Neutralizing Type I Interferon Autoantibodies in Japanese Patients With Severe COVID-19

Purpose Autoantibodies (aAbs) to type I interferons (IFNs) have been found in <1% of individuals under the age of 60 in the general population, with the prevalence increasing among those over 65. Neutralizing autoantibodies (naAbs) to type I IFNs have been found in at least 15% of patients with life-threatening COVID-19 pneumonia in several cohorts of primarily European descent. We aimed to define the prevalence of aAbs to IFN-α2 in 3,456 Japanese controls aged 20–91 and of aAbs and naAbs to IFN-α2 and IFN-ω in 627 Japanese COVID-19 patients aged 0–104, among whom were 170 critical, 235 severe, 112 moderate, 105 mild, and 5 asymptomatic infections. Methods ELISA and ISRE reporter assays were used to detect aAbs and naAbs using E. coli-produced IFNs. Results In an uninfected general Japanese population aged 20–91, we found aAbs in 0.087% of individuals. naAbs to type I IFNs (IFN-α2 and/or IFN-ω, 100 pg/mL) were detected in 10.6% of patients with critical infections, 2.6% of patients with severe infections, and ≤1% of patients with asymptomatic to mild infections. They were higher in COVID-19 patients over 50 (5.8%) than in younger patients (0%) and higher in men (5.5%) than in women (1.1%). A significant but not strong correlation between aAbs and naAbs to IFN-α2 existed (r=−0.307, p-value<0.0001), stressing the importance of measuring naAbs. Conclusion In the largest study focusing on a single ethnic and geographic group, we show that Japanese individuals with pre-existing naAbs have a much higher risk of life-threatening COVID-19 pneumonia.


Introduction
Coronavirus disease 2019 (COVID- 19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The clinical spectrum of COVID-19 varies in severity: approximately 80% of cases are asymptomatic, mild, or moderate (nonhypoxemic pneumonia), 15% of cases involve severe pneumonia (hypoxemia with O2<6 L/min), and 5% involve critical pneumonia (1). This virus is highly contagious and virulent, so the health care systems in endemic areas are in crisis. Therefore, appropriate triage of patients is needed before this infection becomes severe, and it is important to establish the risk factors that predict life-threatening COVID-19 pneumonia.
To date, the strong epidemiological risk factor for life-threatening COVID-19 is age (especially >65 years) (2)(3)(4)(5). In contrast, other factors, such as male sex, cardiovascular disease, etc., are modestly associated with COVID-19 aggravation (6-8). However, there is important interindividual variability in the spectrum of COVID-19 severity, and in some patients, the disease becomes severe without these risk factors. In some of these cases, abnormalities of the innate immune system allowed viral replication in the early stage of the infection, and then excessive production of in ammatory cytokines triggered aggravation (9)(10)(11)(12)(13)(14).
Indeed, abnormalities in the type I interferon (IFN) signaling system have been reported to be strongly involved in COVID-19 aggravation. Zhang et al. reported that inborn errors of TLR3-and IRF7-dependent type I IFN induction and ampli cation could underlie 3.5% of life-threatening cases of COVID-19 pneumonia (15). Asano et al. reported that approximately 1.8% of male patients under the age of 60 years with critical COVID-19 pneumonia had X-linked recessive TLR7 de ciency (16). Even when these genes were not congenitally abnormal, the presence of neutralizing autoantibodies (naAbs) to type I IFNs can predate the infection and become a serious risk factor in COVID-19 aggravation. Bastard et al. reported that 10.2% of patients with life-threatening COVID-19 pneumonia had naAbs to type I IFNs, while 0.33% of healthy individuals and 0% of patients with asymptomatic/mild disease had them (17). Afterward, the authors revealed that 20% of both patients with critical COVID-19 over the age of 80 years and those with fatal COVID-19 of all ages had naAbs to type I IFNs (18). These ndings suggest that approximately 5% of younger patients have a risk of aggravation due to genetic abnormalities associated with type I IFNs, and approximately 20% of older patients have a risk of aggravation due to naAbs to type I IFNs. The naAbs to type I IFNs have also been shown to underlie life-threatening adverse reactions to yellow fever vaccine (YFV) (19). However, the clinical symptoms associated with these naAbs are not apparent in other viral infections unless patients are infected with SARS-CoV-2. Therefore, it is di cult to predict the risk of aggravation due to naAbs to type I IFNs prior to SARS-CoV-2 infection. Although several subsequent studies supported these observations, accumulation of evidence is still required. It may be important to focus on a single ethnic group for precise characterization of the role of antibodies to type I IFNs in COVID-19 aggravation.
The current study aimed to clarify the prevalence of antibodies to type I IFNs in Japanese COVID-19 patients, estimating how much the presence of naAbs contribute as a risk factor in the Japanese population. This study also aimed to reveal the prevalence of aAbs in the uninfected Japanese population and discuss differences among ethnic groups compared to previous reports.

