Biallelic interferon regulatory factor 8 mutation: A complex immunodeficiency syndrome with dendritic cell deficiency, monocytopenia, and immune dysregulation

Background The homozygous K108E mutation of interferon regulatory factor 8 (IRF8) is reported to cause dendritic cell (DC) and monocyte deficiency. However, more widespread immune dysfunction is predicted from the multiple roles ascribed to IRF8 in immune cell development and function. Objective We sought to describe the effect on hematopoiesis and immunity of the compound heterozygous R83C/R291Q mutation of IRF8, which is present in a patient with recurrent viral infection, granuloproliferation, and intracerebral calcification. Methods Variant IRF8 alleles were identified by means of exome sequencing, and their function was tested by using reporter assays. The cellular phenotype was studied in detail by using flow cytometry, functional immunologic assay transcriptional profiling, and antigen receptor profiling. Results Both mutations affected conserved residues, and R291Q is orthologous to R294, which is mutated in the BXH2 IRF8-deficient mouse. R83C showed reduced nuclear translocation, and neither mutant was able to regulate the Ets/IRF composite element or interferon-stimulated response element, whereas R291Q retained BATF/JUN interactions. DC deficiency and monocytopenia were observed in blood, dermis, and lung lavage fluid. Granulocytes were consistently increased, dysplastic, and hypofunctional. Natural killer cell development and maturation were arrested. TH1, TH17, and CD8+ memory T-cell differentiation was significantly reduced, and T cells did not express CXCR3. B-cell development was impaired, with fewer memory cells, reduced class-switching, and lower frequency and complexity of somatic hypermutation. Cell-specific gene expression was widely disturbed in interferon- and IRF8-regulated transcripts. Conclusions This analysis defines the clinical features of human biallelic IRF8 deficiency, revealing a complex immunodeficiency syndrome caused by DC and monocyte deficiency combined with widespread immune dysregulation.

Effective immunity depends on multiple interacting transcription factors (TFs) that govern hematopoietic development and immune cell function. Members of the interferon regulatory factor (IRF) family of TFs have a dual role, interacting with hematopoietic TFs to determine cell fate and with immune signaling molecules to direct cell responses. 1 IRF4 and IRF8 are structurally related family members with an N-terminal DNA-binding domain and C-terminal interferon regulatory factor-associated domain (IAD). 2 They activate or repress transcription at specific DNA motifs in collaboration with other TFs interacting with the IAD.
IRF8 heterodimerization partners include SPI1/PU.1, which binds the Ets-interferon regulatory factor composite element (EICE), frequently activating gene expression. Other IRF family members (IRF1, IRF2, and IRF4) are cofactors at the mostly repressive interferon-stimulated response element (ISRE). 3,4 IRF8 forms a complex with BATF3 and JUN on the activator protein 1-interferon regulatory factor composite element (AICE) to activate gene expression. 5 IRF8 plays a critical role in hematopoietic lineage determination through these interactions, governing the development of granulocytes, dendritic cells (DCs), monocytes, and B cells. 6,7 In myeloid cell development IRF8 is pivotal in determining granulocyte versus monocyte fate. IRF8 competes with CEBPA for binding to chromatin, resulting in attenuation of neutrophil differentiation in granulocyte-monocyte progenitors. 8 IFR8/PU.1 interaction is critical in monocyte differentiation by marking distal enhancers for subsequent activation by the master regulator Kruppel-like factor 4. 9 The Irf8 2/2 mouse develops massive expansion of granulocytes and progression to fatal myeloblastic leukemia. 10 Ly6C 1 and Ly6C 2 monocytes are depleted, but some tissue macrophages persist. 11 The BXH2 mouse is homozygous for the Irf8 R294C mutation in the IAD region of the gene, disrupting association with heterodimerization partners. Monocytes and macrophages are present but rendered hypofunctional in their response to IFN-g, 12 production of inflammatory cytokines, 13 and defense against intracellular infection. 4,14 These 2 models illustrate the dual hematopoietic and immune response roles of IRF8 and their differential sensitivity to IRF8 deficiency.
DCs are critical for activation of the immune response and arise independently of monocytes. All mammalian species have IFN-a-producing plasmacytoid dendritic cells (pDCs) and 2 myeloid or conventional dendritic cell (cDC) populations. 15 cDC1s express C-type lectin containing domain type 9A/ chemokine XC receptor 1 (CLEC9A/XCR1) and are specialized for cross-presentation, whereas cDC2s express signal regulatory protein a (SIRPA) and mediate T H 2 and T H 17 responses. In human subjects they are marked by CD141 and CD1c, respectively. 16 In mice IRF8 specifies DC lineage in cooperation with Id2. 17 IRF8 is also upregulated by E2-2 18 and required for pDC development and function. 19,20 In cDC1s it interacts with BATF3 to maintain terminal differentiation. 21 The Irf8 2/2 mouse lacks cDC1s and has reduced numbers of pDCs, whereas selective loss of cDC1s is observed in the hypomorphic BXH2 mouse. cDC2s are preserved in both IRF8-deficient strains through their reciprocal dependence on IRF4. 22 In B-cell development the balance between IRF8 and IRF4 mirrors that observed in specification of myeloid DCs subsets.
Abbreviations used AICE: Activator protein 1-interferon regulatory factor composite element BAL: Bronchoalveolar lavage cDC: Conventional dendritic cell CDR3: Complementarity-determining region 3 DC: Dendritic cell EICE: Ets-interferon regulatory factor composite element EMSA: Electrophoretic mobility shift assay HA: Hemagglutinin IAD: Interferon regulatory factor-associated domain IRF: Interferon regulatory factor ISRE: Interferon-stimulated response element NK: Natural killer pDC: Plasmacytoid dendritic cell SSC: Side scatter STAT: Signal transducer and activator of transcription TF: Transcription factor TLR: Toll-like receptor Treg: Regulatory T IRF8 and IRF4 act redundantly to drive pre-B-cell development but are required differentially to regulate marginal zone versus follicular fate. IRF8 is important in the germinal center reaction and class-switch recombination, whereas IRF4 is necessary for plasma cell differentiation. 23,24 IRF8 also directly influences the terminal differentiation of natural killer (NK) and T cells. Defects in NK cell maturation and function were reported recently in patients with IRF8 mutations. 25 Investigation of the Irf8 2/2 mouse has identified defects in T-cell subset differentiation. IRF8 is required for T H 1 polarization, CD8 1 effector memory T-cell differentiation, 26,27 and generation of a T H 1-associated subset of regulatory T (Treg) cells, 28 with conflicting reports of the influence of IRF8 on T H 17 polarization. 29,30 Human primary immunodeficiencies resulting from singlegene mutations offer unique opportunities to study the development and regulation of immune cells in vivo. An infant with impaired immunity to mycobacteria, viruses, and fungi because of a homozygous K108E mutation of IRF8 was described previously. 31,32 The human phenotype was broadly reminiscent of Irf8 2/2 mice, except that cytopenia encompassed all DC and monocyte subsets. Analysis was complicated by environmental factors of disseminated BCG infection and cytoreductive therapy, and wider cell-specific immune defects were not studied. Here we describe 2 novel alleles of IRF8 in a compound heterozygous patient who was ambulatory between infective exacerbations. This reveals a complex immunodeficiency syndrome comprising profound hematologic defects with widespread immune dysregulation.

