Autoantibodies neutralizing type I IFNs are present in ~4% of uninfected individuals over 70 years old and account for ~20% of COVID-19 deaths

Circulating autoantibodies (auto-Abs) neutralizing high concentrations (10 ng/mL, in plasma diluted 1 to 10) of IFN-α and/or -ω are found in about 10% of patients with critical COVID-19 pneumonia, but not in subjects with asymptomatic infections. We detect auto-Abs neutralizing 100-fold lower, more physiological, concentrations of IFN-α and/or -ω (100 pg/mL, in 1/10 dilutions of plasma) in 13.6% of 3,595 patients with critical COVID-19, including 21% of 374 patients > 80 years, and 6.5% of 522 patients with severe COVID-19. These antibodies are also detected in 18% of the 1,124 deceased patients (aged 20 days-99 years; mean: 70 years). Moreover, another 1.3% of patients with critical COVID-19 and 0.9% of the deceased patients have auto-Abs neutralizing high concentrations of IFN-β. We also show, in a sample of 34,159 uninfected subjects from the general population, that auto-Abs neutralizing high concentrations of IFN-α and/or -ω are present in 0.18% of individuals between 18 and 69 years, 1.1% between 70 and 79 years, and 3.4% >80 years. Moreover, the proportion of subjects carrying auto-Abs neutralizing lower concentrations is greater in a subsample of 10,778 uninfected individuals: 1% of individuals <70 years, 2.3% between 70 and 80 years, and 6.3% >80 years. By contrast, auto-Abs neutralizing IFN-β do not become more frequent with age. Auto-Abs neutralizing type I IFNs predate SARS-CoV-2 infection and sharply increase in prevalence after the age of 70 years. They account for about 20% of both critical COVID-19 cases in the over-80s, and total fatal COVID-19 cases.


INTRODUCTION
Since the start of the COVID-19 pandemic in December 2019, more than 200 million people have been infected with SARS-CoV-2, resulting in at least 4 million deaths, and probably closer to 7 to 9 million deaths worldwide. Interindividual clinical variability in the course of acute infection is vast, extending from silent or mild infection in about 90% of subjects to pneumonia and respiratory failure, both requiring hospitalization, in less than 10% and 2% of cases, respectively. Age is the major epidemiological risk factor for hospitalization or death from pneumonia, the risk doubling with every five years of age (1,2). The frequencies of critical disease and Circulating autoantibodies (auto-Abs) neutralizing high concentrations (10 ng/mL, in plasma diluted 1 to 10) of IFN-α and/or -ω are found in about 10% of patients with critical COVID-19 pneumonia, but not in subjects with asymptomatic infections. We detect auto-Abs neutralizing 100-fold lower, more physiological, concentrations of IFN-α and/or -ω (100 pg/mL, in 1/10 dilutions of plasma) in 13.6% of 3,595 patients with critical COVID-19, including 21% of 374 patients > 80 years, and 6.5% of 522 patients with severe COVID-19. These antibodies are also detected in 18% of the 1,124 deceased patients (aged 20 days-99 years; mean: 70 years). Moreover, another 1.3% of patients with critical COVID-19 and 0.9% of the deceased patients have auto-Abs neutralizing high concentrations of IFN-β. We also show, in a sample of 34,159 uninfected subjects from the general population, that auto-Abs neutralizing high concentrations of IFN-α and/or -ω are present in 0.18% of individuals between 18 and 69 years, 1.1% between 70 and 79 years, and 3.4% >80 years. Moreover, the proportion of subjects carrying auto-Abs neutralizing lower concentrations is greater in a subsample of 10,778 uninfected individuals: 1% of individuals <70 years, 2.3% between 70 and 80 years, and 6.3% >80 years. By contrast, auto-Abs neutralizing IFN-β do not become more frequent with age. Auto-Abs neutralizing type I IFNs predate SARS-CoV-2 infection and sharply increase in prevalence after the age of 70 years. They account for about 20% of both critical COVID-19 cases in the over-80s, and total fatal COVID-19 cases. death from COVID-19 are higher in men than in women (3)(4)(5). With the COVID Human Genetic Effort (6), we previously reported that inborn errors of TLR3-and IRF7-dependent type I IFN induction and amplification can underlie lifethreatening COVID-19 pneumonia in a small subset of patients (7,8). Autosomal dominant disorders were found in 19 patients, but our cohort also included four previously healthy unrelated adults aged 25 to 50 years with autosomal recessive, complete IRF7 (N=2) or IFNAR1 (N=2) deficiency. These findings indicated that type I IFN immunity is essential for protective immunity to respiratory infection with SARS-CoV-2 but surprisingly redundant otherwise. We also reported that an autoimmune phenocopy of inborn errors of type I IFN-dependent immunity can underlie critical COVID-19 pneumonia (9). Indeed, autoantibodies (auto-Abs) neutralizing 10 ng/mL IFN-α2 and/or -ω were found in the blood of at least 10% of an international cohort of patients with lifethreatening COVID-19 pneumonia, but in none of the tested individuals with asymptomatic or paucisymptomatic infection (9). These auto-Abs were detected in serum or plasma diluted 1/10. The auto-Abs in the patients' undiluted blood can therefore probably neutralize as much as 100 ng/mL IFN-α2 and/or -ω. The 17 subtypes of type I IFNs, including 13 IFNα subtypes, IFN-ω, IFN-β, IFN-ε, and IFN-κ, bind to the same heterodimeric receptor (IFNAR1 and IFNAR2). (10). The 13 IFN-α subtypes and IFN-ω are closely related phylogenetically, while IFN-β, IFN-ε, and IFN-κ are more distant (9). The auto-Abs to IFN-α2 and/or -ω were mostly found in men (95%) and in the elderly (half the patients with antibodies being over the age of 65 years) (9). These findings were later replicated in independent cohorts from Amsterdam, Lyon, Madrid, New Haven, and San Francisco (11)(12)(13)(14)(15)(16). These auto-Abs against type I IFNs were found in about 0.3% of a general population sample of 1,227 subjects collected before the pandemic and aged 20 to 69 years, suggesting that they predated SARS-CoV-2 infection and caused critical COVID-19 rather than being triggered by it (9). Moreover, production of these antibodies can be genetically driven, and can begin during early childhood, as attested by their presence in almost all patients with autoimmune polyendocrine syndrome type-1 (APS-1) due to germline mutations of AIRE (17)(18)(19). APS-1 patients are, indeed, at very high risk of developing severe or critical COVID-19 pneumonia (20,21). These auto-Abs are also found in patients with combined immunodeficiency and hypomorphic mutations of RAG1 or RAG2 (22), in men with immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome and mutations of FOXP3 (23), and in women with incontinentia pigmenti and heterozygous null mutations of X-linked NEMO (9). They are also seen in patients treated with IFN-α or IFN-β (24,25), in patients with systemic lupus erythematosus (26,27), thymoma (28), or with myasthenia gravis (29,30). Finally, they underlie a third of adverse reactions to the 17D live attenuated vaccine against yellow fever virus (YFV), further suggesting that they were present in these patients, as in patients with critical COVID-19, before viral infection (31). Remarkably, for all patients tested, the auto-Abs neutralized the protective effect of ~400 pg/mL IFN-α2 against SARS-CoV-2 or YFV-17D in vitro, even when the plasma was diluted by >1/1,000 (9). As blood IFN-α concentrations during acute asymptomatic or paucisymptomatic SARS-CoV-2 infection typically range from 1 to 100 pg/mL (32,33), and IFN-α levels in the respiratory tract might be even lower yet protective, we hypothesized that auto-Abs neutralizing concentrations of type I IFNs below 10 ng/mL may underlie lifethreatening COVID-19 pneumonia in more than 10% of cases. We also hypothesized that the prevalence of auto-Abs against type I IFNs in the general, uninfected, population may increase with age and that these antibodies may be more common in men than in women.

