Whole-genome CRISPR screening identifies genetic manipulations to reduce immune rejection of stem cell-derived islets

Abstract

Human embryonic stem cells (hESCs) provide opportunities for cell replacement therapy of insulin-dependent diabetes. Therapeutic quantities of human stem cell-derived islets (SC-islets) can be produced by directed differentiation. However, preventing allo-rejection and recurring autoimmunity, without the use of encapsulation or systemic immunosuppressants, remains a challenge. An attractive approach is to transplant SC-islets, genetically modified to reduce the impact of immune rejection. To determine the underlying forces that drive immunogenicity of SC-islets in inflammatory environments, we performed single-cell RNA sequencing (scRNA-seq) and whole-genome CRISPR screen of SC-islets under immune interaction with allogeneic peripheral blood mononuclear cells (PBMCs). Data analysis points to "alarmed" populations of SC-islets that upregulate genes in the interferon (IFN) pathway. The CRISPR screen in vivo confirms that targeting IFNγ-induced mediators has beneficial effects on SC-islet survival under immune attack. Manipulating the IFN response by depleting chemokine ligand 10 (CXCL10) in SC-islet grafts confers improved survival against allo-rejection compared with wild-type grafts in humanized mice. These results offer insights into the nature of immune destruction of SC-islets during allogeneic responses and provide targets for gene editing.

Keywords: CXCL10; T1D; beta cells; chemokine; diabetes; hypo-immunogenicity; immunogenicity; pancreatic islets; transplantation.

Graphical abstract

Figure 1

Single-cell transcriptional profile and whole-genome CRISPR screen of SC-islet grafts in an in vivo humanized model (A) SC-islets or CRISPR library transduced (LT) SC-islets were transplanted in MHC^null NSG mice. Half of each mice cohort was injected with human PBMCs, and human insulin was monitored until graft failure was observed. Grafted cells were then extracted (week 10 post PBMCs) and analyzed by scRNA-seq for gene expression, or by gDNA sequencing for gRNA abundance. (B and C) SC-islet graft failure was assayed in fasted mouse blood by human insulin detection over time, 30 min post glucose. (B) n = 6–8 per group of SC-islet transplanted mice. (C) n = 6 per group of LT SC-islet transplanted mice. (D) Immunofluorescence (IF) staining of kidney SC-islet grafts sections at week 10 after PBMC injection. Bars represent 100 μm in left (×5) and center (×20) and 20 μm in magnified view (right). Kidney (K) and graft (G) margins are outlined. CHGA, chromogranin A. (E and F) scRNA-seq analysis of SC-islet grafts. (E) Volcano plot of differential expressed genes in SC-β and SC-α in hPi versus control grafts. (F) Differential expression of selected genes in different populations, presented as a heatmap. Each row specifies a Z score of the specified gene in all graft samples, in the indicated endocrine population. (G) Analysis of enriched and depleted gene KOs. Rank is plotted against fold changes (hPi versus control) of gRNA counts (×4 integrated per gene) relative to integrated non-targeting (NT) gRNA counts (×941). Significant genes are color coded based on false discovery rate (FDR) as indicated. (H) Boxplot presenting individual gRNAs counts (full model predictions) from mice replicates (n = 6 per condition times n = 4 targeting gRNAs, n = 85 for NT gRNAs, or n = 50 for intergenic gRNAs) with genes of interest with positive and negative enrichment in screen. Box lines represent median values. Dashed line represents mean of NT gRNA counts in control mice. Error bars or shaded areas are mean ± SD; ns, not significant; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001, unpaired two-tailed t test.

Figure 2

Early response of immune-challenged SC-islets profiled by single-cell transcription analysis after co-culture with human allogeneic PBMCs (A) hESC-derived SC-islets were co-cultured with human allogeneic PBMCs (n = 2 donors) for 0, 24, and 48h, followed by scRNA-seq for gene expression. (B) Volcano plot of differential expressed genes in SC-α or SC-β after 24-h co-culture with PBMCs compared with control (t = 0). (C) Pathway analysis and gene set enrichment analysis (GSEA) of upregulated genes in co-cultured SC-β (48 h). (D) Dot plot representing expression of selected inflammatory genes in groups of SC-α and SC-β over time in co-culture with PBMCs. (E) Venn diagrams feature significantly upregulated genes (log2 fold change >1 and adjusted p values <0.05) obtained from in vivo (blue) and in vitro (red) SC-α/SC-β scRNA-seq data (Figures 1 and 2) that are common to CRISPR screen hits (positively enriched in hPi-mice, log2 fold change >1) (green). (F) Violin plots of SC-β timed expression of selected genes. See also Figure S2F. (G) UMAP plots of SC-islet cells expressing CXCL10 or STAT1 over time in co-culture with PBMCs. Specific endocrine cell type clustering is indicated. (H) ELISA for human CXCL10, from supernatant of co-culture of SC-islets and PBMCs. n = 2 donors. Error bars are mean ± SD. Dashed line is the lower detection limit, while any data below it is extrapolated. (I) IF staining of SC-islet clusters ±48-h co-culture with PBMC. C-peptide staining (green) for SC-β and DAPI (blue) for nuclei. Bars represent 100 μm in main panels and 50 μm in magnified panels.