Study approval
The study was performed in accordance with the Declaration of Helsinki and approved by the Newcastle and North Tyneside Research Ethics Committee. Written informed consent was obtained from participants before recruitment.

Whole-exome sequencing
Details of whole-exome sequencing and analysis are described in the Methods section in this article's Online Repository at www.jacionline.org.

Serum autoantibody analysis across human protein microarray
Sera from the patient and 2 age-matched control subjects were incubated on HuProt human protein microarrays (CDI Laboratories, Baltimore, Md). Details are available in the Methods section in this article's Online Repository.

Luciferase reporter assays
Expression constructs encoding human PU.1, Spi-B transcription factor (SPIB), BATF, and IRF1 have been described elsewhere. 33,34 Human JUNB and IRF8 wild-type and mutated cDNA sequences preceded by a hemagglutinin (HA) epitope tag were purchased from GenScript (Piscataway, NJ) and cloned into pIRES2-EGFP expression vectors (Clontech, Mountain View, Calif). Luciferase reporter constructs are described in the Methods section in this article's Online Repository. HeLa cells were seeded at 1 3 10 5 cells/well in 24-well plates with RPMI plus 10% FBS and transfections were performed in triplicate with GeneJuice Transfection Reagent (Novagen, Merck). Twenty nanograms of expression vector as cotransfected with 150 ng of firefly luciferase vector and 1 ng of pRL-CMV Renilla luciferase control. For each condition, the total amount of vector transfected was normalized by inclusion of control pIRES2-EGFP empty vector. Luciferase activity was assayed 24 hours after transfection by using the Promega luciferase dual assay system (Promega, Madison, Wis) and analyzed on a Berthold Lumat LB Luminometer (Berthold Technologies, Bad Wildbad, Germany). For each condition, luciferase activity was expressed relative to the average reading from the empty vector transfections and normalized to 1. Comparable expression of IRF8 wild-type and variant alleles by transected cells was confirmed by means of immunoblotting for the HA epitope tag.

Electrophoretic mobility shift assays
Nuclear extracts were prepared from HeLa cells transfected with expression vectors for PU.1, SPIB, IRF1, BATF, JUNB, IRF8 WT, IRF8 R83C, or IRF8 R291Q, as previously described. 33 Double-stranded DNA probes endlabeled with [g-32 P] ATP using T4 polynucleotide kinase were incubated with appropriate nuclear extract in the presence of poly(dI:dC; Amersham Biosciences, Buckinghamshire, United Kingdom) for 30 minutes at room temperature. The probes are detailed in the Methods section in this article's Online Repository.