COVID-19 patients and individuals in the general population subjected to analysis
We conducted the study at Hiroshima University Hospital, Tokyo Medical and Dental University Medical Hospital, and Osaka City University Hospital. We enrolled 627 COVID-19 patients admitted to our institutes and 3,456 individuals from the general population (Fig. 1A, B, Table 1), which included 1,000 previously reported individuals (18). The detailed materials of the patients and the general population are described in the Supplemental materials and methods.
We assessed the severity of COVID-19 based on the Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia as previously described (17). A critical case was de ned as a case involving mechanical ventilation (including intubation, high ow nasal cannula, continuous positive airway pressure, bilevel positive airway pressure, etc.), septic shock, any other organ failure and/or use of ECMO in the intensive care unit. Severe cases involved oxygen therapy < 6 L/min because of pneumonia.
Moderate cases involved symptoms of mild pneumonia but no requirement for oxygen therapy. Mild cases were de ned as some mild symptoms without pneumonia. Asymptomatic cases did not involve any symptoms.
Neutralization assay of autoantibodies (aAbs) to type I IFNs We performed luciferase reporter assays with reference to previous research (17). The detailed method of neutralization assay is described in the Supplemental materials and methods.

Measurement of aAbs to type I IFNs and IFN-α2 concentration
The detailed methods of ELISA and ProQuantum TM Immunoassay are described in the Supplemental materials and methods.

