Higher COVID-19 pneumonia risk associated with anti-IFN-α than with anti-IFN-ω auto-Abs in children

Auto-Abs neutralizing IFN-α and/or IFN-ω are found in about 10% of children with COVID-19 pneumonia. In uninfected children, auto-Abs neutralizing IFN-α2 are rarer (0.2%) than those neutralizing IFN-ω (2%). Auto-Abs against IFN-α thus confer a greater risk of COVID-19 pneumonia.

These genetic and immunological deficits account for about 20% of cases of critical COVID-19 pneumonia in adults.They provide a general mechanism for pathogenesis of the disease in adults, with insufficient type I IFN immunity during the first days of infection being the key driver of the disease (Campbell et al., 2022;Casanova andAbel, 2021, 2022;Casanova and Anderson, 2023;Cobat et al., 2023;Garcia-Garcia et al., 2023;Samuel, 2023;Su et al., 2023;Zhang et al., 2020a;Zhang et al., 2022a).However, much less is known about life-threatening COVID-19 pneumonia in children.Children are very rarely hospitalized for COVID-19 pneumonia, with the risk of hospitalization being only about 0.1% (O'Driscoll et al., 2021).Recessive inborn errors underlying complete deficiencies of a small set of genes governing type I IFN immunity have been found in ∼10% of an international cohort of children hospitalized for COVID-19 pneumonia (COVID Human Genetic Effort [CHGE]; https://www.covidhge.com),suggesting that the same mechanisms of disease are at work in adults and children (Zhang et al., 2022b).Most children with X-linked recessive TLR7 deficiency, or autosomal recessive IFNAR1, TBK1, STAT2, or TYK2 deficiency infected with SARS-CoV-2 suffered life-threatening COVID-19 pneumonia (Asano et al., 2021;Schmidt et al., 2021;Zhang et al., 2022b).However, the human genetic and immunological determinants of COVID-19 pneumonia in the other 90% of children in this cohort remain unknown.With the CHGE, we recruited children hospitalized for COVID-19 pneumonia, including children with recessive inborn errors affecting type I IFNs (Zhang et al., 2022b).The rare children with autoimmune polyendocrine syndrome type I (APS-1), who harbor high titers of auto-Abs neutralizing type I IFNs from infancy onwards, are also known to be at high risk of life-threatening COVID-19 (Bastard et al., 2020(Bastard et al., , 2021c;;Meisel et al., 2021;Valenzise et al., 2023).Furthermore, two Brazilian children hospitalized for severe COVID-19 were subsequently diagnosed with APS-1, following the identification of such auto-Abs (Schidlowski et al., 2022).We, therefore, tested the hypothesis that some of the unvaccinated children without APS-1 who suffered from COVID-19 pneumonia may also have harbored auto-Abs against type I IFNs before infection with SARS-CoV-2.We also assessed the prevalence of auto-Abs against type I IFNs in an uninfected pediatric population.We thus tested whether the main conclusions drawn with samples from adults, including both uninfected individuals and patients with various SARS-CoV-2 infections of various degrees of severity, also apply to children.

