The Chemagen 4 K extraction kit (CMG-1074) was used according to the manufacturers’ instructions. Genomic DNA (gDNA)from the patients was isolated from whole blood. Targeted sequencing (Illumina Hiseq2500) for primary immunodeficiency (PID) genes was performed using a standard PID panel on Nimblegen v4 data. Annotation was done using RefSeq (release 78) and Alissa Interpret (v5.0). Pathogenic variants in NFKB1 were verified by Sanger sequencing in index patients and family members.

The NFKB1 PDB structure was extracted from the RCSB database (7). The V263L mutation was introduced in the crystal structure using the built-in Pymol mutagenesis tool. To highlight the effect of the V263L mutation upon protein folding, the apo structure of both wild type (WT) and mutant NFKB1 were further processed for unrestrained molecular dynamics (MD) simulations. All preparative and unrestrained MD simulations in explicit solvent were run using the GPU version of the PMEMD engine provided with AMBER18 package. The ff14SB force field is used to describe the protein in the generated complex (8). Missing hydrogens and counter ions for neutralization are added through the LEAP module. Prior to the set-up, all systems are immersed into an octahedral box with TIP3P water molecules, spacing the atoms of the apo protein or protein complex 12 Å from the boundary of the simulation box edges. SHAKE algorithm is applied to all systems to constrain all bond lengths involving hydrogen atoms allowing a 2-fs time-step. The cut-off distance for van der Waals interaction is set at 12 Å. A system of Particle Mesh Ewald (PME) method is used to treat long-range electrostatic interactions. To remove steric clashes, a seven-step minimization procedure involving 1000 steps of steepest descent energy minimization is first performed followed by 4000 steps of conjugate gradient minimization at each step. The system is then slowly heated to 300 K after which the complexes are equilibrated by 2 ns position restraint MD simulations with 10.0 kcal/mol/Ų constant force on the heavy atoms of protein and substrate under NPT condition (1 atm). After this equilibration protocol, unrestrained molecular dynamics is performed on all solvated systems under control of a Berendsen thermostat (1 atm) and a Langevin thermostat (300 K) using periodic boundary conditions applied in all three cartesian directions to mimic the infinity of the system. The nonbonded list is updated every 25 steps, with periodic new random number seeds to prevent simulation synchronization of the trajectories. The trajectories are sampled every 2 ps for analysis in production dynamics. Unrestrained 100 ns MD simulations were finally carried out under NPT conditions. To estimate the stability of the WT compared to the mutant NFKB1 of the ligand with the protein target, the cpptraj module in AmberTools18 was used to additionally calculate root-mean-squared deviation (RMSD) on Cα atoms from the average structure from the MD simulation.

Whole blood was diluted 1:1 with RPMI 1640 and layered over lymphocyte separation medium (LSM) (MP Biomedicals, 0850494-CF). Tubes were centrifuged at 400 x g for 25 minutes and PBMCs were harvested and stored in liquid nitrogen until FACS staining or used immediately for functional assays.

Fresh or thawed PBMCs were plated overnight, and afterwards stimulated for indicated time points with 50 ng/ml+ 1 µg/ml PMA/ionomycin (Sigma-Aldrich) in RPMI 1640 medium (supplemented with FBS 5% + HEPES+ MEM non-essential aminoacids) (GibcoTM). After stimulation, PBMCs were lysed in lysis buffer [50 mM Tris-HCl pH 7.5, 135 mM NaCl, 1.5 mM MgCl2, 1% Triton-X, 10% glycerol, 1X protease inhibitor (Pierce TM Protease Inhibitor, ThermoFisher Scientific) and 1X phosphatase inhibitor (PhosSTOP, Roche)]. Protein concentrations were determined using a Bradford Protein Assay (Bio-Rad). Protein lysate was denaturized in LDS (NuPAGE LDS sample buffer, Novex) and DTT (Bio-Rad) at 70°C for 10 min and was then loaded on a 4-12% Bis-Tris polyacrylamide gel (BoltTM Bis-Tris Plus, Thermo Fisher Scientific) in MOPS buffer (NuPAGE MOPS SDS running buffer, Novex). Separated proteins on the gel were transferred onto a methanol-activated PVDF membrane (GE Healthcare) in transfer buffer [10% methanol, 1X Tris/Glycine Buffer (Bio-Rad)]. After transferring, the PVDF membrane was blocked in 5% milk TBS-T for 30 min at RT and then incubated with primary antibody O/N at 4°C followed by a wash with TBS-T and incubation with a secondary antibody for 1 h at RT. Primary antibodies used for Western blot: rabbit anti-p105/p50 (Cell Signaling Technology, D4P4D), rabbit anti-p-p105 (Cell Signaling Technology, 18E6) and mouse anti-GAPDH (Invitrogen). Secondary antibodies used were conjugated with HRP: goat anti-rabbit (Abcam) and anti-mouse (Rockland). Secondary antibodies were visualized using ECLTM15 Prime Western Blotting Detection Reagent (AmershamTM) with the G:box Chemi-XRQ and quantified using ImageJ.

