Selective predisposition to bacterial infections in IRAK-4–deficient children: IRAK-4–dependent TLRs are otherwise redundant in protective immunity

We report 28 patients with IRAK-4 deficiency. The patients originate from 18 unrelated kindreds and 11 countries (Table I and Fig. 1). All IRAK4 exons, flanking intron regions, and, when appropriate, entire introns, were sequenced in 24 patients (P1–4, 6–13, 15, 17–20, and 22–28). IRAK-4 deficiency was diagnosed on clinical grounds in four deceased relatives (P5, 14, 16, and 21) for whom no biological material was available. The patients of 13 kindreds were apparently homozygous (kindreds A–C, E, F, H–L, and P–R), and those from 5 kindreds were compound heterozygous (D, G, and M–O) for IRAK4 mutations. However, four seemingly homozygous patients from three unrelated families (P2 from kindred B, P7 from kindred F, and P11 and 12 from kindred I) had one parent who did not carry the mutant allele. Fluorescence in situ hybridization with BAC210N13, which covers the entire IRAK4 locus, and the genotyping of polymorphic markers showed that P2 was heterozygous for a large de novo deletion (designated BAC210N13del) encompassing IRAK4 (Fig. S1, top; and not depicted). For P7, using the same BAC as for P2, fluorescence in situ hybridization revealed two signals, consistent with homozygosity owing to segmental uniparental disomy or compound heterozygosity with an undetected deletion encompassing a fraction of IRAK4 (Fig. S1, bottom; and not depicted). Not enough material was available to explore the IRAK4 locus in the deceased patients P11 and 12 from kindred I (8). 3 out of the 14 mutant alleles identified carried nonsense mutations (Y48X, Q293X, and E402X) (1, 3, 4, 6, 8, 9, 11, 36), 3 carried large deletions (1-1096_40+23del, BAC210N13del, and 942-1481_1125+547del), 2 carried splice mutations (1188+520A>G and 1189-1G>T) (12), and 6 carried frameshift insertions and deletions (167_172insA, 573delA, 620_621delAC, 631delG, 821delT, and 1240insA) (1, 4, 7, 34) (Table I and Fig. 2 A). All mutations are predicted to be null, as they create a premature termination codon or delete a large segment of the gene. No missense mutation was found. The 14 mutations were not found in 100 healthy controls sequenced. The Q293X mutant allele was found in homozygotes from six kindreds (C, H, K, P, Q, and R) and compound heterozygotes from four kindreds (B, D, M, and possibly F). The recurrence of this mutation may reflect a mutational hotspot, a founder effect, or both (unpublished data).

We assessed IRAK4 mRNA levels in EBV-transformed B lymphocyte cell lines (B-EBVs; Fig. 2 B) derived from most patients and a healthy control by RT-PCR. The two patients carrying the 573delA mutation died before cell lines could be established (34). Most other patients lacked detectable full-length IRAK4 mRNAs species, presumably because of nonsense-mediated mRNA degradation. However, P7 (mutation Q293X), P8 (mutations 1188+520A>G and 1189-1G>T), P13 (mutation E402X), P19 (mutation 167_172insA), and P22 (mutation Q293X/620-621del) had low levels of detectable full-length IRAK4 mRNA. We then assessed IRAK-4 protein levels in B-EBVs (Fig. 2 C). No IRAK-4 protein was detected in any of the patients tested, even in P7, 8, 13, 19, and 22, all of whom had detectable full-length mRNAs, excluding a potential role of IRAK-4 as a scaffold protein in our patients (40, 41). Finally, we assessed the functional impact of IRAK4 mutations. B-EBVs bearing mutations 821delT (P1), Q293X (P2, 3, and 7), 1188+520A>G/1189-1G>T (P8), E402X (P13), and 1-1096_40+23del (P15) did not respond to TLR7 and 8 agonists, as measured by TNF-α production (Fig. 3 A). SV40-transformed fibroblasts (SV40-fibroblasts) bearing mutations 821delT (P1), Q293X (P2 and 3), 1188+520A>G/1189-1G>T(P8), E402X (PI3), 1-1096_40+23del (P15), Y48X/631delG (P23), and 1240insA/942-1481_1125+547del (P24) did not respond to IL-1β, as assessed by measuring IL-6 production. However, IRAK-4–deficient SV40-fibroblasts did produce IL-6 upon activation by poly(I:C) (Fig. 3 B) (13). Thus, all patients had complete IRAK-4 deficiency and a complete absence of IRAK-4–dependent TIR signaling, owing to the inheritance of two loss-of-expression, loss-of-function IRAK4 alleles.

