Early nasal type I IFN immunity against SARS-CoV-2 is compromised in patients with autoantibodies against type I IFNs

Cell-intrinsic immunity starts with microbial detection through innate sensors. In case of infection by RNA viruses such as SARS-CoV-2, the sensing may involve cytosolic retinoic acid–inducible gene I, melanoma differentiation-associated protein 5 (Schultze and Aschenbrenner, 2021), and endosomal TLR3 (Gao et al., 2021). The activation of these receptors triggers a signaling cascade that leads to the induction of type I and III IFNs (IFN-I and IFN-III). During the early phase of this cascade, the transcription of different IFN-I and IFN-III genes and numerous IFN-stimulated genes (ISGs) is induced. Secreted IFN-I binds to IFN receptors that signal through JAK1 and tyrosine kinase 2 and activate STAT1, STAT2, and IFN regulatory factor 9. The last three then form a complex called ISGF-3 that translocates to the nucleus and binds to IFN-stimulated response elements (ISREs), thereby inducing the expression of hundreds of ISGs, most of which have antiviral activity. Secreted IFN-III binds to another receptor complex composed of IFNLR and IL10RB and activates the same ISGF-3 complex, inducing the expression of a large number of overlapping ISGs (Lazear et al., 2019). ISGF-3 may also participate to the induction of ISGs downstream IFN-II (Platanitis et al., 2019) but is not essential for IFN-γ–dependent immunity (Rosain et al., 2019).

IFN-I and IFN-III participate in the first line of defense against infection, as they promote virus clearance, induce tissue repair, and stimulate the adaptive immune response. However, IFN-I activity has been reported at very low or undetectable levels in the blood (Hadjadj et al., 2020;Trouillet-Assant et al., 2020b) and the nasal epithelium (Ziegler et al., 2021) of some critically ill COVID-19 patients. This defect is explained in some patients by inborn genetic defects of IFN-I induction or amplification (Zhang et al., 2020b) and in others by the presence of neutralizing autoantibodies (auto-Abs) against type I IFNs, mainly IFN-α and/or IFN-ω and, in rare cases, IFN-β (Bastard et al., 2020). These auto-Abs underlie life-threatening pneumonia in at least 10% of intensive care unit (ICU)–admitted COVID-19 patients, as reported by multiple independent cohort studies (Zhang et al., 2020b; Bastard et al., 2020; Troya et al., 2021; Koning et al., 2021; van der Wijst et al., 2021 Preprint). While IFN-I immunity is essential for pulmonary immunity to SARS-CoV-2, the contribution of IFN-III immunity remains unknown.

The delayed production of IFN-I is thought to result in unbalanced immune responses unable to clear the virus (Channappanavar et al., 2016; Carvalho et al., 2021;Blanco-Melo et al., 2020). A two-step model of pathogenesis has proposed that insufficient IFN-I levels and hence insufficient ISGs induction during the first 10 d of infection underly viral spread, thereby accounting for the subsequent pulmonary and systemic inflammation that is characteristic of critical cases in which leukocytes attempt to resolve what IFN-I did not in the first place (Zhang et al., 2020a). However, few studies have performed longitudinal measurements of immunity during the acute phases of infection, especially in mildly symptomatic individuals, and thus the dynamics of IFN-I activity relative to virus infection and other immunological perturbations remains largely elusive. Moreover, early host–pathogen interactions occurring at the primary infection site (i.e., the nasal mucosa) are likely to influence in a major way the clinical outcome of infection (Robinot et al., 2021). While many studies have characterized the systemic immune response against SARS-CoV-2, little is known about the immune events occurring in the nasopharyngeal (NP) mucosa. Two recent RNA-sequencing studies have reported ISGs induction in the nasal mucosa of mildly symptomatic COVID-19 patients at diagnosis (Mick et al., 2020; Ng et al., 2021), but whether or how this induction correlates with viral parameters, the dynamics of the peripheral immune response, the delay after symptom onset, and the clinical outcome remains unclear. Moreover, epithelial cells specifically express the IFN-λ (IFN-III) receptor. This receptor signals through a similar JAK-STAT pathway as the IFN-I receptor complex and induces many of the same ISGs that are regulated by the binding of ISGF-3 to ISREs (Schneider et al., 2014), and both IFN-I and IFN-III can restrict SARS-CoV-2 infection of human airway epithelial cultures (Vanderheiden et al., 2020). However, the relative contributions of IFN-I and IFN-III in the defense against SARS-CoV-2 infection in vivo remain unknown. Here, we set out to fill these gaps in knowledge by studying in parallel mucosal and systemic immunity in COVID-19 patients using different RNA-based technologies and comparing patients with and without auto-Abs neutralizing IFN-I.