Cell culture and functional assays
PBMCs were prepared by means of density gradient centrifugation, and neutrophils were harvested from the red cell/granulocyte layer, according to standard protocols. Skin from the patient and control subjects undergoing mammoplasty surgery was processed, as previously described. 35 Dihydrorhodamine oxidative burst was performed with Phagoburst (Glycotope Biotechnology, Berlin, Germany). PHA responses were determined by means of incorporation of tritiated thymidine after 6 days of culture. Whole-blood cytokine assays were performed, as previously described. 36 Cytokines and chemokines in serum and bronchoalveolar lavage (BAL) fluid supernatants were measured with Luminex by using ProcartaPlex 34plex Immunoassay (eBioscience, San Diego, Calif) on Qiagen Liquichip 200 (Qiagen, Hilden, Germany) running Luminex 100 integrated system software (version 2.3; Luminex, Austin, Tex). Standard curves were constructed to interpolate analytes by using Procartaplex Analyst (version 1.0). The mean of technical duplicates was recorded.
Intracellular cytokine measurement was performed on sorted T-cell subsets. Fifty thousand cells in 200 mL of RPMI plus 10% FBS were stimulated for 3 hours with 0.02 mg/mL phorbol 12-myristate 13-acetate (Sigma-Aldrich) and 0.5 mg/mL ionomycin (Sigma-Aldrich) at 378C, followed by 2 mg/mL brefeldin A (Sigma-Aldrich) for a further 3 hours. Cells were washed and stained for intracellular cytokines after permeabilization with FOXP3 staining buffer (eBioscience).

Flow cytometry and microscopy
PBMCs and skin mononuclear cell preparations or lymphoblastoid cell lines were stained in aliquots of 1 to 3 3 10 6 cells in 50 mL of Dulbecco-PBS with 2% FCS and 0.4% EDTA. Dead cells, usually less than 5%, were

12
-     Baso excluded by using 49-6-diamidino-2-phenylindole dihydrochloride (Partec, G€ orlitz, Germany). Intracellular IRF8 staining was performed after fixation and permeabilization, according to the manufacturer's instructions (eBioscience). Analysis was performed with an LSRFortessa X-20 and sorting with a FACSAria III (BD Biosciences, San Jose, Calif). Data were processed with FlowJo software (Tree Star, Ashland, Ore). Absolute cell counts were obtained with TruCount tubes (BD Biosciences) with 200 mL of whole blood with 900 mL of lysis buffer. Fluorescence microscopy of epidermal sheets and cytospin and Giemsa staining of PBMCs was performed, as previously described, 36 with an Axioplan 2 microscope (Carl Zeiss, Oberkochen, Germany) with an EC Plan-Neofluar 340 NA 0.75 lens. Antibodies are listed in Table E1 in this article's Online Repository at www.jacionline.org.
Deep sequencing of the B-cell IgH complementaritydetermining region 3 Peripheral blood B cells (>60,000) from patients with the 83C/291Q and K108E mutations and 3 age-matched controls were sorted by means of fluorescence-activated cell sorting to greater than 95% purity, and genomic DNA was extracted by using standard methods. High-throughput sequencing of the B-cell receptor heavy chain (IgH) complementarity-determining region 3 (CDR3) was undertaken by using the Adaptive Biotechnologies immuno-SEQ Assay (www.adaptivebiotech.com). Further details are available in the Methods section in this article's Online Repository.

Statistical analysis
Graphs were plotted with Prism software (version 5; GraphPad Software, La Jolla, Calif) and means, SDs, z scores, and Student t test results were calculated within this software. For gene ontology analysis and B-cell receptor IgH gene use, R version 3.3.0 was used, together with the bimaRt, GOstats, and ggplot2 packages.

RESULTS Immunodeficiency
A white male subject born to nonconsanguineous parents had an early history of recurrent severe viral respiratory tract infection with influenza H1N1, rhinovirus, and mycoplasma, often requiring hospitalization and ventilatory support (Fig 1, A, and see the Methods section in this article's Online Repository for clinical details). Developmental delay had been noted, and multifocal bilateral calcification was present at age 3 years (Fig 1, B). Total CD4 1 cell, CD8 1 T-cell, B-cells, and NK cell counts were normal, and T-cell PHA responses were intact (see Table E2 and Fig E4, B, in this article's Online Repository at www.jacionline.org). IgA deficiency was noted during infancy (see Table E3 in this article's Online Repository at www.jacionline.org). Whole-blood cytokine assay revealed absent IL-12 production with reduced IFN-g, IL-6, IL-10, and TNF-a levels in response to BCG, LPS, or LPS and IFN-g (Fig 1, C). No antinuclear antibodies were identified in serum by using BioPlex 2200 screening, but analysis across a human protein microarray (HuProt), representing more than 15,000 genes (see Table E4 in this article's Online Repository at www.jacionline.org), revealed a striking reactivity against immunoglobulin heavy and kappa and lambda light chains compared with 2 age-matched control subjects (Fig 1, D). A number of discrete non-immunoglobulin-targeted antibodies were identified in the patient, and a smaller number were identified in control subjects (Fig 1, E).

