Pulmonary Alveolar Proteinosis and Multiple Infectious Diseases in a Child with Autosomal Recessive Complete IRF8 Deficiency

The patient (II. 2) was a boy born in 2019 to non-consanguineous parents originating from and living in Jujuy, in the North of Argentina (Fig. 1a). His maternal half-sister and his two parents are healthy. He was born at 36 weeks of gestation, with a normal weight and length. Ventriculomegaly was diagnosed on prenatal ultrasound. The patient received BCG and hepatitis B vaccines at birth. He was admitted to the intensive care unit (ICU) at the age of one month, for sepsis and respiratory distress requiring assisted mechanical ventilation. No microorganisms were documented during this first episode, and the patient improved, thanks to empirical broad-spectrum antibiotics. At 4 months of age, he was hospitalized for a second episode of respiratory distress syndrome. Multiplex PCR on nasopharyngeal secretions was positive for a non-SARS-CoV-2 coronavirus but negative for adenovirus, influenza virus, metapneumovirus, Mycoplasma pneumoniae, and Pneumocystis jirovecii. PCR tests on blood for Epstein-Barr virus (EBV), cytomegalovirus (CMV), human herpesvirus 6 (HHV6), and HIV yielded negative results. Acinetobacter baumanii was cultured from bronchoalveolar lavage (BAL). The patient presented one episode of febrile convulsion. A cerebrospinal fluid (CSF) sample collected at the time was found to contain no cells, and normal protein and glucose levels. PCR on CSF for enterovirus, EBV, varicella-zoster virus (VZV), CMV, herpes simplex virus (HSV), and HHV6 yielded negative results. No mycobacterial culture was performed on CSF. The patient was treated with meropenem, vancomycin, colistin, trimethoprim/sulfamethoxazole, and liposomal amphotericin. At the age of 7 months, an axillary adenopathy, 2 × 3 cm in size, was detected at the BCG vaccination site. M. bovis-BCG was isolated from BAL, and the patient was treated with pyrazinamide, ethambutol, rifampicin, and levofloxacin.

A computed tomography (CT) scan of the thorax found mediastinal structures without density alterations, with a thickening of the septal interstitium, overlapping ground glass opacities with a cobblestone pattern and a tendency to consolidate at the base of the lungs (Fig. 1b). These findings were consistent with PAP, which was confirmed by pathological examinations of BAL and of lung biopsy (Fig. 1c). The patient was treated with a bolus of steroids and partial lung washes that were performed weekly at the ICU, with a good response at the start of treatment. The patient also presented facial dysmorphism, with a very wide fontanel, mid facial hypoplasia, bilateral keratoconus, a short nose and neck, retrognathia, and short stature (− 3 SD at the age of one month). Brain CT scan was performed at the age of seven months and showed ventriculomegaly with mega-cisterna magna, cerebellar hypoplasia, mesencephalic, temporal, ventricular, and ependymal calcifications (Fig. 1d). The patient died at the age of 10 months whilst in palliative care for respiratory failure. Overall, he presented susceptibility to life-threatening mycobacterial (BCG-osis) and viral (coronavirus) infections, associated with severe PAP, and cerebral calcifications.

Whole-exome sequencing (WES) was performed for this patient. The homozygosity rate inferred from WES was low (0.37%), consistent with the absence of consanguinity between the two parents. Principal component analysis (PCA) confirmed that the patient was of Latino American descent (Fig. 1e). An analysis of rare variants with a minor allele frequency (MAF) < 10⁻³, predicted to be deleterious, for genes with a low gene damage index, identified 95 variants (Fig. 1f). Only two of these variants concerned a gene known to be involved in IEI, IRF8. One of these heterozygous variants was a single-base pair deletion (c.55del) predicted to create a frameshift (p.D19Tfs*8), and the other was a single-base pair substitution (c.331C > T) that was predicted to result in a nonsense variant (p.R111*) (Supplementary Fig. 1a). Both variants were confirmed by Sanger sequencing (Fig. 1g). The c.55del variant was inherited from the patient’s mother, whereas the c.331C > T variant was inherited from the patient’s father; the patient was therefore compound heterozygous (Fig. 1a and g). Both variants were predicted to be deleterious, with combined annotation-dependent depletion (CADD) scores of 33 for c.55del and 37 for p.R111*. These values are above the 99% mutation significance cutoff (MSC) of 8.2 for IRF8 [2] (Fig. 1h). The p.R111* variant was previously reported in a Chinese patient with AR IRF8 deficiency [11]. However, neither of these variants has ever been reported in database of germline variants such as the public databases ExAC, gnomAD v2.1.1 or v3.1.1 [4], BRAVO/TOPmed freeze 8 [5], and ATAV [6] containing WES or whole-genome sequencing (WGS) data for more than 150,000 individuals, or in the WES data of our in-house cohort of more than 15,000 patients with severe infectious diseases (Fig. 1h). We also screened for copy number variants (CNV), in particular for homozygous or hemizygous large deletions [13]. Using WES data, targeted screening for heterozygous large deletions encompassing TBX1 locus was negative. No CGH or SNP microarray was performed due to the lack of material. No relevant rare or private variants in other genes for which mutations are known to be associated with PAP [14‐16] (namely, ADA, ABCA3, CSF2RA, CSF2RB, GATA2, MARS, NKX21, RAB5B, SFTPB, SFTPC, and SLC7A7), or susceptibility to mycobacterial disease [17] were identified by WES. However, biallelic mutations of IRF8 are known to underlie syndromic MSMD [1, 8‐10]. These results suggest that this patient had AR IRF8 deficiency.

