LIGAND-INDEPENDENT INTERACTION OF THE TYPE I INTERFERON RECEPTOR COMPLEX IS NECESSARY TO OBSERVE ITS BIOLOGICAL ACTIVITY

Abstract

Ectopic coexpression of the two chains of the Type I and Type III interferon (IFN) receptor complexes (IFN-αR1 and IFN-αR2c, or IFN-λR1 and IL-10R2) yielded sensitivity to IFN-alpha or IFN-lambda in only some cells. We found that IFN-αR1 and IFN-αR2c exhibit FRET only when expressed at equivalent and low levels. Expanded clonal cell lines expressing both IFN-αR1 and IFN-αR2c were sensitive to IFN-alpha only when IFN-αR1 and IFN-αR2c exhibited FRET in the absence of human IFN-alpha. Coexpression of RACK-1 or Jak1 enhanced the affinity of the interaction between IFN-αR1 and IFN-αR2c. Both IFN-αR1 and IFN-αR2c exhibited FRET with Jak1 and Tyk2. Together with data showing that disruption of the preassociation between the IFN-gamma receptor chains inhibited its biological activity, we propose that biologically active IFN receptors require ligand-independent juxtaposition of IFN receptor chains assisted by their associated cytosolic proteins.

1. Introduction

Type I interferons (IFNs), released in response to the detection of a viral infection, promote a virus-resistant state in cells and assist the immune system to detect and eliminate viruses and virus-infected cells. All activities of Type I IFNs are mediated by their interaction with the Type I IFN receptor complex [1;2], also known as the IFN-α or IFN-α/β receptor. The IFN-αR2c (also known as IFNAR-2C) chain interacts with not only the tyrosine kinase Jak1 but also the transcription factor Stat2. The IFN-αR2b splice variant encodes a transmembrane receptor chain with a truncated intracellular domain that binds neither Stat2 nor Jak1. The IFN-αR1 (also known as IFNAR-1) chain interacts with the tyrosine kinase Tyk2.

An important question is whether the Type I IFN receptor complex in its resting state is a preassembled entity or is dissociated prior to its nucleation by Type I IFN. Based on crosslinking studies, immunoprecipitations, and studies with soluble receptors, it is currently accepted that the Type I IFN receptor complex is dissociated until ligand is added, and IFN-αR1 and IFN-αR2c interact only indirectly through the ligand [3–7]. these studies are described in depth in Supplementary Text 2. Because these studies require analyzing the receptor outside of its native environment (where bias favoring high-affinity interactions that are environmentally insensitive is introduced), we wanted to use technologies that remove this bias by analyzing receptor structure in their physiological environment (i.e., within intact, where environmentally-sensitive and low-affinity interactions are maintained) in order to obtain more physiologically accurate data.

We successfully utilized fluorescence resonance energy transfer (FRET) to noninvasively probe the structure of the IFN-γ and the interleukin-10 receptor complexes. Observing FRET between IFN-γR1 and IFN-γR2 and between IL-10R1 and IL-10R2 in the absence or presence of ligand, we concluded that the IFN-γ and the IL-10 receptor chains interact in the absence of ligand, inferring that conformational changes accompany the activation of the receptor complex by ligand [8–10]. We confirmed the specificity of these interactions by coexpressing receptor chains from different receptor complexes and by mutating the IFN-γ receptor chains to inhibit biological activity; FRET between receptor chains was reduced substantially in most cases, providing evidence that mutations that eliminate biological activity also inhibit the interaction between IFN-γR1 and IFN-γR2.

Whether IFN-α and IFN-λ receptor complexes are similarly associated in a ligand-independent manner is the subject of these current studies.

2. Materials and Methods

2.1 Genetic engineering and modification of proteins

All restriction endonucleases and DNA-modifying enzymes were obtained from New England Biolabs and used according to the manufacturer’s instructions. Primers were obtained from Integrated DNA Technologies (Coralville, IA) and were used as provided. All primers in this manuscript are written in the 5′-3′ orientation. The synthesis of the various cDNA’s used in this manuscript is described in Supplementary Text 1 of this manuscript and elsewhere [11].

2.2 Interferons

Human IFN-α2a (also known as IFN-αA, #11100-1) and human IFN-γ (#11500-1) were provided by PBL Biological Laboratories (Piscataway, NJ). The biological activity of human IFN-λ3 (recombinant, from conditioned medium of 293T cells, a gift from PBL Biomedical Laboratories) was assayed by the cytopathic effect inhibition assay in HepG2 cells challenged with EMCV. This assay was validated with human IFN-λ2-His (#11820-1). From these assays, half-maximal cytopathic effect inhibition is defined as one unit/mL (U/mL).

2.3 Cell culture

Cell lines were grown in HEPA-filtered incubators at 37 °C, 5% CO₂, and 95% relative humidity. CHO-q3 cells [8] were grown in F12 medium containing 10% fetal calf serum (FCS), while 293T cells were grown in DMEM medium containing 10% FCS and 4 mM glutamine.

To obtain stably transfected cell lines, geneticin (350 μg/mL) was added to the medium three days after transfection to select for transfected cells harboring genomically integrated plasmid until expanded colonies were visible to the naked eye (two to three weeks). A sample of these cells were diluted to single cells in 96-well dishes to isolate clonal colonies. The clones were analyzed by light fluorescent microscope using the FITC and TR channels to visually identify colonies harboring GFP and OFP. The clones mentioned in this manuscript were green and orange when GFP or OFP were respectively directly excited (g+o), were weakly orange when OFP was directly excited (wo), or had such weak fluorescence that it was essentially nonfluorescent (nf). Selected clones were further amplified and screened for their ability to induce MHC Class I or be protected from viral cytotoxicity in response to human IFN-α2.

2.4 Confocal Fluorescence Spectroscopy

FRET between fluorescent protein donor:acceptor pairs ECFP and EYFP in intact cells was measured with an instrument in which a spectrometer is integrated into the sidebox of a confocal microscope, as described previously [9;11–13].

2.5 Data analysis

Spectral deconvolution (the mathematical separation of an observed emission spectrum into several spectra of homogeneous components) was described previously [10;12;13]. The improved analysis of deconvolved spectra, calculation of FRET efficiency, other cellular biophysical quantities, and the inter-FP distances are described elsewhere [11].

2.6 Interferon activity assays

The antiviral assay is based on cytopathic effect inhibition, and was performed as described previously with minor modifications [14;15]. Briefly, 50,000 cells are plated in each well of a 96-well plate. IFN is next added in two-fold serial dilutions throughout a row, and is incubated with the cells for 24 hours. Afterward, vesicular stomatitis virus (VSV, Indiana Strain) is added at 1250 pfu/well. The virus is incubated with the IFN-treated cells for 48 hours or until nearly complete killing of negative-control cells (that were not incubated with IFN) but not of positive control cells (that were incubated with neither IFN nor virus) is observed. Media is removed by inversion of the plates, and cells that survived viral infection are stained with crystal violet. Results are scored visually.

IFN also induces MHC Class I surface antigen in a dose-dependant fashion. The induction of MHC Class I Surface antigen by IFN was performed as previously described [8;15–18].

2.7 Fluorescence-Activated Cell Sorting

Flow cytommetry was performed with a Becton-Dickinson FACScan as previously described [13]. In two-color experiments, the FL1 (530/30 nm barrier emission filter) and FL3 (650 long-pass emission filter) channels were compensated against the FL2 (585/42 barrier emission filter) channel prior to data collection to minimize the contribution of EGFP fluorescence in the FL2 channel and the contribution of StFP fluorescence into the FL1 channel. A positive gate on a forward scatter:side scatter dotplot was used to isolate intact cells.

