The molecular basis for functional plasticity in type I interferon signaling

Feature Review

Focus: Type I interferons

The molecular basis for functional plasticity in type I interferon signaling Gideon Schreiber1 and Jacob Piehler2 1 2

Department of Biological Chemistry, Weizmann Institute of Science, Rehovot 76100, Israel Department of Biology, University of Osnabru¨ck, 49076 Osnabru¨ck, Germany

Type I interferons (IFNs) are best known for their role in innate immunity, but they are also involved in other functions including immunomodulation, restricting proliferation, cancer surveillance, and the regulation of the adaptive immune response. All these responses are mediated through the interaction with a single cell surface receptor, albeit at different ligand and receptor concentrations, ligand subtypes, and time of activation. Here we review the functional plasticity of IFN signaling from a quantitative perspective, showing how variations in different ingredients of the system lead to differential IFN responses and how cells tune the system to maximize efficiency while minimizing detrimental effects. We present a basic model wherein the integrated action of different feedback mechanisms can provide sufficient temporal control to differentially drive cellular decisions. IFNs as a paradigm of cytokine receptor plasticity Cytokine surface receptors recognize signals from the outside and transmit them into the cell, driving a pleiotropy of cellular responses. Signal transmission across the plasma membrane is a result of the cytokine interacting with the extracellular domains of the receptor subunits, placing them in proximity to one another and/or driving conformational changes [1]. Receptor dimerization results in the activation of cytosolic associated Janus family tyrosine kinases (JAKs), which in turn initiate downstream signaling cascades that propagate the signal into the nucleus and regulate gene transcription, mainly via signal transducer and activator of transcription (STAT) proteins [2,3]. The translation of cytokine recognition by the receptor into cellular decisions is highly plastic and in many cases not only depends on the cellular context but also on the molecular determinants of the cytokine–receptor interaction [3]. Thus, numerous instances of differential cellular responses elicited by different cytokines through the same cell surface receptor have been reported [3,4]. Here we focus on providing a critical review of the plasticity of type I IFN signaling, which is encoded by the binding of different IFNs to a shared cell surface receptor comprising two subunits: IFNAR1 and IFNAR2 (Figure 1A) [5]. This, in turn, results in the activation and Corresponding authors: Schreiber, G. ([email protected]); Piehler, J. ([email protected]). 1471-4906/ ß 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.it.2015.01.002

repression of over 2000 genes with a multitude of biological functions [6]. We are particularly intrigued by the nonlinear signaling, where a seemingly similar input (ligand– receptor interaction) promotes a multitude of biological outputs that are activated by different signaling cascades. Some of these outputs have key functions in innate and adaptive immune responses while others have been shown to be detrimental to health, apparently due to cytotoxicity and induced inflammation [7]. Thus, the tight control of this system is of foremost importance. New insights into this system provide a basic model where the actions of known activators, inhibitors, and feedback loops offer sufficient control of the system. The molecular determinants of IFN receptor assembly The human type I IFN system comprises 17 different ligands binding to two receptor subunits. The ligands include 13 IFNa subtypes, IFNb, IFNv, IFNk, and IFNe. IFNe is a unique case as it is a signaling molecule during pregnancy and it is not discussed further here [8]. The various IFNas share 80% sequence homology between them and IFNv and IFNb have 50% and 30% homology, respectively. The divergence of multiple IFNa subtypes occurred separately in different species, supporting the concept that different vertebrate groups have independently expanded their IFN subtypes [9,10]. All type I IFNs bind the receptor subunits at the same location, producing structurally highly similar ternary ligand–receptor complexes [11–14]. The main difference between the various IFN subtypes lies in their binding affinity toward the receptor subunits (Figure 1B). In general, the binding affinity to IFNAR2 is high (of the order of nanomolar affinity) and to IFNAR1 it is low (micromolar affinity) [15–17]. Binding affinity, however, varies by about 1000-fold between the different ligands. The weakest binder to IFNAR2 is IFNa1 and the tightest is IFNb (200 nM versus 0.2 nM affinity) [16,18]. All IFNas bind IFNAR1 with similar low affinities of 1–5 mM, while IFNb binds significantly more tightly, at 50 nM affinity [17,18]. Despite some allosteric crosstalk [19], the binding affinities of IFNs to each receptor subunit are hardly influenced by the interaction with the other receptor subunit [20]. IFNb therefore has overall a substantially higher ‘integral’ affinity to the cell surface receptor compared with all other natural IFNs [18,21,22], which has been suspected to play a decisive role in the pronounced differential cellular activities of IFNb. We have succeeded in reproducing the entire range of binding affinities of natural Trends in Immunology, March 2015, Vol. 36, No. 3

139

Feature Review

Trends in Immunology March 2015, Vol. 36, No. 3

(A)

(B) SD1 IFNAR1 IFNα IFNAR2

SD2

IFNα1

IFNα2

IFNβ

YNS

IFNAR1 750 IFNAR2 220 Anviral 2 Anproliferave 50000

2000 2 0.5 2000

100 0.4 0.2 50

40 2 0.2 20

D1

SD3

SD1 IFNAR1

SD1 IFNAR1 SD2 SD3 SD4

SOCS1

IFNAR2 SD2 D1 SD3

IFNAR2 D1 SD3

D2

D2

D2

SD1 IFNAR1 IFNAR2 SD2 D1 SD3 D2

SD4

nM nM pM pM IFNAR2 D1

SD4

D2

D2

SD4

Tyk2 P P P

SD2

SD4

ND 0.4 2000 ND

SD1 IFNAR1

SD1 IFNAR1

IFNAR2 SD2 D1 SD3 SD4

IFN– 1ant

Jak1 P USP18 STAT2 P STAT1 P

STAT2 STAT1

IRF9

STAT2 IRF9 STAT1

Cytoplasm Nucleous

STAT1 STAT1

Gene inducon GAS

ISRE

STAT1 STAT1

ISGs Robust genes

STAT2 IRF9 STAT1

Engaged Receptor Time for Cell type IFN engagement acvaon specific α, ω, β

Low

Fast

No

Mainly β

High

Slow

Yes

+ ??

Tunable genes

STAT2 IRF9 STAT1 TRENDS in Immunology

Figure 1. Signaling by type I interferons (IFNs). (A) IFN complex formation leads to phosphorylation of the associated Janus family tyrosine kinases (JAKs), which results in the recruitment and phosphorylation of signal transducer and activator of transcription (STAT) proteins. These, together with IFN regulatory factor 9 (IRF9), form complexes that translocate into the nucleus where they act as transcription factors for IFN-induced genes. Suppressor of cytokine signaling (SOCS) and ubiquitin-specific protease 18 (USP18) are negative regulators of the system. (B) Affinity of binding and antiviral and antiproliferative potencies of three natural and two engineered interferons – H57Y, E58N, and Q61S (YNS) and IFN-1ant [25] – are shown. The IFN-induced signal in the nucleus is driving robust and tunable responses.