mRNA was isolated from PBMCs using Trizol reagent (Life Technologies). 500 µL of trizol was added to PBMCs or differentiated macrophages and frozen at -80°C. Later, samples were thawed for further processing. 100 µL chloroform was added and cells were centrifuged (10000 x g) for 15 min at 4°C. Aqueous solution was collected, 1 µL Glycoblues and 200 µL isopropanol were added and samples were incubated at RT for 15 min. Thereafter, samples were centrifuged (10000 x g) for 15 min at 4°C, supernatant was removed, 500 µL of 75% ethanol was added, samples were again centrifuged (10000 x g) for 15 min at 4°C, supernatant was removed and samples were air-dried for 10 min. Finally, the pellet was resuspended in 15 µL elution buffer (MACHEREY-NAGEL). Complementary DNA (cDNA) was synthesized from RNA using the GoScriptTM Reverse Transcription System (Promega) via RT-PCR (SimpliAmp Thermal Cycler, ThermoFisher Scientific).

Forward and reverse primers were purchased from Integrated DNA Technologies (IDT). For NFKB1 forward primer AACAAGAAGTCTTACCCTCAGGT and reverse primer AGATCCCATCCTCACAGTGTTTT were used. For IL1B we used forward primer TGGCAATGAGGATGACTTGT and reverse primer GGAAAGAAGGTGCTCAGGTC. HPRT1 was used as housekeeping gene for all performed qPCR’s. The reaction was performed as followed: 1 µl of primers (0.75 µM), 8 ng of DNA (2 ng/µl) and 1X Fast SYBR TM Green Master Mix (ThermoFisher Scientific) were mixed and plated on a 96-well plate. The plate was run on the StepOneTM Real-Time PCR system (ThermoFisher Scientific) and analyzed with StepOne Software v2.3. For detection of any alternative spliced product in the context of the c.1638-2A>G mutation, we performed a PCR reaction using a forward primer TGGTGAGGTCACTCTAACGTA and reverse primer GGGGTGTGGTTCCATCGTAG. 10 µM of primers and 10 ng of cDNA were added to a PCR reaction mix containing 1X Q5 reaction buffer (New England Biolabs), 10 mM dNTP (Promega), Q5 high GC enhancer (New England Biolabs) and Q5 hot start high fidelity polymerase (New England Biolabs). PCR cycling was performed at the following parameters: One cycle of denaturation (98°C for 30 s), 35 three-segment cycles of amplification (98°C for 10 s, 62°C for 30 s and 72°C for 30 s) and a final cycle of extension (72°C for 2 min). 1X DNA loading dye (Thermo Scientific) was added to the PCR products and loaded on a 1% agarose gel (VWR Life Science). PCR bands were visualized with the Gbox ChemiXRQ.