Cells expressing StFP-labeled or EGFP-labeled receptors that were not treated with antibodies were identified and gated in a FL1:FL2 dotplot as events that were found at different dispersion angles from cells exhibiting only endogenous fluorescence. Gates were drawn to fully exclude cells that contain only endogenous fluorescence, accepting that cells with insignificantly low levels of acceptor and donor fluorescence will also be excluded with this negative gate. These gates were not changed among the various transfected populations. Because CHO-q3 cells constitutively express human Class I MHC surface antigen HLA-B7 on their cell surface, cells treated with primary anti-MHC Class I antibodies and FITC-conjugated secondary antibodies were shifted considerably along the FL1 axis whether they responded to IFN or not; induction by IFN further shifted only the subpopulation that is sensitive to IFN. To score for induction of MHC by IFN, a gate was created so that the left edge of the gate was positioned between the densest part of the IFN-inducible population and the densest part of the uninduced population, where the event density was minimal. The gate was held stationary between untreated and treated cells, and the percentage of cells within each gate was measured. The fraction of cells that induced MHC Class I surface antigen was taken as the difference in the percentage of gated cells in the ligand-treated and ligand-untreated populations. Data were collected and pre-processed with the CellQuest 3.3 software package (Becton-Dickinson) and processed with WinMDI2.9 (Joseph Trotter, The Scripps Research Institute, La Jolla, CA USA) for presentation in this manuscript.

2.8 Transfections

Several transfection systems were employed in this report, depending on the most cost-effective or favored transfection reagent or protocol at the time. In most cases, the expression was optimal after 24–48 hours (optimal OFP expression required 48–72 hours). We used transiently transfected cells two to four days after transfection only if the transfection efficiency exceeded 15%, judged visually by wide-field fluorescence of a sample of transfected cells. Stable transfections in CHO-q3 cells were done with either Polyfect (main manuscript) or custom liposomes (Supplementary Figure 3.

Transient transfections using Polyfect, Superfect (both from Qiagen, Inc.) or Fugene-6 and Fugene-HD (both from Roche) were performed according to the manufacturer’s instructions. Transfections using DDAB:DOPE liposomes and polyethyleneimine (PEI) were performed as described elsewhere [11;19].

3. Results

3.1 Conditional FRET between IFN-αR1 and IFN-αR2c

We easily identified cells transfected with fluorescent IFN-γR1 and IFN-γR2 or with fluorescent IL-10R1 and IL-10R2 in which FRET between the two receptor chains was observed [8–10]. We hypothesized that we would also easily identify cells in which IFN-αR1 and IFN-αR2c exhibit FRET and biological activity. Initial studies using monomeric cyan and yellow fluorescent proteins (ECFP, EYFP) corroborated our hypothesis; these results are explained in Supplementary Text 3 and shown as Supplementary Figures 1–3.

Discovering that using non-jellyfish red-shifted truly monomeric fluorescent proteins gives more precise FRET determinations and removes artefacts arising from interactions between jellyfish fluorescent proteins [11], we used fluorescent protein pairs derived from distinct species (e.g., jellyfish-derived monomeric green or citrine fluorescent proteins, EGFP and CiFP, or anemone-derived teal fluorescent protein, TFP, as donors, and coral-derived orange or strawberry fluorescent proteins, OFP and StFP, as acceptors).

Our initial studies using IFN-αR1/EGFP and IFN-αR2c/OFP argued against a simple preassociation. The results of several experiments overlapped and were pooled together. While most cells expressing IFN-γR1/EGFP and IFN-γR2/OFP exhibit FRET (Fig. 1A), only a fraction of cells coexpressing IFN-αR1/EGFP and IFN-αR2/OFP exhibited FRET efficiency over 10% (Fig. 1B). By resolving these data according to the molar ratio of the two chains, it was clear that while a stoichiometric excess of IFN-γR2 did not inhibit FRET with IFN-γR1, a stoichiometric excess of IFN-αR1 inhibited FRET with IFN-αR2c.

Figure 1.

Inhibition of FRET by moderate overexpression of IFN-αR1 relative to IFN-αR2c. (A) 293T cells were transiently transfected with a plasmid coexpressing IFN-γR2/OFP and IFN-γR1/EGFP, and the FRET efficiency between the two receptor chains was measured in a selection of cells and plotted as a function of the molar ratio of IFN-γR2 to IFN-γR1. (B) 293T cells were transiently transfected with a plasmid coexpressing IFN-αR1/EGFP and IFN-αR2c/OFP, and the FRET efficiency between the two receptor chains was measured in selected cells and plotted as a function of the IFN-αR1:IFN-αR2c molar ratio. Open circles are data obtained from cells whose spectra are indistinguishable from those containing only OFP. The data from (A) and (B) were pooled from multiple experiments. Either IFN-γR2 and IFN-γR1 (C) or IFN-αR1 and IFN-αR2c (D) were tagged with TFP and CiFP (red), OFP and EGFP (black), or StFP and EGFP (green), and FRET efficiency (proportional to the diameter of the open circle) was plotted as a function of the calculated relative IFN-γR2 and IFN-γR1 levels or relative IFN-αR1 and IFN-αR2c levels. The diagonal lines in (C) and (D) denote balanced expression of IFN-γR1 and IFN-γR2 or of IFN-αR1 and IFN-αR2c.

Currently available fluorescent proteins have diverse folding and maturation rates, and impart varying stabilities to fusion proteins. To ensure that these variables do not influence our results, we coexpressed either IFN-γR1 and IFN-γR2 or IFN-αR1 and IFN-αR2c tagged to various donor:acceptor pairs and analyzed these results in aggregate as a bubble plot. Whether the pair was TFP:CiFP (red circles), EGFP:OFP (black circles) or EGFP:StFP (green circles), excess relative levels of IFN-γR2 did not inhibit FRET between IFN-γR2 and IFN-γR1 (Fig. 1C), while an excess of IFN-αR1 rarely yielded strong FRET between IFN-αR1 and IFN-αR2c (Fig. 1D). Thus we conclude that excess levels of IFN-αR1 inhibit FRET between IFN-αR1 and IFN-αR2c.

3.2 Fluorescent IFN-αR1 and IFN-αR2c are biologically functional

FRET was seen between IFN-αR1 and IFN-αR2c only when IFN-αR1 levels do not exceed IFN-αR2c levels. Because we found that mutations to the IFN-γ receptor complex that inhibited FRET also prevented its activation [9], we hypothesized that only in those few cells in which IFN-αR1 and IFN-αR2c interact will functional Type I IFN receptor complexes form. To test this, we transfected IFN-αR1/EGFPm and IFN-αR2c/OFP stably in hamster CHO-q3 cells and isolated individual clones. Because hamster cells do not respond to IFN-α2 unless both human IFN-αR1 and IFN-αR2c are coexpressed [20], CHO-q3 cells are useful to test the biological activity of two fluorescent receptor chains.

Several clones expressing both IFN-αR1/EGFP and IFN-αR2c/OFP were isolated, expanded, and assayed for biological activity to human IFN-α2. Although IFN-α2 did not induce Class I MHC in one clone (clone 5.g+o, cyan, Fig. 2A, left), it induced MHC in a second clone (clone 6.wo, purple, Fig. 2A, right) with comparable relative levels of both IFN-αR1/EGFPm and IFN-αR2c/OFP. We conclude that simply having IFN-αR1 and IFN-αR2 present in the cell is not sufficient for IFN-α to induce signaling activity, even when all other intracellular components are present to propagate signaling. The induction in cells expressing IFN-αR1/EGFPm and IFN-αR2c/OFP was similar in magnitude to that induced in cells expressing IFN-αR1 and IFN-αR2c (Supp. Fig. 3C), suggesting that fluorescently tagged Type I IFN receptors retain full signaling potential.