We analyzed blood leukocyte subsets in 12 IRAK-4–deficient patients. We previously showed that granulocytes, CD14⁺, CD16⁺, and CD14⁺/CD16⁺ monocyte subsets, and MDCs and PDCs, were present in normal numbers in three patients (13). We now report that T cell subsets, including CD4⁺ and CD8⁺, and CD45RA⁺ and CD45RO⁺ T cells, are also present in normal numbers (Table S1), with the possible exception of normal to low levels of T cells in P17 and 18 (36). T cells proliferated normally in response to the mitogen PHA, CD3, and recall antigens in vitro (Table S2). B cells and memory B cells (CD27⁺) were also present in normal numbers (Table S1). Serum Ig levels for IgA were normal in five, high in two (P8 and 11), and low in four (P1, 2, 17, and 18) patients (36). IgG levels were normal in seven and high in four (P7, 8, 11, and 17) patients, and IgM levels were normal in seven, high in three (P7, 11, and 19), and low in one (P2) patients. IgE levels were high in 8 (P1, 7, 8, 11, 13, 15, 17, and 23) out of the 11 patients evaluated (Table S2). Antibody responses to protein antigens were normal in all but two patients, who had slightly low titers (P7 and 15); however, the date of recall vaccination before serological testing was unknown. The antibody response to glycans was impaired in some (P2, 8, 17, 18, and 29) but not all patients, and in response to some but not all pneumococcal and erythrocyte AB antigens (Table S2 and unpublished data) (11, 12, 33). Finally, the surface expression of CD16 and CD56 on NK cells was normal (Table S1). IFN-γ secretion and surface expression of CD107 (degranulation) by the patients' NK cells were normal (unpublished data). Overall, there seemed to be no overt defect of leukocyte development in IRAK-4–deficient patients. Thus, antigen-specific T and B cell responses seemed to be normal, except for an impaired glycan-specific antibody response in at least some patients and against some glycans, and except for an overproduction of IgE in most of the patients tested.

We previously reported that IRAK-4–deficient whole blood cells and PBMCs produce only very small amounts of TNF-α, IL-6, IL-12, G-CSF, GM-CSF, and IFN-γ in vitro in response to all IL-1R and TLR agonists tested (1–9, 11, 12). We wondered whether the induction of other cytokines, chemokines, IFNs, and growth factors was also dependent on IRAK-4 after TLR stimulation. We therefore activated PBMCs from IRAK-4–deficient patients with Pam₃CSK₄ (TLR1/2), Pam₂CSK₄ (TLR2/6), poly(I:C) (a nonspecific TLR3 agonist), LPS (TLR4), flagellin (TLR5), 3M-13 (TLR7), 3M-2 (TLR8), R-848 (TLR7 and 8), and CpG (TLR9) for 24 h. We did not assess TLR10 responses, as there is no known agonist for this receptor (23). Cytokine secretion into the supernatant was assessed using a multiplex cytometry-based system. 11 out of the 25 cytokines assayed were induced and detectable after TLR stimulation in healthy controls. IRAK-4–deficient cells did not respond to seven out of nine agonists for all cytokines tested (Fig. 4). Upon activation with poly(I:C), the patients' PBMCs displayed induction of IL-12, monocyte chemoattractant protein 1, and macrophage inflammatory protein 1β (MIP-1β) to levels similar to those in healthy controls, as well as some induction of IFN-inducible protein 10 (Fig. 4). However, the induction of IL-12 and MIP-1β was weak in both patients and healthy controls (Fig. S2). IL-7 induction was abolished in the patients, whereas other cytokines were not induced in controls. The patients' PBMCs showed detectable IL-8 and MIP-1β (an IFN-inducible cytokine) responses to LPS, but these responses were weaker than those of healthy controls (Fig. 4). The other cytokines were not induced in the patients. These data are reminiscent of our previous observation that IRAK-4–deficient PBMCs respond to poly(I:C) by producing IFN-α protein, and to poly(I:C) and LPS by producing IFN-β mRNA (13). However, whereas LPS responses can be specifically ascribed to TLR4, we recently showed, in TLR3-deficient patients, that the poly(I:C) responses of PBMCs are TLR3-independent (42). These data indicate a broad immunological impact of IRAK-4 deficiency, as the production of 11 key cytokines was completely impaired in response to all TLR agonists, with the exception of a couple of cytokines in response to poly(I:C) and LPS.