Results
Auto-Abs against type I IFNs in 19 of 183 children with COVID-19 pneumonia We studied 183 previously healthy unvaccinated children hospitalized for COVID-19 pneumonia, including eight patients with one of the 15 known recessive IEI affecting type I IFNs (Zhang et al., 2022b) (see study flowchart).The patients were recruited via the CHGE and originated from nine countries (Brazil, France, Italy, Morocco, Saudi Arabia, Spain, Peru, Turkey, and Ukraine).The patients had a median age of 11 years and a mean age of 9 years (range: 0-17 years), and 50% were girls.As previously reported (Bastard et al., 2021a;Gervais et al., 2023), we used plasma or serum samples (diluted 1:10) from the patients for the assessment of anti-IFN-α2 IgG levels by Gyros (Fig. S1 A), and for assessments of neutralization activity against non-glycosylated IFN-α2 and IFN-ω at concentrations of 10 ng/ml and 100 pg/ml, and against glycosylated IFN-β at a concentration of 10 ng/ml (Fig. 1, A and B; and Table 1).The samples were obtained while the patients were hospitalized for COVID-19.The cohort studied here includes 53 of the 112 children previously reported in a study focusing on recessive IEI affecting type I IFNs (Zhang et al., 2022b).Only one of the eight children from our previously published cohort of 12 children with IEI affecting type I IFN immunity (Zhang et al., 2022b) tested had auto-Abs against type I IFNs.This patient had TLR7 deficiency and carried auto-Abs neutralizing 100 pg/ml IFN-ω.No plasma samples were available for the remaining four children from this cohort who were therefore not included in the cohort of 183 children studied here.Conversely, 130 of the children studied here were not included in the cohort investigated in the previous study (Zhang et al., 2022b).Our global cohort of 183 children consisted of 136 cases of critical disease requiring intensive care unit (ICU) hospitalization with high-flow oxygen (>6 L/min) supplementation or mechanical ventilation, 35 cases of severe COVID-19 pneumonia requiring <6 L/min of oxygen supplementation, and 12 with moderate infections that did not require oxygen supplementation.In this context, we identified 10 (5.5%) children with auto-Abs neutralizing IFN-α2: three (1.6%) with auto-Abs against IFN-α2 only, five (2.7%) with auto-Abs against IFN-α2 and IFN-ω, and two (1.1%) with auto-Abs against IFN-α2, IFNω, and IFN-β.In addition, nine children (4.9%) had auto-Abs neutralizing IFN-ω only.In total, 19 children (10.4%) had neutralizing auto-Abs against type I IFNs: 14 with critical, 4 with severe, and 1 with moderate COVID-19 pneumonia (Fig. 1, A and B; and Table 2).Moreover, plasma from 7 (3.8%)children contained auto-Abs that neutralized at least one IFN at a concentration of 10 ng/ml, whereas 12 (6.6%)children had auto-Abs neutralizing IFN only at a concentration of 100 pg/ml (Fig. 1, A and B).All patients with Gyros values over 200 had neutralizing auto-Abs (Fig. S1 B).Auto-Abs neutralizing IFN-α2, IFN-β, and/or IFN-ω were, thus, detected at the onset of COVID-19 pneumonia in 19 of the 183 unvaccinated children studied (10.4%): 10 (5.5%) with auto-Abs neutralizing IFN-α2 (three [1.6%] IFN-α2 only, five [2.7%]IFN-α2 and IFN-ω, and two [1.1%]IFN-α2, IFN-ω, and IFNβ), and nine (4.9%) with auto-Abs neutralizing IFN-ω only.3).They originated from four different countries (France, Pakistan but living in Italy, Spain, and Turkey), with a particularly large proportion of patients from Turkey (79%).Turkish children did not account for a disproportionate number of the individuals with auto-Abs against type I IFNs (79% of Turkish children with antibodies (Abs) versus 66% without; P = 0.31; Fisher's exact test).None of these individuals had previously suffered from other severe viral infections known to be associated with these auto-Abs, such as live attenuated yellow fever viral vaccine disease (Bastard et al., 2021b), West Nile virus encephalitis (Gervais et al., 2023), critical influenza pneumonia (Zhang et al., 2022c), critical MERS pneumonia (Alotaibi et al., 2023), or severe zoster infection (Busnadiego et al., 2022;Mathian et al., 2022;Mogensen et al., 1981;Pozzetto et al., 1984;Walter et al., 2015).The two children under the age of 6 mo may have received the auto-Abs via materno-fetal transmission.APS-1 was excluded clinically (no other clinical features of autoimmunity) in all 19 patients and genetically in all 12 for whom DNA samples were available.Only 1 of the 12 children for whom DNA was available carried any of the known IEI-affecting type I IFNs.This patient produced auto-Abs neutralizing the lower concentration of IFN-ω and had X-linked TLR7 deficiency (Asano et al., 2021).All the children were hospitalized for pneumonia following SARS-CoV-2 infection.Among the 19 children with auto-Abs neutralizing type I IFNs, 1 (5%) with auto-Abs neutralizing IFN-ω only (at a concentration of 100 pg/ml) had moderate COVID-19 pneumonia, 4 children (21%) were hospitalized for severe COVID-19 pneumonia, and 14 children (74%) had critical   3).One of the children with critical disease had cardiological, neurological, cutaneous, and gastrointestinal manifestations of the multisystem inflammatory syndrome in children (MIS-C) (Lee et al., 2023;Sancho-Shimizu et al., 2021).All the patients survived, and all except the patients with additional manifestations had positive SARS-CoV-2 RT-PCR results on samples from the respiratory tract.
The patient with additional manifestations had critical pneumonia and a subsequent positive serological test demonstrated infection.Overall, these findings suggest that auto-Abs against type I IFNs can underlie life-threatening COVID-19 pneumonia in a significant proportion of previously healthy unvaccinated children.

