Biallelic loss-of-function mutation in NIK causes a primary immunodeficiency with multifaceted aberrant lymphoid immunity

Primary immunodeficiency disorders enable identification of genes with crucial roles in the human immune system. Here we study patients suffering from recurrent bacterial, viral and Cryptosporidium infections, and identify a biallelic mutation in the MAP3K14 gene encoding NIK (NF-κB-inducing kinase). Loss of kinase activity of mutant NIK, predicted by in silico analysis and confirmed by functional assays, leads to defective activation of both canonical and non-canonical NF-κB signalling. Patients with mutated NIK exhibit B-cell lymphopenia, decreased frequencies of class-switched memory B cells and hypogammaglobulinemia due to impaired B-cell survival, and impaired ICOSL expression. Although overall T-cell numbers are normal, both follicular helper and memory T cells are perturbed. Natural killer (NK) cells are decreased and exhibit defective activation, leading to impaired formation of NK-cell immunological synapses. Collectively, our data illustrate the non-redundant role for NIK in human immune responses, demonstrating that loss-of-function mutations in NIK can cause multiple aberrations of lymphoid immunity.

P rimary immunodeficiency disorders represent unique models to identify factors essential for host defense and immune homeostasis. In humans, development of mature B cells from immature precursor cells is critically dependent on signalling pathways downstream of B-cell receptor (BCR) and on tumour necrosis factor-a (TNFa) receptor superfamily members including BAFF receptor (BAFFR), TACI and CD40 (reviewed in ref. 1). BAFFR signals are needed to mature beyond the transitional B-cell stage 2 , while lymphotoxin-a1/b2 (LTb) and CD40 ligand (CD40L) are required for thymic and secondary lymphoid organ structure, respectively 3 . CD40-mediated signalling additionally orchestrates processes dependent on CD4 þ T-helper cells such as class-switch recombination (CSR) and somatic hypermutation (SHM) in the germinal centre (GC) reaction and CD8 þ cytotoxic T-cell memory 4 .
BAFFR, CD40 and LTb receptors transmit signals through the non-canonical nuclear factor-kB (NF-kB) pathway (reviewed in ref. 5), which induces proteolytic processing of p100 to p52 (ref. 6). Together with RelB, p52 forms a heterodimer that upon nuclear translocation functions as transcriptional activator of a subset of NF-kB target genes 5 . Processing of p100 depends on the phosphorylation of the serine residues 866 and 870, which is controlled by the MAP3 kinase-kinase-kinase NIK (NF-kB inducing kinase, MAP3K14) 6 through NIK's substrate IkB kinase a (IKKa) 7 . Non-canonical NF-kB signalling is controlled by TNF receptor associated factor (TRAF) proteins TRAF2 and NIK's negative regulator TRAF3, whereby a TRAF3-containing complex continuously targets NIK for degradation under steady-state conditions 5 . On receptor activation, TRAF3 is degraded and NIK protein levels can accumulate, allowing NIK to phosphorylate and activate downstream effectors.
To date, human patients carrying mutations in MAP3K14 have not yet been described. In Nik mutant mice (aly; alymphoplasia) 8,9 and knockout animals 3 , lymph nodes, Peyer's patches as well as splenic and thymic structures are severely disorganized. In addition, B-cell numbers are reduced and immunoglobulin (Ig) serum levels are decreased leading to humoral immunodeficiency. Although the function of NIK in B lymphocytes has been well established, the role of NIK-dependent signalling for T and natural killer (NK) lymphocytes is less well understood.
Here we report a combined immunodeficiency syndrome caused by biallelic mutations in the gene encoding NIK, encompassing B-cell lymphopenia and impaired memory B-cell differentiation. We also identify abnormal NK-cell development and function, as well as aberrant T-cell responses, indicating that biallelic loss-of-function mutations in NIK cause a hitherto unrecognized, pervasive combined immunodeficiency syndrome.

Identification of a homozygous mutation in MAP3K14.
We studied a large consanguineous pedigree with two patients (termed P1 and P2) who showed signs of combined immunodeficiency including recurrent, severe bacterial and viral infections and Cryptosporidium infection ( Supplementary Fig. 1a,b and Supplementary Tables 1 and 2; see Supplementary Note for further clinical course details). Investigation for known genetic aetiologies of defective CSR including CD40 and CD40L deficiencies and gain-of-function PIK3CD mutations 10,11 was performed; however, no mutation was identified.
Immunological assessment in both affected patients revealed decreased immunoglobulin levels (Supplementary Table 1) and decreased numbers of both B and NK cells, while T-cell numbers were within normal age-adjusted ranges (Supplementary Table 3).
