Differential activation of acute phase response factor/Stat3 and Stat1 via the cytoplasmic domain of the interleukin 6 signal transducer gp130. II. Src homology SH2 domains define the specificity of stat factor activation.

Distinct yet overlapping sets of STAT transcription factors are activated by different cytokines. One example is the differential activation of acute phase response factor (APRF, also called Stat3) and Stat1 by interleukin 6 and interferon-gamma. Interleukin 6 activates both factors while, at least in human cells, interferon-gamma recruits only Stat1. Stat1 activation by interferon-gamma is mediated through a cytosolic tyrosine motif, Y440, of the interferon-gamma receptor. In an accompanying paper (Gerhartz, C., Heesel, B., Sasse, J., Hemmann, U., Landgraf, C., Schneider-Mergener, J., Horn, F., Heinrich, P. C., and Graeve, L. (1996) J. Biol. Chem. 271, 12991-12998), we demonstrated that two tyrosine motifs within the cytoplasmic part of the interleukin 6 signal transducer gp130 specifically mediate APRF activation while two others can recruit both APRF and Stat1. By expressing a series of Stat1/APRF domain swap mutants in COS-7 cells, we now determined which domains of Stat1 and APRF are involved in the specific recognition of phosphotyrosine motifs. Our data demonstrate that the SH2 domain is the sole determinant of specific STAT factor recruitment. Furthermore, the SH2 domain of Stat1 is able to recognize two unrelated types of phosphotyrosine motifs, one represented by the interferon-gamma receptor Y440DKPH peptide, and the other by two gp130 YXPQ motifs. By molecular modeling, we propose three-dimensional model structures of the Stat1 and APRF SH2 domains which allow us to explain the different binding preferences of these factors and to predict amino acids crucial for specific peptide recognition.

Interleukin 6 (IL-6) 1 is a multifunctional cytokine synthesized by many different cells after appropriate stimulation. It acts on a wide spectrum of target cells and exerts multiple functions during the immune response, hematopoiesis, neuronal differentiation, and the acute phase reaction (1)(2)(3)(4).
IL-6 acts via a cell surface receptor complex composed of two subunits: an 80-kDa IL-6-binding protein (IL-6 receptor) and a 130-kDa signal transducing protein, gp130 (5)(6)(7). gp130 is also the signal transducing component of the receptors for leukemia inhibitory factor, oncostatin M, ciliary neurotrophic factor, interleukin 11, and cardiotrophin 1 (8,9). Binding of IL-6 to its receptor induces the homodimerization of two gp130 molecules, thereby transducing the signal into the cell (10). The 61-amino acid juxtamembrane region of the 277-amino acid cytoplasmic domain of gp130 was found to be sufficient for mediating a proliferative signal in transfected BAFB03 cells (11). This region contains two short segments referred to as "box 1" and "box 2" which are conserved between members of the hematopoietic cytokine receptor family. Three members of the Janus tyrosine kinase family, Jak1, Jak2, and Tyk2, are found to be constitutively associated with gp130 and are activated and autophosphorylated after IL-6 stimulation (12,13). Recently, Tanner et al. (14) have shown that the conserved box 1 motif within gp130 is required for the association with Jak1 and Jak2. The Jak kinases are believed to mediate the tyrosine phosphorylation of gp130, thereby creating docking sites for SH2 domain containing signaling molecules (15). All STAT factors contain SH2 as well as SH3 domains (16). The SH2 domain has been shown to be involved in both activation and dimerization of STAT1 and STAT2 in response to interferons (17,18). Two members of the STAT family of transcription factors, namely APRF (or STAT3) and STAT1 are rapidly phosphorylated on tyrosine in a number of cell types upon IL-6 stimulation, a process for which Jak1 kinase was found to be essential (12, 19 -21). They homo-or heterodimerize and are then translocated into the nucleus where they bind to specific enhancers within promoter regions of acute phase protein genes (22,23). APRF has also been shown to associate with gp130 (12). Stahl et al. (24) have recently demonstrated that any of the last four carboxyl-terminal tyrosine-containing motifs within the cytoplasmic domain of gp130 is sufficient for the activation of APRF. They proposed the amino acid sequence YXXQ as an APRF binding consensus sequence.