High and intermediate levels of IgG auto-Abs against IFN-α2 and/or IFN-ω in ~20% of patients with critical COVID-19
We recruited a cohort of 3,595 patients hospitalized with critical COVID-19 pneumonia (hereafter referred to as "critical patients", and defined as pneumonia in patients with critical disease, including (i) pulmonary, with high-flow oxygen (> 6 L/min) or mechanical ventilation (continuous positive airway pressure, bilevel positive airway pressure, intubation), (ii) cardiovascular shock, or (iii) any other organ failure requiring admission to an intensive care unit), including 566 patients of our previously described cohort of 987 patients with critical COVID-19 pneumonia for whom residual samples were available (9), 623 individuals with severe COVID-19 pneumonia (with less than 6 L/min of oxygen supplementation, hereafter referred to as "severe patients"), and 1,639 individuals with asymptomatic or paucisymptomatic (mild) upper respiratory tract SARS-CoV-2 infection (the "controls", infected with SARS-CoV-2 (as demonstrated by a positive PCR and/or serological test and/or displaying typical manifestations, such as anosmia/ageusia after exposure to a confirmed COVID-19 case) who remained asymptomatic or developed mild, self-healing, ambulatory disease with no evidence of pneumonia), including 427 samples from the initial control cohort of 663 individuals (9). The patients originated from 38 different countries, across all continents. We did not include patients with moderate pneumonia, who did not receive oxygen therapy (7,9). We searched for auto-Abs against IFN-α2 and -ω, by establishing novel, sensitive, and robust assays for the detection of circulating IgG auto-Abs. We used Gyros technology (34), a high-throughput automated enzyme-linked immunosorbent assay (ELISA)-like assay capable of detecting a large range of auto-Ab levels (Fig. S1A). We  S1B). We then screened newly recruited critical or severe patients and controls from our COVID-19 cohort (Fig. 1A). We found high levels of anti-IFN-α2 and/or anti-IFN-ω auto-Abs in 6.9% of critical patients, 3.4% of patients with severe COVID-19, and only 0.6% of the asymptomatic or paucisymptomatic controls (Fig. 1A). We also found that another 12.7% of patients with critical COVID-19 had intermediate levels of anti-IFN-α2 and/or IFN-ω auto-Abs in Gyros assays (defined as levels >30 and <100, based on the distribution observed in healthy controls), whereas this was the case for 8.6% of patients with severe COVID-19 and 11% of the individuals in our control cohort. Collectively, these findings replicate and extend our previous results and those of other groups (9,(11)(12)(13)(14)(15)35), while suggesting that intermediate levels of auto-Abs against type I IFNs might be neutralizing and underlie critical disease.

