The WD motif-containing protein receptor for activated protein kinase C (RACK1) is required for recruitment and activation of signal transducer and activator of transcription 1 through the type I interferon receptor.

An obligatory step in the activation of Signal Transducers and Activators of Transcription (STATs) by cytokines is their docking to specific receptors via phosphotyrosines. However, this model does not address whether STATs pre-associate with their corresponding receptor or exist free in the cytoplasm before receptor activation. In this report, we demonstrate that pre-association of STAT1 with the receptor is required for type I interferon (IFN) signaling. Interestingly, the interaction between the human type I IFN receptor and STAT1 is not direct but mediated by the adapter protein receptor for activated protein kinase C (RACK1). Disruption of the IFNalpha receptor-RACK1 interaction abolishes not only IFNalpha-induced tyrosine phosphorylation of STAT1 but also activation of STAT2, indicating that RACK1 plays a central role in early signaling through the Jak-STAT pathway. These findings demonstrate the involvement of RACK1 in STAT1 activation and raise the possibility that other STATs may pre-associate with cytokine receptors through similar adapter-STAT-mediated interactions.

Cytokines and interferons (IFNs) 1 bind to receptors of the cytokine receptor superfamily (1)(2)(3), resulting in the activation of kinases of the Jak family and transcription factors designated STATs or Signal Transducers and Activators of Transcription (4 -7). The Jak-STAT pathway has evolved as the paradigm of cytokine and IFN signaling (4 -7). Although STAT can be activated by different cytokines (i.e. STAT1 is activated by IFN␣, IFN␥, IL6, leukemia inhibitory factor, IL10, etc.), studies with knockout mice clearly indicated that their function is well restricted to precise systems. For example, STAT1 is only required for the physiological functions of IFN␣ and IFN␥ (reviewed in Ref. 7).
STATs are recruited to distinct phosphotyrosines within the receptor complex and then are phosphorylated, probably by Jaks, on the highly conserved C-terminal tyrosines (i.e. tyrosine 701 of Stat1), allowing the SH2 domain of one STAT to interact with the phosphorylated tyrosine on another STAT to form homo-or heterodimers. STAT dimers translocate to the nucleus, where they bind specific DNA elements to activate or inhibit transcription of specific genes (reviewed in Refs. 5 and 8).
One distinctive feature in the type I IFN system is that STAT2 is pre-associated with IFN␣R␤L chain (9,10). Activation of STAT2 in response to type I IFNs (IFN␣, ␤, or ) requires the presence of this constitutive site and one or more of the five proximal tyrosines of the ␤L chain (9). However, the mechanism for STAT1 activation by type I IFNs has not been elucidated. It is known that activation of STAT1 requires the previous activation of STAT2 (11), but it has not been determined whether receptor tyrosines are also required for activation. This is in clear contrast to the activation of STAT1 by IFN␥, which requires docking of STAT1 to a phosphorylated tyrosine on the ␣ chain of the receptor (12).
We report here that RACK1 constitutively interacts with non-phosphorylated STAT1 and functions as an adaptor between this factor and the long form of the ␤ subunit of the IFN␣R (IFN␣R␤L). No interaction between RACK1 and other STAT factors was detected. The interaction between IFN␣R␤L and RACK1 is critical for normal STAT activation and IFN signaling. This is supported by the finding that mutations in the RACK1 binding site of IFN␣R␤L, which includes the Box 2 motif, impaired IFN␣-induced tyrosine phosphorylation of STAT1 and STAT2 and the development of the antiviral state.
GST Fusion Proteins and Mammalian Expression Constructs-The different GST fusion proteins encoding different regions of the cytoplasmic domain of IFN␣R␤L have been described previously (25). For mapping of the RACK1 binding site of IFN␣R␤L, a combination of two or three alanine mutations per construct was introduced in the GST␤L300 -375 that contains the minimum region that binds RACK1. GST fusion expression constructs with mutations of the RACK1 site were made by polymerase chain reaction using the Quickchange kit (Stratagene). All mutations were confirmed by sequencing. A GST fusion protein encoding the full-length RACK1 (GST-RACK1) was produced by polymerase chain reaction and subcloned into the pGEX-KG vector. GST fusion proteins were produced in BL-21 cells as described previously (25). Pull-down experiments and immunoblotting were performed using the same procedure described above for immunoprecipitations.