In the IFN␥ receptor ␣-chain a single tyrosine residue, Tyr-440, was found to be functionally important for signal transduction and binding of STAT1 to the receptor (25,26). Mutational studies demonstrated that the important sequence in the IFN␥ receptor is YDXXH, with tyrosine, aspartic acid, and histidine all being crucial for stimulation of major histocompatibility class I expression and IRF-1 induction.
Although STAT1 activation is also triggered via gp130, there is no sequence within the cytoplasmic tail of gp130 resembling the YDKPH motif of the IFN␥ receptor. Therefore, the question arose, how STAT1 recruitment can be triggered by IL-6. Three alternative mechanisms can be envisioned: (i) STAT1 is recruited directly via specific phosphotyrosine modules of gp130, (ii) STAT1 associates with gp130 only indirectly via bound APRF, and (iii) STAT1 associates directly with an activated Jak kinase. In the present study, we therefore identified by DNA binding competition assays with synthetic phosphopeptides and by mutational analysis the specific tyrosine motifs required for APRF and STAT1 activation through gp130. We found that APRF can be activated independently via four tyrosine modules but that only two of these were able to mediate STAT1 activation. The specificity of the activation sequences can be changed by two point mutations in both directions. The identification of a novel phosphotyrosine motif within the cytoplasmic domain of gp130 important for STAT1 activation strongly suggests that STAT1 transiently associates with gp130 independent of APRF.
Recombinant human erythropoietin has been kindly provided by J. Burg and K.-H. Sellinger (Boehringer Mannheim, Penzberg, Germany). The erythropoietin receptor cDNA was a generous gift of Harvey F. Lodish (Boston). STAT cDNAs and antisera were kindly provided by James E. Darnell (New York) and Chris Schindler (New York). A frequently used buffer was phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 10.1 mM Na 2 HPO 4 , 1.76 mM KH 2 PO 4 , pH 7.2). The monoclonal antibody M2 against the FLAG epitope (N-DYKDDDDK-C) was purchased from Kodak (New Haven).

Expression Vectors
Standard cloning procedures were performed as outlined by Sambrook et al. (28).
Construction of the pSVL-Eg⌬S Truncation Mutant-A Bluescript-SK-gp130 vector (pB130), which lacks EcoRV, AccI, ClaI, HindIII, and SalI restriction sites was digested with SphI and BamHI. Into these sites, an oligonucleotide was introduced which encodes the sequence for the FLAG-epitope and contains SphI, SalI, NsiI, BamHI sites, and a stop-codon: 5Ј-CAT GTC GAC TAC AAA GAC GAT GAC GAT AAA TAG-3Ј. The ⌬S mutant was subcloned into the pSVL-Eg expression vector via the EcoRI and BamHI site of the Eg chimera and the pSVL, respectively. The construction of the pSVL-Eg vector coding for the extracellular domain of the murine EpoR, the transmembrane domain, and the cytoplasmic part of gp130, is described elsewhere. 2 Construction of the pSVL-Eg-FLAG-An oligonucleotide 5Ј-CCT AAA AGT TAC TTA CCA CAG ACT GTA CGG CAA GGC GGC TAC ATG CCT CAG G-3Ј containing the sequence of gp130 from amino acids 902-918 and a SphI and SalI site was introduced into the SphI and SalI sites of the pB130 plasmid. The construct was subcloned into pSVL-Eg as described above.
Construction of the pSVL-Eg⌬B-A pBluescript-vector containing the cDNA for the Eg-chimera was digested with BglII and SphI and a linker with the sequence 5Ј-GAT CTG AAA TCA GGC ATG-3Ј was introduced into these sites. The Eg⌬B construct was subcloned into the XbaI and BamHI sites of the pSVL expression vector.
