Humans homozygous for rare or common hypomorphic IL23R variants are prone to tuberculosis

Mendelian susceptibility to mycobacterial diseases (MSMD) is characterized by selective susceptibility to clinical disease due to weakly virulent mycobacteria, such as Bacille Calmette Guérin (BCG) vaccine substrains and environmental mycobacteria (EM) (Boisson-Dupuis et al., 2024; Bustamante, 2020; Casanova and Abel, 2002; Rosain et al., 2018). The genetic etiology of MSMD is known in about half the patients (Casanova and Abel, 2022). Since 1996, 47 genetic disorders involving 22 loci have been reported (Casanova and Abel, 2002; Casanova and Abel, 2022; Rosain et al., 2018; Rosain et al., 2023; Bustamante, 2020; Yang et al., 2020; Martin-Fernandez et al., 2022; Bohlen et al., 2023; Philippot et al., 2023; Boisson-Dupuis et al., 2024; Neehus et al., 2024; Casanova, 2025; Casanova, 2026). All but two of these disorders affect genes involved in IFN-gamma (IFN-γ) production by lymphocytes, responses of mononuclear phagocytes to IFN-γ, or both. CCR2 deficiency affects monocyte migration (Neehus et al., 2024), whereas the mechanism by which inherited ZNFX1 deficiency causes MSMD remains unclear (Le Voyer et al., 2021). The two most common etiologies of MSMD are autosomal recessive (AR) complete IL-12Rβ1 and IL-12p40 deficiencies (Alodayani et al., 2018; Altare et al., 1998a; Altare et al., 1998b; de Beaucoudrey et al., 2010; de Jong et al., 1998; Fieschi et al., 2003; Khavandegar et al., 2024; Melo et al., 2024; Prando et al., 2013). IL-12 is a heterodimer composed of the p35 and p40 subunits that binds to a heterodimeric receptor composed of IL-12Rβ1 and IL-12Rβ2. Remarkably, p40 can also dimerize with p19, forming IL-23, which binds to a heterodimeric receptor composed of IL-12Rβ1 and IL-23R (Oppmann et al., 2000; Parham et al., 2002). Studies of mouse and human IL-12 and IL-23 have suggested that IL-12 is the signature IFN-γ–inducing T_H1 cytokine, whereas IL-23 is the signature IL-17–inducing T_H17 cytokine (Acosta-Rodriguez et al., 2007; Cua and Tato, 2010; Fieschi and Casanova, 2003; Teng et al., 2015; Wilson et al., 2007).

However, patients with AR complete IL-23R deficiency present with MSMD (Martinez-Barricarte et al., 2018; Philippot et al., 2023; Staels et al., 2022); the mechanism involves impaired IL-23–dependent IFN-γ production, particularly by MAIT and Vδ2⁺ γδ T cells (Philippot et al., 2023). Chronic mucocutaneous candidiasis (CMC) in patients with complete IL-23R deficiency has incomplete penetrance due to the low contribution of IL-23–induced IL-17A/F-dependent immunity to CMC (Martinez-Barricarte et al., 2018; Philippot et al., 2023; Staels et al., 2022). In parallel, we reported an enrichment in homozygosity for the common TYK2 P1104A allele in cohorts of patients of European descent with tuberculosis (TB) relative to ethnicity-adjusted controls, and that homozygosity for the TYK2 P1104A variant selectively impairs, but does not abrogate, the IL-23–dependent induction of IFN-γ (Boisson-Dupuis et al., 2018; Kerner et al., 2019; Kerner et al., 2021). Consistently, we found that impaired IL-23–dependent induction of IFN-γ is the only mechanism of mycobacterial disease common to patients with any of the five forms of AR TYK2 deficiency (Ogishi et al., 2022). We also found that the impaired IL-23–dependent induction of IFN-γ underlies mycobacterial disease in X-linked recessive MCTS1 deficiency (Bohlen et al., 2023) and AR ITK deficiency (Ogishi et al., 2023). Overall, these findings challenge the classical IL-12–T_H1/IL-23–T_H17 dichotomy, suggesting that human IL-23 plays an essential role in IFN-γ–mediated host defense against mycobacteria, with complete and partial deficiencies underlying MSMD and TB, respectively. In this context, we hypothesized that other biallelic IL23R variants—null or hypomorphic, rare or common—may underlie MSMD or TB.

We screened our cohort of 24,046 patients with various infectious diseases, including 1,874 patients with TB (pulmonary and extrapulmonary) and 901 patients with MSMD, and cohorts of individuals from the general population (gnomAD v4.1.0, Analysis Tool for Annotated Variants Data Base (ATAVDB), Great Middle East, Iranome, Turkish Variome, and Browse All Variants Online (BRAVO) variant server) for biallelic coding non-synonymous or essential splice site IL23R variants. We identified 19 rare (minor allelic frequency [MAF] < 0.01) and four common (MAF > 0.01) variants (Table 1). Four of these variants (C115Y, c.367+1G>A, E269*, and c.1149-1G>A) had previously been reported to be loss-of-function (LOF) (Martinez-Barricarte et al., 2018; Philippot et al., 2023). We assessed the functional impact of each of the 19 remaining variants—including G300V and V362I, as well as an allele carrying both G300V and V362I—in an overexpression assay. Indeed, individuals homozygous for IL23R G300V in our in-house cohort were also homozygous for the IL23R V362I variant. We transiently transfected human embryonic kidney 293T (HEK) cells with a plasmid containing the WT IL12RB1 cDNA under the control of the PGK promoter, together with a plasmid containing either WT or mutant IL23R cDNA under the control of a UbC promoter. In contrast to previous reports of IL23R LOF variants expressed under the control of strong CMV or EF1a promoters (Martinez-Barricarte et al., 2018; Philippot et al., 2023; Staels et al., 2022), we used the weaker UbC promoter to facilitate assessment of the impact of hypomorphic variants (Fig. 1 A). The HEK cells were cotransfected with a plasmid containing a luciferase reporter gene under the control of five sis-inducible elements and a plasmid encoding Renilla luciferase. Transfection efficiency was similar for HEK cells transfected with WT or mutant IL23R cDNAs, as indicated by the proportion of mCherry⁺ and IL-12Rβ1⁺ cells and the similar levels of IL23R and IL12RB1 mRNAs (Fig. S1 A). We then evaluated the response to stimulation with 10 ng/ml IL-23 by assessing luciferase activity. HEK cells transfected with the mutant IL23R cDNAs had activity levels similar to (Q3H, A55T, R86Q, K150T, V159M, V160A, L193F, A199V, S221F, P306S, L310P, V362I, I373F, Q487H, and S559R) or lower (G300V, R381Q, G149R, and L372F) than that of cells transfected with WT IL23R cDNA (Fig. 1 A) or had no activity at all (C115Y, as a representative of known complete LOF variants). The allele carrying both G300V and V362I behaved like G300V, and the degree of hypomorphism could be ranked as follows, from low to high levels of luciferase activity: L372F < G300V < G149R < R381Q. STAT3 phosphorylation following IL-23 stimulation was assessed by western blotting of lysates prepared from HEK cells transfected with IL12RB1 and either WT or mutant IL23R cDNAs. HEK cells transfected with the four mutated IL23R cDNAs (G300V, R381Q, G149R, and L372F), as identified in the luciferase assay, had lower levels of STAT3 phosphorylation following IL-23 stimulation than cells transfected with the WT IL23R cDNA. No STAT3 phosphorylation upon IL-23 stimulation was detected with the known LOF variant C115Y (Fig. 1 B, Table 1, and Fig. S1 B). Overall, 15 variants were found to be isomorphic, four were hypomorphic (≤50% of WT activity), and the previously reported C115Y variant (Martinez-Barricarte et al., 2018) was LOF for cellular responses to IL-23 in this overexpression setting.

