Type I interferon autoantibody footprints reveal neutralizing mechanisms and allow inhibitory decoy design

Abstract

Autoantibodies neutralizing type I interferons (IFN-Is; IFNα or IFNω) exacerbate severe viral disease, but specific treatments are unavailable. With footprint profiling, we delineate two dominant IFN-I faces commonly recognized by neutralizing IFN-I autoantibody-containing plasmas from aged individuals with HIV-1 and from individuals with severe COVID-19. These faces overlap with IFN-I regions independently essential for engaging the IFNAR1/IFNAR2 heterodimer, and neutralizing plasmas efficiently block the interaction of IFN-I with both receptor subunits in vitro. In contrast, non-neutralizing autoantibody-containing plasmas limit the interaction of IFN-I with only one receptor subunit and display relatively low IFN-I-binding avidities, thus likely hindering neutralizing function. Iterative engineering of signaling-inert mutant IFN-Is (simIFN-Is) retaining dominant autoantibody targets created potent decoys that prevent IFN-I neutralization by autoantibody-containing plasmas and that restore IFN-I-mediated antiviral activity. Additionally, microparticle-coupled simIFN-Is were effective at depleting IFN-I autoantibodies from plasmas, leaving antiviral antibodies unaffected. Our study reveals mechanisms of action for IFN-I autoantibodies and demonstrates a proof-of-concept strategy to alleviate pathogenic effects.

Graphical abstract

Figure 1.

Identification of IFNα R2 residues targeted by anti-IFNα autoAbs. (A) Anti-IFNα IgG levels in plasma samples from Aged (n = 10) and COVID ICU (n = 11) cohorts. Data are expressed as MFI FC values compared with the values of six negative control plasma samples. The dashed line represents the threshold to determine positivity, set as the mean plus 5 standard deviations of the control group. (B) Abilities of plasma samples to functionally neutralize IFNα2 at 0.2 or 10 ng/ml. Dashed lines represent neutralization thresholds that were set as the mean minus five standard deviations of the control group. Data are normalized to the luciferase signal from the control group. Pos indicates data from an anti-IFNα mAb–positive control. The green circle (A and B) indicates plasma C11 with non-neutralizing anti-IFNα IgG. (C) Schematic overview of eGFP/V5-tagged IFNα2 deletion constructs used for initial screening. (D) Western blot reactivity of anti-IFNα IgGs from donor plasma A1 to the IFNα2 deletion constructs shown in C, with anti-V5 IgG used as a loading control. (E and F) Western blot quantification of relative anti-IFNα IgG-binding levels from 20 neutralizing plasma samples to the IFNα2 deletion constructs (E) or to the indicated IFNα2 mutant constructs with amino acid stretches substituted for alanines (F). All original blot data are shown in Fig. S1 and Fig. S2. (G and H) Western blot reactivity of anti-IFNα IgGs from six plasmas to IFNα2 constructs harboring the indicated single amino acid substitutions (G) or IFNα2_R144A/E146A (H), with anti-V5 IgG used as a loading control. (I) Quantification of the data in G and H, along with additional plasmas shown in Fig. S3. For all panels, results shown are representative of at least n = 2 independent experiments. Statistical significance between groups was determined by the Mann–Whitney U test (A and B): ns, not significant; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Fig. S1, Fig. S2, and Fig. S3. FC, fold change. Source data are available for this figure: SourceData F1.

Figure S1.

Western blot reactivity of anti-IFNα IgGs from donor plasmas to IFNα2 deletion constructs. (A) Western blot reactivity of the indicated plasma samples positive (A2, A4, A5 from the Aged cohort; C11 from the COVID ICU cohort) or negative (N1–3) for anti-IFNα IgG, or a mouse anti-IFNα2 mAb, to 0.1 or 0.5 µg of recombinant IFNα2. (B) Western blot reactivity of anti-IFNα IgGs from 20 neutralizing plasmas (A1–A10 from the Aged cohort, C1–C10 from the COVID ICU cohort) to the IFNα2 deletion constructs shown in Fig. 1 C, with anti-V5 IgG used as a loading control. The V5 and A1 panels are also shown in Fig. 1. (C) Previously described structure of the IFNα2 protein (PDB: 3SE3) with identified autoAb-reactive residues 25–48 colored in orange, and identified autoAb-reactive residues 144–157 colored in green. All data shown are representative of at least n = 2 similar experiments. Source data are available for this figure: SourceData FS1.