Auto-Abs neutralizing 10 ng/mL IFN-α2 and/or -ω in almost 10% of the critical patients
We investigated the ability of these auto-Abs to neutralize high concentrations of type I IFNs, as defined in our previous reports (10 ng/mL IFN-α2 or IFN-ω in medium containing 1/10 plasma or serum, the equivalent of 100 ng/mL IFN-α2 or IFN-ω in undiluted plasma). We tested not only the patients with high levels of auto-Abs, as in our previous study (9), but all the available patients with critical COVID-19 (N=3,136), or severe COVID-19 (N=623), and controls (N=1,076) from our expanded cohort. We designed a high-throughput luciferase assay in which we transfected human embryonic kidney (HEK)293T cells with (i) a plasmid containing five IFNstimulated response element (ISRE) repeats and a firefly luciferase reporter, and (ii) a plasmid encoding the Renilla luciferase. We stimulated these cells with an individual recombinant type I IFN (IFN-α2 or IFN-ω), in the presence of plasma diluted 1/10 (plasma 1/10) from patients or controls. We then measured firefly luciferase induction, normalized against Renilla luciferase activity (Fig. 1B). We confirmed the robustness of this assay by comparing the results with our previous pSTAT1 flow cytometry data (9). Consistent results were obtained for all 50 patients tested with both techniques (Fig. S1C, D). We then tested all patients and controls. Most plasma samples with high auto-Ab levels (>100) against IFN-α2 according to the Gyros assay were neutralizing (Fig. S1E). We found that 9.8% (307 of 3,136) of the critical patients tested and 3.53% (22 of 623) of the severe patients had auto-Abs neutralizing IFN-α2 and/or IFN-ω, versus only 0.37% (4 of 1,076) controls ( Fig. 1C) ( Table 1 and Table S1). In the patients with neutralizing auto-Abs, these auto-Abs were able to neutralize both IFN-α2 and IFN-ω in 175 of the 307 critical patients (57%), 6 of the severe patients (27%), and none of the controls; IFN-α2 alone in 106 critical patients (34.5%), 11 severe patients (50%), and only one of the controls (25%); IFNω alone in 26 of critical patients (8.5%), 5 severe patients (22%), and 3 controls (75%) ( Table S1). None of the patients with these auto-Abs had inborn errors of TLR3-or TLR7dependent type I IFN immunity (7,36).
Auto-Abs neutralizing 100 pg/mL IFN-α2 and/or -ω in at least 13.6% of critical patients and 6.8% of severe patients As the amounts of circulating type I IFNs in infected individuals are 100 to 1,000 times lower than the amounts tested previously (32,33), we investigated the neutralization of more physiological concentrations of type I IFNs, by performing assays with 100 pg/mL type I IFN. We observed a robust response in our luciferase system, in the presence of 1/10 dilutions of control plasma (Fig. S1F). The plasma or serum was diluted 1/10, so the concentration neutralized corresponds to 1 ng/mL IFN in circulating whole blood. With diluted plasma samples from a positive control, we gained at least two orders of magnitude of sensitivity in terms of neutralizing activity, providing proof-of-concept that these auto-Abs can neutralize lower, more physiological, amounts of type I IFNs (Fig. 1D, Fig. S1G), lower than the concentrations previously tested by a factor of 100 (9). We then retested all available samples from our extended cohort. Overall, 13.6% of all critical patients tested (N=489 of 3,595), 6.5% (N=34 of 522) of the severe patients, and 1% of the controls (N=17 of 1,639) had circulating auto-Abs that neutralized 100 pg/mL IFN-α2 and/or IFN-ω in plasma 1/10 ( Fig. 1E-G) ( Table 1 and Table  S1). In the patients with neutralizing auto-Abs, these auto-Abs were able to neutralize both IFN-α2 and IFN-ω in 256 of the 489 positive critical patients (52%), 18 of the 34 severe patients (53%), and 1 of the 17 controls (6%); IFN-α2 alone in 104 critical patients (21%), 14 severe patients (41%), and 4 of the controls (23.5%); IFN-ω alone in 129 critical patients (26%), 2 severe patients (6%), and 12 controls (70%) ( Table  S1). Further dilution of a plasma sample from one patient neutralizing 100 pg/mL of type I IFNs led to a loss of neutralizing activity (Fig. 1D, Fig. S1G). Importantly, for four unrelated patients, all of whom suffered from critical COVID-19, including one who died, samples collected before COVID-19 were available and tested positive for neutralizing auto-Abs against type I IFNs. One neutralized IFN-α2 and IFN-ω at a concentration of 10 ng/mL, two neutralized both cytokines at 100 pg/mL and one IFN-ω only at 100 pg/mL (Fig. S1H). The four patients tested therefore had auto-Abs neutralizing 10 ng/mL or 100 pg/mL IFN-α2 and/or -ω before infection with SARS-CoV-2. These four patients, and another two reported in our previous study (9) all, therefore, had auto-Abs neutralizing type I IFNs before infection with SARS-CoV-2. We then assessed the risk, adjusted for age and sex, of having critical or severe disease for subjects carrying auto-Abs against each individual IFN and the different possible combinations. We found that all auto-Abs, except those neutralizing only IFN-ω at a concentration of 10 ng/mL, were highly significant risk factors in comparisons of patients with critical or severe COVID-19 with controls (Table 1 and Table S2).

Auto-Abs neutralize low concentrations of IFN-α2 protective against SARS-CoV-2
We previously reported that plasma diluted 1/100 from patients with auto-Abs against type I IFNs neutralized the ability of IFN-α2 (at a concentration of 20 pM, approximately 400 pg/mL) to block SARS-CoV-2 and YFV-17D replication in Huh-7.5 cells (9,31). Strikingly, this neutralization was seen in all patients tested, even for a 1,000-fold dilution, and, in most patients, it was more potent than the neutralizing effect of a commercially available neutralizing monoclonal Ab (mAb) against IFN-α2. These auto-Abs against type I IFNs were, therefore, able to neutralize IFN-α2 at concentrations well beyond physiological levels. We therefore hypothesized that patients with lower titers of auto-Abs against type I IFNs, which can neutralize 100 pg/mL but not 10 ng/mL in plasma diluted 1/10, would also neutralize the protective effect of IFN-α2 against SARS-CoV-2. We therefore performed our SARS-CoV-2 assay with 5 pM (~100 pg/mL) or 20 pM (~400 pg/mL) IFN-α2, on five samples from patients with lifethreatening COVID-19 and two samples from uninfected elderly individuals with auto-Abs neutralizing 100 pg/mL but not 10 ng/mL IFN-α2. As controls, we tested a commercial mAb against IFN-α2, a sample from a patient with auto-Abs neutralizing 10 ng/mL IFN-α2, and samples from three patients with life-threatening COVID-19 and three healthy controls without detectable auto-Abs against type I IFNs. We found that the 1/100 dilutions of plasma from four of the five critical COVID-19 patients and one of the two elderly individuals with auto-Abs neutralizing 100 pg/mL IFN-α2 were able to neutralize the protective effect of ~400 pg/mL IFN-α2 against SARS-CoV-2, whereas samples from all these individuals fully or partially neutralized ~100 pg/mL IFN-α2 ( Fig.  2A). No such neutralizing effect was observed for any of the auto-Ab-negative controls. Overall, our findings indicate that auto-Abs against type I IFNs capable of neutralizing 100 pg/mL IFN in 1% plasma can block the protective effect of ~100 pg/mL or ~400 pg/mL IFN-α2 against SARS-CoV-2. These findings raise the possibility that even 100-fold lower levels of auto-Abs against type I IFNs, capable of neutralizing lower, physiological concentrations of 10 pg/mL IFN-α2, may be present in an even larger proportion of patients. The testing of this hypothesis will require the development of new, more sensitive methods to screen for neutralization.