The consensus negative selection score (CoNeS = −0.08) and the gene damage index (GDI = 3.6) of IL23R are similar to those of IL12RB1 (CoNeS = 0.45; GDI = 3.79), IL12RB2 (CoNeS = 1.02; GDI = 4.05), and TYK2 (CoNeS = −0.07; GDI = 7.4). These values are consistent with deleterious variants of the gene having the potential to underlie an AR inborn error of immunity (Itan et al., 2015; Rapaport et al., 2020). No biallelic pLOF IL23R variants were found in public databases (gnomAD v4.1.0, ATAVDB, Great Middle East, Iranome, Turkish Variome, and BRAVO). The hypomorphic G300V and L372F variants are rare in the general population (frequency in gnomAD v4.1.0 = 0.0024% and 0.004%, respectively), with the highest frequencies recorded in populations of Middle Eastern ancestry for G300V (frequency in this population in gnomAD v4.1.0 = 0.032%) and East Asian ancestry for L372F (frequency in this population in gnomAD v4.1.0 = 0.118%) (Fig. 1 C, Fig. S2 A, and Table 1). In the Turkish Variome database (Kars et al., 2021), the G300V variant was reported only once in the homozygous state, and there were no individuals homozygous for L372F. In our own database (Human Genetic of Infectious Diseases [HGID] database), L372F is absent, and G300V was reported in the homozygous state in two index individuals of European ancestry. We estimated that the most recent common ancestor carrying the G300V variant lived about 3,078 years ago (95% confidence interval [CI]: 972–11,500 years). The G149R variant is more frequent than G300V and L372F (frequency in gnomAD v4.1.0 = 0.616%), and its frequency is highest in East Asian populations (frequency in this population in gnomAD v4.1.0 = 6%), with a total of 121 homozygotes identified in gnomAD v4.1.0 (0.01%, mainly in East Asian and non-Finnish Europeans) and two unrelated individuals identified in our HGID cohort. The R381Q variant is the most frequent of the hypomorphic variants (frequency in gnomAD v4.1.0 = 5.4%), with the highest frequency observed in Amish, Ashkenazi Jewish, and non-Finnish European populations (frequency in these populations in gnomAD v4.1.0 = 10, 7.2, and 6.4% respectively). It was reported in the homozygous state in 2,839 individuals (0.4%) in the gnomAD v4.1.0 database and in 100 individuals (0.4%) in our HGID cohort (Table S1). Overall, the cumulative frequency of the hypomorphic homozygous variants G149R and R381Q in gnomAD v4.1.0 was 0.4%, ranging from 0.02% in the African population to 1.3% in the Amish population.

Investigations of the impact of the C115Y variant on IL-23R cell surface expression indicated a small decrease in expression in HEK cells (Lay et al., 2023). However, the lack of a functional response of C115Y IL-23R to IL-23 results predominantly from a disruption of IL-23 binding (Lay et al., 2023). We focused on the four newly discovered hypomorphic variants. The G149R and G300V variants localize to the extracellular domain, whereas the L372F and R381Q variants localize to the transmembrane and intracellular domains of IL-23R, respectively (Fig. 1 D) (Parham et al., 2002). The cell surface expression of the G300V and R381Q variants was weaker than that of WT IL-23R, at levels similar to those reported for the C115Y variant (Fig. 2, A and B) (Lay et al., 2023). By contrast, the cell-surface expression of the G149R and L372F variants was similar to that of WT IL-23R (Fig. 2, A and B). Wide-field luminescence imaging demonstrated receptor expression in the form of luminescence originating from the N-terminal NLuc tag fused to the variant or WT IL-23R following the addition of the NLuc substrate furimazine (Fig. 2 B). We previously performed NanoLuciferase bioluminescence resonance energy transfer (NanoBRET) assays between N-terminal NanoLuciferase-tagged WT IL-23R or IL-12Rβ1 and a TAMRA-labeled IL-23 (Lay et al., 2022). These experiments demonstrated a 66-fold increase in the affinity of IL-23–TAMRA for the heterodimeric complex composed of NLuc–IL-23R and unlabeled IL-12Rβ1 (K_D = 27.0 pM) relative to NLuc–IL-23R or NLuc–IL-12Rβ1 expressed individually (K_D = 222.2 nM or 30.1 nM, respectively). Here, we performed comparable experiments with the four IL-23R variants, each tagged at the N terminus with NanoLuc and used for cotransfection with IL-12Rβ1. All four IL-23R variants had binding affinities similar to that of WT IL-23R (Fig. 2 C and Fig. S2 B). Importantly, these K_D values were substantially lower than the concentrations required for binding to IL-23R or IL-12Rβ1 expressed individually (Lay et al., 2022), suggesting that the measurements were unlikely to be confounded by IL-23–TAMRA binding to non-heterodimeric receptor complexes at the higher ligand concentrations used. However, the lower maximum NanoBRET signal observed for G300V and G149R and, to a lesser extent, R381Q and L372F probably reflect changes in the conformation of the receptor altering its orientation or the distance between the NanoLuciferase donor tag and the TAMRA fluorophore acceptor, resulting in a lower efficiency of energy transfer for these mutant receptors. None of the variants significantly altered the ability of IL-23R and IL-12Rβ1 to form dimers in the presence or absence of 5 nM IL-23, although the lower levels of IL-23R at the surface reduced overall levels of dimer formation in both instances (Fig. S2 C). These data suggest that the weaker pSTAT3 responses observed with the G300V, R381Q, G149R, and L372F variants (Fig. 2, A–C; and Fig. S2, B and C) are not due to a lower affinity for IL-23, but instead reflect the lower levels of cell surface expression for some of these IL-23R variants (R381Q and G300V) and/or conformational changes affecting the efficacy of the agonist with respect to STAT3 phosphorylation.

We then used patient-derived cell lines to capture the impact of the full IL23R locus genotypes of the patients in the context of their own genome. We obtained EBV-immortalized B (EBV-B) cells from eight healthy controls, three patients homozygous for G300V, five patients homozygous for R381Q, and two patients homozygous for G149R (IL23R^WT/WT, IL23R^G300V/G300V, IL23R^R381Q/R381Q, and IL23R^G149R/G149R EBV-B cells, respectively). We also included cells from two patients with complete IL-23R deficiency (C115Y: IL23R^−/−) in the analysis. However, we were unable to derive EBV-B cells from an individual homozygous for L372F, as no such individuals are present in our in-house cohort (>24,000 individuals). These EBV-B cells were left unstimulated or were stimulated with IL-23 or with IFN-α2a as a positive control. EBV-B cells from healthy controls (IL23R^WT/WT) responded to IL-23 by STAT3 phosphorylation (Fig. 2, D and E; and Fig. S2 D). In IL23R^R381Q/R381Q, IL23R^G149R/G149R, and IL23R^G300V/G300V EBV-B cells, STAT3 phosphorylation in response to IL-23 was impaired but not abolished. By contrast, STAT3 was phosphorylated in response to IFN-α2a in all cells (Fig. S2 D). In parallel, we assessed the functionality of IL23R variants in T cell blasts derived from 21 healthy controls, 2 IL23R^G300V/G300V, 16 IL23R^R381Q/R381Q, 2 IL23R^G149R/G149R, 2 IL23R^−/−, and 1 IL12RB1^−/− patient. T cell blasts were left unstimulated or were stimulated with IL-23 or IFN-α2a and STAT3 phosphorylation was monitored by intracellular flow cytometry. T cell blasts from IL23R^WT/WT individuals responded to IL-23 by phosphorylating STAT3, whereas no such phosphorylation was observed in cells from IL23R^G300V/G300V, IL23R^R381Q/R381Q, IL23R^G149R/G149R, IL23R^−/−, and IL12Rβ1^−/− patients (Fig. 2 F), as previously shown for IL23R^R381Q/R381Q (Gerbaux et al., 2025). All T cell blasts responded similarly to IFN-α2a. Thus, IL-23 signaling was impaired but not abolished in lymphoid cells derived from patients homozygous for the G149R, G300V, or R381Q IL23R variants.

