Study Subjects

A 10-week old infant presenting with disseminated BCG infection (post-vaccination), oral candidiasis, cachexia and suspected signs of immunodeficiency was admitted to hospital and was investigated (Case Description in Supplementary Information). She underwent multiple rounds of aggressive antibiotic treatments which were only partially effective in improving her health. She was successfully treated by transplantation wtih cord-blood stem cells. She was the second-born child to healthy unrelated Irish parents (kindred A); the elder sibling was well, and all had received BCG vaccination.

Two additional individuals diagnosed with MSMD were also enrolled in the study. Affected individual 1 (kindred B) was born in 1970 to non-consanguineous parents of Italian descent living in Brazil. He was vaccinated with BCG at birth and developed recurrent episodes of lymphadenopathies at age 15 months, 20 years and 30 years; on two such episodes, lymph node biopsies revealed presence of acid-fast bacilli and all such episodes were successfully treated with anti-mycobacterial drug regimens. His mother, father and sister were vaccinated with BCG and did not present with clinical infectious disease suggesting immunodeficiency.

Affected individual 2 (kindred C) was born in 1996 to a non consanguineous family of Italian descent living in Chile, and received BCG vaccination at birth. At one year of age, she developed lymphadenopathy with chronic granulomatous tuberculoid lesions. At two years of age, she presented with multiple lymphadenopathies and fever, requiring hospital admission. Histological analysis of lymph-node biopsy and bacterial culture identified pyrazinamide resistant M. bovis, and the child was successfully treated with adjusted antibiotic treatment administered for 12 months; no subsequent clinical episodes were reported. We obtained written informed consent from participants or their parents, and Institutional Review Boards approved the studies.

Genetic and Transcriptional Analyses

We sequenced the exons and selected non-coding sequences of IRF8 from participants’ genomic DNA following standard PCR amplification with sequence-specific primers (Table S3). We assayed the transcriptional activity of mutant and non-mutant IRF8 through activation of the IL12B and NOS2 promoters using reporter constructs, transiently transfected into RAW 264.7 macrophages, as previously described[16].

Biochemical Assays, Molecular Characterization and Statistical Analysis

See the Supplementary Appendix for details.

Autosomal Recessive IRF8 Deficiency

We evaluated a 10-week-old infant presenting with disseminated BCG disease (post-vaccination), cachexia and oral candidiasis (see Supplementary Appendix). The blood count revealed a strikingly abnormal myeloid compartment with absence of monocytes and basophils and a very high neutrophil count (Fig. 1; Table S1). Examination of peripheral blood using flow cytometry confirmed severe depletion of the CD3/19/56⁻ (lin⁻) HLA-DR⁺ cellular compartment, with total absence of both CD14⁺ and CD16⁺ monocytes (Fig. 1A–C). Furthermore, we could not detect DCs in the blood, whether CD11c⁺ myeloid DCs (CD1c⁺ or CD141⁺) or CD123⁺ plasmacytoid DCs (Fig. 1A–D). The only lin⁻DR⁺ cells proved to be circulating CD34⁺ progenitor cells, which were elevated (Fig. S1, Panel A) and correlated with elevated serum levels of Fms-like tyrosine kinase-3 ligand (Flt3L)(Fig. S1, Panel B). In contrast, B and NK cells were present in normal numbers (Table S2 and Fig. S2), thus excluding the recently described syndrome of autosomal dominant and sporadic monocytopenia[18,19].

Through assays using whole blood, we observed that IL12 production in response to BCG, phytohaemagglutinin (PHA) and lipopolysaccharide was completely absent and IFN-γ production was poor, with similarly poor production of TNF-α, IL-10 and IL-6 (Fig.1F–H). Preincubation of the infant’s cells with IL-12 could partially restore IFN-γ production in response to the same stimuli (Fig. 1G). Biopsies of the infant's bone marrow and axillary lymph node (Fig. S3) showed striking myeloid hyperplasia, and the presence of acid-fast bacilli within granulomata in (a) lymph node. Despite the subject’s profound peripheral monocytopenia, the diseased node was positive for histiocytic markers CD68 and CD163 and bone trabeculae showed evidence of normal osteoclast activity (Fig. S4). In dermal tissues, the density of CD1a⁺ and CD14⁺ DCs was remarkably low (Fig. 1C–D). In contrast, we observed epidermal Langerhans cells (LCs) to be at normal density (Fig. 1E), and detected the occasional Langerin⁺/CD1a⁺ DC in lymph node (Fig. S4). We conclude that the subject had a profound deficit of tissue DCs, blood DCs and monocytes, a variable deficit of tissue macrophages, and normal levels of LCs.