Figure S2.

Western blot reactivity of anti-IFNα IgGs from donor plasmas to IFNα2 mutant constructs with amino acid stretches substituted for alanines. (A) Amino acid sequence alignment of IFNα2 mutant constructs containing stretches of alanine substitutions. Residues substituted for alanine in each construct are highlighted in green. Amino acid numbering refers to the mature form of IFNα2. (B–E) Previously described structures of the IFNα2 protein (PDB: 3SE3) with residues substituted for alanine in each construct colored green. (F) Western blot reactivity of anti-IFNα IgGs from 20 neutralizing donor plasmas (A1–A10 from the Aged cohort, C1–C10 from the COVID ICU cohort) to the IFNα2 alanine mutants shown in panels A–E. Anti-V5 IgG was used as a loading control. All data shown are representative of at least n = 2 similar experiments. Source data are available for this figure: SourceData FS2.

Figure S3.

Fine-mapping of the IFNα2 R2-footprint and generation of an IFNα2 R1 mutant construct. (A–C) Western blot reactivity of anti-IFNα IgGs from 20 neutralizing plasmas (A1–A10 from the Aged cohort, C1–C10 from the COVID ICU cohort) to IFNα2 constructs harboring single amino acid substitutions (A and B) or IFNα2_R144A/E146A (C). Anti-V5 IgG was used as a loading control. The V5, A1, A4, A10, C1, C2, and C3 panels are also shown in Fig. 1. (D) Validation of the HiBiT-based qIP assay for the immunoprecipitation of IFNα2 protein using negative control (anti-V5, anti-IFNω) or positive control (anti-IFNα2) mAbs, as well as plasmas A1 (positive) or N1 (negative). Mean values from n = 3 replicates are shown, and error bars indicate standard deviations. (E and F) Structure of the IFNα2 protein (PDB: 3SE3) with residues substituted for alanine to generate the R1 and R2 mutants colored green (E) or depicted as an amino acid sequence alignment (F), where the numbering refers to the mature form of IFNα2. For all data panels, results are representative of at least n = 2 similar experiments. Source data are available for this figure: SourceData FS3.

Figure 2.

Identification of IFNα R1 residues targeted by anti-IFNα autoAbs. (A) Previously determined structure of rontalizumab bound to IFNα2 (PDB: 4Z5R) with the identified R2-footprint recognized by anti-IFNα autoAbs colored in green. (B) Previously determined structure of sifalimumab bound to IFNα2 (PDB: 4YPG) with a second footprint, denoted as R1, highlighted in green. (C) Ability of Ronta, Sifa, or the indicated plasma samples to immunoprecipitate various HiBiT-tagged IFNα2 mutant proteins. Immunoprecipitation with anti-V5 IgG was used as a normalization control. (D and E) Ability of the indicated plasma samples to immunoprecipitate various HiBiT-tagged IFNα2 mutant proteins. Immunoprecipitation with anti-V5 IgG was used as a normalization control. For all data panels, mean values from n = 3 replicates are shown. Error bars indicate standard deviations. Results shown are representative of at least n = 2 independent experiments. Statistical significance between groups was determined by one-way ANOVA with Dunnett’s multiple comparison correction (C and E) or one-way ANOVA with Tukey’s multiple comparison correction (D): ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Ronta, rontalizumab; Sifa, sifalimumab.

Figure 3.

Comparison of IFNα residues targeted by autoAbs or bound by IFNAR1/IFNAR2. (A) Amino acid sequence alignment of all human IFNα subtypes. Residues that are part of the R1 or R2 autoAb footprints are highlighted in bold, and receptor-binding residues are underlined. Amino acid numbering refers to the mature form of IFNα2. (B) Previously determined structure of the IFNα2-IFNAR1-IFNAR2 protein complex (PDB: 3SE3). The identified R1 and R2 autoAb footprints are colored green. (C and D) Side-by-side comparison of the IFNα2 structure with important R1-footprint residues colored green (C) and important IFNAR1-binding residues colored orange (D). (E and F) Side-by-side comparison of the IFNα2 structure with important R2-footprint residues colored green (E) and important IFNAR2-binding residues colored orange (F).