PBMCs, isolated as described in the section PBMC isolation, were used for monocyte purification using negative selection for CD14++CD16- monocytes (Miltenyi, 130-117-337) according to the manufacturer’s protocol. Before plating, purity of the samples was evaluated on flowcytometry by CD14 staining (>80% of total cells). Monocytes were counted, resuspended in macrophage serum free medium (Gibco, 12065-074) and seeded in a 96 well plate at a density of 1x10⁶ cells/ml in 100 µL with GM-CSF 10 ng/ml (Biolegend, 572902). Medium was changed every 3 days and macrophages were harvested at day 7. Macrophages were primed with lipopolysaccharide (LPS, 1 µg/ml; Sigma) for 6 h to induce the expression of IL1B, and thereafter activation of the NLRP3 inflammasome was induced with adenosine triphosphate (ATP, 5 mM, 45 min). Secretion of the mature IL-1β was measured by ELISA (Abcam, ab46052) according to manufacturer´s recommendations. Cells were further used for RNA extraction as described in the section isolation of mRNA and generation of cDNA.

PBMCs were isolated from heparinized blood samples of human healthy donors using Ficoll-Paque density centrifugation (MP biomedicals), frozen and then stored in liquid nitrogen. Frozen PBMCs were thawed and counted, and cell concentration was adjusted to 5 x 10^6 for each sample.

Cells were plated in a V-bottom 96-well plate, washed once with PBS (Fisher Scientific) and stained with live/dead marker (fixable viability dye eFluor780, eBioscience) and fluorochrome-conjugated antibodies against surface markers ( Supplementary Table E1 ) to identify the following subsets: T cells (CD3⁺, CD4⁺ and CD8⁺), B cells (total CD3^-CD19⁺, naïve CD27^-, memory CD27⁺, switched memory CD27⁺IgM^-IgD^-, transitional CD24⁺CD38⁺ and CD21^low B cells), NK cells (total CD3^-CD56⁺, CD56^brightCD16^-/+, CD56^dimCD16⁺ NK cells) and monocytes (classical CD14⁺CD16^-, intermediate CD14⁺CD16⁺ and non classical CD14^-CD16⁺ monocytes).

Samples were stained for 60 min at 4°C, washed twice in PBS/1% FBS (Tico Europe), and then fixed. Cells were stored overnight at 4°C and were then acquired on a Symphony flow cytometer with Diva software (BD Biosciences). A minimum of 5 x 10^5 events were acquired for each sample. Compensation beads (UltraComp eBeads, ThermoFisher) were used to optimize fluorescence compensation settings for multi-color flow cytometric analysis at a Symphony flow cytometer.

Fresh whole blood (lithium heparin) was incubated with 10 µg/mL polymyxine B (Sigma-Aldrich), CD45-V500 (BD biosciences), and CD14-FITC (BD biosciences) in the absence or presence of 75 ng/mL IL-1b (eBioscience), 1 µg/ml Pam3CKS4 (In vivogen), 3 µg/mL R848 (In vivogen), or 20 ng/mL TNF-a (Peprotech). Cells were incubated for 15 min at 37°C, 5% CO2. Thereafter, the RBCs were lysed, fixed [with Lyse/Fix Buffer (BD biosciences) for 10 minutes at 37°C, 5% CO2], permeabilized (Perm Buffer II, BD biosciences) for 30 minutes on ice, washed with Stain buffer (BD biosciences) and incubated with 1:20 diluted PE Mouse anti-IkBa (BD biosciences).

Plasma cytokines were determined using a multiplex assay (MSD multi spot assay system, proinflammatory panel 1), according to the manufacturer’s instructions.

Direct digital detection of mRNA levels of selected genes from patients’ PBMCs was performed using nCounter^® Analysis System (NanoString Technologies) as described in (9).

Interaction analysis was done following the workflow detailed by Liu et al. (10, 11). The NFKB1 constructs were expressed with N-terminal MAC-tags and three biological replicates were prepared for both AP-MS and BioID, for each cell line.