Figure 2.

Bioactivity and interaction between IFN-αR1/EGFP and IFN-αR2c/OFP. (A) Induction of MHC Class I surface antigens in CHO-q3 clone 5.g+o (left, in light blue) and in clone 6.wo (right, in purple). The filled histograms with thin black borders are from untreated cells, while the unfilled histograms with the thick black border come from cells treated with 1000 units/ml human IFN-α2 for three days. (B) Antiviral activity of human Type I IFNs in CHO-q3 cell clones expressing IFN-αR1/EGFP and IFN-αR2c/OFP. Clones 5.g+o (top), 5.wo (middle) and 5.2nf (bottom) were treated with either serially two-fold diluted (left to right) 200 ng/mL IFN-α2 (first row), serially two-fold diluted 200 ng/mL IFN-β (second row) or no IFN and either no virus (third row, left six wells) or no IFN and VSV (third row, right six wells). Cells surviving viral infection are stained with crystal violet. For (C–E), the data from CHO-q3 clone 5.g+o (light blue) and clone 6.wo (purple) are coplotted with data from 293T cells transiently transfected with plasmids expressing IFN-αR1/EGFP and IFN-αR2c/OFP (black open circles). FRET efficiencies are plotted as a function of (C) IFN-αR1/EGFP relative levels, (D) IFN-αR2c/OFP relative levels, (E) the molar ratio of IFN-αR2c/OFP to IFN-αR1/EGFP, and (F) the relative number of IFN-αR1/EGFP:IFN-αR2c/OFP pairs formed.

Only some MHC-inducible clones exhibited antiviral activity in response to human IFN-α. Though IFN-α2 did not protect clone 5.g+o (Fig. 2B, top) or clone 6.wo (data not shown) from vesicular stomatitis virus (VSV)-induced cytotoxicity, IFN-α2 protected clones 5.wo (Fig. 2B, middle) and 5.2 nf (Fig. 2B, bottom) from VSV-induced cytotoxicity (Fig. 2B). IFN-β had comparable specific activity to IFN-α2 in these cell lines. Among MHC-inducible clones, antiviral activity was not consistent, and was never complete. Nevertheless, IFN-αR1/EGFPm and IFN-αR2c/OFP form an active Type I IFN receptor complex under the proper conditions.

We analyzed FRET in clones 5.g+o and 6.wo and compared data obtained from those clones to data obtained from the transiently transfected population (Fig. 2C–2F, purple, cyan, and unfilled circles, respectively). We found that receptors in most cells from the bioactive clone 6.wo exhibited more FRET than did receptors within the inactive clone 5.g+o. In other cells within clone 6.wo, FRET between the two receptor chains was not observed. Addition of 20 ng/mL IFN-α2 to the cells did not appreciably change the FRET between receptor chains in cells of either clone (data not shown). We show representative spectra from two cells in each population in the absence of IFN-α2, and two different cells in each population in the presence of 20 ng/mL IFN-α2; Supplementary Figure 4 shows the spectra of one of the two cells from each clone with higher FRET efficiency, while Supplementary Figure 5 shows the spectra of the cells with a lower FRET efficiency.

The relative levels of both IFN-αR1/EGFPm (Fig. 2C) and IFN-αR2c/OFP (Fig. 2D) in the two clones were low compared with those seen in transiently transfected populations of 293T cells; this is also observed in hamster cells expressing other cytokine receptor pairs (unpublished observations). However, the FRET efficiency as a function of the molar ratio overlapped among the two clones and the transiently transfected populations (Fig. 2E). Thus even in stably transfected clones, the molar ratio between IFN-αR1 and IFN-αR2c dictated the interaction between IFN-αR1 and IFN-αR2c. Notably, in cells from the bioactive clone 6.wo, the IFN-αR2c chain was in molar excess, while the IFN-αR1/EGFP chain was in molar excess in the cells from the inactive clone 5.g+o (Fig. 2E). To ensure that affinity differences between IFN-αR1/EGFPm and IFN-αR2c/OFP did not correlate with conditional sensitivity, we replotted the data as a function of the number of IFN-αR1/EGFPm:IFN-αR2c/OFP pairs formed (Fig. 2F). The affinity of both clones was similar (~20 relative pairs) but significantly less than that found in most transiently transfected cells (~100 relative pairs), implying that the increased apparent affinity between IFN-αR1/EGFPm and IFN-αR2c/OFP observed in hamster cells does not explain biological sensitivity to the Type I IFN receptor complex. Overall, our data imply that a juxtaposition of IFN-αR1 and IFN-αR2c the produces FRET is a prerequisite to initiate signaling by Type I IFN in cells.

3.3 Coexpression of IFN-αR1 and IFN-αR2c does not guarantee a biological response

We observed that considerable FRET and biological sensitivity to IFN-α2 was observed in only some cells. To confirm that biological activity was seen only in some cells expressing both IFN-αR1 and IFN-αR2c, we correlated the expression of receptor chains with biological functionality in expanded populations of stably transfected hamster CHO-q3 cells.

Because CHO-q3 cells respond to human IFN-γ or to human IFN-λ only when both human IFN-γR1 and human IFN-γR2 or both human IFN-λR1 and human IL-10R2 are coexpressed respectively [16;21], we can analyze similar hindrances in forming biologically active IFN-γ or IFN-λ receptor complexes. Because most cells coexpressing IFN-γR1 and IFN-γR2 exhibited FRET, we expected that most cells expressing IFN-γR1 and IFN-γR2 will be biologically responsive to IFN-γ. Finally, because Type I IFN and IFN-λ initiate similar signaling cascades; we expected that only some cells expressing IFN-λR1 and IL-10R2 are biologically responsive to IFN-λ3.

To visualize the expression of a single chain within a functional complex, we stably cotransfected CHO-q3 cells with one plasmid expressing a receptor chain tagged with the strawberry fluorescent protein (StFP) and a second plasmid expressing its nonfluorescent partner receptor chain (i.e., IFN-γR2/StFP and IFN-γR1, IFN-γR1/StFP and IFN-γR2, IFN-αR1/StFP and IFN-αR2c, and IFN-αR2c/StFP and IFN-αR1). We used two-color flow cytometry to correlate the expression of receptors (by StFP fluorescence) with the biological sensitivity to IFN (by increased binding of FITC-conjugated secondary antibodies). Because the detection of StFP-tagged proteins is less sensitive when FITC-conjugated antibody complexes are bound to the cell surface, we analyzed the expression of fluorescent receptors separately, in identical cells that are not treated with antibodies. In this fashion, we can compare the fraction of cells that express receptors with the fraction of cells that induce MHC Class I surface antigen. Because we designed these experiments so that bioactive receptor complexes contain one fluorescent receptor chain, we assume that all MHC-inducible cells are a subset of the cells that expresses a fluorescent receptor chain.

3.3.1 Most cells expressing IFN-γR1 and IFN-γR2 respond to IFN-γ

As expected, most cells expressing IFN-γR2/StFP or IFN-γR1/StFP were sensitive to IFN-γ (Fig. 3A-F). Of the 26.8% of the population that expressed significant levels of IFN-γR2/StFP (Fig. 3A), 70% of that 26.8% (21.3% - 2.3% = 19.0% of the total population) induced MHC Class I in response to IFN-γ treatment (Fig. 3B, 3C). Similarly, 68% of the 23.6% of cells expressing IFN-γR1/StFP (Fig. 3D) induced MHC Class I surface antigen in response to IFN-γ (18.4% − 2.3% = 16.1% of the total population, Fig 3E, 3F). Thus, the coexpression of IFN-γ receptor chains is usually sufficient to reconstitute full IFN-γ bioactivity in heterologous cells.