IFNs by mutating IFNa2, showing that the variation in binding affinity accounts for much of the variation of different natural IFNs[12,23]. In cell-based assays, IFN activity is most often probed by its ability to induce JAK/STAT phosphorylation and gene induction, as well as antiviral and antiproliferative cellular responses. The antiviral state is observed in all cell lines on activation for a few hours by picomolar IFN concentrations and is maintained even after subsequent removal of the IFN [24]. We therefore designated the antiviral response a ‘robust’ activity of IFNs (Figure 1B). By contrast, the antiproliferative activity of IFNs is celltype specific, requires about 1000-fold higher interferon concentrations, is activated at physiological IFN concentrations only by IFNb, requires continuous receptor activation over prolonged periods (days), and is susceptible to the concentration of cell surface receptors [25,26]. Such IFN activities, which strongly depend on the cellular context, were designated ‘tunable’. Studies of various natural IFNs and IFN mutants have shown a clear linear correlation between binding affinity and antiproliferative activity, as long as the affinity toward IFNAR1 is not reduced below approximately 100 mM [18,25,27]. Antiviral activity, by contrast, is promoted even with extremely weak-binding IFN mutants that do not promote antiproliferative activity. In contrast to antiproliferative activity, the antiviral 140

potency of IFNs is maximized already at medium receptor binding affinity (possessed by most IFNas), with the higher binding affinity of IFNb hardly increasing the antiviral potency above that of IFNa2 or even IFNa1 (Figure 1B) [25,26]. The IFN receptor subunits comprise an extracellular domain containing four and two fibronectin type III domains for IFNAR1 and IFNAR2, respectively, a single transmembrane helix, and a cytoplasmic domain of 100 and 251 amino acids for IFNAR1 and IFNAR2, respectively, that is mostly unstructured (Figure 1A). The structures of the isolated extracellular portions of the receptors and IFNs, as well as the receptors bound to IFNv, IFNa2, and IFNb, have been solved to high resolution [11,12,28]. The recognition of IFNs by IFNAR1 and IFNAR2 has been recently reviewed [29]. To summarize, the high-resolution structures revealed that different type I IFNs bind the receptors at the same location, resulting in a very similar ternary structure independent of the ligand subtype. The variation in amino acid sequence between the different type I IFNs represents a multitude of parallel solutions that support binding at various affinities [12]. Comparing the unbound with the bound structure reveals rearrangements of the membrane-proximal domains of the receptors in relation to the membrane-distal domains on ligand binding. This reordering is particularly obvious for IFNAR1 [12], where a

Feature Review

Trends in Immunology March 2015, Vol. 36, No. 3

first-order rate constant of approximately 1 s 1 was determined [30]. The role of the observed structural rearrangement on the biological activity of the receptors is currently under investigation. Alanine scanning of the IFN–IFNAR2 interface has identified a classical arrangement of binding energies found in protein–protein interfaces, with the hotspot residues in the center surrounded by less important residues and the peripheral residues not contributing significantly to binding [31–33]. Interestingly, the most varied sequence in IFNa is the C-terminal unstructured tail, where little homology is observed between the subtypes. Replacing the tail of IFNa2 with that of IFNa8 increased its binding affinity and biological activity by approximately 15-fold [34]. Alanine scanning of the low affinity IFN–IFNAR1 interface did not reveal any hotspot, with several mutations resulting in a decrease of three- to fivefold in binding affinity [35]. However, replacing Arg120 on IFNa2 with glutamic acid almost completely eliminates binding to IFNAR1 [36]. Surprisingly, mutation of any one of the three clustered amino acids His57, Glu58, and Gln61 to alanine increases binding by about threefold, with the triple HEQ mutant increasing binding by 20-fold [23]. Further optimization using phage display generated the triple mutation H57Y, E58N, and Q61S (YNS), which binds IFNAR1 60-fold tighter than IFNa2 and three fold tighter

than IFNb [37]. The conservation of His, Glu, and Gln at positions 57, 58, and 61 in all IFNa subtypes suggests that weak binding to IFNAR1 is an evolutionarily conserved trait. Strikingly, the largely overlapping activity profiles of YNS and IFNb corroborate that differential IFN activity is directly related to receptor binding affinities (Figure 1B) [23,37]. Assembly and dynamics of the signaling complex in the plasma membrane The striking correlation between the receptor binding properties and activity profiles of IFNs suggests a key role of receptor assembly and dynamics in signaling specificity. The mechanism of cytokine receptor assembly has been debated during the past decade. The original observation of human growth hormone receptor dimerization caused by simultaneous interaction of the ligand with the receptor subunits has long been viewed as the dominant mechanism [38]. This mechanism is in line with the asymmetric binding affinities of the ligand to the receptor subunits, which is also prominent in the case of the IFN receptor. The latter suggests a binding mechanism, where ligand binding to the high-affinity subunit is followed by recruitment of the low-affinity subunit at the plasma membrane (see Figure I in Box 1). However, following the observation of a

Box 1. Spatiotemporal dynamics of IFN receptor assembly and activation Ligand-induced receptor assembly and the formation of dynamic ternary complexes is controlled by several rate constants, as depicted in Figure I. IFN binding to the cell surface is dominated by the approximately tenfold higher association rate constant of the
interaction with IFNAR2 kaB compared with IFNAR1 [62,95]. Subsequently, IFNAR1 is recruited by 2D interactions in the plasma membrane to form the ternary signaling complex. In model membranes, this reaction has been demonstrated to be far from diffusion control despite an approximately 100-fold reduced mobility of the interacting partners [62,95]. The dynamic equilibrium between binary and ternary complexes at the plasma membrane is
determined by the 2D IFNAR1 equilibrium dissociation constant KdT versus the receptor concentration. Thus, the 100-fold higher IFNAR1 binding affinity of IFNb compared with IFNa allows efficient IFNAR1 recruitment at much lower receptor concentrations. Importantly, the total IFN binding affinity to the cell surface receptor depends on the equilibrium between the binary and ternary complexes, because the dissociation probability (and thus the equilibrium dissociation constant of cell surface binding) of IFN simultaneously interacting with IFNAR1 and IFNAR2 is substantially decreased (more than 100-fold for
IFNa2 [17]) compared with the dissociation from IFNAR2 alone kdB .

IFNAR1

KDB IFNAR2

Jak1

KaB KdB

KDT KaT KdT

+

The molecular lifespan of individual signaling complexes is determined by the 2D dissociation rate constant, kdT . Until dissociation occurs, receptor phosphorylation occurs followed by signaling activity. However, in the absence of further stabilizing mechanisms (e.g., endocytosis; cf. Figure 2 in main text), complexes can dissociate and are subject to dephosphorylation. It is possible that dissociated receptor subunits remain signaling competent, which could allow serial activation of multiple receptors by a single ligand [64]. However, plasma membrane mosaicity may introduce further determinants of receptor dynamics. Diffusional encounters by the receptor subunits are more complicated due to the diffusion barriers caused by the cortical actin skeleton [membrane skeleton [MSK]): inside corrals of approximately 100 nm in diameter, fast local diffusion is possible, which is reduced by a factor >10 over micrometer distances due to the reduced probability of crossing the actin-mediated fences [96]. For receptor dimerization, both receptor subunits need to be localized within the same corral. Importantly, transient receptor confinement may also ensure reassociation of dissociated receptor dimers and thus maintain active ternary signaling complexes over longer time periods. Such ‘signaling bursts’ have been theoretically predicted.

KDT Kp

KaT KdT

+

KdP

+

tyk2

TRENDS in Immunology

Figure I. Putative ligand-induced dimerization and dynamics of the interferon (IFN) receptor. After IFN binding to the high-affinity subunit IFNAR2, IFNAR1 is recruited on the membrane to form the ternary signaling complex. Activated by mutual interactions, Jak1 and Tyk2 phosphorylate each other as well as tyrosine residues in the cytoplasmic receptor domains. On dissociation of the complexes, signaling is abrogated by dephosphorylation. More details are provided in Box 1.

