Interferon regulatory factors (IRF) regulate gene expression and immune responses to interferon, and possess N-terminus DNA binding domains (DBD), and C-terminus IRF-association domains (IAD), the latter allowing for dimerization and cofactor interaction (1–3). Mouse models have elucidated the functional roles of interferon regulatory factor 8 (IRF8) in monocyte and dendritic cell differentiation, and inhibition of neutrophil development (1, 4, 5). Irf8 knockout mice show increased viral infections, reduced IL-12p40 and IFN-gamma production, neutrophil expansion and a blast crisis phenotype, and lower numbers of pDCs and CD8α+ mDCs, but not CD11b+DCs (6–8). Mice with IRF8 IAD mutations affecting dimerization retain monocytes, macrophages and pDCs, but develop a chronic myeloid leukemia-like disease and cannot produce IL-12 or IFN-gamma (9).

In mice, IRF8 promotes monocyte and dendritic cell differentiation via its interaction with PU.1 and induction of KLF4. Neutrophil development is hindered by interfering with the transcription factor CEBPA binding chromatin in monocyte-DC progenitors. IRF8 is also critical for pDC function, and terminal differentiation of cDC1 via BATF3 interaction (10, 11). It is essential for IL-12p35, Il-12p40 and IL-18 expression, and its absence impairs Th1 responses, increasing intracellular pathogen susceptibility in Irf8-/- mice (11–15).

Currently, nine human cases of IRF8-related disease are reported, five with autosomal recessive inheritance, and four with 3 de novo dominant negative variants, including a mother-child dyad. Two phenotypes emerge in these reports: an autosomal recessive severe immunodeficiency with impaired myeloid lineage, and a dominant negative form with decreased cDC2s and mycobacterial susceptibility (16). NK cell developmental and functional defects were additionally described in one autosomal recessive case (17).

The first IRF8 deficiency patient reported had biallelic variants leading to disseminated BCG (Mendelian Susceptibility to Mycobacterial Disease (MSMD), and fungal infections (18, 19) They had deficient monocytes and dendritic cells with failure to produce either IL-12 or IFN- γ upon LPS stimulation (13, 20). Subsequent autosomal recessive IRF8 deficiency cases were also reported to have recurring viral infections, BCG susceptibility, absent or decreased monocyte and DC populations, and decreased IL-12 and IFN-γ production (21, 22). In contrast, dominant negative patients showed reduced CD1c+ DCs and IL-12 production, while also demonstrating MSMD (22). Recently, a dominant negative variant was described in mother and child, with decreased pDC and cDC1, but normal cytokine production.

We describe an individual with previously unreported heterozygous IRF8 variant causing a frameshift mutation in the last exon, and cellular deficiencies consistent with previous reports, leading us to conclude the identified IRF8 variant is likely pathogenic. We juxtapose this against a second patient with a different, also previously unreported heterozygous IRF8 variant, with a less convincing clinical and immunologic presentation for inborn error of immunity (IEI), and where the cellular phenotype tested was unaltered. We propose that a combination of specific cellular testing could help interrogate IRF8 variants of uncertain significance (VUS).

Consent from P1 was obtained under IRB 2008–0483 at Cincinnati Children’s Hospital and for P2 under IRB 2018–3937 at McGill University.

Detailed methodologies are provided in the supplemental document.

In P1, the c.1182dup indel in exon 9/9 causes a frameshift at Glu395 replacing the native 27-aa C-terminus with a 75-aa neo-tail and introduces a new stop codon within the last coding exon. Under the canonical last-exon/50-nt rules, the transcript is predicted to escape nonsense mediated decay, suggesting production of an elongated IRF8 protein with an altered C-terminus (24–26). This variant was de novo for P1 and not found in GnomAD ( Figure 1B ). Table 2 lists all variants reported.

For P2, a de novo heterozygous missense variant in IRF8, c.10C>T (p.Arg4Trp) was identified. This variant is present 3 times in gnomAD, and is predicted “disease causing” by MutationTaster and “probably damaging” (score 1.000) by Polyphen-2 ( Figure 1C ). Genetic testing also identified a heteroplasmic VUS in the mitochondrial genome gene MT-TC, (m.5819T>C), present in 5.3% of reads; the variant was also identified in P2’s mother, present in 1.5% of reads (27).