Neutralization of type I IFNs in the absence of detectable auto-Abs against IFN-α2 or -ω
The neutralization assays performed on all patients and controls revealed that some patients with neutralizing activity against 10 ng/mL IFN-α2 and/or IFN-ω, as shown in luciferase assays, did not have high, or even intermediate levels of IgG auto-Abs in Gyros assays (Fig. S1E). We also observed that some patients with neutralizing auto-Abs had low or undetectable levels of auto-Abs in Luminex assays (Fig. S1I). For these individuals, we assessed the prevalence of IgA and IgM auto-Abs against type I IFNs; we found that none of the patients tested (N=12) had detectable titers of IgA or IgM auto-Abs (Fig. S1J). We then tested the alternative hypothesis that these auto-Abs were directed against the IFNAR1 or IFNAR2 chain of type I IFN receptors, assessing the ability of plasma samples from these patients to neutralize IFN-β. None of the samples from these patients neutralized IFN-β, suggesting that the auto-Abs in these patients were not directed against IFNAR1 or IFNAR2 (Fig. S1K). An alternative plausible hypothesis is that the epitope recognized by the auto-Abs might be concealed by the binding of the cytokine to the plate (ELISA), biotinylation of the cytokine (Gyros), or covalent coupling of the cytokine to magnetic beads at lysine residues (Luminex) (19). This observation has important clinical implications, suggesting that a lack of detection of auto-Abs against type I IFNs does not rule out the possibility of such antibodies being present and having neutralization capacity.

Auto-Abs typically neutralize the 13 IFN-α subtypes and/or IFN-ω
In six patients with auto-Abs neutralizing 100 pg/mL but not 10 ng/mL IFN-α2 and/or IFN-ω, we tested the reactivity of the antibodies against the 17 type I IFNs (the 13 IFN-α forms, IFN-ω, IFN-β, IFN-ε, and IFN-κ). Like patients with auto-Abs neutralizing 10 ng/mL type I IFNs (9), those capable of neutralizing only 100 pg/mL had detectable auto-Abs against most of the 13 IFN-α forms and/or IFN-ω, albeit at lower levels (Fig. 2B). Of the six patients with auto-Abs against IFN-α and/or IFN-ω tested, only one also had auto-Abs against IFN-β and none had detectable auto-Abs against IFN-ε or IFN-κ. Overall, the patients with auto-Abs against IFN-α2 and/or IFN-ω capable of neutralizing 100 pg/mL IFN displayed patterns of reactivity to the 17 type I IFNs similar to those reported in previously described patients with auto-Abs neutralizing 10 ng/mL (9). We then set up an assay for assessing neutralization of the 13 IFN-α forms, using our luciferase-based assay. We tested two patients with auto-Abs neutralizing IFN-α2 and IFN-ω, two patients with auto-Abs neutralizing only IFN-α2, and two patients with auto-Abs neutralizing only IFN-ω. Interestingly, we found that the APS-1 patient, and the two patients with auto-Abs neutralizing 10 ng/mL IFN-α2 and IFN-ω were able to neutralize all 13 IFNα subtypes, as were the two patients with neutralizing auto-Abs against IFN-α2. Conversely, in the conditions tested, the two patients with auto-Abs neutralizing IFN-ω only, but not IFN-α2, were not able to neutralize any of the 13 IFN-α subtypes (Fig. 2C). In addition, to confirm that the IgG auto-Abs detected were indeed the cause of the neutralization activity observed, we performed an IgG depletion experiment and found that the removal of the IgG fraction abolished the neutralizing activity, whereas the purified IgG fraction had full neutralizing activity (Fig. S2A). Thus, patients with neutralizing auto-Abs against only IFN-ω do not seem to neutralize any of the 13 IFN-α subtypes, whereas patients with auto-Abs neutralizing IFN-α2 neutralize all these subtypes.

Auto-Abs neutralizing IFN-β in 1.3% of critical patients
We previously reported that auto-Abs neutralizing IFN-β were detected in only two of 101 critical patients with auto-Abs neutralizing 10 ng/mL IFN-α2 and/or IFN-ω (9). Given the potential therapeutic use of IFN-β (37, 38), and the absence of IFN-β-neutralization data for COVID-19 patients, we tested a larger number of patients and controls, including patients without auto-Abs against IFN-α or IFN-ω, for auto-Abs against IFN-β, assessing the levels and neutralizing activity of auto-Abs against 10 ng/mL IFN-β. We screened 1,773 patients with critical COVID-19 pneumonia, and found that 1.3% (N=23) had neutralizing auto-Abs against IFN-β; by contrast, such antibodies were present in none of the 187 severe patients tested and in only two of the 1,044 controls tested (0.18%) (Fig. 2D, S2B and Table S3). Interestingly, only six of the 23 (21.7%) critical patients also had auto-Abs neutralizing IFN-α2 and/or IFN-ω at 100 pg/mL, and none of the controls had such antibodies. Of note, five of these six patients had auto-Abs neutralizing all three cytokines. All the other critical patients and controls had only neutralizing auto-Abs against IFN-β. The presence of neutralizing auto-Abs against IFN-β was significantly associated with critical, but not severe, disease relative to the controls (Table 1, Tables S2-3). Interestingly, Gyros did not appear to be able to detect auto-Abs against IFN-β, perhaps because of the biotinylation of the cytokine hiding the epitope recognized by the auto-Abs. As most (78.3%) of the patients with neutralizing auto-Abs against IFN-β did not have neutralizing auto-Abs against IFN-α2 or IFN-ω, this suggests that auto-Abs against IFN-β alone may also underlie life-threatening COVID-19 (Table 1).

