Multiomics dissection of human RAG deficiency reveals distinctive patterns of immune dysregulation but a common inflammatory signature

Abstract

Human recombination-activating gene (RAG) deficiency can manifest with distinct clinical and immunological phenotypes. By applying a multiomics approach to a large group of RAG-mutated patients, we aimed at characterizing the immunopathology associated with each phenotype. Although defective T and B cell development is common to all phenotypes, patients with hypomorphic RAG variants can generate T and B cells with signatures of immune dysregulation and produce autoantibodies to a broad range of self-antigens, including type I interferons. T helper 2 (T_H2) cell skewing and a prominent inflammatory signature characterize Omenn syndrome, whereas more hypomorphic forms of RAG deficiency are associated with a type 1 immune profile both in blood and tissues. We used cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) analysis to define the cell lineage-specific contribution to the immunopathology of the distinct RAG phenotypes. These insights may help improve the diagnosis and clinical management of the various forms of the disease.

Fig. 1.. Clinical manifestations, type of mutations, and recombination activity in our cohort of RAG-mutated patients.

(A) Pie charts indicating the proportion of RAG-mutated patients in each of the groups (CID, LS, OS, and SCID) who suffered from bacterial, viral, and fungal infections, skin rash, granulomas, cytopenias, or other autoimmune manifestations. (B) Pie charts showing the proportion of missense, frameshift, nonsense, and insertion/deletion variants identified in RAG1 (left) and RAG2 (right) genes in our cohort of patients. (C) Recombination activity of individual mutant RAG variants, as compared with wild-type alleles. The variants are grouped on the basis of the clinical phenotype of the patients: CID (n = 90), LS (n = 63), OS (n = 81), and SCID (n = 54). Results are shown as violin scatter plots with median and quartile. Kruskal-Wallis H test, followed by Dunn’s multiple comparisons test. ****P < 0.0001.

Fig. 2.. Analysis of in vitro T cell development, distal TCRVα gene rearrangement usage, and BM B cell development in RAG-mutated patients.

(A) Graph showing absolute cell counts of T cell subsets in the ATOs. Box-and-whisker plots show median and 5 to 95% confidence intervals for all HDs analyzed (n = 11). Each symbol represents a patient’s sample (CID, n = 8; LS, n = 1; OS, n = 3; SCID, n = 1). (B and C) Representative flow cytometry plots and bar graphs showing the frequency of (B) CD3⁺TCRVα7.2⁺ cells in HDs (n = 45), RAG-mutated patients (CID, n = 35; LS, n = 23; OS, n = 20; SCID, n = 14), and heterozygous individuals (n = 34) and (C) the frequency of CD3⁺TCRVα7.2⁺CD161⁺ MAIT cells in HDs (n = 34), RAG-mutated patients (CID, n = 34; LS, n = 21; OS, n = 19; SCID, n = 14), and heterozygous individuals (n = 31). (D) Stacked bar graph summarizing the frequency of bone marrow (BM) B cell subsets in HDs (n = 5) and RAG-mutated patients (n = 17). (E) Bar graph with scatter dot plots summarizing the frequency of the combined pro-B and pre-B I cell subsets in HDs (n = 5) and RAG-mutated patients (n = 17). Bars [(B) to (E)] show mean values ± SD. (F) Violin plots with scatter dot plots showing the frequency of mature/recirculating B and plasmablasts/plasma cells (PB/PC) among total BM B cells in HDs (n = 5) and patients with SCID + OS (n = 5) and LS + CID (n = 12). Violin plots show median and quartiles. In all panels, data from individual experiments were pooled. Data were analyzed using [(A), (D), and (E)] Mann-Whitney U test, two-tailed, or [(B), (C), and (F)] Kruskal-Wallis H test, followed by Dunn’s multiple comparisons test. **P < 0.005, ***P < 0.001, and ****P < 0.0001.

Fig. 3.. Inflammatory profile in skin lesions of patients with OS and CID and distribution of T_H cell, T_FH cell, and ILC subsets in RAG-mutated patients.