Statistical analysis
The detailed method of statistical analysis is described in the Supplemental materials and methods.
naAbs to type I IFNs were frequently detected in patients with critical COVID-19 ELISA detects aAbs that react with type I IFNs, but it does not always detect their neutralizing activity. We thus measured neutralizing activity against type I IFNs using the ISRE reporter assay in sera from 627 patients with COVID-19 (18). Sera were considered to have neutralizing activity if the induction of ISRE activity, which was normalized to Renilla luciferase activity, was less than 15% of the median values of healthy controls in reference to a previous report (18). The results of the neutralizing assay are summarized in Tables 3, S3 and S4. The prevalence of naAbs to high concentrations (10 ng/mL) of IFN-α2 or IFN-ω was found in 5.9% of critical cases, 2.1% of severe cases, 0.9% of moderate cases, 0% of mild cases, and 0% of asymptomatic cases ( Fig. 2A, Table S3). In detail, 5.9% (95% CI: 3.2-10.5) of critical cases had naAbs to IFN-α2, 4.1% (95% CI: 2.0-8.3) of them to IFN-ω, and 4.1% (95% CI: 2.0-8.3) to both IFN-α2 and IFN-ω. On the other hand, less than 1% of patients with asymptomatic to moderate disease had naAbs to type I IFNs (Fig. 2B, table 3). Among patients over 50 years old, 3.6% (95% CI: 2.2-5.7) had naAbs to IFN-α2, 2.2% (95% CI: 1.2-4.1) to IFN-ω, and 2.2% (95% CI: 1.2-4.1) to both IFN-α2 and IFN-ω. In contrast, none of the patients younger than 50 years old had naAbs to type I IFNs (Fig. 2C, Table 3).
Comparison of the results of the neutralization assay and ELISA.
The neutralization assay can be considered a gold standard in assessing the biological effect of aAbs on type I IFNs. The ISRE reporter assay, especially that involving low concentrations of type I IFNs, gave us a sensitive test to assess naAbs to type I IFNs in COVID-19 patients. However, this assay is not suitable for assessing multiple samples because it requires gene introduction and takes 3 days to complete the whole process. On the other hand, ELISA-based assays enable us to analyze multiple samples with shorter experimental times. We thus compared the results of neutralizing activity against type I IFNs measured by the ISRE reporter assay with the results of aAbs to type I IFNs measured by ELISA. When the presence of naAbs to IFN-α2 was predicted by the results of aAbs to IFN-α2, the sensitivity was 50%, the speci city was 99.3%, the positive predictive value (PPV) was 66.7%, the negative predictive value (NPV) was 98.7% at 10 ng/mL (Fig. 4C), and these two detection methods had a weak negative correlation (a correlation coe cient -0.307 (95% CI: -0.376 -0.234, P value <0.0001)). For the 100 pg/mL condition, the sensitivity was 40%, the speci city was 99.3% (PPV of 66.7% and NPV of 98.0%), and these two detection methods had a weak negative correlation (a correlation coe cient -0.199 [95% CI: -0.273 -0.123, P value <0.0001]) (Fig. 4D). We thus realized that ELISA-based detection of aAbs to IFN-α2 can be an alternative method to enable testing of multiple samples, e.g., screening tests for the general population, and to evaluate antibodies to type I IFNs in sera. In contrast, for IFN-ω, ELISA failed to adequately detect the presence of naAbs to IFN-ω. Indeed, ELISA-based detection of aAbs to IFN-ω pointed out the presence of naAbs to IFN-ω (10 ng/mL condition) with a sensitivity of 10% and speci city of 98.4% (PPV of 9.1% and NPV of 98.5%) (Fig. 4E). Regarding the 100 pg/mL condition, aAbs to IFN-ω only indicated naAbs to IFN-ω with a sensitivity of 9.5% and a speci city of 98.5% (PPV of 18.2% and NPV of 96.9%) (Fig. 4F). Sera from COVID-19 patients with naAbs to IFN-α2 show low concentrations of IFN-α2. We next measured the IFN-α2 concentration in sera from the 627 COVID-19 patients with the ProQuantum TM Human IFN alfa Immunoassay, which is a qPCR-based technique. The level of IFN-α2 in sera with naAbs to IFN-α2 and/or IFN-ω was signi cantly lower. In detail, 24 of 26 serum samples (neutralized at 100 pg/mL) and 15 of 16 serum samples (10 ng/mL) were below their detection limit (< 4 pg/mL). The concentrations of IFN-α2 between sera with and without naAbs were signi cantly different (Wilcoxon rank sum test, 100 pg/mL: p=0.0038, 10 ng/mL: 0.0086) (Fig. 5A, B). These results were consistent with a previous report (17).
Prevalence of aAbs to IFN-α2 in uninfected individuals from the general Japanese population.
Measuring neutralizing activity against type I IFNs was ideal for predicting the risk of COVID-19 aggravation associated with the presence of aAbs to type I IFNs. However, neutralization assays using the ISRE reporter assay are not suitable for testing multiple samples. In the current study, ELISA-based prediction of naAbs to IFN-α2, at least in the 10 ng/mL condition, was thought to be trustworthy with a sensitivity of 50% and a speci city of 99.3%. Therefore, we estimated the prevalence of naAbs to type I IFNs in the Japanese population by detecting aAbs to IFN-α2 by ELISA. We measured aAbs to IFN-α2 in 3,456 Japanese individuals aged 20-91 years and unaffected by COVID-19. In this population, 3 individuals had aAbs to IFN-α2 (0.087% [95% CI: 0.0295-0.255%]) (Fig. 5C). These 3 individuals consisted of an 86-year-old female, a 78-year-old male, and a 42-year-old male.

