Human CARMIL2 deficiency underlies a broader immunological and clinical phenotype than CD28 deficiency

Abstract

Patients with inherited CARMIL2 or CD28 deficiency have defective T cell CD28 signaling, but their immunological and clinical phenotypes remain largely unknown. We show that only one of three CARMIL2 isoforms is produced and functional across leukocyte subsets. Tested mutant CARMIL2 alleles from 89 patients and 52 families impair canonical NF-κB but not AP-1 and NFAT activation in T cells stimulated via CD28. Like CD28-deficient patients, CARMIL2-deficient patients display recalcitrant warts and low blood counts of CD4+ and CD8+ memory T cells and CD4+ TREGs. Unlike CD28-deficient patients, they have low counts of NK cells and memory B cells, and their antibody responses are weak. CARMIL2 deficiency is fully penetrant by the age of 10 yr and is characterized by numerous infections, EBV+ smooth muscle tumors, and mucocutaneous inflammation, including inflammatory bowel disease. Patients with somatic reversions of a mutant allele in CD4+ T cells have milder phenotypes. Our study suggests that CARMIL2 governs immunological pathways beyond CD28.

Graphical abstract

Figure 1.

Genetic analysis demonstrates that the CARMIL2 isoform 3 is canonical and that the isoform 1 is pathogenic. (A) Schematic representation of the three CARMIL2 isoforms. Isoform 1 (ENST00000334583.11, transcript 1) encodes a 1435-amino acid protein and has 38 exons. Isoform 2 (ENST00000545661.5) encodes a 1372-amino acid protein and has 38 exons. Isoform 2 lacks 108 nucleotides present in isoform 1 due to an additional intron within exon 14. It also loses the whole of exon 36 but retains the same open reading frame. Isoform 3 (not reported in the Ensembl database) encodes a 1399-amino acid protein and has 39 exons. Relative to isoform 1, isoform 3 loses part of exon 14, but retains exon 36. (B) CARMIL2 murine and human isoform 1-3 crystal structures. (C) scRNA-seq UMAP clustering of PBMCs from 6 HDs showing normalized CARMIL2 expression in major cell lineages. (D) Reclustering of the NK and T cell cluster from C identifying the different cellular subsets, with superimposed CARMIL2 levels. (E) Bubble graph presenting the percentage of cells for which CARMIL2 transcripts are detected by scRNA-seq for each cluster presented in C and D. Bubble size indicates median relative expression for CARMIL2-positive cells. (F) Genome browser snapshot showing bulk RNA-seq coverage for sequence reads aligned to the CARMIL2 gene. The structure of the two known isoforms is shown at the top; blue rectangles represent exons and connecting lines represent introns. The structure of isoform 3 is also shown; this was the predominant isoform detected in the cells analyzed. The insets show sequence coverage and the number of detected spliced reads for exons 13 to 16 and exons 35 to 37. Monos, monocytes. (G) CARMIL2 expression, assessed by FACS (N = 3) and Western blotting (N = 2) in a CARMIL2-deficient Jurkat T cell line, after transduction with an empty lentivirus, or lentivirus encoding each of the three CARMIL2 isoforms. (H) Phospho-p65 and phospho-SLP76 levels in cells, as described in G, following stimulation with anti-CD3, or anti-CD3/CD28 mAb. (N = 3). (I) Schematic representation of the CARMIL2 gene isoform 1 (ENST00000334583.11), isoform 2 (ENST00000545661.5), and isoform 3, for exon 14. The G>C substitution is indicated in position c.1154 for isoform 1, and c.1149 + 5 for isoforms 2 and 3. (J) Total mRNA was extracted from the T cell blasts of a HD and P42 (c.1149 + 5G>C). Total mRNA was subjected to RT-qPCR for the assessment of total CARMIL2 expression. Data are displayed as 2^−ΔCt values after normalization against endogenous GUS control gene expression. Mean ± SEM of three technical replicates (N = 2). (K) PCR amplification of the full-length CARMIL2 cDNA from T cell blasts from a HD and P42. Electropherograms show an insertion of 108 nucleotides into exon 14 of the patient’s sequence. The predicted impact on the translated protein in P42 corresponds to the exact sequence of isoform 1. (L) CARMIL2 protein levels in T cell blasts, as assessed by intracellular FACS on CD4⁺ T cells for a HD, P42 and P43, both homozygous for the G>C substitution shown in I. Source data are available for this figure: SourceData F1.

Figure 2.