Neutralizing auto-Abs against type I IFNs in at least 20% of critical patients over 80 years of age
We further assessed the percentage of critical COVID-19 patients positive for neutralizing auto-Abs per decade of life and by sex ( Fig. 3A-J, S3A-W) (Tables S1-4). In our previous report, we found that critical COVID-19 patients with auto-Abs neutralizing IFN-α2 or IFN-ω at 10 ng/mL were older (more than half the patients with auto-Abs were over the age of 65 years) and more likely to be male (95% of the antibody carriers were men) (9). These results have been confirmed by other groups, albeit with a smaller proportion of men (11)(12)(13)(14)35). In our expanded cohort of patients with critical COVID-19 pneumonia (N=3,595), the mean age was 61 years and 73% of the patients were men (Fig. 3A, Table S4). We confirmed that critical patients with auto-Abs neutralizing IFN-α and/or IFN-ω at 10 ng/ml were significantly older than those not carrying auto-Abs (mean age [SD] 65.8 years [14.1] versus 61.6 years [15.5], Firth's multivariable logistic regression, P=3x10 −6 ) and more likely to be male (78.5% versus 71%, Firth's multivariable logistic regression, P=0.003). The proportion of critical COVID-19 patients with auto-Abs neutralizing 10 ng/mL IFN-α2 and/or IFN-ω increased continuously, with auto-Abs detected in 5% of patients under the age of 40 years, 6.8% of those between 40 and 49 years of age, 7.1% of those between 50 and 59 years of age, 10.7% of those between 60 and 69 years of age, 12.3% of those between 70 and 79 years, and almost 14% in those over 80 (Fig. 3C-F, S3B-I). In severe patients, the proportion of auto-Abs was much more stable with age ( Fig. S3T-W, Firth's multivariable logistic regression P=0. 16) and sex (Firth's multivariable logistic regression P=0.44). Similar results were obtained for critical COVID-19 patients with auto-Abs neutralizing 100 pg/mL IFN-α2 and/or IFN-ω, but with even higher proportions ( Fig.  3G-J, S3L-S) (Table S1). Indeed, the proportion of patients with auto-Abs ranged from 9.6% of patients below the age of 40 years, to more than 21% of those over 80 (Fig. 3G-J, S3L-S). In men, the proportion of critical COVID-19 patients carrying auto-Abs neutralizing 100 pg/mL IFN-α2 and/or IFN-ω increased to up to 23% over 80 years of age. A very different pattern was seen for auto-Abs neutralizing 10 ng/mL IFN-β, with a more stable proportion of auto-Abs carriers according to age (Fig. S3J, K, Firth's multivariable logistic regression, P=0.68) (Table S3). Overall, the prevalence of auto-Abs neutralizing 10 ng/mL and/or 100 pg/mL IFN-α2 and/or IFN-ω increased sharply with age in critical patients. A striking enrichment in patients with neutralizing auto-Abs against IFN-α2 and/or IFN-ω was observed in the elderly, with more than 20% of patients, and 23% of men, over the age of 80 years with critical COVID-19 having neutralizing auto-Abs against these type I IFNs.

Neutralizing auto-Abs against type I IFNs in at least 18% of deceased patients
The prevalence of auto-Abs against type I IFNs in patients dying from COVID-19 pneumonia is unknown. For the 3,595 patients with critical COVID-19, we analyzed data for the 1,124 who died. These patients were aged 20 days to 99 years (mean age: 71 years), 73% were male, and all had confirmed SARS-CoV-2 infection and critical COVID-19 pneumonia before death (Fig. 4A). In these patients, we analyzed the presence of neutralizing auto-Abs against type I IFNs at concentrations of 10 ng/mL and 100 pg/mL for IFN-α2 and IFN-ω, and at 10 ng/mL for IFN-β ( Fig. 4B-J, S4A-K). We found that 13.3% of the deceased patients carried auto-Abs neutralizing 10 ng/mL IFN-α2 and/or IFN-ω ( Fig. 4B-F, S4A-E). Strikingly, 18.5% carried auto-Abs neutralizing 100 pg/mL of either or both cytokines ( Fig. 4G-J, S4F-I). In addition, 0.9% had auto-Abs neutralizing IFN-β ( Fig. S4J-K). An analysis of the prevalence of neutralizing auto-Abs against type I IFNs in these patients who died of COVID-19 by decade of age revealed a moderate increase with age for auto-Abs neutralizing 10 ng/mL (Firth's multivariable logistic regression P=0.03) or 100 pg/mL (Firth's multivariable logistic regression P=0.01) (Table S1-2). For a type I IFN concentration of 100 pg/mL, the prevalence of auto-Abs neutralizing IFN-α2 and/or IFN-ω was 20% below the age of 40 years, 14% for individuals between 40 and 49 years old, 12.5% for those between 50 and 60 years old, 16.3% for those between 60 and 69 years old, 17.9% for those between 70 and 79 years old, and greater than 23% for those over the age of 80 years. Overall, at least 18% of patients dying from COVID-19 pneumonia have auto-Abs capable of neutralizing 100 pg/mL type I IFNs in plasma 1/10.

Auto-Abs capable of neutralizing IFN-α2 and/or IFN-ω at 10 ng/mL in 0.53%, and at 100 pg/mL in 2.3% of individuals from the general population
We previously tested a sample of 1,227 individuals aged 20 to 65 years from the general population collected in 2015-2017. This sample had an equal sex distribution, and we identified four individuals with auto-Abs against type I IFNs among the 1,227 tested (0.3%), suggesting that the auto-Abs pre-dated COVID-19 (9). These findings were replicated at the University of California San Francisco (UCSF) in a sample of 4,041 subjects aged 4 to 90 years (0.32%) (16). In the current study, we tested a much larger cohort of 34,159 individuals aged 20 to 100 years from the general population, with an equal distribution between the sexes (Fig. 5A). Samples were collected before 2018 for blood donors at the French blood bank (19,966 individuals), the 3C cohort (801) and in 2019 for participants in the French CONSTANCES cohort (8,850) and Cerba HealthCare (4,542). We performed serological tests for SARS-CoV-2 on the samples collected in 2019, and included only the individuals who had not been infected with SARS-CoV-2 in the sample. We used Gyros to screen this whole cohort for IgG auto-Abs against IFN-α2 and IFN-ω (Fig. 5B,  S5A). We did not measure auto-Abs against IFN-β by Gyros. We found that only 0.05% and 4.2% had anti-IFN-α2 and/or anti-IFN-ω auto-Abs above the thresholds of 100 and 30, respectively (Fig. 5B, S5A). We then assessed the ability of these antibodies to neutralize 10 ng/mL IFN-α2 or IFN-ω, for all individuals with a high or intermediate level of IgG auto-Abs against IFN-α2 or IFN-ω. We found 181 individuals with neutralizing auto-Abs, for whom 1/10 dilutions of plasma neutralized 10 ng/mL IFN-α2 and/or IFN-ω, giving an overall prevalence of 0.53% (Fig. 5C-F, S5B-I) (Table S5-6), consistent with our two previous reports (9,16). We may have slightly underestimated the number of positive individuals, as some may have had neutralizing auto-Abs at too low a titer for detection. Next, we assessed the prevalence of auto-Abs neutralizing 10 ng/mL of IFN-β in 9,583 individuals, and found an overall prevalence of 0.26% (Fig. 5G-H) (Table S5-6). Finally, for a subset of 10,778 samples, we further assessed the ability of plasma/serum samples (diluted 1/10) to neutralize 100 pg/mL IFN-α2 and/or IFN-ω in the luciferase assay ( Fig. 5I-J, 6A-H). The prevalence of auto-Abs neutralizing 100 pg/mL IFN-α2 and/or IFN-ω was 2.3% (Table S1).

