Research ArticleCELL BIOLOGY

Temporal expression of MOF acetyltransferase primes transcription factor networks for erythroid fate

See allHide authors and affiliations

Science Advances  20 May 2020:
Vol. 6, no. 21, eaaz4815
DOI: 10.1126/sciadv.aaz4815
  • Fig. 1 Mof exhibits dynamic expression along erythropoiesis, and its reduction has a pronounced impact on erythroid lineage commitment.

    (A) Heatmap representing gene expression patterns of KATs and HDACs along the erythroid trajectory [data from (18)]. (B) Top: Schematic representation of fate-bias analysis, showing the cell fate-bias score (x axis) versus transcript expression (y axis). The fate-bias score indicates the probability of a given cell to belong to target cell types A, B, or C. A score of 1.0 indicates that the given cell has 100% probability of belonging to a target cell type. The example cell expressing Gene X has a 75 to 100% chance of belonging to cell type A but only a 0 to 20% chance of belonging to cell types B or C. In the analysis represented in the bottom panel, we selected erythroid, myeloid, and lymphoid cells as our target cell types. Bottom: Dot plot from 4763 cKit+ cells, showing the cell fate-bias score determined by FateID plotted against RaceID-normalized transcript expression in erythroid [Gata1+Klf1+Gypa+Hba-a1+; cluster 3; erythroid versus Mof expression r = +2 × 10−2; Pearson correlation, P = 4 × 10−2), myeloid (Mpo+Elane+; cluster 1; myeloid versus Mof expression r = −3.6 × 10−2; Pearson correlation, P = 6.3 × 10−3), and lymphoid (Ebf1+Rag1+; cluster 7; lymphoid versus Mof expression r = +6 × 10−3; Pearson correlation, P = 0.32) lineages; original data from (12)]. For t-distributed stochastic neighbor embedding (t-SNE), maps, and cluster annotation see fig. S1 (A and B). (C) Box plot showing the total number of RBC, hemoglobin (HB) quantity, and hematocrit (HCT) percentage (n = 4 per genotype) and scatter plot showing the HB concentration measured using an enzyme-linked immunosorbent assay (ELISA)–based assay (wild type, n = 7; Mof+/−, n = 13; P < 0.001). (D) Left: Box plot showing the total area of the colonies obtained from the single-cell colony-forming unit (sc-CFU) assay on fluorescence-activated cell sorting (FACS)–sorted HSCs (LSK+Flt3CD34CD48CD150+). Right: Box plot showing the total cell number per colony obtained in the sc-CFU assay. (E) Pie charts representing the lineage potency from the sc-CFU assay. (F) Stacked bar plot showing the fraction of lineage output from each clone. Final populations were defined on the basis of cell surface markers: myeloid (Mye; cKitCD11b+, number of cells: 208,961 wild type and 905,455 Mof+/−), erythroid (Ery; cKitTer119+, number of cells: 205,963 wild-type and 27,564 Mof+/−), lymphoid [Lym; cKitIgD+, number of cells: 65,093 wild-type and 63,815 Mof+/−], and progenitor cells (Prog; cKit+, number of cells: 98,988 wild-type and 124,690 Mof+/−). Significant enrichment by Fisher’s exact test was set as P > 0.05. sc-CFU data from 72 wild-type and 147 Mof+/− colonies from three independent animals. (G) Bar plots showing the CFU capacity output of 100 FACS-sorted HSCs (wild-type, n = 6; Mof+/−, n = 3). (H) Same as (G), but using HSCs sorted from Cag-Cre:ERT2Tg/+(control) or Moffl/flCag-Cre:ERT2Tg/+transgene (Mof-iKO) animals. After sorting, cells were cultured with 4-hydroxytamoxifen (4-OHT) to induce Mof depletion in vitro. Bar plots show CFU capacity output of HSCs (control, n = 3; Mof-iKO, n = 3). (I) Serial CFU assay scheme and serial colony formation capacity from (G) or (H). (J) CFU assay scheme and CFU capacity output of FACS-sorted MEPs (LincKithighSca-1IL-7RCD34FcRgII/III) from wild-type or Mof+/− (wild-type, n = 6; Mof+/−, n = 3) animals. (K) same as (J), but from control or Mof-iKO–sorted MEPs, wherein Mof depletion was induced in vitro after sorting and plating. Error bars represent means ± SEM, and biological replicates are represented as the overlaid dots. Experimental significance was determined by one-way analysis of variance (ANOVA), P < 0.05. Related to figs. S1 to S4.

  • Fig. 2 Mof+/− erythroid progenitor cells show altered erythroid identity and low probability to belong to the erythroid trajectory.