This combination of immunodeficiency [10,11], DC deficiency [12–15] and myeloproliferation [20] is strikingly similar to the phenotype of mice harboring loss-of-function mutations in Irf8. On sequencing IRF8 in the subject, we observed a homozygous missense variant predicted to effect a lysine-to-glutamic acid substitution at position 108 (K108E) (Fig. 2A). Both parents were heterozygous for K108E and an unaffected sibling lacked the variant allele (Fig. 2C). We sequenced IRF8 in 454 unrelated persons with clinical susceptibility to mycobacterial infection and did not detect the variant or any other homozygous variants.

Amino-acid position 108 is within the DNA-binding domain of IRF8 (Fig. 2A); the lysine residue is invariant in IRF8 orthologs (Fig. 2B) and is highly conserved among human IRF family members (data not shown). We expressed the mutant IRF8 in cultured mouse macrophages. Immunoblotting studies showed similar expression levels of normal (K108) and mutant (K108E) variants, and a comparatively slower electrophoretic mobility of the mutant variant (Fig. 3E), suggesting that the mutation affects overall protein structure/folding. We tested the ability of normal and mutant isoforms to activate transcription of IRF8 targets — the promoters of IL12B (Fig. 3A) and NOS2 (Fig. 3B) — in mouse macrophages. Although the combination of normal IRF8 and its co-activator IRF1 induced a dose-dependent stimulation of IL12B and NOS2 promoters, the K108E variant was almost inactive (Fig. 3A, 3B), suggesting that K108E abrogates the IRF1-dependent transcriptional activity of IRF8. Moreover, the mutant IRF8 variant bound the IL12B promoter far more weakly than did the normal variant (Fig. 3D).

Molecular modeling using the structure of DNA-bound IRF2 [21] places K108 in a short β strand that runs parallel to the major DNA-binding α helix, suggesting that it makes a critical hydrogen bond with the DNA sugar backbone (3.3A) that facilitates IRF8 docking onto DNA (Fig. 2E-1). Replacement of a positively-charged amino acid (lysine) by one that is negatively charged (glutamic acid) at position 108 would likely prevent the formation of a hydrogen bond (Fig. 2E-2), and cause local repulsion of the supporting IRF8 β strand away from DNA, allowing a water molecule to fill the space. This model predicts loss of DNA binding and thus loss of transactivation.

These findings show that the subject carries a loss-of-function mutation in IRF8. The heterozygous parents were healthy despite having received BCG as neonates, and had normal numbers of peripheral blood monocytes and DCs (Fig. 1), consistent with the autosomal recessive (AR) inheritance of the IRF8^K108E allele.

Autosomal Dominant IRF8 Deficiency

In parallel with our analysis of the infant with recessive disease, we sequenced IRF8 in 454 persons with MSMD of unknown genetic etiology (we ruled out the possibility that these persons carried mutations in several of the genes in which alterations are known to cause MSMD; see Supplementary Appendix). Two unrelated persons from Chile and Brazil and suffering from recurrent episodes of disseminated BCG disease, were each found to carry the same de novo heterozygous mutation, predicted to effect a threonine-to-alanine substitution at position 80 (T80A) of IRF8 (Fig. 2A,D). We genetically confirmed reported biological paternity. That each mutation was de novo suggests that the same T80A allele arose independently on two occasions. We did not detect the T80A variant in 1064 healthy controls [41] of diverse relevant ancestry (See Supplementary Appendix). Amino-acid position 80 of IRF8 is within the DNA-binding domain (Fig. 2A); the threonine residue is strictly conserved across IRF8 orthologs (Fig. 2B) and human paralogous genes (data not shown).