2μl of the reconstituted samples were analyzed using the Evosep One liquid chromatography system coupled to a hybrid trapped ion mobility quadrupole TOF mass spectrometer (Bruker timsTOF Pro) via a CaptiveSpray nano-electrospray ion source. An 8 cm × 150 µm column with 1.5 µm C18 beads (EV1109, Evosep) was used for peptide separation with the 60 samples per day method (21 min gradient time). Mobile phases A and B were 0.1% formic acid in water and 0.1% formic acid in acetonitrile, respectively. The MS analysis was performed in the positive-ion mode using data-dependent acquisition (DDA) in PASEF mode with 10 PASEF scans per topN acquisition cycle.

Raw data acquired in PASEF mode were processed with Fragpipe version 17.1, including MSFragger version 3.4, against the human entries of Uniprot database (UP000005640, 20371 entries). Carbamidomethylation of cysteine residues was used as static modification. Aminoterminal acetylation and oxidation of methionine were used as the dynamic modification. Trypsin was selected as enzyme, and maximum of two missed cleavages were allowed. Label-free quantification parameters were left to default settings. Peptides with false discovery rate (FDR) < 0.05 were exported from the search results.

The resulting protein detections were filtered using SAINT and MAC-tagged GFP control interactions. Interactors with < 0.01 BFDR were chosen as the filtered high-confidence interactors. The triplicate samples were individually bait normalized by the NFKB1 spectral counts and the normalized average spectral counts were then used to made dotplots with ProhitsViz.The filtered BioID results were used to run GO biological process term enrichment analysis using the DAVID bioinformatics resources (PMID: 19131956).

Data with ≥2 independent experiments were pooled (if available) for statistical analysis using a Mann-Whitney U test.

With next generation sequencing, we studied a cohort of patients with broad immunological phenotypes in the adult immunology clinic. Targeted sequencing with a PID panel revealed the presence of 2 heterozygous NFKB1 variants in 2 different families. The clinical and immunological characteristics are summarized in Tables 1 , 2 .

Family 1 ( Figure 1A , Supplementary E1A ) included 2 affected members who presented with recurrent respiratory tract infections and B cell dysfunction. The 47-year-old index patient (1.II.1) met the ESID criteria for CVID and had a more severe course of disease characterized by hospitalizations for pneumonia and pyelonephritis. In addition, as part of the non-infectious CVID spectrum, she suffered from psoriatic arthritis (treated with sulfasalazine and afterwards methotrexate, both stopped before evaluation of CVID because of liver function test disorders), non-malignant intra-abdominal and thoracic lymphoproliferation at the age of 47 with coexistent chronic diarrhea at that time. Lymph node biopsy showed expansion of T and B cells in the interfollicular region but no signs of malignancy; gastro-intestinal biopsies (duodenal, colon) were unremarkable ( Table 1 ). Eventually these symptoms spontaneously resolved without the need of immunomodulating agents. In immunophenotyping, she had low B cells, low switched memory B cells, hypogammaglobulinemia and abnormal pneumococcal vaccine response ( Table 1 ). Her 19-year-old daughter (1.III.1) only displayed mild infections never requiring hospitalization, decreased IgG levels and abnormal pneumococcal vaccine response, with normal B cells. Genetic analysis by targeted, also Sanger-confirmed sequencing showed a heterozygous c.787G>C (p.V263L) mutation in NFKB1.

Family 2 ( Figure 1A and Supplementary E1A ) included 4 affected members with variable presentation (clinical or subclinical). The index patient (2.II.1) was diagnosed with CVID at the age of 55 after investigations for polymyalgia rheumatica. Laboratory results before start of corticosteroids showed marked hypogammaglobulinemia, low switched memory B cells and no responses to pneumococcal vaccine. He did not report any severe infections during child- or adulthood. His 30-year-old daughter (2.III.1) presented at the age of 26 with recurrent sinopulmonary infections requiring antibiotic treatment. She had hypogammaglobulinemia, low switched memory B cells, and no responses to pneumococcal vaccine on laboratory evaluation. At the age of 28, she developed a hitherto unexplained bilateral keratitis refractory to diverse lubricants and ultimately treated with punctal plugs. His 31-year-old son (2.III.2) had a history of recurrent cervical adenopathy at the age of 20 years, for which he underwent an excision biopsy showing reactive hyperplasia. Infectious and autoimmune serology at that time was negative. Laboratory analysis revealed low IgM but normal IgG and IgA. Both siblings from the index patient (2.II.2 and 2.II.3) were asymptomatic and laboratory results only showed abnormalities in 2.II.2 with isolated hypogammaglobulinemia. Targeted sequencing in the index showed a heterozygous splicing mutation c.1638-2A>G, located before exon 16. Sanger sequencing confirmed the presence of this mutation in both his children and siblings.