    (A) t-SNE representation of transcriptome similarities between each cell. After normalization and filtering, 34,111 genes and 1842 cells were analyzed. Sorting strategy, scRNA-seq quality controls, and cluster characterization are shown in fig. S5 (A to D). t-SNE map shows the population annotation based on differentially expressed genes, transcriptome entropy, and key TFs generated by the RaceID3/StemID2 algorithm. (B) Expression of Mof and representative marker genes is highlighted on the t-SNE map from (A). Multipotent stem cells (Cd34 and Gata2), myeloid (Elane and Mpo), erythroid (Gata1 and Car2), B cells (Ebf1 and Ighm), dendritic cells (DCs; Itgax), neutrophils (Ly6g), and macrophage (Lpl). The scale bars show the normalized expression for each gene [for further markers, see figs. S5 (E and F) and S6 (E to K)]. (C) Heatmap showing the log-normalized expression of six key population markers each for “stem cells,” “erythroid,” “myeloid,” and “lymphoid” across all clusters with n > 5 cells. Expression was scaled by genes. (D) Bar plot showing that the top four differentially expressed genes in clusters 3 and 6 are related to erythropoiesis. scRNA-seq data were generated from n = 7 animals in three independent experiments (see Materials and Methods for details). FC, fold change. (E) Box plots showing the erythroblast fate-bias probability of wild-type (gray) and Mof+/− (orange) erythroblast cells (cluster 3) and dormant HSCs (dHSCs; cluster 2). Statistical significance was determined by t test, P < 0.05. (F) Left: Representative FACS dot plot of resident BM erythroblasts. Erythroid populations are shown in green. Right: Bar plot showing total number or total frequency of erythroid progenitor cells in the BM. Data from n = 11 animals per genotype. After normal distribution evaluation, the P values were calculated by unpaired t test or two-tailed Mann-Whitney test. (G) Left: t-SNE projection depicting cells genotype, wild-type (black triangles) and Mof+/−(orange circles) cells. The gray oval highlights cluster 10. Right: Bar plots showing clusters that are significantly enriched in one genotype. P value was determined by Fisher’s exact test. Related to figs. S5 and S6.

  • Fig. 3 Mof expression is associated with global chromatin accessibility and H4K16ac levels.

    (A) Representative flow cytometry histograms showing ATAC-see staining intensities. The dark-gray shading shows the ATAC-see in wild-type animals, and orange shows the ATAC-see for Mof+/−animals. The black line marks ATAC-see oligo autofluorescence, and light gray shading shows the fluorescence minus one (FMO) control in which the ATAC-see oligo was omitted. (B) Summary box plot showing the quantification of the ATAC-see FACS median fluorescence intensities (MFIs) of HSCs (LSK+CD34CD150+), MPPs (LSK+CD34+), CMPs (LinKithighSca1CD34+CD16/32dim), GMPs (LinKithighSca1CD34+CD16/32high), and MEPs (LinKithighSca1CD34CD16/32) from wild-type (gray bars) and Mof+/− (orange bars) mice. Data represent n = 7 independent animals. P values were calculated by two-tailed Mann-Whitney test, with confidence level set as 95%. (C) Representative images of ATAC-see (green), MOF immunofluorescence (red), and 4′,6-diamidino-2-phenylindole (DAPI; blue) in sorted MEPs from n = 3 animals and two independent experiments. Scale bars, 2 μm (see also fig. S8, A and B). (D) Density plot showing the mean number of ATAC-seq peaks found in human HSCs (dash line), MPPs (blue line), MEPs (red line), and erythroblasts (gray line). Data from GSE74912. (E) Box plots showing FACS MFI signal of H4K16ac in HSC (LSK+CD34Flt3CD48CD150+), MPP1 (LSK+Flt3CD34+CD150+CD48), MPP2 (LSK+Flt3CD34+CD150+CD48+), MPP3 (LSK+CD34+Flt3CD150CD48+), CMP, and MEP populations [as in (B)] normalized by their total H4 levels in wild-type and Mof+/− mice. Number of animals n = 4 from two independent experiments. For representative histogram, see fig. S8C. Mean value is shown as a line. P value was determined by two-tailed Mann-Whitney test, with confidence level set as 95%. A.U, arbitrary units. (F) Ex vivo erythropoiesis assay. Bar plot showing reverse transcription quantitative polymerase chain reaction (RT-qPCR) analyses of Mof levels in HSCs isolated from wild-type (gray) or Mof+/−(orange) animals and cultured in methylcellulose-based medium for the indicated number of days. Mof expression was determined by [CT-Mof − CT-Hprt, dCT] and then normalized by the smallest dCT (normalization was conducted considering each genotype and day of interest), ddCT. The fold change is 2ddCT. Each overlaid data point represents the number (n) of independent animals. Error bars represent means ± SEM. Significance was determined after normal distribution analysis by two-tailed Mann-Whitney test. (G) Scheme of CFU experiment. Wild-type, Cag-Cre:ERT2Tg/+(control), or Moffl/flCag-Cre:ERT2Tg/+transgene (Mof-iKO) HSCs were sorted and cultured with 4-OHT in methylcellulose medium for 10 days. After 10 days in culture, ATAC-see signal was evaluated. (H) Left: Representative immunofluorescence image of MOF (red) and ATAC-see (green) from wild-type, Cag-Cre:-ERT2Tg/+, or Moffl/fl Cag-Cre:ERT2Tg/+ (Mof-iKO) mice. Right: MFI of ATAC-see, MOF, and DAPI signal (number of animals, n = 4). After normality test, two-tailed Mann-Whitney test was applied for statistical significance. Related to fig. S7.