Human IFN-γ is essential for antimycobacterial immunity, as most of the known genetic etiologies of MSMD or TB are associated with impaired IFN-γ activity (Boisson-Dupuis et al., 2024; Casanova et al., 2024). We therefore studied the impact of the three IL23R hypomorphic genotypes on IL-23–dependent IFN-γ production. We first compared the distributions of peripheral blood mononuclear cells (PBMCs) from healthy controls, nine IL23R^R381Q/R381Q patients, one IL23R^G300V/G300V patient, one IL23R^G149R/G149R patient, one IL12RB1^−/− patient, and two IL23R^−/− patients by CyTOF with two antibody panels (Fig. S3, A–D). Like patients with AR complete IL-23R deficiency, all the patients studied here had normal counts and frequencies of myeloid and lymphoid subsets, including normal frequencies of helper T cells, natural killer (NK) cells, and MAIT and γδ T cells, which produce large amounts of IFN-γ upon IL-23 stimulation (Martinez-Barricarte et al., 2018; Philippot et al., 2023) (Fig. S3, A–D). We also performed baseline single-cell RNA sequencing (scRNAseq) on two IL23R^G300V/G300V, three IL23R^R381Q/R381Q patients, three TYK2^{P1104A/P1104A}, one IL23R^G149R/G149R, three IL23R^−/−, three IL12RB1^−/− patients, and 11 healthy controls. Clustering analysis revealed comparable numbers and proportions of the 22 discrete transcriptionally defined leukocyte subsets detected in healthy controls and the patients (Fig. 3 A and Fig. S4, A and B), consistent with data derived from flow cytometric immunophenotyping. Gene set enrichment analysis (GSEA) revealed a downregulation of genes regulated by IFN-γ in the classical monocytes of all patients homozygous for hypomorphic IL23R alleles (G149R, G300V, and R381Q), as previously described for IL-23R^−/−, IL-12Rβ1^−/−, and TYK2 P1104A deficiency (Fig. 3 B) (Martinez-Barricarte et al., 2018; Ogishi et al., 2022; Philippot et al., 2023). We hypothesized that homozygosity for a hypomorphic IL23R variant might affect IFNG mRNA induction following the stimulation of patient leukocytes with IL-23 stimulation. We tested this hypothesis with two IL23R^G300V/G300V, one IL23R^G149R/G149R, three IL23R^R381Q/R381Q, two TYK2^{P1104A/P1104A}, two IL23R^−/−, and one IL12RB1^−/− patient, and nine healthy controls, by performing scRNAseq on PBMCs incubated with or without IL-23 for 6 h (Fig. 3, C and D; and Fig. S4, C and D). As previously reported in patients with complete IL-23R or IL-12Rβ1 deficiency, NK, MAIT, and Vδ2⁺ γδ T cells homozygous for the IL23R G149R, IL23R G300V, or TYK2 P1104A variant displayed similar impairments of IFNG mRNA induction upon IL-23 stimulation (Ogishi et al., 2022; Philippot et al., 2023). By contrast, a milder impairment was observed in cells homozygous for the IL23R R381Q variant, consistent with the degree of hypomorphism in the luciferase assay (Fig. 1 A and Fig. 3 D). Overall, IL-23–mediated IFNG mRNA induction was impaired following IL-23 stimulation in MAIT, NK, and Vδ2⁺ γδ T cells from patients homozygous for hypomorphic IL-23R G300V, G149R, and R381Q variants and TYK2^{P1104A/P1104A} individuals.

We then used spectral flow cytometry to assess intracellular IL-23–dependent IFN-γ production by specific lymphocyte subsets: CD4⁺, CD8⁺, NK (bright/dim) and MAIT, and γδ T (Vδ1⁺ and Vδ2⁺) cells. IFN-γ induction in response to IL-23 is barely detectable in CD4⁺, CD8⁺, and Vδ1⁺ γδ T cells, probably because they express IL-23R only weakly, if at all, in the basal state (Camard et al., 2025). As previously described (Bohlen et al., 2023; Philippot et al., 2023), stimulation with IL-18 and high doses of IL-23 stimulation induced a synergistic increase in the frequency of IFN-γ–expressing MAIT, Vδ2⁺ γδ T, and NK cells relative to stimulation with IL-18 only in healthy controls. No such effect was observed with IL23R^G300V/G300V, IL23R^G149R/G149R, and IL12RB1^−/− cells, and a less pronounced defect was observed in IL23R^R381Q/R381Q and TYK2^{P1104A/P1104A} cells (Fig. S5, A–C). However, a clear defect was observed in IFN-γ⁺ cells stimulated with a lower dose of IL-23 (1 ng/ml) for MAIT, Vδ2⁺ γδ T, and NK cells from IL23R^R381Q/R381Q, IL23R^G300V/G300V, IL23R^G149R/G149R, and TYK2^{P1104A/P1104A} individuals, as also observed in IL12RB1^−/− and IL23R^−/− cells, relative to cells from healthy individuals (Fig. 4, A–C). To establish that this does not reflect a generalized inability of the cells to produce cytokines in vitro, the production of IFN-γ and TNF following stimulation with PMA ionomycin was also assessed and was found similar for all genotypes (Fig. 4). After infection with BCG, intracellular IFN-γ induction was impaired in MAIT, Vδ2⁺ γδ T, and NK from IL23R^G300V/G300V, IL23R^G149R/G149R, and IL23R^R381Q/R381Q patients and from a TYK2^{P1104A/P1104A} individual (Fig. 5, A–C). These data suggest that homozygosity for any of the three hypomorphic IL-23R variants (G149R, G300V, and R381Q) impairs IL-23–dependent IFN-γ production ex vivo but does not totally abolish it, as reported for TYK2^{P1104A/P1104A}. In addition, the IL-17A induction induced by PMA ionomycin stimulation on lymphocytes was intermediate between that of healthy individuals and IL12RB1^−/− and IL23R^−/− patients (Fig. 5 D), consistent with the absence of chronic CMC in these patients. The impairment of IL-23–dependent IFN-γ production mostly affected MAIT, Vδ2⁺ γδ T, and NK cells, resulting in a predisposition to TB.

Given the role of IL-23 in inducing IFN-γ secretion, we performed ex vivo experiments to test the hypothesis that homozygosity for the IL23R G300V, G149R, and R381Q variants impairs IL-23–dependent IFN-γ secretion. Homozygosity for IL23R L372F would be expected to have a similar effect, but this hypothesis could not be tested due to the absence of individuals carrying this genotype. We tested IL-23 responses either alone (at low or high concentration) or in combination with IL-1β, IL-18, or BCG in PBMCs from healthy individuals, individuals with IL23R^G300VG300V, IL23R^R381Q/R381Q, or IL23R^G149R/G149R variants, and in IL-23R– and IL-12Rβ1–deficient patients, with IFN-γ concentrations in the supernatant determined after 48 h (Fig. 5 E and Fig. S5 F). PBMCs from individuals with IL23R^G300V/G300V, IL23R^G149R/G149R, or IL23R^R381Q/R381Q genotypes produced significantly less IFN-γ upon IL-23 stimulation, with or without IL-1β, than healthy controls, with a more pronounced decrease observed in patients with complete IL-23R or IL-12Rβ1 deficiency (Fig. S5 D). A similar pattern was observed when PBMCs were stimulated with IL-23 (at low and high doses) in the presence of IL-18 (Fig. 4, D and E; and Fig. S5, E and F). Impaired IFN-γ production by IL23R^R381Q/R381Q, IL23R^G300V/G300V, IL23R^G149R/G149R, and TYK2^{P1104A/P1104A} PBMCs relative to healthy controls was also observed after stimulation with BCG and IL-23 stimulation (Fig. 5 E). As a control, the induction of TNF and IFN-γ by PMA ionomycin was similar for all genotypes. Thus, IL-23–mediated IFN-γ production is impaired in IL23R^R381Q/R381Q, IL23R^G149R/G149R, IL23R^G300V/G300V, and TYK2^{P1104A/P1104A} primary mononuclear cells.