The T80A mutation had no effect on level and stability of IRF8 in immortalized B cell lines derived from the subjects (Fig. 3E), or following expression in transfected macrophages (data not shown). When tested for its ability to transactivate the promoters of IL12B (Fig. 3A) and NOS2 (Fig. 3B) in mouse macrophages, the T80A mutant showed relatively low activity (Fig. 3A, 3B), similar to that of the hypomorphic R294C Irf8 variant of Irf8-deficient BXH2 mutant mice (Fig. 3A, 3B). The T80A mutant showed weak recruitment to the IL12B promoter (20% of the amount of the non-mutant IRF8 protein), suggesting that the mutation interferes with DNA binding (Fig. 3D).

Homology modeling shows that T80 maps to the DNA-binding α helix of IRF8 that fits into the major groove, with its side-chain directed towards the DNA bases (Fig. 2E-3). Molecular dynamics simulations indicate that T80A alters the hydrophobic interface between the protein and DNA, and possibly modulates DNA-binding specificity of IRF8 (Fig. 2E-4). Alanine has a smaller side chain than does threonine, and so may allow entry of a water molecule at the DNA:IRF8 interface. Although the functional data suggest that T80A is severely hypomorphic, we investigated whether the T80A variant is negatively dominant (that is, it interferes with the function of the non-mutant IRF8 protein). We co-expressed the mutant and non-mutant alleles in macrophages (Fig. 3C) and observed decreasing activation of the IL12B (not shown) and NOS2 (Fig. 3C) promoters with increasing levels of mutant (T80A) IRF8 against fixed levels of non-variant IRF8. The effect was T80A-specific; we did not observe it to be associated with K108E (data not shown) or R294C mutant proteins. We conclude that the T80A mutation has a dual effect on IRF8 function and causes autosomal-dominant MSMD.

We observed subtle deficits in the peripheral blood mononuclear cells of the two T80A affected adults. There was no deficiency of circulating lymphocytes and granulocytes (data not shown), monocyte subsets, or BDCA2⁺ CD123⁺ plasmacytoid DC (Fig. 4D). However, within the CD11c⁺ myeloid DC subset, which is normally divided into minor CD141⁺ and major CD1c⁺ subsets, there was marked loss of CD1c⁺DCs (Fig. 4A), while the CD141⁺ subset and total number of CD11c⁺DCs remained intact (Fig. 4B). CD1c⁺ DCs from unaffected persons can produce large amounts of IL-12 when stimulated with the TLR7/8 ligand R848 (compared to PBMC; student t test, p<0.02) whereas their depletion from PMBCs was associated with comparatively low levels of IL-12 production by the remaining PBMCs (p<0.02) (Fig. 4E). PBMCs harboring IRF8^T80A produced one-third the amount of IL-12 produced by control cells in response to R848 (p<0.004)(Fig. 4E). We suggest that depletion of IL-12-producing CD1c⁺ DCs contributes to susceptibility to mycobacterial disease in these individuals. In vitro assays using whole blood from the affected persons showed no detectable impact on production of IFN-γ (Fig. S6), or in the subjects’ PBMCs in response to stimulation by PPD (purified protein derivative), BCG or PHA (Fig. S7), or by PHA-driven T-cell blasts stimulated with IL-12 (data not shown). Production of IL-12 in a whole blood assay in response to BCG (Fig. S6), by monocyte-derived DCs stimulated with CD40L (Fig. S8A), and by EBV-B cells stimulated with phorbol dibutyrate (Fig. S8B) was also normal, indicating that T80A does not cause a generalized defect in IL-12 production.

Because the two healthy heterozygous parents of the infant with autosomal-recessive IRF8 deficiency had normal levels of functional DCs, the marked deficit of CD1c⁺CD11c⁺ DCs was probably caused by the dominant-negative effect of the T80A mutant protein.