141

Feature Review ligand-independent dimer within a crystal structure of the erythropoietin (Epo) receptor ectodomain [39], increasing evidence for the preassembly of several cytokine receptors has emerged [40–43]. For the homodimeric cytokine receptors (growth hormone, Epo), pre-dimerization via the transmembrane domains has been proposed [40–42]. These findings were followed by other heterodimeric cytokine receptors [44–49] including IFNAR [50], for which predimerization was suggested. Interestingly, an important role of the associated JAKs in receptor pre-dimerization has also been proposed [50]. Taken together, these findings suggest that receptor dimerization may be mediated not only by the extracellular domains but also by other parts of the transmembrane receptor. These concepts are in line with recent studies on the epidermal growth factor receptor (EFGR), for which a subtle interplay of interactions, including the transmembrane domains and the kinase domains, was demonstrated [51]. Importantly, these rather weak interactions may only marginally contribute to receptor dimerization but play a key role in JAK activation [52] and in regulating receptor activity [51]. However, these weak interactions may also be responsible for inducing receptor dimerization on overexpression, which is required for most dimerization assays. Autoactivation of receptors caused by receptor overexpression has been reported in some cases (e.g., for the EGFR), which may be caused by shifting the equilibrium toward receptor dimers. Moreover, there is ample evidence that, at physiological receptor expression levels, receptor dimerization is induced by the ligand. This includes the role of receptor expression levels in the observed cell surface binding affinity (Box 1) as well as the bell-shaped dose–response curves observed for homodimeric receptors [53–55]. As is the case for many other cytokine receptors, the physiological expression level of IFNAR1 and IFNAR2 is very low – typically a few hundred copies per cell [15]. Traditional ensemble fluorescence microscopy cannot be applied at such cell surface receptor concentrations, requiring the use of single-molecule imaging techniques. Single-molecule tracking of individual IFNAR1s and IFNAR2s at physiological cell surface expression levels has revealed a random distribution and independent diffusion in the plasma membrane [56,57]. On addition of IFN, a minor decrease in receptor mobility was observed, which probably can be attributed to ligand-induced dimerization of IFNAR1 and IFNAR2 [58]. Similar spatiotemporal dynamics have been observed for fluorescent IFNa2 bound to endogenous IFNAR [59]. By contrast, superresolution imaging revealed strong co-clustering of overexpressed IFNAR1 and IFNAR2 [60], corroborating the potential experimental bias caused by high densities of receptors. While such effects are clearly not detectable at physiological receptor expression levels, transient receptor confinement in actin-dependent membrane microdomains was identified [57], which was also related to STAT recruitment in transient signaling zones [57]. The spatiotemporal organization of receptor subunits and the signaling complex identified by single-molecule tracking suggests an even more complex interplay of interactions and receptor confinement involved in the dynamics of IFN signaling complexes. 142

Trends in Immunology March 2015, Vol. 36, No. 3

Based on available evidence, we propose that, at physiological receptor levels, the molecular interactions of IFN with its receptor subunits dominate receptor dimerization. This assumption is in line with the key role of IFN binding affinity toward IFNAR1, which has been shown to be responsible for the differential activity between IFNa2 and IFNb [23]. Assuming a two-step assembly mechanism as depicted in Figure I in Box 1, the interaction between IFN and IFNAR1 determines the 2D equilibrium between binary (IFNAR2/IFN) and ternary (IFNAR1/IFNAR2/IFN) complexes in the plasma membrane. In the case that the 2D equilibrium dissociation constant (KdT) significantly exceeds the cell surface concentrations of IFNAR1 and IFNAR2, efficient complex formation is abrogated. Importantly, this equilibrium can only partially be compensated by IFN concentrations, as only IFN bound to IFNAR2 contributes to ternary complex formation. In vitro studies of receptor dimerization in artificial membranes suggested that the binding affinities of IFNs toward IFNAR1 limit ternary complex formation at physiological receptor expression levels, while no significant contribution of the transmembrane domain to receptor dimerization was detectable [61–63]. Thus, substantially more efficient ternary complex formation by IFNb compared with IFNa2 can be assumed at physiological receptor expression levels, which is in line with the observation that the differential activity of IFNb compared with IFNa is abrogated on increasing receptor cell surface expression [64]. Notably, increased binding affinity toward IFNAR1 is accompanied by an increased lifespan for individual signaling complexes (cf. Box 1). The exact role of complex lifespan in signaling activity has remained unclear. Indications that IFNa2 may even surpass the potency of IFNb at high receptor expression levels have been interpreted by a model of ‘serial activation’; that is, the ability of a single ligand to sequentially activate multiple receptors (see Figure I in Box 1) [64]. Moreover, differential binding affinities have also been linked to differential receptor endocytosis [23] – probably another key event in the regulation of (differential) IFN activity (see below). Although the exact correlation between molecular interaction properties and the spatiotemporal dynamics of receptor assembly and signal propagation remain to be elucidated, its key role in understanding functional receptor plasticity is clear. Effector activation at the IFNAR signaling complex The cytoplasmic domains of the IFNAR subunits are associated with various effector proteins responsible for propagating the signal to the nucleus to regulate gene activation mainly via the JAK/STAT pathway [65]. Key effector proteins are the JAKs Jak1 and Tyk2, which are associated with IFNAR2 and IFNAR1, respectively. Dimerization of IFNAR1 and IFNAR2 probably promotes interactions between Jak1 and Tyk2, thereby allowing further activation of kinase activity via cross-phosphorylation. Furthermore, several tyrosine residues of IFNAR1 and IFNAR2 are phosphorylated by the activated JAKs, which serve as docking sites for other effector proteins. The hallmark of IFN signaling is the formation of a pSTAT1/pSTAT2 heterodimer, which in complex with IFN regulatory factor 9 (IRF9) forms the transcription

Feature Review factor IFN-stimulated gene factor 3 (ISGF3), which promotes transcription of key ISGs (Figure 1A). While STAT proteins are supposedly recruited to pTyr via their SH2 domains, STAT2 was found to bind IFNAR2 via a constitutive, phosphorylation-independent binding site [66]. STAT1 in turn was shown to be recruited via STAT2 [66]. However, a tyrosine phosphorylation site of IFNAR2 is required for STAT phosphorylation [67]. Recently, livecell micropatterning was established to probe the interaction of STAT proteins with the IFNAR in living cells. This method is based on spatially reorganizing bait proteins in the plasma membrane by means of micropatterned capturing groups on a solid support [68]. Thus, effector interactions with IFNAR can be monitored in situ and quantified with respect to complex stabilities. These studies confirmed constitutive STAT2 binding to IFNAR2 as well as docking of STAT1 via STAT2 [68]. The role of the constitutive STAT2 binding site in STAT1/STAT2 phosphorylation remains under debate, as mutation studies suggested that constitutive STAT2 binding may not be mandatory [69]. Interestingly, increased STAT recruitment to the receptor was observed on formation of the ternary signaling complex [68], which could be caused by cooperative STAT binding to IFNAR2 via constitutive and phosphorylation-dependent sites. Beside pSTAT1/pSTAT2 heterodimers, homodimeric pSTAT1 is also generated on IFNAR activation, which is responsible for the regulation of IFNg-activated sequence (GAS) elements, although STAT2 was reported to be essential for STAT1 phosphorylation [70]. However, uncompromised STAT1 phosphorylation on RNAi-mediated silencing of STAT2 has been observed [71], suggesting more complicated and potentially redundant mechanisms of STAT1 recruitment. While the fundamental mechanism of ISGF3 formation appears to be distinct from classic STAT activation, the control of homo- versus heterodimer formation may play a critical role in the functional plasticity of IFN signaling. Characteristics of robust and tunable IFN activities IFN activities can be divided into robust and tunable, as discussed above [26]. The most obvious robust activity is the antiviral activity of IFNs, which is common to all cells and serves as the first line of defense against viral and other pathogen attacks [8]. This response is a basic constituent of innate immunity and its absence is one of the reasons for the high lethality associated with Ebola virus infection [72]. Interestingly, robust responses are induced with similar potencies by all IFNs despite their highly varying receptor binding affinities. By contrast, tunable responses are induced at physiological IFN concentrations mainly by the high affinity binding IFNb. Thus, a primary cause of receptor plasticity is the different affinity–activity relationships of robust and tunable cellular responses (Figure 2). Examples of the tunable activities of IFNs are their antiproliferative (apoptosis and cell cycle arrest) activities and at least some of their immunomodulatory functions. The robust and tunable activities have a strong signature in IFN-induced gene induction [27]. Robust genes that are rapidly induced by low IFN concentrations for a short