Compound heterozygous IRF8 mutations: R83C and R291Q
Whole-exome sequencing of the patient and both parents identified 2 novel missense mutations in coding regions of IRF8: c.247C>T; p.Arg83Cys and c.872G>A; p.Arg291Gln (Fig 2, A). Four genes with biallelic variants were identified. Of these, IRF8 had the highest combined annotation-dependent depletion scores and was most biologically plausible (see Table  E5 in this article's Online Repository at www.jacionline.org). Both IRF8 mutations were absent from ExAC, dbSNP147, EVS, the NHLBI GO Exome sequencing project, and gnomAD databases and 116 locally sequenced samples. Mutated residues were highly conserved between IRF family members and IRF8 orthologs (Fig 2, B).
In luciferase reporter assays neither R83C nor R291Q activated a derivative of the PSMB8 promoter EICE in the presence of PU.1 or SPIB, and both were unable to repress IRF1-mediated expression through the TAPASIN promoter ISRE (Fig 2, C). However, a difference was observed in reporter constructs containing the IL10 AICE activated by BATF/JUNB. This element is repressed by wild-type IRF8. Here the R83C variant was inactive, but R291Q was comparable with wild-type IRF8. These results were consistent with binding patterns seen in electrophoretic mobility shift assays (EMSAs), where only wild-type IRF8 bound to DNA in the presence of the appropriate cofactor, except for the AICE probe, which retained R291Q binding in the presence of BATF/JUNB (Fig 2, D). Total IRF8 protein expression was normal in patients' lymphoblastoid cell lines and whole-cell lysates of transfected HeLa cells (Fig 2, E and F), but in cellular fractionation experiments R83C was unable to translocate to the nucleus (Fig 2, G). R291Q showed a nuclear-cytoplasmic distribution similar to wild-type protein.

Monocyte and DC deficiency with preserved tissue macrophages
Accurate monocyte and DC profiling was possible only after exclusion of abundant CD15 1 granulocytes in the PBMC fraction (Fig 3, A). This revealed a profound depletion of CD14 1 classical monocytes, CD16 1 nonclassical monocytes, CD123 1 pDCs, CD141 1 cDC1s, and CD1c 1 cDC2s (Fig 3, A-C, and see Fig  E1 in this article's Online Repository at www.jacionline.org). As recently reported, there was a relative excess of CD56 bright immature NK cells and depletion of CD56 dim mature NK cells (see Fig E1, D). In the dermis CD14 1 monocytes-macrophages, CD141 1 cDC1s, and CD1c 1 cDC2s were similarly decreased, whereas lymphocytes and autofluorescent dermal resident macrophages were preserved. Elective BAL fluid was also depleted of DCs and CD14 1 and CD16 1 monocyte-derived cells, control subjects. MFI, Mean fluorescence intensity. E, Basophils (Baso) identified as Lineage 2 HLA-DR 2 CD45 low CD123 1 cells. Absolute count (BD TruCount) in whole blood is shown. Bars represent means 6 SDs of 18 control subjects. F, Heat map showing differential gene expression of 1.5 log 2 or greater from control mean and z score of 2 or greater between the patient with the 83C/291Q mutation and control subjects (n 5 3). Open circles, Genes differentially regulated by interferon; solid squares, genes bound by IRF8. Pie charts show gene ontology (GO) terms significantly (P < .01) enriched after hypergeometric testing of differentially regulated transcripts by using the nCounter Human Immunology V2 panel as the gene universe. = J ALLERGY CLIN IMMUNOL VOLUME 141, NUMBER 6 but side-scatter (SSC) hi CD45 hi CD206 1 alveolar macrophages and SSC low HLA-DR 2 lymphocytes remained in normal numbers. BAL fluid CD15 1 granulocytes were approximately 10-fold more abundant than in control subjects (Fig 3, C). Proliferating epidermal Langerhans cells were also intact (Fig 3, D). BAL fluid cytokine levels were also grossly disturbed (see Fig  E2, A, in this article's Online Repository at www.jacionline.org).

Dysregulation of granulopoiesis
Myeloproliferation consisted of abundant hypogranular neutrophils with reduced density, low light SSC properties, and reduced mRNA expression of a number of primary and secondary granule components (Fig 3, A, and see Fig E2, B). CD34 1 progenitors, including CD38 lo CD45RA 2 CD90 1 hematopoietic stem cells, were consistently mobilized into the peripheral blood (Fig 4, A). Myelopoietins were increased in serum, including the products of several IRF8-bound genes (IL1RA, PDGFB, CCL4, CCL5, IL1B, and IL10), and expression of cathepsin G (CTSG) was increased (Fig 4, B and C). The neutrophil oxidative respiratory burst, tested in the absence of corticosteroid therapy, was reduced in frequency in response to Escherichia coli and reduced in intensity (mean fluorescence intensity) in response to both E coli and phorbol 12-myristate 13-acetate (Fig 4, D). The patient was also basopenic (Fig 4, E). Many transcripts expressed in neutrophils were differentially regulated compared with those in healthy control subjects, including a number of interferon-regulated and IRF8-bound genes (Fig 4, F, and see Tables E5 and E6 in this article's Online Repository at www.jacionline.org).

