Human inherited PD-L1 deficiency is clinically and immunologically less severe than PD-1 deficiency

Abstract

We previously reported two siblings with inherited PD-1 deficiency who died from autoimmune pneumonitis at 3 and 11 years of age after developing other autoimmune manifestations, including type 1 diabetes (T1D). We report here two siblings, aged 10 and 11 years, with neonatal-onset T1D (diagnosed at the ages of 1 day and 7 wk), who are homozygous for a splice-site variant of CD274 (encoding PD-L1). This variant results in the exclusive expression of an alternative, loss-of-function PD-L1 protein isoform in overexpression experiments and in the patients' primary leukocytes. Surprisingly, cytometric immunophenotyping and single-cell RNA sequencing analysis on blood leukocytes showed largely normal development and transcriptional profiles across lymphoid and myeloid subsets in the PD-L1-deficient siblings, contrasting with the extensive dysregulation of both lymphoid and myeloid leukocyte compartments in PD-1 deficiency. Our findings suggest that PD-1 and PD-L1 are essential for preventing early-onset T1D but that, unlike PD-1 deficiency, PD-L1 deficiency does not lead to fatal autoimmunity with extensive leukocytic dysregulation.

Graphical abstract

Figure S1.

Genetic analysis of the two siblings with neonatal-onset type 1 diabetes. (A) Ancestry-level PCA. PC scores were computed from the Human Genome Diversity Project and 1,000 Genomes phase 3 global genomic reference population datasets (Auton et al., 2015; Bergström et al., 2020) and local WGS data for the two affected siblings and their healthy relatives. The PCA plot is colored according to the reported genetic ancestry group (for controls). Individuals from the family studied are represented as yellow crosses outlined in black. (B) Country-level analysis. The previously reported dataset (Henn et al., 2012) was used as a reference. The PCA plot is colored according to the reported country of origin (for controls). Individuals from the family studied are represented as yellow crosses outlined in black.

Figure 1.

Two siblings homozygous for a splice-site variant of CD274. (A) The pedigree. Black symbols indicate affected individuals. Genotypes for the CD274 allele are also shown. WT: wild-type. M: mutant. E?: unknown. (B) Validation of the variant by Sanger sequencing. (C) Gene-level negative selection. PDCD1, CD274, and PDCD1LG2 (encoding PD-1, PD-L1, and PD-L2, respectively) are not under negative selection, as shown by CoNeS score (Rapaport et al., 2021), as also reported for other genes for which mutations underlie AR IEI. CTLA4 is also shown, as an example of another gene under negative selection. (D) Population genetics of PDCD1, CD274, and PDCD1LG2. The MAF and CADD scores for all non-synonymous variants found in the gnomAD database are depicted. All biallelic variants are labeled with their predicted protein-level consequences. The horizontal dotted line indicates the MSC (Itan et al., 2016; Kircher et al., 2014).

Figure 2.

Analysis of the effect of the CD274 splice-site variant on mRNA splicing in an overexpression system. (A) Schematic diagram of an exon-trapping assay. A region of genomic DNA flanking the fourth exon of the canonical CD274 isoform with or without the c.682+1G>A splice-site in the homozygous state was inserted into the pSPL3 vector. The plasmids were used to transfect HEK293T cells and, 24 h later, the spliced mRNA product was recovered by RT-PCR and TOPO cloning, and subjected to Sanger sequencing. (B) Exon trapping. The schematic diagram shows the four types of cDNA identified, with the number of nucleotides in each region indicated. Representative data from two experiments are shown. (C) A schematic diagram of the CD274 mRNA and PD-L1 protein. Exon 1 is omitted because it contains no coding sequence. The red rectangle depicts the 51-amino acid in-frame deletion caused by the c.682+1G>A variant. SP, signal peptide; EC, extracellular domain; TM, transmembrane domain; IC, intracellular domain. Source data are available for this figure: SourceData F2.

Figure 3.