Trends in Immunology March 2015, Vol. 36, No. 3

time are annotated with Gene Ontology biological process terms from the Database for Annotation, Visualization, and Integrated Discovery (DAVID) relating to response to viruses, biotic stimulus, and MHC class I. Tunable genes are annotated with cytokine and chemokine activities, taxis, and inflammatory and antiproliferative activities [25]. Promoter analysis of robust genes has shown a strong signature of the canonical IFN response elements (ISREs), while tunable gene promoters seem to be driven mainly by other response elements [25]. In contrast to robust IFN activity, the tunable activity of IFNs is highly cell-type specific. For example, IFNs induce strong tunable activity, as monitored by antiproliferative activity and specific gene activation, in OVCAR3 (human ovarian carcinoma) cells but not in T47D (breast cancer) cells [25]. As discussed above, the activation of robust activities requires IFN binding to a few cell surface receptors, while tunable activities require IFN binding to most receptors [26]. Reducing the number of surface receptors (for example by using siRNA or by natural processes such as those observed in certain cancer cells [26,73]) hinders cells from activating tunable activities. Conversely, increasing the number of cell surface receptors enables the cell to respond in a tunable manner [64]. Importantly, IFNAR1 binding affinity could be confirmed as a key determinant of tunable IFN activities. We have generated an IFNa2 mutant with strongly reduced IFNAR1 binding but increased IFNAR2 binding (IFN-1ant [36]) that can activate robust responses only, independent of IFN-1ant concentration (Figure 1B) [25]. This observation highlights the key relevance of IFNAR1 recruitment into the ternary complex as a determinant of the initiation of tunable responses. Spatiotemporal regulation of IFN signaling How do ‘robust’ and ‘tunable’ translate at a signaling level? Surprisingly, the potencies of IFNa2 and IFNb with respect to the activation of STAT phosphorylation are very similar despite their different binding affinities. This suggests that additional signaling pathways activated by IFNAR, such as mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K) signaling, are likely to be involved in promoting differential IFN activities. However, specific roles for these pathways in differential IFN activity have remained vague. By contrast, interplay between different feedback mechanisms regulating the assembly and activity of signaling complexes could explain specific features of receptor plasticity [74]. There are at least three layers of negative regulation of IFN signaling: endocytosis and endocytic trafficking [75]; suppressor of cytokine signaling 1 (SOCS1), which is a potent inhibitor of JAKs and binds to phosphorylated Tyk2 [76]; and ubiquitin-specific protease 18 (USP18) binding to the cytosolic domain of IFNAR2 [77]. While endocytosis is induced immediately after receptor activation, SOCS1 and USP18 are ISGs, which are substantially upregulated after around 4 h. Inadequate activity of any one of these three systems, as well as constant high expression of IFNs, is associated with various diseases [74,78–82]. Among these feedback regulators, endocytosis and USP18 have been proposed to critically contribute to differential 143

Feature Review

Trends in Immunology March 2015, Vol. 36, No. 3

kon koff Jak1

Tyk2

USP18

STAT1/ STAT2

Receptor dimerizaon Integral binding affinity Ternary complex stability Regulaon by USP18

Confinement by the MSK Transient signaling zones

ke

Log acvity

IFNβ Tunable IFNα2 Robust

Endocytosis Irreversible ligand binding Enhanced local concentraon Specific environment/spaal control

Log affinity

TRENDS in Immunology

Figure 2. Control of signaling specificity by the interplay of spatiotemporal regulatory mechanisms. Receptor dimerization is determined by (local) receptor concentrations and by the negative feedback regulator ubiquitin-specific protease 18 (USP18), which thus control equilibrium ligand binding to the cell surface and ternary complex formation (cf. Box 1). These parameters regulate long-term signaling activity, which is responsible for tunable cellular responses including cytotoxicity and immunomodulatory activities. Via endocytosis, signaling complexes are irreversibly removed from equilibrium ligand binding at the plasma membrane. If the endocytosis rate ke significantly exceeds koff, a further increase of IFN binding affinity cannot increase the number of signaling complexes. This feature can explain the highly robust antiviral activity of IFNs over a very large range of receptor binding affinities. Transient confinement of signaling complexes by the membrane skeleton (MSK) may regulate both receptor assembly and endocytosis. The inset depicts the differential correlation of robust and tunable activities with ligand binding affinity.

IFN activity (Figure 2). Endocytosis of IFN signaling complexes was demonstrated to be initiated via ubiquitination of IFNAR1, which exposes an endocytic motif masked by Tyk2 in the inactive state [83,84]. Interestingly, receptor endocytosis does not immediately abrogate signaling and even appears to enhance STAT phosphorylation activity [85,86]. While immediate regulation of receptor activity by endocytosis has not been directly demonstrated (Box 2), endocytosis also potentially contributes to receptor assembly: while the receptor subunits are highly diluted in the plasma membrane, uptake into early endosomes locally concentrates the receptor and the IFN into extremely small compartments. Receptor assembly can be considered essentially irreversible in endosomes, while reversible binding applies at the plasma membrane. Under these conditions, endocytosis of intact signaling complexes could account for the similar potencies of IFNa2 and IFNb in the early phase of receptor activation, despite the approximately 1000-fold difference in their receptor binding affinities. As this higher binding affinity is caused by increased complex stability, endocytosis rather than ligand dissociation limits the lifespan of the ligand–receptor complex at the plasma membrane (Figure 2). Thus, steady-state binding, endocytosis, 144

and degradation/recycling rather than the law of mass action determine the total concentration of active signaling complexes. Moreover, nuclear translocation of pSTAT could be facilitated by endosomal uptake (cf. Box 2), which could also account for the decreased pSTAT levels observed on inhibition of endocytosis. However, the final fate of endocytosed IFNAR is lysosomal degradation [87], which possibly contributes to downregulation of IFN signaling activity. Another key regulator of IFN signaling is USP18 which is upregulated on IFN stimulation and responsible for efficient receptor desensitization after the first wave of gene induction [88]. Interestingly, the ISG15-specific protease activity of USP18 is not required for IFNAR desensitization. Rather, binding to IFNAR2 was shown to be responsible for its activity as a negative feedback regulator [88]. Interestingly, USP18 desensitizes receptor activation by IFNa2 much more efficiently compared with IFNb [77,89], identifying USP18 as a key regulator of differential activity. Intriguingly, desensitization by USP18 was demonstrated to be tightly linked to IFNAR1 binding affinity, as the IFNa2 HEQ variant (see above) also showed reduced desensitization compared with wild type IFNa2 [89]. Ligand-dependent IFNAR desensitization by USP18 was