Both variants of the patient were located in the DNA-binding domain of the IRF8 protein (Fig. 1i). We studied their impact by transiently transfecting human embryonic kidney (HEK)293 T cells with the wild-type IRF8 (WT; NM_002198. 2), or mutant c.55del or p.R111* cDNAs, and with the three previously described LOF mutants as controls (p.K108E [8, 9], p.R83C [10], p.R291Q [10]). Constructs were epitope-tagged at the carboxy-terminus (V5). Transfection was efficient, as shown by qPCR, with similar levels of IRF8 cDNA detected in HEK293T cells transfected with the WT and mutant alleles (Supplementary Fig. 1b). Immunoblotting of cell extracts with either the monoclonal IRF8 antibody recognizing p.G85 at the amino-terminus or an antibody recognizing the carboxyterminal V5 tag showed that the WT form, and the p.K108E, p.R83C, and p.R291Q variants were produced at the expected molecular weight (MW) (Fig. 2a). Surprisingly, a band was detected at a molecular weight slightly lower than expected for the c.55del variant, whereas no protein was detected at all for the p.R111* variant (Fig. 2a). Two ATG codons downstream from c.55del and upstream from p.R111* were identified and predicted in silico to lead to a re-initiation of translation (encoding p.M22 and p.M35, respectively) [11]. We tested this hypothesis, by mutating the corresponding methionine codons (p.M22 and p.M35) to create a methionine-to-alanine (ATG > GCG) substitution. In the WT-IRF8 cDNA, the mutation of p.M22 (WT/M22A), p.M35 (WT/M35A), or both (WT/M22A/M35A) had no effect on the MW of the protein produced (Fig. 2a). Conversely, in c.55del cDNA, the mutation of p.M22 (c.55del/M22A) resulted in protein with a slightly lower MW, whereas the mutation of p.M35 (c.55del/M35A) resulted in protein of similar MW to the c.55del protein (Fig. 2a). The substitution of both methionine residues (c.55del/M22A/M35A) completely abolished protein production (Fig. 2a). Thus, in overexpression systems, the c.55del mutation leads to a re-initiation of translation at the downstream ATG encoding p.M22. The resulting protein is truncated by 21 amino acids and has a lower MW (p.M1_S21del). The c.55del variant displayed impaired accumulation in the nucleus, like the previously described p.K108E mutant [9] as shown by analysis of cytoplasmic and nuclear extracts from HEK293T cells (Supplementary Fig. 1c) and by confocal microscopy on HeLa cells (Fig. 2b). In luciferase assays on HEK293T cells, the c.55del and p.R111* proteins, like the three previously reported mutants (p.K108E [8, 9], p.R83C [10], p.R291Q [10]), were completely unable to inhibit IRF1-driven ISRE-transcriptional activity (Fig. 2c). These results indicate that both the variants from the patient are LOF and suggest that the patient displayed AR complete IRF8 deficiency.