Germline CARMIL2 mutations. (A) Schematic representation of the CARMIL2 gene (isoform 3), with the 49 mutations studied. Functional domains are also indicated: PH, pleckstrin homology domain; HDD, homodimerization domain; CPI, capping protein-interacting domain; PRR, proline-rich repeat domain; C-cap, C-terminal cap of the LRR; N-cap, N-terminal cap of the LRR. (B) Missense and in-frame deletion mutations mapped onto the crystal structure of CARMIL2 isoform 3, modeled from murine Rltpr (4K17) with SWISS-MODEL (Waterhouse et al., 2018). The c.871G>C splicing variant is not shown. (C) Consensus negative score of CARMIL2 and its distribution for genes causing inborn errors of immunity (IEI), according to disease mode of inheritance. AD, autosomal dominant; AR, autosomal recessive.

Figure S1.

In vitro validation of CARMIL2 variants. (A) CADD score (y axis) plotted against MAF (x axis) for the homozygous variants present in the gnomAD database (http://gnomad.broadinstitute.org). The variants found in gnomAD and tested functionally are annotated in black, and the mutations confirmed in CARMIL2-deficient patients are annotated in red. MSC, mutation significance cutoff; NR, not reported. (B) Total mRNA was extracted from the T cell blasts of P83 and a HD, amplified with RT-PCR, and subjected to Sanger sequencing. (C) Total mRNA was extracted from the T cell blasts of a HD and P31, and subjected to RT-PCR, followed by TA cloning. PCR amplification of the CARMIL2 cDNA is shown, with GAPDH cDNA as a positive control. (D) Exon-trapping experiments were conducted for the c.871G>C allele. Control genomic DNA was inserted into the pspl3 plasmid (WT) and subjected to site-directed mutagenesis to obtain the c.871G>C-encoding plasmid (Mut). Total mRNA was extracted from COS-7 cells transfected with the WT and Mut plasmids and subjected to RT-PCR. PCR amplification of the CARMIL2 cDNA, with aberrant splicing detected in 100% of the screened colonies. HIV-Tat are exons from the pspl3 plasmid. N = 1. (E) Exon-trapping was performed on genomic DNA extracted from the T cell blasts of a HD and P4 (c.1128C>T) and subjected to TA cloning. Schematic representation of the PCR products, showing abnormal splicing in 100% of the screened colonies, relative to control cells. (F) Total mRNA was extracted from the T cell blasts of a HD and P1 (c.1578C>T) and subjected to RT-PCR. PCR amplification and Sanger sequencing of the CARMIL2 cDNA showed an aberrant product in P1. N = 1. (G) Total mRNA was extracted from the T cell blasts of three HDs, P1, and P31. Total mRNA was subjected to RT-qPCR for the assessment of total CARMIL2 expression. Data are displayed as 2^−ΔCt values after normalization according to endogenous GUS control gene expression. The bar represents the mean value in controls. N = 1. (H) Western blot analysis of CARMIL2 levels in total cell extracts from HEK293T cells transfected with a pcDNA3.1 plasmid, either empty (EV) or containing WT isoform 3 (WT), or mutant forms found in CARMIL2-deficient patients. Two Abs were used: an Ab against the N-terminus of CARMIL2, and an Ab against GAPDH. The data shown are representative of two independent experiments. (I) Western blot analysis of CARMIL2 levels in total cell extracts from non-transduced CARMIL2 KO Jurkat T cells, or after transduction with an empty lentivirus (EV), a lentivirus encoding WT isoform 3 (WT), or the missense and in-frame deletion variants identified in CARMIL2-deficient individuals (N = 2). Source data are available for this figure: SourceData FS1.

Figure 3.

In vitro and ex vivo validation of CARMIL2 alleles. (A) Western blot analysis of CARMIL2 levels in total cell extracts from HEK293T cells transfected with a pcDNA3.1 plasmid, either empty (EV), or containing the WT isoform 3 (WT) or mutant forms, including all the missense and in-frame deletion variants found in CARMIL2-deficient individuals. Two Abs were used: an Ab against the C-terminus of CARMIL2 (EM-53), and an Ab against GAPDH. The data shown are representative of two independent experiments. (B) CARMIL2 protein levels in Jurkat T cells, as assessed by intracellular FACS, after transduction with an empty lentivirus, or a lentivirus encoding the WT isoform 3 (WT) or the missense mutants found in CARMIL2-deficient individuals. Mean fluorescence intensity (MFI; top panel) and FACS histograms (lower panel) for CARMIL2 (N = 2). (C) Phospho-p65 in Jurkat T cells, as described in B, following stimulation with anti-CD3 mAb with or without anti-CD28 mAb. The data represent the ratio of the MFI after subtraction of the MFI in the not-stimulated cells. The bar represents the mean. Error bars represent the SD. N = 3. (D) CARMIL2 protein levels, as assessed by intracellular FACS, in CD4⁺ and CD8⁺ T cells, for a representative HD, and patients homozygous for five missense mutations, compound heterozygous for three missense, frameshift and splice-site mutations, two in-frame deletion mutations, and one stop codon mutation. Source data are available for this figure: SourceData F3.