Feature Review

Trends in Immunology March 2015, Vol. 36, No. 3

correlated with reduced cell surface binding affinity of IFNa2 but not IFNa2 HEQ, suggesting that USP18 regulates IFN signaling at the level of receptor assembly. In combination with receptor endocytosis, such regulation of receptor binding affinity could provide the mechanistic bases for the functional plasticity of IFN receptors.

response (which is a robust function of IFN) and the second is autocrine response amplification, where secreted IFN buffers the stochastic nature of the single-cell response to viruses. A different outcome is observed for the IFN tunable responses. Stochastic gene expression results in only 60– 75% of WISH (human amniotic epithelial) cells treated with very high concentrations of IFN undergoing apoptosis. A direct link was observed between average receptor number and percentage of cells undergoing apoptosis [26,64]. Single-cell analysis has shown that WISH cells require a minimum number of surface receptors to be able to activate tunable functions (such as apoptosis) and that stochastic cell-to-cell variation in the number of receptors dictates whether a specific cell will have sufficient receptor numbers to be able to induce tunable responses [26]. Basically, tunable responses require the system to transmit signals to its maximum capacity (a high concentration of tightly binding IFN occupying most cell surface receptors to generate a maximum signal). A reduction in any one of these parameters in the single cell will affect its tunable signaling. This allows cell-to-cell variability as well as cell type-specific signaling of tunable functions. Thus, for example, OVCAR3 cells promote a strong tunable signal while this is almost absent in T47D cells [25,26]. By contrast, the stochastic nature of receptor numbers is much less important for robust functions, as only a few receptors and weak signaling are required for close-to-optimal activation [25,26]. These observations highlight the key relevance of stochasticity in governing some but not all cell functions and that the requirement for different levels of receptor engagement for short- versus long-term cellular responses is a primary cause of functional receptor plasticity. Interestingly, the percentage of cells undergoing apoptosis increased to >95% and the time of activation shortened from days to hours when cFLIP (an inhibitor protein of caspase-8) was downregulated using gene knock down, yet without affecting the IFN EC50 [93]. The magnitude of apoptosis is a direct consequence of the balance between cFLIP and caspase-8, and as such is also stochastic. This shows that not only the initiation of tunable IFN functions dictates their outcome, but also their execution.

The stochastic nature of IFN activities The ability of cells to accurately control gene expression levels in response to extracellular signals is limited by the inherently stochastic nature of signal transduction and transcriptional regulation. Cell-to-cell variability in expression (noise) changes as a function of transcription factor activity, which leads to variability at the level of the individual cell [90,91]. Key steps of virus-induced signal transduction, IFNb expression, and induction of IFN-stimulated genes are stochastic events in individual cells. The heterogeneity of IFN and IFN-induced gene production is of cellular origin, arising from intrinsic noise, and thus is temporal and unpredictable [92]. However, the antiviral response is activated in all cells despite the stochastic response of single cells to viruses. This is achieved through a combination of two mechanisms. One is the very low threshold for the initiation of the antiviral

Translating differential activity into a digital code The assembly of published experimental quantitative data obtained with WISH cells provides a view of the complexity of the temporal sequence of events of IFN-induced responses and the key role of receptor engagement levels. In Figure 3, we plot the timeline of the various activities and their magnitudes resulting from two regimens of activation: 1 pM IFNa2 (A) and 1 nM IFNb (B). Low IFNa2 activates only robust activities, while high IFNb activates both robust and tunable activities. Due to feedback mechanisms, IFN induces wave-like responses, with receptor complexation and STAT phosphorylation occurring within minutes. Interestingly, STAT phosphorylation peaks after around 30 min and then declines, independent of the magnitude of treatment (pSTAT1 declines faster than pSTAT2). The decline in pSTAT, caused by negative feedback regulation including receptor endocytosis and

Box 2. Role of receptor endocytosis in regulating cytokine signaling Cytokine receptors undergo endocytosis via clathrin-dependent and clathrin-independent pathways and highly specific features of endocytic trafficking have been observed for different receptors [86,97,98]. While these characteristic features suggest specific biological functions, the exact regulatory mechanisms exerted via receptor endocytosis remain debated. Originally, endocytosis was considered to control downregulation of cytokine signaling by depleting receptors and/or associated cytokines via lysosomal degradation. Enhanced signaling activity was achieved by altering endosomal sorting via engineered cytokines promoting recycling rather than degradation [99]. A key role of Epo receptor shuttling to control the dynamic range of signaling was recently demonstrated by experimental studies combined with quantitative simulations [100]. However, beside the control of total concentration by endosomal trafficking, several observations suggest that cytokine receptor complexes remain signaling competent in endosomes. Thus, efficient STAT3 association with endosomal vesicles was observed and regulatory control of cytosolic transport and nuclear translocation as well as post-translational modification via endosomal trafficking was proposed [101,102]. Strikingly, reduced STAT1 phosphorylation was observed for IFNAR but not IFNGR signaling on inhibition of endocytosis [86]. Based on these findings, control of signaling specificity by providing a specific signaling environment in endosomes was proposed [85]. In this context, membrane microdomains have been speculated to play an important role by controlling the spatiotemporal organization of signaling complexes at the plasma membrane. Beyond these still rather hypothetical concepts of specific endosomal signaling pathways, uptake of intact cytokine receptor complexes into endosomes may also contribute to increase the number of active signaling complexes and to enhance signaling by increasing the local concentration. Steady-state conditions of continuous receptor uptake into signaling-competent endosomes followed by degradation and/or recycling (see Figure 2 in main text) may be responsible for the highly robust, affinity-independent early signaling response observed for many cytokines. By integrating these possible functions into the spatiotemporal regulation of signaling activity, a more detailed, quantitative understanding of these processes is required to unravel the cellular basis of functional receptor plasticity.

145

Feature Review

(A)

an antiproliferative response, for which longer times and a stronger signal are required.

1pM IFNα2

1

Bound receptors pSTAT1 pSTAT2 Surface receptors R1 Surface receptors R2 Genes robust Genes tunable USP18 transcript AV acvity AP acvity

0.8

Fracon

Trends in Immunology March 2015, Vol. 36, No. 3

0.6 0.4 0.2

(B)

1nM IFNβ

1

Fracon

0.8 0.6 0.4 0.2

1

10

30 min

100

4 hr

1000 24 hr

104

Time (min) TRENDS in Immunology

Figure 3. Timeline of receptor, effector, and gene activation during type I interferon (IFN) signaling. Two regimens of activation are presented: low-level activation by 1 pM IFNa2 (A) and high-level activation by 1 nM IFNb (B). Most of the data were obtained with WISH cells [12,17,18,23–27,34,35,37,62,64,89,103], to avoid the complexity of using different cell lines that drive different levels of responsiveness.