Construction of pSVL-EgY759, -Y767, -Y␥440 -Oligonucleotides containing the NsiI and SphI sites and the sequence encoding 7 amino acids around tyrosines 759 and 767 of gp130 and 440 of the IFN␥ receptor were inserted into pB130. These constructs were digested with SphI and BamHI and ligated with the FLAG-oligonucleotide (see above). The constructs were subcloned into pSVL-Eg as described.
Construction of pSVL-EgY915stop-An oligonucleotide coding for the Y915 motif containing a NsiI, BamHI site and a STOP codon was introduced in the NsiI and BamHI sites of the pB130⌬S plasmid. The mutant was subcloned into pSVL-Eg as described. The sequences of the oligonucleotides coding for the tyrosine modules can be obtained upon request.

Transfection
Transfection of cells was carried out using the Gene Pulser TM from Bio-Rad Laboratories (Mü nchen, Germany). 2 ϫ 10 6 cells in 0.8 ml of DMEM were co-transfected with 10 g of pSVL vector containing either the STAT1 or APRF cDNA plus 20 g of pSVL vector containing EpoR/gp130 chimera cDNAs using a voltage of 230 V and a capacity of 960 F. Three days after transfection, cells were used for additional studies.

Synthesis of Nonphosphorylated and Tyrosine-phosphorylated Peptides
All peptides were synthesized as amides on a multiple peptide synthesizer (Abimed AMS 422, Langenfeld, Germany) according to the standard Fmoc machine protocols using TentaGel S RAM resin (50 M) (Rapp Polymere, Tü bingen, Germany) and PyBOP (benzotriazole-1-yloxytripyrrolidinophosphonium hexafluorophosphate) activation (Novabiochem, Bad Soden, Germany). Side chain protection groups were Trityl (Asn, Gln, His, Cys), Pmc (Arg), tert-butyl (Ser, Thr, Tyr, Asp, Glu), and t-Boc (Lys, Trp). For the synthesis of the phosphorylated peptides, Fmoc-Tyr(PO 3 H 2 )-OH (Novabiochem, Bad Soden, Germany) was used. The peptides were cleaved from the resin and deprotected by treatment with 750 mg of phenol, 250 l of ethanedithiol, 500 l of thioanisole, 500 l of water in 10 ml of trifluoroacetic acid for 4 h. After precipitating the peptides with cold t-butyl methylether, the samples were washed six times with t-butyl methylether, dissolved in 5% acetic acid, and freeze-dried. All products were purified to greater than 80% purity by preparative high performance liquid chromatography.

Preparation of Nuclear Extracts from HepG2 and COS-7 Cells
Nuclear extracts from HepG2 cells were prepared as described previously (22). Nuclear extracts of COS-7 cells were prepared as described by Andrews and Faller (29). The amount of protein was measured with a Bio-Rad TM protein assay.

Electrophoretic Mobility Shift Assay (EMSA)
EMSAs were performed as described previously (22) with the following modifications. For nuclear extracts from COS-7 cells a gel shift incubation buffer without KCl was used. We used a double-stranded 32 P-labeled mutated SIE-oligonucleotide from the c-fos promoter (m67SIE: 5Ј-GAT CCG GGA GGG ATT TAC GGG GAA ATG CTG-3Ј) (30). The protein-DNA complexes were separated on a 4.5% polyacrylamide gel containing 7.5% glycerol in 0.25-fold TBE at 20 V/cm for 4 h. Gels were fixed in 10% methanol, 10% acetic acid, and 80% water for 1 h, dried, and autoradiographed.
For the supershift assay, nuclear extracts, gel shift mixture, and antisera (in a final dilution of 1:500) were incubated for 30 min at room temperature. Then the double-stranded 32 P-labeled m67SIE-oligonucleotide was added and the EMSAs were performed.

Incubation with Synthetic Peptides
Nuclear extracts, gel shift mixture, and peptides, in the concentrations as indicated in the figures, were incubated for 30 min at room temperature, the radioactive m67SIE-probe was added, and EMSAs were performed.