Figure S2.

Functional assays in Jurkat T cells, and counts of leukocytes in blood. (A) Phospho-p65 levels in CARMIL2 KO Jurkat T cells, as assessed by intracellular FACS, after transduction with an empty lentivirus (EV), a lentivirus encoding CARMIL2 WT isoform 3 (WT), or the indicated missense mutants identified in CARMIL2-deficient individuals, in the absence of stimulation or following stimulation with the mAb indicated. MFI and histograms are shown. The bar represents the mean. Error bars represent the SD. N = 3. (B) Phospho-p65 levels in CARMIL2 KO Jurkat T cells, as assessed by intracellular FACS, after transduction with an empty lentivirus, a lentivirus encoding WT isoform 3 (WT), or homozygous missense variants found in gnomAD (MAF higher than 10⁻⁵, and a CADD score above 20), following stimulation with anti-CD3 mAb with or without anti-CD28 mAb. Histograms are plotted, with the bar representing the mean. Error bars represent the SD. N = 3. CyTOF was conducted after the exclusion of dead cells from fresh blood isolated from 38 adult and 11 pediatric healthy controls, 9 adults and 7 children with CARMIL2 deficiency, and 3 CD28-deficient patients. (C−M) Box-and-whisker plots are shown for the counts of each subset of leukocytes. The whiskers indicate the maximum and minimum values. The bars represent the mean value. P values indicate significant differences (in Mann–Whitney tests, or t tests): *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. T_N, naive T cells; T_FH, follicular helper T cells.

Figure 4.

CARMIL2 deficiency affects lymphocyte development more profoundly than CD28 deficiency. CyTOF was conducted after the exclusion of dead cells from fresh blood isolated from 38 adult and 11 pediatric healthy controls, 9 adults and 7 children with CARMIL2 deficiency, and 3 CD28-deficient patients. (A) UMAP plots are presented with legends indicating the different leukocyte subsets, as defined by surface markers. (B−G) Box-and-whisker plots are shown for the counts of each subset of leukocytes. The whiskers indicate the maximum and minimum values, with the bars indicating the mean. P values indicate significant differences (Mann–Whitney test, or t test): *, P < 0.05; **, P < 0.01; ****, P < 0.0001. T_N, naive T cell; T_FH, follicular helper T cell. (H) Phospho-p65 levels, as assessed by intracellular FACS on CD4⁺ T cells from a HD and P8 (N41Kfs*47), upon stimulation with anti-CD3 mAb with or without anti-CD28 mAb, or PMA. (I) Phospho-p65 levels, as assessed by intracellular FACS on T cell blasts from P11 (Q817*) transduced with an empty lentivirus or a lentivirus encoding the WT isoform 3 (WT), following stimulation with anti-CD3 mAb with or without anti-CD28 mAb, or PMA. Results are shown for CD4-expressing cells also positive for GFP. (J and K) Phospho-p65 levels, as assessed by intracellular FACS after PMA stimulation in T cell blasts from a HD and P2 (L636Afs*39) transduced with an empty lentivirus (EV), a lentivirus encoding WT isoform 3 (WT), or the variants indicated, identified in CARMIL2-deficient individuals. Results are shown for CD4-expressing cells also positive for GFP. Representative results from two experiments are shown. (L and M) Proliferation of sorted naive (defined as CD3⁺ CD45RA⁺ CCR7⁺ cells) CD4⁺ (L) and CD8⁺ (M) T cells following 4 d of incubation with anti-CD3 mAb, anti-CD3 + anti-CD28 mAb, or P/I. Proliferation was assessed by CFSE dilution. Representative flow plots for P51 and a HD are depicted. The bar represents the mean. Error bars represent the SD.

Figure S3.