degradation (Box 2) as well as SOCS activity, constitutes the first layer of negative feedback on the system. Gene activation, which follows the assembly of ISGF3, is observed from 2 h of continuous IFN induction onward, independent of the type of IFN, a time point where pSTAT is already declining. Interestingly, the very small wave of pSTAT induced by 1 pM IFNa2 is sufficient to drive sufficient transcription of robust (but not tunable) genes, thus yielding full antiviral protection and showing the sensitivity of the system. Tunable gene expression starts later, is maximized at 16–24 h of continuous IFN stimulation, and is not induced by 1 pM IFNa2 but only by 1 nM IFNb. Interestingly, phosphorylated STAT1 and STAT2 can no longer be detected at this time point and thus uncoupling between STAT phosphorylation and tunable gene transcription is observed. However, the formation of ISGF3 is essential for tunable gene induction, suggesting a second layer of gene regulatory control possibly related to the transcriptional activity of unphosphorylated STATs [94]. USP18 expression is upregulated similarly to the other robust genes. As a result, from approximately 8 h onward an additional negative feedback is induced that reduces complex assembly at the plasma membrane and, particularly, abrogates signaling with the 1 pM IFNa treatment. In the absence of USP18, one sees significantly stronger transcription induced by IFNas. However, the stronger-binding IFNb has the ability to partially overcome the USP18-induced inhibition. In line with the massive induction of robust genes even with 1 pM IFNa2, it is already sufficient to induce an antiviral response after 4 h of induction. Conversely, it is not sufficient to induce 146

Concluding remarks The molecular and cellular mechanisms controlling specificity in pleiotropic signaling by many cytokines have remained enigmatic and controversially debated in this field. For IFNs as a paradigm for cytokine receptor plasticity, compelling determinants are emerging, which we have summarized in this review. Based on comprehensive, quantitative structure–function analyses, affinity–activity correlations, and analysis of gene expression patterns, together with enormous progress in mechanistic understanding of receptor assembly, signal activations, and regulatory mechanisms, strikingly simple principles have been identified. Our concept proposes that, rather than qualitative differences in signaling pathways, the temporal evolution of signaling activity plays a decisive role in cellular decision making. In combination with very different signaling thresholds for different cellular responses, differential activity can be explained by the different abilities of IFNs to maintain receptor activation. For this concept, feedback mechanisms that modulate receptor activity over time in a ligand affinity-dependent manner are key. This quantitative understanding of differential IFN activity has important implications for the future development of IFNs as drugs, both to increase their benefits and reduce their harmful side effects. Thus, IFNs with highly specific activities have been engineered [25] and the efficacy of cell marker-specific targeting IFNs could be tremendously enhanced [59]. Moreover, the exposure of patients treated with IFN can be systematically optimized. For particular therapies, we can now ask whether a more constant drug concentration, achieved by frequent injections or by extending the in situ IFN half-life, is beneficial and which type of IFN-induced activity is sought. These considerations may shift the way we develop cytokine drugs in the future and choose

Box 3. Outstanding questions IFNs have a prominent role in many aspects of innate and acquired immunity. However, despite almost 60 years of research in the field, we are still surprised by the complex integration of molecular information into cellular decision making. While some basic mechanistic concepts of IFN receptor plasticity are emerging, several key questions remain unsolved.  What is the molecular mechanism of STAT activation at the cytosolic IFNAR signaling complex?  How is STAT activation spatiotemporally regulated in the context of different feedback mechanisms (receptor assembly, endocytosis) and further post-translational modifications?  How do IFN signaling and gene transcription proceed in the absence of STAT phosphorylation?  What mechanisms are responsible for differentially controlling robust and tunable gene expression?  Which are the specific cellular determinants regulating cellspecific decision making? Answering these questions will require a much more detailed picture of the IFNAR signaling interactome and transcriptome. Moreover, novel probes and reporters to quantitatively monitor protein interactions and conformational dynamics at the IFNAR signaling complex as well as signal propagation and gene transcription in situ will be essential.

Feature Review their mode of administration. However, the exact molecular and cellular mechanisms controlling the temporal evolution of IFN signaling remains largely unclear (Box 3). In particular, a much better understanding how IFN signaling proceeds after potent abrogation of STAT phosphorylation by negative feedback regulators and how tunable genes are specifically induced will be essential. This will require not only a detailed spatiotemporal inventory of the IFNAR interactome and transcriptome, but also novel tools to monitor downstream signaling events with subcellular resolution on multiple temporal scales. Acknowledgments Financial support by the Deutsche Forschungsgemeinschaft (SFB944) to J.P., the I-CORE Program of the Planning and Budgeting Committee and the Israel Science Foundation (Grant No 1775/12), and Grant No 2011093 from the United States–Israel Binational Science Foundation (BSF) to G.S. is gratefully acknowledged.

References 1 Stroud, R.M. and Wells, J.A. (2004) Mechanistic diversity of cytokine receptor signaling across cell membranes. Sci. STKE 2004, re7 2 de Weerd, N.A. and Nguyen, T. (2012) The interferons and their receptors – distribution and regulation. Immunol. Cell Biol. 90, 483–491 3 Moraga, I. et al. (2014) Multifarious determinants of cytokine receptor signaling specificity. Adv. Immunol. 121, 1–39 4 Broughton, S.E. et al. (2012) Cytokine receptor activation at the cell surface. Curr. Opin. Struct. Biol. 22, 350–359 5 Uze, G. et al. (2007) The receptor of the type I interferon family. Curr. Top. Microbiol. Immunol. 316, 71–95 6 Samarajiwa, S.A. et al. (2009) INTERFEROME: the database of interferon regulated genes. Nucleic Acids Res. 37, D852–D857 7 Ivashkiv, L.B. and Donlin, L.T. (2014) Regulation of type I interferon responses. Nat. Rev. Immunol. 14, 36–49 8 Hertzog, P.J. and Williams, B.R. (2013) Fine tuning type I interferon responses. Cytokine Growth Factor Rev. 24, 217–225 9 Woelk, C.H. et al. (2007) Evolution of the interferon alpha gene family in eutherian mammals. Gene 397, 38–50 10 Harari, D. et al. (2014) Bridging the species divide: transgenic mice humanized for type-I interferon response. PLoS ONE 9, e84259 11 de Weerd, N.A. et al. (2013) Structural basis of a unique interferonbeta signaling axis mediated via the receptor IFNAR1. Nat. Immunol. 14, 901–907 12 Thomas, C. et al. (2011) Structural linkage between ligand discrimination and receptor activation by type I interferons. Cell 146, 621–632 13 Chill, J.H. et al. (2003) The human type I interferon receptor: NMR structure reveals the molecular basis of ligand binding. Structure (Camb.) 11, 791–802 14 Li, Z. et al. (2008) The EM structure of a type I interferon–receptor complex reveals a novel mechanism for cytokine signaling. J. Mol. Biol. 377, 715–724 15 Cohen, B. et al. (1995) Ligand-induced association of the type I interferon receptor components. Mol. Cell. Biol. 15, 4208–4214 16 Piehler, J. and Schreiber, G. (1999) Mutational and structural analysis of the binding interface between type I interferons and their receptor IFNAR2. J. Mol. Biol. 294, 223–237 17 Lamken, P. et al. (2004) Ligand-induced assembling of the type I interferon receptor on supported lipid bilayers. J. Mol. Biol. 341, 303–318 18 Lavoie, T.B. et al. (2011) Binding and activity of all human alpha interferon subtypes. Cytokine 56, 282–289 19 Akabayov, S.R. et al. (2010) NMR mapping of the IFNAR1-EC binding site on IFNa2 reveals allosteric changes in the IFNAR2-EC binding site. Biochemistry 49, 687–695 20 Waichman, S. et al. (2010) Functional immobilization and patterning of proteins by an enzymatic transfer reaction. Anal. Chem. 82, 1478–1485 21 Jaks, E. et al. (2007) Differential receptor subunit affinities of type I interferons govern differential signal activation. J. Mol. Biol. 366, 525–539

