The Comparative Roles of Suppressor of Cytokine Signaling-1 and -3 in the Inhibition and Desensitization of Cytokine Signaling*

Negative feedback is a mechanism commonly employed in biological processes as a means of maintaining homeostasis. We have investigated the roles of suppressor of cytokine signaling (SOCS) proteins in regulating the kinetics of negative feedback in response to cytokine signaling. In mouse livers and bone marrow-derived macrophages, both interferon-γ (IFNγ) and interleukin-6 (IL-6) rapidly induced the tyrosine phosphorylation of signal transducer and activator of transcription-1 (STAT1) and STAT3. STAT3 tyrosine phosphorylation was bi-phasic in response to continuous IL-6 signaling. In macrophages lacking Socs3, however, continuous IL-6 signaling induced uniformly high levels of STAT3 tyrosine phosphorylation, and early IL-6-inducible genes were inappropriately expressed at intermediate time points. SOCS3 therefore imposes bi-phasic kinetics upon IL-6 signaling. Compared with Socs3 mRNA, Socs1 mRNA was induced relatively slowly, and SOCS1 simply attenuated the duration of IFNγ signaling. Surprisingly, heightened Socs1 mRNA expression but minimal STAT1 tyrosine phosphorylation was observed after prolonged stimulation with IFNγ, indicating that STAT1 may not play a large role in inducing Socs1 mRNA during steady-state IFNγ signaling. We also demonstrate that both SOCS1 and SOCS3 can desensitize primary bone marrow-derived macrophages to IFNγ and IL-6 signaling, respectively. Consistent with the kinetics with which Socs1 and Socs3 mRNAs were induced, SOCS3 desensitized cells to IL-6 rapidly, whereas SOCS1-mediated desensitization to IFNγ occurred at later time points. The kinetics with which SOCS proteins are induced by cytokine may therefore be a parameter that is “hard-wired” into specific cytokine signaling pathways as a means of tailoring the kinetics with which cells become desensitized.

Interleukin-6 (IL-6) 2 and interferon-␥ (IFN␥) are cytokines with key roles in regulating the immune response. Signaling by IL-6 and IFN␥ begins at the surface of the cell where the cytokines associate with their respective receptor complexes (1). The IFN␥-receptor complex consists of the ligand-binding IFN␥-receptor ␣ subunit and the signal-transduc-ing IFN␥-receptor ␤ subunit, whereas the IL-6-receptor complex consists of the ligand-binding IL-6-receptor ␣ subunit and the signal-transducing glycoprotein-130 subunit (1). Associated with the receptor subunits are the Janus kinases (JAKs), which become activated upon receptor dimerization, and phosphorylate signal transducers and activators of transcription (STAT) transcription factors (2,3). Both IFN␥ and IL-6 can activate JAK1 (4 -6) and JAK2 (7)(8)(9)(10), and subsequently STAT1 (11,12) and STAT3 (6,13,14). The phosphorylated STATs (P-STATs) then dimerize and are transported to the nucleus, where they bind to promoter sequences, and activate the transcription of a suite of genes, among which are those encoding the suppressor of cytokine signaling-1 (SOCS1) and SOCS3 proteins (15,16). STAT3 activates the transcription of Socs3 mRNA by associating with the 5Ј promoter region of the Socs3 gene (17). The regulation of Socs1 mRNA by STAT1 is poorly understood, and it remains to be clarified whether STAT1 directly activates the transcription of Socs1. Interferon regulatory factor-1, which is induced by STAT1, does, however, activate the transcription of Socs1 (18). SOCS1 and SOCS3 proteins can both inhibit JAK phosphorylation of STAT, thus creating a negative feedback loop that attenuates cytokine signal transduction, although the mechanisms by which they act appear to differ. Whereas SOCS1 functions by binding directly to JAK proteins, SOCS3 inhibits signaling by binding to phosphorylated tyrosine sites on the cytoplasmic domain of the receptor (19,20).
Although the signaling pathways activated by IFN␥ and IL-6 share common features, their biological outcomes are quite different. IFN␥ acts on cells to induce inflammatory and anti-viral responses, whereas IL-6 acts on cells to regulate the production of acute phase proteins (primarily in hepatocytes) as well as the growth and differentiation of a variety of cell types. Classically, this paradox has been explained by an understanding that IFN␥ primarily signals via STAT1, whereas IL-6 primarily signals via STAT3 (21). Physiologically, SOCS1 is the main inducible inhibitor of IFN␥ signaling and in neonatal mice prevents the establishment of a lethal inflammatory disease characterized by heightened levels of circulating IFN␥, fatty degeneration of the liver, and necrosis of the liver and other organs (22). Mice lacking Socs3 die during embryogenesis (23); however, the generation of mice conditionally deficient for Socs3 revealed SOCS3 to be a potent physiological suppressor of IL-6 but not IFN␥ signaling (24 -26). Surprisingly, IL-6 signaling in the absence of SOCS3 resulted in the prolonged phosphorylation of STAT1 and a qualitative shift in the transcriptional response of IL-6 toward that typical of IFN␥ (16,24,25). This indicates that signaling by IFN␥ and IL-6 is fundamentally similar, and that the activity of SOCS3 is required to sculpt the cellular response to IL-6 and thereby discriminate it from that of IFN␥.
