Figure 1

Cohesin Mutants Impair Hematopoietic Differentiation

(A) Schematic diagram of the cohesin complex with four components (RAD21, SMC1A, SMC3, and STAG2). Mutations found in AML (in the Cancer Genome Atlas Research Network or our own tissue bank) in each cohesin component are indicated by triangles. Specific mutations used for studies reported here are indicated. (^∗) indicates nonsense mutations.

(B) TF-1 cells were infected with lentiviruses encoding doxycycline (DOX)-inducible cohesin WT or mutant variants and GFP. Erythroid differentiation of parental TF-1 cells and variants was determined by flow cytometry for GPA expression after 2 initial days of DOX treatment and 8 days of EPO and DOX treatment. Relative expression is shown as mean fluorescence intensity (MFI) of GPA normalized to IgG isotype control. ^∗∗p < 0.01.

(C) Expression of fetal hemoglobin was determined by qRT-PCR for cells treated as in (B). Values are normalized to No DOX controls. ^∗∗p < 0.01.

(D) Expression of KLF-1 was determined by qRT-PCR for cells treated as in (B). Values are normalized to No DOX controls. ^∗∗p < 0.01.

(E) TF-1 cells were treated as in (B) and then washed out of DOX and maintained in EPO media for an additional 4 days. Erythroid differentiation of parental TF-1 cells and variants was determined by flow cytometry for GPA expression. Relative expression is shown as MFI of GPA normalized to IgG isotype control. No statistically significant differences were detected.

(F) Expression of fetal hemoglobin was determined by qRT-PCR for cells treated as in (E). Values are normalized to TF-1 control. No statistically significant differences were detected.

(G) Expression of KLF-1 was determined by qRT-PCR for cells treated as in (E). Values are normalized to TF-1 control. No statistically significant differences were detected.

Figure 2

Cohesin Mutants Impair Myeloid, Erythroid, and Stem Cell Differentiation of Primary Human HSPCs

(A) Schematic of in vitro differentiation experiments using human CD34-enriched HSPCs from cord blood.

(B) HSPCs were infected with lentiviruses encoding GFP alone (control) or GFP in addition to the indicated cohesin variants. 72 hr post-infection, GFP+ cells were isolated by FACS and cultured in HSPC-retention medium. 6 days later, cells were analyzed for expression of progenitor markers CD34 and CD38. Representative FACS plots are shown.

(C) Summary of data from three independent experiments as described in (B); the percentage of CD34+ cells was normalized to GFP control. Unpaired Student’s t test was used to determine statistical significance between WT and mutant populations. ^∗∗p < 0.01.

(D) HSPCs isolated and lentivirally transduced as in (B) were cultured in myeloid differentiation medium. 6 days later, cells were analyzed for expression of myeloid markers CD33 and CD14. Representative FACS plots are shown.

(E) Summary of data from three independent experiments as described in (D); the percentage of CD33+/CD14+ cells was normalized to GFP control. Unpaired Student’s t test was used to determine statistical significance between WT and mutant populations. ^∗∗p < 0.01.

(F) HSPCs isolated and lentivirally transduced as in (B) were cultured in erythroid differentiation medium. 6 days later, cells were analyzed for expression of erythroid markers CD71 and GPA. Representative FACS plots are shown.

(G) Summary of data from three independent experiments as described in (F); the percentage of GPA+/CD71+ cells was normalized to GFP control. Unpaired Student’s t test was used to determine statistical significance between WT and mutant populations. ^∗∗p < 0.01.

Figure 3

Knockdown of RAD21 Impairs Hematopoietic Differentiation and Induces Myeloid Skewing In Vivo

(A) Human CD34-enriched cord blood HSPCs were infected with constitutive RAD21 shRNA or scramble control RFP-encoding lentiviral vectors. 72 hr post-infection ∼50,000 cells were transplanted by intrafemoral injection into NSG mice (three mice per condition). 7 weeks post-transplant, bone marrow aspirates were analyzed for human engraftment (human CD45+) and RFP+ expression. No statistically significant differences were detected.

(B) Human RFP+ engrafted cells from (A) were analyzed for expression of progenitor markers CD34 and CD38. Representative FACS plots are shown.