Auto-Abs neutralize all 12 IFN-α subtypes
We assessed the neutralization of the 12 individual IFN-α subtypes at the intermediate concentration of 1 ng/ml.There are 13 IFNA loci, but only 12 IFN-α proteins, as the products of IFNA1 and IFNA13 are identical (Moreau et al., 2023).Interestingly, for all patients with auto-Abs neutralizing 10 ng/ml IFN-α2, all 12 IFN-α subtypes were neutralized (Fig. 2 A).None of the patients with auto-Abs against IFN-ω but without detectable auto-Abs against IFN-α2 displayed neutralization of any of the 12 IFN-α subtypes.However, the patient tested for whom IFN-α2 neutralization was observed at 100 pg/ml, but not 10 ng/ml, also displayed neutralization at a concentration of 1 ng/ml, and neutralization of most of the other IFN-α subtypes (no neutralization of IFN-α4/5/10).We did not assess the neutralization of these IFNs at 100 pg/ml.These findings are consistent with the high degree of similarity between the 12 IFN-α subtypes (Manry et al., 2011) and the presence of a B cell epitope recognized by the auto-Abs in a conserved region of these IFNs (Meyer et al., 2016).It also suggests that patients with auto-Abs neutralizing all IFN-α subtypes might be at higher risk of severe viral disease.
Here, we considered four human type I IFNs (IFN-α2a/b, IFN-α14, IFN-ω, and IFN-β) normally produced and secreted as glycosylated proteins.IFN-α2b is produced as an O-glycosylated form, whereas IFN-α2a is present in two forms, one fully and the other partially O-glycosylated.By contrast, IFN-α14, IFN-ω, and IFN-β are produced as N-glycosylated forms.We, therefore, investigated the effects of glycosylation on the recognition of these proteins by auto-Abs by determining whether auto-Abs recognized glycosylated but not unglycosylated IFNs or vice versa.We, therefore, tested the neutralization of glycosylated forms of IFN-α2b and IFN-ω produced in mammalian cells.We first determined the optimal experimental set-up.We found that the optimal concentration for testing was 1 ng/ml for glycosylated type I IFN (Fig. S3 A).We tested 183 children, including 19 with auto-Abs against type I IFNs.Most of the 19 patients with auto-Abs neutralizing unglycosylated IFN-α2 or IFN-ω also displayed neutralization of the glycosylated forms.However, three patients had auto-Abs that neutralized the unglycosylated but not the glycosylated form of IFN-ω, and three had auto-Abs neutralizing the unglycosylated but not the glycosylated form of IFN-α2, at a concentration of 1 ng/ml (Fig. 2,  B-E).Interestingly, we also found two patients with auto-Abs neutralizing the glycosylated forms of both IFN-ω and IFN-α2b but not the unglycosylated form of either cytokine, and one patient with auto-Abs neutralizing the glycosylated form of IFN-α2b but not the unglycosylated form.Another two patients had auto-Abs neutralizing the glycosylated form of IFN-ω but not the unglycosylated form.We did not test the unglycosylated form of IFN-β.Serum from the six patients with auto-Abs neutralizing 10 ng/ml unglycosylated IFN-α2 also neutralized all 12 IFN-α subtypes and glycosylated IFN-α2 at a concentration of 1 ng/ml.Interestingly, 2 of the 183 patients tested had auto-Abs neutralizing the glycosylated but not the unglycosylated form of IFN-α2b or IFN-ω, whereas 3 had auto-Abs neutralizing the unglycosylated forms of both cytokines but not the glycosylated forms.These findings suggest that it may be useful to assess the neutralization of glycosylated forms of IFN-α2a and IFN-ω as a means of identifying previously unrecognized patients.
Auto-Abs neutralizing type I IFNs are rare in children from the general population We previously tested large adult cohorts, comprising a total of 39,198 individuals, to assess the prevalence of auto-Abs against type I IFNs in the uninfected general population (Bastard et al., 2020(Bastard et al., , 2021a;;van der Wijst et al., 2021).The prevalence of auto-Abs neutralizing 10 ng/ml (or 100 pg/ml) IFN-α2 or IFN-ω was found to increase significantly with age, with the detection of such Abs in 0.17% (1.1%) of individuals under and >1.4% (4.4%) of those over the age of 70 years, making a major contribution to the higher risk of life-threatening COVID-19 in the elderly population (Manry et al., 2022).The prevalence of auto-Abs against IFN-β was lower and remained stable across age groups (0.26%) (Bastard et al., 2020).Interestingly, auto-Abs neutralizing IFN-α2 (at the lower concentration of 100 pg/ml), regardless of the presence or absence of auto-Abs neutralizing IFN-ω, were found in 0.3% of individuals under the age of 70 years, whereas those neutralizing 100 pg/ml IFN-ω were found in 0.9% of this population (Table 4).Strikingly, the prevalence of auto-Abs neutralizing IFN-α2 increased eightfold after the age of 65 years, whereas the prevalence of auto-Abs neutralizing IFN-ω increases only 2.5-fold (Fig. 3 C, Fig. 4, A-H; Fig. S3, B-H; Fig. S4, A-H; and Table 4).In men, the increase in the prevalence of auto-Abs against IFN-α2 was even greater, >10-fold after the age of 65 years.We therefore assessed the prevalence of these auto-Abs in 2,267 children from the general population with samples collected before the pandemic and, therefore, before any possibility of infection with SARS-CoV-2.Samples were collected in Belgium (n = 126), Canada (n = 161), Estonia (n = 288), Spain (n = 1,685), and Pakistan (n = 7).The children studied were aged 0-18 years (median and mean ages: 10 and 9 years, respectively), with an equal distribution between the sexes (56% were girls).Interestingly, in children, auto-Abs neutralizing IFN-α2 were exceedingly rare.Indeed, only one child (0.04%) had auto-Abs neutralizing IFN-α2 at 10 ng/ml and only three children (0.1%) had auto-Abs neutralizing this cytokine at a concentration of 100 pg/ml.The auto-Abs of these three children also neutralized IFN-ω.By contrast, auto-Abs neutralizing IFN-ω alone were found in a much higher proportion of uninfected children.Indeed, eight children (0.35%) had auto-Abs  neutralizing IFN-ω at 10 ng/ml, whereas an additional 38 (1.7%) had auto-Abs neutralizing IFN-ω at 100 pg/ml (Fig. 3, A and B).
We also identified one girl, aged 1.5 years, with auto-Abs neutralizing only glycosylated IFN-β at a concentration of 10 ng/ml (0.04%) (Fig. 3 A).Finally, we tested two additional independent cohorts of healthy children.None of the individuals of a cohort of 249 healthy children aged 0-18 years (median: 9 years) from Japan tested positive (Fig. S5 A).These pediatric controls included 34 individuals who had had mild COVID infection and did not harbor auto-Abs against type I IFNs.A cohort of 200 healthy children from Estonia (all aged 8-9 years) included only three individuals with auto-Abs neutralizing IFN-ω at a concentration of 100 pg/ml (1.5%) (Fig. S5 B).Overall, 0.17% of uninfected children from our cohort had auto-Abs neutralizing IFN-α2 (4 of 2,267, including 3 at low and 1 at high concentration), whereas 0.04% had auto-Abs neutralizing glycosylated IFN-β (1 of 2,267, at 10 ng/ml) and 2% had auto-Abs neutralizing IFN-ω only (46 of 2,267, including 38 at low and 8 at high concentration).The neutralization of two IFNs simultaneously was exceedingly rare and restricted to IFN-α2 and IFN-ω at the lower concentration in three patients (0.1%).

