Human interferon-ϵ and interferon-κ exhibit low potency and low affinity for cell-surface IFNAR and the poxvirus antagonist B18R

IFNϵ and IFNκ are interferons that induce microbial immunity at mucosal surfaces and in the skin. They are members of the type-I interferon (IFN) family, which consists of 16 different IFNs, that all signal through the common IFNAR1/IFNAR2 receptor complex. Although IFNϵ and IFNκ have unique expression and functional properties, their biophysical properties have not been extensively studied. In this report, we describe the expression, purification, and characterization of recombinant human IFNϵ and IFNκ. In cellular assays, IFNϵ and IFNκ exhibit ∼1000-fold lower potency than IFNα2 and IFNω. The reduced potency of IFNϵ and IFNκ are consistent with their weak affinity for the IFNAR2 receptor chain. Despite reduced IFNAR2-binding affinities, IFNϵ and IFNκ exhibit affinities for the IFNAR1 chain that are similar to other IFN subtypes. As observed for cellular IFNAR2 receptor, the poxvirus antagonist, B18R, also exhibits reduced affinity for IFNϵ and IFNκ, relative to the other IFNs. Taken together, our data suggest IFNϵ and IFNκ are specialized IFNs that have evolved to weakly bind to the IFNAR2 chain, which allows innate protection of the mucosa and skin and limits neutralization of IFNϵ and IFNκ biological activities by viral IFN antagonists.

IFN⑀ and IFN are part of the human type-I interferon (IFN) 2 family that consists of 16 different cytokines whose signaling properties are critical for the control and elimination of microbial pathogens (1)(2)(3)(4)(5). IFNs activate innate immunity through the induction of IFN-stimulated genes (ISGs) that exhibit antiviral activity (2,6,7). IFNs also promote adaptive immunity through up-regulation of major histocompatibility complex and the induction of chemokines (8 -11). They also control cell growth and apoptosis (12,13), B-cell lineage commitment (14), and induction of T regulatory cells (15,16). Due to the importance of IFNs, medical applications of IFNs have been devel-oped including the treatment of viral infections (17), cancer (18), and multiple sclerosis (19).
The biological activities of all 16 IFNs are initiated upon binding to the cell-surface receptors IFNAR1 and IFNAR2. IFN-IFNAR interactions activate JAK1 and TYK2 kinases and the transcription factors STAT1 and STAT2 (20 -23). JAK/ STAT signaling, and additional kinases and transcription factors (24), ultimately induce IFN gene expression programs that protect the host from virus, bacteria, and even fungi (5,7,25,26). Due to the critical role IFNs play in protecting the host from infection, many pathogens produce proteins that block IFN activity at multiple steps in the IFN signaling pathway. For example, Dengue, West Nile, and Zika viruses disrupt IFN-mediated STAT2 signaling (27,28). In contrast, poxviruses encode IFN-binding proteins that neutralize IFN activity by binding to secreted IFNs, which prevents them from engaging cell-surface IFNAR1 and IFNAR2 (29). The IFN-binding proteins, B18R and B19R, were identified in vaccinia virus strains Western Reserve and Copenhagen, respectively (30,31). B18R/B19R encode the same secreted ϳ65-kDa glycoprotein that promiscuously binds to all of the type-I IFNs (32). Deletion of B18R from vaccinia virus resulted in an attenuated virus, emphasizing the importance of the type-I IFNs in controlling vaccinia virus infection (30,31).
IFN⑀ and IFN are unique from the other IFNs based on their amino acid sequences and limited expression in mucosa and skin. IFN⑀ and IFN share 35% sequence identity with one another and the 14 other IFNs. The additional 14 IFNs consist of 12 IFN␣ subtypes, IFN, and IFN␤. The IFN␣s share 77-95% sequence identity with one another. IFN␤ exhibits 35% sequence identify with all other IFNs, whereas IFN shares the highest sequence identity (60%) with both IFN⑀//␤ and the IFN␣s. Based on sequence and structural comparisons, all IFNs exhibit an ␣-helical fold consisting of five helices, which are labeled from the N terminus as helix A, B, C, D, E, and F (33)(34)(35). A unique feature of IFN⑀ is that it encodes an 18-amino acid C-terminal tail following helix F, whereas IFN has a 13-amino acid peptide insertion between helices D and E.
In addition to these novel sequence features, IFN⑀ and IFN are expressed predominantly in the female reproductive track (FRT) and keratinocytes, respectively (26,36). In fact, analysis of the IFN⑀ gene identified putative progesterone-binding sites in the promoter, suggesting hormones regulate IFN⑀ expression (37). Subsequent studies confirmed IFN⑀ expression is induced by hormones but not by viral infection (e.g. TLRs) like other IFNs (26). Consistent with IFN⑀'s expression in the reproductive tract, IFN⑀ is able to induce important restriction factors that prevent HIV-1 infection (38,39). In fact, HIV-1 negative female sex workers express high levels of IFN⑀ in their cervical tissues (40). IFN is constitutively expressed in keratinocytes, which are found in skin and in the mucosa of the FRT. Thus, IFN and IFN⑀ are both found in the FRT, but IFN is inducible by virus and dsRNA, whereas IFN⑀ is not (36). Although it appears that IFN⑀ appears to play an important role in immunity against HIV-1, IFN expression is rapidly reduced in keratinocytes that are infected with human papilloma virus strains that induce cervical cancer (41). These data strongly argue that IFN⑀ and IFN are essential components to the host response against pathogens in the FRT, whereas the specific role of IFN, produced by keratinocytes in the skin, remains to be determined.
Despite a critical role of IFN⑀ and IFN signaling in mucosa and skin, their interactions with the IFNARs and their functional activities have not been extensively characterized. To address this issue, we have expressed human IFN⑀ and IFN for comparative biophysical and functional studies with other IFN family members (often called IFN subtypes). Studies with purified IFN⑀ and IFN reveal they induce ISFG3-mediated gene expression that is ϳ1000-fold weaker than IFN␣2 or IFN. The weaker potency of IFN⑀ and IFN is consistent with their reduced affinities for the IFNAR2 receptor chain. However, IFN⑀ and IFN exhibit IFNAR1 binding affinity that is similar to IFN␣2 and IFN.
Because poxviruses cause disseminated infections of the skin and mucosa and can block type-I IFN signaling, we evaluated the ability of B18R to neutralize IFN⑀ and IFN cellular activity. Subsequent kinetic binding studies determined IFN⑀, IFN, as well as IFN␣1, exhibit reduced binding to B18R, relative to the other IFNs. Sequence and structural models of IFN⑀/-IFNAR2 and IFN⑀/-B18R complexes identified residues responsible for the disrupted IFNAR2 and B18R binding phenotypes. Our data suggests IFN⑀ and IFN have evolved to exhibit reduced IFNAR2 binding affinity and biological potency optimized for tissue-specific expression and escape from viral type-I IFN antagonists.