We then assessed the role of IRAK-4 in TLR signaling pathways in discrete leukocyte cell populations. Cell subsets other than granulocytes and DCs were purified by cell sorting (purity >99.5%). More than 95% of the granulocytes purified on Ficoll were CD15⁺. The response of DCs (MDCs and PDCs) was tested in PBMCs. We assessed the CD62L shedding of granulocytes from four healthy controls and four IRAK-4–deficient patients after activation with Pam₃CSK₄, Pam₂CSK₄, LPS, flagellin, 3M-13, 3M-2, R-848, and TNF-α (10). The response to all TLR agonists was impaired in the granulocytes of all four patients tested (Fig. 5 A). CD14⁺ monocytes from healthy controls responded to TLR1–8 agonists but not to TLR9 agonists. The monocytes of IRAK-4–deficient patients did not respond to these agonists, with the possible exception of very weak TNF-α production upon LPS stimulation (Fig. 5 B). Finally, we tested MDCs and PDCs by stimulating PBMCs from seven healthy donors and three IRAK-4–deficient patients with the TLR agonists Pam₃CSK₄, Pam₂CSK₄, poly(I:C), LPS, flagellin, 3M-13, 3M-2, R-848, and CpG for 3 h. We assessed TNF-α and MIP-1β production for MDCs (Lin-1⁻, HLA-DR⁺, and CD123^low) and PDCs (Lin-1⁻, HLA-DR⁺, and CD123^high) by intracellular staining. In healthy individuals, MDCs responded to all of the TLR agonists tested, except the TLR9 agonist, with the induction of TNF-α and MIP1-β. In contrast, only upon activation with poly(I:C) (nonspecific TLR3 agonist) and LPS (TLR4), did MDCs from the patients display normal levels of MIP1-β induction and some induction of TNF-α. PDCs from healthy individuals responded only to agonists of TLR7 and 9, whereas IRAK-4–deficient PDCs did not respond to any of the agonists tested (Fig. 5, C–F). As poly(I:C) activation in MDCs appears to be TLR3 independent (42), we further evaluated the production of TNF-α and the up-regulation of IFN-inducible surface-expressed CD40, CD80, and CD86 by in vitro MDDCs, which respond to poly(I:C) in a TLR3-dependent manner (42). MDDCs from healthy controls responded normally to the TLR agonists Pam₃CSK₄, Pam₂CSK₄, poly(I:C), LPS, flagellin, and 3M-2. In contrast, the patients' MDDCs did not respond to Pam₃CSK₄, Pam₂CSK₄, flagellin, and 3M-2. However, IRAK-4–deficient MDDCs showed a weak but not abolished TNF-α response and normal induction of CD40, CD80, and CD86 upon activation with poly(I:C) (TLR3). Normal induction of CD40, CD80, and CD86 was also observed upon activation with LPS (TLR4) (Fig. 5, G and H). These data indicate that the IRAK-4–deficient individual myeloid cell subsets tested displayed no response to most TLR agonists, with the exception of normal responses to poly(I:C) and LPS detected in MDCs for MIP-1β, an IFN type I–inducible cytokine, and in MDDCs for CD40, CD80, and CD86, which are induced by type I IFNs and TNF-α.

We then tested the TLR responses of the B, T, and NK lymphoid cell subsets. The subsets were purified by cell sorting (purity >99.5%). CD19⁺ B cells were activated by incubation with the TLR agonists Pam₃CSK₄, Pam₂CSK₄, poly(I:C), LPS, flagellin, 3M-13, 3M-2, R-848, and CpG for 24 h, and their response was measured by assessing IL-10 production. Highly purified control B cells showed a unique pattern of activation, with no response to agonists of TLR1/2, TLR2/6, TLR3, TLR4, TLR5, and TLR8, and only weak IL-10 production in response to TLR7, TLR7 and TLR8, and TLR9 agonists (Fig. 6 A and not depicted). In contrast, no response to these TLR agonists was observed in the three IRAK-4–deficient patients tested (Fig. 6 A). Moreover, the response to TLR7 and 9, as measured by cell-surface expression of CD40, CD80, and CD86 after 3 d of incubation with IL-4 and various TLR agonists, was also impaired in the patients' B cells (Fig. 6 B) (13). CD3⁺ T cells from healthy individuals were activated by Pam₃CSK₄, Pam₂CSK₄, poly(I:C), LPS, flagellin, 3M-13, 3M-2, R-848, and CpG. Control T cells displayed a weak but detectable response to Pam₃CSK₄ and flagellin in terms of IFN-γ production, whereas T cells from IRAK-4–deficient patients were not activated by any of the TLR agonists (Fig. 6 C). Finally, control NK cells were shown to respond to TLR3, 7, and TLR7 and 8 agonists in terms of IFN-γ production, but no response was observed in NK cells from IRAK-4–deficient patients (Fig. 6 D). NK cells respond to poly(I:C) through TLR3 (42), suggesting that at least some TLR3 pathways are IRAK-4 dependent. These data indicate that the three major blood lymphoid subsets require IRAK-4 for TLR responses, including TLR3 responses in NK cells.