The G300V rare variant was present in the homozygous state in five individuals from two unrelated Turkish families in our TB cohort (kinship coefficient of zero) (Fig. S2 E). Four of these individuals suffered from TB, none had MSMD, and the fifth remained asymptomatic. The R381Q variant is present in the homozygous state in 98 individuals from the HGID cohort: seven had TB (four North Africans, one European, one person of Middle Eastern ancestry, and one Latin American), two had MSMD, and the remaining 89 were either healthy or suffered from other infections (Table S1). Two individuals from the HGID cohort (one of Middle Eastern ancestry and one Asian) were homozygous for the G149R allele, but neither had TB or MSMD (Table S1). There were no homozygotes for L372F in our entire in-house HGID database. We first focused on the variant with the strongest hypomorphic effect, G300V, comparing the frequency of homozygosity for this variant between 1,874 TB patients of diverse ethnic origins and 20,883 individuals from the in-house HGID database (“control cohort”) who were either healthy or suffering from non-intramacrophagic diseases. The proportion of G300V homozygotes was 2.1 × 10⁻³ in the TB cohort and 4.8 × 10⁻⁵ in the control cohort, giving an adjusted odds ratio (OR) for TB of OR [95% CI] = 26.5 [4.2–167.0]; P value = 4.9 × 10⁻⁴ on logistic regression (see Materials and methods; Table S2). We then performed an activity-weighted burden test (see Materials and methods, with the weight inversely proportional to the residual activity of the variant), which demonstrated a significant enrichment in homozygosity for hypomorphic IL23R variants (G300V, R381Q, G149R, or L372F) among TB cases (P = 3.8 × 10⁻²) (Table S2). This enrichment was even stronger among individuals with inferred “Middle Eastern” ancestry (see Materials and methods section) (N cases = 404 and N controls = 1,871), in whom R381Q is more frequent than in other populations (frequency in the gnomAD v4 Middle Eastern population = 5.1%) (P = 9.7 × 10⁻⁵). By contrast, no significant enrichment in homozygosity for G300V, R381Q, G149R, or L372F was detected in cohorts of MSMD or CMC patients, as suggested in other studies (Gerbaux et al., 2025) (Table S2), and no enrichment in heterozygosity for these variants was detected in any of the cohorts, suggesting that these variants implicated in TB display recessive inheritance. As a negative control, we performed the same test on the 15 variants identified as isomorphic in the luciferase assay; we found no enrichment in these variants in the TB cohort (P = 0.48). It should be noted that the low to moderate prevalence of TB in some regions of the world (e.g., Europe and the Middle East) implies that most of the controls living in these regions included in this analysis are unlikely to have been exposed to and infected with Mycobacteriumtuberculosis. The calculated enrichment in homozygosity for hypomorphic variants among TB patients for G300V, R381Q, G149R, or L372F is, therefore, probably an underestimation of the true risk of TB in homozygous individuals upon infection. Our data suggests that homozygosity for the G300V, R381Q, G149R, or L372F variants confers a predisposition to TB.

We identified rare (G149R, G300V, and L372F) and common (R381Q) IL23R variants that were hypomorphic. Homozygosity for each of these variants causes partial, rather than complete, IL-23R deficiency, with IL-23–dependent STAT3 phosphorylation impaired but not abolished. The underlying mechanism involves low levels of surface expression and/or conformational changes, with no effect on affinity for IL-23. Cells from individuals homozygous for the IL23R G300V, G149R, or R381Q variant display impaired IL-23–dependent IFN-γ production. This is probably also true for IL23R^L372F/L372F, but we were unable to test cells from an individual with this genotype due to its rarity. A higher proportion of TB patients homozygous for G300V than of those homozygous for R381Q had extrapulmonary disease and an earlier disease onset. However, these observations are based on a limited number of individuals. The defect affects principally the innate-like (MAIT and Vδ2⁺ γδ T) and innate (NK) lymphocyte subsets, which are normally the cells with the best response to IL-23 stimulation (Hunter, 2005; Martinez-Barricarte et al., 2018; Ogishi et al., 2022; Philippot et al., 2023). The cellular phenotypes of individuals with the rare IL23R^G300V/G300V and IL23R^G149R/G149R genotypes or the common IL23R^R381Q/R381Q genotype resemble that of TYK2^{P1104A/P1104A} patients (Boisson-Dupuis et al., 2018; Ogishi et al., 2022). The residual IL-23–dependent IFN-γ production in patients homozygous for the IL23R G300V, G149R, and R381Q variants is sufficient to protect most of these individuals against weakly virulent mycobacteria (BCG and EM), contrasting with patients with a complete form of recessive IL-23R deficiency (Martinez-Barricarte et al., 2018; Philippot et al., 2023; Staels et al., 2022). Like TYK2^{P1104A/P1104A} (Boisson-Dupuis et al., 2018; Ogishi et al., 2022), IL23R^R381Q/R381Q has a very low penetrance for MSMD. By contrast, the residual IL-23–dependent IFN-γ production is not sufficient to protect homozygotes against M. tuberculosis, which is at least 1,000 times more virulent than BCG and EM (Boisson-Dupuis et al., 2011; Fieschi et al., 2003). Consistently, no enrichment in these rare and common IL23R variants was detected in the MSMD cohort. There was also no enrichment in homozygosity for these variants in the CMC cohort, consistent with the low penetrance of CMC in patients with complete IL-23R deficiency (only a third of IL-23R^−/− patients had CMC). The residual IL-23–dependent IL-17 immunity in homozygotes is sufficient to ensure antifungal immunity.

Like the TYK2 P1104A variant, the IL23R R381Q variant is most frequent in populations of European ancestry (MAF: 0.06) (Boisson-Dupuis et al., 2018; Kerner et al., 2019; Kerner et al., 2021; Kerner et al., 2023). Moreover, these two alleles have evolved under negative selection in European populations over the last 4,500 years. Indeed, the frequency of IL23R R381Q has fallen from 10% in the Bronze age and 15% in the Neolithic to 5.5% today (Kerner et al., 2023). The frequency of TYK2 P1104A has fallen from 10% in the Bronze age and 15% in the Neolithic to 2.9% today (Kerner et al., 2021; Kerner et al., 2023). These evolutionary trajectories are probably associated with the high burden of TB in Europeans between the Bronze age and the middle of the 20^th century (Kerner et al., 2021; Kerner et al., 2023). Both the TYK2 P1104A and IL23R R381Q variants have a protective effect against inflammatory bowel diseases (ORs ranging from 0.1 to 0.3 and 0.2 to 0.7, respectively) (Boisson-Dupuis et al., 2018; Duerr et al., 2006; Momozawa et al., 2011). These observations suggest that the higher risk of inflammatory disorders in post-Neolithic Europeans may be due, in part, to the negative selection of alleles such as TYK2 P1104A and IL23R R381Q that weaken protective immunity to primary infection by attenuating the inflammatory response (Boisson-Dupuis et al., 2018; Fieschi et al., 2003). The risk of TB is probably largely underestimated in our study, as few present-day Europeans are exposed to M. tuberculosis. The current clinical penetrance of partial IL-23R deficiency for TB in Europe is therefore probably low, and the impact of IL23R R381Q homozygosity is probably lower than that of TYK2 P1104A at cellular level (Fig. 3 D). Genetic testing for these IL23R variants should nevertheless be offered to TB patients, especially those of European ancestry, as treatment with recombinant IFN-γ may be of therapeutic benefit (Casanova et al., 2024).

Informed consent for participation in this study was obtained in accordance with local regulations, with approval from the institutional review board. The study was approved by the institutional ethics committees of The Rockefeller University and Necker Hospital for Sick Children and was performed in accordance with the requirements of these bodies. The experiments described here were performed in France and the United States of America, in accordance with local regulations.

Genomic DNA was extracted from whole-blood samples from the patients and their relatives. WES was performed as previously described (Ogishi et al., 2022; Ogishi et al., 2025). MAFs in the general population, as reported in gnomAD database v4.1, and precomputed combined annotation–dependent depletion scores (v1.7) were used for variant filtering. The mutation significance cutoff was calculated as previously described (Itan et al., 2016). For the verification and familial segregation of variants, exons and flanking regions were amplified from DNA with DreamTaq DNA polymerase. They were then sequenced by the Sanger method with the Big Dye Terminator v3.1 kit (Thermo Fisher Scientific) and subjected to capillary electrophoresis (#A30469, Applied Biosystems 3500xL system, Thermo Fisher Scientific).

We performed an enrichment analysis on the G300V, R381Q, and G149R variants of IL-23R in our cohort of 1,874 TB patients and 20,883 healthy controls or patients with non-intramacrophagic infectious diseases. Association analyses were performed with regenie assuming a recessive mode of inheritance (Mbatchou et al., 2021). This method builds a whole-genome regression model based on common variants to correct for the effects of relatedness and population stratification. We used the approximate Firth P value when the logistic regression P value was below 0.05. A weight was applied to variants based on the ratio of WT to mutant protein activity to ensure that variants with more severe effects were given higher weights, as suggested by Wu et al. (2011). Analyses were adjusted for sex and the first 10 principal components (PC) of the PC analysis (PCA). PCA was performed with Plink v1.9 (Purcell et al., 2007) on WES data with a pruned set of 17,934 exonic variants (maximum linkage disequilibrium between each pair of SNPs <0.4) with a MAF >1% and a call rate >99%. We inferred genetic ancestry from the PCA, using a reference set of individuals, as previously described (Belkadi et al., 2016). Individuals with a prediction confidence >80% for similarity to individuals with Middle Eastern ancestry were retained for Middle East-specific analysis.