Expression and characterization of IFN⑀ and IFN proteins
Expression plasmids encoding human IFN⑀ and IFN protein sequences, which also encode C-terminal histidine tags, were synthesized using optimized codons. Expression studies were performed in Escherichia coli where IFN⑀ and IFN formed insoluble inclusion bodies. The guanidine-solubilized IFNs were refolded by rapid dilution into a refolding buffer. The refolded IFNs were purified using a 2-step strategy consisting of nickel affinity and cation exchange chromatography. Endotoxin levels for IFN⑀ and IFN were less than 1 EU/g, which is within the range of values observed in IFN␣ preparations obtained from commercial sources (42). SDS-PAGE gel analysis of the final purified IFN⑀ and IFN protein preparations is shown in Fig. 1. The molecular weight of IFN⑀ on SDS-PAGE gels matched its theoretical molecular weight of 23,249. However, IFN ran larger than expected (ϳ25,000 kDa) in SDS-PAGE gels, suggesting the peptide could be a frameshift product (43) or exhibit aberrant gel migration. To resolve these possibilities, MS was performed on the samples, which demonstrated the molecular masses were consistent with full-length IFN⑀ and IFN proteins without the N-terminal initiating methionine residues ( Table 1).
The IFN⑀ and IFN amino acid sequences include three and five cysteine residues, respectively. Based on prior structural and biophysical analysis of other IFNs (42,44), IFN⑀ and IFN are predicted to contain one free cysteine and form one and two disulfide bonds, respectively, in their folded forms. To confirm disulfide bond formation had occurred during the refolding process, IFN⑀ and IFN were treated with iodoacetamide (IA) in the presence, or absence, of the reducing agent dithiothreitol (DTT, Table 1). IA selectively binds to free cysteines, resulting in an increase in mass of 57 Da. Upon IA treatment in the absence DTT, the mass of IFN⑀ and IFN increased by 57 Da, which is consistent with one free cysteine in the folded proteins (Table 1). However, in the presence of DTT, the mass of IFN⑀ and IFN increased by three IA and five IA mass units, respectively. This is consistent with the protection of two (IFN⑀) and four (IFN) cysteines, due to disulfide bond formation, in the folded proteins (Table 1). Thus, the MS data are consistent with the predicted IFN⑀ and IFN disulfide bonding patterns.