We aimed to study the dynamics of nasal and blood transcriptome in patients during SARS-CoV-2 infection. For this purpose, we constituted a cohort of health care workers presenting mild COVID-19 symptoms (n = 44 patients; Trouillet-Assant et al., 2020a) and no respiratory coinfection and for whom the delay between symptom onset and SARS-CoV-2 molecular screening was ≤7 d (Table 1). We collected blood samples and nasal swabs weekly for 4 wk, starting at diagnosis (four different visits [V1–V4]; see Fig. S1, which outlines the study); a group of uninfected healthy controls (HCs) was also included (n = 5). We first measured the transcript levels of 800 genes associated with different immune pathways in blood cells using the NanoString technology (“host response” panel). We compared changes in gene expression between mildly symptomatic patients at each time point of the study (V1–V4) and HCs in order to identify differentially expressed genes (DEGs; Table S2). Most changes in gene expression were found at V1 (i.e., at diagnosis of COVID-19;Fig. S2 A). A principal-component analysis confirmed this result and additionally showed that all patients at V3 and V4 and two thirds of patients at V2 grouped together with HCs, while one third of V2 samples clustered between V1 and the other samples (Fig. 1 A). Thus, changes in gene expression in leukocytes mainly occurred before seroconversion, which generally occurs between V1 and V2 (Fig. S2 B). The top functional pathways associated with DEG at V1 according to the nSolver NanoString analysis software were ISGs, including IFN-I– and IFN-III–dependent ISGs (i.e., ISGs driven by ISGF-3 binding to ISRE) and IFN-II–dependent ISGs (i.e., ISGs mainly driven by γ-activated factor binding to gamma IFN activation sites; Fig. 1 B). To monitor the representation of these pathways in a longitudinal manner, we computed an expression score based on the combined expression of different ISGs (see Materials and methods). As shown in Fig. 1, C and D, the type I/III ISGs (ISG-I/III) and ISG-II scores were maximal at V1 and lower for other time points, even though they were still significantly higher at V2 compared with HC samples. Moreover, other pathways were significantly enriched at V1, including DNA/RNA sensing, MHC-I antigen presentation, lysosome, myeloid inflammation, and nucleotide oligomerization domain–like receptor signaling; for all these pathways, the scores were significantly higher at V1 compared with HCs and similar between other time points and HCs (Fig. S3). The ISG-I/III score at V1 was associated with the normalized viral load determined at diagnosis (Fig. 1 E; Spearman ρ = 0.55, P < 0.0001), but not with age, body mass index, or delay after symptoms (Fig. S2 C). Altogether, these data highlight the predominance of ISGs among the genes induced in blood cells in response to SARS-CoV-2 infection, the transient nature of this induction, and its association with viral load.

Next, we aimed to study the regulation and the role of IFN-I/III signaling in the nasal mucosa in response to SARS-CoV-2 infection. For this, we took advantage of the nasal swabs used for standard molecular SARS-CoV-2 screening. However, measuring the host response in such samples is technically challenging, as the amount of cellular material is extremely limited. To overcome this problem, we used the FilmArray technology, a nested PCR-based technique allowing rapid and sensitive measurement of ISGs transcripts that has been recently described (Mommert et al., 2021) and that we adapted for nasal swabs. This technique is advantageous, as it is semiautomated and also designed for clinical routine, providing an IFN-I/III score based on the quantification of four ISGs regulated through ISGF3-dependent ISREs (i.e.,IFI27, IFI44L, RSAD2, andIFIT1; Hernandez et al., 2018). We monitored the nasal IFN-I/III score at diagnosis in 23 SARS-CoV-2–infected health care workers and compared this score with the blood IFN-I/III score and the serum level of IFN-α₂ from the same patients. There was a strong correlation between nasal and blood IFN-I/III scores (Fig. 2 A) and between the nasal IFN-I/III score and serum IFN-α₂ levels (Fig. 2 B; Spearman ρ > 0.75 for both comparisons). We also measured the serum level of IFN-λ₁, even though this cytokine is believed to act in a local manner (Park and Iwasaki, 2020). IFN-λ₁ was detectable in the serum of many patients, and its levels were also correlated with the nasal and blood IFN-I/III scores, but the correlation coefficient for these relations was lower than those of IFN-α₂ and IFN-I/III scores (Fig. S4). We then compared the dynamics of the IFN-I/III score in the blood versus nasal mucosa, taking advantage of the availability of serial and paired blood samples and nasal swabs. Both parameters followed very similar kinetics, peaking at the earliest time points after symptom onset and rapidly decreasing thereafter, becoming virtually negative 14 d after symptom onset (Fig. 2 C). This overlap confirms the coordination of nasal and systemic IFN-I/III antiviral defenses and shows that the nasal IFN-I/III score measured using FilmArray directly from swab samples is a faithful proxy of the blood IFN-I/III score in patients.