RESULTS
In order to elucidate through which tyrosine module STAT1 is activated upon IL-6 stimulation, we first used a phosphopeptide competition assay. Such an approach has recently been used successfully in studying the activation of IL-4STAT/ STAT6 (31). Since the SH2 domain of a STAT factor is believed to mediate both, binding to the receptor at phosphotyrosine sites and dimerization with a second phosphorylated STAT factor (18), incubation of a STAT dimer with a large molar excess of relevant phosphotyrosine peptides should lead to the dimer dissociation and thereby prevention of its binding to the specific DNA element.
We examined by electrophoretic mobility shift assays (EMSAs) the inhibitory effects of several synthetic peptides containing each one of the six tyrosine residues of the intracellular domain of gp130 in a nonphosphorylated and a phosphorylated form (Fig. 1A). In addition, the following peptides were analyzed: Y␥, which contains tyrosine 440 of the IFN␥ receptor (32); YAPRF, which contains tyrosine 705 of APRF believed to play a critical role in dimerization (19); YRHE, which contains tyrosine 767 but has a point mutation from glutamine (Q) to glutamic acid (E) at position ϩ3 and Yn, which contains the identical amino acids as peptide Y767 but in a random formation (Fig. 1A).
Nuclear extracts from HepG2 cells stimulated with IL-6 were analyzed using an oligonucleotide with high affinity to both APRF and STAT1 (19,30). EMSAs of control extracts show three complexes (Fig. 1B, left lanes) which are (from top to bottom), APRF homodimers, APRF-STAT1 heterodimers, and STAT1 homodimers (19). Four of the six phosphopeptides from gp130 were able to specifically inhibit the formation of DNAprotein complexes: Y P 767, Y P 814, Y P 905, and Y P 915 (Fig. 1B), whereas Y P 683 and Y P 759 had no effect (data not shown). Peptides Y P 767 and Y P 905 were effective already at the lowest concentration used (30 M), phosphopeptides Y P 814 and Y P 915 inhibited the DNA binding of STATs only when used at higher concentrations. The phosphopeptide Y P RHE which has the same sequence as Y P 767 except for a point mutation Gln 3 Glu at position ϩ3 showed almost no competition indicating that the glutamine in position ϩ3 from tyrosine 767 is important for APRF activation. Peptide Y P ␥ from the IFN␥ receptor preferentially abolished the association with the DNA of STAT1 and only at high concentrations reduced that of APRF. Interestingly, the phosphopeptide from APRF (YAPRF) also competed STAT1 and APRF DNA binding. Random peptide Y P n showed no competition, nor did the nonphosphorylated peptides (Fig.  1B, right lanes).
To more precisely identify the tyrosine residues of gp130 necessary for STAT1 recruitment, we constructed a series of chimeric receptors consisting of the extracellular domain of the erythropoietin receptor (EpoR) and transmembrane and cytoplasmic parts of gp130 (Fig. 2). These chimeric receptors allowed us to study the activation of the STAT factors independently of endogenous gp130 in transiently transfected COS-7 cells which do not express the EpoR. The expression of the different chimeric receptors as analyzed by Western blotting using an antibody to the FLAG epitope was found to be comparable (data not shown). Three days after transfection, cells were stimulated with Epo, nuclear extracts were prepared and analyzed by EMSAs.