The age of the most recent common ancestor carrying the G300V variant was estimated using the EstiAge method (Genin et al., 2004), assuming a generation time of 27 years (Wang et al., 2023). In brief, the EstiAge method assumes that all affected individuals inherited the variant from a single common ancestor who introduced it generations ago. The number of generations is estimated from the length of the haplotype shared around the variant by locating the most likely recombination points on the ancestral haplotype across the different samples.

The generation of N-terminal NanoLuciferase-tagged WT IL-23R (NLuc IL-23R), HaloTag WT IL-12Rβ1 receptor (HaloTag IL-12Rβ1), and untagged IL-12Rβ1 constructs has been described elsewhere (Lay et al., 2022). The R381Q IL23R mutant was generated from the NL-IL23R plasmid with a Phusion site-directed mutagenesis kit (Thermo Fisher Scientific) according to the manufacturers’ instructions. The oligonucleotide primers used (Merck) were the following:

Forward primer 5′-ATT​TAA​CAG​ATC​ATT​CCA​AAC​TGG​GAT​TAA​AAG​AAG​G-3′

Reverse primer 5′-ATC​CCA​ATC​AAA​GAA​AGA​ATT​GAC​AAC-3′.

The PCR products were then digested with Dpn1 for 15 min at 37°C to eliminate methylated (template) DNA. The mutated R381Q IL23 sequence was then excised from the PCR-generated plasmid vector and cloned back into the original template vector. Sequences were confirmed by Sanger sequencing (DEEPSEQ, University of Nottingham).

The NanoLuciferase-tagged IL23R plasmids containing mutations G300V and L372F were generated by site-directed mutagenesis with the Q5 Site-Directed Mutagenesis Kit (#E0554S; NEB) and the following primers:

G300V:

Forward primer: 5′-TCA​AGA​AAC​AGTGAAA​AGG​TAC​TGG-3′

Reverse primer: 5′-CAT​CTC​ACT​TGA​AAT​ACG​TAC-3′.

L372F:

Forward primer: 5′-AAT​TCT​TTC​TTTTATT​GGG​ATA​TTT​AAC​AG-3′

Reverse primer: 5′-GAC​AAC​ATA​ACA​GCA​AAG-3′.

In all the above primer sequences, the underlined nucleotides were mutated.

The NanoLuciferase-tagged IL23R plasmid containing the G149R mutation was generated by site-directed mutagenesis and assembly cloning. Briefly, two separate PCR were performed with the primers below, with the WT NLuc IL23 plasmid as a template. The PCR products were purified and assembled with the NEBuilder HiFi DNA Assembly Cloning Kit (#E5520S; NEB) according to the manufacturer’s instructions.

G149R:

Forward primer 1: 5′-CTG​GAA​TGC​TCGCAAG​CTC​ACC​TAC-3′

Reverse primer 1: 5′-GGA​GCG​AAC​GAC​CTA​CAC​CGA​ACT​GAG​ATA​CCT​ACA​GCG-3′

Forward primer 2: 5′-CGC​TGT​AGG​TAT​CTC​AGT​TCG​GTG​TAG​GTC​GTT​CGC​TCC-3′

Reverse primer 2: 5′-GTA​GGT​GAG​CTTGCGAGC​ATT​CCA​G-3′.

In all the above primer sequences, the underlined nucleotides were mutated.

All the NanoLuciferase-tagged IL23R receptor plasmids and variants of them were verified by whole-plasmid sequencing (Plasmidsaurus, London, using Oxford Nanopore Technology).

PBMCs were isolated by Ficoll-Hypaque density centrifugation (Amersham Pharmacia Biotech). Human embryonic kidney cells (HEK293T [CRL3216; ATCC]) and EBV-B cells were cultured in DMEM or RPMI-1640 medium (Gibco), respectively, supplemented with 10% FCS (Sigma-Aldrich). For T-blast induction, PBMCs were cultured in ImmunoCult-XF T Cell Expansion Medium (STEMCELL) in the presence of ImmunoCultTM Human CD3/CD28/CD2 T cell activator (12.5 μl.ml⁻¹) (STEMCELL) and human recombinant IL-2 (100 ng/ml, Novartis).

Total RNA was extracted with the Quick-RNA Microprep kit (Zymo) and reverse-transcribed to generate cDNA with the High-Capacity RNA-to-cDNA kit (Applied Biosystems). Quantitative PCR was then performed on the RNA with the Applied Biosystems probes/primers specific to IL23R-FAM (Hs00332759_m1) and β-glucuronidase-VIC (4326320E) for normalization. Results are expressed according to the ΔCt method.

HEK cells were used to seed a six-well plate (10578911; Corning Costar) at a density of 90,000 cells per well. The cells were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The next day, cells were transfected with a 4:1 ratio of WT or mutant NLuc IL-23R and IL-12Rβ1 cDNA with FuGENE HD (Promega Corporation), used according to the manufacturer’s instructions (3:1 total cDNA-to-reagent ratio). Cells were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The transfected cells were then plated at a density of 20,000 or 30,000 cells per well on poly-D-lysine–coated white, flat-bottomed 96-well plates (655098; Greiner Bio-One) and incubated overnight at 37°C under an atmosphere containing 5% CO₂. The plating medium was then removed, and the cells were incubated with various concentrations of IL-23–TAMRA (Lay et al., 2022) in HEPES-buffered saline solution (HBSS; 2 mM sodium pyruvate, 145 mM NaCl, 10 mM D-glucose, 5 mM KCl, 1 mM MgSO₄·7H₂O, 10 mM HEPES, 1.3 mM CaCl₂, and 1.5 mM NaHCO₃ in double-distilled water, pH 7.45) supplemented with 0.1% BSA in the presence or absence of unlabeled IL-23 (Lay et al., 2022) (100 nM). IL-23–TAMRA was generated as previously described (Lay et al., 2022). Cells were incubated for 60 min at 37°C without CO_2, and the NanoLuciferase substrate furimazine was then added to all wells (Promega Corporation; final concentration of 7.7 μM). Plates were incubated for 5 min at room temperature, and the emitted luminescence and fluorescence were then simultaneously detected with a PHERAstar FS plate reader (BMG LabTech) fitted with a 450 ± 30 nm band-pass (luminescence emission [NLuc]) and >550 nm long-pass (fluorescence emission [IL-23–TAMRA]) filters.

HEK cells were used to seed a 6-well plate (10578911; Corning Costar) at a density of 90,000 cells per well. They were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The cells were then transfected with a 4:1 ratio of WT or mutant NLuc IL23R and IL12RB1 cDNA with FuGENE HD (Promega Corporation), according to the manufacturer’s instructions (3:1 cDNA-to-reagent ratio). Cells were incubated overnight at 37°C under an atmosphere containing 5% CO₂ and were then used to seed poly-D-lysine–coated white, flat-bottomed 96-well plates (655098; Greiner Bio-One) at a density of 20,000 or 30,000 cells per well. The plates were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The culture medium was then removed and replaced with 7.7 μM of furimazine in HBSS (10 min at 37°C) in the presence or absence of NanoLuc extracellular inhibitor (60 μM; Promega Corporation). Luminescence emissions were then detected with a PHERAstar FS.

HEK cells were used to seed poly-D-lysine–coated 5-mm Cellview 4-quadrant culture dishes (Greiner Bio-one) with a 10-mm glass coverslip bottom at a density of 80,000 cells per quadrant. The plates were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The cells were then transfected with a 4:1 ratio of WT or mutant NLuc IL23R and HaloTag IL12RB1 cDNA with FuGENE HD (Promega Corporation), according to the manufacturer’s instructions (3:1 total cDNA-to-reagent ratio). Cells were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The plating medium was then removed and replaced with HBSS containing furimazine (1:400 final dilution). Luminescence images were acquired with an Olympus LuminoView 200 microscope fitted with a 60× NA1.42 oil immersion objective using a 0.5× tube lens and a C9100-23B IMAGE EMX2 camera (Hamamatsu, Japan) with an exposure time of 20 s (25 gain).