Trends in Immunology March 2015, Vol. 36, No. 3

22 Kuruganti, S. et al. (2014) Production and characterization of thirteen human type-I interferon-alpha subtypes. Protein Expr. Purif. 103C, 75–83 23 Jaitin, D.A. et al. (2006) Inquiring into the differential action of interferons (IFNs): an IFN-a2 mutant with enhanced affinity to IFNAR1 is functionally similar to IFN-b. Mol. Cell. Biol. 26, 1888– 1897 24 Jaitin, D.A. and Schreiber, G. (2007) Upregulation of a small subset of genes drives type I interferon-induced antiviral memory. J. Interferon Cytokine Res. 27, 653–664 25 Levin, D. et al. (2014) Multifaceted activities of type I interferon are revealed by a receptor antagonist. Sci. Signal. 7, ra50 26 Levin, D. et al. (2011) Stochastic receptor expression determines cell fate upon interferon treatment. Mol. Cell. Biol. 31, 3252–3266 27 Kalie, E. et al. (2008) The stability of the ternary interferon–receptor complex rather than the affinity to the individual subunits dictates differential biological activities. J. Biol. Chem. 283, 32925– 32936 28 Chill, J.H. et al. (2002) The human interferon receptor: NMR-based modeling, mapping of the IFN-a2 binding site, and observed ligandinduced tightening. Biochemistry 41, 3575–3585 29 Piehler, J. et al. (2012) Structural and dynamic determinants of type I interferon receptor assembly and their functional interpretation. Immunol. Rev. 250, 317–334 30 Strunk, J.J. et al. (2008) Probing protein conformations by in situ noncovalent fluorescence labeling. Bioconjug. Chem. 20, 41–46 31 Conte, L.L. et al. (1999) The atomic structure of protein–protein recognition sites. J. Mol. Biol. 285, 2177–2198 32 Roisman, L.C. et al. (2001) Structure of the interferon–receptor complex determined by distance constraints from double-mutant cycles and flexible docking. Proc. Natl. Acad. Sci. U.S.A. 98, 13231– 13236 33 Piehler, J. et al. (2000) New structural and functional aspects of the type I interferon–receptor interaction revealed by comprehensive mutational analysis of the binding interface. J. Biol. Chem. 275, 40425–40433 34 Slutzki, M. et al. (2006) Variations in the unstructured C-terminal tail of interferons contribute to differential receptor binding and biological activity. J. Mol. Biol. 360, 1019–1030 35 Roisman, L.C. et al. (2005) Mutational analysis of the IFNAR1 binding site on IFNa2 reveals the architecture of a weak ligand– receptor binding-site. J. Mol. Biol. 353, 271–281 36 Pan, M. et al. (2008) Mutation of the IFNAR-1 receptor binding site of human IFN-a2 generates type I IFN competitive antagonists. Biochemistry 47, 12018–12027 37 Kalie, E. et al. (2007) An interferon a2 mutant optimized by phage display for IFNAR1 binding confers specifically enhanced antitumor activities. J. Biol. Chem. 282, 11602–11611 38 Cunningham, B.C. et al. (1991) Dimerization of the extracellular domain of the human growth hormone receptor by a single hormone molecule. Science 254, 821–825 39 Livnah, O. et al. (1999) Crystallographic evidence for preformed dimers of erythropoietin receptor before ligand activation. Science 283, 987–990 40 Remy, I. et al. (1999) Erythropoietin receptor activation by a ligandinduced conformation change. Science 283, 990–993 41 Constantinescu, S.N. et al. (2001) Ligand-independent oligomerization of cell-surface erythropoietin receptor is mediated by the transmembrane domain. Proc. Natl. Acad. Sci. U.S.A. 98, 4379–4384 42 Brown, R.J. et al. (2005) Model for growth hormone receptor activation based on subunit rotation within a receptor dimer. Nat. Struct. Mol. Biol. 12, 814–821 43 Gent, J. et al. (2002) Ligand-independent growth hormone receptor dimerization occurs in the endoplasmic reticulum and is required for ubiquitin system-dependent endocytosis. Proc. Natl. Acad. Sci. U.S.A. 99, 9858–9863 44 Damjanovich, S. et al. (1997) Preassembly of interleukin 2 (IL-2) receptor subunits on resting Kit 225 K6 T cells and their modulation by IL-2, IL-7, and IL-15: a fluorescence resonance energy transfer study. Proc. Natl. Acad. Sci. U.S.A. 94, 13134–13139 45 Tenhumberg, S. et al. (2006) gp130 dimerization in the absence of ligand: preformed cytokine receptor complexes. Biochem. Biophys. Res. Commun. 346, 649–657 147

Feature Review 46 Zaks-Zilberman, M. et al. (2008) Interleukin-5 receptor subunit oligomerization and rearrangement revealed by fluorescence resonance energy transfer imaging. J. Biol. Chem. 283, 13398–13406 47 Krause, C.D. et al. (2006) Preassembly and ligand-induced restructuring of the chains of the IFN-gamma receptor complex: the roles of Jak kinases, Stat1 and the receptor chains. Cell Res. 16, 55–69 48 Krause, C.D. et al. (2002) Seeing the light: preassembly and ligandinduced changes of the interferon gamma receptor complex in cells. Mol. Cell. Proteomics 1, 805–815 49 Krause, C.D. et al. (2006) Interactions among the components of the interleukin-10 receptor complex. Biochem. Biophys. Res. Commun. 340, 377–385 50 Krause, C.D. et al. (2013) Ligand-independent interaction of the type I interferon receptor complex is necessary to observe its biological activity. Cytokine 64, 286–297 51 Endres, N.F. et al. (2013) Conformational coupling across the plasma membrane in activation of the EGF receptor. Cell 152, 543–556 52 Endres, N.F. et al. (2011) Regulation of the catalytic activity of the EGF receptor. Curr. Opin. Struct. Biol. 21, 777–784 53 Wells, J.A. (1996) Binding in the growth hormone receptor complex. Proc. Natl. Acad. Sci. U.S.A. 93, 1–6 54 Matthews, D.J. et al. (1996) A sequential dimerization mechanism for erythropoietin receptor activation. Proc. Natl. Acad. Sci. U.S.A. 93, 9471–9476 55 Whitty, A. and Borysenko, C.W. (1999) Small molecule cytokine mimetics. Chem. Biol. 6, R107–R118 56 You, C. et al. (2013) Electrostatically controlled quantum dot monofunctionalization for interrogating the dynamics of protein complexes in living Cells. ACS Chem. Biol. 8, 320–326 57 You, C. et al. (2014) Dynamic submicroscopic signaling zones revealed by pair correlation tracking and localization microscopy. Anal. Chem. 86, 8593–8602 58 You, C. et al. (2010) Self-controlled monofunctionalization of quantum dots for multiplexed protein tracking in live cells. Angew. Chem. Int. Ed. Engl. 49, 4108–4112 59 Garcin, G. et al. (2014) High efficiency cell-specific targeting of cytokine activity. Nat. Commun. 5, 3016 60 Wilmes, S. et al. (2012) Triple-color super-resolution imaging of live cells: resolving submicroscopic receptor organization in the plasma membrane. Angew. Chem. Int. Ed. Engl. 51, 4868–4871 61 Gavutis, M. et al. (2006) Determination of the two-dimensional interaction rate constants of a cytokine receptor complex. Biophys. J. 90, 3345–3355 62 Gavutis, M. et al. (2005) Lateral ligand–receptor interactions on membranes probed by simultaneous fluorescence–interference detection. Biophys. J. 88, 4289–4302 63 Roder, F. et al. (2014) Rapid transfer of transmembrane proteins for single molecule dimerization assays in polymer-supported membranes. ACS Chem. Biol. 9, 2479–2484 64 Moraga, I. et al. (2009) Receptor density is key to the a2/b interferon differential activities. Mol. Cell. Biol. 29, 4778–4787 65 Stark, G.R. and Darnell, J.E.J. (2012) The JAK–STAT pathway at twenty. Immunity 36, 503–514 66 Li, X. et al. (1997) Functional subdomains of STAT2 required for preassociation with the alpha interferon receptor and for signaling. Mol. Cell. Biol. 17, 2048–2056 67 Wagner, T.C. et al. (2002) Interferon signaling is dependent on specific tyrosines located within the intracellular domain of IFNAR2c. Expression of IFNAR2c tyrosine mutants in U5A cells. J. Biol. Chem. 277, 1493–1499 68 Lochte, S. et al. (2014) Live cell micropatterning reveals the dynamics of signaling complexes at the plasma membrane. J. Cell Biol. 207, 407–418 69 Nguyen, V.P. et al. (2002) Stat2 binding to the interferon-alpha receptor 2 subunit is not required for interferon-alpha signaling. J. Biol. Chem. 277, 9713–9721 70 Leung, S. et al. (1995) Role of STAT2 in the alpha interferon signaling pathway. Mol. Cell. Biol. 15, 1312–1317 71 Romero-Weaver, A.L. et al. (2010) Resistance to IFN-alpha-induced apoptosis is linked to a loss of STAT2. Mol. Cancer Res. 8, 80–92 72 Xu, W. et al. (2014) Ebola virus VP24 targets a unique NLS binding site on karyopherin alpha 5 to selectively compete with nuclear import of phosphorylated STAT1. Cell Host Microbe 16, 187–200 148