Frequencies of leukocytes in blood, and functional assays in primary CD4⁺ T cells. CyTOF was conducted after the exclusion of dead cells from fresh blood isolated from 38 adult and 11 pediatric healthy controls, 9 adults and 7 children with CARMIL2 deficiency, and 3 CD28-deficient patients. (A−K) Box-and-whisker plots are shown for the frequencies of each subset of leukocytes. The whiskers indicate the maximum and minimum values. The bars represent the mean. P values indicate significant differences (Mann-Whitney tests, or t tests): *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. T_N, naive T cells; T_FH, follicular helper T cells. (L) CARMIL2 levels (upper panel), as assessed by intracellular FACS on T cell blasts from a HD and P3 (L636Afs*39) transduced with an empty lentivirus (EV), or lentiviruses encoding CARMIL2 isoforms 1, 2, and 3. (M) Phospho-p65 levels (lower panel), as assessed by intracellular FACS on T cell blasts from P3 (L636Afs*39) transduced with an empty lentivirus (EV), or lentiviruses encoding CARMIL2 isoforms 1, 2, and 3 with or without stimulation with PMA (representative results from two experiments are shown). (N) CARMIL2 levels, as assessed by intracellular FACS on T cell blasts from a HD and from P2 (L636Afs*39) transduced with an empty lentivirus (EV), or lentiviruses encoding the missense and in-frame deletion variants found in CARMIL2-deficient individuals. N = 1.

Figure S4.

T cell function is impaired in patients. (A and B) T cell proliferation in CD4⁺ (A) and CD8⁺ (B) T cells from 17 CARMIL2-deficient patients and 19 HD, upon stimulation with anti-CD3 mAb, anti-CD3 + anti-CD28 mAb, or P/I. Representative flow plots for the proliferation of CD4⁺ (A) and CD8⁺ (B) T cells from P71 and a HD are shown. Proliferation was assessed by CFSE dilution. The bars represent the mean. The error bars represent the SD. P values were calculated using Kruskal–Wallis test alongside Dunn’s correction for multiple comparisons with *, P < 0.05; **, P < 0.01; ***, P <0.001; and ****, P < 0.0001. (C and D) MFI of CD25 in CD4⁺ T cells (C) and CD8⁺ T cells (D) from PBMCs upon stimulation with anti-CD3 mAb, anti-CD3 + anti-CD28 mAb, or P/I. (E–H) Proliferation of CD4⁺ T cells (E) and CD8⁺ T cells (G) within PBMCs, or sorted naive (defined as CD3⁺ CD45RA⁺ CCR7⁺ cells) CD4⁺ T cells (F) and CD8⁺ T cells (H) following 4 d of incubation with the abovementioned stimuli and the addition of IL-2 (500 IU/ml). Representative flow plots for P71 and a HD are depicted. The bars represent mean. The error bars represent the SD. (I and J) MFI of CD25 in CD4⁺ T cells (I) and CD8⁺ T cells (J) stimulated with additional IL-2 (500 IU/ml) in the PBMCs of eight CARMIL2-deficient patients and five HDs. The bars represent mean. The error bars represent the SD. Statistical analysis was performed with Kruskal–Wallis tests and Dunn’s correction for multiple comparisons with *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

Figure 5.

CARMIL2 regulates the expression of a restricted set of genes downstream from CD28 signaling. Impact of CARMIL2 deficiency on the transcriptome of sorted naive CD4⁺ T cells. RNA-seq data. Heatmap showing log₂ FC in expression in stimulated naive CD4⁺ T cells at 2 h. (A) Only genes differentially expressed in response to stimulation in controls (adjusted P < 0.05 and | log₂ FC | > 2) upon simulation with PMA are shown (PMA dependent). (B) Only genes both PMA dependent and differentially expressed in CARMIL2-deficient patients (adjusted P < 0.05 and log₂ FC < −2) are shown (PMA and CARMIL2 dependent). (C) Heatmap showing the genes differentially expressed in control naive CD4⁺ T cells (adjusted P < 0.05 and | log₂ FC | > 1) upon stimulation with anti-CD3 + anti-CD28 Abs versus anti-CD3 mAb alone (CD28-dependent genes). (D) Heatmap showing the CD28-dependent genes (defined in C) differentially expressed in CARMIL2-deficient patients relative to controls (adjusted P < 0.05 and log₂ FC < −1). (E) Extraction from the RNA-seq data of representative target transcripts of the NF-κB (NFKB1), AP-1 (FOSL2), and NFAT (FASLG) pathways upon stimulation as indicated. (F) RT-qPCR on naive CD4⁺ T cells from four controls, three CARMIL2-deficient patients, and one CD28-deficient patient at 0, 2, and 8 h of activation with anti-CD3 mAb with or without anti-CD28, or with anti-CD2 mAb. Data are displayed as 2^−ΔCt after normalization against GUS expression. Statistical analysis was performed at 8 h for IL2, NFKB1, and FASLG, and at 2 h for FOSL2. Error bars represent the SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001. N = 1.