Study of the Adaptor Function of RACK1 Using a Wheat Germ in Vitro Translation-STAT1 and RACK1 were produced by a T7 wheat germ in vitro transcription/translation kit (Promega) following the manufacturer's procedure. [ 35 S]methionine-labeled STAT1 and RACK1 proteins alone or in combination were incubated with GST-␤L overnight, washed, and analyzed by SDS-polyacrylamide gel electrophoresis as described for immunoprecipitations.
Expression of IFN␣R␤L Constructs in Mammalian Cells-Mammalian expression constructs with mutations of the RACK1 site of IFN␣R␤L (amino acids 302, 304, and 305) were made by polymerase chain reaction using the Quickchange kit (Stratagene). All mutations were confirmed by sequencing. Constructs were subcloned into the pLXSN retroviral vector and transfected into the 2 packaging cell line, and retrovirus-containing supernatants were used to transduce LpR␣ cells (L-929 cells stably expressing the human ␣ chain of the receptor; Ref. 25). Stable transfectants were selected in medium containing G-418 (500 g/ml) and hygromycin B (500 g/ml), and positive clones were screened by fluorescence-activated cell sorter using the IFNaR␤1 monoclonal antibody (26) that recognizes IFN␣R␤L.

RACK1 Specifically
Associates with STAT1-It has been suggested that proteins containing WD repeats may serve as scaffold or adaptor proteins (27). Because RACK1 interacts with a region of IFN␣R␤L (amino acids 300 -346) (13) that is required for activation of STATs and the antiviral response (25), we hypothesized that RACK1 could recruit STAT1 to the receptor complex. We first performed coimmunoprecipitation experiments using an anti-RACK1 monoclonal antibody to test for an interaction between RACK1 and STAT1. Fig. 1A shows that the anti-RACK1 antibody coimmunoprecipitated STAT1 (upper panel, lane 6) but not STAT3 (lower panel, lane 6) or STAT2 (data not shown) present in unstimulated U266 cell lysates. This interaction is specific, because it cannot be detected by non-immune IgM used as negative control (Fig. 1A, lane 5). The anti-STAT1 and -STAT3 antibodies also coprecipitated STAT3 and STAT1, respectively, as previously reported (28). However, the strong signal for STAT3 detected in anti-STAT1 immunoprecipitates (Fig. 1A, lower panel, lane 2) may correspond in part to incomplete stripping of the membrane after STAT1 immunoblotting. This result strongly suggests that STAT1 specifically interacts with RACK1. It should be noticed, however, that we have not consistently been able to coimmunoprecipitate RACK1 using anti-STAT1 antibodies. One possible explanation is that both antibodies recognize epitopes close to the C-terminal part of the protein, where the RACK1 binding site may be located, and therefore disrupt the interaction.
To further characterize the RACK1-STAT interactions, a GST fusion protein that encoded the full-length RACK1 protein was used to determine whether other STAT proteins bind RACK1. Fig. 1B shows that GST-RACK1, but not GST alone, binds to STAT1. However, GST-RACK1 failed to bind STAT2, -3, -4, -5, or -6 ( Fig. 1, C-F), confirming that the interaction between RACK1 and STAT1 is specific. Interestingly, GST-␤L also bound STAT1 present in cell lysates (Fig. 1B). This result differs from our previous observation using STAT1 produced in wheat germ in vitro translation systems, in which no interaction between IFN␣R␤L and STAT1 was detected (9). One possible explanation is that a protein such as RACK1 present in cell lysates, but absent in the wheat germ in vitro translation system, may serve as an adapter between IFN␣R␤L and STAT1 (see below).