Discussion
The current study investigated aAbs and naAbs to type I IFNs in 627 patients with COVID-19 before the Delta variant became predominant. This is the second largest study on the scale of the samples, the largest study focusing on a single ethnic group, and the rst in Asia. To minimize selection bias, we collected sera from COVID-19 patients from three geographically different areas (Tokyo, Osaka and Hiroshima) in Japan. The prevalence of naAbs to type I IFNs was high among patients with critical disease, elderly patients, and male COVID-19 patients. These observations were consistent with a previous study (17), providing strong evidence to support the risk of COVID-19 aggravation in individuals with naAbs to type I IFNs. The modest risk factors that are well known thus far are male sex (OR = 1.457) (7), cardiovascular disease (adjusted risk = 2.6) (6), chronic pulmonary disease (OR = 1.089) (7), diabetes mellitus with chronic complications (rate ratio = 1.295) (7). In contrast, the risk of COVID-19 aggravation among individuals with naAbs to type I IFNs was estimated to be higher (100 pg/mL of IFN-α2 and IFN-ω OR = 5.5, IFN-α2 or IFN-ω OR = 4.7, among patients over 50 years) than those with these modest risk factors. As shown in this study and a previous study (3,5,18), the prevalence of naAbs to type I IFNs is increased with older age and increases throughout life, which might be one of the reasons why age is the most striking epidemiological risk factor. On the other hand, naAbs to type I IFNs were also identi ed in 1% or less of patients with mild to moderate COVID-19. Therefore, although the presence of naAbs to type I IFNs is a strong risk factor for aggravation, a small portion of patients with these naAbs can have mild symptoms. This was also suggested by another case report (28).
It has been reported that only approximately 1% of the patients with naAbs to type I IFNs have naAbs to IFN-β (18). Therefore, IFN-β therapy might be one of the bene cial effects in cases with naAbs to type I IFNs (29)(30)(31)(32). In addition, the removal of naAbs to type I IFNs with plasma exchange may be bene cial in the treatment of COVID-19 patients (33). Since these treatments may be effective only in the early stage of the infection (21), rapid tests to evaluate naAbs to type I IFNs are necessary for appropriate therapeutic interventions. The neutralization assay of aAbs to type I IFNs is the gold standard in assessing functional autoantibodies; however, it is not practical to use for rapid testing or for multiple samples due to its complexity. We thus compared the results of ELISA and the neutralization assay using the ISRE reporter assay and identi ed a certain correlation between aAbs and naAbs for IFN-α2. On the other hand, no correlation was observed for IFN-ω. A previous study pointed out a strong association between the severity of COVID-19 and naAbs to IFN-α2; in contrast, the risk of aggravation by naAbs to IFN-ω alone was not clear (18). We thus performed a systematic study by ELISA in 3,456 individuals without COVID-19 and revealed that 0.087% of them were positive for aAbs to IFN-α2. Since aAbs to IFN-α2 predicted the presence of naAbs to IFN-α2 with sensitivities of 50% (10 ng/dL condition) and 40% (100 pg/mL condition), the prevalence of naAbs to IFN-α2 was assumed to be 0.17-0.22%. This prevalence in the general population in Japan was slightly lower than that in a previous international study (0.33%) (17). In agreement that the prevalence of naAbs in patients with critical disease in Japan (10.6%) was lower than that (13.6%) in a previous international study, the prevalence of naAbs to type I IFNs might be slightly lower, or at least equivalent, in the Japanese population.
One of the limitations of this study in the general population was that the prevalence of naAbs to type I IFNs was estimated based on the results of ELISA. To accurately estimate the prevalence, the development of a simple method for assessing neutralizing activity that enables multisample processing is required. The current neutralization assay using the ISRE reporter assay requires gene transfection for each assay. On the other hand, the establishment of a stable cell line expressing the ISRE reporter gene may partially overcome this problem. In our study, some patients did not exhibit neutralizing activity against type I IFNs even if they had a high titer of aAbs to type I IFNs. This observation was also reported in another study (27) and was thought to be a false positive. This was partly because the detected aAbs can bind to target cytokines without neutralizing activity. The other possible explanation is that these aAbs may have neutralizing activity at concentrations lower than 100 pg/mL of stimulation. The neutralizing assay in this current study used 10% sera, so naAbs to 100 pg/mL of cytokines could neutralize 1000 pg/mL of cytokines. The IFN-α2 concentrations of most patients in this study were below 100 pg/mL in sera. Therefore, the neutralization assay using 10 pg/mL cytokine stimulation might be worth trying. These possibilities are thought to be topics for future analysis. Despite having some limitations, this was the rst study to characterize the relationship between naAbs to type I IFNs and COVID-19 aggravation in a Japanese population and the second largest study on this theme, providing strong evidence to support the contribution of naAbs to type I IFNs to the risk of COVID-19 aggravation. Research (FRM, EA20170638020) and by the MD-PhD program of the Imagine Institute (with the support of the Fondation Bettencourt-Schueller).