HEK cells were used to seed a 6-well plate (10578911; Corning Costar) at a density of 90,000 cells per well. They were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The cells were then transfected with a 4:1 ratio of WT or mutant NLuc IL-23R and HaloTag IL-12Rβ1 cDNA with FuGENE HD (Promega Corporation), according to the manufacturer’s instructions (3:1 cDNA-to-reagent ratio), and incubated overnight at 37°C under an atmosphere containing 5% CO₂. The transfected cells were used to seed poly-D-lysine–coated white, flat-bottomed 96-well plates (655098; Greiner Bio-One) at a density of 20,000–30,000 cell per well. They were incubated overnight at 37°C under an atmosphere containing 5% CO₂. The culture medium was then removed and replaced with DMEM/10% FBS supplemented with the membrane-impermeant HaloTag 618 ligand (Promega Corporation) at a concentration of 100 nM. The cells were incubated for 5 h at 37°C under an atmosphere containing 5% CO₂ and were then washed three times HBSS/0.1% BSA and incubated with HBSS/0.1% BSA containing 5 nM IL-23 for 60 min at 37°C. Furimazine was then added to all wells (7.7 μM final concentration), and the plate was allowed to stand for 5 min before the simultaneous measurement of luminescence and fluorescence emissions with a PHERAstar FS fitted with 460 ± 80 nm band-pass (luminescence [NLuc]) and >610 nm long-pass filters (fluorescence [HaloTag 618]).

Plasmids containing the WT human IL23R (#RG211477; OriGene) ORF were obtained, the tag was removed, and the variants studied here were generated by site-directed mutagenesis with specific primers and CloneAmp HiFi PCR premix (Takara). These ORFs included the IL23R S221F variant. The ORFs were introduced into the pLenti III-UbC-mCherry plasmid.

HEK cells were grown in DMEM supplemented with 10% FCS for 24 h before transfection. We assessed the impact of the mutation on IL-23R function by transfecting cells with the pLenti III-UbC-mCherry (empty vector or plasmid, EV) or the same plasmid backbone containing one of the IL23R ORFs (25 ng/well for a 96-well plate), pGL4.47(luc2P/SIE/Hygro) (Promega Corporation) reporter plasmids (100 ng/well), the pRL-SV40 plasmid (10 ng/well), and the PGK-IL12RB1 plasmid (encoding WT IL-12Rβ1—50 ng/well) in the presence of X-tremeGENE 9 DNA Transfection Reagent (Roche). The medium was removed 36 h after transfection and replaced with DMEM supplemented with 10% FCS or the indicated cytokines for 16 h. Experiments were performed with technical duplicates, and the promoter activity of each well is expressed as firefly luciferase activity/Renilla luciferase activity normalized against the WT signal after stimulation with 10 ng/ml IL-23 taken as 100% activity.

Levels of pSTAT3 were assessed in HEK or EBV-B cells. After 36 h of transfection (6 h with transfection medium followed by 30 h in DMEM containing 10% FCS), cells (1 × 10⁶) were incubated for 30 min with IL-23 (1 or 10 ng/ml, R&D) or IFN-α2a (1 ng/ml, Miltenyi Biotec). Cells were washed in PBS and lysed in RIPA lysis buffer (Thermo Fisher Scientific) supplemented with protease inhibitors (Roche mini EDTA-free, one tablet in 10 ml), phosphatase inhibitors (Roche PhosSTOP, one tablet in 10 ml), and benzonase (50 U/ml). Lysates were clarified and protein concentration was determined in a Bradford assay. Equal amounts of protein were subjected to SDS-PAGE, and the resulting bands were transferred to a PVDF membrane with 0.2 μm pores. Membranes were subjected to Ponceau staining, incubated in 5% skim milk in PBST for 1 h, briefly rinsed with PBST, and then incubated overnight at 4°C in primary antibody solution (5% BSA PBST or 5% skim milk PBST). Membranes were then washed three times, for 15 min each, in PBST, incubated in secondary antibody solution (1:2,000 dilution in 5% skim milk PBST) for 1 h at room temperature, and then washed again three times, for 15 min each, in PBST. Finally, chemiluminescence was detected with ECL reagents and a Bio-Rad ChemiDoc. After acquisition, the membrane was incubated in stripping buffer for 15 min (Restore Western Blot Stripping Buffer, Thermo Fisher Scientific) and rinsed in PBST. The induction of pSTAT3 induction is expressed as a percentage of the pSTAT3/STAT3 for the WT after stimulation with IL-23.

Cells (1 × 10⁶) were stained for 5 min at room temperature with the Aqua Dead cell viability marker (Thermo Fisher Scientific) and were then incubated for 30 min at 4°C with PE-conjugated anti-CD212 (IL-12Rβ1) antibody (Cat: 556065; BD Biosciences, Clone: 2.4E6, 1:5 dilution). The cells were analyzed in PE-Texas Red-, BV421-, and PE-compensated channels on an LSRFortessa X-20 flow cytometer (Beckman Coulter), and the results were analyzed with FlowJo software.

We used 100,000 PBMCs to seed RPMI supplemented with 10% human serum in each well of a round-bottomed 96-well plate. The cells were stimulated by incubation for 48 h with human recombinant IL-1β (2.5 ng/ml, R&D) and human recombinant IL-23 (0.001–100 ng/ml, R&D). PMA (Cat: P1585-1MG; Sigma-Aldrich, 8 ng/ml) + ionomycin (Cat: 56092-81-0; Sigma-Aldrich, 10-5 M) were added for the last 24 h only. The supernatants were collected and subjected to LEGENDplex multiplex ELISA with Human Inflammation Panel 1 (740809; BioLegend), according to the manufacturer’s instructions.

Freshly thawed PBMCs from healthy controls, IL23R^G300V/G300V, IL23R^R381Q/R381Q, IL23R^G149R/G149R, TYK2^{P1104A/P1104A}, IL-12Rβ1-, and IL23R-deficient patients were dispensed into a U-bottomed 96-well plate at a density of 3 × 10⁵ cells per well, in 200 μl lymphocyte medium per well, as previously described (Philippot et al., 2023; Yang et al., 2020). Cells were incubated in the presence or absence of live BCG, at a multiplicity of infection of 1, with or without recombinant human IL-12 (5 ng/ml, R&D Systems), recombinant human IL-18 (25 ng/ml, R&D Systems), and/or recombinant human IL-23 (100 ng/ml, 1290-IL R&D Systems) or IFN-γ (5,000 IU/ml). After 40 h of stimulation, GolgiPlug (555029; BD Biosciences; 1:1,000 dilution) was added to each well to inhibit cytokine secretion. Cells were also costimulated for 24 h with IL-23 (100, 10, or 1 ng/ml) and IL-18 (200 ng/ml). PMA ionomycin was added 6 h before the end of incubation. The cells were collected by centrifugation for flow cytometry analysis. They were stained with the Zombie NIR Fixable Viability Kit (BioLegend; 1:2,000 dilution) at room temperature for 15 min and then stained on ice for 30 min with a surface-staining panel containing FcR blocking reagent (Miltenyi Biotec; 1:50 dilution), anti-CD3-Alexa Fluor 532 (Clone: UCHT1, 58-0038-42; eBioscience; 1:50 dilution), anti-γδTCR-FITC (Clone:B1.1, 11-9959-41; eBioscience; 1:50 dilution), anti-Vδ2-APC/Fire 750 (Clone:B6, 331419; BioLegend; 1:100 dilution), anti-CD56-BV605 (Clone: 5.1H11, 362537; BioLegend; 1:100 dilution), anti-CD4-BV750 (Clone: SK3, 566356; BD Biosciences; 1:800 dilution), anti-CD8a-Pacific Blue (Clone: SK1, 344717; BioLegend; 1:100 dilution), anti-Vα7.2 TCR-APC (Clone: 3C10, 351708; BioLegend; 1:100 dilution), anti-Vα24-Jα18-PE/Cy7 Clone: 6B11, 342912; BioLegend; 1:100 dilution), anti-CD20-BV785 (Clone: 2H7, 302356; BioLegend; 1:200 dilution), and anti-PD- 1-PE (Clone: MIH4, 12-9969-42; eBioscience; 1:100 dilution) antibodies. Cells were fixed by incubation with 2% paraformaldehyde in PBS on ice for 15 min. They were then permeabilized/stained by incubation overnight at −20°C in the permeabilization buffer from the Nuclear Transcription Factor Buffer Set (BioLegend) with an intracellular cytokine panel containing FcR blocking reagent (Miltenyi Biotec; 1:50 dilution), anti-IFN-γ-BV711 (Clone: 4 S.B3, 502540; BioLegend; 1:50 dilution), anti-TNF-BV510 (Clone: MAb11, 502950; BioLegend; 1:50 dilution), and anti-IL-10-PE/Dazzle594 (Clone: JES3-19F1, 506812; BioLegend; 1:50 dilution) antibodies. As a positive control, cells in a separate well were stimulated by incubation with PMA (Sigma-Aldrich; 25 ng/ml) and ionomycin (Sigma-Aldrich; 500 nM) for 1 h without GolgiPlug and then for 7 h with GolgiPlug (for intracellular cytokine staining). Cells were acquired with an Aurora cytometer (Cytek). Data were manually gated with FlowJo.