Figure 3.

MHC induction in CHO-q3 cells stably transfected with plasmids encoding cDNA’s of various interferon receptors. Populations of CHO-q3 cells were obtained that stably expressed the following receptor pairs: (A–C) IFN-γR1 and IFN-γR2/StFP, (D-F) IFN-γR1/StFP and IFN-γR2, (G–I) IFN-αR1/StFP and IFN-αR2c, (J–L) IFN-αR1 and IFN-αR2c/StFP, (M–O) IFN-αR1/EGFP and IFN-αR2c/StFP, and (P-R) IFN-λR1/EGFP and IL-10R2/StFP. These cells were either untreated (A, D, G, J, M, and P) or were treated with primary anti-MHC antibody and FITC-conjugated secondary anti-murine IgG antibody after three days treatment with either no ligand (B, E, H, K, N, and Q) or with 1000 U/mL IFN-γ (C, F), 1000 U/mL IFN-α2 (I, L, and O), or 4 U/mL IFN-λ3 (R). The percentage of total cells that were within each gate is indicated near each gate. To calculate the percentage of cells that demonstrate induced MHC Class I surface antigen, the percentage gated without ligand treatment was subtracted from the percentage gated after ligand treatment. Note that the fluorescence of GFP-tagged receptors is obscured by the presence of high levels of FITC-conjugated antibodies bound to constitutively expressed MHC Class I surface antigen.

3.3.2 Few cells containing IFN-αR1 and IFN-αR2c respond to IFN-α2

Contrasting the above results with IFN-γ receptors, only small fractions of cells expressing IFN-αR1/StFP or IFN-αR2c/StFP were sensitive to IFN-α2 (Fig. 3G-L). Of the 7.1% of the population that expressed significant levels of IFN-αR1/StFP (Fig. 3G), only 26% of that 7.1% (4.7% − 2.8% = 1.9%) was sensitive to IFN-α2 (Fig. 3H, I). Similarly, of the 18.3% of cells expressing IFN-αR2c/StFP (Fig. 3J), only 6.7% of that 18.3% (4.3% − 3.2% = 1.1% of the total population) was sensitive to IFN-α2 (Fig. 3K, L). This result implies that simply coexpressing IFN-αR2c and IFN-αR1 ectopically does not generate bioactivity in response to IFN-α2 in all hamster cells, even though both fluorescent receptor chains are biologically active.

3.3.3 Few cells harboring IFN-λR1 and IL-10R2 are sensitive to IFN-λ3

To see whether coexpression of IFN-λR1 and IL-10R2 (like IFN-αR1 and IFN-αR2c) confers a biological response to IFN-λ3 in only a fraction of cells coexpressing both chains, we coexpressed IFN-αR1/EGFP and IFN-αR2c/StFP or coexpressed IFN-λR1/EGFP and IL-10R2/StFP, and compared the fraction of cells that were sensitive to IFN-α2 or IFN-λ3 respectively to the fraction of cells that express EGFP-tagged receptor and StFP-tagged receptor. In both populations, significant expression of only one chain was observed, suggesting that expression of the other chain was poor or undetectable. In the presented representative experiment, 10.9% of cells expressed observable levels of IFN-αR2c/StFP, and a distinct 25.5% of the population expressed detectable levels of IFN-αR1/EGFP (Fig. 3M); analogously, 10.6% of cells expressed visible relative levels of IL-10R2/StFP, and a distinct 14.9% of cells express significant levels of IFN-λR1/EGFP (Fig. 3P).

As seen above with Type I IFN receptors, only a fraction of cells within each population contained biologically active receptor. IFN-α2 induced MHC Class I in only 3.3% of cells transfected with IFN-αR1/EGFP and IFN-αR2c/StFP (14.7% -11.4%, Fig. 3N, 3O), while IFN-λ3 induced MHC Class I antigens in only 3.7% of cells transfected with IFN-λR1/EGFP and IL-10R2/StFP (6.0% - 2.3%, Fig. 3Q, 3R). These percentages are far less than the expression of any single chain. We observed stronger induction of Class I MHC by IFN-λ3 than by IFN-α2.

Nevertheless, we isolated clones expressing both IFN-αR1/EGFP and IFN-αR2c/StFP or expressing both IFN-λR1/EGFP and IL-10R2/StFP that exhibited anti-EMCV activity in response to IFN-α2 or IFN-λ3 respectively (data not shown), Thus Type III IFN receptors, like Type I IFN receptors, are only conditionally active.

3.4 Other interactions between Type I IFN receptor chains

We also investigated interactions between IFN-αR1 and IFN-αR2b, a primate-specific splice variant of IFN-αR2c that uses an alternate nonfunctional intracellular domain that does not bind Jak1 or Stat2. We present this data in Supplementary Text 4 and Supp. Fig. 6. Summarily, IFN-αR1 interacts slightly stronger with IFN-αR2c than with IFN-αR2b.

We also investigated interactions between IFN-αR1 chains and between IFN-αR2c chains, based on the observations of FRET between IFN-γR2 chains and between IFN-γR1 chains, and our supposition that they are largely nonspecific due to their low affinity [11]. We found that while homodimeric interactions among the IFN-γ and IFN-λ receptor chains are weaker than is the heterodimeric interaction between IFN-λR1 and IL-10R2, the affinity of homodimeric interactions between Type I IFN receptor chains is similar to that between IFN-αR2c and IFN-αR1, suggesting that the observed FRET between IFN-αR1 and IFN-αR2c arises largely from nonspecific interactions in most cells; these data are explained in more detail in Supplementary Text 5 and Supp. Figs. 7 and 8.

3.5 Intracellular proteins binding IFN-αR1 and IFN-αR2c

Because the interaction between IFN-αR1 and IFN-αR2c is very weak, we surmised that, similar to what was observed within the IFN-γ receptor complex, intracellular proteins (like RACK-1 and the Janus kinases that bind IFN-αR1 and IFN-αR2c) promote the specific association and permit signaling to propagate from IFN-αR1 and IFN-αR2c in cells.

3.5.1 Interactions between receptor chains and Janus kinases

We not only investigated FRET between IFN-αR1 and Tyk2 or between IFN-αR2c and Jak1, pairs known to coimmunoprecipitate in vitro [22–27], but also between IFN-αR1 and Jak1 and IFN-αR2c and Tyk2 to see whether cross-interactions occur in cells that have not been observed upon extraction from cells. To accomplish this, we coexpressed either IFN-αR1/StFP or IFN-αR2c/StFP and either Jak1/CiFPm or Tyk2/CiFPm in 293T cells, and measured FRET efficiencies between the protein pairs in various representative cells within the four populations.

As shown in Figs 4A and 4B, significant FRET was seen with all four combinations of Type I IFN receptor and Janus kinase. IFN-αR1 displays similar cellular affinity (~ 100 relative complexes/cell) for both Jak1 (Fig. 4A, grey circles) and Tyk2 (Fig. 4A, black circles). With the exception of a few cells possessing higher FRET efficiency between Tyk2 and IFN-αR1 (the interaction observed in vitro, black circles), the two datasets overlapped completely. Although IFN-αR2c exhibited FRET with both Jak1 (Fig. 4B, black circles) and Tyk2 (Fig. 4B, grey circles), IFN-αR2c interacted somewhat stronger with Tyk2 (~ 30 relative complexes/cell) than with Jak1 (the interaction observed in vitro, ~ 50 relative complexes/cell). The optimal FRET efficiency for each pair in most cells in all four populations is about 0.30 (or 0.60 for a few cells expressing IFN-αR1 and Tyk2).

