent transactivator when overexpressed by transient transfection where small numbers of transfected cells probably express high levels of Stat2 and p48, the Stat2-p48 complex is a poor transactivator when expressed at physiological levels.
Stat2-p48 Displays Lower DNA Binding Affinity and Different Sequence Specificity Compared with ISGF3-To explore possible reasons for the poor transcriptional activity observed for Stat2-p48 complexes, the stability of the Stat2-p48 and ISGF3 protein-DNA complexes was determined from dissociation rate measurements (22). Stat2-p48, ISGF3, and p48 complexes bound to DNA were challenged with a 500-fold molar excess of unlabeled ISRE oligonucleotide. At serial time points after addition of competitor, aliquots were removed and loaded directly onto a running polyacrylamide gel (Fig. 4A). Quantitation of the decay of each complex over time by PhosphorImager analysis indicated that dissociation of the ISGF3-ISRE complex was slow, displaying a half-life between 10 and 30 min. In contrast, the Stat2-p48-ISRE interaction was much less stable, with a half-life of Ͻ1 min, similar to the short half-life observed for p48 bound to the ISRE alone (22). These results suggest that the presence of Stat1 greatly increases the low intrinsic affinity of p48 for DNA, whereas Stat2 homodimers are unable to stabilize this interaction. This difference in affinity of Ͼ20-fold is consistent with the approximately 30-fold lower levels of induction of endogenous ISG54 gene expression in Stat2-p48 cells (Fig. 3C).
The increased affinity of ISGF3 for DNA over that of p48 alone correlates with an altered sequence specificity (22). p48 binds a 9-nucleotide core element, whereas ISGF3 makes additional base-specific contacts extending beyond the core sequence. The contribution of Stat1 and Stat2 to this altered binding specificity was determined by gel shift using oligonucleotide probes with altered ISRE sequences (Fig. 4B). Complex binding to wild type sequences was compared with binding to oligonucleotides containing either a C89T or C101T mutation. These mutations were chosen to disrupt two potential GAS half-sites (TTC) flanking the core ISRE element that could be contact points for Stat proteins (47). As shown previously (22), ISGF3 failed to bind the T89 oligonucleotide (lanes 1-3), whereas Stat2-p48 was capable of binding all three sequence elements (compare lanes 4 -6). This binding specificity is similar to that defined for p48 alone (22), indicating that the higher affinity and altered recognition specificity of ISGF3 result from additional protein/nucleotide contacts contributed by Stat1 but not by Stat2. Additional evidence for this idea came from examining ISRE binding characteristics of Stat1-p48 complexes that form in response to IFN-␥ (28). Both the C89T and the C101T mutations prevented Stat1-p48 complexes from binding DNA (lanes 7-9), indicating that Stat1 either directly or indirectly contacts DNA to alter the stability and specificity of p48 binding.

DISCUSSION
Both IFN-␣ and IFN-␥ activate transcription by inducing the tyrosine phosphorylation of Stats. Once phosphorylated, these proteins can form a number of distinct transcription factor complexes. These include Stat1 homodimers that bind GAS sequences in response to either IFN type, Stat1 homodimers that bind ISRE sites in conjunction with p48, predominantly in response to IFN-␥, and Stat1-Stat2 heterodimers that bind the ISRE in conjunction with p48 (ISGF3). Since Stat2 is only phosphorylated in response to IFN-␣, this complex only forms in IFN-␣-treated cells. We now show that another p48-containing complex can form in IFN-␣-treated cells, composed of Stat2 homodimers without Stat1. However, this complex displayed only limited affinity for DNA.
Analysis of the different Stat complexes capable of interacting with p48 suggests a model for transcription factor assembly. The heteromeric ISGF3 complex composed of Stat1, Stat2 and p48 binds a composite DNA sequence defined by a p48 recognition element plus a flanking sequence not contacted by p48 directly. This flanking sequence resembles a half-site of the GAS element recognized by Stat1 homodimers, which can be considered a spaced palindrome of two TTC half-sites (47). The capability to bind spaced palindromic half-sites is inherent in Stat1 dimers, presumably due to a DNA binding domain created by juxtaposition of protein regions between amino acids 400 and 500 (46,48). Therefore, it would seem likely that DNA binding by ISGF3 relies on the same Stat1 domain, but in this case dimerized with an analogous region of Stat2. Although Stat dimers are incapable of binding to a single half-site, a low affinity interaction at a half-site would be stabilized through protein-protein contacts with DNA-bound p48. Additional evidence for Stat1-Stat2 dimers being capable of forming a functional DNA binding domain has been provided recently by Li et al. (49), who showed that such dimers could bind a palindromic GAS sequence. These Stat1-Stat2 complexes bind a DNA sequence also recognized by Stat1 homodimers, indicating that Stat2 does not affect the specificity of Stat1, probably because it fails to contribute directly to DNA binding. Likewise, analysis of ISGF3 formation by protein-DNA cross-linking failed to detect stable interaction between Stat2 and DNA (25).
