A Genome-wide CRISPR Screen Reveals a Role for the Non-canonical Nucleosome-Remodeling BAF Complex in Foxp3 Expression and Regulatory T Cell Function

Abstract

Regulatory T (Treg) cells play a pivotal role in suppressing auto-reactive T cells and maintaining immune homeostasis. Treg cell development and function are dependent on the transcription factor Foxp3. Here, we performed a genome-wide CRISPR loss-of-function screen to identify Foxp3 regulators in mouse primary Treg cells. Foxp3 regulators were enriched in genes encoding subunits of the SWI/SNF nucleosome-remodeling and SAGA chromatin-modifying complexes. Among the three SWI/SNF-related complexes, the Brd9-containing non-canonical (nc) BAF complex promoted Foxp3 expression, whereas the PBAF complex was repressive. Chemical-induced degradation of Brd9 led to reduced Foxp3 expression and reduced Treg cell function in vitro. Brd9 ablation compromised Treg cell function in inflammatory disease and tumor immunity in vivo. Furthermore, Brd9 promoted Foxp3 binding and expression of a subset of Foxp3 target genes. Our findings provide an unbiased analysis of the genetic networks regulating Foxp3 and reveal ncBAF as a target for therapeutic manipulation of Treg cell function.

Keywords: Brd9; CRISPR; Foxp3; GBAF; SWI/SNF; Treg; ncBAF; screen.

Graphical abstract

Figure 1

A Genome-wide CRISPR Screen in Treg Cells (A) Workflow of the CRISPR screen in Treg cells. (B–D) Validation of the CRISPR screen conditions. (B) Fluorescence-activated cell sorting (FACS) plots showing Foxp3 expression in Treg cells after sgRNA targeting of Foxp3 (sgFoxp3), the positive regulator Cbfb (sgCbfb), and the negative regulator Dnmt1 (sgDnmt1). Red and green gates were set based on Foxp3 low 20% and high 20% in sgNT Treg cells, respectively. (C and D) Mean fluorescence intensity (MFI) of Foxp3 (C) and relative log2 fold change of the cell count (D), comparing Foxp3^lo with Foxp3^hi after deletion of the indicated target gene (n = 3 per group; data represent mean ± SD). See also Figures S1 and S2.

Figure 2

Identification of Foxp3 Regulators in Treg Cells (A and B) A scatterplot of the Treg cell screen results, showing positive regulators (A) and negative regulators (B). Genes that met the cutoff criteria (p < 0.01 and LFC > ±0.5) are shown as red dots for positive regulators and as green dots for negative regulators. (C) Distribution of sgRNA LFC comparing Foxp3^lo with Foxp3^hi. Red stripes represent sgRNAs from positive Foxp3 regulators, whereas green stripes represent sgRNAs from negative Foxp3 regulators. (D) Venn diagram showing the overlap of Foxp3 regulators with genes involved in cell contraction or expansion. (E and F) Gene Ontology analysis of positive Foxp3 regulators (E) and negative Foxp3 regulators (F). See also Figures S3 and S4 and Tables S1, S2, S3, and S4.

Figure 3

The Three SWI/SNF Complex Assemblies Have Distinct Regulatory Roles for Foxp3 Expression in Treg Cells (A) A diagram showing three different variants of SWI/SNF complexes: BAF, ncBAF, and PBAF. BAF-specific subunits (Arid1a and Dpf1–Dpf3) are colored blue, ncBAF-specific subunits (Brd9, Smarcd1, Gltscr1l, and Gltscr1) are colored orange, and PBAF-specific subunits (Pbrm1, Arid2, Brd7, and Phf10) are colored green. Shared components among complexes are colored gray. Also shown is an immunoprecipitation assay of Arid1a, Brd9, Phf10, and Smarca4 in Treg cells. The co-precipitated proteins were probed for shared subunits (Smarca4, Smarcc1, and Smarcb1), BAF-specific Arid1a, ncBAF-specific Brd9, and PBAF-specific Pbrm1. (B) FACS histogram of Foxp3 expression in Treg cells after sgRNA targeting of the indicated SWI/SNF subunits. (C) MFI of Foxp3 after sgRNA targeting of the indicated SWI/SNF subunits. Data represent mean and standard deviation of biological replicates (n = 3–21). (D) Principal-component analysis of RNA-seq data collected from Treg cells transduced with guides against the indicated SWI/SNF subunits. In cases where two independent guides were used to target a gene, the second guide for targeting the gene is indicated as “-2.” (E) MFI of Foxp3 expression in Treg cells after treatment with DMSO or 0.16–10 μM dBRD9 for 4 days. Data represent mean ± SD. Statistical analyses were performed using unpaired two-tailed Student’s t test (non-significant [ns], p ≥ 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001). See also Figure S5.