In addition to their role as inhibitors of cytokine signaling, a role for SOCS proteins in the desensitization of cytokine signaling was recently postulated (27). Desensitization, as opposed to inhibition, refers to the process by which an initial signaling event causes the cell to become refractory to subsequent signals. In the case of IFN␥ signaling, the T-cell protein-tyrosine phosphatase has been shown to potently desensitize cells to repeated stimulation (28). Although SOCS1 and SOCS3 are attractive candidates as mediators of desensitization to IFN␥ and IL-6, respectively, detection of this activity has so far proved elusive (27).
In this study, we have explored the relationship between the kinetics of STAT tyrosine phosphorylation and the kinetics of Socs mRNA expression for both IFN␥ and IL-6 signaling. We demonstrate marked differences in the kinetics with which SOCS1 and SOCS3 inhibit IFN␥ and IL-6 signaling, respectively, and, furthermore, we show that both SOCS1 and SOCS3 can desensitize cells to IFN␥ and IL-6 signaling, respectively.

MATERIALS AND METHODS
Mouse Injections-C57BL/6 mice housed at The Walter and Eliza Hall Institute animal facility received an intraperitoneal injection at 6 -8 weeks of age with 100 l of either saline, recombinant murine IL-6 (500 g/kg, a gift from Dr. Richard Simpson, Ludwig Institute for Cancer Research, Melbourne, Australia), or IFN␥ (250 g/kg weight, Peprotech, Rocky Hill, NJ). Mice were asphyxiated in CO 2 , and livers were dissected and frozen in liquid nitrogen at various time points following injection. Experiments were performed with the approval of the Melbourne Health Research Directorate Animal Ethics Committee.
RNA Isolation and cDNA Synthesis-Total RNA for real-time PCR and microarray analysis was isolated from cells using the RNeasy kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. Single-stranded cDNA was synthesized from 20 g of total RNA (for microarray analysis) or Ͻ1 g of total RNA (for real-time PCR) in a 20-l reaction for 1 h at 42°C using 100 units of Superscript II (Invitrogen) and 5 M T7 oligo(dT) primer (synthesized by GeneWorks, Thebarton, SA, Australia) as described in the GeneChip Technical Manual (Affymetrix, Santa Clara, CA). For real-time PCR, cDNA products were diluted to 80 l in Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 8.0). Double-stranded cDNA was synthesized for microarray analysis from 20 l of single-stranded cDNA products according to Affymetrix methodology and was purified with Phase-Lock Gels (Eppendorf, Hamburg, Germany).
Real-time Quantitative PCR-Real-time quantitative PCR (Q-PCR) was carried out on single-stranded cDNA samples using the QuantiTect SYBR Green PCR kit (Qiagen) in 10-l reactions containing 0.5 pmol of forward and reverse primers, 5 l of QuantiTect Master Mix, and 4 l of cDNA. Amplification was performed according to Qiagen's recommendations with a 7900HT sequence detection system (Applied Biosystems, Foster City, CA). Primer sequences were as follows: porphobilinogen deaminase (Pbgd)-forward: CCTGGTTGTTCACTCCCTGA; Pbgdreverse: CAACAGCATCACAAGGGTTTT. Socs1-forward: GTGGT-TGTGGAGGGTGAGAT; Socs1-reverse: CCCAGACACAAGCTGC-TACA. Socs3-forward: TGAGCGTCAAGACCCAGTCG; Socs3reverse: CACAGTCGAAGCGGGGAACT. Ttp-forward: CCTCTGC-AACTCTGGTCTCC; Ttp-reverse: TTCTCAGGAGAGGCTTCGAC. Gadd45␥-forward: GCATCCTCATTTCGAATCC; Gadd45␥-reverse: CACCCAGTCGTTGAAGCTG. Expression was determined for the gene of interest relative to a standard curve created by serial dilution of the PCR product and was normalized against the expression of the housekeeping gene Pbgd. Microarray and real-time PCR analysis confirmed Pbgd to be stably expressed in livers and primary macrophages treated with IFN␥ or IL-6 (data not shown). Where actinomycin-D was used to inhibit RNA transcription, and it was not appropriate to normalize to the expression level of Pbgd, gene expression was instead normalized to the number of cells plated.