  • Fig. 4 Identification of cell type-specific MOF targets and its association with chromatin status.

    (A) Heatmap showing the log2(fold change) MOF ChIP-seq versus input enrichment on combined MACS2 peaks in HSCs and MEPs. Three unsupervised k-means clusters were generated, and the peak center was used as reference point ±5 kb (see Materials and Methods). The regions were ordered according to signal intensity and kept the same in HSCs and MEPs. Cluster 1 (#1) contains 13,150, cluster 2 (#2) 16,942, and cluster 3 (#3) 17,782 targets. (B) Box plot displaying the MOF ChIP enrichment intensities on all MACS2 peaks per k-means cluster and on random regions. Enrichment scores were calculated using deepTools2 multiBigwigSummary. For ChIP-seq peaks characterization, see fig. S8 (A to C), and for MACS2 regions, see Supplementary Data. (C) Heatmap showing the log2(fold change) ChIP versus input enrichment in HSCs and HPC7 cells. The peak center was used as the reference point while plotting the signal ±5 kb. ChIP-seq profiles were generated for MOF in HSC (gray) and HPC7 (blue), as well as those for H4K16ac in HPC7 (blue). p300 data from HPC7 was from (31) (purple). (D) Box plot showing the normalized transcript expression from MOF-bound genes or random genes across the scRNA-seq clusters. (E) Left: MOF ChIP-seq genome browser snapshots at Runx1 and Cdk8 loci (common bound peak) obtained from sorted HSCs of wild-type or Mof+/− mice. Right: Runx1 and Cdk8 RNA levels in sorted HSCs. dCT was determined by [CT-target gene − CT-bTrc] and then normalized by the smallest dCT (normalization was conducted considering each genotype), ddCT. The fold change is 2ddCT. After normality test, significance was evaluated by Mann-Whitney test (see fig. S8, D and F for MOF ChIP-seq analysis in Mof+/− HSCs). (F) TF affinity prediction (TRAP) motif analysis on MOF peaks from cluster 1 or cluster 2. The top-scoring motifs were chosen on the basis of combined P value. (G) Stacked plot showing the overlap between MOF target regions in HSCs, MEPs, or equally sized random regions enrichment in bulk ATAC-seq peaks (data from Encode). P value was calculated by Fisher’s exact test using the BEDtools function FisherBED. (H) Overlap between extracted k-mers from single-cell ATAC-seq (scATAC-seq) HSCs or MEPs [data from (38)] and MOF-TSS flanking k-mers in HSC and MEP (see Materials and Methods for details). Enrichment was calculated by Fisher’s exact test, and odds ratios are shown in the figure. (I) Subway map showing scATAC-seq pseudotime trajectory. Dots represent single cells, and colors represent cell population [data from (38)]. (J) Subway map showing three k-mers from MOF peaks in HSC or (K) MEP-associated k-mers in the pseudotime trajectory. Scale shows the scATAC-seq z-score. Pseudotime analyses were conducted by STREAM (39). (L) Bar plots showing the P value of MOF-peaks k-mers along HSC-MPP (S2-S0), MPP-MEP (S0-S3), GMP/LMPP-CLP (S4-S5), or GMP/LMPP-pDC (S4-S6) trajectories. One-way ANOVA determined experimental significance. Related to fig. S8.

  • Fig. 5 GFI1B binds the Mof promoter and leads to Mof activation.