Next, we examined the possibility of using the nasal IFN-I/III score as a biomarker of SARS-CoV-2 infection when combined with SARS-CoV-2 RT-PCR. To this aim, we first tested the association between the nasal SARS-CoV-2 viral load and the nasal ISG-I/III score, both determined longitudinally and paired for each time point. There was a significant correlation between both parameters (Spearman ρ = 0.59 [0.42; 0.73], P < 0.0001; Fig. 3 A). Additionally, based on iterative comparisons, we found that a nasal viral load of 4.40 log₁₀ copies [cp]/10⁶ cells discriminated best between patients with high and low IFN-I/III scores; accordingly, most patients with viral loads above 4.40 log₁₀ cp/10⁶ cells had high nasal (Fig. 3 B) and blood (Fig. 3 C) IFN-I/III scores. A few patients had high ISGs scores despite low virus dose and vice versa, which suggests that other parameters than the virus dose, such as genetic variation, could influence ISGs induction upon infection. We then performed virus culture assays directly from nasal swabs in Vero cells as previously described (Bal et al., 2021) in order to evaluate the putative association between the nasal IFN-I/III score and the isolation of infectious virus from the swab. We found that positive virus cultures were usually associated with high median nasal IFN-I/III scores and, reciprocally, that negative cultures were almost always associated with low median IFN-I/III scores (Fig. 3 D). We then evaluated the interest of using the IFN-I/III score as a surrogate biomarker of replicative SARS-CoV-2 infection through a receiver-operating characteristic curve. The area (95% confidence interval) under the receiver-operating characteristic curve was 0.84 (0.7; 0.99), revealing a good performance of the IFN-I/III score to predict the virus replication capacity with a cut-point score at 6.75 (Fig. 3 E). Of note, one patient had a strong nasal IFN-I/III score at V1 (33.2) despite a low viral load <4.4 log₁₀ cp/10⁶ cells (in red in Fig. 3, A and B). For this patient, V1 occurred only 1 d after symptom onset (unlike most of the other patients from the cohort), which suggests that the patient was probably in the ascending phase of viral replication at V1. Accordingly, for this patient, the IFN-I/III score remained high at V2 (43.2) and was associated with a high viral load (>4.4 log₁₀ cp/10⁶ cells); of note, the viral culture was positive for both V1 and V2 time points. Lastly, to further evaluate the relevance of using the IFN-I/III score as a marker of replicative infection, we performed comparative measurements of the IFN-I/III score and of the viral RNA in conditions of limiting material mimicked by 10-fold serial dilutions of positive swabs. The Ct (cycle threshold) values of the viral PCR rapidly increased upon swab dilutions, while the IFN-I/III scores remained relatively stable and consistently higher than the 6.75 value previously shown to discriminate between swabs containing or not replicative virus (Fig. 3 F). Altogether, our data highlight the relevance of using the nasal IFN-I/III score as a biomarker of active SARS-CoV-2 infection in combination with SARS-CoV-2 RT-PCR and as a tool to probe the mucosal pathogenesis of COVID-19 at the molecular and cellular levels.

A fraction of critically ill COVID-19 patients have low or undetectable blood IFN-I levels (Hadjadj et al., 2020;Trouillet-Assant et al., 2020b), which can be due to either inborn errors of IFN-I immunity (Zhang et al., 2020b), the presence of circulating auto-Abs that neutralize IFN-α and IFN-ω, or both (Bastard et al., 2020). These auto-Abs could also theoretically block mucosal IFN-I, but antibodies do not diffuse as well in tissues as they do in the blood (Iwasaki, 2017), and in some situations, they can be sequestrated within the vascular compartment (Chen et al., 2018). We thus tested the relationships among disease severity, the presence of IFN-I auto-Abs, viral load, and nasal IFN-I/III score. We obtained blood and nasal swabs at ICU admission from 26 critically ill patients (Table 1). The presence of neutralizing auto-Abs against IFN-α and IFN-ω in serum was tested using a novel highly sensitive ELISA (Goncalves et al., 2021). We detected IFN-α and/or IFN-ω auto-Abs in the sera of 8 out of 26 critically ill patients and none of the mildly symptomatic patients (Fig. 4). While most patients without IFN-I auto-Abs and a viral load >4.4 log₁₀ cp/10⁶ cells had a very high ISGs score (Fig. 4 A), patients with IFN-I auto-Abs had very low nasal IFN-I/III scores, irrespective of viral load, except for three patients who displayed auto-Abs against IFN-α₂ only and for whom we assumed that IFN-ω could compensate the blockade of IFN-α₂. The other five patients had auto-Abs against both IFN-α and IFN-ω (Fig. 4 B) These data suggest that IFN-I auto-Abs could block IFN-I in the nasal mucosa as well, which was tested by assessing the presence of neutralizing IFN-I auto-Abs in nasal swabs in patients for whom serum auto-Abs levels were also determined. Interestingly, nasal IFN-I auto-Abs were detected in half of the patients who had serum auto-Abs (Table S3). This is consistent with their previous detection in a bronchoalveolar lavage (de Prost et al., 2021). The serum level of IFN-I auto-Abs was not correlated to that of the swab, but differences in the quality of the swab may explain this discrepancy.