(C) Summary of data from 3 independent experiments as described in (B); the percentage of CD34+ cells was normalized to non-transduced control. Unpaired Student’s t test was used to determine statistical significance between scramble shRNA and RAD21 shRNA populations. ^∗∗p < 0.01.

(D) Human RFP+ engrafted cells from (A) were analyzed for expression of B lymphoid (CD19 & CD20) and myeloid markers (CD33). Representative FACS plots are shown.

(E) Summary of data from three independent experiments as described in (D). The percentage of CD33+ cells is shown. Unpaired Student’s t test was used to determine statistical significance between non-transduced and scramble shRNA versus RAD21 shRNA populations. ^∗∗p < 0.01.

Figure 4

Cohesin Mutants Induce Myeloid Skewing, Increase Serial Replating, and Enforce Stem Cell Gene Expression Programs in Human HSPCs

(A) Human CD34-enriched HSPCs were infected with lentiviruses encoding GFP alone (control) or GFP in addition to the indicated cohesin variants. 72 hr post-infection, GFP+ cells were isolated by FACS and cultured in methylcellulose for colony-forming assays. Every 14 days, colonies were scored for morphology and cells (1,000–10,000) were replated up to five cycles. The number of colonies per 1,000 cells in the first plating is indicated on the left; the right indicates the number of colonies per 1,000 cells in the subsequent platings. nd, none detected. Unpaired Student’s t test was used to determine statistical significance between WT and mutant populations. ^∗p < 0.05 and ^∗∗p < 0.01.

(B) The morphological colony-types from the primary plating of cells as described in (A) are indicated. Statistically significant differences in the percent of CFU-GM were detected. ^∗∗p < 0.01.

(C) Cells from the methylcellulose colonies from (B) were isolated and analyzed for expression of myeloid marker (CD33) or erythroid marker (GPA). Representative FACS plots are shown. Data are representative of three independent experiments.

(D) Human CD34-enriched HSPCs were infected with control, cohesin WT, or cohesin mutant GFP-encoding lentiviruses. 72 hr later, GFP+ transduced cells were FACS purified and subject to gene expression analysis on Affymetrix microarrays. The resulting expression data was analyzed by unsupervised hierarchical clustering and the Pearson correlation dendogram is shown here.

(E) Gene expression microarray data as described in (D) was utilized in Gene Set Enrichment Analysis (GSEA) to identify gene sets enriched in cohesin mutant compared to cohesin WT and control samples in both the molecular signatures database (MSigDB) and manually curated leukemia stem cell gene sets.

(F) GSEA identified gene sets enriched in WT and control samples compared to cohesin mutants in the MSigDB.

Figure 5

Cohesin Mutants Impair Human HSPC Differentiation in a Cell-Context-Dependent Manner

(A) Human CD34-enriched cord blood was sorted for HSPC subpopulations (HSCs, MPPs, LMPPs, CMPs, MEPs, and GMPs) (See Figure S5 for sorting strategy). Each subpopulation was then infected with lentiviruses encoding GFP alone (control) or GFP in addition to the indicated cohesin variants. The virally transduced cells were cultured in myeloid differentiation medium, as in Figure 2 . 8 days later, the cells were analyzed by flow cytometry for expression of myeloid markers CD33 and CD14. Representative FACS plots of GFP+ cells are shown.

(B) Summary of data from three independent experiments as described in (A); the percentage of CD14+ cells was normalized to GFP control. Scramble shRNA and RAD21 shRNA transduced HSPC subpopulations ( Figure S5 B) are included in the summary. Unpaired Student’s t test was used to determine statistical significance between WT and mutant populations. ^∗p < 0.05 and ^∗∗p < 0.01.

(C) HSPC subpopulations isolated and lentivirally transduced as in (A) were cultured in erythroid differentiation medium, as in Figure 2 . 8 days later, the cells were analyzed for expression of erythroid markers CD71 and GPA. Representative FACS plots of GFP+ cells are shown.

(D) Summary of data from three independent experiments as described in (C); the percentage of CD71+ cells was normalized to GFP control. Scramble shRNA and RAD21 shRNA transduced HSPC subpopulations ( Figure S5 C) are included in the summary. Unpaired Student’s t test was used to determine statistical significance between cohesin WT and mutant populations. ^∗∗p < 0.01.