(A and B) Representative [hematoxylin and eosin (H&E), CD4/CD8, GATA3, and T-bet] and RNAscope (CXCL9 and IFNG) of skin biopsies from a patient with (A) OS and (B) CID. Images are ×20 magnification, except for IFNG (×40). Scale bars, 50 μm. (C) Absolute eosinophil counts in the peripheral blood of patients with CID (n = 31), LS (n = 19), OS (n = 39), and SCID (n = 23). (D) IgE levels in the plasma of patients with CID (n = 28), LS (n = 16), OS (n = 30), and SCID (n = 16). The gray area in (C) and (D) indicates normal range in HDs. (E) CD162⁺CD4⁺CD45RO⁺ cell frequency in HDs (n = 53) and patients with CID (n = 31), LS (n = 20), and OS (n = 27). (F) CD162⁺CD8⁺CD45RO⁺ cell frequency in HDs (n = 52) and patients with CID (n = 31), LS (n = 21), and OS (n = 26). (G) T_H cell subset distribution in HDs (n = 64) and patients with CID (n = 29), LS (n = 19), and OS (n = 28). (H) T_FH cell subset distribution in HDs (n = 62) and patients with CID (n = 29), LS (n = 18), and OS (n = 27). (I) ILC subset distribution in HDs (n = 29) and patients with CID (n = 28), LS (n = 18), OS (n = 22), and SCID (n = 9). [(C) to (I)] Data from multiple independent experiments were pooled. Bars [(G) to (I)] show mean values ± SEM. Data were analyzed using [(C) to (F)] Kruskal-Wallis H test with Dunn’s multiple comparisons test. [(G) to (I)] Asterisks reflect significant differences with HDs as a result of mixed-effects model with Geisser-Greenhouse correction and Benjamini, Krieger, and Yekutieli’s multiple comparisons test. *P < 0.05, **P < 0.005, ***P < 0.001, and ****P < 0.0001.

Fig. 4.. Broad inflammatory signature of RAG-mutated patients.

(A) Radar chart depicting 22 biomarkers in RAG-mutated patients and HDs. The graph shows the median value for each biomarker in each group. The radar chart is scaled to the maximum and minimum value for each biomarker. Data were pooled from multiple independent experiments. VEGF, vascular endothelial growth factor. (B to E) Proteomic analysis in RAG-mutated patients and HDs. Graphs show the top 25 plasma proteins up-regulated in each group of RAG-mutated patients (n = 5 individuals per group) versus HDs: (B) CID, (C) LS, (D) OS, and (E) SCID. Each group of patients was compared with an age-matched group of HDs (n = 3 to 6). Results for all groups were obtained in one experiment. Top up-regulated proteins were identified by selecting all proteins with FDR < 0.05 and P < 0.05 (two-tailed t test) and then ordering them according to increased fold changes expressed in a log₂ scale (log₂FC). Heatmaps inside each graph show the most significantly enriched pathways for the group comparison, and the statistical significance is expressed as −log(P value). FGF2, fibroblast growth factor 2; MMP9, matrix metalloproteinase 9; GDF15, growth and differentiation factor 15; CRP, complement-reactive protein; ICAM-1, intercellular adhesion molecule–1; CSF1, colony-stimulating factor 1; TLR2, Toll-like receptor 2; FTH1.FTL, ferritin heavy chain 1.ferritin light chain; RPS3A, ribosomal protein S3A; FCRL3, Fc receptor like 3; PRTN3, proteinase 3; FUT5, fucosyltransferase 5; FCN1, ficolin 1; SELE, E-selectin; HGF, hepatocyte growth factor; FCGR3B, Fc gamma receptor IIIb; SECTM1, secreted and transmembrane 1; DPP7, dipeptidylpeptidase 7; LTA4H, leukotriene A4 hydrolase; SNA25, synaptosome associated protein 25; ENPP7, ectonucleotide phosphatase/phosphodiesterase 3; PIGR, polymeric immunoglobulin receptor; ASGR1, asialoglycoprotein receptor 1; FSTL3, follistatin like 3; TGM3, transglutaminase 3; AMIGO2, adhesion molecule with Ig like domain 2; REN, renin; RPSA, ribosomal protein SA; CTSZ, cathepsin Z; REG4, regenerating family member 4; FASLG, Fas ligand; CDKN1B, cyclin dependent kinase inhibitor 1B; HIST3H2A, histone gene cluster 3 H2A; H2AFZ, histone 2A family member Z; PI3, peptidase inhibitor 3; IMPDH1, inosine monophosphate dehydrogenase 1; PFDN5, prefoldin subunit 5; ANXA2, annexin A2; DCTPP1, DCTP pyrophosphatase 1; HSPA1A, heat shock protein family A member 1A; CHIT1, chitinase 1; HAVCR2, hepatis A virus cellular receptor 2; VWF, von Willebrand factor; ISG15, interferon stimulated gene 15; IDS, iduronate 2-sulfatase; GOT1, glutamic-oxaloacetic transaminase 1; MPO, myeloperoxidase; CTSH, cathepsin H; YWHAE, tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein epsilon; NSFL1C, NSFL1 cofactor; FCGR3B, Fc gamma receptor IIIb; LGALS3BP, galectin 3 binding protein; SPP1, secreted phosphoprotein 1; PAPPA, pappalysin 1; SIGLEC14, sialic acid binding Ig like lectin 14.