Thus, our data suggest that (1) IFN-I is the main inducer of strong ISGs responses in the nasal mucosa, while IFN-III seems insufficient; (2) IFN-I auto-Abs compromise early antiviral defenses in the nasal mucosa of critically ill patients, especially when auto-Abs against both IFN-α₂ and IFN-ω are present; and (3) IFN-ω is involved in nasal IFN-I immunity.

To assess the impact of IFN-I auto-Abs in the early defense against SARS-CoV-2 infection in a more direct manner, we used a previously validated (Pizzorno et al., 2020) and predictive reconstituted HAE model of SARS-CoV-2 infection. In this model, HAEs were treated (24 h before and 1 h after infection) with recombinant IFN-α₂ in the presence or absence of prepandemic serum from HCs or patients with biallelic loss-of-function variants of AIRE (autoimmune regulator locus). The latter suffer from autoimmune polyendocrine syndrome type 1 (APS-1) and produce a broad range of auto-Abs, including neutralizing IFN-I auto-Abs (Bruserud et al., 2016; Levin, 2006; Meager et al., 2006). This range is very similar to what is found in a subset of patients with life-threatening COVID-19 (Bastard et al., 2020), which allows testing the impact of IFN-I auto-Abs without the confounding effects of anti–SARS-CoV-2 neutralizing Abs. While recombinant IFN-α₂ significantly inhibited viral replication in the presence of serum from individuals without auto-Abs, this antiviral effect was completely abrogated in the presence of neutralizing auto-Abs from APS-1 individuals, as the viral titers were comparable to those of the untreated control (Fig. 5, A and B). These results were further confirmed by the reduced transepithelial electric resistance values (Fig. 5 C), which are used as reliable surrogate of epithelium integrity (Pizzorno et al., 2020). By contrast, the APS-1 serum had no effect on viral replication when applied in the absence of IFN-I. The presence of APS-1 serum was also associated with lower IFN-α₂–induced IFN-induced protein 44–like (IFI44L) expression (Fig. 5 D). Moreover, the use of a serum from a non–APS-1 individual with IFN-I auto-Abs not previously infected by SARS-CoV-2 showed the same results as those obtained with the APS-1 serum (Fig. S5), excluding the possibility that other auto-Abs not directed against IFN-I in the APS-1 serum may also influence results in the HAE model.

We then directly compared the antiviral effects and ISGs induction capacity of exogenous IFN-α₂, IFN-β₁, IFN-λ₁, IFN-λ₂, IFN-λ₃, and IFN-ω and the impact of IFN-I auto-Abs on these effects in the HAE model. A high concentration of all recombinant IFNs inhibited viral replication (Fig. 6, A–C) and induced IFI44L expression (Fig. 6 D); however, neutralizing auto-Abs abrogated only the effect of IFN-α₂ and IFN-ω and not that of IFN-β₁, IFN-λ₁, IFN-λ₂, or IFN-λ₃ (Fig. 6, A–D). Altogether, these results suggest that in the nasal mucosa of infected patients, IFN-α₂ and IFN-ω have a prominent role in the antiviral response that cannot be compensated by other IFNs, probably not expressed at sufficient levels. Moreover, our results functionally validate the neutralizing activity of IFN-I auto-Abs, which further suggests that those Abs could compromise the nasal IFN-I response in a subset of critically ill patients.

We investigated herein anti–SARS-CoV-2 IFN-I immunity in a longitudinal way in patients with mild COVID-19 manifestations. This immunity is not well characterized, as most longitudinal studies have focused on hospitalized patients (Lucas et al., 2020; Liu et al., 2021; Szabo et al., 2021; Bernardes et al., 2020). IFN-I/III ISGs and IFN-II ISGs were the main groups of genes up-regulated in the blood during infection. This finding confirms the impact of IFNs in the antiviral immune response, which is also supported by previous observations according to which severe COVID-19 patients develop auto-Abs against IFN-I or have inborn defects of IFN-I induction, amplification, or response (Bastard et al., 2020). Interestingly, other patients with these inborn defects in IFN-I induction or with auto-Abs against IFN-I are prone to other viral infections, for which there is incomplete penetrance (Bastard et al., 2021b; 2021c; Duncan et al., 2015; Gothe et al., 2020; Hernandez et al., 2019), while the penetrance of COVID-19 critical forms is probably very high. Our data show that in most patients with mild COVID-19, the up-regulation of ISGs was transient, maximal at diagnosis, and coordinated with antigen presentation, which presumably helps initiating the ensuing adaptive immune response. The ISGs wave indeed preceded seroconversion, which occurred on average 14 d after symptom onset, consistent with previous studies (Premkumar et al., 2020;Rydyznski Moderbacher et al., 2020).