Dysregulation of lymphoid gene expression
A total of 125 transcripts were differentially regulated in sorted lymphoid cells (73 upregulated and 54 downregulated; see Fig  E3, A, in this article's Online Repository at www.jacionline. org). A core set of 11 genes varied in expression in all lineages. These included upregulation of the antibacterial protein lactoferrin, NFIL3, and CCR1 and downregulation of TNFAIP3, CCL3, CCL4, HLA-DRA, CXCR4, IRF1, CDKN1A, and DUSP4. More than half of all differentially expressed genes showed interferon-dependent transcription, and approximately a third (45/125) were IRF8 bound according to homology mapping of ChIP-seq data in the mouse. 4 Within the NanoString code set, a number of pathways were significantly differentially regulated (see Fig E3, B, and Table E6). The NanoString panel contained 82 of the DEGs identified by RNA sequencing in the patient with the K108E mutation. Of these, 36 were differentially expressed by both patients, and 15 were IRF8 bound (see Table E7).

Impairment of CD8 1 memory differentiation and T H 1/T H 17 polarization
T-cell numbers (CD4 1 , CD8 1 , and Treg cells) and PHA responses were normal (see Fig E4, A and B, in this article's Online Repository at www.jacionline.org). CD4 memory differentiation was normal (see Fig E4, C), but CD8 1 CD45RA 2 CD27effector memory cells were reduced in number in blood and BAL fluid (Fig 5, A and B, 37 and see Fig E4, D). 38 T H 1 and T H 17 differentiation was impaired and T H 2 differentiation was increased by cytokine secretion and intracellular staining (Fig 5, C and D). Production of GM-CSF by CD4 1 and CD8 1 T cells was increased (Fig 5, C), but CD8 1 T cells also produced significantly less IFN-g (Fig 5, D). T-bet and signal transducer and activator of transcription 1 (STAT1) expression were also decreased in bulk sorted T cells (Fig 5, D). CXCR3 was virtually undetectable (<2% expression) on CD4 1 T cells, CD8 1 T cells, and Treg cells isolated from the blood and BAL fluid of the patient (Fig 5, E, and see Fig E4, E).

Impaired B-cell maturation and B-cell receptor diversification
Although numbers of CD38 1 CD27 2 transitional B cells and CD38 hi CD27 1 plasmablasts were normal, numbers of both IgD 1 CD27 1 and IgD 2 CD27 1 memory cells were reduced (Fig 6, A and B). 39 Total IgG, IgM, and IgE levels were normal, but IgA and IgG 2 levels were reduced (Fig 6, C). Although the patient mounted normal specific IgG responses to childhood vaccines, antibodies to HiB and Pneumococcus species were not durable and decreased to nonprotective levels by the age of 3 years (see Table E3). Deep sequencing of the immunoglobulin heavy chain (IgH) CDR3 was performed on B-cell DNA isolated from both patients with the R83C/R291Q and those with the K108E/K108E mutations. Clonality and successful template rearrangement were normal (see Fig E5, A and B, in this article's Online Repository at www.jacionline.org), but CDR3 length was shorter in both productive and nonproductive templates of the patients (Fig 6, D, and see Fig E5, C and D), which is consistent with reduced numbers of nontemplated N1 and N2 nucleotide insertions (Fig 6, E). Both patients had proportionally fewer productive templates that had undergone somatic hypermutation (Fig 6, F) and a truncated distribution of somatic hypermutation counts with fewer mutated bases per template (Fig 6, G). A survey of VH gene expression identified restricted gene use compared with age-matched control subjects (Fig 6, H).