Microarray Time-course Analysis-Biotin-labeled cRNA was synthesized from double-stranded cDNA with the BioArray HighYield RNA Transcript Labeling Kit (Enzo, New York, NY). Samples were cleaned with the RNeasy kit (Qiagen), and 15 g of each cRNA sample was fragmented then hybridized overnight to MGU74Av2 Murine Gene-Chips (Affymetrix) following the manufacturer's instructions. Gene-Chips were stained with phycoerythrin-streptavidin (Molecular Probes, Eugene, OR) and scanned with an Affymetrix GeneChip scanner. Normalization and background correction was performed using the "Affy" package (www.bioconductor.org) of the "R" programming environment. Genes with a change in expression Ͼ1.7-fold between treated and control samples in two consecutive samples were considered to be differentially expressed. This cut-off was selected based on visual inspection of normalized quantile-quantile plots of the expression differences between pairs of microarrays. Genes displaying differential expression between control samples (either non-treated or PBS-treated) were removed. GeneCluster2 (31) 3 was then employed to identify sets of differentially expressed genes with common expression profiles.

RESULTS
Differential Induction of STAT1 and STAT3 Tyrosine Phosphorylation by IFN␥ and IL-6 in the Liver-Gene deletion studies in mice have revealed the importance of SOCS1 for regulating IFN␥ and SOCS3 for regulating IL-6 signaling, however little is known of the comparative induction of SOCS proteins by IFN␥ and IL-6. To investigate the in vivo activation of STAT1 and STAT3 and expression of Socs1 and Socs3 in response to these cytokines, we injected C57BL/6 mice with IFN␥ or IL-6 and examined proteins in the livers of mice after various time periods. Immunoprecipitation (9) and Western blotting demonstrated that, in response to either cytokine, tyrosine phosphorylation (a readout of STAT activation) of both STAT1 and STAT3 was maximal or nearmaximal by 15 min after injection (Fig. 1, A and B). STAT3 tyrosine phosphorylation declined to near basal levels after 4 h in response to either cytokine, and a much higher level was detected following injection of IL-6 compared with injection of IFN␥ (Fig. 1A). In contrast, while the maximal levels of STAT1 tyrosine phosphorylation were similar following injections of both cytokines, phosphorylation declined to basal levels after 4 h in response to IFN␥ but after only 1 h in response to IL-6 ( Fig. 1B). These differences in the phosphorylation of STAT1 and STAT3 by IFN␥ and IL-6 are consistent with the role of STAT1 as an important mediator of the response to IFN␥ and the role of STAT3 as an important mediator of the response to IL-6 (21).
IFN␥ induced the robust expression of SOCS1 protein, with levels reaching a maximum at 2-4 h after injection of IFN␥, at a time when STAT1 and STAT3 phosphorylation was waning ( Fig. 1C; compare with Fig. 1, A and B). SOCS1 protein expression was maintained at a high level for at least 8 h, and declined to near basal levels by 16 h. In contrast, whereas the kinetics of induction of SOCS1 by IL-6 were similar to those induced by IFN␥, the magnitude was greatly reduced. In the case of SOCS3, the opposite appeared true; IL-6 induced the expression of SOCS3 protein very rapidly, with maximal levels detected 30 min to 1 h following injection, at a time that correlates with a decline in STAT1/STAT3 tyrosine phosphorylation ( Fig. 1D; compare with Fig. 1,  A and B), whereas IFN␥ induced a barely detectable amount of SOCS3 (Fig. 1D).
We confirmed the pattern of SOCS1 and SOCS3 protein expression in response to IFN␥ and IL-6 at the mRNA level using a Q-PCR assay (Fig. 1E). Again, IFN␥ induced Socs1 to greater levels than did IL-6, whereas IL-6 induced Socs3 to greater levels than did IFN␥. IL-6 induced both Socs1 and Socs3 mRNA to a similar extent, but more SOCS3 protein than SOCS1 protein, indicating that synthesis of SOCS3 protein may be more efficient than synthesis of SOCS1 protein. The kinetics of Socs3 mRNA and SOCS3 protein expression were similar, suggesting that the SOCS3 protein is highly labile, a finding that is consistent with the reported half-life of this protein (33). Expression of SOCS1 protein, however, appeared to persist at 4 h and 8 h, despite declining levels of Socs1 mRNA, suggesting that the half-life of SOCS1 during active cytokine signaling in vivo is substantially greater than that of SOCS3.