We showed that the mucosal IFN-I/III signature was correlated with the blood IFN-I signature. Two recent studies have also reported a strong induction of ISGs in the nasal mucosa of SARS-CoV-2–infected patients, which was even higher compared with patients infected with other viruses, such as influenza or seasonal coronaviruses (Mick et al., 2020; Ng et al., 2021), confirming that a high ISGs expression is a hallmark of anti–SARS-CoV-2 mucosal immunity (Pizzorno et al., 2020). Interestingly, another study has also reported that the nasal ISGs expression was higher in pediatric than in adult patients infected with SARS-CoV-2 (Pierce et al., 2021), which could explain the differences in disease severity observed between these two populations. However, the pediatric patients included in the latter study were in fact symptomatic and presenting to the emergency department (Pierce et al., 2021), suggesting that their immune response was not comparable to typically asymptomatically infected children.

Using FilmArray technology, we found that the nasal IFN-I/III score was commensurate to the viral load and to virus infectivity, which is interesting, since SARS-CoV-2 RT-PCR results are often difficult to interpret in clinical routine when Ct values are high and in the absence of clinical manifestation (Han et al., 2021). The nasal ISG-I/III score could therefore be used as a biomarker of active SARS-CoV-2 infection in combination with SARS-CoV-2 RT-PCR to rapidly help in the identification of patients at risk of virus transmission and reciprocally help to avoid quarantine measures in patients who do not or no longer represent a possible source of contamination.

We also found that a sizeable fraction of critically ill patients had a low nasal IFN-I/III score despite a high viral load. These data extend previous findings that the blood IFN-I signature or the serum IFN-I concentration was very low in a subset of critically ill patients (Trouillet-Assant et al., 2020b; Bastard et al., 2020). Thus, the measurement of the nasal IFN-I/III score could be used to help stratify patients and identify patients at risk of developing severe symptoms at early stages of infection. Many COVID-19 therapeutic treatments are currently being tested (for a review, see Canedo-Marroquín et al., 2020), and it is likely that their efficacy will depend on their timely administration during the infection (Park and Iwasaki, 2020), which prompts the identification of early markers of disease severity. We also detected IFN-I auto-Abs in both serum and nasal swabs in a fraction of critically ill patients, in particular in patients with a low nasal IFN-I/III score despite a high viral load. These data suggest that IFN-I auto-Abs could explain the low nasal IFN score in a subset of patients and thus that IFNs-I are the main drivers responsible for inducing protective ISGs signatures in the nasal epithelium compared with IFNs-III. The observation that a few patients with IFN-α₂ auto-Abs without IFN-ω auto-Abs had a high nasal IFN-I/III score suggests a redundancy between IFN-α and IFN-ω in this process. The blockade of IFN-I signaling at the site of virus infection could help virus dissemination by preventing the IFN-I system to render epithelial and other ACE2-expressing cells refractory to infection. We demonstrated this experimentally using a reconstructed epithelium that we infected with SARS-CoV-2 and in which we showed that serum auto-Abs could block the effects of both IFN-α and IFN-ω. A recent study has reported the presence of IFN-I auto-Abs in tracheal aspirates as well (de Prost et al., 2021), suggesting that auto-Abs could block IFN-I signaling at multiple sites. We observed that some critically ill patients had a fairly low viral load despite the presence of anti–IFN-I auto-Abs. This could be explained by later sampling for these patients, associated with a spread of the infection to the lower lungs (Wang et al., 2020).

As recently discussed (Stertz and Hale, 2021), many questions remain, such as how auto-Abs that are produced at modest titers in the serum (Goncalves et al., 2021) block IFN-I at mucosal sites where their diffusion is likely rather limited, even though we detected them in the present study. Moreover, auto-Abs are difficult to manage in clinical practice (de Prost et al., 2021; Bastard et al., 2021a), and it is unclear which therapies should be administered to COVID-19 patients upon detection of IFN-I auto-Abs. Our findings suggest that early therapy with IFN-β, which is apparently not neutralized in most patients with auto-Abs against IFN-I and the response to which is not affected in most patients with inborn defects of IFN-I, may prevent the development of critical disease. An alternative is IFN-III, which displayed antiviral activity in our in vitro system regardless of the presence of IFN-I auto-Abs. Other studies have shown its efficacy in animal models (Park and Iwasaki, 2020). Many clinical trials on the prophylactic and therapeutic use of IFNs for COVID-19 care are ongoing (Sallard et al., 2020) and might bring answers to these questions.

HCs were recruited among donors to the Lyon blood transfusion center (Etablissement Français du Sang) and were considered as a control group. According to French procedures, a written nonopposition to the use of donated blood for research purposes was obtained from HCs. The donors’ personal data were anonymized before transfer to our research laboratory. We obtained approval from the local ethical committee and the French ministry of research (DC-2008-64) for handling and conservation of these samples.