To first investigate whether both STAT1 and APRF activa- 1. Inhibition of APRF and STAT1 DNA binding activity by phosphopeptides derived from gp130, IFN␥R, and APRF. A, synthetic peptides with amino acid sequences corresponding to tyrosine modules of gp130, IFN␥R, and APRF were synthesized using the multiple peptide synthesizer AMS 422 (Abimed). The peptides include all 6 tyrosine motifs of the gp130 cytoplasmic tail (Y683, Y759, Y767, Y814, Y905, Y915), tyrosine 440 of the IFN␥R crucial for STAT1 activation (Y␥), and tyrosine 705 of APRF involved in dimerization (YAPRF). As controls, a nonsense peptide (Yn) with a randomized sequence of peptide Y767 and a peptide containing the sequence of Y767 with the amino acid exchange Gln 3 Glu (YRHE) were synthesized. All peptides were synthesized in both nonphosphorylated and tyrosine-phosphorylated forms. B, HepG2 cells were stimulated for 7 min with 20 ϫ 10 3 B9 units/ml of IL-6. Nuclear extracts were prepared as described under "Experimental Procedures." Nuclear extracts were incubated for 30 min at room temperature with synthetic peptides in M concentrations as indicated. Samples were then mixed with the double-stranded 32 Plabeled mutated SIE probe of the c-fos promoter (5Ј-GAT CCG GGA GGG ATT TAC GGG GAA ATG CTG-3Ј), and EMSAs were performed. The DNA-protein complexes formed were separated from the free probe by electrophoresis on a native 4.5% polyacrylamide gel. tion can be triggered through the EpoR/gp130 chimera (Eg), we compared the STAT factor activation by Epo in transfected COS-7 cells with the one stimulated by an IL-6⅐soluble IL-6R complex in untransfected cells. Stimulation via the endogenous gp130 in nontransfected COS-7 cells by soluble IL-6R and IL-6 resulted in the formation of three DNA-protein complexes, designated a, b, and c corresponding to APRF homodimers, APRF-STAT1 heterodimers, and STAT1 homodimers, respectively (Fig. 3) (19, 33). After transient transfection of COS-7 cells with cDNAs coding for the chimera containing the complete cytoplasmic domain of gp130 (Eg) and murine APRF, complex a and, to a minor extent, complex b were formed (Fig.  3). Both disappeared after preincubation with APRF antiserum, whereas treatment with STAT1 antibodies had no effect. Nuclear extracts from COS-7 cells expressing the Egchimera and STAT1 showed only complex c. Anti-APRF treatment had no effect, whereas incubation with anti-STAT1 led to a supershift. From these results we conclude that both APRF and STAT1 can be activated via the Eg-chimera. However, our observations indicated that the endogenous levels of the two STAT factors in COS-7 cells are quite different. Stimulation of endogenous gp130 preferentially activated STAT1 but only small amounts of APRF. This is in contrast to the typical IL-6 response in HepG2 and other cells (Fig. 1B). Similarly, in COS-7 cells that were transfected with the Eg-chimera alone, preferential STAT1 activation was observed. 2 Also, Wen and colleagues (34) found only low endogenous amounts of APRF in COS-7 cells. Therefore, all subsequent studies were performed with either APRF or STAT1 coexpression.
To investigate which of the tyrosine residues within the cytoplasmic domain of gp130 are responsible for the activation of APRF and/or STAT1, we performed coexpressions of the different chimeras (Fig. 2) with either APRF or STAT1. In control cells which were transfected with the pSVL vector and either APRF-or STAT1-cDNAs, no activation of STAT factors was detected (Fig. 4). The full-length chimera (Eg) was able to activate both APRF and, to a minor extent, STAT1. In contrast, the deletion mutant Eg⌬S, which lacks the two carboxyl-terminal tyrosine residues of gp130 resulted in the activation of exogenous APRF, whereas coexpressed STAT1 could not be recruited. Deletion mutant Eg⌬B which includes only the box 1 and 2 regions of gp130 and therefore lacks the five carboxyl-terminal tyrosine residues, activated neither APRF nor STAT1. These results indicate that the last two tyrosine residues are required for STAT1 recruitment.