IFN⑀ and IFN have disrupted IFNAR2-binding properties
Three preparations of IFN⑀ and IFN were characterized for their ability to bind to soluble IFNARs (Fig. 2). The IFNs were injected over Biacore chips coupled with IFNAR1-FC, IFNAR2-FC, or an IFNAR1/IFNAR2-FC heterodimer, as previously described (45). Binding is reported as receptor occupancy (RO) for each IFN. For comparison with other IFNs, binding studies were also performed with IFN subtypes, IFN␣2 and IFN.

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
(RO ϭ 11%). In addition to binding to the single IFNARs, we evaluated IFN binding to an IFNAR1/IFNAR2-FC heterodimer, which positions IFNAR1 and IFNAR2 close to one another in space through the FC heterodimer (45). IFN⑀ and IFN binding to IFNAR1/IFNAR2-FC remained low, relative to IFN␣2 and IFN. However, IFN binding, in particular, was significantly enhanced (RO ϭ 36%) relative to the IFNAR2 and IFNAR1 single receptor experiments.

IFN⑀ and IFN weakly induce ISGF3-mediated gene expression in reporter cells
The biological activity of IFN⑀ and IFN were compared against IFN and IFN␣2 using a reporter cell line (HL116), which contains a firefly luciferase gene downstream of the IFI6 promoter. Dose-response curves were generated from at least six independent measurements, to derive half-maximal effective concentrations (EC 50 , Fig. 2B, Table 2). Consistent with the receptor binding data, IFN⑀ (60 nM Ϯ 26 nM) and IFN (22 Ϯ 16 nM) exhibited ϳ1000-fold lower EC 50 values, compared with IFN␣2 (11 Ϯ 3 pM) or IFN (6 Ϯ 1 pM). The dose-response curves were repeated using a different reporter cell line that contains the ISG54 promoter, with similar results.

Distinct neutralization of IFN⑀ and IFN biological activity
To further understand how IFN⑀ and IFN engage cell-surface IFNARs, IFN⑀and IFN-induced gene expression was monitored in reporter cells treated with a series of reagents that block IFN biological activity (Fig. 3). For comparative purposes, IFN␣2a and IFN were also included in the analysis. To validate that reporter activity was due to IFN⑀ and IFN receptor binding, the pan-anti-IFN␣ neutralizing antibody (IFN␣ NAb) was added to the assay. Consistent with its specificity profile, the IFN␣ NAb efficiently blocked IFN␣2 reporter activity, but not the activity of IFN, IFN⑀, or IFN. To neutralize the activity of all four IFNs, the IFNs were incubated with the poxvirus antagonist B18R, which has been reported to block the activity of all IFN subtypes (31,32). As expected, B18R efficiently neutralized IFN␣2 and IFN reporter activity. However, IFN⑀ and IFN activity was only blocked at the higher concentration of B18R tested. Thus, IFN⑀ and IFN are less sensitive to neutralization by B18R than IFN␣2 and IFN.
IFN⑀and IFN-mediated bioactivity was also studied in the presence of NAbs against cell-surface IFNAR1 and IFNAR2. The IFNAR1 NAb was equally effective in blocking IFN␣2, IFN⑀, and IFN activity, but was much less effective in blocking IFN activity. The anti-IFNAR2 NAb exhibited a different blocking profile than the IFNAR1 NAb. The IFNAR2 NAb most efficiently blocked IFN␣2 activity at both high and low concentrations tested. Neutralization of IFN⑀ and IFN activity was intermediate, relative to IFN␣2 neutralization, whereas IFN activity was largely insensitive to IFNAR2 NAb inhibition. These data suggest neutralization by the IFNAR2 NAb is sensitive to differences between the IFN subtypes that go beyond receptor affinity. Soluble IFNAR1-FC and IFNAR2-FC proteins were also used to block IFN⑀ and IFN activity. As expected based on its low affinity, IFNAR1-FC was unable to block the activity of any of the IFNs. In contrast, soluble IFNAR2-FC blocked IFN␣2 and IFN activity, but was unable to block IFN⑀ or IFN activity. Thus, IFNAR1-FC and IFNAR2-FCmediated neutralization was consistent with IFN-IFNAR receptor affinities.
IFN⑀ and IFN binding to the IFNAR1/IFNAR2-FC was also determined ( Table 2). The IFNAR1/IFNAR2-FC provides a biochemical mimic of the cell-surface IFNAR1/IFNAR2 heterodimer and the resulting affinities are due to IFN binding to both IFNAR1 and IFNAR2 (45). As a control, IFN␣2 bound to the IFNAR1/IFNAR2-FC with a K D of 54 pM. Notably, this affinity constant is on the same order of magnitude of IFN␣2's EC 50 (11 pM) in the ISFG3 reporter assays (Fig. 2B, Table 2). Similarly, IFN exhibits a K D of 25 pM for IFNAR1/IFNAR2-FC and an EC 50 value of 6 pM in the reporter assay. In contrast to the picomolar affinities observed for IFN␣2 and IFN, IFN⑀ and IFN exhibit nanomolar K D values for the IFNAR1/IF-NAR2-FC (IFN⑀ K D ϭ 3.5 ϫ 10 Ϫ9 and IFN K D ϭ 2 ϫ 10 Ϫ9 ). These affinity values are consistent with the reduced activity of IFN⑀ and IFN in the reporter assays. The results obtained with the IFNAR1/IFNAR2-FC also show a 24 -66-fold reduction in the k a values of the interactions. Thus, IFN⑀ and IFN bind with reduced affinity to IFNAR2, relative to other IFN subtypes, such as IFN␣2 and IFN.