Figure 4

Brd9 Deletion Reduces Foxp3 Binding at CNS0 and CNS2 Enhancers and a Subset of Foxp3 Target Sites (A) Genome browser tracks of Smarca4, Brd9, and Phf10 ChIP-seq and ATAC-seq signals as well as Foxp3 ChIP-seq in sgNT, sgFoxp3, sgBrd9, and sgPbrm1 Treg cells and Foxp3 in DMSO- and dBRD9-treated Treg cells (2.5 μM dBRD9 for 4 days). The Foxp3 locus is shown with CNS0 and CNS2 enhancers, indicated as gray ovals. (B) Heatmap of Foxp3, Brd9, Smarca4, and Phf10 ChIP-seq and ATAC-seq signals ± 3 kb, centered on Foxp3-bound sites in Treg, ranked according to Foxp3 read density. (C) Venn diagram of the overlap between ChIP-seq peaks in Treg cells for Brd9, Foxp3, and Phf10 (hypergeometric p of Brd9:Foxp3 overlap = e⁻²⁷⁶⁶⁵, hypergeometric p of PHF10:Foxp3 overlap = e⁻¹⁷¹⁸⁵, hypergeometric p of Brd9:PHF10 overlap = e⁻¹⁴²¹⁷). (D) Heatmap of Foxp3 ChIP-seq signals in sgNT, sgFoxp3, sgBrd9, and sgPbrm1 Treg cells ± 3 kb, centered on Foxp3-bound sites in sgNT, ranked according to read density. (E) Venn diagram of the overlap (hypergeometric p = e^−11,653) between sites that significantly lose Foxp3 binding (FC 1.5, Poisson p < 0.0001) in sgFoxp3 and sgBrd9, overlaid on all Foxp3-bound sites in sgNT (gray). (F) Histogram of Foxp3 ChIP read density ± 1 kb surrounding the peak center of sites that significantly lose Foxp3 binding in sgFoxp3 and sgBrd9 (n = 1,699) in sgNT, sgFoxp3, sgBrd9, and sgPbrm1. (G) As in (E) but for sites that lose H3K27ac (FC 1.5, Poisson p < 0.0001, hypergeometric p of overlap = e^−7,938). (H) As in (F) but for H3K27ac ChIP read density. (I) As in (D) but for Foxp3 ChIP-seq signals in DMSO- and dBRD9-treated Treg cells at all Foxp3-bound sites in DMSO. (J) As in (E), but for sites that significantly lose Foxp3 binding in dBRD9 treated Treg cells versus DMSO (FC 1.5, Poisson p < 0.0001). (K) As in (F) but for DMSO- and dBRD9-treated cells. (L) As in (F) but for Treg cells transduced with sgNT or sgBrd9, with ectopic expression of the MIGR vector control or Foxp3. See also Figure S6.

Figure 5

Brd9 Co-regulates the Expression of Foxp3 and a Subset of Foxp3 Target Genes (A) Volcano plot of log2 fold change RNA expression in sgFoxp3 versus sgNT Treg cells versus adjusted p value (Benjamini-Hochberg). The numbers of down and up genes are indicated and colored blue and red, respectively. (B) Significance of enrichment of Foxp3-dependent genes in each Gene Ontology. (C) Pie chart of Foxp3 and Brd9 binding by ChIP-seq for Foxp3-dependent genes. (D) Gene set enrichment analysis (GSEA) enrichment plot for up and down genes in sgBrd9 versus sgNT compared with RNA-seq data of genes that significantly change in sgFoxp3 versus sgNT Treg cells. ES, enrichment score; NES, normalized enrichment score; FWER, family-wise error rate. (E) As in (D) but for up and down genes in dBRD9 versus DMSO Treg cells. (F) GSEA of the sgFoxp3 versus sgNT RNA-seq data; the plot shows the FWER p value versus the NES. See also Figure S6 and Table S5.

Figure 6

The ncBAF Complex Regulates Treg Cell Suppressor Function In Vitro and In Vivo (A) In vitro suppression assay of Treg cells with sgRNA targeting of Brd9, Smarcd1, Pbrm1, and Phf10. sgNT was used as a non-targeting control (n = 3 per group; data represent ± SD). (B) In vitro suppression assay of sgBrd9 or sgNT with ectopic expression of Foxp3 or the control vector MIGR (n = 3 per group; data represent ± SD). (C–F) Experiment to measure function of sgNT or sgBrd9 Treg cells relative to “no Treg” cells in a T cell transfer-induced colitis model. (C) Experimental procedure. (D) Body weight loss. (E) Colon histology (left) and colitis scores (right). (F) Percentage of Foxp3⁺ cells in the transferred CD45.2⁺CD4⁺ Treg population at the endpoint (n = 4–7 per group; data represent mean ± SEM). Statistical analyses were performed using unpaired two-tailed Student’s t test (ns, p ≥ 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001). See also Figure S7.

Figure 7

Targeting Brd9 in Treg Cells Improves Anti-tumor Immunity (A) Experimental procedure to measure the function of sgNT or sgBrd9 Treg cells relative to “no Treg” cells in the MC38 tumor model. (B) Tumor growth curve. (C) Tumor weight at the endpoint. (D and E) Bar graph of total CD4 T cell (D) and CD8 T cell (E) percentage in the CD45⁺ immune cell population. (F and G) Bar graph of the IFN-γ⁺ cell percentage in CD4 T cells (F) and CD8 T cells (G). (H) Bar graph of the CD4⁺GFP⁺Foxp3⁺ donor cell percentage in CD4 T cells. (I) Ratio of CD8:Treg cells. (J) Bar graph of the Foxp3^– ex-Treg cell percentage in the transferred Treg population marked by the GFP reporter. (K) Bar graph of Foxp3^–IFN-γ⁺ cell percentage in the transferred Treg population (n = 5–7 per group; data represent mean ± SEM). Statistical analyses were performed using unpaired two-tailed Student’s t test (ns, p ≥ 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001). See also Figure S7.