In total, 28 IRAK-4–deficient patients from 18 families were studied, including the 7 patients (P21–27) from 5 families described in this study for the first time (Table I and Fig. 1). Most IRAK-4–deficient patients had had at least one Gram-positive bacterial infection: 22 out of the 28 (79%) had had invasive disease caused by S. pneumoniae (meningitis, septicemia, or arthritis), and 9 out of the 28 (32%) had suffered severe disease caused by S. aureus (meningitis, septicemia, or liver abscess; Table I). If we also take into account peripheral staphylococcal disease (cellulitis and subcutaneous abscess), 14 patients could be considered particularly susceptible to S. aureus. One patient (P20) had had no major infectious disease. This patient is 25 mo old and was diagnosed with IRAK-4 deficiency as a neonate. He was placed on IgG substitution and antibiotic prophylaxis shortly after birth. Seven patients also suffered from severe Gram-negative bacterial infections, which were invasive in four cases (Shigella sonnei and P. aeruginosa) and peripheral in four cases (Escherichia coli, Serratia marcescens, Neisseria meningitidis, and P. aeruginosa). As previously reported in a smaller series (13), no severe viral, fungal, or parasitic infections were observed in the patients. Most patients developed their first invasive infection before the age of 2 yr (20 out of 28; 71%), often before the age of 6 mo (9 out of 28; 32%) and in the neonatal period (4 out of 28; 14%), when maternal antibodies are still present. Remarkably, no invasive infection was documented in the six patients over the age of 14 yr (P2, 14 yr; P4, 24 yr; P7, 32 yr; P17 and 18, 27 yr; and P24, 16 yr), even in the absence of prophylaxis (P2, 4, 7, 17, and 18; n = 5; Fig. 7 A) (4, 6, 36). 12 patients died of invasive Gram-positive infections, all before the age of 8 yr and most before the age of 2 yr (Fig. 7 B). IRAK-4 deficiency is thus associated with a selective predisposition to pyogenic bacterial infections, mostly caused by Gram-positive bacteria (S. pneumoniae in particular and S. aureus to a lesser extent), and clinical status and outcome both improve with age. The detailed clinical features of IRAK-4 deficiency will be reported elsewhere (unpublished data).

The 28 patients reported in this study suffered from complete IRAK-4 deficiency. The patients had been exposed to an extremely diverse range of microorganisms, including many potential viral, bacterial, and fungal pathogens, as well as parasites (Tables S2 and S3). However, IRAK-4–deficient patients presented a strikingly narrow infectious phenotype (Table I), similar to the three patients initially reported (1). 27 patients suffered from invasive infectious disease, typically caused by Gram-positive S. pneumoniae (n = 22; 79%) and/or S. aureus (n = 9; 32%). Seven patients (25%) also presented severe infections with Gram-negative bacteria (P. aeruginosa, N. meningitidis, S. sonnei, and S. marcescens). 15 patients had peripheral infectious disease. When identified, the causal pathogens were S. aureus, P. aeruginosa, and Streptococcus species. The susceptibility of IRAK-4–deficient patients to S. aureus is consistent with that observed in IRAK-4– and MyD88-deficient mice (31, 43). MyD88-deficient mice are susceptible to P. aeruginosa (44) and, in some models, to S. pneumoniae (45, 46). Intriguingly, the 28 IRAK-4–deficient patients were not particularly susceptible to most other microorganisms, including common viruses (e.g., herpes viruses, enteroviruses, adenoviruses, and papillomaviruses), and widespread bacteria (e.g., Listeria, Mycobacterium, and Enterobacteriaceae), parasites (e.g., Toxoplasma), and fungi (e.g., Cryptococcus, Pneumocystis, Candida, and Aspergillus). As five of these patients have had no prophylaxis for 60 patient years (Fig. 7 B) (4, 6, 36), the resistance to most microbes observed is unlikely to be caused by the early death of some patients or to the prophylactic treatment of the survivors. Ascertainment bias cannot be excluded, but remains unlikely, as 10 affected relatives with causal mutations shared the case-definition clinical phenotype of index cases. In contrast, MyD88-deficient mice were found to be susceptible to mouse CMV (47), HSV-1(48), Listeria monocytogenes (49, 50), Mycobacterium avium (51), Toxoplasma gondii (52), Cryptococcus neoformans (53), Candida albicans, and Aspergillus fumigatus (54), among other relevant infections (37–39).