The patient displayed constant marked leukocytosis due to the proliferation of polymorphonuclear neutrophils (Fig. 3a). Monocytes were not identified with any certainty on conventional hemograms (probably confused with large lymphocytes) (Fig. 3a), and the complete absence of these cells was demonstrated by flow cytometry (Fig. 3b). Bone marrow aspiration was performed, and no signs of lymphoma or leukemia were detected. The patient had slightly high plasma FTL3 ligand levels (Supplementary Fig. 1d). Circulating pDC and cDC1 were also found to be absent by flow cytometry (Fig. 3b). The number of NK cells was within the normal range (Fig. 3c), and the distribution of CD56^br and CD56^dim cells was normal (Fig. 3d). The patient displayed mild CD4⁺ T lymphopenia (Fig. 3c), with a normal percentage of naïve CD4⁺ T cells (Table 1). T_H17, T_H1, and Treg CD4⁺ T cells were present at their usual percentages, whereas the percentage of T_HF cells was slightly lower than normal (Table 1). CD8⁺ T cells were present within the normal range (Fig. 3c). The patient presented B-cell lymphopenia (Fig. 3c) and had low levels of memory B-cell subsets (CD27⁺, CD27⁺IgD⁺, and CD27⁺IgD⁻) for his age (Table 1). Platelet levels were either low or in the low end of the normal range (Fig. 3a). Despite an absence of pDCs, plasma IFN-α levels were at upper range of controls in the patient (Supplementary Fig. 1d). Following stimulation with IL-12, phytohemagglutinin (PHA)-activated peripheral blood mononuclear cells (PBMCs) or PHA-blasts from the patient displayed normal levels of intracellular IFN-γ production by CD8⁺ and CD8⁻ T cells (Supplementary Fig. 1e). Whole-blood activation with LPS + IL-12 resulted in an impaired IFN-γ production (data not shown). Consistent with a complete absence of circulating myeloid mononuclear cells, following activation with LPS and IFN-γ, whole blood from the patient was completely unable to produce IL-1β, IL-10, IL-12p70, and IL-23, and TNF levels were low (Fig. 1e). Screening for anti-GM-CSF and anti-type I IFN autoantibodies, which can underlie PAP and severe COVID-19 pneumonia [18], respectively, was negative in a plasma sample of the patient collected at the age of nine months (Supplementary Fig. 1f and 1g).

Genomic DNA was extracted from whole-blood samples with the QIAamp DNA Mini Kit (QIAGEN). Exons were captured with the Agilent SureSelect Human All Exon V5 kit and sequenced on an Illumina platform with a read length of 100 base pairs and a main depth coverage of 100 × . Paired-end reads were mapped onto the human reference genome GRCh37 with the Burrows-Wheeler aligner (BWA), and the following analysis was performed as recommended by the Broad Institute guidelines, with the Genome Analysis Toolkit (GATK). Duplicate reads were marked with PICARD, and variants were called with “Haplotype Caller” from GATK (100,341 variants were obtained). We annotated the final VCF file with SnpEff/SnpSift. The following databases were included: dbSNP, ExAC, 1000 Genomes, ClinVar, and the prediction tools Polyphen2, SIFT, and Mutation Taster. Variant prioritization was performed with the web-based server B_Platform (Bitgenia, www. bitgenia. com). Candidate variants were analyzed for presence in the set of candidate genes, population frequency, possible impact on the protein sequence, and previous clinical reports in databases.

Empty vector or plasmid (EV) pcDNA3.1 was obtained from Thermo Fisher Scientific, and pCMV6 EV and IRF1-WT-DDK tagged plasmid (#RCPS100001 and #RC203500, respectively) were obtained from Origene. IRF8-WT-V5-tagged (NM_002163) was inserted not the vector by cloning. Constructs carrying single-nucleotide mutant alleles were generated from this plasmid by mutagenesis with the appropriate primers, with the Pfu Ultra II Fusion HS DNA (#600,674, Agilent) polymerase or CloneAmpHiFi PCR Premix (#639,298, Takara) followed by digestion with DpnI (#R0176L, New England Biolab). Plasmids were amplified in competent cells of E. coli (#C3019H, New England Biolab) and were purified with a maxiprep kit (#12,663, Qiagen). HEK293T cells were transiently transfected with the various constructs at a concentration of 2.5 µg/mL, in the presence of Lipofectamine LTX reagent with PLUS (#15,338,100, Thermo Fisher Scientific) in accordance with the manufacturer’s instructions.