Continuous IL-6 Signaling in Bone Marrow-derived Macrophages Produces a Bi-phasic Response-The in vivo analysis of signaling confirmed the reciprocal relationship of IFN␥ with SOCS1, and IL-6 with SOCS3. However, the intrinsic variability when working with mice, and the difficulty in precisely regulating cytokine levels following injection led us to investigate the kinetics of signaling in more tractable systems: cultures of primary cells. We initially examined IL-6 signaling in Socs3deficient primary hepatocytes, but like Socs3-deficient mouse embryonic fibroblasts (27), these cells exhibited signs of basal STAT3 activation (albeit to a lesser extent), including low level STAT3 tyrosine phosphorylation and induction of IL-6-inducible genes (data not shown). Fortunately, Socs3-deficient primary bone marrow-derived macrophages did not show signs of basal STAT3 activation ( Fig. 2A,  lane 10).
In bone marrow-derived macrophages stimulated with IL-6, tyrosine phosphorylation of STAT3 was observed during two distinct phases in response to IL-6 ( Fig. 2, A and B). The first phase occurred between 15 and 30 min, and the second from 120 min onward. The decline in STAT3 phosphorylation from the initial peak at 15 min was rapid, and near-basal P-STAT3 levels were achieved by 60 min, suggesting that the half-life of STAT3 phosphorylation must be very short during this period. Expression of Socs3 mRNA was maximal at 30 min, 15 min after maximal STAT3 activation was observed (Fig. 3A). This was followed by a rapid reduction of Socs3 mRNA expression, with minimal expression observed between 60 to 90 min, and subsequent re-establishment of expression by 180 min.
SOCS3 Is Responsible for Bi-phasic Tyrosine Phosphorylation of STAT3 and Expression of Genes in Response to IL-6-The observation of two separate phases of signaling by IL-6 in bone marrow-derived macrophages suggested that negative feedback by SOCS3 might impose a period of oscillation upon signaling by IL-6. To test this hypothesis, levels of phosphorylated STAT3 were determined in response to IL-6 in Socs3 Ϫ/⌬ bone marrow-derived macrophages (Fig. 2, A and B). In the absence of Socs3, STAT3 phosphorylation was strongly induced after 15 min and was not substantially reduced for the remainder of the time course. SOCS3 is therefore responsible for imposing bi-phasic kinetics upon the signaling of IL-6.  To determine how the expression levels of IL-6-responsive genes are affected by SOCS3-mediated regulation of IL-6 signaling, the responses of the IL-6-inducible genes growth arrest and DNA damage-inducible gene 45␥ (Gadd45␥) (34) and tristetraprolin (Ttp) (35), were examined following stimulation with IL-6 in bone marrow-derived macrophages. In wild-type cells, the response of Gadd45␥ mRNA expression to IL-6 was bi-phasic, with peaks of expression occurring at 30 and 180 min. In the absence of Socs3, however, Gadd45␥ mRNA expression was sustained at a high level in response to IL-6 ( Fig. 3B). Expression of Ttp mRNA in wild-type cells peaked at 30 min and was sustained at a low level for the remainder of the time course (Fig. 3C). In Socs3 Ϫ/⌬ cells, however, stimulation with IL-6 resulted in heightened levels of Ttp mRNA expression for the duration of the time course. These results suggest that loss of Socs3 results in the prolonged expression of genes that are normally induced by IL-6 at early time points and in the heightened and uniform expression of genes that are normally induced by IL-6 with bi-phasic kinetics.
Disparate Kinetics of STAT1 Tyrosine Phosphorylation and Socs1 Expression in Response to Continuous IFN␥ Signaling-Although the role of SOCS1 in regulating the kinetics of STAT1 tyrosine phosphorylation in response to IFN␥ has been explored previously (30), it is unknown how the kinetics with which SOCS1 and SOCS3 regulate cytokine signaling compare. We therefore examined the tyrosine phosphorylation of STAT1 and expression of Socs1 mRNA in response to IFN␥ in bone marrow-derived Ifn␥ Ϫ/Ϫ and Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ macrophages. Consistent with previous observations (30), STAT1 tyrosine phosphorylation in response to IFN␥ was prolonged in the absence of SOCS1 (Fig. 4A). The effects of SOCS1 were not apparent until ϳ90 min after stimulation with IFN␥, ϳ60 min later than the effects of SOCS3 were first observed in response to IL-6 (compare Fig. 2, A and B, with Fig. 4, A and B). The pattern of Socs1 mRNA expression in Ifn␥ Ϫ/Ϫ cells was quite different to the pattern of STAT1 phosphorylation (Fig. 4, compare B with C). Whereas STAT1 tyrosine phosphorylation occurred rapidly and declined to near-basal levels by 240 min, expression of Socs1 mRNA was induced slowly, but remained at near-maximal levels from 60 min until at least 480 min. This was surprising, given that, after 240 min, STAT1 tyrosine phosphorylation remained low for the duration of the time course. These results seem to differ from the response to injection of IFN␥ that is observed in whole livers (Fig. 1, B and E), probably because the livers were taken from mice subjected to a single stimulation of IFN␥, whereas IFN␥ signaling in the bone marrowderived macrophages was continuous.