As decreased immunoglobulin levels and B-cell numbers suggested impaired B-cell development and function, we performed flow cytometry-based immunophenotyping to assess the relative frequencies of CD27 þ memory B-cell populations. Both patients showed a relative reduction of total CD19 þ B cells in peripheral blood (Fig. 1a). Absolute blood cell counts revealed B lymphopenia in P2, while B-cell numbers in P1 were in the agematched lower normal range (Supplementary Table 3). Patients had decreased CD19 þ CD27 þ IgD þ marginal zone-like/innate B cells and CD19 þ CD27 þ IgD À class-switched memory B cells compared with controls 12 , suggesting defects in late stages of B-cell development and activation (Fig. 1a).
Given the consanguineous background, an autosomal-recessive inheritance mode was assumed. To unveil the presumed monogenetic cause of disease, single-nucleotide polymorphism (SNP) array-based homozygosity mapping of P1 and P2 ( Fig. 1b and Supplementary Table 4) was combined with exome sequencing (ES) (Fig. 1c) of P1. Single-nucleotide variants (SNVs) and insertion/deletion variants resulting from ES were filtered for those present inside homozygous candidate intervals shared between both affected patients. Synonymous and non-coding variants were excluded. We identified a single homozygous variant on chromosome 17q21 in MAP3K14 (c. C1694G, p. Pro565Arg) present in both patients ( Fig. 1d and Supplementary Fig. 1), which was not detected in dbSNPbuild137, 1000Genomes, ENSEMBL, UCSC, NCBI or EVS (Exome Variant Server) public SNP databases. The Pro565 residue of NIK is located within the kinase domain of the protein (Fig. 1e) and is highly conserved throughout evolution (Fig. 1f). The exchange from proline to arginine at this position was predicted as highly deleterious using the functional prediction algorithms Polyphen-2 and SIFT with maximum scores (1.0 and 0.0, respectively).
Effects of NIK Pro565Arg on kinase function. Pro565 forms part of the APE motif within a helix in the activation segment of the kinase 13 . This motif is conserved in NIK from various vertebrate phyla as well as in orthologous serine/threonine kinases (Fig. 1f). An exchange of a non-polar, conformationally rigid amino acid as in NIK Pro565Arg may have an impact on protein folding and function. The protein stability analysis tool CUPSAT predicted that the overall stability of NIK Pro565Arg may be compromised (Supplementary Table 5). Coarse-grained molecular dynamics simulation of NIK wild-type and NIK Pro565Arg showed conformational changes within the kinase domain. Notably, the nearby Thr559 residue has been reported to form a hydrogen bond with Lys517 in the catalytic loop (Fig. 2a) and mutation of Thr559 has been found to reduce kinase activity 14,15 . In the NIK wild-type simulation, Pro565 remains buried within the protein in the vicinity of the ATP-coordinating centre, allowing hydrogen bond formation between Lys517 and Thr559 (Fig. 2a). In the NIK Pro565Arg simulation, the mutated arginine residue transitions towards the protein surface, thereby increasingly contacting the surrounding solvent and repositioning adjacent helices. In the simulation, Arg565 prevents Thr559 from forming a hydrogen bond with Lys517 in the ATP-coordinating centre, thereby impairing the kinase activity of NIK (Fig. 2b,c).
To experimentally assess the effect of the mutation, we analysed the kinase activity of NIK Pro565Arg compared with NIK wild-type and the catalytically inactive mutant NIK Lys429Ala/Lys430Ala (ref. 16) by testing NIK-dependent phosphorylation of IKKa. Recombinantly expressed NIK wild-type , but not NIK Pro565Arg or NIK Lys429Ala/Lys430Ala , could phosphorylate both endogenous (Fig. 2d) and co-expressed IKKa ( Supplementary  Fig. 2)  represents a loss-of-function mutation with respect to abolished kinase activity towards its direct target IKKa.
Defective non-canonical and canonical NF-jB signalling. Processing of p100 into p52 and nuclear translocation of the p52/RelB complex is an essential step following NIK activation 6 . Therefore, we studied activity of the non-canonical NF-kB pathway in response to activation with BAFF and LTb, respectively. In patient-derived Epstein-Barr virus-immortalized lymphoblastoid cell lines (B-LCL), total NIK protein levels (which are tightly controlled via proteolysis 5 ) were unaffected (Fig. 3a). However, levels of its immediate downstream target IKKa were elevated and p100 accumulated already before BAFF-mediated BAFFR ligation (Fig. 3a), possibly reflecting pre-activation of the NF-kB pathway in B-LCL by viral proteins 17 . Despite p100 accumulation, p52 protein levels were decreased (Fig. 3a), resulting in a severely reduced nuclear content of p52 and a lower nuclear content of RelB in patient-derived cells (Fig. 3b), demonstrating functional insufficiency of the non-canonical NF-kB pathway in patient-derived cells.