DISCUSSION
This report describes the hematologic defects and immune dysregulation caused by compound heterozygosity of 2 new defective IRF8 alleles, R83C and R291Q, the latter orthologous to R294C, a hypomorphic mutation responsible for the BXH2 mouse phenotype. Similar to the previously reported child carrying IRF8 K108E/K108E , the patient, who was studied in the absence of active infection, had granuloproliferation with profound DC and monocyte deficiency. Further analysis revealed dysregulation of B, NK, and T cells. Although it cannot be proved that lymphoid defects are due to intrinsic loss of IRF8, a number of findings are in accordance with known direct effects of IRF8. These observations shed new light on the mechanisms of immunodeficiency in patients with biallelic IRF8 mutations, which now include maturation and functional deficits of NK cells; deficiency of T H 1, T H 17, and CD8 effector memory responses; and blunted somatic hypermutation, class-switching, and memory B-cell formation.
At the phenotypic level, analysis of this compound heterozygous patient suggests that susceptibility to viral infection, myeloproliferation, and developmental delay associated with cerebral calcification are consistent features of biallelic IRF8 mutation because they were seen similarly in the patient with homozygous K108E mutation. 31 The previous case also    37 analyzed as a percentage of total CD8 1 T cells. B, Proportion of CD27 2 CD45RA 2 CD8 1 T cells in the PB of the patient (black column) versus control subjects (gray column) and BAL fluid of the patient (black column) versus published mean of healthy control subjects (star). C, Cytokine secretion into culture medium, as determined by using Luminex technology by purified CD4 1 and CD8 1 T cells from the patient and 4 healthy control subjects after phorbol 12-myristate 13-acetate (PMA)/ionomycin stimulation. Bars represent means 6 95% CIs. D, Intracellular cytokine production by purified CD4 1 and CD8 1 T cells from the patient and 4 healthy control subjects after PMA/ionomycin stimulation and selected gene expression analysis in the patient with the 83C/291Q mutation (black) versus 5 age-matched control subjects, as determined by using NanoString nCounter technology (Human Immunology V2 panel). Bars represent means 6 95% CIs. E, Flow cytometric analysis of CXCR3 expression on CD4 1 , CD8 1 , and CD4 1 CD127 2 CD25 1 Treg cells from a representative control subject and the patient, with the percentage of CXCR3 1 cells indicated. Summary of CXCR3 expression in blood (filled circles) and BAL fluid (open circles) CD4 1 , CD8 1 , and CD4 1 CD127 2 CD25 1 Treg cell subsets from the patient with the 83C/291Q mutation versus 6 healthy control subjects (gray circles). Bars represent means 6 95% CIs. *P < .05, **P < .01, and ***P < .001. The left column shows populations defined by IgD and CD27: naive (gate 1), nonswitched memory (gate 2), switched memory (gate 3), and CD27 memory (gate4). The right column shows populations defined by CD38 and CD27 to identify transitional (gate 5), naive mature (gate 6), mature activated (gate 7), memory (gate 8), and plasmablastic (gate 9) cells. B, PB B-cell subsets defined by CD27 and IgD expression as a proportion of total B cells from the patient with 83C/291Q mutations (gray outlined dots) and 3 local agematched control subjects (gray triangles) plotted against an age-specific (18 months to 4 years) normal range. 39 Bars represent means and ranges. C, Serum immunoglobulin isotype levels (IgG, IgA, IgM, and IgE) over time. Horizontal lines represent upper and lower limits of normal. The right plot shows serum levels of IgG subtypes at age 6 months. Gray bars show the reference range. D-H, B-cell receptor (BCR) IgH CDR3 region sequencing of genomic DNA from purified PB B cells (Adaptive Biotechnologies' Immuno-SEQ Assay) from the patient with the 83C/291Q mutations (red), the patient with the K108E mutations (blue), and 3 age-matched control subjects (gray). Fig 6, D, Summary of template generation showing the percentage of productive (Prod), out-of-frame (OoF), or stop templates generated. Mean CDR3 length (nucleotides; Fig 6, E) and mean number of inserted untemplated nucleotides (N1 and N2; Fig 6, F) in out-of-frame and inframe rearrangements. Fig 6, G, Percentage of templates with 1 or more mutated bases. Fig 6, H, Frequency of mutated bases per in-frame template expressed as the percentage of total mutations. Bars represent means and SDs (Fig 6, E-H). Statistics were calculated by using t tests as follows: *P < .05 and **P < .01.
indicates high susceptibility to mycobacterial infection. Although the compound heterozygous patient was less critically ill, having not been exposed to BCG vaccine in infancy, immunodeficiency was severe enough to warrant hematopoietic stem cell transplantation at age 4 years.
The novel IRF8 variants, affecting conserved residues, were functionally defective in vitro. R83, which is in the DNAbinding domain of the protein, forms hydrogen bonds with the guanine of the GAAA core. 40,41 R83 belongs to one of the 2 basic residue clusters that form the bipartite nuclear localization signal of IRF8, 42 explaining the defective nuclear translocation of R83C in cellular localization assays, a feature shared with K108E mutant protein. 32 R291 is one of only 12 invariant residues in the IADs of IRFs. Crystal structures of IRF3 and IRF5 show that this residue, lying between helix 2 and helix 3, directly interacts with Ets factor and IRF partners in the formation of DNAbinding dimers. 43 Mutation of the orthologous murine residue 294 in the BXH2 mouse blocks the function of IRF8 in complementation assays and affects the ability of IRF8 to interact with other TFs through the IAD. 44 Luciferase reporter assays and EMSAs showed that both variants were unable to activate EICE with PU.1 or SPIB and were unable to compete with IRF1 at the ISRE. R83C was unable to bind to BATF/JUN at the AICE but R291Q maintained activity. Although this suggests that R291Q might function effectively as a partner of BATF to facilitate the terminal stages of cDC1 development, 21 this cannot be interrogated because the failure to coactivate SPI1/PU.1-directed genes blocks DC and monocyte development at an earlier stage. Taken together, the molecular characteristics of the mutations and the phenotypic similarities with the patient carrying null K108E mutations suggest that compound heterozygous R83C and R291Q mutations result in IRF8 activity of less than the threshold for function. Because IRF8 contains a superenhancer region, gene dose is likely to be critical for normal function.