Analysis of the PD-L1 protein with in-frame deletion in an overexpression system. (A and B) PD-L1 protein levels. Raji B-lymphoma cells were lentivirally transduced with cDNA encoding the WT or a mutant PD-L1 isoform, or with EV, and were then subjected to selection on puromycin. PD-L1 protein levels were determined by (A) immunoblotting and (B) flow cytometry with monoclonal antibodies (mAb) against PD-L1. In B, a vertical dotted line within a histogram indicates the median. Representative results from two experiments are shown. (C and D) PD-1:PD-L1-mediated suppression assay. (C) Schematic diagram. HuT78 T-lymphoma cells lentivirally transduced with EV or with WT PD-1 were cocultured with Raji cells transduced with EV or a WT or mutant PD-L1 isoform for 24 h without stimulation or with blinatumomab (CD3-CD19 bispecific antibody, BiTE). Secretion inhibitors were added for the last 6 h. IFN-γ production was quantified by intracellular flow cytometry. The effect of anti-PD-L1 neutralizing mAb (equivalent to atezolizumab) or its isotype control was also assessed in this system. (D) Summary plot. The readout (percentage of IFN-γ⁺ HuT78 cells) was normalized against the mean in the “BiTE plus anti-PD-L1 antibody” group. Results from two independent experiments with 12 technical replicates in total were compiled. Statistical significance was determined for differences between EV and each PD-L1 construct in BiTE-stimulated conditions by two-tailed Wilcoxon’s rank sum tests with FDR adjustment. n.s., not significant. ****, P < 0.0001. Source data are available for this figure: SourceData F3.

Figure 4.

Analysis of endogenously expressed CD274 mRNA and PD-L1 protein in the patients’ leukocytes. (A, B, and D) Bulk RNASeq analysis. PBMCs from the two PD-L1-deficient siblings (ages 11 and 10 years), and adult and age-matched controls were either left non-stimulated or were stimulated with lipopolysaccharide (LPS), anti-CD2/CD3/CD28 mAb cocktail, or phorbol 12-myristate 13-acetate and ionomycin (P/I) for 24 h. (A) A schematic diagram of the CD274 mRNA exon 3-4-5 splice junctions in the cells of healthy donors (canonical) and the patients (alternative). (B) Ratio of read counts supporting the canonical and alternative exon 4–5 splice junction to read counts for the exon 3–4 splice junction. (C) RT-PCR products with a primer pair amplifying the whole CD274 coding sequence derived from the total RNA of PBMCs stimulated with LPS for 24 h. (D) Expression levels (transcripts per million reads; TPM) for each CD274 exon. (E) Western blot analysis for PD-L1 in PBMCs. PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and two healthy controls were either left non-stimulated or were stimulated with PHA overnight. Cell lysates were either left untreated or were treated with PNGase F, as indicated. (F) Densitometry results for the western blot shown in E. Values are normalized against the density of the loading control (HSP90). (G) Surface PD-L1 expression. PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years) and adult and age-matched controls were either left non-stimulated or were stimulated with IFN-α2, LPS, or anti-CD3/CD28 mAb-conjugated beads for 24 h. The level of PD-L1 expression on the surface of the cells of the different leukocyte subsets was determined by flow cytometry with two different mAbs against human PD-L1. (H) IFN-γ neutralization assay. PBMCs from healthy controls were either left non-stimulated or were stimulated with anti-CD3/CD28 mAb-conjugated beads for 24 h in the presence of anti-IFN-γ neutralizing mAb or its isotype control. PD-L1 levels were determined by flow cytometry with the 29E.2A3 clone. The horizontal dotted line indicates the level of background fluorescence determined with an isotype control for 29E.2A3. In B and G, bars represent the mean and SEM. In A–D and G, the experiments were performed once. In E and F, representative data from two experiments are shown. In H, results from three experiments (six donors in total) with technical duplicates are compiled. Statistical significance was determined for differences between IFN-γ neutralization and isotype control in anti-CD3/CD28-stimulated conditions by two-tailed paired Wilcoxon signed rank tests with FDR adjustment. **, P <0.01. Source data are available for this figure: SourceData F4.

Figure S2.

Analysis of CD274 mRNA in the patients’ leukocytes. (A) Schematic diagram of the design of RT-PCR primers and TaqMan probes. (B) RT-PCR on the CD274 CDS from total RNA extracted from PBMCs stimulated with LPS for 24 h. The 153nt deletion observed in bulk RNASeq data was confirmed by Sanger sequencing. (C) Bulk RNASeq for stimulated PBMCs. PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), an age-matched control, and several healthy adult controls were either left unstimulated or were stimulated with LPS, anti-CD2/3/28 mAb cocktail, or PMA/ionomycin (P/I) for 24 h. Total RNA was used to prepare libraries for bulk RNASeq. TPM is shown for each CD274 exon. (D) Quantitative PCR on the cDNA derived from the total RNA extracted from the LPS-stimulated PBMCs analyzed in C. GUSB was used as an endogenous control. Ratios of results for two TaqMab probes, targeting the exon 1–2 or 6–7 junction, are shown. The non-stimulated conditions were analyzed twice (technical replicates). (E) Bulk RNASeq on whole-blood leukocytes. Freshly drawn venous blood samples from the two PD-L1-deficient siblings (aged 11 and 10 years) and age-matched controls were used for total RNA extraction. Globin-depleted total RNA was used for sequencing. TPMs per region (CD274 exons 1–4 and 5–7) are shown. (F) IFN-γ neutralization assay. PBMCs from healthy controls were either left non-stimulated or were stimulated with anti-CD3/CD28 mAb-conjugated beads for 24 h in the presence of anti-IFN-γ neutralizing mAb or its isotype control. PD-L1 levels were determined by flow cytometry with the 29E.2A3 clone. The fold-change decrease in PD-L1 MFI was calculated. In A–E, the experiments were performed once. In F, results from three experiments (six donors in total) with technical duplicates are compiled. The statistical significance of the difference between anti-IFN-γ and isotype control treatments was determined for each set of conditions in two-tailed Wilcoxon’s signed-rank tests with FDR adjustment. n.s., not significant. **, P < 0.01.