RACK1 Functions as an Adaptor between STAT1 and the IFN␣R-We reasoned that if RACK1 links STAT1 to IFN␣R␤L, deletions or mutations in IFN␣R␤L that decrease RACK1 binding should also decrease the association of STAT1 with the receptor. To test this hypothesis, we performed pull-down experiments using GST fusion proteins containing different regions of the cytoplasmic domain of IFN␣R␤L. Fig. 2A, top panel shows that GST fusion proteins encoding the entire cytoplasmic domain (lane 7, GST␤L-wt) and proteins truncated at amino acids 462, 375, and 346 (lanes 3-5), but not at amino acids 265-299 (lane 2), were also able to interact with STAT1. Similarly, a GST fusion protein encoding amino acids 300 -515 ( Fig. 2A, lane 6, GST␤L300 -515) and therefore lacking the first 35 amino acids of the cytoplasmic domain (265-299) also interacts with STAT1. The same GST-␤L fusion proteins that bound STAT1 also interacted with RACK1 ( Fig. 2A, lower panel). This result indicates that the minimal interaction domain for STAT1 corresponds to amino acids 300-346 of IFN␣R␤L and is identical to the RACK1 binding site ( Fig. 2A, lower panel 1. RACK-1 interacts with STAT1. A, U-266 (2 ϫ 10 7 cells/ immunoprecipitation (IP)) cell lysates were immunoprecipitated with the indicated polyclonal antibodies against STAT proteins (anti-STAT1, -STAT2, or -STAT3) or normal rabbit (NR) serum. RACK1 was immunoprecipitated using a specific anti-RACK1 monoclonal antibody or control monoclonal antibody. Immunoblotting was sequentially performed using anti-STAT1, -STAT3, and -STAT2 (data not shown) antibodies. B-F, pull-downs with GST-RACK1 were performed to determine whether RACK1 interacts with different STATs. U-266 (B, C, and E) or Daudi (F) cells were used as a source of the indicated STAT proteins. STAT4 was produced by in vitro translation (D). Immunoblotting was performed with the indicated anti-STAT antibodies. WB, Western blot.
We next performed an alanine scan of this region to further define the RACK1 and STAT1 binding sites. Although no individual mutation completely abolished RACK1 or STAT1 bind-ing to GST-␤L, some mutations produced a decrease in binding of STAT1 to IFN␣R␤L that paralleled the decrease in binding of RACK1 to this receptor chain (Fig. 2B). The most intense reduction in binding was observed when amino acids within the region 302-305 and 314 -327 of IFN␣R␤L (Fig. 2B, lanes 2 and  6 -8) were mutated to alanine. The overlapping in RACK1 and STAT1 binding sites strongly supports the concept that RACK1 functions as an adaptor between IFN␣R␤L and STAT1.
Although the interaction between IFN␣R␤L and STAT1 is detected in cellular lysates (Figs. 1B and 2A), the association between these proteins is not observed when STAT1 is produced in wheat germ lysates (9). A possible explanation for this is that RACK1 functions as an adaptor between IFN␣R␤L and STAT1 and that the wheat germ homolog of RACK1 fails to interact with IFN␣R␤L, STAT1, or both. Therefore, we assessed the ability of GST-␤L to bind STAT1 produced alone or together with RACK1 using a wheat germ in vitro translation system. Fig. 2C shows that GST-␤L interacts with RACK1 but not STAT1 when these proteins are produced separately (Fig.  2C, lanes 2 and 6). However, when RACK1 and STAT1 are in vitro translated together GST-␤L pulls down STAT1 (Fig. 2C,  lane 10). Thus, IFN␣R␤L and STAT1 interact only when RACK1 is present, strongly suggesting that RACK1 functions as an adaptor between IFN␣R␤L and STAT1.