The mechanism of granuloproliferation in patients with IRF8 deficiency might be multifactorial, with increased production and mobilization but reduced apoptosis of progenitors and neutrophils. First, through cooperation with PU.1 and inhibition of CEBPA at the molecular level, IRF8 is responsible for the balanced generation of myeloid cells in the granulocytemonocyte progenitor compartment. 45 Spontaneous granuloproliferation caused by the unopposed action of CEBPA is a consistent finding in both patients with the biallelic IRF8 mutations, and BXH2 and Irf8 2/2 mouse models.
Second, it was observed that myeloid progenitors, including hematopoietic stem cells, were mobilized into the blood. CXCL12/CXCR4-mediated retention in the bone marrow might have been attenuated because of increased expression of cathepsin G, which degrades CXCL12, and reduced expression of CXCR4, an IRF8 target gene. 46,47 Levels of growth factors known to mobilize progenitors, including stem cell factor, 48 hepatocyte growth factor, 49 and vascular endothelial growth factor, 50 were also increased.
Finally, Irf8 2/2 progenitors are hyperresponsive to granulocyte colony-stimulating factor 51 and defective in their inability to upregulate BAX and Fas ligand2mediated apoptosis during termination of emergency myelopoiesis. 52,53 In keeping with this, apoptosis-related genes were downregulated in the patient's neutrophils.
The expanded neutrophil population was comprised of cells reminiscent of those seen in myelodysplasia; hypogranular with low-light SSC properties, 54 defects in primary and secondary granules, 55 and defective respiratory burst responses. 56 Excessive granulopoiesis appeared to contribute to respiratory complications. At rest, the BAL fluid contained high numbers of granulocytes associated with increased inflammatory cytokine and chemokine levels, including TNF-a, IL-8, CCL2, CCL3, CXCL12, IL-6 and IL-1b. During respiratory tract infection, treatment with glucocorticoids exacerbated the neutrophil response, often leading to a clinical deterioration, whereas treatment with DNAse was effective, which is consistent with lysis of excessive extracellular neutrophil DNA. 57 In contrast to the neutrophils, basophils were reduced, which is in keeping with a known function of IRF8. 58 A much broader cytopenia encompassing all monocytes and DCs was seen in both biallelic human subjects compared with the full Irf8 knockout mouse, which still retains some monocytes and pDCs and has intact cDC2s. 20 The homozygous R294C BXH2 mouse and heterozygous Irf8 knockout mouse are only deficient in cDC1s, which require IRF8 for terminal differentiation. 20 Such selective defects have not yet been observed in human subjects. However, other alleles of IRF8 have been described that do not impair DC and monocyte development to the same degree. Notably, the heterozygous T80A mutation, although described as dominant negative in reporter assays, had a very modest effect on DC development with apparent loss of cDC2s. 31 This pattern is unexpected based on mouse models in which cDC2s are always preserved, but the appearance of atypical CD11c 1 CD1c 2 cells in the human subject suggests that T80A disturbs hematopoiesis by an allele-specific mechanism that has not yet been elucidated. Additional IRF8 variants recently described by Mace et al 25 also showed more subtle deficiency in monocytes and DCs but pronounced effects on NK cell development. These variants were localized to the IAD (P224L and A201V). Neither allele was compromised in reporter assays by using the EICE with PU.1 or SPIB or the ISRE with IRF1, suggesting that the loss of protein-protein interactions was responsible for lineage-restricted functional deficits. It is highly likely that other IRF8 variants will come to light with allele-specific effects on hematopoiesis and immunity. Precise interspecies mapping might not always be possible because of the complexity of interactions between IRF8, IRF4, and other TFs.
Preservation of Langerhans cells and macrophages in states of severe monocytopenia, such as GATA-2 and IRF8 deficiency, remains a key demonstration of the ability of human tissue macrophages to survive without continual replenishment by monocytes. 31,59 These observations were extended to the alveolar space, which contained large numbers of macrophages. At least partial preservation of macrophages and Langerhans cells is seen in the BXH2 and Irf8 2/2 mouse models, 60,61 although functional deficits are apparent. 4,12,62 The contribution of macrophage dysfunction to the human phenotype of IRF8 deficiency remains to be determined, in particular whether the developmental delay and intracerebral calcification seen in patients bearing IRF8 K108E/K108E and IRF8 R83C/R291Q are related to the reported role of Irf8 in murine microglial development and function. 63,64 There was no evidence for disordered calcium metabolism, but we cannot exclude infectious complications or dysregulated interferon signaling as contributors to this phenotype.
Lymphocyte development and function in the setting of IRF8 deficiency can be altered through both direct cell-intrinsic effects and indirect effects because of defective hematopoiesis and external influences, such as infection. Analysis of the lymphoid compartment showed transcriptomic dysregulation with phenotypic and functional defects in NK, T, and B cells.