Distinct binding of IFN⑀ and IFN to the poxvirus antagonist B18R
IFN⑀ and IFN exhibit unique B18R neutralization profiles, relative to IFN␣2 and IFN (Fig. 3). To study this further, IFN⑀/ B18R and IFN/B18R interactions were compared with the other 14 IFNs using SPR (Fig. 5). The 16 IFNs were injected at two concentrations (high, 50 nM; low, 5 nM), over a Biacore chip coupled with B18R from poxvirus strain Copenhagen (Fig. 5).

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
Consistent with the reduced ability of B18R to neutralize IFN⑀ and IFN activity in the reporter assay (Fig. 3), B18R bound 80 and 64% of IFN⑀ and IFN, respectively, when injected at the high concentration. However, the percentage dropped to 14% for IFN⑀ and 7% for IFN when the IFNs were injected at the low concentration (Fig. 5B). Interestingly, the B18R binding profile of IFN␣1 mimicked IFN⑀ and IFN yielding 52% binding at high concentration and 11% at the low concentration.
Based on the results of the binding screen (Fig. 5, A and B), four IFNs that span the strong B18R binder group (IFN␣14 and IFN␣4), and the weak binders (IFN␣1 and IFN⑀), were subjected to further kinetic analysis (Fig. 5). Elucidating their binding parameters revealed the strong and weak binders were distinguished by their association rate constants (k a , Table 2). Specifically, the poor binders had small k a values (k a ϭ 1.5 ϫ 10 5 to 5.5 ϫ 10 5 M Ϫ1 s Ϫ1 ), whereas the k a values of the strong binders were 10-to 73-fold faster (k a ϭ 5.8 ϫ 10 6 to 1 ϫ 10 7 M Ϫ1 s Ϫ1 ). Importantly, all four IFNs tested exhibited very slow (t1 ⁄ 2 ϳ 10 h) dissociation rates that differ from one another by no more than 2-fold. Thus, as observed for IFNAR2 interactions, the B18R neutralization profiles of IFN⑀ and IFN correlate with reduced association rates, which reduce the ability of B18R to bind and neutralize their biological activity (Fig. 3). These observations suggest IFN⑀, IFN, and even IFN␣1, may have a selective advantage during poxvirus infections by partially escaping B18R-mediated IFN neutralization.

A molecular model to explain IFN⑀/ binding to IFNAR2 and B18R
SPR analysis revealed IFN⑀, IFN, as well as IFN␣1, have reduced association rate constants for B18R and IFNAR2, relative to the other type-I IFNs. To identify regions of IFN⑀, IFN,