Sharp increase in the prevalence of auto-Abs against IFN-α2 and/or IFN-ω after the age of 70 years in the general population
We then assessed the percentage of individuals from the general population positive for neutralizing auto-Abs per decade of life and by sex. Strikingly, we noted that the prevalence of auto-Abs neutralizing 10 ng/mL type I IFN was more than 10 times higher in individuals over the age of 70 years than in those below this age (Firth's multivariable logistic regression, P<10 −13 ) ( Fig. 5C-F, S5B-I) (Table S5-6). The prevalence of auto-Abs capable of neutralizing 10 ng/mL IFN-α2 and/or IFN-ω was 0.17% in individuals below 70 years of age, 0.9% in individuals between 70 and 75 years of age, 1.6% between the ages of 75 and 80 years and more than 4% between the ages of 80 and 85 years. Intriguingly, after 85 years, the prevalence of these antibodies decreased to about 2.6%. These findings were replicated independently in two cohorts of 703 and 376 elderly individuals from Estonia and Japan, tested with Luciferase-based immunoprecipitation assay (LIPS) and ELISA assays, respectively (Fig. S5J, K). A strong increase in the prevalence of auto-Abs neutralizing 100 pg/mL IFN-α2 and/or IFN-ω was observed with age ( Fig. 6A-H, S6A-D), with the prevalence almost doubling with every five years from 65 to 85 years of age. Indeed, 0.87% of individuals between the ages of 65 and 70 years, 1.73% of those between 70 and 75 years, and 7.1% of those between 75 and 80 years were positive for auto-Abs. Interestingly, there was an overall decrease in the prevalence of auto-Abs after 85 years of age, especially in men. By contrast, the prevalence of auto-Abs neutralizing IFN-β did not vary significantly with age (Fig. 5G, H) (Table  S4). We then assessed the risk, adjusted for age and sex, of having critical or severe disease, for subjects carrying auto-Abs against each individual IFN and the different possible combinations, relative to the general population. We also found that all auto-Abs were highly significant risk factors in comparisons of patients with critical or severe COVID-19 with the general population (Table 1 and Table S2). The strongest association was again that for auto-Abs neutralizing both IFN-α2 and IFN-ω at 10 ng/mL (OR=30, P<1 x10 −13 ), followed by those neutralizing IFN-α2 +/− IFN-ω at 10 ng/mL (OR=20, P<10 −13 ), and IFN-ω +/− IFN-α2 at 10 ng/mL (OR =15, P<10 −13 ) ( Table 1). Auto-Abs neutralizing both IFN-α2 and IFN-ω at 100 pg/mL were also highly significant risk factors (OR [95% CI]=12 [9][10][11][12][13][14][15][16], P<10 −13 ) (Table 1). Overall, these findings indicate that there is a sharp increase in the prevalence of auto-Abs neutralizing type I IFNs with age in elderly uninfected individuals, with at least 4% of those over the age of 70 years positive for auto-Abs against IFN-α2 and/or IFN-ω, and that these auto-Abs pre-date COVID-19.

DISCUSSION
We report that at least 20% of patients over 80 years of age with life-threatening COVID-19 pneumonia carry circulating auto-Abs neutralizing 100 pg/mL IFN-α2 and/or IFNω, and that such antibodies are present in more than 13.6% of patients of all ages with this condition. Some of these auto-Abs are not identified by immunoassays and are only detectable by a neutralization assay. In addition, at least 18% of deceased individuals in most age groups were found to have such auto-Abs. We also report that auto-Abs against IFN-β are found in about 1.3% and 0.9% of critical and deceased patients, most of whom do not have auto-Abs against IFN-α2 and/or IFNω. In all four patients tested for whom pre-COVID-19 samples were also available, the auto-Abs against IFN-α2 and/or IFN-ω were clearly present before SARS-CoV-2 infection, as in patients with APS-1 (9,20), and in two other previously described patients (9). Importantly, auto-Abs capable of neutralizing high concentrations of type I IFNs have been found in patients without inborn errors of TLR3-or TLR7-dependent type I IFN immunity (7,36), suggesting that both inborn errors and auto-Abs are independently causal of critical disease. It is also striking that inborn errors are more common in patients under the age of 60 years, whereas auto-Abs are more common in patients over the age of 70 years. We also report that the prevalence of auto-Abs neutralizing 10 ng/mL (and 100 pg/mL) type I IFNs, except for IFN-β, increases significantly with age in the general population, with 0.17% (1.1%) of individuals positive for these antibodies before the age of 70 years, and more than 1.4% (4.4%) positive after the age of 70 years, with a prevalence of 4.2% (7.1%) between the ages of 80 and 85 years.
These auto-Abs provide an explanation for the major increase in the risk of critical COVID-19 in the elderly. This increase with age is consistent with studies of various auto-Abs since the 1960s (39)(40)(41)(42)(43). These auto-Abs appear to have remained clinically silent in these individuals until SARS-CoV-2 infection. Our results also suggest that the neutralization of only one type I IFN (IFN-α2, IFN-ω, or IFN-β) can underlie life-threatening COVID-19 (Table 1, Tables S1-S3). Auto-Abs neutralizing 10 ng/mL IFN-β have a frequency only about one tenth that of auto-Abs neutralizing the same concentrations of IFN-α2 and/or IFN-ω ( Table 1, Table S3). We have shown that auto-Abs neutralizing 100 pg/mL type I IFN in plasma diluted 1/10, corresponding to the neutralization of 1 ng/mL IFN in vivo, can account for at least 18% of deaths and more than 20% of critical cases in the elderly >80 years of age. It is tempting to speculate that an even greater proportion of lifethreatening COVID-19 cases are due to auto-Abs neutralizing lower, physiological concentrations of type I IFNs. In vitro, concentrations of type I IFN as low as 100 pg/mL can impair SARS-CoV-2 replication in epithelial cells ( Fig. 2A). Moreover, the levels of type I IFN detected in the blood of patients with acute and benign SARS-CoV-2 infections are in the range of 1 to 100 pg/mL (32,33).
Our findings have immediate clinical applications. First, it is quick and easy to test for auto-Abs against type I IFNs in patients infected with SARS-CoV-2. Screening for these antibodies is even possible in the general population before infection. The type I IFN-neutralizing activity of these antibodies is a better read-out than their mere detection, which can be falsely negative. Tests should be performed for auto-Abs against at least three individual IFNs: IFN-α2, IFNω, and IFN-β. Particular attention should be paid to elderly individuals, and patients with known autoimmune or genetic conditions associated with auto-Abs against type I IFNs (17-20, 22, 23, 26-29). Second, patients with auto-Abs against type I IFN should be vaccinated against COVID-19 as a priority. Third, live attenuated vaccines, including YFV-17D and vaccines using the YFV-17D backbone against SARS-CoV-2, should not be given to patients with auto-Abs (31,44). Fourth, these patients appeared to be healthy before SARS-CoV-2 infection, but they should also be carefully followed for other viral illnesses, as exemplified by adverse reactions to YFV-17D (31). Fifth, in cases of SARS-CoV-2 infection in unvaccinated individuals with auto-Abs against type I IFNs, the patients should be hospitalized for prompt management. Early treatment with monoclonal antibodies (45,46) can be administered in patients without symptoms of severe COVID-19 pneumonia, and IFN-β can be administered in the absence of both pneumonia and auto-Abs against 38). Rescue treatment by plasma exchange is another therapeutic option in patients who already have pneumonia (47).
Sixth, blood products, especially plasma, should be screened for anti-IFN auto-Abs and any products containing such antibodies should be excluded from donation (13). Plasma from donors convalescing from COVID-19 should be tested for such auto-Abs (13). Seventh, given the documented innocuity and potential efficacy of a single injection, early therapy with IFN-β may be considered for the contacts of contagious subjects or during the first week after infection, even in the absence of, or before the documentation of auto-Abs against type I IFNs, in elderly patients, who have a higher risk of critical pneumonia and auto-Abs against IFN-α2 and IFN-ω, but not IFN-β (48). Another possibility would be the administration of monoclonal antibodies that can neutralize SARS-CoV-2 (45,46). Finally, it will be important to decipher the mechanism underlying the development of these auto-Abs, which may differ in patients over and under 65 years of age. Overall, our findings show that auto-Abs neutralizing concentrations of type I IFN lower than previously reported (9,(11)(12)(13)(14)(15)(16), but still higher than physiological concentrations, are common in the elderly population. Their prevalence increases with age in the uninfected general population, reaching more than 4% of individuals after the age of 70 years. They underlie about 20% of cases of critical COVID-19 pneumonia in patients over the age of 80 years, and about 20% of total COVID-19 deaths. We previously reported that they can underlie severe adverse reactions to the yellow fever live attenuated virus (31). It is tempting to speculate that they may also underlie other severe viral diseases, especially in the elderly.