Each of the three polypeptides comprising ISGF3 appears to provide a distinct function. p48 contributes most of the DNA binding specificity by recognizing the core sequence of the ISRE (22). Stat2 contains a transactivation domain that is essential for transcriptional activity of ISGF3 (29). Stat1, which also contains a transactivation domain that functions in Stat1 homodimer complexes (32), does not appear to contribute significantly to the transcriptional potency of ISGF3 (29). Rather, Stat1 contributes necessary contacts with DNA which raise the affinity of ISGF3 for DNA above a minimal threshold provided by p48 alone. Indeed, the Stat1␤ isoform, which lacks the transactivation domain used by Stat1 homodimers, is capable of stabilizing ISGF3 formation (32).
Further evidence that the Stat1 DNA binding domain is involved in interactions at the ISRE was obtained from analysis of the recognition specificity of the ISGF3-like complex composed of Stat1 homodimers and p48. Again, GAS-like halfsites appear involved in stabilizing the protein-DNA interaction. Interestingly, the Stat1-p48 complex contacted two TTC half-sites on either side of the p48 recognition sequence, unlike ISGF3 where Stat1 interacted with only the 3Ј TTC. It is possible that the presence of Stat1 homodimers in this complex forms a symmetric complex, whereas Stat1-Stat2 heterodimers form an asymmetric complex that restricts Stat1 binding to a single side of p48. Similar asymmetric interactions of heterodimeric transcription factors have been reported for NF-AT⅐AP-1 complexes on the interleukin-2 receptor enhancer (50).
All Stat proteins, with the exception of Stat2, appear to bind FIG. 4. ISGF3, Stat2-p48, and Stat1-p48 complexes display distinct interactions with the ISRE. Panel A, Stat2-p48 complexes are less stable on DNA than ISGF3. Protein-DNA complexes were formed between ISGF3, Stat2-p48, or p48 alone and labeled ISRE oligonucleotide. The stability of these complexes was determined by measuring the amount of residual labeled DNA which remained bound to protein at serial times following addition of unlabeled ISRE oligonucleotide. Portions of each binding reaction were removed at the indicated times (min) and loaded directly onto a running polyacrylamide gel. Samples were loaded at differing times throughout the experiment and were therefore electrophoresed for different lengths of time, resulting in somewhat different mobilities of the same complexes from different time points. Panel B, distinct DNA binding specificities of ISGF3-like complexes. The DNA binding specificity of ISGF3 (lanes 1-3), Stat2-p48 (lanes 4 -6), or Stat1-p48 (lanes 7-9) was measured using wild type and point-mutated ISRE probes labeled to equal specific activities, as indicated. Only the ISGF3-like complexes are shown. palindromic DNA sequences as homodimers. It is possible that Stat2 homodimers also could bind DNA on their own. Indeed, Stat2-containing complexes have been detected which bind genomic DNA, but neither the nature of the complex nor the sequence of the DNA has been characterized (51). Our data would argue that, even if Stat2 is capable of binding DNA, it is specific for a distinct sequence other than the ISRE. The presence of Stat2 in an ISGF3 complex did not appear to add nucleotide contacts to those provided by Stat1. Also, Stat2 homodimers bound to DNA in conjunction with p48 appeared to have a DNA binding specificity identical to that of p48 alone. The presence of Stat2 did not increase the affinity for DNA, again suggesting that Stat2 was not in contact with DNA. On the other hand, it is entirely possible that Stat2 homodimers are capable of binding to a DNA sequence distinct from either the ISRE or the GAS elements.
The biological significance of Stat2-p48 complexes is difficult to assess. Although this transcription factor is capable of being formed in response to IFN-␣, its low DNA binding affinity for an ISRE sequence precludes transcriptional responses at modest protein concentrations. Thus, U3A cells that lack Stat1 but express Stat2 do not detectably express ISRE-containing genes (32). Similarly, we and others have produced mouse strains lacking Stat1, and these animals are also impaired for transcription of IFN-␣-stimulated genes (52,53). However, in both U3A cells and in Stat1Ϫ/Ϫ mice, constitutive levels of p48 are reduced relative to wild type and neither p48 nor Stat2 levels can be induced in response to IFN treatment. Therefore, active Stat2-p48 complexes would not be expected to be detected. It remains possible that natural situations of abundant Stat2 and p48 exist where Stat2-p48 complexes could mediate a weak transcriptional response to IFN-␣. It also remains possible that alternative DNA sequences or alternative adaptor proteins exist which allow Stat2 homodimers to be recruited to DNA and activate transcription without the direct participation of Stat1.