Characteristics of children from the general population with auto-Abs against type I IFNs
We found that 2,267 children from the general population tested included 4 with auto-Abs neutralizing IFN-α2 (0.2%) and 45 with auto-Abs neutralizing IFN-ω (2%).Three (6%) of these children had auto-Abs neutralizing both IFN-α2 and IFN-ω.
These three children were aged 8, 11, and 13 years, and all three were boys.The individual with auto-Abs neutralizing IFN-α2 only was a 9-year-old girl.Finally, the median age of the 42 children with auto-Abs neutralizing IFN-ω only was 8 years, and 31 (67%) of these children were boys.We also tested a cohort of 145 samples from children hospitalized for bacterial infections.Only one of these patients (0.7%) harbored auto-Abs against type I IFNs (against IFN-ω only) with neutralizing activity against a concentration of 100 pg/ml.None of these children had auto-Abs neutralizing IFN-α2 or IFN-β.None of the children with auto-Abs tested had any remarkable medical antecedents, despite having been infected with many viruses, as shown by Virscan analyses on nine of the positive children (Fig. 3 D).However, it should be noted that infections with influenza viruses or common coronaviruses were not investigated with Virscan.These findings probably attest to the higher tonic type I IFN activity in children than in adults (Loske et al., 2021;Pierangeli et al., 2022;Pierce et al., 2021).Overall, we found that auto-Abs neutralizing IFN-α2 were very rare in children from the general population.By contrast, auto-Abs neutralizing IFN-ω only, at the lower concentration, were less rare (2%) and mostly found in boys, at a rate slightly higher than that for young adults under the age of 40 years (45/2,267 [2%] in children versus 17/1,251 [1.4%] in adults between 18 and 40 years old, P = 0.28).

Risk of life-threatening COVID-19 in children with auto-Abs against type I IFNs
We then assessed the risk of COVID-19 pneumonia (hospitalization for hypoxemic pneumonia, including severe or critical pneumonia) in children carrying auto-Abs capable of neutralizing different concentrations and combinations of type I IFNs, relative to uninfected children from the general population, as previously reported for COVID-19 and influenza in adults (Bastard et al., 2021a;Manry et al., 2022).All types of auto-Ab combinations were highly significant risk factors when patients with severe or critical COVID-19 pneumonia were compared with the general population (Fig. 5 and Table 2).The strongest association with severe or critical pneumonia was that for children with auto-Abs neutralizing both IFN-α2 and IFN-ω at a concentration of 10 ng/ml (OR [95% CI] = 122.8 [12.8-16,364.8],P = 6 × 10 −6 ; OR, odds ratio; CI, confidence interval).Auto-Abs neutralizing IFN-α2 and IFN-ω at a lower concentration of 100 pg/ml were also highly significant risk factors (OR [95% CI] = 27.9 [8.2-116.5],P = 4 × 10 −7 ), whereas auto-Abs neutralizing IFN-α2 or IFN-ω were weaker risk factors (OR [95% CI] = 5.5 [3.1-9.6] at 100 pg/ml and OR [95% CI] = 12.9 [4.6-35.9]at 10 ng/ ml), with the OR for auto-Abs neutralizing high concentrations of IFN-α2 or IFN-ω significantly higher than that for auto-Abs neutralizing low concentrations of IFN-α2 or IFN-ω (P = 0.006).2).Overall, these findings indicate that auto-Abs against type I IFNs can be found in previously healthy children in whom they are a major risk factor for hypoxemic COVID-19 pneumonia, particularly if they neutralize IFN-α2.Indeed, auto-Abs against IFN-α confer the highest risk of life-threatening COVID-19.Interestingly, the ORs for critical COVID-19 were also higher for adults with auto-Abs neutralizing IFN-α than for those neutralizing IFN-ω, in comparison to adults with critical COVID-19 and asymptomatic infected adult controls (P = 0.008; Table 5).Moreover, the risk  2 for children, Table 5 for adults).