Next, we tested the effect of NIK Pro565Arg in NF-kB activation upon LTb stimulation, which is strictly dependent on NIK 18 . LTb can activate non-canonical NF-kB signalling as well as the IKKa-IKKb-NEMO complex mediating nuclear translocation of Distance from lower telomere of Chr 17

Y F P K D L R G T E I Y M S P E V I L C R G H S T K A D I Y S L G A T L I H M Q T G T P 355 T G M K S V T G T P Y W M S P E V I S G E G Y G R K A D I W S V A C T V V E M L T E K P 569 . . Y T H E V V T L W Y R A P E I L L G C K Y Y S T A V D I W S L G C I F A E M V T R R 197
II-2 (Parent)  canonical NF-kB complexes 19 . Stimulation of patient fibroblasts with the NIK-independent canonical NF-kB activator TNFa led to a rapid decay of IkBa, a hallmark event of canonical NF-kB signalling (Fig. 3c). In contrast, patient cells failed to induce IkBa decay after LTb stimulation (Fig. 3d). Consistent with these findings, immunofluorescent staining showed defective nuclear translocation of both p52 and p50 in patient primary fibroblasts upon LTb stimulation (Fig. 4a,b).
To demonstrate the causative role of NIK Pro565Arg for deficient non-canonical NF-kB signalling, we performed retroviral-mediated gene transfer of MAP3K14 into patient fibroblasts. Expression of NIK wild-type reactivated non-canonical NF-kB signalling manifesting in nuclear translocation of p52 ( Fig. 4c,d), demonstrating that the presence of functional NIK protein is the limiting factor for p100 activation.
Although NIK may have IKKa-independent functions 20 , the main function of the protein comprises the catalytic activity as a kinase and activation of the signalling cascade leading to NF-kB translocation. As the NIK Pro565Arg mutant is catalytically inactive and therefore deleterious to these functions, the phenotype caused by NIK Pro565Arg is hereafter also referred to as functional NIK deficiency.
Reduced survival of mature B cells. The spontaneous Map3k14 mouse aly mutant 8,9 and Map3k14 knockout mice 3 show reduced mature B-cell numbers and decreased Ig serum levels, resulting in defects in both antibody and cellular immune responses. In addition, non-canonical NF-kB signalling mediated by Nik controls CSR, in particular to IgA isotype 21 . Accordingly, patients bearing NIK Pro565Arg had severely reduced total B-cell counts and impaired generation of CD27 þ IgD À class-switched memory B cells in the peripheral blood (Fig. 1a), leading to chronically reduced IgA titres in both patients and to reduced IgG levels in P1 at the age of 10 months (Supplementary Table 1) prompting intravenous immunoglobulin substitution. This led us to test whether NIK is involved in CSR and SHM, processes essential for the generation of high-affinity antibodies.
To study the occurrence of SHM in B cells, we analysed the mutation frequency in rearranged variable regions of the Ig heavy chain (IGHV) genes by cloning and sequencing the IGHV3 and IGHV4 rearranged gene families of both gand a-chain immunoglobulin transcripts (Cg and Ca). The percentage of mutations within the analysed IGHV regions was significantly reduced in NIK Pro565Arg -bearing patients compared with agematched healthy donors, although not as severely as in CD40Ldeficient patients who showed near-complete absence of mutations in IGHV Ca and completely lacked Cg transcripts ( Fig. 5a) 22 .
Next, we investigated activation and CSR capacity of NIK Pro565Arg B cells by stimulating peripheral blood mononuclear cells (PBMCs) with a range of stimuli. Patient B cells were able to respond to stimulation with CD40L and IL4 by upregulating the activation markers CD95 and CD69, as well as the costimulatory molecule CD86, although to a lesser extent than B cells from a healthy donor (Fig. 5b). Notably, patient B cells were only partially able to upregulate the activation marker CD25, suggesting impaired IL2-mediated survival and proliferation 23 . As NIK Pro565Arg B cells were largely able to upregulate the aforementioned activation markers, we further tested their proliferation capacity and ability to undergo CSR in response to CD40L and IL4 stimulation. Indeed, we observed a progressive increase in the percentage of DAPI (4 0 ,6-diamidino-2phenylindole)-negative blasts in NIK Pro565Arg mutant cells ( Supplementary Fig. 3a) over a course of 9 days, consistent with activation and proliferation of cells upon stimulation with CD40L and IL4 (ref. 24). Furthermore, B cells with mutated NIK underwent CSR to IgG in vitro, albeit with reduced frequency at day 6 when compared with controls ( Fig. 5c), probably due to delayed lymphocyte proliferation ( Supplementary Fig. 3a). Concomitant to proliferative outgrowth of lymphocytes at day 9, These data demonstrated the inability of NIK Pro565Arg B cells to respond to BCR, TLR9, IL21R and/or BAFFR stimulation. Only CD40L/IL4 stimulation, known to mediate survival and proliferation of primary B cells 25,26 , could induce cell proliferation ( Fig. 5d and Supplementary Fig. 3a), suggesting that intact NIK is required to relay signals essential for survival and proliferation of activated mature B-cell populations. Given the partial phenotypic overlap of functional NIK deficiency and IL21 (receptor) deficiency 27,28 (for example, colitis, susceptibility to Cryptosporidium infection, hypogammaglobulinemia and decreased frequencies of class-switched B cells, defective antigen-specific T-cell proliferation and impaired NKcell cytotoxicity), we sought to exclude an involvement of NIK in IL21-mediated signalling in B cells. Indeed, although classical signalling via STAT3 (signal transducer and activator of transcription 3) phosphorylation was readily observed after stimulating sorted B cells with IL21, no activation of p100 processing could be detected ( Supplementary Fig. 5).