Fig. 5.. Characterization of the autoantibody profile in the plasma of RAG-mutated patients.

(A) Heatmap showing the top 50 peptides to which autoantibody reactivity was detected in plasma from RAG-mutated patients using the HuProt assay. Each column represents a patient sample: CID, n = 4; LS, n = 3; OS, n = 3; APECED, n = 22; thymoma, n = 20; HD, n = 10. Results for all groups were obtained in one experiment. PPP1R1B, protein phosphatase 1 regulatory inhibitory subunit 1B; CRYGS, crystallin gamma S; NECAP1, NECAP endocytosis associated protein 1; SERPINB6, serpin family B member 6; CHMP3, charged multivesicular body protein 3; LYSMD2, LysM domain containing protein 2; ESPN, espin; RNF8, ring finger protein 8; IFNA8, interferon alpha 8; TPSAB1, tryptase alpha/beta 1; APIG1, adaptor related protein complex 1 subunit gamma 1; USP7, ubiquitin specific peptidase 7; AP2A2, adaptor protein complex 2 subunit alpha 2; FICD, FIC domain containing protein; SERPINB3, serpin family B member 3; TAF10, TATA-box binding protein associated factor 10; EFHD2, EF-Hand domain family member D2; HES6, HES family BHLH transcription factor 6; CDC45, cell division cycle 45; UBA1, ubiquitin like modifier activating enzyme 1; PPP4R2, protein phosphatase 4 regulatory protein 2; PAGE2B, PAGE family member 2B; RFTN1, raftin lipid raft linker 1; MACROD2, mono-ADP ribosylhydrolase 2; RSPH1, radial spoke head component 1; ATP4A, ATPase H+/K+ transporting subunit alpha; SCG2, secretogranin 2; IQWD1, IQ motif and WD repeat-containing protein 1; CUEDC2_FRAG, CUE domain containing 2; ITGA4B7, integrin subunit alpha 4 beta 7; SNRK, SNF related kinase; HECTD3, HECT domain 3 ubiquitin protein ligase 3; SEPSECS, SEP (O-phosphoserine) TRNA:Sec (selenocysteine) TRNA synthase; FCHSD1, FCH and double SH3 domains 1; PSME3, proteasome activator subunit 3; LAMTOR5, late endosomal/lysosomal adaptor MAPK and MTOR activator 5; TP63, tumor protein P63; KLHL7, Kelch like family member 7; RBM38, RNA binding motif protein 38; KCTD14, potassium channel tetramerization domain containing 14; SOHLH2, spermatogenesis and oogenesis specific basic helix-loop-helix 2; ANXA11, annexin A 11; NT5C1A, 5′-nucleotidase cytosolic IA. (B to E) Box-and-whisker plots with scatter dot plots showing the fluorescence intensity of (B) anti–IFN-α, (C) anti–IFN-ω, and (D) anti–IFN-β antibodies detected in the plasma of HDs (n = 43) and patients with CID (n = 38), LS (n = 29), OS (n = 30), and SCID (n = 29). (E) Box-and-whisker plots with scatter dot plots showing the quantification of anti–IFN-λ antibodies as reactive luciferase units in the plasma of HDs (n = 67) and patients with CID (n = 33), LS (n = 22), OS (n = 24), and SCID (n = 23). [(B) to (E)] Data from multiple independent experiments were pooled. The dashed lines in (B) to (E) represent the upper limit of normal defined as means + 3 SD in HDs. Asterisks in (B) to (E) reflect P values results of Kruskal-Wallis H test, followed by Dunn’s multiple comparisons test. *P < 0.05, **P < 0.005, ***P < 0.001, and ****P < 0.0001.