Discussion
We report that auto-Abs neutralizing unglycosylated and glycosylated type I IFN-α2 or IFN-β are exceedingly rare (0.17% and 0.04%) in children from the general population, whereas about 0.35% of these children harbor auto-Abs neutralizing IFN-ω at a concentration of 10 ng/ml, and up to 2% of them harbor auto-Abs neutralizing this cytokine at 100 pg/ml.This relatively high proportion of individuals with auto-Abs neutralizing IFN-ω is similar to that found in young adults under the age of 40 years (Bastard et al., 2021a;Manry et al., 2022;van der Wijst et al., 2021).Nevertheless, all these proportions are significantly different from those for adults under the age of 70 years, in whom auto-Abs neutralizing IFN-α2 were found in only 0.3%, auto-Abs neutralizing IFN-ω were found in about 0.9%, and auto-Abs neutralizing IFN-β were found in about 0.3% (Bastard et al., 2020(Bastard et al., , 2021a;;Manry et al., 2022;van der Wijst et al., 2021).Auto-Abs against glycosylated IFN-α2 and IFN-ω were not investigated in the studies on adults (Bastard et al., 2020(Bastard et al., , 2021a;;Manry et al., 2022;van der Wijst et al., 2021), but the similar frequencies of auto-Abs against glycosylated and unglycosylated forms suggest that this is not a major bias.The prevalence of auto-Abs against the three type I IFNs therefore appears to remain stable over time, with two exceptions of different magnitudes.First, the levels of auto-Abs against IFN-ω decrease slightly during middle age.Second, the levels of auto-Abs against IFN-α2, and to a lesser extent IFN-ω (but not IFN-β), suddenly begin to increase after the age of 65 years (Bastard et al., 2020(Bastard et al., , 2021a;;Manry et al., 2022;van der Wijst et al., 2021).The prevalence of auto-Abs neutralizing IFN-α2 increased eightfold after the age of 65 years (the increase was even as great as 10fold in men), a much greater increase than was observed for the prevalence of auto-Abs neutralizing IFN-ω only (which increased only 2.5-fold).Overall, auto-Abs neutralizing 1 ng/ml glycosylated IFN-β are exceedingly rare in children (0.04%), rare in adults below 65 years of age (0.3%), and not more common in the elderly (0.18%).Auto-Abs neutralizing at least 100 pg/ml glycosylated IFN-α2 are very rare in children (0.17%), rare in adults (0.3%), and much more common in the elderly (2.6%).Finally, auto-Abs neutralizing at least 100 pg/ml IFN-ω are less rare in children (2%) than in adults (0.9%) and display a lesser increase in prevalence in the elderly (2%).
The levels of auto-Ab against type I IFNs do not appear to be markedly lower in the youngest children (those under the age of 5 years).This suggests that the production of these pathogenic auto-Abs, especially those against IFN-α2 or IFN-β, in children and young adults may have a germline genetic etiology.In support of this hypothesis, several inborn errors are known to underlie the occurrence of these auto-Abs.Indeed, (i) most, if not all patients with APS-1 and biallelic deleterious variants of AIRE produce such auto-Abs from early childhood onward, as do some patients with dominant-negative variants of AIRE (Ahonen et al., 1990;Bastard et al., 2021c;Meyer et al., 2016;Oftedal et al., 2015Oftedal et al., , 2023;;Ossart et al., 2018); (ii) about a third of women suffering from incontinentia pigmenti due to heterozygosity for loss-of-function mutations of IKBKG harbor these auto-Abs (Bastard et al., 2020;Rosain, 2023); (iii) most patients heterozygous for NFKB2 variants that are gain-of-function for IκBδ activity and loss-of-function for p52 activity, and patients with recessive deficiencies of NIK or RELB have such auto-Abs (Le Voyer, 2023); and (iv) patients with autosomal dominant IKZF2 (Helios) deficiency (Hetemäki et al., 2021a), biallelic RAG1 or RAG2 hypomorphic variants (Chen et al., 2014;Walter et al., 2015) or FOXP3 deficiency (Rosenberg et al., 2018) also carry these auto-Abs.New inborn errors underlying the production of these auto-Abs are expected to be discovered in the future.Other germline genetic etiologies may underlie the production of auto-Abs against IFN-α or IFN-ω, particularly in children and young adults.It is also surprising that the prevalence of anti-IFN-ω auto-Abs seems to decrease slightly with age, before increasing again, together with the prevalence of anti-IFN-α auto-Abs, after the age of 65 years.Auto-Abs arising after the age of 65 years are less likely to be caused by germline variants; their production may be due to somatic variants, epigenetic changes in hematopoietic or non-hematopoietic cell lineages, or thymic lesions, such as thymomas (Cheng et al., 2010;Meager et al., 2003;Rapnouil et al., 2023, Preprint;Shiono et al., 2003).The Table 5. Increase in the risk of critical COVID-19 in adult patients with auto-Abs against IFN-α (neutralizing 10 ng/ml) versus IFN-ω (neutralizing at 10 ng/ml) (from Bastard et al., 2021a) Type I IFN auto-Abs in adults (and amount of type I IFN neutralized in plasma diluted 1/10)

