Differential stoichiometry changes of Mediator complex subunits during erythropoiesis.

(A) Western blot showing the relative abundance of Mediator subunits over time (Days 0-20) in primary human CD34+ erythroid culture. β-actin was used as a loading control, and NCOA4, which is highly expressed in terminal erythropoiesis, served as a positive control. (B) FACS sorting strategy for isolating indicated cell populations and corresponding micrographs by benzidine-Giemsa staining. (C) Western blot showing the relative abundance of Mediator subunits in FACS-sorted cell populations representing different developmental stages (HSPC: CD71-CD235a-; early erythroblasts, EEB: CD71+CD235a+; late erythroblasts, LEB: CD71-CD235a+) in human CD34+ erythroid culture with β-actin as a loading control, and NCOA4 as a positive control. (D) Immunofluorescence imaging of MED1 (left) or MED26 (right) in mouse fetal liver erythroblasts. CD71+Ter119- cells are early erythroblasts (shown by dotted arrows), and CD71+Ter119+ cells are late erythroblasts (shown by white solid arrows). Hoechst 33342 staining was used to indicate the nuclear area.

MED26 regulates all stages of erythropoiesis.

(A) Representative images of mouse embryos of Med26flox/flox;wt/wt and Med26flox/flox;EpoR-Cre/wt at E14.5. Scale bar=5 mm. (B-I) Analyses conducted in Med26flox/flox;Mx1-Cre/wt mice. In these panels, “cKO” refers to Med26flox/flox;Mx1-Cre/wt mice. (B) Western blot demonstrating Med26 knockout efficiency in mouse spleen cells with GAPDH as a loading control. (C) Giemsa staining of peripheral blood smears from control and cKO mice. The arrow indicates reticulocytes. For panels (B and C), tests and phenotypes were analyzed 3 days after administering 10 µg/g pIpC. (D) Timeline of the mice bone marrow transplantation procedure. Bone marrow cells were collected from CD45.2+ control or cKO mice. Subsequently, 10 million cells were injected into 10Gy (5Gy+5Gy) irradiated myeloablative CD45.1+ recipient mice. After a two-week period of hematopoietic reconstitution, the mice received two injections of 100 µg pIpC (totaling 200 µg) to induce Med26 knockout. Peripheral blood samples were collected weekly from weeks 2-4 for CBC analysis and chimera percentage assessment. (E) Survival rate of recipient mice after Med26 knockout. The p value was calculated by Kaplan-Meier log rank test, *, p < 0.05. (C) Giemsa staining of peripheral blood smears from control and cKO mice. Scale bar=10 µm. (F-I) The graphs depict the dynamics of erythroid reconstitution before and after Med26 knockout. Data are represented as the mean ± SD (Control group: n=6; cKO group: n=9). Statistical significance is denoted as n.s., not significant, *, p < 0.05, **, p < 0.01, ***, p < 0.001, ****, p < 0.0001 as determined by Student’s t-test. (J) Western blot demonstrating MED26 overexpression efficiency in the ex vivo human CD34+ erythroid differentiation system on Day 8 with GAPDH as a loading control. (K) Pictures of CD34+ cell pellets upon MED26 overexpression on Day 10 and Day 12. The dotted circles indicate the cell pellet positions. (L) FACS analysis to detect two erythroid markers (CD71 and CD235a) upon MED26 overexpression. Data are represented as the mean ± SD (n=3).

RNA polymerase Ⅱ exhibits transcription pausing at MED26-enriched loci.

(A) Heatmaps from CUT&Tag assays showing the number and distribution of MED1, MED26, GATA1 and GATA2 common peaks; MED1 unique peaks, and MED26 unique peaks on Day 4 in the primary human CD34+-derived erythroid culture. (B) The ratio of MED26 to MED1 signals on all the transcription start sites (TSSs) with MED1 and/or MED26 occupancy in CUT&Tag assays of primary human erythroblasts on Day 4. Each dot represents one gene. (C) Schematic defining the MED26-enriched or MED26-poor Mediator loci. (D) Heatmaps showing RNA Pol Ⅱ and PRO-seq signal of MED26-enriched or MED26-poor genes from −3 kb of the TSS to the +3 kb of transcription end site (TES). (E) IGV visualization showing examples of MED26-enriched (top) and MED26-poor (bottom) loci. RPB1 is the largest component of RNA Pol Ⅱ. (F) Boxplots comparing the pausing indices of MED26-enriched or MED26-poor loci. The pausing indices were calculated from RPB1 CUT&Tag or PRO-seq. The p values were calculated using the two-sided Wilcoxon rank-sum test. (G) Model representing the different transcription behaviors of MED26-enriched or MED26-poor condensates.