As NIK is an integral component of the non-canonical NF-kB pathway downstream of the BAFFR, which plays a key role in mature B-cell survival 29 , we investigated whether functional NIK deficiency resembles phenotypes found in BAFFR deficiency 2,30 .
Similar to findings in Baffr À / À mice 30 , B cells of both patients showed lower cell surface expression of CD21, involved in prosurvival signalling on B cells 31 (Fig. 5e). This observation prompted us to investigate whether the NIK Pro565Arg mutant affects the expression of anti-apoptotic genes BCL2, BCL2L1 and MCL1 by quantitative reverse transcriptase-PCR in sorted naïve mature CD19 þ CD27 À IgD þ B cells. From the transcripts tested, BCL2 expression was markedly downregulated in patient naive B cells compared with heterozygous parent and healthy donor controls ( Fig. 5f and Supplementary Fig. 4b). To exclude that functional NIK deficiency ablated BAFFR expression, we analysed surface BAFFR levels by flow cytometry on PBMCs stimulated in vitro with CD40L and IL4 for 9 days. BAFFR expression on B cells from P1 was comparable to B cells from control parent or healthy donor (Supplementary Fig. 4c).
BAFFR-deficient patients display a partial block in development beyond the transitional CD19 þ CD21 low/intermediate B-cell stage 2 . As NIK Pro565Arg patient peripheral B cells had overall a CD19 þ CD21 low/intermediate phenotype, we tested the expression of the alternative transitional B-cell markers IgM, CD10, CD38 and CD5. Transitional B cells (defined as CD19 þ IgM hi IgD þ and CD19 þ CD23 À CD27 À CD5 þ IgM hi , respectively) were increased, in particular the CD19 þ CD38 þ CD10 À transitional T2 population, indicating a partial block in B-cell maturation ( Supplementary Fig. 4d,e). Taken together, the NIK Pro565Arg lossof-function mutant causes a partial block of B-cell development between transitional and naive mature B-cell stages accompanied by impaired survival of mature peripheral B cells. ARTICLE Aberrant T-cell phenotype and antigen-specific proliferation. Next, we assessed the effect of NIK Pro565Arg on T cells. Overall CD3 þ CD4 þ helper T-cell and CD3 þ CD8 þ cytotoxic T-cell subset distribution was unaffected ( Supplementary Fig. 6a), T-cell receptor Vb repertoires were polyclonal ( Supplementary Fig. 7) and regulatory T cells were unaltered ( Supplementary Fig. 6b).