So why are the infectious phenotypes of MyD88/IRAK-4–deficient mice and IRAK-4–deficient humans so different? An overrepresentation of MyD88 deficiency with respect to IRAK-4 deficiency in mouse studies may be involved, although IRAK-4– and MyD88-deficient mice, when infected by the same pathogens, are indistinguishable (31, 43). We provide an experimental demonstration in this paper that the occurrence of human-specific IRAK-4–independent TLR pathways is not involved. We show that IRAK-4–deficient PBMCs do not secrete any of 11 cytokines tested when stimulated with agonists of TLR1, 2, 5, 6, 7, 8, and 9. The TLR4 response was abolished for all but two cytokines, which were weakly induced. One of these two cytokines was the IFN-inducible MIP-1β, consistent with the IFN-β mRNA response to LPS in IRAK-4–deficient PBMCs (13). IRAK-4–deficient PBMCs also responded to poly(I:C), producing IFN-inducible monocyte chemoattractant protein 1 and IFN-inducible protein 10, as expected from the previously reported induction of IFN-α, -β, and -λ in IRAK-4–deficient PBMCs and fibroblasts (13). However, poly(I:C) activates PBMCs normally in patients with TLR3 deficiency (42), making it difficult to infer conclusions about TLR3 responses from the data for poly(I:C) stimulation. In any event, the MyD88- and IRAK-4–independent TLR3 and TLR4 pathways, present in mice, cannot account for humans being more resistant (13, 15). The “conventional” MyD88-dependent pathway downstream from TLRs appears to be strictly IRAK-4–dependent in humans; no detectable leakiness can apparently account for the narrow infectious phenotype. We cannot, however, exclude the possibility that other TLR-inducible genes may be IRAK-4 independent.

We further excluded the possibility that human IRAK-4 deficiency may be milder than mouse MyD88/IRAK-4 deficiency owing to the occurrence of human-specific IRAK-4–independent TLR pathways in discrete leukocyte subsets, as suggested by the normal induction of both IL-6 and IFN-β/λ in IRAK-4–deficient fibroblasts (13). We showed that IRAK-4 deficiency impaired the TLR responses of all lymphoid and myeloid leukocyte subsets tested ex vivo, including granulocytes, monocytes, PDCs, MDCs, NK, T, and B cells. With the exception of the induction of IFN-inducible MIP-1β production in MDCs in response to poly(I:C) and LPS (Fig. 5, E and F), there was no detectable TLR response in individual subsets. The LPS response is TLR4 dependent, whereas the poly(I:C) response in MDCs appears to be TLR3 independent (42). Even IRAK-4–deficient NK cells did not respond to poly(I:C), suggesting that responses to poly(I:C) in NK cells are largely TLR3- (42) and IRAK-4–dependent. Moreover, MDDCs generated in vitro did not respond to TLR agonists, with the exception of poly(I:C) and LPS. The poly(I:C)-triggered induction of TNF-α, CD40, CD80, and CD86 in MDDCs was IRAK-4 independent (Fig. 5, G and H) and seemed to be TLR3 dependent (42). These data extend previous findings (1, 13) and show that human IRAK-4 plays a nonredundant role in the conventional TLR signaling pathway in at least seven major leukocyte subsets. In contrast, IRAK-4 may be dispensable for the “alternative,” TRIF-dependent pathways downstream from TLR3 (for IFNs and other cytokines) and TLR4 (for IFNs). Obviously, we cannot formally exclude the possibility that specific leukocyte subsets in certain tissues (55) and nonleukocyte cell types (56–59) display IRAK-4–independent TLR responses involved in host defense.

There are, therefore, no overt immunological differences between MyD88/IRAK-4–deficient mice and IRAK-4–deficient patients. Nonetheless, MyD88 and IRAK-4 are critical for protective immunity to numerous pathogens in the mouse, whereas IRAK-4 is largely redundant for protective immunity in humans. Intrinsic differences between mice and humans, affecting receptors other than TLRs, may account for the observed discrepancies. There may be non-TLRs governing the innate immune recognition of pathogens in humans but not in mice. An alternative, complementary hypothesis is that immunity to infection in animals is studied in experimental conditions, whereas immunity to infection in humans operates in natural conditions, accounting for considerable differences in the hosts, microbes, and routes of infection (60, 61). The human model can be used to define the function of host genes in a natural ecosystem in which species live and undergo selection. The ecologically relevant and evolutionarily selected function of human IRAK4 appears to be narrower than predicted from experimental studies in the mouse. This is reminiscent of the narrow infectious phenotype of patients with mycobacterial disease and mutations in the IL-12–IFN-γ circuit (62), or patients with herpes simplex encephalitis and mutations in the TLR3–UNC-93B pathway (42, 63). In any event, whether owing to species differences or to the conditions of infection, our findings for this series of IRAK-4–deficient patients strongly suggest that human IRAK-4–dependent TLRs are redundant for protective immunity to most microbes.