Study design
We enrolled, from 38 countries across all continents, 3,595 patients with proven critical COVID-19, 623 patients with severe COVID-19, 1,639 asymptomatic or paucisymptomatic individuals with proven COVID-19, and 34,159 healthy controls in this study. We collected plasma or serum samples for all these individuals to test by immunoassay for the presence of IgG auto-Abs to type I IFNs. All subjects were recruited according to protocols approved by local institutional review boards (IRBs).

COVID-19 classification
The severity of COVID-19 was assessed for each patient as follows (7,9). "Critical COVID-19 pneumonia" was defined as pneumonia developing in patients with critical disease, whether pulmonary, with high-flow oxygen, mechanical ventilation (Continuous positive airway pressure, bilevel positive airway pressure, intubation), septic shock, or with damage to any other organ requiring admission to the intensive care unit. "Severe COVID-19" was defined as pneumonia developing in patients requiring low-flow oxygen (<6 L/min). The controls were individuals infected with SARS-CoV-2 (as demonstrated by a positive PCR and/or serological test and/or displaying typical symptoms, such as anosmia/ageusia after exposure to a confirmed COVID-19 case) who remained asymptomatic or developed mild, selfhealing, ambulatory disease with no evidence of pneumonia. Multiplex particle-based assay Serum/plasma samples were screened for autoantibodies (auto-Abs) against IFN-α2 and IFN-ω in a multiplex particlebased assay, in which magnetic beads with differential fluorescence were covalently coupled to recombinant human proteins (2.5 μg/reaction). Beads were combined and incubated with 1/100-diluted serum/plasma samples for 30 min. Each sample was tested once. The beads were then washed and incubated with PE-labeled goat anti-human IgG (1 μg/mL) for an additional 30 min. They were then washed again and used for a multiplex assay on a Bio-Plex X200 instrument.

Detection of anti-cytokine autoantibodies
Enzyme-linked immunosorbent assays (ELISA) ELISA was performed as previously described. In brief, 96-well ELISA plates (MaxiSorp; Thermo Fisher Scientific) were coated by incubation overnight at 4°C with 2 μg/mL rhIFN-α2 (Miltenyi Biotec, ref. number 130-108-984), and rhIFN-ω (Merck, ref. number SRP3061). Plates were then washed (PBS 0.005% Tween), blocked by incubation with 5% nonfat milk powder in the same buffer, washed, and incubated with 1:50 dilutions of plasma from the patients or controls for 2 hours at room temperature (or with specific mAbs as positive controls). Each sample was tested once. Plates were thoroughly washed. Horseradish peroxidase (HRP)conjugated Fc-specific IgG fractions from polyclonal goat antiserum against human IgG, IgM or IgA (Nordic Immunological Laboratories) were added to a final concentration of 2 μg/mL. Plates were incubated for 1 hour at room temperature and washed. Substrate was added and the optical density (OD) was measured. A similar protocol was used to test for

Functional evaluation of anti-cytokine autoantibodies
Luciferase reporter assays The blocking activity of anti-IFN-α2 and anti-IFN-ω auto-Abs was determined with a reporter luciferase activity. Briefly, HEK293T cells were transfected with a plasmid containing the firefly luciferase gene under the control of the human ISRE promoter in the pGL4.45 backbone, and a plasmid constitutively expressing Renilla luciferase for normalization (pRL-SV40). Cells were transfected in the presence of the X-tremeGene9 transfection reagent (Sigma-Aldrich, ref. Samples were considered neutralizing if luciferase induction, normalized against Renilla luciferase activity, was below 15% of the median values for controls tested the same day. A similar protocol was used to test for auto-Abs against 12 subtypes of IFN-α, except that we used cytokines from PBL Assay Science (catalog #11002-1) at 1 ng/mL for stimulation.