Socs1 and Socs3 mRNAs Are Degraded at Similar Rates-One possible explanation for the discrepancy between STAT1 phosphorylation and Socs1 mRNA expression after 240 min of IFN␥ signaling could be that Socs1 mRNA is degraded relatively slowly. To test this, we compared the stability of Socs1 mRNA with the stability of Socs3 mRNA, which presumably has a short half-life because it exhibits rapid bi-phasic kinetics in response to IL-6. As shown in Fig. 5A, actinomycin-D efficiently suppresses synthesis of Socs3 mRNA in response to IL-6. To assess the rates at which Socs1 mRNA and Socs3 mRNA are degraded, cells were stimulated with either IFN␥ (at 0 min) or IL-6 (at 30 min) and then treated with actinomycin-D (at 60 min). No difference was observed in the kinetics with which mRNAs for Socs1 and Socs3 are degraded in the absence of transcription (Fig. 5B). By 120 min (60 min after treatment with actinomycin-D), levels of both Socs1 mRNA and Socs3 mRNA were ϳ25% of the maximal level, indicating that the mRNAs of both Socs1 and Socs3 have a half-life of ϳ30 min (Fig. 5B). Because this value is relatively short, we conclude that a high rate of transcription must be required to sustain the levels of Socs1 mRNA shown in Fig. 4C at between 240 and 480 min of continuous IFN␥ signaling. Because the corresponding levels of STAT1 phosphorylation were minimal (Fig. 4, A and B), it seems likely that some other IFN␥responsive transcription factor (such as interferon regulatory factor-1) is responsible for the majority of Socs1 mRNA transcription observed at these points.
Uniform Regulation by SOCS1 of the Transcriptional Response to IFN␥ in the Liver-Given the dramatic inflammatory response to IFN␥ of mice deficient for Socs1 (22), we were interested in determining how SOCS1 regulates the response to IFN␥ at a genome-wide level, and in particular, whether distinct sets of genes respond differently to the loss of SOCS1 following stimulation with IFN␥. Sets of IFN␥-inducible genes with abnormally dramatic responses to the loss of Socs1, for instance, could provide additional insight into the development of the inflammatory disease to which neonatal Socs1 Ϫ/Ϫ mice succumb (22). Microarrays were used to profile the transcriptional responses of livers from Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ and Ifn␥ Ϫ/Ϫ mice subjected to various durations of stimulation with IFN␥, and the expression profiles of genes found to be differentially regulated due to loss of Socs1 were clustered into six common groups using GeneCluster2 (31). 3 The number of cluster groups used, six, was determined through a trial-and-error approach to be the minimum number of groups required for the segregation of genes with visually distinct kinetic profiles. Heat plots were then generated to visualize the average transcriptional profile of each group of genes. Importantly, changing the parameters specifying the number of clusters recognized by GeneCluster2 made little difference to the relationships observed between the responses to IFN␥ of livers from Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ and Ifn␥ Ϫ/Ϫ mice.
Heat plots for the response to IFN␥ of genes in Ifn␥ Ϫ/Ϫ livers (relative to a 0-h, non-stimulated control) revealed that initial induction occurs at 1-2 h and that expression was sustained until between 12 and 16 h ("␥ Ϫ/Ϫ -0 h" heat plots, Fig. 6, A-C). Heat plots were also generated to compare the responses to IFN␥ of Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ livers with the response to IFN␥ of Ifn␥ Ϫ/Ϫ livers (this is the expression comparison by which genes were clustered). The time and intensity at which the IFN␥inducible genes were initially induced by IFN␥ did not appear to be affected by loss of Socs1 ("S1␥ Ϫ/Ϫ -␥ Ϫ/Ϫ " heat plots, Fig. 6, A-C; compare with "␥ Ϫ/Ϫ -0 h" heat plots). Loss of Socs1 did, however, result in the heightened expression of IFN␥-responsive genes at later time points. Consistent with the previously reported role of SOCS1 in regulating the duration of STAT1 activation (30), the expression of genes grouped into clusters A and B was prolonged in the absence of SOCS1; it is uncertain whether the same is true for the genes grouped into cluster C, because sufficient discriminatory time points were not available. Somewhat surprisingly, the duration over which the effects of SOCS1 were observed in each of the groups of IFN␥-inducible genes appeared to be markedly uniform, suggesting that different sets of IFN␥-inducible genes are not differentially regulated by the actions of SOCS1, at least in response to a single injection of IFN␥.