Figure 3

Immunogenicity of CXCL10 expressing SC-islets (A) Transduced SC-islets with lentiviruses carrying Cas9 + gRNA (KO) or a given open reading frame (ORF) insert (overexpression [OE]), were co-cultured with allogeneic PBMCs. (B) Flow cytometry for %TUNEL+ (apoptotic) SC-β cells (C-peptide+), following 48-h PBMC co-culture. Apoptosis was calculated by fraction from baseline (%TUNEL without PBMC). gRNA lentivirus transduced SC-islets were compared with non-targeting (NT) gRNA, and OE transduced SC-islets were compared with eGFP OE. n = 3 for ×5 PBMC donors (left; KO), n = 2–3 for ×2 donors (right; OE). (C) Blocking antibodies prior to/with co-cultures: PBMCs with anti-CXCR3, or SC-islets with anti-TLR4, or anti-CXCL10 during co-culture. (D) Flow cytometry analysis for apoptotic SC-β, following 48-h PBMC co-culture. n = 3 for ×2–6 donors. (E) PBMCs were labeled with cell trace violet (CTV) prior to co-culture. Following a 48-h co-culture, PBMCs were separated and allowed to grow for 7 days. CD3⁺ were gated for the CTV-negative fraction of divided cells. n = 5 for ×3 donors. Error bars are mean ± SD. ns, not significant; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001, unpaired two-tailed t test.

Figure 4

Generation and performance of CXCL10 KO and STAT1 KO hESC lines (A and B) Scheme of targeting the (A) CXCL10 or (B) STAT1 locus in hESCs using CRISPR. Red and blue arrows are PCR primers for genotyping as shown in Figure S4A. (C) Flow cytometry of intracellular CXCL10 protein in WT/C10G SC-islets and SC-β (C-peptide+) ± rhIFNγ for 48 h n = 3–5. (D) CXCL10 ELISA of supernatants from ±rhIFNγ-treated WT/C10G/ST1L SC-islets. Dashed line is the lower detection limit, while any data below it is extrapolated. (E) Flow cytometry for protein expression in rhIFNγ-treated GAPDH-luciferase (GL) or ST1L SC-islets. n = 3–4. (F–J) Gene-modified (GM; C10G/ST1L) and control (WT/GL) lines were differentiated into SC-islets, and co-cultured with human PBMCs or purified T cells/NK cells. Apoptosis was calculated by fraction from baseline (%TUNEL without PBMCs). (G) Apoptotic WT or C10G SC-β cells (n = 4 for ×6 PBMC donors, n = 2–3 ×2 T cell donors, n = 4 × 4 NK cell donors). (H) Apoptotic GL or ST1L SC-β cells (n = 4 for ×2 PBMC or NK cell). (I and J) Proliferated CD3 T cell following co-culture with indicated GM SC-islets (I) n = 9 for ×5 donors and (J) n = 9 for ×2 donors). Error bars are mean ± SD. ns, not significant; ^∗p < 0.05; ^∗∗p < 0.01, unpaired two-tailed t-test.

Figure 5

CXCL10 KO SC-islet grafts evade alloimmune attack in humanized mice (A) WT or C10G SC-islets were transplanted into MHC^null NSG mice (n = 10 from each line). n = 6–7 mice from each group injected with human PBMCs (n = 2 human donors), while the remainder served as control (n = 3 per group). (B) Graft failure at week 11 after PBMC injections, as measured by human insulin in fasted mice plasma, 30 min after glucose injection to fasted mice. Data presented as fold increase from t = 0 before PBMC injections. (C) Flow cytometry of SC-α (glucagon+/C-peptide−) and SC-β (glucagon−/C-peptide+) in extracted grafts at week 18 post PBMC injection. n = 3–4 mice per group. (D) Flow cytometry of human T cells in hPi-mouse graft infiltrating at week 18 post PBMC injection. n = 3–5 mice per group. Error bars are mean ± SD. ns, not significant; ^∗p < 0.05; ^∗∗p < 0.01, unpaired two-tailed t test.