IRAK-4 seems to be crucial for protective immunity to Gram-positive S. pneumoniae and S. aureus and a few Gram-negative bacteria. It remains unknown whether invasive bacterial disease in patients with IRAK-4 deficiency results from an upstream impairment of IL-1R and TLR signaling or a combination of both pathways, from the defective induction of one or a combination of specific target genes downstream, or a combination of upstream and downstream defects. Impaired IL-1R and TLR2 signaling may play a role in the observed infections. Indeed, studies of experimental infection models in knockout mice have indicated that defense against S. pneumoniae and S. aureus may depend on IL-1R (64, 65), TLR2 (43, 66), and, for S. pneumoniae, perhaps also TLR4 (67, 68). Interestingly, the role of TLR2 in mouse defense against S. pneumoniae has been called into question in some experimental conditions (69, 70). Impaired stimulation of TLR7, 8, and 9 is probably not involved in predisposition to pneumococcal disease, as UNC-93B–deficient patients with impaired TLR3, 7, 8, and 9 signaling do not suffer from invasive pneumococcal disease (63). The impaired production of IL-6–inducible molecules, such as C-reactive protein (CrP), may also be involved. IRAK-4–deficient cells produce small amounts of IL-6 in vitro upon activation with IL-1β and TLR agonists. Moreover, most patients have weak or delayed acute inflammatory responses in vivo (low serum CrP levels in particular) (34, 71). As CrP contributes to the clearance of S. pneumoniae (72, 73), susceptibility to S. pneumoniae may be enhanced by the delayed increase in CrP levels. The contribution of individual molecules upstream or downstream from IRAK-4 to infectious phenotypes should be clarified by the identification of new patients with mutations in the corresponding genes (74).

Despite conferring selective susceptibility to only a few bacteria, IRAK-4 deficiency is life-threatening in infancy and childhood, with a mortality rate of 43% in our series. Most, if not all, patients would have probably died in the absence of antibiotic treatment. Strikingly, although IRAK-4 is absolutely vital in childhood, infections become rarer with age, with no deaths recorded after the age of 8 yr and no invasive infection after the age of 14 yr, even in the absence of antibiotic or IgG prophylaxis for more than 60 patient years (4, 6, 36). This dramatic improvement with age may be accounted for by the modest impact, if any, of IRAK-4 deficiency on antigen-specific T and B lymphocyte responses. Human T cells do not need IRAK-4 for activation by OKT3 in vitro (Table S2), in contrast to the results obtained for mice in a previous report (32) and in accordance with a more recent study (75). Moreover, our patients displayed no detectable global defect of protein antigen–specific T and B cell responses. However, most of the patients displayed IgE overproduction, and some patients have been shown to have weak antibody responses to a subset of glycan antigens (11, 12, 33). A more thorough investigation of B cells and antibody responses in IRAK-4–deficient patients is therefore currently underway (unpublished data). Our data are consistent with the apparently intact primary and secondary antigen-specific responses in mice with MyD88 deficiency, TRIF deficiency, or both (76, 77). Adaptive immunity may therefore progressively compensate for the poor innate immunity in our patients. An alternative and complementary hypothesis, accounting for the clinical improvement of IRAK-4–deficient patients with age, is that innate immune responses may also mature with age (78, 79). As shown in this study, the TIR pathway, including TLR responses in particular, remains dependent on IRAK-4 with age, but the maturation of other innate pathways may gradually compensate for the lack of TIR–IRAK-4 signaling.

Our study was conducted according to the principles expressed in the Helsinki Declaration, with informed consent obtained from each patient or the patient's family. The study was approved by the Comité d'Éthique, CCPPRB, Hôpital Necker–Enfants Malades.

Genomic DNA was isolated from whole blood cells or from B-EBVs. The cells were lysed by incubation overnight at 37°C in extraction buffer (10 mM Tris, 0.1 M EDTA, 0.5% SDS, 1 mg/ml proteinase K) and subjected to phenol/chloroform extraction. DNA was precipitated in ethanol. Amplified PCR products were analyzed by electrophoresis in a 1% agarose gel purified by centrifugation through superfine resin (Sephadex G-50; GE Healthcare), sequenced by dideoxynucleotide termination with the BigDye terminator kit (Applied Biosystems), and analyzed on an ABI Prism 3730 apparatus (Applied Biosystems).

