Interleukin-6 Signaling via Four Transcription Factors Binding Palindromic Enhancers of Different Genes*

Interferons (IFNs), as well as some interleukins, growth factors, and hormones, all induce tyrosine phosphorylation of STATl and additional transcription fac- tors of similar sizes. These factors are activated to trans-locate to nucleus and bind to enhancers of consensus sequence 'ITnCnnnAA (y-IFN activated sequence-like enhancers). In mammary cells or hybridoma B9 cells, four distinct tyrosine-phosphorylated transcription complexes activated by interleukin-6 (IL-6) and IFN-fi were observed: pIRFA and complexes I, 11, and I11 (of increasing electrophoretic mobility). The factors have unequal affinities for enhancers of different genes; they are activated with distinct kinetics and to different ex-tents by IL-6 and IFNs. The pIFWA band isolated from IL-Cstimulated B9 hybridoma cells revealed three DNA-interacting components: two large subunits of 91 and 98 kDa, as well as a small component of 46 kDa not seen in other complexes analyzed. One of the large pIRFA subunits may be APRFBTAT3, since pIRFA reacted with anti-APRF antibodies as do complexes I and 11. However, pIRFA did not react with antibodies to STAT1, indicat- ing STATl is not the other large component of pIRFA. Complex 11, which reacted to anti-acute phase response factor antibodies also reacted to anti-STAT1 antibodies, whereas complex I11 reacted only to anti-STAT1 and was the only complex resistant to N-ethylmaleimide. By its multimeric subunit structure and its cytokine

APRF, isolated from livers of IL-6-treated rats by its affinity to a2M sequences, is structurally related to STATl and is tyrosine-phosphorylated in response to IL-6 (18). The same factor, called also STAT3, is tyrosine-phosphorylated in response to EGF (19). We have studied an effect of IL-6 seen in various cell types of lymphoid, myeloid, or epithelial origin, which is the rapid induction of t h e gene encoding transcription factor IRF-1 (20)(21)(22). The IRF-1 gene is controlled by a palindromic IFN-yresponsive enhancer, pIRE (23), which also mediates the effects of IL-6 (22). The main DNA-protein complex observed in IFNy-treated mammary cells using this sequence contains STAT1, while an additional pIRFA complex is the major one in IL-6treated cells (22). Here we show that the pIRFA complex differs in its subunit structure and antibody reaction from the three complexes (A, B, and C) reported previously on GAS sequences (19). Enhancer sequences, as found in the IRF-1 (231, ICSBP (241, u2M (171, and FcyR genes (25)(26)(27), display a TTTC...GAAA palindrome. We compared these DNA sequences with consensus GAS sequences (TTnCnnnAA) for their capacity to bind 1L-6-and IFN-activated factors.

MATERIALS AND METHODS
Mobility Shift Assays-Mobility shift assays were done as detailed (22). Briefly, human breast carcinoma T47D-07 cells were subcultured and 1 day later were treated by human rIL-6 (20 ng = 100 unitdml), rIFN-p, or rIFN-y (500 IU/ml) (all Chinese hamster ovary-produced). B9 hybridoma cells were grown in an incubator as described. Alternatively B9 hybridoma cells were injected into BALB/C mice together with 50 unitdm1 IL-6, and ascitic fluid cells obtained every 3 4 days were either frozen (untreated) or concentrated to 15 x lo6 cells/ml and treated for 45 min in a 5% CO, incubator with 200 units/ml IL-6. From frozen cell pellets, cytosol was extracted in the low salt buffer W (which has antiproteases, and antiphosphatases 50 m M NaF, 0.1 m M sodium vanadate, 10 m M sodium molybdate) and then nuclear extracts made with 0.4 M NaC1, buffer W (22).
The pIREIIRF-1 oligonucleotides 5'-gatcCTGATTTCCCCGAAAT-GACGG, pIRE/a,M oligonucleotides, 5'-gatcCTGATTTCTGGGAAAT-GACGG (in bold, the sequence found in the rat a,M enhancer or in the FcyR enhancer (opposite orientation)), and ICSBP oligonucleotides 5'-gatcGTGAWCTCGGAAAGAGAG, and their complements, were synthesized and end-labeled by [y-32PlATP with T4 polynucleotide kinase to 2 x lo3 cpdfmol. Competitors are shown in Table I. Mobility shift reaction mixtures of 20 p1 (22) with equal amounts of extract proteins (5-10 pg) and 2 x lo4 cpm of DNA probes were incubated 15 min at 25 "C before gel electrophoresis 2-3 h at 175 V, subject to either autoradiography or exposure in a Fujix BAS1000 Phospho-Imager.