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
and IFN␣1 that could explain their unique binding properties, we aligned the amino acid sequences of all of the IFNs. This analysis revealed IFN is the only IFN that has replaced the conserved IFNAR2-binding residue, Arg-56, with an asparagine (Fig. 6E). Structural and biophysical studies confirm Arg-56 makes extensive interactions with IFNAR2 and an R56A mutant drastically disrupts the IFN-IFNAR2 interaction (46,47). These data suggest the change of Arg-56 to an asparagine in IFN could be sufficient to explain its reduced affinity for IFNAR2.
To define the impact of the IFN Arg to Asn change on B18R binding, we generated a three-dimensional model of the human IFN␣2/B18R complex based on the structure of the murine ectromelia virus type-I IFN antagonist, C12R (PDB 3OQ3). The B18R model is highly reliable because C12R shares 89% sequence identity with B18R (Fig. 6). Our analysis focused on the C-terminal domain of B18R (D3 domain 91% sequence identity with C12R), which forms contacts with the IFNAR2binding site of the IFNs (Fig. 6B). Consistent with our hypothesis, Arg-56 in the IFN␣2/B18R complex forms a salt bridge with glutamate 297 (C12R numbering) that would be significantly disrupted by replacing it with an asparagine as found in IFN. Because electrostatic interactions drive association rate constants, we looked for additional salt bridge interactions that are conserved in the high affinity IFN␣2/IFNAR2 and IFN␣2/ B18R complexes. This analysis identified two structurally conserved salt bridge interactions, formed with IFN␣2 residues Arg-35 and Arg-36 (to IFNAR2 Glu-190 and B18R Asp-260) and Arg-172 (to IFNAR2 Glu-77 and B18R Asp-276). Arg-172, located in helix F, is conserved in all 16 IFNs. In contrast, Arg-35 and Arg-36 are not conserved in either IFN⑀ (Gln-35/ Glu-36) or IFN (Trp-35/Gln-36), suggesting these residues contribute to the reduced IFNAR2 and B18R association rates observed for IFN⑀ and IFN.
This analysis suggests that IFN⑀ and IFN have reduced affinities for IFNAR2 and B18R because they have replaced Arg-35, Arg-36, and Arg-56 (IFN) with alternative residues that are suboptimal for IFNAR2 and B18R binding. Although this provides a possible explanation for disruption of IFN⑀/ binding, all three arginine residues are conserved in the sequence of IFN␣1. Thus, other residues must be responsible for the reduced affinity of IFN␣1 for IFNAR2 and B18R. Further sequence analysis identified three residues that are unique to IFN␣1, relative to the high affinity B18R binders (Fig. 6). The three residues in the high affinity IFNs (Arg-45, Phe-50, and Lys-54) are replaced with Ser-45, Ser-50, and Met-54 in IFN␣1. Interestingly, these residues are all located adjacent to Arg-56 in the AB loop of IFN␣1. This suggests IFN␣1 alters IFNAR2

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
affinity by indirectly changing the conformation of Arg-56, and/or stability of the entire AB loop, rather than replacing Arg-56 with an Asn, as observed for IFN. Notably, the same three residues are also different in the sequences of IFN⑀ and IFN, relative to the high affinity IFNs (Fig. 6). This suggests IFN␣1, IFN⑀, and IFN modulate IFNAR2 and B18R affinity through salt bridge/hydrogen bond interactions that particularly manipulate Arg-56-mediated interactions.
The activity profiles of three IFN mutants (IFN-M1, -M2, and -M3, Fig. 6) confirmed the importance of the N56R substitution for increasing IFN biological potency (Fig. 7B, Table 2). However, the mutants also demonstrated that the conformation/biophysical properties of the IFN N-terminal region (e.g. helix A, AB loop, and helix B) are distinct from other IFNs. For example, making the S45R/V50F double mutant in IFN (IFN-M1), the same mutations that significantly increased the bioactivity and receptor binding affinity of IFN␣1, resulted in decreased bioactivity (Fig. 7B). In contrast, the IFN-M2 triple mutant (S45R/V50F/N56R) increased ISGF3 bioactivity (6-fold) and IFNAR1/2 binding affinity (5-fold), relative to IFN. This demonstrates IFN is not optimized to induce maximal bioactivity, or to bind with the highest affinity possible to the IFNAR receptors. Although accurate rate constants could  Table 2.
We also tested if additional amino acid changes (W35R and Q36R) in IFNM2 (e.g. IFN-M3) would further increase IFN bioactivity. However, IFN-M3 exhibited the poorest bioactivity of the three IFN mutants tested (Fig. 7B), suggesting these mutations may significantly disrupt the structure of IFN's IFNAR-binding sites.