RNA was extracted from B-EBV and SV40-fibroblasts in TRIzol (Invitrogen), and cDNA was prepared using reverse transcriptase (SuperScript II; Invitrogen) for RT-PCR, according to the manufacturer's instructions. Proteins for Western blotting were extracted from B-EBV and SV40-fibroblasts, and Western blots were probed with rabbit antibodies against IRAK−4 (Tularik) and GAPDH (Santa Cruz Biotechnology, Inc.).

TLR agonists and cytokines were used at the following final concentrations, unless otherwise indicated: synthetic triacylated lipopeptide (PAM₃CSK₄, agonist of TLR1/2; Invivogen), 100 ng/ml; synthetic diacylated lipopeptide (PAM₂CSK₄, agonist of TLR2/6; Invivogen), 100 ng/ml; poly(I:C) (a synthetic analogue of dsRNA, polyinosine-polycytidylic acid, and nonspecific TLR3 agonist; Invivogen), 25 μg/ml; LPS (Re 595 from Salmonella minnesota, agonist of TLR-4; Sigma-Aldrich), 100 ng/ml; flagellin (TLR5 agonist; Invivogen), 1 μg/ml; 3M-13 (TLR7 agonist) and 3M-2 (TLR8 agonist; both provided by 3M Pharmaceuticals), 3 μg/ml each; R-848, resiquimod hydrochloride (TLR7 and TLR8 agonist; provided by PharmaTech), 3 μg/ml; and unmethylated CpG DNA CpG-C (C274; 5′-TCGTCGAACGTTCGAGATGAT-3′; TLR9 agonist; provided by R. Coffman and F. Barrat, Dynavax Technologies, Berkeley, CA), 3 μg/ml. Polymyxin B was used at 10 μg/ml (Sigma-Aldrich).

We suspended 10⁶ B-EBV cells per well in RPMI 1640 (Invitrogen) supplemented with 10% FCS (Invitrogen) and activated them by incubation with 3M-13, 3M-2, R-848, and 10⁻⁷ M PMA plus 10⁻⁵ M ionomycin (Sigma-Aldrich) for 24 h. 10⁵ SV40-fibroblast cells per well were seeded in DMEM (Invitrogen) supplemented with 10% FCS in 24-well plates. Cells were activated with 20 ng/ml TNF-α (R&D Systems), 10 ng/ml IL-1β (R&D Systems) and 10⁻⁷ M PMA plus 10⁻⁵ M ionomycin the next day. The supernatants were harvested after 24 h of activation.

ELISA determinations of TNF-α, IL-6, and IL-10 in cell culture supernatants were performed with a kit (PeliPair reagent set; Sanquin), according to the manufacturer's instructions. Optical density was determined by an automated ELISA reader (MR5000; Thermolab Systems). We used a fluorescence-based assay (a human cytokine 25-plex antibody bead kit) that can detect 25 cytokines (LHC0009; Biosource International) for the simultaneous determination of multiple cytokines. Fluorescence was measured with a 100 IS system (Luminex Corporation). The assay and analysis were performed according to the manufacturer's instructions.

Blood samples from healthy controls or patients were collected into heparin-containing tubes, and PBMCs and granulocytes were separated by Ficoll-gradient centrifugation. The patients were of different ages when the experiments were performed, ranging from 7 to 32 yr old. For granulocyte isolation, erythrocytes were lysed and washed twice in PBS. More than 95% of the granulocytes purified on Ficoll were CD15⁺. We did not purify granulocytes by flow cytometry, as the surface expression and TLR-induced shedding of L-selectin (CD62L) were not better detected (unpublished data).The PBMC preparation was enriched in T cells, B cells, monocytes, and NK cells by magnetic bead isolation using anti-CD3, -CD19, -CD14, and -CD56 microbeads (Miltenyi Biotec), according to the manufacturer's instructions. Purified T cells were labeled with anti-CD3–FITC (BD Biosciences), B cells with anti-CD19–PE (BD Biosciences), monocytes with anti-CD14–FITC (BD Biosciences), and NK cells with anti-CD3–FITC/anti-CD56–PE (BD Biosciences) antibodies, and sorting was performed on a flow cytometer (FACSVantage; BD Biosciences). The isolated cells were cultured in RPMI 1640 supplemented with 10% FCS, with immediate TLR agonist stimulation. We added 100 U/ml IL-2 to cultures of purified T cells. Purified B cells were suspended in RPMI 1640 supplemented with 10% FCS at a density of 10⁶ cells/ml. Cells were stimulated with TLR agonists together with 100 U/ml IL-4 for 3 d.