Anti-STAT1 antibodies were obtained from rabbits immunized with an Escherichia coli-produced fused protein containing amino acids 598-705 of p91 ISGF3 (28) cloned in pGEX-3X vector (Pharmacia Biotech Inc.) and purified on glutathione columns (22). Ammonium sulfateprecipitated immunoglobulins from control and immune sera (1 pl) were preincubated with 5 pg of extract protein in 5 pl for 30 min at 0 "C before adding the 32P-labeled probe. Rabbit anti-STAT91N from Transduction Laboratories (Lexington, KY), anti-NF-IL-6 from Dr T. Kishi-mot0 (Osaka, Japan), anti-ISGF3y p48 produced as above, and monoclonal anti-phosphotyrosine antibodies PT-66 (BioYeda, Kiryat Weizmann, Israel) were used similarly. For N-ethylmaleimide (NEM) reaction, nuclear extracts were preincubated in 10 m M NEM (Sigma) for 10   W Cross-linking-For UV cross-linking, nuclear proteins were extracted from murine hybridoma B9 cells first starved of IL-6 by resuspension at 2 x lo5 celldml for 5 h in fresh medium (22) and then treated for 1 h with IL-6 (20 ng/ml). The mobility shift assay was done with 15 pg of protein from B9 nuclear extracts and lo5 cpm of pIRE/a,M probe (core FcyR) or pIREfiRF1 probe, and the part of the gel corresponding to the pIRFA complex was exposed to W light (302 nm) for 30 min. Analysis of the DNA-linked proteins was done by electrophoresis on a 7.5% SDS-polyacrylamide gel.

Binding of Four IL-6-activated Factors to pIRE Sequences-
The pIRE sequences of two members of the IRF gene family, IRF-1 and ICSBP, differ by 2 bases in the central trinucleotide spacer separating the palindrome (Table I). A sequence that deviates from IRF-1 by only 1 of these bases was chosen so as to be similar to the FcyR gene enhancer but also to the core of the acute phase response element found in the rat (but not the human) a,M gene (17) and has been designated pIRE/a,M (Table I). Nuclear extracts from human mammary carcinoma T47D cells treated with IL-6 (22) displayed different sets of induced protein complexes with the three probes (Fig. 1). The slowest migrating pIRFA complex was predominantly formed with pIREiIRF-1 (Fig. lA,  The GAS sequence of the IFN-y-induced GTP-binding protein gene, which binds mainly the STATl factor (291, has been shown previously not to compete for the formation of the complex pIRFA with the pIRE/IRF-1 probe (221, but it competes for formation of complex 111, shown to contain p91 STATl (22). The sequence specificity of the four IL-6-dependent complexes forming on pIRE/a,M oligonucleotides (as in Fig. lA, lane 6) was examined by competitor titration (Table 11). The GAS consensus Ly6E sequence ( 2 ) had a high affinity for complex I11 but not for pIRFA, competitor levels that displaced more than 75% of complex I11 had no effect on pIRFA. When Ly6E was used as a probe, no pIRFA complex formed in the mobility shift assay (not shown). Complexes I and I1 were competed by LyGE, albeit less than complex 111, and also displayed higher affinities for pIRE/a,M than for IRF-1 (Table 11).