RACK1 Interacts Specifically with the Non-phosphorylated Form of STAT1-Although the experiments presented above demonstrate that the non-activated forms of IFN␣R␤L, RACK1, and STAT1 form a complex, they do not address whether RACK1 interacts with the phosphorylated form of STAT1. This is an important issue because once STAT1 is phosphorylated it must detach from the receptor to form a DNA-binding complex in association with STAT2 and p48. To address this issue, STAT1 phosphorylation was induced by treating U-266 cells with IFN␣ for 15 min. Then, we assessed the ability of GST-RACK1 and/or GST-IFN␣R␤L to associate with tyrosine-phosphorylated STAT1, as determined by immunoblotting with the antiphosphotyrosine antibody 4G10. Immunoprecipitations with an anti-STAT1 serum or GST alone were used as positive and negative controls, respectively. The resultant precipitates were divided in equal parts, resolved in separate gels, and immunoblotted using either anti-phosphotyrosine or anti-STAT1 antibodies. Fig. 3A shows that neither GST-␤L nor GST-RACK1 can precipitate the tyrosine-phosphorylated fraction of STAT1 after IFN␣ treatment (lower panel, lanes 6 and 7), but both bind non-phosphorylated STAT1 in control cells (upper panel, lanes 2 and 3) as well as the non-phosphorylated fraction after IFN␣ treatment (upper panel, lanes 6 and 7). As expected, the anti-STAT1 antibody precipitates the phosphorylated and nonphosphorylated forms of STAT1 (Fig. 3A, lanes 4 and 8). Identical results were obtained in similar experiments in which the same membrane was first immunoblotted with anti-phosphotyrosine and then anti-STAT1 antibodies and the converse (data not shown). These results demonstrate that RACK1 interacts only with the non-phosphorylated form of STAT1 and support the concept that STAT1 dissociates from the IFN␣R␤L-RACK1 complex after becoming phosphorylated to form a DNA-binding complex. This was further demonstrated by the finding that 20 min after IFN␣ treatment almost all STAT1 was localized to the nucleus (Fig. 3B, panel f), whereas RACK1 fluorescence increased and remained in the cytoplasm (panel e). In untreated cells, RACK1 is detected in the cytoplasm, as previously reported (13), whereas STAT1 was present in both cytoplasm and nucleus (Fig. 3B, panels  c and d, respectively). The specificity of the immunofluorescence procedure is demonstrated by the complete lack of signal when normal rabbit serum and anti-IgM were used as negative controls for STAT1 and RACK1 (Fig. 3B, panels a and b), respectively.  (25) were used to map the STAT1 and RACK1 binding sites. GST alone and an anti-RACK1 antibody were used as negative and positive controls, respectively. Lysates obtained from U-266 cells were used as a source of STAT1 and RACK1. RACK1 and STAT1 were detected by immunoblotting with specific monoclonal antibodies (Transduction Laboratories). The residues of IFN␣R␤L encoded by each GST fusion protein are indicated. WB, Western blot; CTRL, control. B, mapping of the RACK1 binding site within the 300 -375 region of IFN␣R␤L. Upper panel, pull-down experiments were performed with GST␤L300 -375 carrying combined mutations of the indicated amino acids to alanine. U-266 cell lysates were used as a source of RACK1 and STAT1. Precipitates were resolved using SDS-polyacrylamide gel electrophoresis, transferred to polyvinylidene difluoride, and blotted with anti-STAT1 and -RACK1 monoclonal antibodies. Lower panel, the same GST fusion proteins used for the pull-down experiment were analyzed by SDS-polyacrylamide gel electrophoresis and stained with Coomassie Blue as a control. C, RACK-1 links STAT1 to IFN␣R␤L. [ 35 S]Methionine-labeled RACK1 and STAT1 were produced alone (lanes 1-4 and 5-8, respectively) or in combination (lanes 9 -12) using a T7 in vitro transcription/translation kit. In vitro translated STAT1 and RACK1 were incubated with GST or GST-␤L or immunoprecipitated with anti-STAT1 and anti-RACK1 antibodies as positive controls. In independent experiments, low amounts of STAT1 were coprecipitated by the anti-RACK1 antibody when in vitro translated STAT1 was used as input. This is probably due to some degree of interaction between STAT1 and the RACK1 homolog present in wheat germ, because RACK1 is relatively conserved among species.