Three groups of genes with variable expression profiles were also identified by GeneCluster2 (data not shown). Several of the genes included in these groups were later determined by Q-PCR analysis to be regulated in the liver with a high degree of variability and not specifically in response to IFN␥ (data not shown).

SOCS3 Desensitizes Bone Marrow-derived Macrophages to
Repeated Stimulation with IL-6-SOCS proteins are tempting candidates as mediators of cytokine desensitization. However, previous attempts to detect SOCS3-mediated desensitization of cytokine signaling in Socs3-deficient mouse embryonic fibroblasts were hampered by the constitutive activation of STAT3 in these cells (27). The kinetics with which IL-6 and IFN␥ induced the tyrosine phosphorylation of STAT proteins and expression of Socs mRNAs suggested that both SOCS1 and SOCS3 might be able to desensitize cells to repeated stimulation with cytokine. Furthermore, given the differences in the kinetic relationships between STAT activation and Socs mRNA expression, it appeared that SOCS1 may be substantially more potent at desensitizing cells to repeated IFN␥ stimulation than SOCS3 would be at desensitizing cells to repeated IL-6 stimulation. To determine whether SOCS3 can desensitize IL-6 signaling, we employed the experimental approach taken by Fischer et al. (27), in which cells are subjected to a pulse of cytokine, rested for various periods of time, and then subjected to a second pulse prior to analysis.
In C57BL/6 bone marrow-derived macrophages, a 15-min pulse of IL-6 resulted in the tyrosine phosphorylation of STAT3 at 15 and 30 min (Fig. 7A). In the absence of Socs3, tyrosine phosphorylation of STAT3 in response to a 15-min pulse of IL-6 was extended until at least 90 min, with low level phosphorylation also apparent at later time points (Fig.  7A). To determine whether a pulse of IL-6 could desensitize cells to subsequent IL-6 signaling, cells were stimulated again after various periods of time had elapsed. Consistent with the previous observations of Fischer et al., pre-treatment of C57BL/6 cells with IL-6 resulted in the desensitization of cells to a further pulse of IL-6 after 60, 90, 120, and 240 min (Fig. 8A). There appeared to be some alleviation of the desensitization at 90 and 120 min, although this was not consistent. A similar pattern of pronounced desensitization was not observed in cells deficient for Socs3, indicating that SOCS3 desensitizes cells to IL-6 signaling (Fig. 8A). Limited desensitization to IL-6 was, however, apparent for all re-stimulated Socs3-deficient cells, and was more pronounced at later time points, indicating that factors other than SOCS3 are required for long term desensitization to IL-6 (Fig. 8A). and Socs3 mRNA (shaded boxes). B, 2.5 ϫ 10 5 C57BL/6 bone marrow-derived macrophages were stimulated with IFN␥ (for 60 min) or IL-6 (for 30 min) prior to the addition of actinomycin-D (at 60 min). Levels of Socs1 or Socs3 mRNA were determined by Q-PCR for cells stimulated with IFN␥ or IL-6, respectively. FIGURE 6. Cluster analysis of the role of SOCS1 in regulating IFN␥-inducible gene expression. Ifn␥ Ϫ/Ϫ and Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ mice were injected with IFN␥ or PBS (to provide controls), and livers were taken at various times afterward for transcriptional profiling by microarray. Data have been deposited in the Gene Expression Omnibus (GEO, www.ncbi.nlm.nih.gov/geo/), and have the GEO Series accession id GSE4232. Genes that were differentially expressed in the livers of IFN␥-injected Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ mice relative to the livers of IFN␥-injected Ifn␥ Ϫ/Ϫ mice (at the same time point after injection) were clustered into sets of commonly differentially expressed genes using GeneCluster2. While only a single liver was used for each time point, differential expression was defined as relative induction or suppression of expression over at least two consecutive time points. Average expression profiles of clusters of genes are displayed as heat plots, where red represents high relative expression, and blue represents low relative expression. Heat plots labeled "␥ Ϫ/Ϫ -0 h" depict the response of livers from IFN␥-injected Ifn␥ Ϫ/Ϫ mice, relative to the liver of a non-injected control. Heat plots labeled "S1␥ Ϫ/Ϫ -␥ Ϫ/Ϫ " depict the response of livers from IFN␥-injected Ifn␥ Ϫ/Ϫ Socs1 Ϫ/Ϫ , relative to the livers of IFN␥-injected Ifn␥ Ϫ/Ϫ mice.