Trends in Immunology March 2015, Vol. 36, No. 3

73 Wagner, T.C. et al. (2004) Interferon receptor expression regulates the antiproliferative effects of interferons on cancer cells and solid tumors. Int. J. Cancer 111, 32–42 74 Schneider, W.M. et al. (2014) Interferon-stimulated genes: a complex web of host defenses. Annu. Rev. Immunol. 32, 513–545 75 Zheng, H. et al. (2011) Ligand-stimulated downregulation of the alpha interferon receptor: role of protein kinase D2. Mol. Cell. Biol. 31, 710–720 76 Piganis, R.A. et al. (2011) Suppressor of cytokine signaling (SOCS) 1 inhibits type I interferon (IFN) signaling via the interferon alpha receptor (IFNAR1)-associated tyrosine kinase Tyk2. J. Biol. Chem. 286, 33811–33818 77 Francois-Newton, V. et al. (2012) USP18 establishes the transcriptional and anti-proliferative interferon alpha/beta differential. Biochem. J. 15, 509–516 78 Sandling, J.K. et al. (2011) A candidate gene study of the type I interferon pathway implicates IKBKE and IL8 as risk loci for SLE. Eur. J. Hum. Genet. 19, 479–484 79 Baruch, K. et al. (2014) Aging-induced type I interferon response at the choroid plexus negatively affects brain function. Science 346, 89– 93 80 Mayer-Barber, K.D. et al. (2014) Host-directed therapy of tuberculosis based on interleukin-1 and type I interferon crosstalk. Nature 511, 99–103 81 Sandler, N.G. et al. (2014) Type I interferon responses in rhesus macaques prevent SIV infection and slow disease progression. Nature 511, 601–605 82 Bhattacharya, S. et al. (2014) Triggering ubiquitination of IFNAR1 protects tissues from inflammatory injury. EMBO Mol. Med. 6, 384– 397 83 Kumar, K.G. et al. (2007) Site-specific ubiquitination exposes a linear motif to promote interferon-a receptor endocytosis. J. Cell Biol. 179, 935–950 84 Kumar, K.G. et al. (2008) Basal ubiquitin-independent internalization of interferon alpha receptor is prevented by Tyk2-mediated masking of a linear endocytic motif. J. Biol. Chem. 283, 18566–18572 85 Claudinon, J. et al. (2007) Interfering with interferon receptor sorting and trafficking: impact on signaling. Biochimie 89, 735–743 86 Marchetti, M. et al. (2006) Stat-mediated signaling induced by type I and type II interferons (IFNs) is differentially controlled through lipid microdomain association and clathrin-dependent endocytosis of IFN receptors. Mol. Biol. Cell 17, 2896–2909 87 Kumar, K.G. et al. (2003) SCF(HOS) ubiquitin ligase mediates the ligand-induced down-regulation of the interferon-alpha receptor. EMBO J. 22, 5480–5490 88 Malakhova, O.A. et al. (2006) UBP43 is a novel regulator of interferon signaling independent of its ISG15 isopeptidase activity. EMBO J. 25, 2358–2367 89 Francois-Newton, V. et al. (2011) USP18-based negative feedback control is induced by type I and type III interferons and specifically inactivates interferon alpha response. PLoS ONE 6, e22200 90 Carey, L.B. et al. (2013) Promoter sequence determines the relationship between expression level and noise. PLoS Biol. 11, e1001528 91 Gat-Viks, I. et al. (2013) Deciphering molecular circuits from genetic variation underlying transcriptional responsiveness to stimuli. Nat. Biotechnol. 31, 342–349 92 Rand, U. et al. (2012) Multi-layered stochasticity and paracrine signal propagation shape the type-I interferon response. Mol. Syst. Biol. 8, 584 93 Apelbaum, A. et al. (2013) Type I interferons induce apoptosis by balancing cFLIP and caspase-8 independent of death ligands. Mol. Cell. Biol. 33, 800–814 94 Cheon, H. and Stark, G.R. (2009) Unphosphorylated STAT1 prolongs the expression of interferon-induced immune regulatory genes. Proc. Natl. Acad. Sci. U.S.A. 106, 9373–9378 95 Gavutis, M. et al. (2006) Probing 2-dimensional protein–protein interactions on model membranes. Nat. Protoc. 1, 2091–2103 96 Fujiwara, T. et al. (2002) Phospholipids undergo hop diffusion in compartmentalized cell membrane. J. Cell Biol. 157, 1071–1081 97 Gesbert, F. et al. (2004) Clathrin-independent endocytosis and signalling of interleukin 2 receptors IL-2R endocytosis and signalling. Curr. Top. Microbiol. Immunol. 286, 119–148

Feature Review 98 Blouin, C.M. and Lamaze, C. (2013) Interferon gamma receptor: the beginning of the journey. Front. Immunol. 4, 267 99 Fallon, E.M. et al. (2000) Increased endosomal sorting of ligand to recycling enhances potency of an interleukin-2 analog. J. Biol. Chem. 275, 6790–6797 100 Becker, V. et al. (2010) Covering a broad dynamic range: information processing at the erythropoietin receptor. Science 328, 1404–1408

Trends in Immunology March 2015, Vol. 36, No. 3

101 Bild, A.H. et al. (2002) Cytoplasmic transport of Stat3 by receptormediated endocytosis. EMBO J. 21, 3255–3263 102 German, C.L. et al. (2011) The STAT3 beacon: IL-6 recurrently activates STAT 3 from endosomal structures. Exp. Cell Res. 317, 1955–1969 103 Lata, S. et al. (2006) Monitoring the dynamics of ligand–receptor complexes on model membranes. J. Am. Chem. Soc. 128, 6–7

149