Upon stimulation of PBMCs with T-cell proliferation stimuli such as anti-CD3 antibody (clone OKT3), phorbol 12-myristate 13-acetate, Staphylococcal enterotoxin A, Staphylococcal enterotoxin B or phytohaemagglutinin, normal proliferative responses were observed (Fig. 6a). In contrast, when the antigen-specific stimuli tetanus toxoid or purified protein derivatives of Mycobacterium tuberculosis were used, proliferative responses were severely reduced (Fig. 6b) despite prior tetanus and Bacillus Calmette-Guérin vaccination. These observations prompted us to assess the presence of naive and memory T-cell subsets in both patients. The relative proportions of CD4 þ effector memory T cells (T EM ) (markers CD4 þ CD45RA À CCR7 À or CD4 þ CD45RA À CD27 À , respectively) were comparable between patients and healthy controls and were within the normal agematched range 32 ( Supplementary Fig. 6c,d and Supplementary Table 3). Relative numbers of CD8 þ memory T cells from both patients were also within the highly variable normal range 32 . However, P1 exhibited a remarkable expansion of CD8 þ T EM cells and terminally differentiated effector memory T cells (T EMRA , identified as CD8 þ CD45RA þ CD27 À or CD8 þ CD45RA þ CCR7 À ; Supplementary Fig. 6c,d), possibly attributable to the persistent cytomegalovirus (CMV) viremia in P1 (Supplementary Table 2). Increased IL7R/CD127 expression on CD8 þ T cells identifies long-lived memory cells 33,34 . Similar to Nik-deficient mice 35 , we found dramatically reduced CD127 expression on CD8 þ memory T cells, on CD8 þ T CM and on CD8 þ T EM in both P1 and P2 (Fig. 6c), pointing towards impaired memory responses to viral infections observed in P1 ( Fig. 6b and Supplementary Table 2). Interaction of inducible co-receptor ICOS with ICOS ligand (ICOSL) is important for the differentiation of follicular helper T FH cells and for memory responses of both T and B cells 36 . T FH cells localize to GC reactions within secondary lymphoid organs where they interact with B cells to aid antibody production and maturation 37 . T FH cell numbers are reduced in ICOS-deficient common variable immunodeficiency and in CD40L or CD40 deficiency 37 . Indeed, we found decreased proportions of T FH cells (identified as CXCR5 þ CD45RA À ) in both patients compared with an age-matched healthy donor (Fig. 6d) and with previously reported healthy donors 38 . As T FH cell development is dependent on ICOSL expression on B cells which is controlled by noncanonical NF-kB signalling 39 , we hypothesized that NIK Pro565Arg causes reduced ICOSL expression leading to impaired T FH generation. Thus, we stimulated PBMCs with CD40L for 36 h and monitored ICOSL expression by flow cytometry. Patient, control parent and healthy donor B cells responded to CD40L stimulation by inducing CD69 expression; however, although control cells were able to upregulate ICOSL, the patient cells failed to do so (Fig. 6e). In summary, NIK Pro565Arg patients exhibit defective differentiation into T FH and impaired function of memory T-cell subsets.

Decreased numbers and functional impairment of NK cells.
Consistently low NK-cell numbers in both patients (Supplementary Table 3) along with the susceptibility to CMV, for which NK-cell-mediated defense is relevant, prompted a detailed phenotypic analysis of NK cells. Both patients bearing the NIK Pro565Arg mutant had low numbers of NK cells in the peripheral blood, particularly P1 (Fig. 7a). Despite their low frequency, NK cells showed normal expression of the cell-surface markers perforin, CD16, CD69, CD57 and NKG2C, hallmarks of acquisition of cytotoxic function (Supplementary Fig. 8). Both CD56 bright and CD56 dim NK cells, representing subsequent stages of NK cell development, were present ( Fig. 7a and Supplementary  Fig. 8). In addition, patients expressed markers associated with pre-terminal NK-cell developmental stages, including CD117, CD27, CD11a, KIR2DL4 and CD94, at levels comparable to healthy donors ( Supplementary Fig. 8). The notable exception was CD62L, which was expressed by a markedly lower proportion of NK cells with NIK Pro565Arg (Fig. 7b).
To determine whether the NK cells present were able to exert cytolytic function and cytokine secretion, we performed activation with phorbol 12-myristate 13-acetate and ionomycin. Notably, production of interferon-g and TNFa were markedly reduced in patient NK cells, in comparison with stimulated healthy donor cells (Fig. 7c). Although patient NK cells expressed comparable levels of perforin compared with  Supplementary Fig. 8), they degranulated at a significantly lower frequency as measured by the expression of CD107a (LAMP1) on the cell surface after stimulation (Fig. 7c).
To further define the cytolytic potential of these cells, we evaluated key components of cytotoxicity by quantitative confocal microscopy. Patient-derived NK cells failed to accumulate F-actin at the immunological synapse following incubation with susceptible targets (Fig. 7d,e). In addition, lytic granules failed to polarize to the lytic synapse (Fig. 7e). In concert with the flow cytometric analyses, these data suggest a marked inability of NIK Pro565Arg NK cells to become activated and exert cytolytic function.

Discussion
TNFa receptor family signalling is essential for B-cell immunity in humans as illustrated by deficiencies in CD40L, CD40 and BAFFR 40 . NIK is an integral component of the non-canonical NF-kB pathway downstream of these receptors 5 .
Studies in the mouse aly mutant 8,9 and Nik knockout mice 3 described B-cell deficiency due to disorganized lymph nodes, Peyer's patches and splenic architecture, accompanied by B-cell lymphopenia and low serum Ig levels due to compromised CSR and SHM 41 . We here identify patients with biallelic mutation in NIK, leading to loss-of-function of the kinase function of NIK. We show that human functional NIK deficiency recapitulates phenotypes described in the mouse studies including B-cell lymphopenia, impaired CSR and SHM, decreased marginal zone and memory B cells, and hypogammaglobulinemia. Although ethic considerations prevented us from obtaining patient biopsies to further investigate secondary lymphoid organ structures, the absence of lymph nodes on repeated clinical examinations suggests that secondary lymphoid organs may be disturbed on loss-of-function of NIK, similar to the observations in mouse models 3,8,41 .