Discussion
Genes encoding IFN⑀ and IFN protein sequences were identified over a decade ago (36,37), yet only recently has their importance in host defense, especially in the FRT, come to light (26,38,41). Although animal studies are progressing, studies that directly characterize the purified human proteins are lacking. At least one reason for the paucity of information about IFN⑀ and IFN is they cannot be expressed using protocols suitable for other IFNs (42,47). In addition, as functional studies about IFN⑀ and IFN are beginning to be reported, the pro-teins used in the assays often have not been biochemically evaluated. In addition, high quality antibodies for robust detection of IFN⑀ and IFN by flow cytometry, ELISA, or other methods, are not available (48). To address these issues, we have identified protocols for producing purified, and functionally active, IFN⑀ and IFN in E. coli. We focused on optimizing the E. coli expression/refolding system because continuing functional, biophysical, and structural studies require significant amounts of protein.
A major question we sought to address in this report is how IFN⑀ and IFN bind to the IFNAR1 and IFNAR2 receptors, compared with previously characterized IFNs. These analyses demonstrated both IFN⑀ and IFN have greatly reduced affinity for the IFNAR2 chain, yet retain IFNAR1 affinities similar to other IFNs. In addition to evaluating binding to the single IFNAR chains, we also characterized IFN⑀/ binding to an IFNAR1/IFNAR2-FC, which provides a soluble mimic of the cell-surface receptor heterodimer. All three analyses support the conclusion that the reduced potency of IFN⑀ and IFN is due to reduced receptor binding, predominantly IFNAR2. We followed up these studies by blocking IFN⑀/IFN functional Binding is reported as fraction RU bound, which corresponds to the SPR RU observed upon IFN injection over the B19R-FC surface, divided by the theoretical RU max of the surface. The type-III IFN proteins (IL29, IL28A, IL28B), which do not bind to B18R, were used as negative controls. Based on the IFN-B18R binding profiles in A and B, kinetic analyses were performed for selected strong B18R binders (IFN␣14 and IFN␣4) and weak B18R binders (IFN⑀ and IFN␣1). The experimental (black) and calculated (colored) sensorgrams for strong and weak binders are shown. The rate constants derived from the sensorgrams are found in Table 2.

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
activity with a variety of antagonists against the cell-surface receptors, soluble receptors, or using the poxvirus antagonist B18R. The main finding of this analysis was IFN⑀ and IFN exhibited a different neutralization profile against B18R, compared with IFN␣2 or IFN. To explore this in greater detail, the interaction of all 16 IFNs with B18R were evaluated by SPR. As suggested in the cell-based neutralization studies, B18R bound weakly to IFN⑀ and IFN. Surprisingly, IFN␣1 also exhibited a B18R binding profile that matched IFN⑀ and IFN. The common biophysical property between IFN⑀, IFN, and IFN␣1 is that they all interact with the IFNAR2 chain with much lower affinity than IFN␣2, IFN (this study), or other IFNs (49 -51). Rate constant analysis revealed the three low affinity IFNs, IFN⑀, IFN, and IFN␣1, all share reduced association rates for IFNAR2, but different dissociation rates. In particular, IFN␣1 dissociates very fast (k d ϭ 0.18 s Ϫ1 ) from IFNAR2, whereas IFN⑀ dissociates from IFNAR2 similar to IFN␣2 and IFN (k d ϭ 0.01 s Ϫ1 ), and IFN dissociates very slowly (k d ϭ 0.002 s Ϫ1 ) from IFNAR2, as observed for IFN␤ (51).
The amino acid sequences of IFN⑀, IFN, and IFN␣1 share ϳ30% identity with one another. Thus, global analysis of the sequences do not explain how these three IFNs bind weakly to IFNAR2 and B18R, relative to other IFNs. However, analysis of the sequences that form the IFNAR2-binding site, which were identified from the IFN/IFNAR1/IFNAR2 crystal structure (46), and a structural model of B18R derived from the ectromelia virus C12R structure (PDB 3OQ3), revealed a series of conserved contacts in both interfaces. Overall, the modeling and mutagenesis data suggest the three low affinity IFNs have evolved a strategy to modulate IFNAR2 and B18R binding that is beneficial in protecting individuals from pathogen challenge. These data support a unified molecular mechanism of manipulating the AB loop to reduce the potency of IFN⑀, IFN, and IFN␣1, relative to the other IFNs. These molecular changes may also allow IFN⑀, IFN, and IFN␣1 to partially escape neutralization by poxvirus. In the case of IFN, it is interesting that cancer-inducing strains of human papilloma virus have found an alternative strategy, inhibition of IFN expression, to evade IFN antiviral activity (41).
The biochemical properties of murine IFN⑀ (murIFN⑀) were recently characterized (52). Interestingly, murine and human IFN⑀ share 59% sequence identity, which is higher than human IFN⑀ shares with any other human type-I IFN. The similarity of the murine and human IFN⑀ sequences, at least partially, explains why murIFN⑀ is active on human cells (52). As reported in our study for human IFN⑀, murIFN⑀ also exhibits higher affinity for the murine IFNAR1 chain than for the murine IFNAR2 chain. However, in the context of the entire murine type-I IFN family, other murine IFN-IFNAR interactions appear to be very different from what is observed for the human IFNs. For example, the K D for murine IFN␣1/IFNAR2 interaction is 2.2 nM (52), whereas our analysis of the human IFN␣1-IFNAR2 interaction results in a K D of 353 nM, which is similar to the value reported by Jaks et al. (49). Interestingly, the high affinity of murine IFN␣1 for IFNAR2 is consistent with murine IFN␣1 exhibiting the same residue combinations identified for the high affinity (strong binding) human IFNs (Fig.  6E). Thus, there does not appear to be a direct correspondence in binding or function between human IFN␣1 and murine IFN␣1. However, both humans and mice have evolved IFNs that exhibit very different affinities for the IFNAR1 and IFNAR2 chains.
There is some evidence to suggest weak IFN-IFNAR2 interactions, as observed for IFN⑀, IFN, and IFN␣1, may result in distinct signaling properties of the IFNs, beyond simply a weaker response. In an extreme case, cells from IFNAR2 knockout mice, when treated with IFN␤, could still induce inflammatory signals through the IFNAR1 chain (53). However, murine IFN␤ exhibits very high affinity for the IFNAR1 chain (49), sug- Figure 6. Molecular model of unique IFN⑀/ and IFN␣1 residues that modulate IFNAR2 and B18R binding. A and B, schematic models and structures of the (A) IFN␣2/IFNAR1/IFNAR2 complex (PDB 3SE3) and the (B) IFN␣2/ B18R complex (IFN␣2 from PDB 3S9D, and B18R derived from the C12R structure, PDB 3OQ3). C, ribbon diagram of the IFN␣2 backbone with key residues that regulate IFNAR2 and B18R binding affinity, as discussed in the text, shown in yellow, orange, and magenta. D, superposition of IFNs in IFN/IFNAR2 and IFN/B18R structures with B18R D3 domain in red and the IFNAR2 D1 domain cyan. Three negatively charged residues conserved in IFNAR2 (green) and B18R (yellow) are positioned by IFN␣ arginines: Arg-35, Arg-56, and Argϭ172 (magenta). E, amino acids within the AB loop region of IFNs that distinguish strong and weak IFNAR2 and B18R binders. Amino acid substitutions made to produce IFN␣1 and IFN mutants are shown in bold.