In order to further investigate which tyrosine residue directs the activation of STAT1, chimeric molecules containing box 1 and box 2 plus one of the five distal tyrosine modules of gp130 were tested (Fig. 2). For comparison, a construct containing box 1 and box 2 of gp130 and the tyrosine module of the IFN␥ receptor was used (EgY␥440). Stimulation of chimeric receptor EgY759 did not result in the activation of any of the two STAT factors (Fig. 4). Recently, it was shown that this tyrosine residue binds the phosphotyrosine phosphatase PTP1D/syp (24). In contrast, the chimeras containing tyrosine modules Y767 and Y814 predominantly activated APRF. However, a weak activation of endogenous STAT1 also must have occurred since in both cases small amounts of the APRF/STAT1 heterodimer were formed. The chimeric receptors containing Y905 and Y915, however, were capable of stimulating both STAT proteins; nevertheless, they were still more active toward APRF (Fig. 4, lower panel). The last amino acid in gp130 is the glutamine in position ϩ3 from tyrosine 915. To exclude that the amino acid residues of the FLAG epitope that is fused to the Y915 module in our construct artificially create an activation motif for STATs, we also constructed a chimeric receptor lacking this epitope. This protein showed the same activation pattern as EgY915 with the FLAG epitope (data not shown). Thus, our data show that tyrosine residues 767, 814, 905, and 915 recruit APRF, whereas tyrosines 905 and 915 can also recruit STAT1.
A mutant receptor EgY␥440 containing the tyrosine module of the IFN␥ receptor crucial for the activation of STAT1 by IFN␥ (25) stimulated DNA binding mainly of STAT1 (Fig. 4). Only in cells overexpressing APRF, also low amounts of APRF and APRF/STAT1 heterodimer were activated. This result demonstrates that the specificity of STAT activation is determined neither by the box 1 and box 2 region nor by the kinases involved (Jak1 and Jak2 in both cases) but rather by the tyrosine motifs within the cytoplasmic domain of the signal transducer.
Stahl et al. (24) have recently suggested that the consensus sequence for APRF activation is YXXQ. For the STAT1 recruitment by the IFN␥ receptor, the sequence YDXXH has been proposed (25). Both gp130 tyrosine modules, namely module Y905 and Y915, that are capable of activating STAT1 comprise the sequence YXPQ. To further define the amino acid requirements of these three different types of motifs, we introduced various point mutations. The tyrosine module Y814 was found to have the highest selectivity of activation toward APRF (Fig.  4). Since its sequence YFKQN varies by 3 amino acids from the sequence of the IFN␥ module YDKPH which shows the best selectivity toward STAT1, we constructed chimeras containing point mutations within the Y814 module of gp130 or the Y440 module of the IFN␥R (Fig. 5). Each module contained a centrally located tyrosine flanked on both sides by 5 amino acids of the authentic sequence of the gp130 or IFN␥R, respectively. The motif Y814 (YFKQN) was mutated to YDKQN, YDKPN, and YDKQH. Likewise, we changed the IFN␥R sequence (YD-KPH) into YFKPH, YDKQH, and YFKQH which corresponds to the 4-amino acid motif of Y814 in gp130. When nuclear extracts from COS-7 cells expressing these chimeras were tested in EMSAs, we found that changing the Y814 motif (YFKQN) into YDKQN resulted in a major loss of activated APRF and a complete disappearance of the small amount of activated STAT1 (Fig. 6A). The mutation of YFKQN to YDKPN led to a chimeric receptor which caused reduced APRF but an enhanced STAT1 activation. However, full activation of STAT1 was obtained by the motif containing the double mutant YFKQN 3 YDKQH. Therefore, by exchanging 2 amino acids downstream of the tyrosine at position ϩ1 and ϩ4 we were able to switch the selectivity completely from APRF to STAT1.
When the reverse exchanges were analyzed, it was found that mutation of Y␥440DKPH into Y␥440FKPH resulted in the loss of STAT1 activation, whereas the change of Y␥440DKPH into Y␥440DKQH had no effect (Fig. 6B). However, mutation of Y␥440DKPH into Y␥440FKQH which creates the 4-amino acid sequence of Y814, resulted in a total change of the activation selectivity. The STAT1 homodimer could not be activated any longer, whereas the APRF homodimer was formed to the same extent as it was via authentic Y814.