Fig. 6.. Analysis of B cell subset distribution and dysreactive B cells in peripheral blood of RAG-mutated patients.

(A) Stacked bar graph showing the distribution of naïve and memory B cell subsets in the peripheral blood of HDs (n = 7) and patients with CID (n = 16) and LS (n = 5). Bars show mean values ± SEM. (B) Representative plots showing the frequency of CD19⁺CD21^lo cells gated on total CD19⁺ cells. (C and D) Bar graphs with scatter dot plots summarizing the frequency of the CD21^lo in total B cells (C) and the mean fluorescence intensity (MFI) of CD19 in CD19⁺CD21^lo cells (D). (E and F) Graphs show the correlation between the frequency of CD21^lo B cells and the concentration of (E) CXCL9 or (F) the frequency of T_FH1 cells. Linear regression curve is shown in each graph along with the Spearman correlation factor (r) and P value. (G to I) Bar graphs showing the frequency of atypical B cells: (G) aNaB, (H) DN2, and (I) aMeB cells. (J) Representative plots showing the frequency of 9G4-positive cells (9G4⁺, 9G4^hi, and 9G4^int) gated on total CD19⁺ cells. (K to M) Bar graphs showing the frequency of (K) 9G4⁺, (L) 9G4^int, and (M) 9G4^hi cells gated on total CD19⁺ B cells. (N and O) Graphs showing (N) the frequency of 9G4⁺ cells among CD19⁺ and CD19⁺CD21^lo cells and (O) the frequency of CD21^lo cells among CD19⁺ and 9G4⁺ cells in patients with CID and LS. The lines connect the values found in the same individual. [(A) to (D) and (G) to (O)] Data were pooled from two independent experiments. [(C), (D), (G) to (I), and (K) to (M)] Bars represent means ± SD. (A) Mixed-effects model with Geisser-Greenhouse correction and Benjamini, Krieger, and Yekutieli’s multiple comparisons test. [(C), (D), (G) to (I), and (K) to (M)] Mann-Whitney U test, two-tailed. [(N) and (O)] Spearman correlation test. *P < 0.05, **P < 0.005, ***P < 0.001, and ****P < 0.0001.

Fig. 7.. Single-cell CITE-seq profiling of peripheral blood cells of RAG-mutated patients.

(A) Uniform Manifold Approximation and Projection (UMAP) visualization of single-cell clusters based on surface protein expression profiles, obtained from PBMCs of HDs and patients combined. (B) Visualization of cells from each of the patient groups projected onto the same UMAP. (C) Distribution of the different CD4 T cell, CD8 T cell, NK cell, B cell, and monocyte subclusters in HDs and in each of the patient groups. (D) Cell type–specific GSEA map obtained by comparing each group of patients versus the HDs for the indicated cell type. Hallmark gene sets/pathways significantly enriched in RAG-mutated patients as compared with HDs are represented in each row, whereas each column indicates the cell type and patient group for which a difference versus HDs has been observed. Dot color denotes normalized gene set enrichment score, and size indicates −log₁₀(adjusted P value). P values were computed from GSEA test of the whole gene sets and adjusted using the Benjamini-Hochberg method. NES, normalized enrichment score. (E) Violin plots showing the single-cell distribution of gene set score of selected Hallmark pathways (TNF-α signaling via NF-κB on the left and IFN-γ response on the right) in HDs and in each group of RAG-mutated patients in monocytes (top row) and NK cells (bottom row). (F) Average expression and percentage of cells expressing the transcription factors (TBX21 and GATA3) among CD4 T memory cells in each of the patient groups and in HDs. Results were from a single experiment in all groups of patients and controls.