OR [95% CI] P value
Anti-IFN-α2 and anti-IFN-ω auto-Abs (10 ng/ml) 67.6In comparisons of adult patients with critical COVID-19 and asymptomatic infected adult controls, ORs are higher in individuals with auto-Abs against IFN-α than in those with auto-Abs against IFN-ω only, particularly for Abs neutralizing IFN-α at a concentration of 10 ng/ml.This increase in risk for patients with auto-Abs against IFN-α is highly significant.differences in prevalence with age also suggest that the pathogenesis of auto-Abs against type I IFNs may differ, in each age group, between auto-Abs neutralizing IFN-α2, IFN-β, and IFN-ω.We also report that at least 10% of the children hospitalized for COVID-19 pneumonia studied had neutralizing auto-Abs against type I IFNs, as reported by other groups in smaller cohorts (Abolhassani et al., 2022).We also show that these auto-Abs neutralized normally glycosylated IFN-α2a/b, IFN-α14, IFNω, and IFN-β.Only glycosylated IFN-β was tested in our previous studies (Bastard et al., 2020(Bastard et al., , 2021a(Bastard et al., , 2021b(Bastard et al., , 2021c(Bastard et al., , 2022)).The risk of life-threatening COVID-19 pneumonia in children with auto-Abs neutralizing type I IFNs is very high, as previously reported for adults (Barcenas-Morales et al., 2016;Bastard et al., 2021a;Puel et al., 2022).The very high risk of life-threatening COVID-19 pneumonia in children harboring auto-Abs against type I IFNs is consistent with that in children with recessive IEI affecting the type I IFN pathway (Zhang et al., 2022b).In the light of our screening of uninfected children, for the combinations tested, auto-Abs against IFN-α conferred a significantly higher risk of life-threatening COVID-19 than auto-Abs against IFN-ω, regardless of the concentration of cytokine neutralized.A similar pattern has also been observed in adults.The risk of critical COVID-19 in adults with auto-Abs neutralizing IFN-α (regardless of their ability to neutralize IFN-ω) is much higher than that in adults with auto-Abs neutralizing IFN-ω only (P = 0.008 at 100 pg/ml and P = 0.0006 at 10 ng/ml; Table 5).Furthermore, in adults with critical COVID-19 pneumonia, the prevalence of auto-Abs neutralizing IFN-α2 at 10 ng/ml doubles after the age of 60 years (5.6% before 60 years versus 11.2% after 60 years).
As previously observed in adults (Manry et al., 2022), the risk of life-threatening COVID-19 is also higher for children carrying auto-Abs neutralizing high concentrations of IFN-α2 and/or IFN-ω than for children carrying only auto-Abs neutralizing low concentrations, further suggesting that auto-Abs neutralizing high concentrations of IFN-α2 and/or IFN-ω have a more deleterious impact on COVID-19 outcomes.The risk of lifethreatening COVID-19 is even higher in patients with auto-Abs neutralizing both IFN-αs and IFN-ω, further suggesting that the IFN-α subtypes and IFN-ω may not be completely redundant in the context of COVID-19.In addition, the risk of hypoxemic pneumonia increased with the concentration of type I IFNs neutralized.The risk of other viral diseases is unclear, although severe influenza has been reported in several children with auto-Abs against type I IFNs (Walter et al., 2015;Zhang et al., 2022c).Children may have higher tonic or virus-induced type I IFN levels than adults in the tissues in which these molecules are most needed as an initial barrier, such as the naso-epithelial barrier for COVID-19 (Alfi et al., 2021;Beer et al., 2022;Hatton et al., 2021;Lopez et al., 2021;Ziegler et al., 2021).
Despite our discovery of recessive IEI of type I IFN immunity in about 10% of the children studied and of auto-Abs against type I IFNs in another 10%, the cause of severe COVID-19 pneumonia remains unexplained in most children.Other auto-Abs (against type III IFNs, for example) (Vanker et al., 2023), or other IEI, possibly, but not necessarily affecting type I IFNs, might explain these remaining cases.However, our findings already have broad clinical implications.Children hospitalized for COVID-19 pneumonia should be tested for auto-Abs against type I IFNs as targeted therapies can be proposed.When effective against the circulating strains, mAbs neutralizing the virus can be effective if administered promptly (Gupta et al., 2021), as recently shown for an IRF9-deficient child during the first wave of the epidemic (Levy et al., 2021) and other patients with IEI (Johnson et al., 2023).Antiviral compounds, such as remdesivir (Beigel et al., 2020;Gottlieb et al., 2021), molnupiravir (Jayk Bernal et al., 2021), or nirmatrelvir plus ritonavir (Hammond et al., 2022), may also be of benefit in these patients, provided that they are administered sufficiently early in the course of infection.Likewise, early recombinant IFN-β therapy may be considered to prevent the development of hypoxemic pneumonia in patients whose auto-Abs do not neutralize IFN-β (Monk et al., 2021;Vinh et al., 2021).Nasal IFN-α2b could also be considered in patients without auto-Abs or IEI affecting the response to type I IFNs (Zhou et al., 2023).Treatment with type III IFNs is another possibility (Sokal et al., 2023).In the most severe cases, a combination of these therapies with plasmapheresis may be proposed (Bastard et al., 2021c).
Children with auto-Abs against type I IFNs should be followed prospectively.In the general population, it is not entirely clear which group of children should be screened because of the high risk of such auto-Abs.Children with IEI should certainly be screened, particularly those with known genetic etiologies of auto-Abs against type I IFNs (Ahonen et al., 1990;Chen et al., 2014;Oftedal et al., 2015Oftedal et al., , 2023;;Walter et al., 2015;Meyer et al., 2016;Ossart et al., 2018;Bastard et al., 2021c;Hetemäki et al., 2021a;Le Voyer, 2023;Rosain, 2023).Children with a history of unusually severe viral infection should also be tested, as the clinical phenotype of anti-type I IFN auto-Ab production is expanding to include other severe viral diseases (Bastard et al., 2021b(Bastard et al., , 2022;;Gervais et al., 2023;Zhang et al., 2022c).Children with auto-Abs neutralizing type I IFNs should be vaccinated against SARS-CoV-2 and influenza, but not with live-attenuated vaccines (Bastard et al., 2021b).Finally, it would be of interest to conduct pilot studies of the screening of selected populations, such as children with autoimmune conditions (e.g., lupus erythematosus, which is associated with these auto-Abs in adults) (Gupta et al., 2016;Kisand et al., 2010;Londe et al., 2023;Mathian et al., 2022;Panem et al., 1982).Many questions remain unanswered.Severe viral infections might occur at higher frequency in individuals with these auto-Abs.By inference from the known risk of severe adverse reaction to the live-attenuated virus vaccine against yellow fever in adults with auto-Abs against type I IFNs (Bastard et al., 2021b), children with these auto-Abs should not receive this vaccine.Surprisingly, MMR vaccination seems to be well tolerated in APS-1 patients despite the presence of high levels of auto-Abs against type I IFNs.Other live attenuated vaccines (against varicella-zoster virus or monkey pox, for example) should probably be avoided due to the unknown risk of adverse reaction.By contrast, children with these auto-Abs would benefit from RNA vaccination against SARS-CoV-2 and boosters as they are able to mount an Ab response capable of neutralizing the virus (Bastard et al., 2022;Sokal et al., 2023;Wolff et al., 2023).Finally, the follow-up of these children will also be of interest as the changes in the levels of these auto-Abs over time and their association with other viral, tumoral, and autoimmune diseases remain unclear.Lifelong follow-up of this cohort should provide answers to these questions and help to improve the clinical management of these children.