SOCS1 Desensitizes Bone Marrow-derived Macrophages to Repeated
Stimulation with IFN␥-Having found that the induction of SOCS3 by IL-6 could desensitize cells to subsequent signaling by IL-6, we wondered whether the induction of SOCS1 by IFN␥ could desensitize cells to subsequent signaling by IFN␥. In response to a 15-min pulse of IFN␥, STAT1 tyrosine phosphorylation was observed until 60 min in Ifn␥ Ϫ/Ϫ cells, and until 360 min in Socs1 Ϫ/Ϫ Ifn␥ Ϫ/Ϫ cells (Fig. 7B). Desensitization of Ifn␥ Ϫ/Ϫ macrophages (as measured by STAT1 tyrosine phosphorylation) following 15 min of IFN␥ signaling was observed after 120 min and until 360 min (Fig. 8B). In Socs1 Ϫ/Ϫ Ifn␥ Ϫ/Ϫ macrophages, desensitization to IFN␥ occurred after 240 min, probably a result of the activity of the phosphatase SHP-2 (36). Thus, SOCS1 clearly desensitizes cells to re-stimulation with IFN␥ at 120 min; however, it is difficult to clarify the specific role of SOCS1 in desensitizing cells to IFN␥ at later times. Consistent with the relative kinetics of SOCS1 and SOCS3-mediated inhibition of cytokine signaling (as determined above), desensitization of cytokine signaling was mediated much more rapidly by SOCS3 than by SOCS1 (compare Fig. 8A with 7B).

DISCUSSION
Despite the physiologically distinct functions of IFN␥ and IL-6, the signaling mechanisms by which they elicit their cellular responses are similar, and in the case of IL-6 are distinguished at least in part through the activity of SOCS3 (16,24,25). To further clarify the importance of SOCS proteins in regulating signaling by IFN␥ and IL-6, we compared the roles of SOCS1 and SOCS3 in regulating the kinetics of cytokine FIGURE 7. Kinetics of STAT3 and STAT1 tyrosine phosphorylation in response to pulsed IL-6 or IFN␥. A, C57BL/6 and Socs3 Ϫ/⌬ bone marrow-derived macrophages were stimulated with IL-6 and washed in MTPBS after 15 min. Cells were then left in fresh media and lysed at the indicated times. Lysates were analyzed by immunoprecipitation followed by Western blotting for levels of tyrosinephosphorylated STAT3 (P-STAT3), and total STAT3. B, Ifn␥ Ϫ/Ϫ and Socs1 Ϫ/Ϫ Ifn␥ Ϫ/Ϫ bone marrow-derived macrophages were stimulated with IFN␥ and washed in MTPBS after 15 min. Cells were then left in fresh media and lysed at the indicated times. Lysates were analyzed by immunoprecipitation followed by Western blot for levels of tyrosinephosphorylated STAT1 (P-STAT1), and total STAT1. signal transduction. Consistent with the known physiological activities of SOCS proteins, Socs1 mRNA and SOCS1 protein were induced to a greater extent by IFN␥ than by IL-6, whereas Socs3 mRNA and SOCS3 protein were induced more by IL-6 than by IFN␥. Both IFN␥ and IL-6 rapidly induced the activation of STATs, and in the absence of Socs1 and Socs3, respectively, continuous cytokine stimulation resulted in the sustained activation of STATs.
In C57BL/6 bone marrow-derived macrophages, regulation of continuous IL-6-signaling by SOCS3 was bi-phasic, with the first peak of IL-6 signaling appearing rapidly and transiently. Bi-phasic STAT3 activation has been reported in response to a variety of stimuli, including granulocyte-colony stimulating factor (37), leukemia inhibitory factor (38), IL-6 (27,38), and lipopolysaccharide (39). Interestingly, each of these stimuli activates the glycoprotein-130/STAT3 or the granulocytecolony stimulating factor-receptor/STAT3 signaling pathways, which are known to be inhibited by SOCS3 in vivo (24 -26, 29, 40, 41). To ascertain whether negative feedback by SOCS3 is necessary for the biphasic response to IL-6, we examined IL-6 signaling in Socs3-deficient macrophages, and found the response to IL-6 of these cells to be maximal and constant. Therefore, SOCS3 is required for suppression of the early phase of IL-6 signaling. It follows that the second phase of IL-6 signaling seen in wild-type cells is probably due to the degradation of the SOCS3 protein synthesized during the first phase of signaling, and the subsequent re-establishment of STAT3 activation (a model for this is shown in Fig. 9A).
As was the case for STAT3 in response to IL-6, tyrosine phosphorylation of STAT1 was detected in Ifn␥ Ϫ/Ϫ macrophages immediately after stimulation with IFN␥. Yet despite the rapid phosphorylation of STAT1, the induction of Socs1 mRNA was slow. In contrast to the bi-phasic manner in which IL-6 signaling was regulated by SOCS3, IFN␥ signaling was simply attenuated in duration by SOCS1, a finding consistent with previous observations (30). Using microarray expression profiling of an IFN␥ time course, all genes regulated by SOCS1 were regulated with uniform kinetics, suggesting that separate sets of genes are not differentially regulated by SOCS1.