Figure 4.

Mechanism of IFNα neutralization by anti-IFNα autoAbs. (A and B) Optimization of BLI assays to detect association of IFNα1 with IFNAR1 (A) or IFNAR2 (B). NSB: no receptor loaded, association phase with 200 nM IFNα1. Zoom-in images of the association phases are shown. (C) Abilities of rontalizumab and sifalimumab to neutralize IFNα2 at 10 or 0.2 ng/ml. Dashed lines represent neutralization thresholds set at 75% reduction compared with control donor samples. Data are normalized to the luciferase signal from the control group. Mean values and standard deviations are shown and are representative of at least n = 2 independent experiments. (D) IFNα2 avidity indexes as determined for rontalizumab and sifalimumab at different urea concentrations. The dashed line indicates the threshold for high avidity (>0.6). (E) Blocking of the interaction between IFNα1 and IFNAR1 or IFNAR2 by Ronta and Sifa. (F) Blocking of the interaction between IFNα1 and IFNAR1 or IFNAR2 by four plasmas containing neutralizing anti-IFNα autoAbs from the COVID ICU cohort. For all data panels, results shown are representative of at least n = 2 similar experiments. Ronta, rontalizumab; Sifa, sifalimumab; NSB, nonspecific binding control.

Figure 5.

Mechanism of IFNω neutralization by anti-IFNω autoAbs. (A) Anti-IFNω IgG levels in plasma samples from Aged (n = 10) and COVID ICU (n = 11) cohorts. Data are expressed as MFI FC values compared with the values of six negative control plasma samples. The dashed line represents the threshold to determine positivity, set as the mean plus 5 standard deviations of the control group. (B) Abilities of plasma samples to functionally neutralize IFNω at 0.2 or 10 ng/ml. Data are normalized to the luciferase signal from the control group. Dashed lines represent neutralization thresholds that were set as the mean minus 5 standard deviations of the control group. Pos indicates data from an anti-IFNω mAb–positive control. The orange circle (A and B) indicates plasma C4 with non-neutralizing anti-IFNω IgG. (C and D) Optimization of the concentration of IFNω required for association with IFNAR1 (C) and IFNAR2 (D) in BLI assays. Zoom-in images of the association phases are shown. (E) BLI assays to demonstrate inhibition of the interaction between IFNω and IFNAR1 or IFNAR2 by two plasmas harboring neutralizing anti-IFNω IgGs from the COVID ICU cohort. For all data panels, results are representative of at least n = 2 independent experiments. Statistical significance between groups was determined by the Mann–Whitney U test (A and B): ns, not significant; *P < 0.05; ***P < 0.001; ****P < 0.0001. FC, fold change.

Figure 6.

Non-neutralizing anti-IFN-I autoAbs display relatively low avidities. (A) BLI assays to assess inhibition of the interaction between IFNα1 and IFNAR1 or IFNAR2 by plasma C11 harboring non-neutralizing anti-IFNα IgGs. (B) BLI assays to assess inhibition of the interaction between IFNω and IFNAR1 or IFNAR2 by plasma C4 harboring non-neutralizing anti-IFNω IgGs. (C) Immunostimulatory activity of plasmas C11 and C4 compared with a HD plasma and 1,000 U/ml IFNα2. Data are normalized to the luciferase signal from untreated cells. Mean values from n = 2 replicates are shown. Error bars indicate standard deviations. (D) IFNα2 avidity indexes as determined for Ronta, three plasmas with neutralizing anti-IFNα IgGs (C1–C3), and plasma C11 with non-neutralizing anti-IFNα IgGs at different urea concentrations. (E) IFNω avidity indexes as determined for three plasmas with neutralizing anti-IFNω IgGs (C7, C8, C10) and plasma C4 with non-neutralizing anti-IFNω IgGs at different urea concentrations. The dashed lines (D and E) indicate the threshold for high avidity (>0.6). For all data panels, results are representative of at least n = 2 similar experiments. Ronta, rontalizumab. HD, healthy donor.