Transcription pausing is essential for erythropoiesis.

(A) Heatmap showing the relative interaction strengths of the indicated proteins baited with MED1 or MED26 in HEK293 cells. The relative interaction strength was calculated according to the NSAFs from mass spectrometry data (see Methods). (B) Coimmunoprecipitation of HEK293 cells transfected with 3×Flag-MED1, 3×Flag-MED26, or 3×Flag-EV (control plasmid). Following immunopurification by Flag antibody, western blotting was performed on the complexes using antibodies against components of the pausing complexes (NELF-A, NELF-D, PAF1, and LEO1), the Mediator CDK8 kinase module (CCNC1), and the elongation complex (CDK9 and CCNT1). (C) Heatmaps showing MED26 and PAF1 signal distribution of −3∼+3 kb regions around peak centers in wild type or MED26 KO K562 cells. (D-E) Boxplots comparing the MED26 to MED1 signal ratio and the PI on Day 4 and Day 16 of the human CD34+-derived erythroid culture. Day 4 ex vivo cultured cells primarily consisted of HSPCs and some erythroid progentiors, while day 16 cells predominantly included late erythroblasts. (The p values are calculated by paired Wilcox test. Nonerythroid genes and erythroid genes were defined based on published data50). (F) IGV visualization of the CUT&Tag signals of PRO-seq, MED1, and MED26 at the RPS9 (nonerythroid gene) and HBB (erythroid gene) loci. (G) FACS analysis detecting erythroid differentiation upon treatment with 30 µM DRB (5,6-dichloro-1-beta-D-ribofuranosylbenzimidazole) in Day 6 primary human erythroblasts, performed on Day 10. (Left) Cell death rate after DRB treatment, assessed by PI staining. (Right) Accelerated erythroid differentiation detected by the percentage of CD235a-positive cells. Data are represented as the mean ± SD (n=5). The p values were calculated using an unpaired two-tailed Student’s t test, and significant differences are marked by asterisks: ****p<0.0001, n.s., not significant.

MED26 exhibits a strong condensate-forming capacity and its condensate-forming domain is sufficient to accelerate erythroid differentiation.

(A) FACS analysis to detect erythroid markers (CD71, CD235a) upon full-length or truncated MED26 overexpression in ex vivo human CD34+ erythroid culture. Data are represented as the mean ± SD (n=3). The p values were calculated using an unpaired two-tailed Student’s t test, and significant differences are marked by asterisks: ***p<0.001, n.s., not significant. (B) In vitro phase separation assay of the indicated MED26 truncations fused with EGFP in a low salt buffer. (C) (Top) Schematic diagram of the OptoDroplet assay. The candidate protein was fused with mCherry and CRY2. Upon blue light stimulation, only disordered proteins can form aggregates. (Bottom) Time-lapse images of various Mediator subunits analyzed by the OptoDroplet assay. The subunits are indicated on the left, and subcellular localization is indicated on the right. (D) Western blot showing the relative abundance of BRD4 and RPB1 over time (Days 0-20) in human CD34+ erythroid culture with β-actin as a loading control. (E) Droplet formation assays of gradient-diluted MED26 and MED1 IDR proteins in low-salt buffer without PEG. (F) Model representing transcription condensate component switching during erythroid differentiation.

MED26 accelerates erythropoiesis by recruiting pausing factors through its phase separation domain.

(A) In vitro phase separation assay of EGFP-fused MED26 1-480 and mCherry-PAF1-400-531 with a buffer containing 5% PEG-8000. (B) Representative images of MED26-KO K562 cells transfected with MED26-truncations-EGFP and PAF1-mCherry. The white dotted circle indicates the nuclear area. (C-D) qPCR validation of the relative expression of PAF1and MED26 in indicated CD34+ cells. qPCR data are represented as the mean ± SD (n=3). (E) FACS analysis to detect two erythroid markers (CD71 and CD235a) upon simultaneous PAF1 knockdown and MED26 over-expression. Data are represented as the mean ± SD (n=3). (F) Model representing the transcription features of erythroid and nonerythroid genes during different differentiation stages. The p values were calculated using an unpaired two-tailed Student’s t test, and significant differences are marked by asterisks: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.