The steady-state outcomes of signaling by IFN␥ and IL-6 were remarkably dissimilar. After prolonged IL-6 signaling, medium levels of STAT3 activation, Socs3 mRNA expression, and presumably SOCS3 protein expression were observed. Surprisingly, prolonged IFN␥ signaling resulted in high levels of Socs1 mRNA expression, and low levels of STAT1 tyrosine phosphorylation. We calculated the half-life of Socs1 mRNA to be ϳ30 min, which is clearly too short to account for the substantial duration over which heightened Socs1 mRNA expression but minimal STAT1 activation was observed (from 240 to 480 min). This suggests that an IFN␥-inducible transcription factor other than STAT1 may be responsible for sustaining Socs1 mRNA expression after prolonged IFN␥ signaling. One likely candidate is interferon regulatory factor-1, because it is induced by STAT1 (32,42,43) and can associate with the Socs1 promoter to activate the transcription of Socs1 mRNA (18). A model for the proposed relationship between STAT1 activation, Socs1 mRNA expression, and SOCS1 protein expression, is given in Fig. 9B.
We demonstrated that both SOCS3 and SOCS1 desensitized bone marrow-derived macrophages to signaling by IL-6 and IFN␥, respectively, with the activity of SOCS1 occurring later than the activity of SOCS3. This is consistent with the finding that IL-6 induces the expression of Socs3 mRNA earlier than IFN␥ induces the expression of Socs1 mRNA. As such, the kinetics with which cytokines induce SOCS proteins appears to be a parameter that is tailored to different cytokines to ensure that cells are desensitized to the cytokine with appropriate kinet-ics. Why cells need to be desensitized to IL-6 earlier than they are to IFN␥ is not currently apparent, although it is clear that this mechanism also allows cells to become re-sensitized to IL-6 more quickly than to IFN␥. It could be speculated that IFN␥ signaling is more dangerous to the health of the cell and the organism than is IL-6 signaling, so the desensitization of cells to IFN␥ may be more critical. Consistent with this idea, desensitization in macrophages lacking Socs genes was more pronounced for IFN␥ signaling than for IL-6 signaling. This result also highlighted the important role that factors other than SOCS1, such as T-cell protein-tyrosine phosphatase for instance (28), also have in desensitizing cells to IFN␥.
In conclusion, both SOCS1 and SOCS3 are able to desensitize cells to signaling by IFN␥ and IL-6, respectively. The kinetics with which SOCS proteins are induced by cytokine is a critical factor in determining how the cell responds. IFN␥ appears to induce SOCS1 protein expression relatively slowly, leading to slow inhibition and desensitization of signaling. In contrast, IL-6 appears to induce SOCS3 protein expression rapidly, resulting in bi-phasic IL-6 signaling, and relatively rapid desensitization. It should be interesting to see how the kinetics with which SOCS proteins are induced determines the kinetics of desensitization for cytokines other than IFN␥ and IL-6. FIGURE 9. A kinetic model of IFN␥ and IL-6 signal transduction dynamics for STAT activity, Socs mRNA expression, and SOCS protein expression. A, at the initiation of IL-6 stimulation, SOCS3 protein is not present, and STAT3 tyrosine phosphorylation rapidly attains a maximal level. After ϳ1 h, the large concentration of phosphorylated STAT3 in the cell leads to a maximal induction of Socs3 mRNA and the accumulation of SOCS3 protein. As the concentration of SOCS3 increases to a maximal level, maximal signal attenuation, including degradation of phosphorylated STAT3 ensues. After 2 h, Socs3 mRNA concentrations decline in the absence of STAT3 transcriptional activity, and SOCS3 protein concentration begins to decline. After 3 h, the concentration of SOCS3 protein is sufficiently reduced to allow IL-6 signaling to resume. Concentrations of tyrosine-phosphorylated STAT3 and Socs3 mRNA again accumulate, eventually reaching a steady-state relationship with the concentration of SOCS3 protein. B, at the initiation of IFN␥ stimulation, SOCS1 protein is not present, and STAT1 tyrosine phosphorylation rapidly attains a maximal concentration. Over 2-3 h, Socs1 mRNA and SOCS1 protein accumulate, and STAT1 tyrosine phosphorylation is reduced to low levels. STAT1 tyrosine phosphorylation and Socs1 mRNA expression are maintained at these levels until at least 4 h. Because the half-life of Socs1 mRNA is fairly short, an IFN␥-inducible transcription factor other than STAT1 (such as interferon regulatory factor-1) is probably responsible for the transcription of Socs1 mRNA during this stage of the response to IFN␥.