Con icts of Interest
The authors declare no competing interests.

Availability of data and material
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Code Availability
Not applicable.

Authors' contributions
Shohei Eto, Miyuki Tsumura and Yoko Mizoguchi performed ELISA experiment, Neutralizing assay and measured IFN-a2 concentration. Shohei Eto prepared the rst draft. Shintaro Nagashima and Junko Tanaka collected samples of general population before the appearance of COVID-19 and revised the draft. Yoko Nukui, Kenichi Kashimada, Keisuke Okamoto, Akifumi Endo, Kohsuke Imai, Hirokazu Kanegane, Tomohiro Morio, Yu Nakagama, Yasutoshi Kido, Hidenori Ohnishi, Masanori Ito, and Hiroki Ohge, collected samples of patients with COVID-19 and general population after the appearance of COVID-19 and revised the draft. Paul Bastard and Jean-Laurent Casanova analyzed and interpreted the data and revised it critically for important intellectual content. Satoshi Okada designed and supervised the study and approved the nal version of the draft.

Ethics approval
This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committees and Institutional Review Board of Hiroshima University.

Consent to participate
Informed consent was obtained from all individual participants included in the study.

Consent for publication
Included subjects or their representatives have consented to publication of their data.

Figure 3
Page 19/20 naAbs to type I IFNs detected by the neutralization assay in 627 patients with COVID-19 at a cytokine concentration of 100 pg/mL. Detection of naAbs to type I IFNs of patients with COVID-19 according to its severity with the neutralization assay (n=627). Antibodies against IFN-α2 and IFN-ω were measured as type I IFNs. The cutoff value of the neutralization assay was 15%. A Dot plot of the neutralization assay stimulated by 100 pg/mL of type I IFNs. The prevalence of naAbs was high in patients with critical COVID-19. B The prevalence of naAbs to type I IFNs of patients with COVID-19 according to its severity. The prevalence of naAbs was high in patients with critical COVID-19. C The prevalence of naAbs to type I IFNs in patients with COVID-19 according to age. The prevalence of naAbs was high in patients over 50 years old. D Neutralizing activity against type I IFNs was compared between IFN-α2 and IFN-ω stimulated by 100 pg/mL. E Odds ratio for critical COVID-19 relative to other milder severities. The odds ratio was high, especially for patients over 50 years old.

Figure 4
Comparison of neutralization assays between 10 ng/mL and 100 pg/mL and ELISA and neutralization assays of 627 patients with COVID-19. A, B Neutralizing activity against type I IFNs was compared between type I IFN concentrations of 100 pg/mL and 10 ng/mL stimulated by IFN-α2 (A) or IFN-ω (B). C-F aAbs to type I IFNs by ELISA were compared with naAbs by the neutralization assay at concentrations of 10 ng/mL IFN-α2 (C), 100 pg/mL IFN-α2 (D), 10 ng/mL IFN-ω (E) and 100 pg/mL IFN-ω (F). The cutoff value of ELISA was 0.5 (O.D.) and that of the neutralization assay was 15%.

Figure 5
IFN-α2 concentration of patients with COVID-19 and prevalence of aAbs to IFN-α2 in 3,456 individuals in the general population. The IFN-α2 concentration in most of the patients with naAbs to IFN-α2 and/or IFNω was below the limit of quanti cation (<4 pg/mL). A Patients with naAbs to 100 pg/mL of IFN-α2 and/or IFN-ω (n=26) and patients without naAbs (n=601) were compared. B Patients with naAbs to 10 ng/mL of IFN-α2 and/or IFN-ω (n=16) and patients without naAbs (n=611) were compared. C aAbs to IFN-α2 in the general population were detected using ELISA. Overall, the prevalence of aAbs to IFN-α2 was 0.087% in 3,456 individuals.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. ESM.docx