Study design
We enrolled 183 patients with proven COVID-19 pneumonia from nine countries (Brazil, France, Italy, Morocco, Saudi Arabia, Spain, Peru, Turkey, and Ukraine) in this study.We collected plasma or serum samples for all these individuals for immunoassay testing for the presence of auto

COVID-19 classification
The severity of COVID-19 was assessed for each patient, as follows (Bastard et al., 2020;Zhang et al., 2020b): "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 ICU."Severe COVID-19" was defined as pneumonia developing in patients requiring low-flow oxygen (<6 L/min) supplementation.

Statistics
OR and P values for the effect of auto-Abs neutralizing each type I IFN on critical or severe COVID-19 using patients with asymptomatic/mild disease or the general population as controls, adjusted for age in three classes (≤5 years old, (5-10 years old], and (10-18 years old]) and sex, were estimated by means of Firth's bias-corrected logistic regression (Firth, 1993;Heinze and Schemper, 2002) as implemented in the "logistf" R package (https://rdrr.io/cran/logistf/).The risks of critical or severe COVID-19 for carriers of different combinations of neutralizing auto-Abs were compared by Firth's logistic regression adjusted for age in three classes and sex, as described above, in the subsample of individuals (cases and individuals from the general population) carrying the auto-Ab combinations compared.The standard error of the mean for the prevalence of neutralizing auto-Abs against each type I IFN by age group and sex was estimated with the Agresti-Coull approximation (Agresti and Coull, 1998).Phage immunoprecipitation sequencing (PhIP-Seq) The reactivity of circulating Abs against common pathogens in plasma samples from patients and healthy controls was analyzed by PhIP-Seq, as previously described (Hasan et al., 2021).Pooled human plasma for IVIg (Privigen CSL Behring AG), human IgGdepleted serum (supplier no.HPLASERGFA5ML; Molecular Innovations, Inc.), and plasma samples from unrelated healthy children were included as controls.PhIP-Seq was carried out as previously described but with the following modifications.Total IgG levels were determined with the Human IgG total ELISA Ready-SET-Go kit (Thermo Fisher Scientific) and diluted samples containing 4 mg total IgG were incubated at 4°C overnight with 2 × 10 10 plaque-forming units of a modified version of the original VirScan phage library.Specifically, the T7 phage library used here for peptide display contained the same viral peptides as the original VirScan phage library plus additional peptides derived from protein sequences of various microbial B cell antigens available from the Immune Epitope Database (https:// www.iedb.org).For the computational analysis and background correction, the phage library was sequenced before (input library sample) and after immunoprecipitation with beads alone (mock IP).Single-end sequencing was performed with the NextSeq500 system (Illumina) to generate ∼2 million reads per sample and ∼20 million reads for the input library samples.Reads were mapped onto the original library sequences with Bowtie 2 and read counts were adjusted according to library size.A zero-inflated generalized Poisson model was used to estimate the P values to reflect enrichment for each of the peptides.We considered peptides to be significantly enriched only if the −log 10 P value was at least 2.3 in all replicates.Species-specific score values were computed for each serum or plasma sample by counting the significantly enriched peptides for a given species with a continuous subsequence of no more than seven residues, the estimated size of a linear epitope, in common.We corrected for the nonspecific binding of peptides to the capture matrix by also calculating species-specific background score values by counting the peptides displaying enrichment to the 90th percentile for the mock IP samples.These peptides were used for background subtraction.