Tyrosine module Y905 (YLPQ) is one of the two motifs which was able to recruit both STAT factors. It differs from Y814 (YFKQ) by 2 amino acids, a leucine at ϩ1 and a proline at ϩ2. To study the importance of these two residues, we constructed chimeric receptors with the mutations YLPQ 3 YFPQ and YLPQ 3 YLKQ (Fig. 5). The mutant EgY905FPQ activated APRF in a manner comparable to the authentic module; the STAT1 homodimer, however, was found to be reduced (Fig. 7). The mutant EgY905LKQ showed reduced APRF and an even more diminished STAT1 activation. This indicates that the hydrophobic leucine at position ϩ1 and more importantly the proline at position ϩ2 are involved in the recruitment of STAT1.

DISCUSSION
In this study we have identified the tyrosine-containing motifs in the cytoplasmic domain of the IL-6 signal transducer gp130 which are responsible for the differential activation of the transcription factors APRF and STAT1. It is currently believed that upon binding of IL-6 to its receptor a homodimerization of the signal transducer gp130 is induced whereby associated kinases of the Jak family become activated. These kinases then phosphorylate and activate each other which in turn leads to the tyrosine phosphorylation of the signal transducer. The phosphotyrosine residues within gp130 are thought to create docking sites for signaling molecules containing SH2 domains, such as the STAT factors. To date, two of the seven cloned STATs were found to be activated after IL-6 in HepG2 and other cells, APRF and STAT1 (23). Since STAT1 but not APRF is activated upon IFN␥ stimulation (15), whereas both cytokines lead to an activation of Jak1 and Jak2 (12,13,35), it is likely that the specificity for the STATs is not determined by the Jak(s) involved, but by the tyrosine motifs of the signal transducer chain. Our results with the chimeric receptor EgY␥440 which showed that the activation sequence of the IFN␥R for STAT1 can be transferred to a truncated gp130 molecule and still is functional supports this idea.
In accordance with the results of Stahl and co-workers (24), we found in this study that four of the six tyrosine modules of gp130, namely Y767, Y814, Y905, and Y915, are able to recruit APRF. In COS-7 cells, this observation was made only after coexpression of APRF; endogenous APRF could only weakly be activated by a complex of IL-6 and the soluble IL-6 receptor (Fig. 3) or by the Eg-chimera (data not shown). This result is most likely due to the fact that COS-7 cells express only small amounts of APRF, in contrast to STAT1 which is activated even without cotransfection (Fig. 3).
We have also demonstrated that two of the four tyrosine modules, namely Y905 and Y915, are capable of prominently activating STAT1 (Fig. 4). In contrast, Stahl et al. (24) did not observe tyrosine phosphorylation of this transcription factor. Since tyrosine phosphorylation of STATs was shown to be a prerequisite for nuclear transport and binding to DNA (36), our finding is not in accordance with the one of Stahl et al. (24). One possibility to explain these differences might be that the assay used in our studies (EMSA) more sensitively measures active STAT1 compared to the detection of phosphotyrosines in a Western blot. Alternatively, different expression levels of endogenous STAT1 in COS-7 cells could be responsible for this discrepancy.
The ability of modules Y905 and Y915 to recruit STAT1 in the absence of activated APRF strongly suggests that STAT1 via its SH2 domain binds directly to the phosphorylated gp130 dimer and not indirectly via APRF. Our results also rule out the possibility that STAT1 recruitment occurs directly via an active Jak kinase since the mutant Eg⌬B which still contains the kinase binding region showed no activation. Thus, we propose that the activation of APRF and STAT1 via the cytoplasmic domain of gp130 is mediated by multiple independent docking sites. It could be hypothesized that for the formation of heterodimers either both factors bind to the same signal transducer chain at different sites or different factors bind to corresponding sites within the dimerized gp130. Future studies have to address these questions.
All modules of gp130 that were able to activate APRF contain the 4-amino acid sequence YXXQ (Ref. 24 and this study), the two modules which also efficiently activate STAT1 the sequence YXPQ. For STAT1 activation through the IFN␥R, the consensus sequence YDXXH was proposed (25). To demonstrate that these 4-or 5-amino acid sequences are not only necessary but also sufficient for determining the specificity toward APRF and STAT1, we introduced a number of point mutations into tyrosine modules Y814 and Y␥440 in order to change their activation characteristics. In both cases, the exchange of only 2 amino acids resulted in a complete reversal of specificity.