A prospective longitudinal cohort study was conducted at the Hospices Civils de Lyon (university hospital of Lyon, France) including health care workers aged >18 yr, having given their written consent, accepting a 6-wk follow-up, and displaying at least one of the following symptoms suggestive of a SARS-CoV-2 infection: fever, respiratory symptoms, headache, anosmia, and ageusia (Trouillet-Assant et al., 2020a). Exclusion criteria were being pregnant or lactating. The COVID-19 diagnosis was confirmed by quantitative RT-PCR (Cobas SARS-CoV-2 Test; Roche Diagnostics). The NP swab used were either in Copan Universal Transport Medium or in Cobas PCR medium tube. Clinical and microbiological data were collected for all included health care workers. Patients with a positive RT-PCR result at inclusion (V1) came back weekly (V2, V3, and V4) for blood and NP sampling until negativation (assessed by RT-PCR). Ethics approval was obtained from the national review board for biomedical research in April 2020 (Comité de Protection des Personnes Sud Méditerranée I, Marseille, France; ID RCB 2020-A00932-37), and the clinical study was registered at www.clinicaltrials.gov (NCT04341142, v3).

All critically ill patients, admitted to the ICU, were included in the MIR-COVID study. This study was registered to the Commission nationale de l’informatique et des libertés (French national data protection commission) under the number 20–097 and was approved by an ethical committee for biomedical research (Comité de Protection des Personnes HCL) under the number 20-41. In agreement with the General Data Protection Regulation (Regulation [EU] 2016/679 and Directive 95/46/EC) and the French data protection law (law 78-17 on 06/01/1978 and Décret 2019-536 on 29/05/2019), we obtained consent from each patient or his/her next of kin. During the ICU stay, blood collected on EDTA tubes were used for anti-IFN antibody investigation as previously described. For this study, nasal swabs used for molecular SARS-CoV-2 screening at hospital admission were collected for additional experiments.

Total RNA was extracted from PAXgene tubes using the Maxwell 16 LEV simplyRNA Blood kit (Promega), following the manufacturer's guidelines. RNA quantity was determined using a Nanodrop (Thermo Fisher Scientific). 200 ng total RNA was hybridized with the nCounter Host Response panel (#LBL-10805-01; NanoString) and counted on an nCounter FLEX platform according to the manufacturer's guidelines. Table S1 provides more information on the panel and the genes measured. Raw counts were normalized using internal positive standards and 12 housekeeping genes. Using ROSALIND NanoString Gene Expression packages (https://rosalind.onramp.bio/), we performed differential expression analysis and gene set analysis with a cutoff at ±1.5-fold change and an adjusted P value < 0.05. Gene expression signatures were generated by summing for each of the 60 signatures encompassed in the panel (LBL-10805-01_nCounter_Host_Reposne_Gene_List annotation file) the normalized counts of each gene divided by the median of expression in the control group. Statistical analysis and graphics were performed using Prism 9 software (https://www.graphpad.com).

The first prototype of the IFN pouch encompasses four ISGs (IFN α inducible protein 27, IFI44L, IFN-induced protein with tetratricopeptide repeats 1, radical S-adenosyl methionine domain containing 2) and three housekeeping genes (hypoxanthine phosphoribosyltransferase 1, peptidylprolyl isomerase B, and 2,4-dienoyl-CoA reductase 1) for signal normalization. 100 µl PAXgene blood or nasal pharyngeal swab samples was tested using the IFN prototype (Tawfik et al., 2020) according to the manufacturer’s instructions. In brief, the pouches were hydrated with the hydration solution supplied with the kit. The PAXgene blood or nasal pharyngeal swab samples were mixed with 800 µl of the sample buffer provided with the kit and directly injected into the pouch and ran on FilmArray 2.0 and FilmArray Torch instruments (BioFire Diagnostics). Results were delivered in <1 h. Using a research version of the instrument, real-time quantification cycle values and post-amplification melt peaks were determined for each assay. The normalized expression values of each assay were then computed using the internal reference genes. The nasal pharyngeal ISGs score was calculated using the same method applied for PAXgene sample as previously described (Pescarmona et al., 2019).

Serum IFN-α₂ concentrations (fg/ml) from plasma were measured by single-molecule array using a commercial kit for IFN-α₂ quantification (Quanterix). The assay was based on a three-step protocol using an HD-1 Analyzer (Quanterix) and required 3 h of processing time from sample to result acquisition (Wilson et al., 2016). Plasma IFN-l concentrations were determined using Meso Scale Discovery U-PLEX platform.

The SARS-CoV-2 load was determined from the nasal pharyngeal swabs using SARS-CoV-2 R-gene kit (bioMérieux). Briefly, nucleic acid extraction was performed from 0.2 ml nasal pharyngeal swabs on NUCLISENS easyMAG, and amplification was performed using Bio-Rad CFX96. The viral load was quantified using four internally developed quantification standards (QS1–QS4 at 2.5 × 10⁶, 2.5 × 10⁵, 2.5 × 10⁴, and 2.5 × 10³ cp/ml, respectively, of SARS-CoV-2 N gene DNA). These quantification standards were quality checked and quantified using the Nanodrop spectrophotometer (Thermo Fisher Scientific) and Applied Biosystems QuantStudio 3D Digital PCR. In parallel, nasal pharyngeal swabs were tested using the CELL Control R-GENE kit (amplification of the HPRT1 housekeeping gene) that contains two quantification standards (QS1 at 10⁴ cp/µl DNA standard that is 50,000 cells/PCR [i.e., 1.25 × 10⁶ cells/ml in our conditions] and QS2 at 10³ cp/µl DNA standard that is 5,000 cells/PCR [i.e., 1.25 × 10⁵ cells/ml in our conditions]) to normalize the viral load according to the sampling quality (log10 [(number of SARS-CoV-2 copies per milliliter/number of cells per milliliter) × 10⁶ cells per milliliter]).