pSTAT1 induction in PBMC
The blocking activity of anti-IFN-α2 and anti-IFN-ω auto-Abs was determined by assessing STAT1 phosphorylation in healthy control cells following stimulation with the appropriate cytokines in the presence of 10% healthy control or patient serum/plasma. Surface-stained healthy control PBMCs (350,000/reaction) were cultured in serum-free RPMI medium with 10% healthy control or patient serum/plasma and were either left unstimulated or were stimulated with IFN-α2 or IFN-ω (10 ng/mL) for 15 min at 37°C. Each sample was tested once. Cells were fixed, permeabilized, and stained for intranuclear phospho-STAT1 (Y701). Cells were acquired on a BD LSRFortessa cytometer with gating on CD14 + monocytes and the data were analyzed with FlowJo software.
Luciferase-based immunoprecipitation assay (LIPS) Levels of autoantibodies against IFN-α subtypes were measured in luciferase-based immunoprecipitation assay (LIPS), as previously described (9). IFNA1, IFNA2, IFNA8, and IFNA21 sequences were inserted into a modified pPK-CMV-F4 fusion vector (PromoCell GmbH, Germany), in which the firefly luciferase replaced the NanoLuc luciferase (Promega, USA). The resulting constructs were used to transfect HEK293 cells and the IFNA-luciferase fusion proteins were collected in the tissue culture supernatant. For autoantibody screening, we combined 2x10 6 luminescence units (LU) of IFNA1, IFNA2, IFNA8 and IFNA21 in a single IP reaction mixture (pool 1), and IFNA4, IFNA5, IFNA6 and IFNA7 in another IP reaction mixture (pool 2). Serum samples were incubated with Protein G agarose beads (Exalpha Biologicals, USA) at room temperature for 1 hour in a 96-well microfilter plate (Merck Millipore, Germany), and we then added 2x10 6 luminescence units (LU) of antigen and incubated for another hour. Each sample was tested once. The plate was washed with a vacuum system and Nano-Glo® Luciferase Assay Reagent (Promega, USA) was added. Luminescence intensity was measured with a VICTOR X Multilabel Plate Reader (PerkinElmer Life Sciences, USA). The results are expressed in arbitrary units (AU), as a fold-difference relative to the mean of the negative control samples.

IgG purification
We demonstrated that the IFN-α2 or IFN-ω neutralizing activity observed was due to auto-Abs and not another plasma factor, by depleting IgG from the plasma with a protein G buffer (Pierce Protein G IgG Binding Buffer, 21011) and column (NAb Protein G Spin Columns, 89953). All buffers were homemade: glycine 0.1 M pH=2.7, Tris 1.5 M pH = 8. Total plasma was loaded onto the column. Each sample was tested once. Purified IgG were then concentrated (Pierce Protein Concentrators PES, 50K MWCO, 88504). Without eluting the IgG, the flow-through fraction (IgG-depleted) was then collected and compared to total plasma in the luciferase neutralization assay.

Statistical analysis
Odds ratios (OR) and P-values for the effect of auto-Abs neutralizing each type I IFN on critical or severe COVID-19, using asymptomatic/mild patients or the general population as controls and adjusted on age in years and sex, were estimated by means of Firth's bias-corrected logistic regression (49,50) as implemented in the "logistf" R package (https://rdrr.io/cran/logistf/). Effect of age (quantitative in years or binary +/− 65 years) and sex on the presence of neutralizing auto-Abs in each cohort (critical, severe, deceased and general population) was tested by multivariable Firth's bias-corrected logistic regression. The standard error of the prevalence of neutralizing auto-Abs to each type I IFN per age groups and sex were estimated using the Agresti-Coull approximation (51).

SARS-CoV-2 experiment
SARS-CoV-2 strain USA-WA1/2020 was obtained from BEI Resources and amplified in Caco-2 cells at 37°C. Viral titers were measured on Huh-7.5 hepatoma cells in a standard plaque assay. Caco-2 (H. sapiens, sex: male, colon epithelial) and Huh-7.5 cells (H. sapiens, sex: male, liver epithelial) were cultured in DMEM supplemented with 1% nonessential amino acids (NEAA) and 10% fetal bovine serum (FBS) at 37°C, under an atmosphere containing 5% CO2. Both cell lines have been tested negative for contamination with mycoplasma. SARS-CoV-2 experiments were performed as follows. Huh-7.5 cells were used to seed 96-well plates at a density of 7.5x10 3 cells/well. The following day, plasma samples or a commercial anti-IFN-α2 antibody (catalog number 21100-1; R&D Systems) were diluted to 1% and incubated with 5 pM (~100 pg/mL) or 20 pM (~400 pg/mL) recombinant IFN-α2 (catalog number 11101-2; R&D systems) for 1 hour at 37°C (dilutions: plasma samples = 1/100 and anti-IFN-α2 antibody = 1/1,000). Molar ratio was calculated according to the manufacturer's datasheet and with http://molbiol.ru/eng/scripts/01_04.html. Following this incubation period, the cell culture medium was removed from the 96-well plates by aspiration and replaced with the plasma/anti-IFN-α2 antibody and IFN-α2 mixture. Each sample was tested once, in triplicate. The plates were incubated overnight and the plasma/anti-IFN-α2 antibody plus IFN-α2 mixture was removed by aspiration. The cells were washed once with PBS to remove potential anti-SARS-CoV-2neutralizing antibodies and fresh medium was then added. Cells were then infected with SARS-CoV-2 by directly adding the virus to the wells. Cells infected at a MOI of 0.05 PFU/cell and incubated at 33°C for 48 hours. The cells were fixed with 7% formaldehyde, stained for SARS-CoV-2 with an anti-N antibody (catalog no. GTX135357; GeneTex), imaged and analyzed as previously described (9).         anti-IFN-β auto-Abs only (10 ng/mL) 0.96% 5  0.043 anti-IFN-β auto-Abs (10 ng/mL) and, anti-IFN-α2 and/or anti-IFN-ω auto-Abs (100 pg/mL) 0.