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
gesting a possible mechanism for IFN␤-mediated activation of IFNAR1 that is not conserved by IFN⑀, IFN, or IFN␣1. Thus, if these IFNs induce IFNAR1-specific signals, it must occur by a distinct mechanism relative to a previously described mechanism for murine IFN␤. Although differences in signaling outputs by IFN⑀ and IFN require further studies, it is clear that human IFN⑀ can inhibit HIV-1 at several steps in its replication cycle (38,39). Furthermore, studies in mice demonstrate IFN⑀ controls Chlamydia and herpes simplex virus 2 (26). These data suggest that, just as IFN␣2 was formulated as an anti-hepatitis C therapy, the unique biochemical properties of IFN⑀ and IFN may be useful to protect women from a variety of pathogens that colonize surfaces of the reproductive tract (5).

Protein expression and refolding
DNA sequences encoding the mature IFN⑀ and IFN protein sequences were synthesized with optimized codons for expression in E. coli (ATUM). The codon-optimized cDNAs were subcloned into the PET21b plasmid. All mutagenesis was performed using the QuikChange site-directed mutagenesis kit (Stratagene). The plasmids were transformed into E. coli BL21-CodonPlus (DE3)-RIPL cells (Agilent) for expression. Cultures were grown at 37°C for ϳ3.5 h to A 600 values of 0.6 -0.8, before induction with 1 mM isopropyl 1-thio-␤-D-galactopyranoside. After induction, cultures were grown for an additional 3 h at 37°C before they were harvested by centrifugation. Protein expression of IFN␣1 was as previously described (42).
Expression of IFN⑀ and IFN resulted in the formation of insoluble inclusion bodies. The inclusion bodies were solubilized in 6 M guanidine HCl and full-length denatured IFN⑀ and IFN were purified by nickel affinity chromatography (Takara). The denatured proteins were subsequently refolded by a rapid 1:10 dilution into a refolding buffer consisting of 0.1 M Tris-HCl, pH 8, 50 mM NaCl, 2.5 mM EDTA, 0.2 mM oxidized GSH, 2 mM reduced GSH, and 0.8 M arginine. The refolding mixture was incubated for 18 h at 10°C and then dialyzed into 20 mM Tris, pH 7.5, and 20 mM NaCl. Refolded IFN⑀ and IFN were subsequently purified using SP cation exchange chromatography (GE Healthcare), which resulted in highly purified IFN⑀ and IFN preparations. Residual endotoxin was removed from the samples using the High Capacity Endotoxin Removal Spin Columns (Pierce). Endotoxin levels were determined using a Limulus Amoebocyte Lysate Endotoxin Quantitation Kit (Pierce).