Total protein extracts were prepared by mixing cells with modified radioimmunoprecipitation assay buffer supplemented with protease inhibitors (Complete, Roche) and phosphatase inhibitor cocktail (PhosphoStop, Roche), 0.1 mM dithiothreitol (DTT; Life Technologies), 10⁻³ mM Na₃VO₄, and 1 mM PMSF, and incubating for 40 min on ice. The cytoplasmic and nuclear contents of the cells were separated with NE-PER nuclear and cytoplasmic extraction reagents (#78,835, Thermo Fisher Scientific). Equal amounts of protein, according to a Bradford protein assay (#5,000,002, Biorad), were resolved by SDS-PAGE in a Criterion TGX 10% precast gel (#5,671,033, #5,671,034, or #5,671,035, Biorad), and the resulting bands were transferred to a nitrocellulose membrane (#1,704,159, Biorad). Membranes were probed with antibodies directed against IRF8 (unconjugated, clone D20D8, #5628, Cell Signaling), the V5 tag (HRP-coupled, Thermo Fisher Scientific), lamin A/C (HRP-conjugated, clone E-1, #sc-376248, Santa Cruz), or GADPH (unconjugated, clone D16H11, #5174, Cell Signaling). Unconjugated antibodies were detected by incubation with HRP-conjugated goat anti-rabbit IgG (H + L) antibodies (#1,706,515, Biorad). Binding was detected by incubation with the Clarity Western ECL substrate (Biorad, #1,705,061) or SuperSignal West Femto (Thermo Fisher Scientific, #34,096) with ChemiDoc MP (Biorad). The Spectra Multicolor Broad-Range Protein Ladder (#26,623, Thermo Fisher Scientific) or the Chameleon Duo Prestained Protein Ladder (#928–6000, Licor) was used to provide molecular weight markers. Membranes were stripped with Restore Western Blot Stripping Buffer (#21,063, Thermo Fisher Scientific). Images were analyzed with Imagine Lab 6. 0. 1 build 34 (Bio-Rad Laboratories).

We used the ISRE3 reporter plasmid (pGL4. 10[luc2] backbone, Promega #E6651), as previously described [25]. HEK293T cells in 96-well plates were transiently transfected with the (ISRE) reporter plasmid (100 ng/well and 100 µL of DMEM-10% FBS medium) and the pRL-SV40 plasmid (Promega, #E2231, 40 ng/well), with or without the IRF1 WT p.CMV6 plasmid (50 ng/well), with or without IRF8-WT or mutant forms, in the presence of the Lipofectamine LTX kit (Thermo Fisher Scientific, #15,338–100) and with or without EV, to ensure that the same amount of plasmid was present in each well, in accordance with the manufacturer’s instructions. Cells were used for the ISRE assay with the Dual-Luciferase system kit (Promega #E1980), according to the manufacturer’s protocol, 24 h after transfection. Signal intensity was determined with a Victor X4 plate reader (Perkin Elmer). Experiments were performed in triplicate, and results are expressed as dual reporter activity.

Phenotypic analyses were performed on PBMCs with standard flow cytometry (FC) methods [26], on an eight-color FACS Canto II instrument (Becton Dickinson, San Jose, CA, USA) equipped with FACSDiva software (BD). Monoclonal antibodies (mAbs) — directed against CD3, CD4, CD11B, CD14, CD16, CD19, CD20, CD21, CD24, CD25, CD27, CD38, CD39, CD45, CD45RA, CD56, CD127, CD185 (CXCR5), IgD, IFNγ, and IL-17 — were purchased from Becton Dickinson (BD), San Jose, CA, USA. Antibodies against CD1c, CD141, CD303, CD304, and CD45v500 were acquired from Miltenyi Biotec, Bergisch Gladbach, Germany. The CD13 and CD11b staining patterns of peripheral mature neutrophils were analyzed, and monocyte subsets were identified on the basis of their expression of CD14 and CD16. DCs were identified on the basis of MHC class II (HLA-DR) expression by lineage negative (LIN −) cells (DR + LIN −) and were counted. The LIN population was defined by markers specific for the T-cell (CD3⁺), B-cell (CD19⁺, CD20⁺), NK-cell (CD56⁺), and monocyte (CD14⁺) lineages. DC subsets were identified on the basis of blood DC antigens (BDCA) 1–4 in the DR⁺ LIN⁻ population: CD1c (BDCA-1) and CD141 (BDCA-3), which were used to recognize cDCs, and CD303 (BDCA-2) and CD304 (BDCA-4), which were used to identify pDCs. DC percentages are expressed relative to total CD45⁺ leukocytes. Treg cells were identified as a CD127^dimCD25^high population on CD3⁺CD4⁺ cells. For the quantification of T_H17 and T_H1 cells, two million PBMCs were left unstimulated or were stimulated with phorbol 12-myristate 13-acetate (PMA, 100 ng/mL) and ionomicin (1 µg/mL) (Sigma Aldrich) for 15 h at 37 °C, under an atmosphere containing 5% CO₂, in the presence of brefeldin A (10 µg/mL). Cells were fixed and permeabilized with Cytofix/Cytoperm and Perm/Wash buffer from Becton Dickinson (BD) and stained with antibodies directed against CD3, CD4, IFN-γ, and IL-17. We analyzed the proportions of T_H17 and T_H1 cells among CD4⁺ T cells.