Further differentiation between the complexes was shown by reaction with NEM. Cell extracts treated with NEM (see "Materials and Methods") lost complex 11, I, and pIRFA (Fig. lB, lane 91, while complex 111 was NEM-resistant (Fig. lB, lane  10). The predominant pIRFA seen on the pIRE/IRF-1 probe ( Fig. 1 B , compare lane 16 to lane 12) was NEM-sensitive as well. Using this probe, NEM had no effect on complex I11 either after IL-6 or after IFN treatments (Fig. lB, lanes 16-18). The DNA binding activity of ISGF3, the IFN-@-activated factor (29) that contains STAT1, STAT2, and an ISGF3r p48 component, is also NEM-sensitive, but the STATl proteins of ISGF3a (p91 and p84) were not reported to be NEM-sensitive. The appearance of complexes I11 and pIRFA is rapid, within 5 min in T47D cells (221, and does not require protein synthesis (Fig. lC, lanes 7-10). In line with activation of STATl binding to GAS elements by IFN-y through tyrosine phosphorylation (2, ll), anti-phosphotyrosine antibodies blocked the binding of complex I11 to the IRF-1 probe when added to extracts of cells stimulated either with IL-6 or with IFN-y (Fig. lC, compare  lanes 2 and 3 with lanes 5 and 6). The pIRFA activity induced by IL-6 was likewise eliminated by anti-Tyr(P) addition (Fig.  lC, compare lane 2 with lane 5), indicating that pIRFAcontains tyrosine-phosphorylated protein(s). Treatment of mammary cells by the tyrosine kinase inhibitor genistein (30) before IL-6 addition prevented the appearance of the STATl and pIRFA complexes (Fig. lC, compare lanes 11-14 with lanes 15-18 or compare lanes 19 and 20 with 23 and 241, indicating that such kinase activity is also needed for IL-6 to trigger pIRFA binding. At 10 min after IL-6, pIRFA was localized in the nuclear fraction, whereas the IL-6-induced complex I11 (STAT11 was still present in the cytosolic fraction (Fig. E , lanes 11-14 and 19- 22 1.
Antipeptide antibodies against APRF/STAT3 (18) were assayed using IL-6-treated extracts of the mouse hybridoma B9 cells (Fig. 2 A ) or of the T47 mammary cells (data not shown). B9 cell extracts show the same four complexes as T47 extracts with pIRE/a,M probes, but much stronger signals are obtained with B9 cells. APRF antibodies eliminated (supershifted) complexes 11, I, and pIRFA (Fig. 2 A ) , indicating that APRF or a similar protein enters in the composition of these complexes. By contrast complex I11 was not displaced by APRF antibodies (Fig. 2 A ) . The same effects are observed with T47 cell extracts and pIRE/a,M probe (data not shown).
Among other antibodies to known factors tested, neither anti-NF-IL-6 (35) nor anti-ISGF3y p48 (36) had an effect on pIRFA or any of the other three complexes (data not shown).
Composition of pIRFA Factor-The protein composition of pIRFA formed on pIRE/IRFl or pIRE/a,M oligonucleotides was examined by U V cross-linking (see "Materials and Methods") on preparative mobility shift assay with nuclear extracts of IL-6-stimulated hybridoma B9 cells (Fig. 2 B ) . The pIRFA complex contained two large cross-linked components of about 91 and 98 kDa, as well as a smaller component a t 46 kDa (Fig.  2B ). In complex I, which forms efficiently on pIRE/a,M probes, the 46-kDa component was not observed, whereas the 91-98-kDa bands were present (data not shown). In the IFN-yinduced complex 111, a 91-kDa protein was cross-linked to DNA, but no small component was detectable (data not shown). Thus pIRFA is another type of factor on the pathway of IL-6 or IFN action, since it has both large and small DNA-cross-linked proteins. Of the pIRFA components interacting with DNA, one may be APRF or a related protein. Among these or other subunits of pIRFA, at least one is NEM-sensitive and one or more is tyrosine-phosphorylated as shown above. DISCUSSION Our data show that the four protein complexes assembling on the pIRE/a,M sequence in response to IL-6 have distinct properties. Functionally, they diverge in their affinities for individual GAS enhancers conserving the TTnCnnnAA consensus and clearly discriminate between several pIRE sequences despite conservation of the TTTC...GAAApalindrome. Comparison of the pIRE from IRF-1, ICSBP, and a,M(FcyR) genes shows how single-base differences in the spacer can affect factor binding. The change between IRF-1 and pIRE/a,M allows formation of the additional complexes I and 11, and the additional single-base change to ICSBP prevents pIRFA formation while allowing complexes I and I1 to assemble. All the pIRE probes used here bind STATl in line with having GAS consensus sequences. Among the four factors, pIRFA displays the lowest affinity for GAS consensus Ly6E and seems more specific for palindromic pIRE sequences, such as in the IRF-1 and a,M(FcyR) gene enhancers. However, some base changes in the spacer greatly reduce its affinity, as seen with the ICSBP pIRE enhancer. The early and strong induction of pIRFA by IL-6 could correlate with activation of the IRF-1 gene, as observed in mammary and lymphoid cells (22) and also in myeloid cells, in which ICSBP was not induced (21,22). The FcyR protein was induced by IL-6 in myeloid M1 cells (32).