Figure S3.

Immunophenotyping analysis of PD-L1-deficient leukocytes. Freshly drawn whole-blood leukocytes were analyzed by flow cytometry. (A) Absolute cell numbers were determined with Trucount Absolute Counting Tubes. (B–D) Frequencies of the given leukocyte subsets within each parental subset (indicated at the top of the plots). (E) Gating strategy for circulating T_FH (cT_FH) cells. (F) Representative plots for cT_FH cells. (G) Percentage of cT_FH cells. In A–D and G, bars represent the mean and SEM.

Figure 5.

Immunophenotyping analysis of PD-1- and PD-L1-deficient leukocytes. Freshly thawed PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and adult and age-matched controls were immunophenotyped by flow cytometry. PBMCs from the previously described PD-1-deficient child (aged 11 years) were also analyzed with the same panel. (A) Proportions of leukocyte subsets in PBMCs. (B) CD4⁺ αβ T lymphocyte subsets. (C) CD8⁺ αβ T lymphocyte subsets. (D and E) Proportions of activated T lymphocyte subsets in (D) PBMCs and (E) each parental subset. Bars represent the mean and SEM.

Figure 6.

Analysis of blood αβTCR repertoire in the PD-1- and PD-L1-deficient patients. The complementarity-determining region 3 (CDR3) sequences in the TRAV and TRBV regions were reconstructed with MiXCR from bulk RNASeq datasets for whole-blood leukocytes from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and adult and age-matched controls. For TRBV, the previously published Adaptive ImmunoSeq data for genomic DNA from the whole-blood leukocytes of the PD-1-deficient child (aged 10 years), his healthy brother (aged 6 years), and three healthy controls are also shown for comparison. (A) CDR3 length and physicochemical properties. The median values for each individual are shown. (B and C) CDR3 clonotype diversity. (D and E) Properties of the distribution of CDR3 clonotype sizes. Bars represent the mean and SEM.

Figure 7.

Single-cell transcriptomic analysis of PD-1- and PD-L1-deficient leukocyte subsets. scRNASeq was performed on cryopreserved PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and adult and age-matched controls. Previously generated datasets for healthy and diseased controls, including the PD-1-deficient child and his healthy brother, were also integrated into the analysis (Lee et al., 2023; Ogishi et al., 2021). (A) Clustering. Graph-based clustering was conducted after the removal of batch effects with Harmony (Korsunsky et al., 2019). Clusters were identified with SingleR (Aran et al., 2019) guided by the Monaco datasets (Monaco et al., 2019), followed by manual inspection. (B) Representative gene expression profiles. (C) Pseudobulk DE analysis. Individuals with PD-1 or PD-L1 deficiency were compared with age-matched controls, including the PD-1-deficient patient’s brother and an age-matched control for the PD-L1-deficient siblings. DE genes (DEGs) were defined as genes with FDR-adjusted P values <0.1. cDC1 was omitted because too few cells were captured for the PD-1-deficient patient. The numbers of DEGs per cell type are shown on a bar chart. (D) DEGs common to other monogenic etiologies of autoimmune or autoinflammatory disorders. For each condition, patients with monogenic disease were compared with age-matched controls. The numbers of DEGs (FDR-adjusted P value <0.1) common to (1) PD-1 or PD-L1 deficiency and (2) one of the four known monogenic forms of autoimmunity or autoinflammatory diseases are shown for each cell type. (E) DEGs common to the PD-1- and PD-L1-deficient leukocyte subsets. Here, DEGs are defined as genes with |log₂FC| > 2 relative to age-matched controls. (F) Geneset overrepresentation analysis. DEGs upregulated in the classical or non-classical monocytes of PD-1- and PD-L1-deficient patients relative to age-matched controls were projected onto the Gene Ontology Molecular Function (GO MF) gene sets. GO MF gene sets for which significant enrichment was detected are shown. (G and H) SCENIC regulon activity analysis (Aibar et al., 2017) on (G) Vδ2⁺ γδ T cells and (H) monocytes (classical and non-classical combined). Single-cell regulon activities were aggregated to obtain a mean level of activity per cell type and per individual. The regulons most strongly differentially regulated in individuals with PD-1 and PD-L1 deficiencies relative to age-matched controls, as determined by two-tailed Wilcoxon’s rank sum test, are shown. Bars represent the mean and SEM.