The two mutations identified in NFKB1 affected different domains of the protein structure ( Figure 1B ). The c.787G>C mutation (Family 1) led to a substitution of a conserved valine to a leucin at the 263 residue (p.V263L) in the Rel Homology Domain (RHD) of the NFKB1 protein. An intronic mutation in Family 2, c.1638-2A>G, was located before exon 16. According to the database splicing consensus single nucleotide variant (dbscSNV) the intronic mutation was predicted to alter splicing (ADA score 0.99 and RF score 0.94). Both mutations were not previously reported in public genomic databases (gnomad, ExAC and 1000 genomes). Based on in silico prediction scores (CADD, PolyPhen 2, SIFT), they were predicted to be damaging to the protein structure or function ( Supplementary E1B , CADD score). Structural modelling of the p.V263L mutant showed a disruption of the C-terminal β-barrel ( Supplementary E1C , right panel, dotted circle 1) which could trigger an unfolding event. Additionally, an upstream loop is converted into a twisted β-sheet exposing hydrophobic residues to the water interface which could promote protein aggregation ( Supplementary E1C , right panel, dotted circle 2). Furthermore, a single-turn α-helical structure is destabilized. In conclusion, these findings predict that both mutations are likely deleterious.

We assessed the transcriptional levels of NFKB1 by qPCR and protein expression of p50/p105 in PBMCs of affected and unaffected family members ( Figures 2A–C and Supplementary E2 ). Phosphorylation of p105 was examined after stimulation of PBMCs with phorbol myristate acetate and ionomycin. In the first family, the p.V256L mutation had no overall effect on mRNA expression, although there was considerable interindividual variation. The index patient 1.II.1, with the most severe phenotype, had slightly reduced levels (75% compared to healthy controls), while her daughter who was mildly affected produced normal levels of NFKB1 mRNA ( Figure 2A ). Protein levels of p50/p105 were reduced in both of them, albeit more significantly in 1.II.1 compared to healthy controls. Phosphorylation of p105 was reduced only in 1.II.1 ( Figures 2B, C ). In the second family, all carriers of the splicing variant c.1638-2A>G had 30-50% of transcriptional levels compared to healthy controls, suggesting nonsense mediated decay of a putative alternatively spliced mRNA ( Figure 2A ). An alternatively spliced product was not detected by PCR on cDNA using primers spanning exon 14-19 ( Supplementary E3 ). As expected, protein levels of p50/p105 were reduced in all patients, except for 2.II.3, who was unaffected ( Figures 2B, C ). Phosphorylation of p105 was reduced in all members (2.III.1 not tested), except for the unaffected member 2.II.3. To study if our NFKB1 haploinsufficiency patients had any upstream defect in canonical NF-κB signaling, we assessed IκBα degradation on CD3+ T cells and CD14⁺ monocytes, using different stimuli, as a surrogate marker ( Figures 3A, B ). These experiments showed that IκBα was degraded to a similar degree in all patients compared to healthy controls.