Figure 6

Cohesin Mutants Exhibit Altered Chromatin Accessibility at Transcriptional Regulatory Elements, but Increased Accessibility and Binding at HSPC TF Motifs

(A) Human CD34-enriched HSPCs were infected with lentiviruses encoding cohesin WT or mutants in addition to GFP. 72 hr later, GFP+ cells were isolated by FACS and subject to ATAC-seq. Average diagram of genome-wide chromatin accessibility at TSS regions (2 kb window) comparing cohesin WT (RAD21 WT and SMC1A WT) and cohesin mutants (RAD21 mutant and SMC1A mutant) (left panel) is shown. Heat map of ATAC-seq signal intensity at all TSS regions (2 kb window) in cohesin WT and mutants (right panel) is also shown.

(B) V-plot analysis of ATAC-seq fragments near TSSs (1 kb window) in cohesin WT (RAD21 WT and SMC1A WT) and cohesin mutants (RAD21 mutant and SMC1A mutant). The x axis represents the distance between the centers of the fragments to the TSSs. The y axis represents the fragment length. The color (scaled from 0 to 9) represents the intensity of the ATAC-seq signal at the coordinate of this xy plane.

(C) Enrichment of TF motifs in the peaks that gain open accessibility in RAD21 mutant-expressing TF1 cells (left panel), RAD21 mutant-expressing CD34+ cord blood cells (middle panel), and SMC1A mutant-expressing CD34+ cord blood cells (right panel). The y axis is –log₁₀(p value) of a motif enrichment test, which is sorted from largest to smallest. The x axis is the sorted motif rank. The ETS, GATA, and Runt families are indicated by green, red, and blue, respectively.

(D) Illustrative UCSC genome browser track of normalized ATAC-seq signal at the CD34 locus. K562 ENCODE ChIP-seq data for GATA2 and RAD21 are overlayed, as well as K562 ChromHMM indicating a higher signal at an active promoter and strong enhancer of CD34.

(E) PIQ footprinting analysis for ERG, GATA2, and RUNX1 indicates a higher likelihood of TF occupancy in RAD21 mutant HSPCs compared to RAD21 WT. ^∗∗∗∗p < 0.0001.

(F) TF-1 RAD21 WT and RAD21 Q592^∗ were induced with DOX for 6 days and then subjected to ChIP-seq for GATA2 and RUNX1. Average diagrams for ChIP-seq signal (values normalized to input control) at GATA2 and RUNX1 motif are shown here. The top row indicates all genome-wide GATA2 motifs, and the bottom row indicates all GATA2 motifs in ATAC-seq peaks (from above).

Figure 7

Cohesin-Mutant-Induced Stem Cell Programs Are Dependent on ERG, GATA2, and RUNX1

(A) TF expression was determined by qRT-PCR to determine the knockdown efficiency of two independent shRNAs for each of the indicated TFs in FACS-purified RFP+ CD34-enriched cord blood. The shRNA exhibiting the strongest knockdown for each TF was used for subsequent experiments.

(B) CD34-enriched cord blood HSPCs were double transduced with lentiviruses encoding scramble or TF-targeting shRNAs (RFP+, from A) and cohesin WT or mutants (GFP+). 6 days later, cells were analyzed for RFP and GFP expression by flow cytometry and representative plots are presented.

(C) HSPCs isolated and lentivirally transduced as in (B) were cultured in HSPC retention medium as in Figure 2 . 6 days later, cells were analyzed for expression of progenitor markers CD34 and CD38. Representative FACS plots are shown. Key comparisons highlighting ERG, GATA2, and RUNX1 knockdown in the setting of mutant RAD21 or SMC1A are indicated in red boxes.

(D) Summary of data from three independent experiments as described in (C); the percentage of CD34+ cells was normalized to scramble shRNA only control. Unpaired Student’s t test was used to determine statistical significance between cohesin WT and mutant populations. ^∗∗p < 0.01.

(E) Schematic model: mutant cohesin impairs hematopoietic differentiation and enforces stem cell programs through the modulation of ERG, GATA2, and RUNX1 chromatin accessibility, expression, and activity.