Mass spectrometry
Mass analyses were performed using MALDI-TOF MS. Briefly, samples were analyzed in the positive mode on a Voyager Elite mass spectrometer with delayed extraction technology (PerSeptive Biosystems, Framingham MA). The acceleration voltage was set at 25 kV and 100 laser shots were summed. Experiments were performed using sinapinic acid (Sigma) at 5 mg/ml dissolved in acetonitrile, 0.1% TFA (1:1) as the matrix. The mass spectrometer was calibrated using apomyoglobin (Sigma). Samples were diluted 1:10 with matrix, and 1 l was  Table 2. SPR sensorgrams for IFN␣1-M1 binding to B18R are shown (C). Experimental (black) and calculated (blue) sensorgrams, used to derive kinetic parameters in Table 2, are shown. Kinetic parameters for IFN and IFN-M2/B18R interactions could not be uniquely determined from the sensorgrams. To compare their binding properties, 5 nM concentrations of the IFNs were injected over the B18R-FC surface (D). Binding is reported as fraction RU bound as described in the legend to Fig. 5.

Human IFN⑀ and IFN exhibit low affinity for IFNARs and B18R
pipetted onto a stainless steel 96-spot plate for analysis. The data were smoothed and processed using the Data Explorer Software (AB Sciex).

IFI6 reporter assay
HT1080 cells, containing a luciferase reporter downstream of the IFN inducible IFI6 promoter (HL116 cells), were used to characterize IFN activity. HL116 cells (4 ϫ 10 4 cells) were plated in white opaque 96-well plates and incubated overnight at 37°C. IFN were added to the cells the next morning and incubated for 5 h at 37°C. Following the 5-h incubation, the plates were moved to room temperature for 10 min, followed by the addition of 50 l of luciferase assay reagent (Steady-Glo, Promega) to each well. Luminescence was measured on a Biotek Synergy 2 plate reader and the dose-response curves were analyzed using PRISM with a three-parameter fit (GraphPad Inc.).

ISG54 reporter assay
HEK-Blue IFN-␣/␤ cells (Invivogen) were used to characterize IFN-mediated stimulation of the ISG54 promoter. Activation of the ISG54 promoter results in dose-dependent secretion of embryonic alkaline phosphatase (SEAP). For the assay, plates were seeded with 50,000 cells/well and incubated overnight. The following day, the cells were stimulated with IFNs (IFN␣2 and IFN from PBL Assay Science) in a total volume of 100 l and incubated for 18 h. Following the 18-h incubation, SEAP levels are quantified by mixing 40 l of cell supernatant with 160 l of quanti-blue substrate (Invivogen), followed by an additional 5-min incubation. The plates were then read using a Spectramax plate reader at 650 nm.

Surface plasmon resonance
SPR experiments were performed using a Biacore T200 (GE Healthcare) at 25°C using a running buffer consisting of 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.0125% P20, and 125 g/ml of bovine serum albumin (BSA). Monomeric IFNAR1-FChk, IFNAR2-FChk, and the IFNAR1/IFNAR2-FChk heterodimer proteins (as described in Ref. 45) were captured onto CM-5 sensor chips using an anti-murine FC antibody (Ab, GE Healthcare). All SPR experiments were performed in duplicate and double referenced (e.g. sensorgram data were subtracted from a control surface and from a buffer blank injection) as previously described (54). The control surface for all experiments con-sisted of the capture Ab. Approximately 100 -500 response units (RU) of the IFNAR-FChks were captured onto the chip surfaces. Fresh IFNAR-FChks were coupled to the surfaces for each injection. The surfaces were regenerated between injections with a 3-min injection of 10 mM glycine, pH 1.7. The buffer flow rate for all studies was 50 l/min. Sensorgrams were globally fit to a 1:1 binding model using Biacore T-200 evaluation software version 1.0.
Receptor occupancies were calculated from the formula (RU obs /RU max ϫ 100) where RU obs is the RU value observed at the end of 90-s injections of the IFNs over each IFNAR-FC. RU max values were derived by multiplying the number of RUs of each IFNAR-FC coupled to the chip surface by the ratio of IFN and IFNAR-FC molecular weights (M r ) used in the experiment (e.g. RU max ϭ RUs of IFNAR2-FC coupled ϫ (IFN MW/ IFNAR2-FC MW)). IFN-B18R interactions were characterized by capturing B18R-FC (R&D Systems) on CM-5 sensor chips using the anti-murine FC Ab. IFN subtypes, expressed as described in Ref. 42, were injected over the B18R-FC surface. IFN-B18R-FC interaction screening was performed at IFN concentrations of 50 and 5 nM using a flow rate of 40 l/min. For kinetic analyses, IFNs were injected over the B18R-FC surface for 2 min and dissociation was monitored for 20 min. The resulting sensorgrams were globally fit to a 1:1 binding model using Biacore T-200 evaluation software version 1.0.