Introduction of a negative charge by aspartate at position ϩ1 of Y814 in all cases strongly impaired APRF activation (Fig.  6A). Aspartate is found at this position in the IFN␥R, but by itself is not capable of activating STAT1 in the context of Y814. This is only achieved in a double mutant which contains aspartate at ϩ1 and histidine at ϩ4. This combination was crucial in all mutants tested, exchange of the charged amino acids by neutral ones (phenylalanine and asparagine, respectively) was not tolerated. Thus, the sequence YDXXH is both necessary and sufficient for activation of STAT1 even in the context of the gp130 Y814 module.
Starting with the Y␥440 module, we were able to change the specificity from STAT1 to APRF by introducing a phenylalanine at position ϩ1 and a glutamine at position ϩ3. As mentioned above, position ϩ1 does not tolerate a negatively charged residue. Within the four tyrosine motifs from gp130 and the two STAT factor motifs responsible for homo-or heterodimerization (APRF, Y705LKTK; STAT1, Y701IKTE) at this position either a hydrophobic amino acid (Phe, Leu, Met, Ile) or a positively charged arginine was found (Fig. 2B). At position ϩ2, a positive charge (histidine or lysine) or a proline was observed. Thus, for APRF activation, we propose the consensus sequence YX h,p X h,p Q (h ϭ hydrophobic; p ϭ positive charge).
The most obvious difference between the two phosphotyrosine modules of gp130 that were capable of recruiting STAT1 and module Y␥440 is the lack of charged amino acids in the gp130 modules. The mutation of residues at postion ϩ1 and ϩ2 in Y905 demonstrated that both contribute to the specificity for STAT1 and that the introduction of a positive charge (lysine) at position ϩ2 destroys the affinity for STAT1. In the module Y␥440, a lysine at this position is tolerated; however, only in combination with a negative charge at postion ϩ1. Thus for STAT1 activation via a module that also activates APRF, the consensus sequence YX h PQ is proposed. These interpretations are consistent with the models for the SH2 domains of APRF and STAT1 that are proposed. 2 In conclusion, our results strongly indicate that the specificity for either STAT1 or APRF is determined by 4-to 5-amino acid sequence motifs independent of the environment in which these are located. In the future, more detailed mutational studies and affinity measurements using plasmon resonance will help to define more precisely these STAT activation motifs.
The studies with the synthetic phosphopeptides only partially confirmed the data found by transfection of the chimeric receptors. The gp130 phosphopeptides Y P 767 and Y P 905 most efficiently inhibited the DNA binding of both STAT factors. A differential activation of STAT1 by module Y905 as was found by the studies of chimeras could not have been predicted from this result. Likewise, phosphopeptide Y P ␥ inactivated both STAT1 and APRF dimers, whereas the corresponding chimera activated only STAT1. Probably, these in vitro competition assays can measure only large affinity differences whereas already subtle ones may be important for the in vivo choice of STATs. Thus, studies of this kind have to be interpreted with care and should be confirmed by independent methods. Interestingly, the phosphopeptide containing Y705 of APRF also is capable of dissociating the STAT dimers, suggesting that this phosphotyrosine residue is actually responsible for the dimerization.
In addition to IL-6, a number of other cytokines were recently described to activate APRF. A search in a protein data library for sequences that agree with the consensus sequences found for the activation of either APRF or APRF and STAT1 revealed appropriate motifs in the cytoplasmic domains of receptors for leukemia inhibitory factor, granulocyte colony-stimulating factor, interleukin 10 and 12, interferon-␣, and thrombopoietin (Fig. 8). Interestingly, such a motif was also found in the ␤-chain of the mouse, but not the human IFN␥R. This might explain why some authors (using murine cells) have reported that APRF is activated upon IFN␥ stimulation, whereas in human cells this was not the case (23,37).