P1 had chronic mild monocytosis ( Table 1 ) and an absence of IL-3Ra/CD123 expressing DCs, while CD11c positive DCs were comparable to controls ( Figures 1D, E ). P2, run on a different occasion, showed DCs comparable to simultaneously-run controls ( Figure 1E ). P1 showed significantly elevated CD21low B cells (15%, Table 1 ). This is observed in immune deficiencies, especially those with autoimmunity as well as hypogammaglobulinemia, and have been suggested to indicate chronic activation of the adaptive immune system (28–31). This supports the postulation of an immune-mediated chronic disease in P1. P1 also showed reduced CD31+ recent thymic emigrant T cells (9%, Table 1 ) (32). These are highly proliferative and replenish the peripheral T cell pool. The significantly reduced level in P1 at that age indicates lowered thymic output, suggesting immune exhaustion (33, 34).

We tested functional cell responses via stimulation with LPS targeting monocytes and DCs, PHA targeting T cells, or mock stimulation. P1, P2, and the mother of P2 showed high cytokine levels when unstimulated, suggesting some pre-existing inflammatory condition, with increased baseline secretion of IL-1β, IL-6, IL-10, and TNF-α compared to control ( Figures 2A, D ). Upon LPS stimulation, P1 had comparable levels of IL-1β, IL-6, and IL-10 upregulation, reduced IFN- γ and TNF-α, and undetectable IL-12p70 ( Figure 2B ). PHA stimulation showed comparable IL-1β and IL-6 levels but low IFN-γ, and no increase in other readouts ( Figure 2C ). P2 and their mother showed reduced IL-12p70 levels when stimulated with LPS ( Figure 2E ). There were comparable values for all readouts upon PHA stimulation ( Figure 2F ).

We present two patients with novel monoallelic IRF8 variants – c.1182dup (p.Glu395Argfs*75) and c.10C>T (p.Arg4Trp) – the former with a clinical presentation strongly suggestive of an underlying IEI, and the latter with a clinical presentation that was plausible for, but less strongly indicative of an IEI. We investigated whether these variants were the likely cause of the patients’ clinical features.

P1’s clinical and immunologic features strongly suggested a monogenic immune disorder, but clinical exome testing only revealed several VUS. Thus cellular studies were performed. We found increased neutrophils and monocytes, an absence of IL-3Ra/CD123 DCs, a high percentage of memory T cells, low recent thymic emigrants, expanded T follicular helper cells, and mostly naïve B cells with increased CD21low expression reminiscent of severe immune dysregulation patients (28, 35). All other lymphocyte populations studied were comparable to controls tested, including NK cells (17). Upon LPS stimulation, which targets monocytes and DCs, secreted IL-12p70, and, to a lesser degree, IFN- γ, were deficient (13, 14). Upon PHA stimulation, IL-10, IL-12p70, and TNF-α were not upregulated, indicating additional T cell response deficiencies, reminiscent of previously reported IRF8 disease cases (20, 36).

The IRF8 transcription factor is essential for monocyte and DC development and function, and in this case, specifically CD123+ DCs (35, 37–39). AlphaFold/TED places the IRF8 DNA binding domain (aa 6–118) and IAD (aa 201–384) upstream of the variant. Prior work shows that this tail contributes to autoinhibitory control of the IAD and to partner-regulated chromatin engagement (18, 36). Consistent with this, C-terminal IRF8 mutations (e.g. G388S) and engineered C-tail extensions can diminish promoter binding, impair nuclear localization, and act dominant-negatively (24, 25, 40). We infer that p.Glu395Argfs*75 leads to loss of normal IRF8 function by disrupting C-terminal regulatory cues that govern cooperative binding with PU.1 and C/EBPα and the transcription of IRF8 targets. IRF members, together with PU.1, also upregulate B cell related genes such as CD20 and Ig light chain enhancers, possibly explaining the naïve B cell phenotype, impaired class switching, and reduced immunoglobulin secretion via CD70 surface expression (41–44). The total absence of pDCs in P1 was more reminiscent of the K108E, R83C/R291Q, or R111* autosomal recessive cases, than the milder T80A dominant negative cases, which demonstrated selective depletion of CD11c+CD1c+ DCs (18, 22, 45, 46).