Figure S4.

Single-cell transcriptomic analysis. scRNASeq was performed on cryopreserved PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and adult and age-matched controls. Previously generated datasets for healthy and diseased controls, including the PD-1-deficient child and his healthy brother, were also integrated into the analysis (Lee et al., 2023; Ogishi et al., 2021). Cell subsets were identified by unsupervised clustering followed by automated (i.e., SingleR) and manual annotation. (A) Frequencies of transcriptionally determined leukocyte subsets. (B–D) Pseudobulk DE analysis was performed to compare individuals with PD-1 or PD-L1 deficiency and age-matched controls, including the PD-1-deficient patient’s brother and an age-matched control for the PD-L1-deficient siblings. DE genes (DEGs) were defined as genes with |log₂FC| > 2 relative to age-matched controls. (B) Geneset overrepresentation analysis. DEGs upregulated in Vδ2⁺ γδ T cells from PD-1- and PD-L1-deficient patients relative to age-matched controls were projected onto the gene ontology (GO) gene sets (BP for biological process, MF for molecular function, and CC for cellular component). GO gene sets for which significant enrichment was detected are shown. (C and D) Gene network plots for (C) Vδ2⁺ γδ T cells and (D) monocytes (classical and non-classical combined). DEGs contributing to a given GO term are connected by edges.

Figure 8.

Analysis of the cellular responses of PD-1- and PD-L1-deficient leukocytes in vitro. (A–E) PBMC stimulation assay. PBMCs from the two PD-L1-deficient siblings (aged 11 and 10 years), their mother, and adult and age-matched controls were either left non-stimulated or were stimulated for 24 h. Bulk RNASeq was performed. (A) PCA. (B) GSEA. Genes were ranked based on their fold-change induction (stimulated versus non-stimulated) in PD-L1-deficient cells relative to either healthy controls or the heterozygous mother. Only significant results (FDR-adjusted P value <0.05) from 50 Hallmark gene sets are shown. Gene sets were reordered by hierarchical clustering for visualization purposes. (C) Differential gene induction. Genes related to cytokines or their receptors downregulated in PD-L1-deficient cells relative to control cells are labeled. (D and E) Transcription factor (TF) activity inference analysis based on the CollecTRI gene regulatory network database. (D) PCA. (E) Scaled activity for the TFs known to regulate IFNG mRNA levels lying in the top 30 for the loading of PC1 in D. (F and G) T-blast stimulation assay. (F) T-blasts were stimulated for 4 h in the presence of secretion inhibitors. Cytokine production was quantified by intracellular flow cytometry. Technical duplicates were prepared. (G) T-blasts were stimulated for 4 h, and cytokine secretion was quantified by multiplex ELISA. The age-matched controls include the healthy brother of the PD-1-deficient child. In A–E, the experiments were performed once. In F, representative data from two experiments are shown. In G, data from three experiments with technical replicates for PD-1-deficient cells (14 replicates for N = 1) and PD-L1-deficient cells (duplicates for N = 2) are compiled. In A, D, F, and G, the bars represent the mean and SEM. In F, statistical significance was determined for differences between all healthy controls combined and the PD-1/PD-L1-deficient patients combined by two-tailed Wilcoxon’s rank sum tests with FDR adjustment. *, P < 0.05.

Figure S5.

Analysis of the cellular responses of PD-1- and PD-L1-deficient T lymphocytes in vitro. T-blasts from healthy donors, a PD-1-deficient patient, and the PD-L1-deficient siblings and their heterozygous mother were either left non-stimulated or were stimulated with anti-CD2/CD3/CD28 mAb cocktail or PMA/ionomycin (P/I) for 4 h. (A–C) Intracellular cytokine levels were measured by flow cytometry. (D) Secreted cytokine levels were measured by multiplex ELISA. In B–D, bars represent the mean and SEM. In B, representative data from two experiments are shown. In C, the experiment was performed once. In D, data from three experiments with technical replicates for PD-1-deficient cells (14 replicates for N = 1) and PD-L1-deficient cells (duplicates for N = 2) are compiled.