We assessed NLRP3 inflammasome activity on differentiated macrophages both by qPCR (IL1B transcriptional level) and ELISA (IL-1β secretion) of affected members in both families ( Figures 4A, B ), and interferon stimulated gene (ISG) expression by NanoString on PBMCs ( Figure 4C ). Serum cytokines were determined on plasma samples ( Figure 4D ). In family 1, only the index patient (1.II.1) had inflammatory symptoms (psoriatic arthritis), while her daughter did not present with any inflammatory feature (till present). NLRP3 inflammasome activity upon stimulation of differentiated macrophages was higher in both patients (1.II.1, 1.III.1). ISG signature was only determined for the index patient (1.II.1) and upregulated compared to a matched healthy control. Consistent with an inflammatory phenotype, the majority of proinflammatory cytokines (IL-6, TNFα and IFN-γ) were increased in patient 1.II.1, while remaining normal in patient 1.III.1. Two affected patients of family 2 (2.II.1 and 2.III.1) had inflammatory symptoms (polymyalgia rheumatica and bilateral idiopathic keratitis). NLRP3 inflammasome activity was persistently higher for both compared to healthy controls. Serum cytokine measurements showed comparable levels of proinflammatory cytokines IL1β, IL-6, TNFα, IL-8 and IFN-γ for 2.II.1 (sample obtained at a time of remission, yet no sample was available during disease activity for comparison) and high levels for 2.III.1 (sample obtained during active keratitis) compared to healthy controls. Not surprisingly, pro-inflammatory cytokines determined on the lacrimal fluid of patient 2.III.1 were higher compared to a matched healthy control ( Supplementary E4 )

We assessed the B cell compartment of affected patients using flow cytometry ( Supplementary E5 ). Results of routine clinical immunophenotyping are found in Table 2 . In general, affected patients had lower B (CD19+), memory B (CD27+), switched memory B (CD27+IgM-IgD-), and high CD21low B cells ( Figure 5 ). In our patients, no differences were seen within the general NK cell compartment ( Supplementary E6 , E7 ), and maturation (assessed by CD27 and CD57) on CD56dimCD16+ cells did not differ from healthy controls ( Supplementary E6 , E6 ). In addition, the monocyte compartment ( Supplementary E6 , E8 ) was similar to healthy controls in subsets (classical-intermediate-nonclassical) and activation markers (CD86, CD38 on classical monocytes).

To study the changes in protein interactions, caused by p.V263L, we analyzed the stable (AP-MS) and proximal (BioID) interactomes of V263L and WT NFKB1 by generating stable inducible cell lines, expressing the studied NFKB1 constructs tagged with both HA affinity tag and BirA biotin ligase tag. The AP-MS and BioID analysis showed that WT NFKB1 interacted with well-known NFKB1 interactors including the NFKB1 family members, NFKB1 inhibitors and several proteins associated with NFKB1 regulation ( Figure 6 ). In both analyses, V263L strikingly led to a reduction or complete loss for almost all interactions compared to the WT ( Figure 6A and Supplementary E9 ). However, the loss of interaction with the proteasome subunits in AP-MS was not confirmed in the BioID analysis and V263L maintained preserved interactions with NFKB2 and TF65. Gained interactions of V263L with hypoxia-inducible factor 1-alpha inhibitor (HIF1N) and NOTCH1 were observed in both analyses. HIF1N hydroxylates Asn678 within ankyrin repeat domains (ARD) of NFKB1 (12, 13). So far, the role of this hydroxylation has not been elucidated (12). NOTCH1 has been linked to the inhibition of NF-κB activity as its N-terminal portion inhibited p50 DNA binding and interacted specifically with p50 subunit, but not p65 of NF-κB (14). Additional interaction partners exclusive to V263L by the BioID analysis, not retrieved in the AP-MS analysis, can be found in Supplementary E9 . Next, using the BioID interactors from WT and V263L NFKB1, we performed gene ontology (GO) enrichment analysis to investigate in which biological processes they were involved compared to a population database set (total of 19349 genes). Figure 6B shows the top 10 gene sets within the biological processes enriched for WT and V263L with p-values, gene counts and fold enrichment values. As expected for WT, interaction proteins are encoded by genes involved in NF-κB signaling, translation and transcription, viral immunity and posttranslational modification. For V263L, 8 out of the 10 most significantly enriched biological processes overlapped with those observed for WT. Although the mutant NFKB1 had a decreased binding affinity towards various proteins involved in NF-κB signaling, seen in the AP-MS ( Figure 6A ) and BioID ( Supplementary E9 ) analysis, it still interacted with proteins involved in similar biological processes to WT.