Viral culture was performed following interim biosafety guidelines established by the World Health Organization [WHO/WPE/GIH/2020.3] from nasal pharyngeal swabs in Universal Transport Medium. RT-PCR–positive nasal pharyngeal swabs were inoculated on confluent Vero cells (ATCC CCL-81) in Eagle’s minimum essential media supplemented with 2% penicillin-streptomycin, 1% L-glutamine, and 2% inactivated fetal bovine serum. Plates were incubated at 33°C (which was determined to be the optimal temperature for virus cultures) and 5% CO₂ for 96 h. Cytopathic effects were monitored daily; samples were harvested when positive, while negative samples at 96 h underwent subculture on new plates. Culture supernatants were sampled at 2 h after inoculation, at 96 h, and after an additional 96 h on subculture. RNA from supernatants were extracted by the automated MGISP-960 workstation using MGI Easy Magnetic Beads Virus DNA/RNA Extraction Kit (MGI Tech), and SARS-CoV-2 detection was performed using TaqPath COVID-19 CE-IVD RT-PCR kit on a QuantStudio 5 System (Applied Biosystems and Thermo Fisher Scientific).

The presence of anti–SARS-CoV-2 antibodies was evaluated using the Wantai Ab assay detecting total antibodies against the receptor-binding domain of the S protein.

The presence of anti–IFN-α₂ antibodies was investigated using a commercially available kit (Thermo Fisher Scientific;Goncalves et al., 2021). The neutralizing capacity of patient sera against IFN-α was then evaluated as previously described (Bastard et al., 2021c).

MucilAir nasal HAE reconstituted from human primary cells was provided by Epithelix SARL and maintained in air–liquid interphase at 37°C and 5% CO₂ in specific culture medium in Costar Transwell inserts (Corning) according to the manufacturer’s instructions. 1 d before infection (d−1), HAEs were mock treated or treated through the basolateral pole with a 2 ng/ml or 20 ng/ml final dilution of recombinant IFN-α_2a, IFN-β₁, IFN-λ₁, IFN-λ₂, IFN-λ₃ (PBL Assay Science), or IFN-ω (Merck) in 700 µl MucilAir culture medium. To assess the potential neutralizing effect of anti–IFN-I antibodies, recombinant IFN-α_2a was preincubated (37°C, 1 h) before addition to HAE with a 1% final dilution of inactivated (56°C, 30 min) pre–COVID-19-pandemic patient sera, containing or not anti–IFN-α antibodies collected from autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy patients and HCs, respectively. A third serum, obtained from a non–autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy patient who tested positive for auto-Abs against type I IFNs, was also used as control. 24 h after IFN treatment (d0) in the presence or not of sera, the apical poles of HAE were gently washed twice with warm OptiMEM medium (GIBCO BRL and Thermo Fisher Scientific) and then infected with a 150 µl dilution of SARS-CoV-2 virus in OptiMEM medium at a multiplicity of infection of 0.1, as previously described (Pizzorno et al., 2020). Basolateral treatment with recombinant IFN (with or without sera) was repeated 1 hpi (hour postinfection) in the same conditions as on d−1. Variations in transepithelial electrical resistance (TEER) were measured using a dedicated volt-ohm meter (EVOM2, Epithelial Volt/Ohm Meter) and expressed in Ohm/cm². At 54 hpi, HAE apical poles were washed with warm OptiMEM and collected in two tubes, one for TCID50 viral titration and one for quantification of viral genome by RT-PCR. HAE cells were harvested in RLT buffer (Qiagen), and total RNA was extracted using the RNeasy Mini Kit (Qiagen) for subsequent gene expression analyses.

Nonparametric Wilcoxon–Mann–Whitney tests and Spearman's correlation tests were performed for all comparisons, unless otherwise specified. A χ² test was performed to analyze contingency tables. To calculate the virus load threshold that better divided patients into two ISGs groups, we used the lowest P value of nonparametric significance test (Wilcoxon rank sum) iteratively applied to the ISGs score, split over normalized nasal viral load thresholds ranging from minimum (0.60) to maximum detected values (9.00) with a step of 0.01. All statistical analyses were conducted using R (the R Foundation; https://www.r-project.org/foundation/, version 3.6.1) and GraphPad Prism 8.3.0 software. A P value of < 0.05 was considered statistically significant. Correlations were only performed on dually detected samples.