Deep immunophenotyping was performed by mass cytometry (CyTOF). We used 200 μl fresh whole blood from the patients and controls. We used a previously described custom-designed panel (Momenilandi et al., 2024), according to Standard BioTools’s instructions. Cells were frozen at −80°C after overnight dead-cell staining, and acquisition was performed on a Helios mass cytometer (Fluidigm). All the samples were processed within 24–36 h of sampling. Data analysis was performed with OMIQ software.

PBMCs were thawed and plated at a density of 50,000–300,000 cells per well in a U-bottom 96-well plate. IL-23 was added to a final concentration of 100 ng/ml. After 6 h of incubation at 37°C, non-stimulated and stimulated cells were recovered, washed three times in PBS supplemented with 0.5% FBS, filtered through a 40-µm-pore MACS strainer, and subjected to scRNAseq.

For baseline analysis, scRNAseq data were integrated with previously published data for eight controls and data for three newly studied healthy controls, together with samples from IL-12Rβ1^−/− (n = 3), TYK2^P1104/P1104 (n = 3), IL-23R^−/− (n = 3), IL-23R ^R381Q/R381Q (n = 3), IL-23R^G300V/G300V (n = 2), and IL-23R^G149R/G149R (n = 1) patients. For the stimulation analysis, paired non-stimulated and IL-23–stimulated samples (data for the 6-h time point) were analyzed, including six previously published adult controls, two newly studied healthy controls, and patient samples: IL-12Rβ1^−/− (n = 1), TYK2^P1104/P1104 (n = 2), IL-23R^−/− (n = 2), IL-23R^R381Q/R381Q (n = 3), IL-23R^G300V/G300V (n = 2), and IL-23R^G149R/G149R (n = 1).

Quality control was performed by filtering cells based on the percent mitochondrial gene expression, the total number of transcripts detected (UMIs), and the number of genes expressed per cell, with standard thresholds. Cell type annotation was performed by anchor-based label transfer in Seurat (Hao et al., 2021). A previously integrated PBMC reference (canonical correlation analysis, CCA) was processed by PCA and UMAP. Anchors were identified between the CCA-integrated reference and PCA-transformed query data, making it possible to transfer cell type labels and projections onto the reference UMAP space. UMAP was used for visualization. Gene expression was quantified in Seurat (Hao et al., 2021). Pseudobulk expression profiles were generated by summing counts per sample and cell type with the SingleCellExperiment framework (Amezquita et al., 2020) and muscat (Crowell et al., 2020). PCA was performed on variance-stabilized counts, with batch effects corrected with removeBatchEffect from limma (Ritchie et al., 2015). Differential expression analysis was performed with DESeq2 (Love et al., 2014), and log₂ fold-change shrinkage for gene ranking was applied with apeglm (Zhu et al., 2019). GSEA was performed with fgsea (Korotkevich et al., 2021, Preprint) and gene sets from the Molecular Signatures Database (Liberzon et al., 2011).

We calculated the induction of IFNG mRNA following IL-23 stimulation using the housekeeping genes HKG, C1orf43, CHMP2A, EMC7, GPI, PSMB2, PSMB4, RAB7A, REEP5, SNRPD3, VCP, and VPS29, with geometric mean normalization for each sample.

Statistical analyses were performed with GraphPad Prism v8.4.3 and v10.5.0. Paired t tests (intrareceptor BRET experiments), one-way ANOVA with Tukey’s post hoc correction for multiple comparisons (IL-23R–IL12Rβ1 interaction experiments), and nonparametric Mann–Whitney U tests were used as appropriate to assess statistical significance. A P value <0.05 was considered statistically significant, *; P < 0.01, **; P < 0.001, ***; and P < 0.0001, ****. PCA was performed with Plink v1.9 software on WES data, using 16,730 exonic variants with a MAF >1% and a call rate >98%.

For cell surface expression experiments, the emitted luminescence measured in the presence of the extracellular NanoLuciferase inhibitor is expressed as a percentage of that measured in the absence of the inhibitor (100%). For IL-12Rβ1 interaction experiments, WT NLuc IL-23R/HaloTag IL-12Rβ1 responses in the absence of IL-23 were set to 100%, and subsequent BRET ratios were normalized against this value and expressed as a percentage ± SEM.

Fig. S1 shows IL12RB1 and IL23R expression at the mRNA and protein levels in transfected HEK cells, along with STAT3 phosphorylation after IL-23 stimulation in cells expressing different IL23R variants. Fig. S2 displays the predicted population-level impact of IL23R variants based on AlphaMissense, as well as their functional effects on IL-23 binding and STAT3 phosphorylation assessed by western blotting in homozygous EBV-B cells. These EBV-B cell lines were generated from the individuals mentioned. PBMC immunophenotyping of individuals homozygous for hypomorphic IL23R variants is shown in Fig. S3. Analysis of the single-cell sequencing of PBMCs from individuals homozygous for hypomorphic IL23R variants is shown in Fig. S4. Fig. S5 shows impaired IFN-γ induction in PBMCs from individuals homozygous for hypomorphic IL23R variants. Table S1 reports IL23R variants in the diverse cohorts of patients, and Table S2 shows the association in the HGID cohort for G149R, G300V, and R381Q variants (no homozygotes for L372F were found in the HGID cohort) for the recessive model.

The data underlying the figures are available in the published article and its online supplemental material. In addition, scRNAseq data are available with the following BioProject ID: PRJNA1449438 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1449438).

We thank the patients and their families for placing their trust in us. We thank the members of both branches of the Laboratory of Human Genetics of Infectious Diseases. We thank Tatiana Kochetkov for technical assistance; Yelena Nemirovskaya, Mark Woollett, Dana Liu, Amyrat Geraldo, Soraya Boucherit, Abigail Bejou, Maya Chrabieh, and Lazaro Lorenzo for administrative assistance. We thank the Vecteurs Viraux et Transfert de Gènes (VVTG) platform of the “Necker Enfants Malades” Institute for the EBV-B cell line. We thank the Flow Cytometry Resource Center at The Rockefeller University. We thank the National Institutes of Health (NIH) Tetramer Core Facility for providing the MR1 tetramer, which was developed jointly with Dr. James McCluskey, Dr. Jamie Rossjohn, and Dr. David Fairlie. We are indebted to the “Biobanc de l’Hospital Infantil Sant Joan de Déu per a la Investigación,” a member of the ISCIII National Network of Biobanks, Spain, for sample and data procurement. Graphical abstract was created with Biorender.com (https://app.biorender.com/citation/69d3f0d806893ae4a332727b).