Figure 4.

Interactions between Type I IFN receptors and cytosolic proteins. All data are presented as obtained FRET efficiencies as a function of the number of donor:acceptor pairs formed. (A) A plasmid encoding IFN-αR1/StFP was coexpressed with a plasmid encoding either Jak1/CiFP (grey circles) or Tyk2/CiFP (black circles). (B) A plasmid encoding IFN-αR2c/StFP was coexpressed with a plasmid encoding either Jak1/CiFP (black circles) or Tyk2/CiFP (grey circles). Note that for (A) and (B) black circles derive from pairs known to coimmunoprecipitate in vitro. (C) FRET between IFN-αR1/CiFP and IFN-αR2c/StFP was measured in the absence of transfected RACK1 (black circles), in the presence of ectopically expressed RACK-1 (red circles) and in the presence of RACK-1/TFP (green circles). IFN-αR2c/StFP was also coexpressed with a mismatched receptor chain, IFN-γR2/CiFP (black unfilled circles). (D) Cytosolic proteins enhance the interaction between IFN-αR1 and IFN-αR2c. FRET between IFN-αR1 and IFN-αR2c was measured and if necessary, converted to an equivalent FRET efficiency between EGFP and OFP. Unfilled blue circles are replotted from Figure 2F, and are from cells in which IFN-αR1/EGFP and IFN-αR2c/OFP were coexpressed. Grey-filled blue circles are replotted from Fig. 2F, and are from cells in which IFN-αR1/EGFP and IFN-αR2c/OFP were coexpressed with Jak1. Black circles are replotted from Fig. 4C and are from cells expressing only IFN-αR1/CiFP and IFN-αR2c/StFP. Green circles are replotted from Fig. 4C and are from cells expressing IFN-αR1/CiFP, IFN-αR2c/StFP and RACK-1/TFP. Purple circles are from cells expressing IFN-αR1/OFP, IFN-αR2c/ChFP, Jak1/TFP and Tyk2/CiFP.

3.5.2 RACK-1 influences the interaction between IFN-αR1 and IFN-αR2c

RACK-1 is a scaffolding protein that couples the Type I IFN receptor chains to their signaling components by directly interacting with IFN-αR2c, Tyk2 and Jak1 as well as Stat1 [28;29]. Hypothesizing that RACK-1 promotes an association between IFN-αR1 and IFN-αR2c, we coexpressed IFN-αR1/CiFP and IFN-αR2c/StFP in the absence of RACK-1 and in the presence of either RACK-1 or RACK-1/TFP and searched for differences among the three transiently transfected populations.

With the exception of one cell in which a high FRET efficiency was seen with relatively few numbers of IFN-αR1:IFN-αR2c pairs, the absence of RACK-1 (Fig. 4C, black circles) resuled primarily in low-affinity receptors possessing an average affinity of about 500 pairs/cell; under similar conditions, the interaction between IFN-γR1 and IFN-γR2 was half-optimal at 50–100 pairs/cell [11]. In the presence of RACK-1, whether expressed without a fluorescent protein (Fig. 4C, red circles) or with TFP attached to the COOH-terminus of RACK-1 (Fig. 4C, green circles), the affinity of IFN-αR1 and IFN-αR2c increased by about 3-fold to about 100–200 pairs/cell. The optimal FRET efficiency between IFN-αR1/CiFP and IFN-αR2c/StFP increased slightly, if one again ignores cells possessing high FRET efficiencies at lower numbers of IFN-αR1:IFN-αR2c pairs. As a negative control, we coexpressed IFN-γR2/CiFP with IFN-αR2c/StFP (Fig. 4C, unfilled black circles); significant FRET between the mismatched receptor chains was never observed, and resembled the FRET seen between most cells expressing IFN-αR1/EGFP and IFN-αR2c/OFP.

3.5.3 Jak1 and Tyk2 influences the interaction between IFN-αR1 and IFN-αR2c

Because interactions between either IFN-αR1 or IFN-αR2c and either Jak1 or Tyk2 are visible spectroscopically, we hypothesized that Jak1 and Tyk2 influences the interaction between IFN-αR1 and IFN-αR2c. Thus we replotted the data from Figs. 2F and 4C (except for cells expressing nonfluorescent RACK-1), and included data in which IFN-αR1/OFP and IFN-αR2c/ChFP was coexpressed with Jak1/TFP and Tyk2/CiFP [30]. The calculated FRET efficiencies of the original data between IFN-αR1 and IFN-αR2c were converted to the predicted FRET efficiency between EGFP and OFP to make the datasets more comparable [11].

We found that the coexpression of both Jak1 and Tyk2 increased the FRET efficiency between IFN-αR1 and IFN-αR2c and significantly enhanced their affinity of interaction to at most 40–80 relative pairs/cell (Fig. 4D, purple circles). These data overlapped the two datapoints in Fig. 4C where high FRET efficiencies are seen with relatively few IFN-αR1:IFN-αR2c pairs, suggesting that in those two cells, endogenous Jak1 and Tyk2 may be present in many of these complexes.

To assess whether Tyk2 or Jak1 promotes stronger FRET between IFN-αR1 and IFN-αR2c by themselves, we coexpressed either nonfluorescent Jak1 or Tyk2 with IFN-αR1/EGFP and IFN-αR2c/OFP. Although the coexpression of Tyk2 alone did not influence FRET between IFN-αR1 and IFN-αR2c (data not shown), the coexpression of Jak1 alone increased the FRET efficiency of a few cells expressing IFN-αR1 and IFN-αR2c (Fig. 4D, grey circles with blue borders); we speculate that in those few cells, the levels of Jak1 approximated those of IFN-αR1:IFN-αR2c complexes.

3.6 Summary of results

Overall, we found that overexpression of IFN-αR1 disrupts FRET between IFN-αR1 and IFN-αR2c. Expression of IFN-αR1 and IFN-αR2c alone results only in a low-affinity interaction or colocalization and is not sufficient to restore biological activity — IFN-αR1 and IFN-αR2c must be able to juxtapose in the absence of ligand in order to exhibit biological activity in the presence of ligand. Coexpression of RACK-1 or Jak1 allows IFN-αR1 and IFN-αR2c to interact; Tyk2 further improves the interaction between IFN-αR1 and IFNαR2c. Similar requirements may hold in order for IFN-λR1 and IL-10R2 to interact and respond to IFN-λ3. The requirement of transmembrane receptor chains to interact in the absence of ligand in order to initiate signaling in the presence of ligand may apply to many cytokine receptor complexes.

4. Discussion

4.1 The Type I IFN receptor is conditionally preassociated

Studies analyzed in vitro conclude that IFN-αR1 and IFN-αR2c interact only indirectly and only in the presence of Type I IFN [3–6]. Our data studying IFN-αR1 and IFN-αR2c in their native environment gave conflicting results. Only in a few cells were IFN-αR1 and IFN-αR2c juxtaposed; in most cells (especially when levels of IFN-αR1 exceeded those of IFN-αR2c; Fig. 1B, 1D), IFN-αR1 and IFN-αR2c exhibited FRET only at high levels of both chains (Fig. 2F, 4C, 4D).