To extend the murine studies on NIK function in B cells, we investigated the survival properties of patient peripheral blood B cells. In vitro stimulation of BAFFR together with BCR and TLR9 resulted in absence of B cells, suggesting a profound survival defect. This is supported by earlier studies showing that Nik overexpression or expression of the NikDT3 mutant in mice (resistant to Traf3-mediated degradation) leads to increased survival of B cells 42 . Here we find significantly decreased expression of the anti-apoptotic gene BCL2 in peripheral NIK Pro565Arg B cells, leading to reduced survival. This is most probably the result of impaired BAFF signalling, as NIK is an integral molecule downstream of BAFFR required for B-cell survival 2,29,30 . Recently, a common variant in BAFFR has been shown to modulate NF-kB signalling albeit without effects on survival and subset composition of B cells 43 . Although we identified this variant in our patients, its presence could not explain the B-cell defects described in this study. BAFFR signalling also activates the expression of the B-cell maturation marker CD21 on transitional B cells 2,30,44 . CD21-deficient mice display severely impaired GC B-cell development and T-celldependent B-cell responses 45 due to reduced GC B-cell survival 46 . CD21 deficiency in humans leads to reduced class-switched memory B cells and hypogammaglobulinemia 47 . Therefore, the reduced CD21 expression on peripheral patient B cells we observed may provide an additional explanation for B-cell survival defects in NIK Pro565Arg patients.
Stimulation with CD40L and IL4, a cytokine with potent antiapoptotic activity mediated by Stat6-dependent upregulation of Bcl-xL 26 , led to CSR with nearly normal frequency, but delayed kinetics, indicating that at least a proportion of patient B cells was responsive to these stimuli, and that the CSR process itself is largely functional. As CD40 stimulation can signal both via canonical and non-canonical NF-kB pathways leading to AID expression and Igg germline transcription 48 , our findings indicate that CD40L-mediated, NIK-independent NF-kB signalling contributes to CSR. Delayed CSR kinetics might also be explained by an increased proportion of transitional B cells in patient PBMCs, which reacted more slowly to the B-cell activation stimuli, similar to BAFF-deficient B cells 2 .
As the clinical phentoype suggested a combined immunodeficiency and because recent studies have focused on the role of NIK in T cells 35,49,50 , we aimed at investigating T-cell functions and T-cell interplay with B cells in patients carrying NIK Pro565Arg . Previous studies indicate that NIK-dependent NF-kB signalling is required for ICOSL expression on activated B cells, directing T FH differentiation via interaction with ICOS 39  ARTICLE leads to the formation of GCs, structures essential for generation of high-affinity antibody responses. Abrogated ICOSL upregulation on B cells and reduced T FH cells in the here described patients probably contribute to impaired GC formation; however, in situ analysis of GCs were precluded by ethical constraints. Another lymphocyte communication process dependent on CD40 signalling via CD4 þ T-cell help is CD8 þ T-cell memory maintenance. Subsequent stimulation of IL7R expression is characteristic and essential for CD8 þ memory T-cell survival 4 . Therefore, lack of IL7R expression on both Nik-deficient mouse 35 and human (this study) CD8 þ memory T cells may contribute to the inability of T cells with NIK Pro565Arg to respond to tetanus toxoid and tuberculin despite prior vaccination. Recently, data from ICOSL-deficient patients has implicated ICOSL costimulation in maintenance of memory populations 51 , possibly providing an additional explanation for memory defects in patients carrying NIK Pro565Arg .
Patients exhibiting defects in canonical NF-kB signalling such as NEMO 52,53 or IKKb deficiency 54 and IkBa hypermorphism 55 also show antibody deficiencies. Consistent with multiple roles of canonical NF-kB signalling, they show pleiotropic defects throughout the adaptive and innate immune system and developmental defects. Patients with heterozygous mutations in NFKB2/p100 have recently been described with B-cell deficiency and autoimmunity 56 . Although showing similar manifestations, functional NIK deficiency is more severe than the heterozygous NFKB2 mutation in humans (this study) and mice 5 . This may be due to the increased p100 levels in NIK Pro565Arg B-LCL that we detected, potentially caused by compensatory increase of NIKindependent canonical NF-kB signalling, which can upregulate p100 expression 19 . In line with this, unprocessed p100 is known to specifically sequester and inhibit RelB 5 , leading to a severe signalling defect.
CD40-dependent IL12 secretion by monocytes is crucial in the defense against mycobacteria. The disseminated Bacillus Calmette-Guérin infection observed in P1 illustrates mycobacterial susceptibility similar to NEMO 57 and IKKb-deficient patients 54 . As NIK can signal via both non-canonical and canonical pathway, NF-kB response to mycobacteria may depend on NIK, suggesting that functional NIK deficiency causes defective CD40 signalling in monocytes as well.