Granulocytes were isolated as described in the previous section, activated with TLR agonists, stained with anti-CD62L–FITC (BD Biosciences) antibody, and analyzed by flow cytometry, as previously described (10).

PBMCs were suspended at a final density of 2 × 10⁶ cells/ml in RPMI supplemented with 10% FCS. They were incubated at 37°C, under an atmosphere containing 5% CO₂, and stimulated with TLR agonists. 10 μg/ml brefeldin A was added after 1 h of activation. After 3.5 h of activation, cells were washed and stained with anti-Lin1–FITC (BD Biosciences), anti-HLADR–PerCP (BD Biosciences), and anti-CD123–PE-Cy7 (e-Bioscience) antibodies. For intracellular staining, PBMCs were permeabilized with the Cytofix/Cytoperm kit (BD Biosciences), according to the manufacturer's instructions. Anti–TNF-α–allophycocyanin (BD Biosciences) and anti–MIP-1β–PE (BD Biosciences) antibodies were used to assess the response of MDCs and PDCs to TLR agonists. PBMCs were also incubated with the respective isotype controls, and cells were acquired on a three-laser flow cytometer (LSR system; BD Biosciences). MDCs were defined as Lin-1⁻, HLA-DR⁺, and CD123^low, and PDCs were defined as Lin-1⁻, HLA-DR⁺, and CD123^high. For analysis, the quadrant for each individual tested was set such that 98% of PBMCs incubated with the respective isotype controls were negative for nonspecific staining.

MDDCs were prepared as previously described (80). In brief, PBMCs were suspended in RPMI 1640 supplemented with 10% FCS, plated in cell culture flasks, and incubated for 1 h. Monocytes attached to the bottom of the culture flask and nonadherent cells were removed with medium. Monocytes were then cultured in RPMI 1640 supplemented with 10% FCS, 25 ng/ml GM-CSF, and 100 U/ml IL-4. GM-CSF and IL-4 were added to the medium every other day to maintain their initial concentrations. On day 7 or 8, some of the MDDCs were stained for CD1a and CD14. Living cells and cell debris were distinguished by forward/side scatter. More than 95% of living cells were CD1a⁺, and no CD14⁺ cells were detected. On day 7 or 8, MDDCs were suspended in RPMI 1640 supplemented with 10% FCS at a density of 2 × 10⁵cells/ml, and supernatants were collected after 24 h of activation. The up-regulation of surface markers was assessed by collecting MDDCs and staining them with anti-CD1a–PE (BD Biosciences), anti-CD40–FITC (BD Biosciences), anti-CD80–FITC (BD Biosciences), and anti-CD86–FITC (BD Biosciences) antibodies.

Patients were immunized against diphtheria and tetanus in accordance with international recommendations. Nine patients received multiple injections of glycan antigens (nonconjugated [“Pneumo23”] and conjugated [“Prevenar”] antipneumococcal vaccine), and their specific antibody titers were subsequently monitored in detail.

Fig. S1 demonstrates a deletion of the IRAK4 locus on one allele in P2 and the presence of both IRAK4 loci in P7. Fig. S2 shows the detailed results for each of the 11 cytokines for which a response to TLR agonists in healthy controls could be detected by multiplex assay. Table S1 shows blood leukocyte subsets in IRAK-4–deficient patients. Table S2 highlights T cell proliferation, Ig levels, and humoral responses to recall antigens and to glycans in IRAK-4–deficient patients. Table S3 offers the serology of patients to common viruses.

We thank all members of the laboratory for helpful discussions, and Catherine Bidalled, Martine Courat, and Tony Leclerc for secretarial and technical assistance. We would particularly like to thank the patients and their families, whose trust, support, and cooperation were essential for collection of the data used in this study.

H. von Bernuth was supported by grants from the Deutsche Forschungsgemeinschaft (VO 995/1-1 and VO 995/1-2), from the European Ph.D. program of the San Raffaele Institute, and from the program “Legs Poix” of the Parisian Universities. L. Maródi was supported by a grant from the Hungarian Research Fund (OTKA 49017), and A. Puel was supported by a grant from the European Union (QLK2-CT-2002-00846). The Laboratory of Human Genetics of Infectious Diseases is supported by the March of Dimes, the BNP Paribas Foundation, the Dana Foundation, and the Schlumberger Foundation. J.L. Casanova is an International Scholar of the Howard Hughes Medical Institute.

The authors have no conflicting financial interests.