Fig. S1 is a graphical outline of the study. Fig. S2 shows a heatmap of genes differentially expressed between COVID-19 patients at different time points after symptom onset. It also shows seroconversion relative to time after symptom onset and the relations between the IFN score and clinical or virological parameters. Fig. S3 shows the dynamics of blood expression of transcripts related to different immune pathways during SARS-CoV-2 infection. Fig. S4 shows the correlation between blood IFN-I/III scores and serum IFN-λ₁ levels. Fig. S5 shows the effect of a serum from a non–APS-1 patient containing IFN-I auto-Abs on the antiviral activity of IFN-I in the HAE model of SARS-CoV-2 infection. Table S1 lists all genes whose expression is measured in the host response NanoString panel and functional annotation of these genes. Table S2 lists genes differentially expressed between blood cells in mild COVID-19 patients at diagnosis and blood cells from HCs.Table S3 is a contingency table relative to IFN-α auto-Ab detection from nasal and plasma paired samples from mildly symptomatic and critically ill COVID-19 patients.

We thank all patients for their participation. We also thank all personnel of the occupational health and medicine department of Hospices Civils de Lyon who contributed to sample collection. Human biological samples and associated data were obtained from NeuroBioTec (Centre de Ressources Biologiques des Hospices Civils de Lyon, Lyon, France; Biobank BB-0033-00046). We thank Karima Brahami and all members of the Direction de la Recherche Clinique et de l’innovation (Hospices Civils de Lyon) for their reactivity and Emilie Laurent and Cedrine Milesi (VirPath team) for their technical support. The authors would like to thank Hélène Boyer for a helpful review of the manuscript.

T. Walzer’s laboratory is supported by the Agence Nationale de la Recherche, the Fondation ARC (équipe labellisée), and the Institut National du Cancer and receives institutional grants from Institut National de la Santé et de la Recherche Médicale, Centre National de la Recherche Scientifique, Claude Bernard Lyon 1 University, and École normale supérieure de Lyon. The VirPath team is supported by REACTing, Institut National de la Santé et de la Recherche Médicale, Centre National de la Recherche Scientifique, the Agence Nationale de la Recherche, and Institut Mérieux (Mérieux Research grant). This study was supported by the Hospices Civils de Lyon and the Fondation des Hospices Civils de Lyon. The Laboratory of Human Genetics of Infectious Diseases is supported by the National Institutes of Health (R01AI088364), the National Center for Advancing Translational Sciences, National Institutes of Health Clinical and Translational Science Award program (UL1 TR001866), an Emergent Ventures Fast Grant, the Mercatus Center at George Mason University, the Fisher Center for Alzheimer’s Research Foundation, the Meyer Foundation, the JPB Foundation, the Agence Nationale de la Recherche under the “Investments for the Future” program (ANR-10-IAHU-01), the Integrative Biology of Emerging Infectious Diseases Laboratory of Excellence (ANR-10-LABX-62-IBEID), the French Foundation for Medical Research (EQU201903007798), the French Foundation for Medical Research and Agence Nationale de la Recherche GENCOVID project (ANR-20-COVI-0003), Agence Nationale de la Recherche Nord-Sud (ANRS-COV05), Agence Nationale de la Recherche AABIFNCOV project (ANR-20-C11-0001), the Square Foundation, Grandir - Fonds de solidarité pour l’enfance, the SCOR Corporate Foundation for Science, the Howard Hughes Medical Institute, the Rockefeller University, the St. Giles Foundation, Institut National de la Santé et de la Recherche Médicale, and the University of Paris.

Author contributions: J. Lopez contributed to conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, and writing (original draft). M. Mommert contributed to conceptualization, data curation, formal analysis, validation, investigation, visualization, and methodology. W. Mouton contributed to conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, and writing (original draft, review and editing). A. Pizzorno contributed to conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, and writing (original draft, review and editing). K. Brengel-Pesce contributed to conceptualization, supervision, validation, investigation, and methodology. M. Mezidi contributed resources. J.-B. Fassier contributed resources and contributed to funding acquisition. M. Villard contributed to investigation and methodology. B. Lina contributed to methodology and funding acquisition, J.-C. Richard contributed resources. V. Cheynet contributed to data curation, formal analysis, supervision, validation, and methodology. B. Padey contributed to formal analysis and methodology. V. Duliere contributed to formal analysis and methodology. T. Julien contributed to formal analysis and methodology. S. Paul contributed resource. P. Bastard contributed to conceptualization, formal analysis, validation, investigation, visualization, methodology, and writing (original draft). A. Belot contributed resources and contributed to conceptualization. A. Bal contributed to conceptualization, formal analysis, validation, and methodology. J.-L. Casanova contributed to conceptualization and writing (original draft, review and editing). M. Rosa-Calatrava contributed to methodology and funding acquisition and writing (original draft, review and editing). F. Morfin contributed to conceptualization, data curation, formal analysis, validation, investigation, and methodology. T. Walzer contributed to conceptualization, methodology, investigation, visualization, and writing (original draft, review and editing). S. Trouillet-Assant contributed to conceptualization, resources, funding acquisition, investigation, visualization, methodology, project administration, and writing (original draft, review and editing).