Recently, impaired NK-cell function has been recognized for several primary immunodeficiencies (reviewed in ref. 58) most often manifesting in susceptibility to infection by herpesviruses (for example, CMV). NK-cells were persistently decreased in patients carrying NIK Pro565Arg . Interestingly, CD62L expression, a marker linked to mature NK-cell subsets with stronger cytolytic functions 59 , was downregulated on patient NK cells. Indeed, although they acquired appropriate developmental and maturity markers, including perforin, patient cells failed to become activated as shown by impaired cytokine secretion, degranulation and polarization of lytic granules towards targets. NIK has not been linked to NK-cell function so far. As canonical NF-kB signalling has been implicated in NK-cell activation 60 , these signals may depend on NIK and therefore mirror NK-cell phenotypes such as those found in CD40L or NEMO deficiencies.
Interestingly, despite the potential defects in secondary lymphoid organ organization discussed above, allogeneic haematopoietic stem cell transplantation (aHSCT) improved the overall condition of patient P1, suggesting that functional NIK deficiency is amenable to aHSCT to at least partially correct the disease. The fatal outcome for P2 (who had received aHSCT without prior conditioning and succumbed shortly after a second aHSCT including conditioning) suggests that at least reduced-intensity conditioning is indicated for successful treatment using aHSCT.
In sum, we identify functional NIK deficiency as a novel, pervasive combined primary immunodeficiency syndrome. Our data revealed an unexpectedly broad range of phenotypic aberrations affecting B-, T-and NK-lineages, and thus highlight essential roles for NIK and adequate control of non-canonical NF-kB signalling for generation and maintenance of the human immune system.

Methods
Patient and ethics. This study has been approved by the ethics committee at the Medical University of Vienna, Austria. Biological material was obtained on informed consent in accordance with the Declaration of Helsinki. Clinical data from the patients was provided in anonymized form by the responsible physician(s). The patients were evaluated, followed up and treated at the Department of Immunology at Ankara University in Turkey, covered by the local ethics agreement.
Homozygosity mapping. Affymetrix 6.0 SNP-based homozygosity mapping was performed in both patients and in both parents of P1, to map homozygous intervals common to both patients but not present in the parents.
Ten microlitres of 50 ng ml À 1 DNA from the patients were used. The protocol was carried out according to the Affymetrix Genome-Wide Human SNP Nsp/Sty 6.0 protocol. The results were analysed using the Affymetrix Genotyping Console software and PLINK whole genome data analysis toolset (http://pngu.mgh. harvard.edu/Bpurcell/plink/) as previously described 61 . The data analysis was carried out as previously described 62 using Burrows-Wheeler Aligner to align the reads to the human genome 19. Insertion/deletion realignment was performed as well as GATK (Genome Analysis Toolkit) base quality score recalibration. For SNV and insertion/deletion calling, Unified Genotyper and GATK Variant quality score recalibration was performed. SNV and insertion/deletion lists were uploaded to SeattleSeq Annotation database. Known variants (present in 1000Genomes or dbSNP build 137, date of accession: 2 January 2012) were excluded and the lists were filtered for nonsense, missense and splicesite variants present within the homozygous regions detected in both patients and absent in the parents. In addition, ENSEMBL, UCSC, NCBI and EVS public SNP databases (date of accession: 20 February 2012) were interrogated for presence of the variant.
The validation and segregation of the variants found in the final hit list from ES were performed using capillary sequencing on genomic DNA from both patients and family members as described below.
Variant validation by capillary sequencing. Primers for the variants detected with whole ES were designed using ExonPrimer software from the Helmholtz Center Munich (http://ihg.gsf.de/ihg/ExonPrimer.html) and PrimerZ 63 , respectively, and ordered from Sigma Aldrich, Austria. PCR amplification of the detected variants was performed using Expand High Fidelity PCR System (Roche, Basel, Switzerland).
Capillary sequencing of amplicons was performed on the Applied Biosystems 3130xl Genetic Analyzer capillary sequencer running 3130xl Genetic Analyzer Data Collection Software v3.0, using Big Dye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Germany). Sequence Analysis Software Version 5.2 was used for analysis of the sequences and heterozygous signals with ambiguity code were indicated when 25% of the signal intensity was exceeded. Reads were aligned to reference sequences using the Sequencher software, version 4.10.1.