Importantly, P1 received BCG vaccination in infancy and yet had no early or delayed adverse reaction to this live vaccine, with negative staining for acid-fast bacillus (AFB) on repeated tests, on multiple tissue samples. The monoallelic IRF8 variant identified was therefore first viewed skeptically, as MSMD, including disseminated infection with minimally pathogenic mycobacteria like Bacillus Calmette-Guérin, was reported as a hallmark feature in cases of autosomal dominant and recessive IRF8 deficiency (18, 21). Indeed, environmental mycobacteria and BCG vaccines are usually eliminated effectively by macrophages activated by T-cells in healthy hosts, and previously, it was demonstrated that individuals with impaired mononuclear phagocyte development due to IRF8 variants were vulnerable to disseminated BCG in infancy or childhood, following vaccination at birth (18, 47). P2 had no history of having received BCG, nor did they present with evidence of mycobacterial disease. However, they did have a history of early life repeated viral infection, albeit less dramatic when compared to P1. P2 also presented with failure to thrive, with associated inflammatory changes in the duodenum involving epithelioid histiocytic inflammation without well-formed granulomas.

For P2, AlphaFold/TED places the IRF8 DNA-binding domain (aa 6–118) downstream of Arg4. The R4W substitution occurs at the N-terminus of the structured fold. The Arg to Trp change is predicted to perturb local electrostatics and DNA-contact geometry. Splicing prediction indicates altered exonic splicing regulatory elements with ~27% risk of impact (SPiP score 0.224) (48). Complementary in silico metrics show missense impact (CADD phred 23.7; MPA ‘moderate’), while splice-site–focused tools are largely negative (SpliceAI AG/AL/DG/DL all 0.00; AbSplice <0.01). The variant lies 11 bp from the acceptor of exon 2 and outside a UniProt-defined domain, extremely rare in gnomAD v4 (genomes ~1.3×10⁻⁵; exomes ~6.8×10⁻⁷; no homozygotes), and is currently a VUS in ClinVar. Taken together, c.10C>T could compromise IRF8 function through combined structural perturbation of the DBD N-terminus and carries a non-negligible risk of splicing dysregulation, but more data are needed to more definitively characterize the effects of this variant (49).

Significant heterogeneity in IRF8 patients has been documented, highlighted by a family with a significant reduction in NK cell number and function, and another with periodontal disease, indicating that there are yet to be determined roles of IRF8 (17, 40, 50, 51). While P1 did display reduced NK numbers over the years, NK phenotype was comparable to controls, although NK function was not tested. Dominant negative IRF8 variants (c.1279dupT) have most recently also been noted to express reduced pDCs but normal cytokine expression (45). While we cannot rule out the missense variant effect in P2, with the available tests performed interrogating DC numbers and cytokine production after stimulation, we did not find changes to support immune impact of this variant similar to previously reported cases. However, P2’s early life history of repeated viral infections, failure to thrive, and poorly formed granulomatous changes on duodenal biopsy are notable, and do raise the question of an underlying IEI, perhaps related to their IRF8 variant causing an as of yet undefined immune impact.

In summary, we report two patients with novel heterozygous de novo IRF8 variants (c.1182dup and c.10C>T). Data presented suggests the former variant is likely pathogenic, with features similar to both autosomal recessive and autosomal dominant IRF8 cases, including absent pDCs and lack of IL-12p70 production. However, the absence of mycobacterial infections despite mycobacterial exposure is unlike other reported cases thus far, and either indicates a wider disease heterogeneity or a confounding factor yet to be determined. P2, with less convincing symptoms of IEI and with apparently unaltered cellular phenotype, demonstrates the utility of performing combination tests to interrogate the pathogenicity of unknown variants. Further functional characterization of the two presented variants will be beneficial to better understand their impact on human immunity in our patients.