Figure 7.

Development of a simIFNα to counteract anti-IFNα autoAbs. (A) IFNα2 amino acid sequence with residues targeted by autoAbs in bold, and residues selected for substitution that are bound by IFNAR1/IFNAR2 underlined. Numbering refers to the mature form of IFNα2. (B) Immunostimulatory activity of the indicated IFNα2 mutants compared with IFNα2_WT on AIR cells at 16 h after stimulation. Input IFNα2 amounts were first normalized to 10⁷ HiBiT luciferase units before the stimulatory activity of each mutant was titrated out using fourfold serial dilutions. The dilutions at which each protein induced AIR cell activity 10-fold over baseline were calculated by nonlinear regression curve fitting using GraphPad Prism 10. Data represent mean values from n = 2 replicates. ND: not detectable. Blue bars indicate mutants studied further. (C) Immunostimulatory activity of IFNα2_R33A and IFNα2_R120E mutants using HEK293T cell supernatants at the indicated dilutions. Data are normalized to the luciferase signal from unstimulated cells. Data represent mean values from n = 3 replicates. (D) Western blot analysis of Avi-tagged IFNα2_R33A/R120E protein, as compared to IFNα2_WT, in cell and supernatant fractions from transfected HEK293T cells. β-Actin was used as a loading/specificity control. (E) Immunostimulatory activity of IFNα2_R33A/R120E, as compared to IFNα2_WT, on AIR cells at the indicated dilution and at 16 h after stimulation. Data are normalized to the luciferase signal obtained from the unstimulated control. Data represent mean values from n = 3 replicates. (F) Comparison of anti-IFNα IgG autoAb reactivity with IFNα2_WT and IFNα2_R33A/R120E proteins using the HiBiT-based qIP assay for six plasmas. Data represent mean values from n = 3 replicates. (G) Neutralization of 1 ng/ml IFNα2 activity on AIR cells by six plasmas, and inhibition of neutralization by preincubation of plasmas with simIFNα. Data are normalized to the luciferase signals from the mock plasma–treated conditions. (H) Relative levels of anti-IFNα IgG autoAbs and virus-specific IgG antibodies (HIV: BG505 SOSIP; COVID: S2) before (pre-) and after plasma depletion using control microparticle beads, or microparticle beads coupled to simIFNα. Data are expressed as MFI FC values made relative to the values derived from six negative control (healthy donor) plasma samples without anti-IFN-I or anti-virus IgG. (I) IFNα2 neutralization activities of the indicated anti-IFNα IgG autoAb–positive plasma samples before and after depletion as described in H. Data are normalized to the luciferase signal from a healthy control plasma–treated condition. Dashed lines in G and I indicate neutralization thresholds, set at 25% activity relative to the IFN-only condition (G) or healthy donor control (I). For all data panels, results shown are representative of at least n = 2 similar experiments. Statistical significance between groups was determined using unpaired t tests (F, H, and I): ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Fig. S4. FC, fold change. Source data are available for this figure: SourceData F7.

Figure S4.

Design of simIFNα retaining the mapped R1 and R2 autoAb footprints. (A) Previously described structure of the IFNα2 protein (PDB: 3SE3) with residues in the R1- and R2-footprints colored green. (B) Previously described structure of the IFNα2 protein (PDB: 3SE3) with residues selected for mutational analysis, based on their role in mediating interactions with IFNAR1/IFNAR2, colored orange. (C) Schematic overview of the assay to determine whether simIFN-Is can block neutralization of IFN-I by anti-IFN-I autoAbs.

Figure 8.