In silico analyses and modelling of NIK protein structure. The algorithms SIFT 64  To obtain a hypothesis about the change in protein structure and dynamics of the NIK variant compared with the wild-type, molecular dynamics (MD) simulations of NIK wild-type and NIK Pro565Arg were performed using the coarsegrained model FREADY 67 implemented in MOIL 68 molecular modelling package. We initiated the MD simulations from the crystal structure of NIK (PDB ID 4G3D, chain A 14 ) and let it run for 50 ns at 300 K. In the MD simulations, residues farther than 9 Å from the mutated residue were fixed to the experimental structure. Protein structures were aligned and visualized using MacPyMol (The PyMOL Molecular Graphics System, Version 1.3 Schrödinger, LLC).
T and NK cells were evaluated using around 1 Â 10 6 PBMCs. The analysis of the B-lymphocyte compartment was performed using around 4 Â 10 6 cells as previously described 2,69,70 . NK cells were stained as described previously 71 . For intracellular staining, PBMCs were activated for 3 h with 50 mg ml À 1 ionomycin and 5 mg ml À 1 phorbol myristate acetate in the presence of Brefeldin A and antibodies to CD107a and cell surface markers 72 . Cells were then fixed and permeabilized with Cytofix/Cytoperm solution (BD Biosciences).
All analyses were performed using FlowJo X (TreeStar Inc.) and data were graphed with Prism 6.0 (GraphPad Software).
Flow cytometry-based sorting of peripheral B-cell populations stained with anti-CD19-PerCP-Cy5.5, anti-CD27-BV421, anti-CD3-APC-H7 and anti-IgD-FITC as described above was performed on ultra-high-speed six-way digital cell sorter from Beckmann Coulter at the Medical University Vienna Flow Cytometry Core Facility.
Quantitative real-time PCR analysis. Extraction of RNA from sorted B cells was performed using RNeasy kit from Qiagen, first-strand complementary DNA synthesis was done using Expand Reverse Transcriptase from Roche using both oligo-dT and random hexamer primers, and gene expression was analysed by quantitative PCR using Kappa Sybr Fast qPCR MasterMix ABI Bioprism from Kappa Biosystems on 7900HT Fast Real-Time PCR System from Applied Biosciences according to manufacturers' instructions.
Primary fibroblasts and HEK293 cells were cultured in glucose-rich DMEM (PAA), supplemented and cultured as above.
Amplification and sequence analysis of IGH transcripts. IgA and IgG transcripts were amplified from cDNA of thawed PBMCs using subgroup-specific forward primers in the leader sequence of IGHV3 and IGHV4 in combination with a Ca or Cg consensus reverse primer 22,74 . All PCR products were cloned into the pGEM-T easy vector (Promega) and prepared for sequencing on an ABIPRISM 3130xl (Applied Biosystems). Obtained sequences were analysed with the IMGT database (http://www.imgt.org/) for IGHV, IGHD and IGHJ use, and mutation analysis 75 .
T-cell proliferation analysis. T-cell proliferation assays were carried out as described previously 62,76 .
Sequences were acquired using an ABI 3130xl Sequencer (ABI Applied Biosystems) and analysed using GeneMapper software version 4.0.
Reconstitution assay. cDNA encoding for wild-type human MAP3K14 was cloned into a bicistronic retroviral pMMP vector coexpressing MAP3K14 and enhanced green fluorescent protein (eGFP) marker gene via IRES sequence. RD114-pseudotyped retroviral particles were generated by transfection into HEK293 cells using the calcium chloride transfection method (8 mg retroviral vector DNA, 12 mg gag/pol DNA, 5 mg RD114 DNA) in the presence of 25 mM chloroquin (Sigma-Aldrich, C6628) for 12 h. Supernatants containing viral particles were collected after 24, 36 and 48 h. Viral titration was performed on HT-1080 cells. Patient and normal donors fibroblast cells were transduced with retroviral particles in the presence of 8 mg ml À 1 polybrene (Santa Cruz, sc-134220) for 12 h. Transduction efficiency was determined by eGFP expression by FACS analysis and was between 45% and 70%. After transduction, immunofluorescence studies were performed as above with additional staining against GFP used at a dilution of 1:100 (antibody sc-69779, Santa Cruz). Slides were visualized as above. Data was graphed using Prism 6.0 (GraphPad Software).
Confocal microscopy of immunological synapses. Cell conjugates were formed for the evaluation of the immunological synapse by confocal microscopy as previously described 78 . Following fixation and permeabilization, conjugates were incubated with anti-perforin Alexa Fluor 488 (Biolegend) at a dilution of 1:50 and Phalloidin Alexa Fluor 568 at a dilution of 1:100. Images were acquired on a Leica SP8 laser scanning confocal microscope. Excitation was by tunable white light laser and detection of emission by hybrid gallidium (HyD) detectors. Acquisition was controlled by Leica LASAF software and images were subsequently exported to Volocity software (PerkinElmer) for analysis. Data were graphed using Prism 6.0 (GraphPad Software).