The Interaction between RACK1 and IFN␣R␤L Is Critical for
Activation of STAT1, STAT2, and the Antiviral Response-To further determine the importance of the interaction between IFN␣R␤L, RACK1, and STAT1 in IFN␣ signaling, we expressed the human IFN␣R␤L chain with mutations of the RACK1 binding site in mouse L-929 cells. Several stable clones expressing the mutant IFN␣R␤L chain (designated ␣␤L⌬R1) were selected by fluorescence-activated cell sorter analysis (Fig. 4). We next tested whether disruption of the interaction between the receptor and RACK1 would prevent the activation of STAT1. Human IFN␣2 induced significantly lower levels of STAT1 phosphorylation in cells expressing mutations of the RACK1 binding site of IFN␣R␤L (Fig. 5A, lanes 3 and 6). Interestingly, tyrosine phosphorylation of STAT2 was also reduced by the mutation of the RACK1 site (Fig. 5A, lanes 3 and  6). Normal tyrosine phosphorylation of STAT1 and STAT2 was detected when cells were treated with murine IFN␣4 (Fig. 5A,   lanes 2 and 5), demonstrating that the STAT pathway was functional when activated through the endogenous mouse receptor. The decrease in tyrosine phosphorylation of STAT1 and STAT2 in response to huIFN␣ treatment was not due to a defect in kinase activation, because tyrosine phosphorylation of Jak1 was normal (Fig. 5A, lower panel). These results indicate that the interaction between IFN␣R␤L and RACK1 is not only important for normal activation of STAT1 but also for STAT2 tyrosine phosphorylation through a mechanism that remains to be elucidated (see "Discussion").
We next studied whether huIFN␣2 was able to elicit an antiviral response in cells expressing a mutation of the RACK1 binding site of IFN␣R␤L (Fig. 5B, ␣␤L⌬R1, clones 21 and 24). Fig. 5B shows that huIFN␣2 induced significantly lower levels of protection against encephalomyocarditis virus than did murine IFN␣4, which activates the endogenous murine receptor, in two independent clones expressing mutations of the RACK1 FIG. 3. RACK1 interacts with nontyrosine-phosphorylated STAT1. A, U-266 cells were treated with huIFN␣2 or left untreated as described in the legend to Fig. 1. Cell lysates were used for pulldowns with control GST (lanes 1 and 5), GST-␤L (lanes 2 and 6), GST-RACK1 (lanes 3 and 7), or control anti-STAT1 serum (lanes 4 and 8). Anti-STAT1 or antiphosphotyrosine (pTyr) monoclonal antibodies (upper and lower panels, respectively) were used for immunoblotting. WB, Western blot. B, the cellular localization of RACK1 (a, c, and e) and STAT1 (b, d, and f) in ECV304 endothelial cells treated with IFN␣ (e and f) or left untreated (a-d) was studied using confocal microscopy as previously described (13). Alexa 568-labeled (red; a, c, and d) anti-mouse IgG heavy and light chain and Alexa 488-labeled (green; b, d, and f) goat anti-rabbit secondary antibodies were used to detect RACK1 and STAT1, respectively. Non-immune IgM and normal rabbit serum (NR) were used as negative controls (a and b). Translocation of STAT1 was achieved by treatment with huIFN␣ for 20 min (e and f). Nuclear staining is indicated (arrows).
site. The level of protection detected was also lower than that induced by huIFN␣2 in cells expressing the wild type receptor. These results demonstrate that recruitment of RACK1 to the IFN␣R complex is critical for the activation of STAT1 and STAT2 and for the induction of an antiviral state.

DISCUSSION
The results presented in this study demonstrate that the adaptor protein RACK1 links STAT1 to the human type I IFN receptor. The interaction between the receptor and RACK1 is required for activation of STAT1 and the induction of an antiviral state by huIFN␣. The RACK1-mediated interaction between STAT1 and IFN␣R␤L and the direct association of STAT2 with the distal part of the same chain (9, 10) demonstrate that activation of the STAT pathway by type I IFNs requires the pre-association of STAT factors with the receptor.
These findings raise the question whether a model in which STATs pre-associate with cytokine receptors through adaptor proteins containing a WD motif also applies to other cytokine systems. It has been recently reported that the WD motifcontaining protein StIP (STAT3-Interacting Protein) interacts with several STATs and JAKs (29). However, there are differences between RACK1 and StIP. 1) StIP interacts with more than one STAT and Jak, suggesting that it is not STAT or cytokine specific, and 2) it has not been addressed whether StIP associates directly with cytokine receptors. Nevertheless, it is tempting to speculate that StIP and RACK1 are members of a novel family of proteins involved in Jak-STAT signaling. The concept that other STAT-specific adaptors may exist is also supported by reports indicating that activation of STAT5 by growth hormone occurs in the presence of growth hormone receptors devoid of all tyrosines (30,31).