Development of a simIFNω to counteract anti-IFNω autoAbs. (A) Structure of the IFNα2 protein (PDB: 3SE3) with amino acid substitutions R33A and R120E (simIFNα) colored green. (B) Structure of the IFNω protein (PDB: 3SE4) with amino acid substitutions mimicking those of simIFNα colored green. (C) Amino acid sequence alignment of IFNα2 and IFNω with residues substituted to generate simIFN-Is colored green. Amino acid numbering refers to the mature form of IFNω. (D) Western blot analysis of HiBiT-tagged IFNω_R35A/R123E protein, as compared to IFNω_WT, in cell and supernatant fractions from transfected HEK293T cells. β-Actin was used as a loading/specificity control. (E) Immunostimulatory activity of IFNω_R35A/R123E, as compared to IFNω_WT, on AIR cells at the indicated dilution and at 16 h after stimulation. Data are normalized to the luciferase signal from unstimulated control cells. Data represent mean values from n = 3 replicates. (F) Validation of the HiBiT-based qIP assay for the immunoprecipitation of the IFNω protein using negative control (anti-V5, anti-IFNα2) or positive control (anti-IFNω) antibodies, as well as plasmas A6 (positive) or N1 (negative). Data represent mean values from n = 3 replicates. (G) Comparison of anti-IFNω IgG autoAb reactivity to IFNω_WT and IFNω_R35A/R123E proteins using the HiBiT-based qIP assay for six plasmas. Data represent mean values from n = 3 replicates. (H) Neutralization of 1 ng/ml IFNω activity on AIR cells by six plasmas, and inhibition of neutralization by preincubation of plasmas with simIFNω. Data are normalized to the luciferase signals from the mock plasma–treated conditions. Dashed lines indicate neutralization thresholds, set at 25% activity relative to the IFN-only condition. For all data panels, results are representative of at least n = 3 similar experiments. Statistical significance between groups was determined by unpaired t tests (G): ns, not significant. See also Fig. S4. Source data are available for this figure: SourceData F8.

Figure S5.

simIFNα and virus replication. (A) Schematic overview of the assay to determine whether simIFN-Is restore the antiviral function of IFN-I in the presence of plasma anti-IFN-I autoAbs. (B) simIFNα alone has no antiviral activity. Replication kinetics of four GFP-expressing viruses in A549 cells that were pretreated for 16 h with the conditions indicated (simIFNα used). Cells were inoculated with an MOI of 0.01 FFU/cell for PIV5-GFP, 0.1 FFU/cell for PIV2-GFP, 0.03 FFU/cell for MeV-GFP, and 0.1 TCID₅₀/cell for RSV-GFP. The GFP signal was monitored with the IncuCyte live-cell imaging system as a surrogate readout for viral replication. Mean values from n = 3 replicates are shown. Error bars indicate standard deviations.

Figure 9.

simIFN-I can restore the antiviral activity of IFN-I in the presence of neutralizing anti-IFN-I autoAbs. (A) Replication kinetics of five GFP-expressing viruses in A549 cells that were pretreated for 16 h with the conditions indicated (simIFNα used with IFNα-neutralizing plasma A1). Cells were inoculated with an MOI of 0.1 PFU/cell for H5N1-GFP, 0.01 FFU/cell for PIV5-GFP, 0.1 FFU/cell for PIV2-GFP, 0.03 FFU/cell for MeV-GFP, and 0.1 TCID₅₀/cell for RSV-GFP. The GFP signal was monitored with the IncuCyte live-cell imaging system as a surrogate readout for viral replication. Data represent mean values from n = 3 replicates, and error bars indicate standard deviations. (B) AUC values of data shown in A. Data represent mean values from n = 3 independent experiments, and error bars indicate standard deviations. (C) AUC values of H5N1-GFP replication in A549 cells that were pretreated with similar conditions as described in A, using additional IFNα-neutralizing donor plasma samples as indicated. Data represent mean values from n = 3 independent experiments, and error bars indicate standard deviations. (D) Fluorescence images of A549 cells infected with H5N1-GFP from the experiments described in C. Images are from 12 h after inoculation, where virus replication had reached peak total integrated intensity values. The scale bar represents 200 μm. For all data panels, results are representative of n = 3 similar experiments. Statistical significance between groups was determined by unpaired t tests (B and C): ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. See also Fig. S5. AUC, area under the curve.