4.2 Activation requires ligand-independent receptor chain interaction

Importantly, a ligand-independant juxtaposition of IFN-αR1 and IFN-αR2c is required for a biological response to ligand to be observed. In CHO-q3 cells where IFN-αR1 and IFN-αR2c are coexpressed but did not exhibit FRET, treatment with IFN-α2 did not elicit significant FRET nor did it result in biological activity; however, when IFN-αR1 and IFN-αR2c exhibited FRET, IFN-α2 initiates biological activity (Fig. 2). Analysis of populations of stably transfected CHO-q3 cells paralleled our interaction data: only a fraction of cells expressing IFN-αR1 and IFN-αR2c induced MHC Class I surface antigens.

Because Type I receptor chains that cannot juxtapose in the absence of IFN-α will not interact nor initiate signaling in the presence of IFN-α, our data argue against structural models in which ligand assembles a receptor complex from dissociated receptor chains and initiates signaling from an active receptor complex simply by nucleating receptor chains. Because Type I IFN binds both IFN-αR1 and IFN-αR2 in cells that are insensitive to IFN with kinetics similar to those found in cells that are sensitive to IFN [31], we hypothesize that a component that helps to maintain association of the Type I IFN receptor complex plays an obligate role in permitting signal transduction.

Ectopically expressed IFN-αR1 and IFN-αR2c are biologically active, even when tagged with fluorescent proteins (Figs. 2, 3), implying that under the right conditions, biologically active fluorescent Type I IFN receptor complexes can be formed. We thus sought to understand which cellular conditions to avoid and which to promote in order to encourage ligand-independant receptor chain interactions and to observe biological activity.

4.2.1 Stoichiometry-based receptor activation

A stoichiometric excess of IFN-αR1 should be avoided. A stoichiometric excess of IFN-αR1 (at least three-fold) inhibited FRET observed between IFN-αR1 and IFN-αR2c (Figs. 1B, 1D, 2E). In contrast, a five-fold higher stoichiometric excess of IFN-γR2 (homologous to IFN-αR1) did not inhibit FRET between IFN-γR2 and IFN-γR1 (Figs. 1A, 1C). It would be interesting if the rapid turnover of IFN-αR1 in cells when Tyk2 is not bound [32], when the receptor is not signaling [33;34] or when Type I IFN signaling is initiated [35] exists to prevent inhibition of the Type I IFN receptor complexes by excessive levels of endogenous IFN-αR1.

Others have studied how variations in relative levels of IFN-αR1 and IFN-αR2 influence Type I IFN bioactivity. In one study [36], it was revealed that relatively high levels of IFN-αR1 compared to that of IFN-αR2c are required for IFN-α to have specific activity resembling that of IFN-β in 2fTGH cells. However, ectopic expression of IFN-αR2 was more effective than ectopic expression of IFN-αR1 even though both were transcribed from the same promoter. The obtained IFN-α-sensitive clones with ectopically expressed IFN-α receptor chains had either balanced levels of IFN-αR1 and IFN-αR2c (both low or both high), or had high levels of IFN-αR2c relative to that of IFN-αR1. Although no clones with a high level of IFN-αR1 and a low level of IFN-αR2c were described in that study, the same research group previously isolated a WISH clonal cell line in which IFN-αR1 was overexpressed that was sensitive to IFN-α2 and IFN-β [37]. In a second study [38], strong expression of IFN-αR2c in hepatic carcinoma cell lines was required to observe strong specific activity of IFN-α2 or IFN-α8; expression of IFN-αR1 protein was consistently poor in these cell lines.

Realizing that the direct determination of the levels or the ratio of label-free receptor chains is difficult, we suspect that cell lines expressing a definitive excess of label-free IFN-αR1 were not obtained, especially if obtaining clones sensitive to IFN-α or IFN-β were the end goal. We obtained a stoichiometric excess of IFN-αR1 by coexpressing IFN-αR1/EGFP with IFN-αR2c/OFP; the expression of the latter is most likely hindered by the slow folding of the OFP relative to the EGFP [11;39].

We do not understand how an excess of IFN-αR1 prevented association of IFN-αR2c with stoichiometric levels of IFN-αR1 in the cell. Perhaps there is a limited amount of Tyk2 (or other IFN-αR1 associated protein) under our conditions, and removal of this protein from IFN-αR2c by mass action hinders the interaction between IFN-αR1 and IFN-αR2c. Other than Tyk2, PRMT1 is known to bind IFN-αR1 in the absence of signaling [40],

Interestingly, extreme overexpression of IFN-γR2 relative to IFN-γR1 reduced the magnitude of MHC Class I induction. MHC induction was stronger in cells expressing lower relative levels of IFN-γR2/StFP than in cells expressing higher levels of IFN-γR2/StFP (Fig. 3C). To contrast this difference, a tenfold difference in IFN-γR1/StFP expression did not alter the magnitude of MHC Class I induction (Fig. 3F). We found this decrease in response by comparing the “slope” of cells that did not induce MHC Class I antigens to those that did induce these antigens. Supplementary Figure 7 (left and right) contains an enlargement of panels C and F from Figure 3 respectively with illustrative diagonal lines to emphasize these “slopes”. However, we acknowledge that the decrease in biological response when IFN-γR2 levels were higher can be explained by lower levels of IFN-γR1:IFN-γR2 complexes if IFN-γR1 levels are reduced to accommodate ultra-high levels of IFN-γR2.

4.2.2 Inclusion of cytosolic proteins in the preassembled Type I IFN receptor complex

Even in cells in which IFN-αR1 was not overexpressed, the coexpression of IFN-αR1 and IFN-αR2c did not guarantee a biological response. Analysis of the FRET between IFN-αR1 and IFN-αR2c as a function of the number of IFN-αR1:IFN-αR2c pairs revealed that in most cells, IFN-αR1 and IFN-αR2c exhibit FRET with an affinity resembling completely nonspecific interactions (Fig. 4, Supp. Fig. 10). The affinity of the interaction between IFN-αR1 and IFN-αR2c in q3 cells that harbored active receptors was considerably lower (Fig. 2F), implying that the interaction between IFN-αR1 and IFN-αR2c somehow acquired specificity, most likely as a consequence of the interaction of IFN-αR1 and IFN-αR2c with endogenous components.

One endogenous protein that we tested was RACK-1, a scaffolding protein that juxtaposes the Jak/Stat components with IFN-αR1 and IFN-αR2c [28;29]. RACK-1 improved the apparent affinity between the two chains by at least five-fold (Fig. 4C). Separate experiments confirmed that RACK-1/TFP, IFN-αR1/CiFP and IFN-αR2c/StFP were all present in comparable levels (not shown).

Jak1 is thought to play both an enzymatic role and a structural role in the IFN-γ receptor complex [41]. Supporting this, we previously showed that Jak1 partially mediates the interaction between IFN-γR1 and IFN-γR2 [9], and may enhance the affinity of the interaction between IFN-γR1 and IFN-γR2 [11]. Because the affinity between IFN-αR1 and IFN-αR2c was slightly higher than that between IFN-αR1 and IFN-αR2b (Supp. Fig. 6), we hypothesized that Jak1 supports an interaction between IFN-αR1 and IFN-αR2c; we confirmed this by coexpressing Jak1, IFN-αR1/EGFP and IFN-αR2c/OFP, and observing significant FRET between IFN-αR1/EGFP and IFN-αR2c/OFP with an apparent affinity of about 100 relative pairs/cell (Fig. 4D). Thus Jak1 allows IFN-αR1 and IFN-αR2c to maintain an interaction.