It was reported (3, 19) that the high affinity SIEm67 sequence, a mutant of the c-fos SIE, forms three complexes, A, B, and C, with factors by EGF in A341 cells. Complex A has been found to contain the STAT3 factor, which is also activated to bind to the same sequence by EGF (19) and is identical to the APRF factor, related to the effect of IL-6 on certain APP genes (18). Differential binding of the EGF-dependent factors to SIE (m67) oligonucleotides, wild-type c-fos SIE and Ly6E were observed, indicating that the slower migrating complexes A and B form better on SIEm67 than on Ly6E sequences (3) and are only partially (complex B) or not at all (complex A) blocked by antibodies to STAT1, while the latter eliminated complex C (13). By their electrophoretic migration, reaction with antibodies, and preferential binding sequences, the SIEm67 A and B bands may be akin to the ICSBP and pIRE/a,M(FcyR) complexes I and 11, while complex I11 corresponds to complex C. STATl binds to DNA as a dimer (33), and it was proposed that STATl is able to dimerize with an EGF-inducible factor of the same family (STAT3) to form complex B, while complex Awould be a dimer of STAT3/APRF (3, 19). Although our antibodies to anti-STAT1 (598-705) also recognize the p84 ISGF3a protein, which is derived from the same gene as STAT1, by differential splicing (33), the p84/STATl is not detected on Western blots of our cell extracts at these times of treatment (data not shown). It is thus unlikely that either of the two small complexes I11 and I1 contains this protein.
The additional complex pIRFA forming on some of the palindromic pIRE enhancers, has a mobility lower than that of complex I and appears as a newly defined oligomeric factor. We show that pIRFA is composed of three proteins, one of 98 kDa, one of 91 kDa, and one of 46 kDa. These results were obtained with B9 cells, most sensitive to IL-6, and where the concentration of the factor pIRFA is much higher than in the mammary cell line T47 or hepatoma cell line HepG2 (data not shown). The composition of pIRFA suggests that a heterodimer of STAT3/ APRF and another large component can associate with a small DNA-binding protein, which would explain why it migrates more slowly. Either of the two large proteins could be APRF since pIRFA is shifted by anti-APRF antibodies. The 91-kDa component of pIRFA is probably not ISGF3 p91/STAT1, and the 46-kDa component is distinct from p48 ISGF3y because pIRFA does not react with anti-STAT1 antibodies or anti-p48 antibodies. In addition to STATl and STATB/APRF, two other proteins of the same family were cloned (19,34). Their participation to pIRFA remains to be tested. Akira et al. (18) mentioned the copurification of other large molecular mass components with AF'RF. In extracts of cells treated by IFN-y, a 43-kDa protein, with no large protein, was reported t o be cross-linked to a 39-base pair fragment of the FcyR promoter (25). However, the IFN-y-induced complexes reacted with STATl antibodies (25) and were therefore different from pIRFA. Isolation of the 46-kDa protein from B9 cells is in progress.
This pIRFA factor is intensely and rapidly activated by IL-6, as well as, although to a lesser extent and more slowly, by IFN-P. This factor could be of critical importance in generating the different patterns of gene activation resulting from the action of these cytokines as compared to growth factors such as EGF. As a consequence of activation of multiple factors with discrete target sequence affinities, cellular genes controlled by closely related but not identical pIRE or GAS enhancers have the possibility to respond specifically to stimulation by a given cytokine. For example, we have observed that in transfected mammary cells, IL-6 activates much less luciferase expression driven by GAS/GTP-binding protein than that driven by pIRE/ IRF-1 or pIRE/a,M, whereas this difference was not seen when the same cells were stimulated by IFN-y (22). Such heterogeneous gene responses observable in a given cell stimulated by one cytokine are expected to be amplified when comparing diverse cell types. Indeed, comparison of various cells has shown that different combinations of Jakl, Jak2, and Tyk2 tyrosine kinases can associate with and may activate the IL-6Aeukemia inhibitory factor/oncostatin Wciliary neurotropic factor gp130 receptor component (9). Hence, a large diversity of transcriptional responses to individual cytokines, growth factors, and hormones can be expected to be generated by what at first sight appears as a common mechanism of action through GAS enhancers. In addition to cell-and receptor-specific variations, the enhancer sequence specificity of the multiple pIRE and GAS binding factors is obviously an important element in generating this diversity.