The alternative to a general model in which all STATs are pre-associated with cytokine receptors is that the only system that requires such pre-association is the type I IFN pathway. In this scenario, the recruitment of STAT1 through RACK1 may reflect a more stringent regulation of STAT activation due to the effect of type I IFNs on cell proliferation, or may preclude the need for phosphorylation of a specific tyrosine on the receptor for STAT1 activation. Unfortunately, we have not been able to address the latter question, because activation of STAT1 is dependent on the previous activation of STAT2 (11). However, addition of single tyrosines to IFN␣R␤L constructs in which all tyrosines had been substituted for alanines failed to affect STAT1 phosphorylation independently of STAT2 phosphorylation. This result raises the possibility that, unlike the IFN␥ receptor, where phosphorylation of tyrosine 440 of the ␣ chain is critical for STAT1 phosphorylation (12,32), tyrosine phosphorylation of the IFN␣R is not critical for STAT1 activation.
Our data also suggest that once STAT1 is tyrosine-phosphorylated, it can dissociate from RACK1 and the receptor complex. This is supported by the finding that RACK1 interacts only with the inactive (non-phosphorylated) form of STAT1 and that soon after type I IFN stimulation STAT1 is almost exclusively detected in the nucleus, whereas RACK1 remains in the cytoplasm. These data suggest that RACK1 should not be part of the ISGF3 or ␥-activated factor complex and does not translocate to the nucleus. This and our previous finding that RACK1 is not tyrosine-phosphorylated (13) further support the concept that RACK1 is an adaptor or scaffold protein important in targeting specific signaling components such as STAT1 to the appropriate subcellular compartment for their activation.
It should be pointed out that the region of IFN␣R␤L that Immunoblotting was first performed with the antiphosphotyrosine (pTyr) antibody 4G10 (upper panel), followed by stripping and reblotting with the precipitating antibodies (lower panels; STAT1, STAT2, and Jak1). IP, immunoprecipitation; WB, Western blot. B, the ability of IFN to protect cells against the cytopathic effect (CPE) of encephalomyocarditis virus was determined using a standard antiviral assay. Cells were pre-incubated with concentrations of IFNs ranging from 1-500 units/ml in Dulbecco's modified Eagle's medium containing 2% fetal bovine serum for 18 h. The medium was removed and replaced with medium containing a dilution of encephalomyocarditis virus stock (1/10,000) that killed 100% of the cells in 24 h. Cell viability was determined 24 h later using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide. mu, murine.
interacts with RACK1 appears to overlap, at least in part, with the Box 2 domain. It has been proposed that this motif could play a role in the activation of Jak1 by the IL2R␤ chain (33). Our results suggest that the interaction between RACK1 and the Box 2 motif could be important for the recruitment of specific signaling proteins such as STAT1 and/or for providing the appropriate receptor configuration that allows Jaks to activate signaling components such as STATs. Unfortunately, the high levels of endogenous RACK1 make it extremely difficult to express potential dominant-negative mutants in which the interaction between RACK1 and STAT1 has been disrupted. Nevertheless, these experiments are important to determine whether the impaired tyrosine phosphorylation of STAT1 and STAT2 observed when RACK1 cannot interact with the receptor is due only to the failure to recruit STAT1 or to the fact that RACK1 may also recruit other signaling proteins. Either mechanism may explain the finding that RACK1 is also required for efficient phosphorylation of STAT2. Thus, the biological significance of RACK1 could go beyond the recruitment of STAT1. This is also suggested by the finding that the Box 2 motif is important in signaling by cytokine receptors in which STAT1 is not required for biological activity.
Finally, it should be pointed out that RACK1, as well as the ␤-subunit of G-proteins, binds pleckstrin homology domains, the SH2 domain of src, and protein kinase C ␤, raising the possibility that these or other proteins with similar motifs may be recruited by RACK1 to cytokine receptors. We are currently addressing the possibility that RACK1 recruits ubiquitously expressed proteins activated by type I IFNs such as insulin receptor substrate-phosphatidylinositol 3-kinase, Akt, and Fyn (34 -36) 2 to the receptor.