Figure 6.

B cell function is impaired in individuals with CARMIL2 deficiency. (A) Frequencies of CARMIL2-deficient patients with serum Ig levels (IgG, IgA, and IgM) below the normal range. (B) Frequencies of CARMIL2-deficient patients with low serological titers of Abs against booster vaccines for tetanus toxoid, diphtheria, and pneumococcus. (C) Adjusted virus scores (Virus Scoreadj) for samples from 20 adult in-house controls, 23 CARMIL2-deficient patients, and 7 CD28-deficient patients, together with mock IP samples, IgG-depleted serum, and IVIg. Virus species for which at least one sample was seropositive (i.e., with Ab levels above our species-specific significance cutoff) are shown (y axis). The heatmap plot shows the z-score values for each sample on a color gradient; in blue if Abs were detected but the counts of non-homologous enriched peptides were below our significance cutoff values, and in purple to red if the z-score values were at least 1× higher (purple) or >2× higher (red) than our significance cutoff values. The bar plot on the bottom illustrates the size of the Ab repertoire for a given sample: the precise number of different species for which peptide enrichment was observed (gray) and the number of different species for which the z-score values passed the significance cutoff (blue). (D) Principal component analysis (PCA) scatter plot for virus scores, color-coded for different groups of samples.

Figure 7.

Clinical manifestations of CARMIL2 deficiency. (A and B) Penetrance of clinical symptoms (A) and of infectious versus non-infectious complications (B) by age in 86 CARMIL2-deficient patients. (C) Clinical description of the first symptoms at disease onset. (D) Kaplan–Meier curve depicting the survival of CARMIL2-deficient patients who did not undergo hematopoietic stem cell transplantation, with the small vertical bars indicating age at last follow-up. The dashed lines indicate 95% confidence intervals. (E and F) Clinical description of infectious complications (E) and non-infectious complications (F) in 87 CARMIL2-deficient individuals at evaluation. CMC, chronic mucocutaneous candidiasis; URTI, upper respiratory tract infections; LRTI, lower respiratory tract infections; Eo, eosinophilic.

Figure 8.

Evidence for somatic reversions. (A) FACS plots (left) and histograms (right) showing CARMIL2 expression in the subsets of naive and memory CD4⁺ T cells, as defined by the surface expression of CD27 and CD45RA, in P57 and a HD. (B) Electropherograms show the DNA sequences in the sorted CARMIL2-negative (upper panel) and CARMIL2-expressing (lower panel) T cell blasts from P57. We detected a somatic reversion at the c.958 + 1 locus in the CARMIL2-expressing cells. (C) Frequencies of CARMIL2-expressing revertant naive and memory CD4⁺ T cells in five patients with CARMIL2 deficiency, as defined by the surface expression of CD27 and CD45RA, determined by FACS. (D) FACS plots showing T_REG, as defined by the expression of CD25 and FOXP3, in P57 and a HD. (E) Frequencies of T_REG, naive (T_N), and subsets of memory CD4⁺ T cells in CARMIL2-deficient patients with (colored dots) and without (black dots) somatic reversion events.

Figure S5.

Validation of CARMIL2 somatic reversions. (A) CARMIL2 protein levels in PBMCs subsets of CD4⁺ T cells from a HD, P1 (c.1578C>T), and P10 (Q817*). A somatic event is suspected in CD45RA⁻CCR7⁺ T_CM and CD45RA⁻CCR7⁻ T_EM CD4⁺ T cells from P1 as shown by comparison with CD45RA⁺CCR7⁺ naive CD4⁺ T cells. (B) Sanger sequencing of sorted CD45RA⁻CARMIL2⁺ cells from P1 confirmed a somatic heterozygous C>G substitution at the c.1578 locus (p.C526W). (C) CARMIL2 protein levels in Jurkat T cells, as assessed by intracellular FACS, following transduction with an empty lentivirus (EV), WT isoform 3 (WT), or the p.C526W missense mutant. MFI is shown. (D) Phospho-p65 in Jurkat T cells, as described in C, following stimulation with anti-CD3 mAb with or without anti-CD28 mAb. MFI is shown. N = 2. (E) Heterozygous reversion to the WT sequence in CARMIL2-reexpressing CD4⁺ T cell blasts from P53. (F) Restoration of the reading frame in CARMIL2-reexpressing CD4⁺ T cell blasts from patient 78 through insertion of an additional nucleotide (G*) at cDNA-position 897, which is absent in the T cell blasts without CARMIL2 expression. (G) Evidence of somatic reversion in the CD8⁺ T_CM compartment of P53.