However, the coexpression of Tyk2 and Jak1 with IFN-αR1 and IFN-αR2c dramatically increased the apparent affinity and the observed FRET efficiency between IFN-αR1 and IFN-αR2c (Fig. 4D). We hypothesize that not only the transmembrane receptor chains but also Jak1, Tyk2 and also RACK-1 need be present in comparable levels for full receptor activity to be observed. Notably, it has been shown that Janus kinases such as Tyk2 and Jak1 interact independantly of receptor activation [42]. A structural model encompassing these requirements is shown in Fig. 5. In summary, IFN-αR1 and IFN-αR2 do not directly interact strongly (Fig. 5A, green and red, respectively) unless other proteins like RACK-1 (Fig. 5B, purple), Jak1 and Tyk2 (Fig. 5C, green-yellow and blue, respectively) are present to anchor the intracellular domains together.

Figure 5.

Assembly and activation of the Type I IFN receptor complex. (A) The two transmembrane receptor chains IFN-αR1 (green) and IFN-αR2c (red) interact very weakly by themselves; the observed FRET is essentially mediated by nonspecific juxtapositions brought about by diffusion through the plasma membrane (brown and white bar). (B) RACK-1 (purple) immunoprecipitates with IFN-αR2c, and promotes the interaction between IFN-αR1 and IFN-αR2c, possibly by slowing their dissociation. (C) A complex of IFN-αR1 and Tyk2 (blue) can interact reversibly with a complex of IFN-αR2c and Jak1 (green-yellow) within the plasma membrane to form the Type I IFN receptor complex. This tetrameric complex is considerably more stable than that stabilized by RACK-1 alone. (D) We combine these scenarios so that the physiological Type I IFN receptor complex forms when a complex of IFN-αR1 and Tyk2 interacts with a complex of IFN-αR2c, Jak1, and RACK-1. The initial complex formed (E) is unstable until the intracellular proteins interact (likely initiated by Jak1). The resulting complex (F) is stable enough to allow dissociation of the extracellular domains (G). (H) Type I IFN (light blue) activates the Type I IFN receptor complex by stably contacting IFN-αR2c, and then contacting IFN-αR1 (stably or reversibly, depending on the subtype of Type I IFN bound). The resulting conformational change allows the enzymatic activation of Tyk2 and Jak1 and results in tyrosine phosphorylations (black dots) at two sites on IFN-αR1, two sites on IFN-αR2c, and on Jak1 and Tyk2.

4.3 Structural model of the activation of the Type I IFN receptor complex

We now combine our data here with that from others published and recently reviewed [2;43], allows us to forward the following model for how we believe the Type I IFN receptor complex is assembled and activated. IFN-αR1, and its constitutively bound Tyk2 kinase, reversibly interacts with IFN-αR2c and its constitutively bound kinase Jak1 (Fig. 5D). We place RACK-1 with IFN-αR2c based on immunoprecipitation experiments [28].

Because (1) many receptors bind Jak1 or Tyk2, (2) many proteins are known to bind RACK-1, and (3) the receptor complex must be preassembled to respond to Type I IFN, specificity must be involved to form the Type I IFN receptor complex in the first place. We believe that, as in the IFN-γ receptor complex, elements in the extracellular domains of IFN-αR1 and IFN-αR2c participate in the initial interaction between the IFN-αR1:Tyk2 complex and the IFN-αR2c:Jak1 complex (Fig. 5E). We hypothesize native specificity between IFN-αR1 and IFN-αR2c exists even though the interaction is very weak because when we use ECFP and EYFP (that apparently interact with each other while attached to IFN-αR1 and to IFN-αR2c), IFN-αR1 and IFN-αR2c interact with higher affinity than do two IFN-αR1 chains or two IFN-αR2c chains (Supp. Fig. 7A, 8A).

The interactions among Jak1 and Tyk2 and the intracellular domains of IFN-αR2c and IFN-αR1 are of higher affinity than are interactions between IFN-αR2c and IFN-αR1 (Table 1). We therefore believe that once the initial complex in Fig. 5E is formed, interactions among the intracellular domains occur that stabilize the preassembled complex (Fig. 5F). Because Jak1 alone increased FRET between IFN-αR1/EGFP and IFN-αR2c/OFP (Fig. 4D), we propose that the initial intracellular interaction is mediated by Jak1, that contacts IFN-αR1 and/or Tyk2. Tyk2 and RACK-1 then participate by an unknown molecular mechanism.

Table 1.

Affinities of various interactions among the components of the Type I and Type III IFN receptor complexes.

These data are summarized from Figs. 2F, 4, Supp. Figs. 2B, 4B, 5, 6, and experiments not displayed in this manuscript. The asterisk above EC₅₀ denotes relative values. If the apparent affinity is diffuse, a range is given; if subpopulations with different affinities are present, individual constants are delimited by commas; a value in parentheses indicates rarely seen data. The data are separated by types of interacting proteins.

Donor	Acceptor	extra proteins	cell line	λexc.	EC50* (pairs/cell)
interactions between accessory chains
IL-10R2/CiFP	IL-10R2/StFP		293T	488	(300),1000–2000
IFN-αR1/ECFP	IFN-αR1/EYFP		293T	442	300–2000
IFN-αR1/CiFP	IFN-αR1/StFP		293T	514	(80),300–1000
interactions between ligand-binding chains
IFN-λR1/CiFP	IFN-λR1/StFP		293T	488	(100),1000
IFN-αR2c/ECFP	IFN-αR2c/EYFP		293T	442	300–2000
IFN-αR2c/CiFP	IFN-αR2c/StFP		293T	514	40–500
IFN-αR2c/ECFP	IFN-αR2b/EYFP		293T	442	300–2000
IFN-αR2b/ECFP	IFN-αR2b/EYFP		293T	442	300–2000
interactions between ligand-binding chain and accessory chain
IFN-λR1/EGFP	IL-10R2/StFP		CHO-q3	488	3–10
IFN-λR1/CiFP	IL-10R2/StFP		293T	488	(70), 200
IFN-αR1/ECFP	IFN-αR2b/EYFP		293T	442	200–2000
IFN-αR1/ECFP	IFN-αR2c/EYFP		293T	442	70–800
IFN-αR1/EYFP	IFN-αR2c/ECFP		293T	442	(4),30–300
IFN-αR1/EGFP	IFN-αR2c/OFP		293T	442	500–1000
IFN-αR1/EGFP	IFN-αR2c/StFP		CHO-q3	488	10–50
IFN-αR1/CiFP	IFN-αR2c/StFP		293T	514	500–1000
IFN-αR1/CiFP	IFN-αR2c/StFP	RACK-1	293T	514	60–300
IFN-αR1/CiFP	IFN-αR2c/StFP	RACK-1/TFP	293T	442	50–400
IFN-αR1/OFP	IFN-αR2c/ChFP	Jak1/TFP, Tyk2/CiFP	293T	442	20–100
interactions between Janus kinase and receptor chains
Jak1/CiFP	IFN-αR1/StFP		293T	514	(30),300
Jak1/CiFP	IFN-αR2c/StFP		293T	514	50–200
Tyk2/CiFP	IFN-αR1/StFP		293T	514	30–300
Tyk2/CiFP	IFN-αR2c/StFP		293T	514	20–80

These intracellular interactions are necessary to keep the receptor preassociated because the extracellular domains readily dissociate (Fig. 5G). There is firm evidence to support this: (1) interactions between soluble Type I IFN receptors or membrane-tethered extracellular domains is not observed in the absence of ligand [6;44–46], (2) neither crosslinking nor co-immunoprecipitation of IFN-αR1 to IFN-αR2c is observed in the absence of Type I IFN [4], and (3) interactions between IFN-αR1/CiFP and IFN-αR2c/StFP were not significantly greater than the presumably nonspecific interactions seen between IFN-αR1 chains or between IFN-αR2c chains (Supp. Figs. 7B, 8B).