Figure 2 .
Figure 2. Auto-Abs neutralize all 12 IFN-α subtypes and the glycosylated IFNs.(A) Results for the neutralization of 12 individual IFN-α subtypes, IFN-ω or IFN-β, from children with auto-Abs (n = 8), healthy controls (n = 2), or auto-Ab positive patients (n = 2).Relative luciferase activity is shown (ISRE dual luciferase activity, with normalization against Renilla luciferase activity, and to the non-stimulated [NS] condition) after stimulation with the various type I IFNs at a concentration of 1 ng/ml in the presence of plasma (1/10 dilution).All samples were tested once.RLA: relative luciferase activity.(B-E) Plots representing the

Table 4 .Figure 3 .
Figure 3. Neutralizing auto-Abs against IFN-α2 and/or IFN-ω in children from the general population and their serological evaluation.(A) Results for the neutralization of 10 ng/ml IFN-α2, IFN-ω, or IFN-β in the presence of plasma (1/10 dilution) from children from the general population (n = 2,267).Relative luciferase activity is shown (ISRE dual luciferase activity, with normalization against Renilla luciferase activity).All samples were tested once.RLA: relative luciferase activity.(B) Neutralization of 100 pg/ml IFN-α2 or IFN-ω in the presence of plasma (1/10 dilution) from children from the general population (n = 2,267).(C) Prevalence of auto-Abs neutralizing type I IFNs, distributed by age, in individuals from the general population.(D) Serological evaluation of auto-Abpositive children: Virscan results for children (n = 9) with auto-Abs against type I IFNs.

Figure 5 .
Figure 5. OR for COVID-19 pneumonia.(A) Bar plot of the calculated OR for COVID-19 pneumonia in children with auto-Abs against type I IFNs.Adjusted ORs and 95% CI were computed by penalized profile likelihood.ORs and P values were estimated by means of Firth's bias-corrected logistic regression.(B) Bar plot of the calculated OR for life-threatening pneumonia in children with auto-Abs against type I IFNs.Adjusted ORs and 95% CI were computed by penalized profile likelihood.ORs and P values were estimated by means of Firth's bias-corrected logistic regression.
Fig. S1 describes neutralizing auto-Abs against type I IFNs in children with life-threatening COVID-19.Fig. S2 describes neutralizing auto-Abs against IFN-α2 and/or IFN-ω in children and adults with life-threatening COVID-19.Fig. S3 describes neutralizing auto-Abs against glycosylated type I IFNs, and proportion of children and adults from the general population with neutralizing auto-Abs against type I IFNs.Fig. S4 describes neutralizing auto-Abs against IFN-α2 or IFN-ω in children from the general population.Fig. S5 describes neutralizing auto-Abs against IFN-α2 and/or IFN-ω in children from the general population in Estonia and Japan.

Table 1 .
Auto-Abs neutralized by the serum from the 19 patients

Table 2 .
Clinical and demographic information for the 19 pediatric patients with COVID-19 disease and auto-Abs neutralizing type I IFNs F: female; M: male.Bastard et al.Journal of Experimental Medicine 6 of 21 Auto-Abs neutralizing IFN-α and IFN-ω in children with COVID-19 https://doi.org/10.1084/jem.20231353

Table 3 .
Numbers of cases, proportion, and OR for COVID-19 pneumonia in pediatric patients Bastard et al.Journal of Experimental Medicine 7 of 21 Auto-Abs neutralizing IFN-α and IFN-ω in children with COVID-19 https://doi.org/10.1084/jem.20231353 Estonia and Japan, and the cohorts of patients with bacterial infections were established independently in Spain.All individuals were recruited according to protocols approved by local institutional review boards (IRBs).Written informed consent was obtained in the country of residence of each patient.Experiments were conducted in France and the United States in accordance with local regulations and with the approval of the IRB of the Institut National de la Santé et de la Recherche Médicale and the Rockefeller University, respectively.Approval was obtained from the French Ethics Committee (Comité de Protection des Personnes), the French National Agency for Medicine and Health Product Safety, the Institut National de la Santé et de la Recherche Médicale in Paris, Multilabel Plate Reader (PerkinElmer Life Sciences).Firefly luciferase activity values were normalized against Renilla luciferase activity values.These values were then normalized against the median level of induction for non-neutralizing samples and expressed as a percentage.Samples were considered neutralizing if luciferase induction after normalization against Renilla luciferase activity was below 15% the median value for controls tested the same day.
Bastard et al.Journal of Experimental Medicine 15 of 21 Auto-Abs neutralizing IFN-α and IFN-ω in children with COVID-19 https://doi.org/10.1084/jem.20231353 Disclosures: Q. Philippot reported personal fees from Gilead outside the submitted work.E. Javouhey reported non-financial support from Biomerieux during the conduct of the study.Y. Tandjaoui-Lambiotte reported personal fees from STANE, QUANTIQ, and Digital Medical Hub outside the submitted work.I. Meyts reported grants from CSL-Behring, other from Boehringer-Ingelheim, and other from Takeda outside the submitted work.D.C. Vinh reported personal fees from Moderna, As-traZeneca, Merck Canada, Novartis Canada, Qu Biologics, and Takeda outside the submitted work; in addition, D.C. Vinh had a patent number 40101099 licensed; and is supported by the Fonds de recherche du Québec-Santé clinician-scientist senior scholar award.A.-F. Mentis reported "other" from BGI Genomics, and grants from ELIDEK and the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call Research-Create-Innovate (project code: T2EDK-02222) outside the submitted work.S. Trouillet-Assant reported nonfinancial support and grants from bioMérieux outside the submitted work.No other disclosures were reported.