Whole-blood samples were diluted 1/2 with RPMI 1640 (GIBCO BRL) and stimulated with lipopolysaccharide (LPS) from Salmonella Minnesota (100 ng/mL, Sigma-Aldrich), PMA (10⁻⁷ M; Sigma-Aldrich) with ionomycin (I) (10⁻⁵ M; Sigma-Aldrich), IL-12 (GIBCO BRL), and IFN-γ (Imukin). Patient samples and samples from two age-matched healthy controls were run in parallel. Supernatants were recovered after 48 h and were used for enzyme-linked immunosorbent assays (ELISA) for human TNF and human IFN-γ (Thermo Fischer Scientific). ELISA was performed on plasma from the patient, GATA2-deficient patients [20] and healthy donors, for human Flt-3 ligand (FLT3L) (R&D Systems, DFK00) in accordance with the kit manufacturer’s instructions. Similarly, plasma IFN-α levels were measured by SIMOA on patient, and healthy donors, in accordance with the kit manufacturer’s instructions (Quanterix). SIMOA was also performed on plasma from the patient, and healthy donors, in accordance with the kit manufacturer’s instructions. The production of myeloid cytokines was assessed after 24 h of stimulation with LPS, with or without IFN-γ, with LEGENDplex™ Human Inflammation Panel 1 (#740,809, BioLegend).

PHA blasts from the patient and an age matched control were generated stimulating 1 × 10⁶ PBMC with PHA (Gibco, diluted 1/50 in RPMI 20% ABS Penicillin/Streptomycin/Glutamine) plus recombinant IL2 (Roche, 50 IU) during 48 h. After two washes with RPMI, mediumcells were divided in two aliquots and incubated either with RPMI 2% FCS (basal tube) or 0.1 µg/ml recombinant IL12 R&D (stimulated tube) for another 24 h. Then, both basal and stimulated tubes were re-stimulated with PMA 100 ng/ml and ionomycin (1 µg/ml) in RPMI 20% ABS P/S/G for 5 h in the presence of 10 µg/ml Brefeldin A (Sigma). Cells from basal and simulated tubes were then washed twice, lysed with Becton Dickinson (BD) FACS lysing, and permeabilized with BD FACS permeabilizing solution 2. Permeabilized cells were stained with IFN-γ FITC, CD3 PerCP, and CD8 PE (BD) for 30 min; washed twice with PBS /BSA; and acquired in a FACS Canto II. Percentages of IFN-γ positive cells from basal and IL12 stimulated blast cells were analyzed on gated CD3 + CD8 + and CD3 + CD8- populations on FlowJo.

Recombinant E. coli-derived human (rh)GM-CSF (#215-GMP, R&D Systems) was first biotinylated with EZ-Link Sulfo-NHS-LC-Biotin (#A39257, Thermo Fisher Scientific), according to the manufacturer’s instructions, with a biotin-to-protein molar ratio of 1:12. The detection reagent contained a secondary antibody (Alexa Fluor 647 goat anti-human IgG (#A21445, Thermo Fisher Scientific)) diluted in Rexxip F (#P0004825, Gyros Protein Technologies; 1:500 dilution of the 2 mg/ml stock to yield a final concentration of 4 μg/ml). Buffer phosphate-buffered saline, 0.01% Tween 20 (PBS-T), and Gyros Wash buffer (#P0020087, Gyros Protein Technologies) were prepared according to the manufacturer’s instructions. Plasma or serum samples were then diluted 1:100 in 0.01% PBS-T and tested with the Bioaffy 1000 CD (#P0004253, Gyros Protein Technologies) and the Gyrolab xPand (#P0020520, Gyros Protein Technologies). Cleaning cycles were performed in 20% ethanol.

Screening of autoantibodies to type I IFN-α2a 100 pg/mL, IFN-β 10 ng/mL, and IFN-ω 100 pg/mL in plasma diluted 1 to 10 were performed as previously described using a dual interferon stimulated response element (ISRE) luciferase system in HEK293T cells [18]. Results were normalized to non-stimulated condition.