We hypothesize that dissociated extracellular domains allow Type I IFN (light blue) to interact directly with IFN-αR1 and IFN-αR2c (Fig. 5H). Kinetic studies have revealed that the principal and most stable contact of Type I IFN is with IFN-αR2 because the K_a of IFN-α2 with soluble, antibody-immobilized, surface-immobilized, or cell-associated IFN-αR2 is 3–10 nM [44;47;48] while the K_a of IFN-α2 with surface-immobilized IFN-αR1 (5 μM) is a thousand-fold higher; contact of IFN with IFN-αR1 likely occurs as an IFN-αR2:IFN complex [44;45].

Activation of the tyrosine kinases Jak1 and Tyk2, phosphorylation of the intracellular domains and of the tyrosine kinases (Fig 5H, black circles), and initiation of signaling cascades likely occur immediately after Type I IFN binds to the extracellular domains of the receptor complex; rigidification of the extracellular domain of IFN-αR1 and other conformational changes in the extracellular domains of both IFN-αR1 and IFN-αR2 by Type I IFN [49–51] likely propagate a conformation that promotes the immediate activation of the intracellular domains.

4.4 Conditional induction of antiviral activity

We found that antiviral activity was not consistently observed in CHO-q3 cell clones when MHC Class I induction was observed. Evidence is mounting that paths other than the Jak/Stat pathway (that is required for MHC Class I induction) need to be activated for antiviral activity to be observed. For example, the activation and tyrosine phosphorylation of Jak1, known to vary among Type I IFN subtypes, must be formidable to observe antiviral activity [52]; knockout of Akt1 and Akt2 (that phosphorylates mTOR that then phosphorylates S6 kinase) inhibited antiviral activity but did not alter MHC induction [53]. In correlation, a subset of monoclonal antibodies raised against IFN-αR2 were raised that inhibited neither binding of Type I IFN nor Jak/Stat signaling, but inhibited antiviral and antiproliferative activities induced by Type I IFNs [54].

However, the deficient antiviral activity in our system may arise from another phenomenon. First, anticytopathic protection was never complete (Fig. 2B), implying that some cells in each isolated clone are not sensitive to Type I IFN. An effective inhibition of the cytopathic effect requires that all cells be fully responsive to Type I IFN; nonresponsive cells would permit more expansion of the virus within each subpopulation than would be expected. This may also explain why only very high doses of IFN-α2 or IFN-β were required to observe viable cells (over 150 nM, Fig. 2B). Consistent with a nonuniform biological response, not all cells within the bioresponsive CHO-q3 clone 6.wo exhibited FRET between the receptor chains (Fig. 2C–F). To corroborate this, there should be evidence that the Type I IFN receptor complex can form one of two different structures from cell to cell in a population. We present data supporting this prediction elsewhere [30].

4.5 The Type III IFN receptor complex

Similar to that observed with IFN-αR1 and IFN-αR2c, only some cells expressing both IFN-λR1 and IL-10R2 exhibit biological activity in response to human IFN-λ3 (Fig. 3M–R). Both the IFN-λ receptor complex and the Type I IFN receptor complex induce antiviral states within cells, and both receptor complexes use Tyk2, Jak1, and Stat2 to propagate signaling [1]. We therefore tested whether the assembly of the IFN-λ receptor complex matches that of the Type I IFN receptor complex. Although IFN-λR1 and IL-10R2 exhibit FRET with an apparent affinity greater than that found between IFN-λR1 chains or between IL-10R2 chains (Supp. Fig. 7C, 8C), the affinity was lower than that found between IFN-γR1 and IFN-γR2, approximating that of other membrane proteins that should not interact (Supp. Fig. 10). We propose that IFN-λR1 and IL-10R2 must be expressed with other intracellular proteins like Jak1 and Tyk2 in order to consistently and stably interact in the absence of IFN-λ and observe signaling upon IFN-λ treatment.

It must be stated that although the Type I IFN and IFN-λ receptor complexes have structural, molecular, and signaling similarities, the genetic induction differs considerably between the two receptor complexes [55]. Supporting the latter, we observed that Class I MHC induction is considerably stronger with IFN-λ3 than with IFN-α2, and is comparable to the induction seen with IFN-γ (Figs. 3C, 3F, 3O, 3R).

4.6 Class 2 Cytokine receptor complexes are preassociated

We present data demonstrating that the biologically active Type I IFN receptor complex is a preassembled entity of at least IFN-αR1, IFN-αR2c, RACK-1, Jak1 and Jak2 in its resting state. The IFN-λ receptor complex also appears to be preassociated (Supp. Fig. 7C, 8C). Preassembly was also observed in the IFN-γ and IL-10 receptor complexes [8–10]. Overall, these data support our hypothesis that Class 2 cytokine receptor complexes are preassociated entities [1;56;57].

A visual display of the various relative binding affinities of the components of the three IFN receptor complexes from Table 1 is shown in Supplementary Figure 10 (red, light blue and green bars, respectively) reveals a mechanism by which Class II cytokine receptor complexes are preassembled. The weakest interactions among the receptor complex are between the receptor chains themselves (purple labels). Although the interaction between a receptor chain and a Janus kinase (orange labels) is stronger than that between two receptor chains, the interaction between the two chains is much stronger when both Janus kinases are present (blue labels), implying that interactions involving the Janus kinases maintain much of the preassociation of each receptor complex.

Nonspecific interactions (yellow-green bars) and homodimeric interactions are of similar affinity, and both sets of interactions (black labels) have lower apparent affinity than those found within most receptor complexes. We therefore believe that homodimeric receptor interactions are almost completely nonspecific.

4.7 Biological activity of receptor complexes requires preassociation

We have observed a correlation between receptor complex preassociation and biological activity. Mutations to either IFN-γR1 or IFN-γR2 that inhibited the preassociation of the two chains also disrupted biological activity [9]. We isolated two CHO-q3 cells lines expressing comparable levels of IFN-αR1/EGFP and IFN-αR2c/OFP. The clone exhibiting stronger FRET efficiency had more IFN-αR2 expressed than IFN-αR1 and was biologically active (Fig. 2), while the clone that was biologically inactive had poor FRET efficiency and a stoichiometric excess of IFN-αR1. Preassociation of IFN-αR1 and IFN-αR2c is disrupted by excessive relative levels of IFN-αR1; this disruption also prevented biological activity. We now hypothesize that a ligand-independant interaction or juxtaposition of Class 2 Cytokine receptor chains is a requirement for biological activity to be initiated by the receptor complex when bound by ligand. The correlation of preassociation and biological activity suggests that new opportunities to inhibit signaling by IFNs and IL-10-family cytokines can be sought by inhibiting the interaction between their receptor chains.

HIGHLIGHTS.

1. Most cells expressing IFN-αR1 and IFN-αR2c from cDNAs have poor FRET.

2. Most cells expressing IFN-αR1 and IFN-αR2c from cDNAs do not respond to IFN-α.

3. Stable clones exhibiting FRET between IFN-αR1 and IFN-αR2c are biologically active.

4. RACK-1 and Jak-1 promote the interaction between IFN-αR1 and IFN-αR2c.

Abbreviations Used

CiFP

enhanced citrine fluorescent protein

ECFP

enhanced cyan fluorescent protein

EGFP

enhanced green fluorescent protein

EYFP

enhanced yellow fluorescent protein

FRET

fluorescence resonance energy transfer

IFN

interferon

IFN-α

interferon-alpha

IFN-γ

interferon-gamma

IFN-λ

interferon-lambda

OFP

orange fluorescent protein

RACK-1

receptor for activated protein kinase-1

StFP

strawberry fluorescent protein