Transcriptomic analysis of purified lymphocyte populations revealed enrichment within upregulated genes for response to the pathogen pathways (LTF, CXCL1, CD27, and TNFRSF) and immune response (NFIL3 and CCR1), with downregulated genes involved in cytokine signaling (including IFN-g), cytokine production (CCL3 and CCL4), Toll-like receptor (TLR) signaling (including TLR3, TLR4, and TLR5), cell activation (HLA-DR), and regulation of cell cycle and apoptosis (DUSP4 and CDKN1A). Aberrations in TNF signaling were apparent with upregulation of both TNF and TNF receptor superfamily members but universal downregulation of IRF8-bound TNF-induced TNFAIP3, a negative regulator of cytokinemediated immune and inflammatory responses. IRF1 expression was also reduced, further contributing to loss of gene expression governed by the IRF8/IRF1 regulome. 4 Defects in NK cell maturation and function were reported recently as an example of more widespread immune dysregulation caused by IRF8 mutation by using data from the IRF8 K108E/K108E and IRF8 R83C/R291Q patients. Both show an increased proportion of immature CD56 bright NK cells, suggesting impaired NK maturation. Defective NK cell-mediated cytotoxicity was also reported in a patient with mutations in the IAD region (201V/224L) who had severe EBV infection without significant DC or monocyte deficiency. 25 In the T-cell lineage profound defects in T H 1, T H 17, and CD8 effector memory development were observed, together with almost complete absence of CXCR3, the chemokine receptor for CXCL9, CXCL10, and CXCL11, which plays a vital role in the positioning of activated T cells in tissues. 65 IRF8 is upregulated during T-cell activation, 28 and experiments with Irf8 2/2 antigen-specific T cells indicate a number of cell-intrinsic regulatory roles for IRF8, including integration of g-chain cytokine and T-cell receptor signaling pathways, secretion of IFN-g and cytotoxicity. 26 IRF8 also regulates STAT1 through mutual promoter binding 66,67 and represses the GM-CSF gene in T cells. 68 In Itgax (CD11c) conditional Irf8 knockouts, T-cell dysfunction was proposed to be largely extrinsic through the loss of cDC1, although recombination was also observed in lymphoid cells, which express CD11c when activated. 69 Evidence for an intrinsic role of IRF8 in T H 17 differentiation is conflicting. 69 It is likely that the strength of T-cell receptor signaling is compromised significantly both by the lack of DCs and intrinsic defects including attenuated STAT1-mediated signaling, especially critical for IFN-g-driven T H 1 differentiation. Both contribute potentially to the failure of T H 1, T H 17, and CD8 effector memory differentiation and lack of CXCR3 expression, leading to compromised peripheral tissue T-cell homeostasis. Transcriptomic analysis of purified T cells showed global downregulation of pathways involved in cell signaling, cytokine production, and cell activation, which is consistent with gene expression analysis of whole PBMCs by RNA sequencing in the IRF8 K108E/108E patient. 32 In B cells lineage-specific knockout in mice indicates that IRF8 is required at several stages of B-cell development, including maintenance of central tolerance in the bone marrow, differentiation of follicular B cells, germinal center formation, antibody affinity maturation, and memory cell formation. 7,23 Correlates of these intrinsic effects were observed in humans. In both IRF8 R83C/R291Q and IRF8 K108E/K108E patients, the extent and complexity of somatic hypermutation was reduced, a factor directly associated with serious respiratory infection in patients with common variable immunodeficiency. 70 Analysis of CDR3 sequences in both patients compared to age-matched control subjects showed shorter lengths with fewer N1 and N2 insertions. Although terminal deoxynucleotidyl transferase, which is responsible for N insertions, is not known to be a direct target of IRF8, impairment is seen in other conditions with perturbed hematopoiesis and aberrant B-cell development. 71 The restricted VH gene repertoire might also reflect abnormal V(D)J recombination within the bone marrow 72 or represent defective secondary receptor editing. 73 Although shorter in-frame CDR3 regions were appropriately selected for compared with out-of-frame sequences, suggesting intact central B-cell tolerance, 74 the presence of widespread anti-immunoglobulin reactivity across a human proteome microarray was striking in the serum of the patient carrying IRF8 R83C/R291Q compared with age-matched control subjects. The clinical significance of autoreactive antibodies in this age group is uncertain, but in the absence of autoimmune clinical symptoms, the anti-immunoglobulin activity might represent an appropriate response to recurrent infections. 75 Recent data indicate that IgA deficiency might be explained by a lack of DC-B-cell interaction in Peyer patches. 76 In summary, this report defines 2 new variants of IRF8 associated with defective function in a compound heterozygous patient. Profound cytopenia of DCs and monocytes in the context of granuloproliferation was confirmed as a consistent feature of immunodeficiency compounded by intrinsic and extrinsic dysregulation of lymphoid cell development, maturation, and responses. These observations are broadly consistent with the notion that IRF8 controls a set of multilineage functions that hone protective immunity to viral infection and intracellular pathogens. These range from governing the hematopoietic development of cross-presenting cDC1s, interferon-producing pDCs, and monocytes, to the fine-tuning of NK maturation, T H 1 and T H 17 cell polarization, effector memory development, and durable production of high-affinity class-switched antibody. This is achieved through multilevel and multilineage control of gene expression, the failure of which is mirrored by a complex syndrome of immunodeficiency.