The HGID laboratory was funded in part by the National Institute of Allergy and Infectious Diseases, NIH (grant numbers R01AI095983 to J.-L. Casanova and J. Bustamante and R01AI127564 to J.-L. Casanova and A. Puel), the National Center for Research Resources and the National Center for Advancing Sciences, NIH (grant number UL1TR001866 for J.-L. Casanova), The Rockefeller University, the St. Giles Foundation, the Howard Hughes Medical Institute, Institut National de la Santé et de la Recherche Médicale (INSERM), Paris Cité University, the French Agence Nationale de la Recherche (ANR) under the France 2030 program (ANR-10-IAHU-01), the Integrative Biology of Emerging Infectious Diseases Laboratory of Excellence (ANR-10-LABX-62-IBEID), the ANR project MAFMACRO (ANR-22-CE92-0008 to J. Bustamante), the Agence Nationale de Recherches sur le Sida et les Hépatites Virales (ANRS) projects ECTZ170784-ANRS0073 and ANRS0787-ECTZ323676 to S. Boisson-Dupuis, the French Foundation for Medical Research (FRM) (EQU202503020018), the European Union’s Horizon 2020 Research and Innovation Program under grant agreement no. 824110 (EASI-genomics), the Square Foundation, Grandir - Fonds de Solidarité pour L’enfance, William E. Ford, General Atlantic’s Chairman and Chief Executive Officer, Gabriel Caillaux, General Atlantic’s Co-President, Managing Director, and Head of Business at EMEA, and the General Atlantic Foundation. C. Conil was supported by an ANRS PhD fellowship (ECTZ293125), and Q. Philippot was supported by the Assistance Publique Hôpitaux de Paris (Année Recherche) and by the MD-PhD program of INSERM (Ecole de l’INSERM Liliane Bettencourt). J. Bohlen was supported by fellowships from the European Molecular Biology Organization and Marie Curie Research. E. Feredj was supported by the FRM PhD program, Poste de Thèse Pour Interne et Assistant (FDM202406018939), “bourse d’aide à la mobilité” de l’Institut Servier, Collège des Universitaires des Maladies Infectieuses et Tropicales, European Society of Clinical Microbiology and Infectious Diseases, and “bourse de mobilité internationale” Réunion Interdisciplinaire de Chimiothérapie Anti-Infectieuse. H. Li was supported by Labex-IBEID and an ANRS PhD fellowship (ECTZ293911). M. Momenilandi was supported by an Imagine Institute Postdoctoral Prize. M. Ogish was supported by the David Rockefeller Graduate Program, the Funai Foundation for Information Technology, the Honjo International Scholarship Foundation, and the New York Hideyo Noguchi Memorial Society. S.J. Hill, L.E. Kilpatrick, and S. Platt are funded by the Medical Research Council, UK (grant no. MR/W016176/1). C.S. Lay was supported by a Biotechnology and Biological Sciences Research Council Industrial Case Studentship with GSK to C.S. Lay, P.D. Craggs, and S.J. Hill (grant no. BB/S507027/1). This study was supported by the project PI15/01094 to L. Abel, integrated in the Plan Nacional de I+D+I and cofinanced by the ISCIII – Subdirección General de Evaluación y Formento de la Investigación Sanitaria – and the Fondo Europeo de Desarrollo Regional. I. Meyts is funded by Fonds Wetenschappelijk Onderzoek (FWO) Vlaanderen G1805826N. A.A. Arias, M. Tejada-Giraldo, A. Baena, and L.F. Barrera were supported by the Ministerio de Ciencia, Tecnología e Innovación MINCIENCIAS (111584467551/CT 415-2020), the Jeffrey Modell Foundation, and the Fundación Diana García de Olarte para las Inmunodeficiencias Primarias (Colombia). C.S. Ma and S.G. Tangye are supported by Investigator Grants awarded by the National Health and Medical Research Council of Australia (C.S. Ma: 2017463 [level 1]; S.G. Tangye: 1176665 [level 3]).

Author contributions: Diana Olguín Calderón: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, and editing. Laura E. Kilpatrick: formal analysis, funding acquisition, investigation, visualization, and writing—original draft, review, and editing. Clément Conil: data curation, formal analysis, software, validation, and writing—original draft, review, and editing. Quentin Philippot: conceptualization, formal analysis, investigation, and writing—original draft, review, and editing. Masato Ogishi: investigation, methodology, software, and writing—review and editing. Joseph Vellutini: data curation, formal analysis, resources, software, and visualization. Ji Eun Han: methodology, resources, and validation. Narelle Keating: investigation and visualization. Hailun Li: investigation. Geetha Rao: formal analysis and investigation. Jonathan Bohlen: conceptualization, investigation, methodology, supervision, validation, and writing—review and editing. Charles S. Lay: investigation and methodology. Simon Platt: investigation, methodology, and writing—original draft. Gaspard Kerner: formal analysis. Elsa Feredj: investigation. Jessica N. Peel: writing—review and editing. Mana Momenilandi: investigation and resources. Yoann Seeleuthner: data curation and resources. Candice Lainé: investigation and resources. Camille Soudée: investigation. Claire Leloup: investigation. Cecile Debuisson: resources. Fanny Lanternier: investigation and visualization. Samuel Bitoun: data curation, formal analysis, investigation, validation, and writing—review and editing. Stephan Pavy: resources. Xavier Mariette: resources and writing—review and editing. Aniss Rafik: conceptualization, investigation, and resources. Hanaa Skhoun: conceptualization, investigation, and resources. Hanane El Ouazzani: conceptualization, investigation, and resources. Ismael Abderahmani-Ghorfi: conceptualization, investigation, and resources. Jamila El-Baghdadi: conceptualization, investigation, and resources. Andrés Baena: resources. Manuela Tejada-Giraldo: investigation. Luis Fernando Barrera: investigation and resources. Andrés Augusto Arias: investigation and resources. Giovanna Fabio: resources and visualization. Maria Carrabba: investigation and resources. Melike Emiroglu: data curation, resources, and supervision. Liliana Bezrodnik: resources. Loubna El Zein: investigation and resources. Hassan Hammoud: resources. Peter K. Gregersen: data curation. Benjamin Terrier: data curation and writing—review and editing. Rafael Leon Lopez: resources. Marion Touzet: investigation. Vincent Pestre: investigation and resources. Marlène Pasquet: resources. Lars Rogge: investigation, resources, and writing—review and editing. Michael Fayon: investigation and validation. François Galode: resources. Eric Jeziorski: resources. Darragh Duffy: resources and writing—review and editing. Lluis Quintana-Murci: resources and validation. Etienne Patin: resources, Charlotte Cunningham-Rundles: resources. Isabelle Meyts: data curation, funding acquisition, investigation, resources, and writing—review and editing. Shen-Ying Zhang: resources and writing—review and editing. Qian Zhang: resources. Emmanuelle Jouanguy: resources and writing—review and editing. Bertrand Boisson: resources. Jérémie Rosain: resources. Vivien Béziat: formal analysis. Mohammad Shahrooei: conceptualization, data curation, formal analysis, investigation, and validation. Seyed Alireza Mahdaviani: conceptualization, investigation, project administration, validation, and writing—review and editing. Nima Rezaei: conceptualization, data curation, investigation, methodology, project administration, supervision, validation, and writing—review and editing. Nima Parvaneh: investigation and resources. Zahra Chavoshzadeh: resources. Niloufar Yazdanpanah: resources. Nathalie Aladjidi: resources. Antoni Noguera-Julian: investigation and writing—review and editing. Ana Esteve-Solé: investigation, resources, and writing—review and editing. Laia Alsina: data curation, investigation, validation, and writing—review and editing. Davood Mansouri: resources. Sevgi Keles: data curation and investigation. Mediha Gonenc Ortakoylu: Conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—original draft, review, and editing. Deniz Aygun: data curation. Esra Yucel: resources. Ayca Kiykim: resources and writing—review and editing. Yildiz Camcioglu: data curation. Cindy S. Ma: funding acquisition, methodology, supervision, and writing—review and editing. Stuart G. Tangye: funding acquisition, project administration, supervision, validation, and writing—review and editing. Peng Zhang: data curation, formal analysis, and software. Laurent Abel: formal analysis and writing—review and editing, Peter D. Craggs: funding acquisition and project administration. Jean-Laurent Casanova: conceptualization, funding acquisition, project administration, resources, supervision, validation, and writing—original draft, review, and editing. Aurélie Cobat: methodology, supervision, validation, and writing—review and editing. Anne Puel: conceptualization, funding acquisition, resources, supervision, and writing—original draft, review, and editing. Jacinta Bustamante: conceptualization, formal analysis, funding acquisition, investigation, methodology, resources, supervision, and writing—original draft, review, and editing. Stephen J. Hill: funding acquisition, investigation, visualization, and writing—original draft, review, and editing. Stéphanie Boisson-Dupuis: conceptualization, data curation, formal analysis, funding acquisition, investigation, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.