    (A) Find individually motif occurrence analysis showing GFI1B consensus sequence enrichment at the mouse and human Mof promoters. (B) Genome browser snapshot of RUNX1 and GFI1B ChIP-seq (40) at the Gfi1b and Mof genes in HPC7 cells. (C) ChIP-qPCR analysis of RUNX1 at the Gfi1b locus and GFI1B at the Mof locus in HPC7 cells. Floating plots showing the average of three independent replicates represented by data points. Enrichment was calculated as percentage of input. (D) Bar plot showing Gfi1b mRNA levels in LSK+ cells from wild-type (gray) and Mof+/− (orange) animals. Normality distribution was scored by Shapiro-Wilk normality test, followed by two-tailed Mann-Whitney test for statistical significance. (E) Left: Representative immunofluorescence showing MOF and DAPI signal in HPC7 cells in wild-type [scr small interfering RNA (siRNA), siRNA control] or upon Gfi1b knockdown (KD; Gfi1b siRNA). Right: Box plot showing the MOF MFI in HPC7 cells upon Gfi1b or scr siRNA treatment. (F) Left: Scheme showing the wild-type Mof promoter construct, in which 790-nt upstream of the Mof 5′ untranslated region (5′UTR) was cloned, and the Mof mutant promoter construct, in which the GFI1b binding region (39 nt) was scrambled. Right: HPC7, K562, Wehi3, and human embryonic kidney (HEK) 293T cells were transfected with luciferase under the expression of either the wild-type or mutant Mof promoter constructs and subjected to either Gfi1b siRNA or Gfi1b ectopic expression (+Gfi1b). Box plot shows the median, maximum, minimum, and the interquartile range for Mof luciferase activity. Mof promoter activity was normalized over the minimal promoter activity. (G) Scheme of the single-cell CFU assay from wild-type (scr siRNA), Runx1 KD, and Gfi1b KD in sorted HSCs. (H) Dot plot showing the total number of colonies observed from wild-type (scr siRNA, gray), Runx1 (dark blue), or Gfi1b (light blue) KD HSCs. (I) Stack plot showing the colony types from (H), overall significance is measured by unpaired ANOVA followed by Kruskal-Wallis test, and colony enrichment is calculated by χ2 test, with the significance threshold set as P < 0.01. (J) Bar plots showing RT-qPCR for Mof, Runx1, and Gfi1b after 10 days in culture. Expression was normalized to Hprt (n = 3). Significance is measured by unpaired ANOVA followed by Kruskal-Wallis test, and P values are calculated by Dunn’s test multiple comparisons test relative to the scr siRNA group. Related to fig. S9A.

  • Fig. 6 Erythroid trajectory perturbation can be rescued by Gata1 expression or by rebalancing acetylation with HDACi.

    (A) t-SNE map showing the erythroid fate bias. Fate bias was calculated per cell (dot in the t-SNE), and colors indicate the power of the bias, 1 being the highest (red) and 0 the lowest (blue) probability of becoming an erythroid cell. (B) Self-organizing map (SOM) generated from the erythroid trajectory (principal curve). For SOM ordering, 264 wild-type and 215 Mof+/− cells were extracted from the principal curve and fitted to cells with a fate bias of >0.4. The SOM shows the cells in pseudotemporal order on the x axis and transcripts that have similar expression profiles along the trajectory aggregated into nodes on the y axis (Pearson’s correlation coefficient, >0.85), where eventually, every transcript corresponds to a line. Node composition and order are the same for wild-type and Mof+/− trajectories (for analysis workflow, see fig. S9B, and the list of genes/nodes and differentially expressed genes along the trajectory is in Supplementary Data). (C) Box plot showing the mean expression of all genes belonging to SOM nodes 5, 30, or 42 in the dHSC (cluster 2), aHSC (cluster 5), MPP (cluster 4), MEP (cluster 6), and erythroblast (cluster 3) populations identified in our scRNA-seq dataset (Fig. 2A). (D) Graphical scheme showing the experimental design of Mof rescue experiments. Control or Mof-iKO HSCs (LSK+CD34Ftl3CD48CD150+) were sorted, transfected with pTreg3g-Mof or pTreg3g-MofE350Q (MOF catalytic mutant), and cultured with 4-OHT in methylcellulose medium for 10 days. After 10 days in culture, colony formation and molecular profiles were evaluated. (E) Bar plots showing RT-qPCR analysis of Mof, Runx1, and Gfi1b levels in control (white bars) and Mof-iKO (red bars) cells. Levels were determined by dCT (CT-target gene − CT-Hprt) and then normalized by the smallest dCT, ddCT. The fold change is 2ddCT. Experimental P values were calculated by one-way ANOVA. Overlaid dots indicate the number of animals (n = 3). (F) Stacked bar plot showing colony profile from transfected HSCs. Error bars represent means ± SEM from n = 4 independent biological replicates. Circles indicate the mice genotype: control (black) and Mof-iKO (red). Experimental P value was determined by one-way ANOVA, and statistical significance was set as P < 0.05. For the total number of colonies, see fig. S9D. The “+” signal under the plot indicates the transfection condition. (G) Graphical scheme showing the experimental design of Gata1 rescue experiments. Control or Mof-iKO HSCs were sorted, transfected with the pCAG-Gata1 expression vector to enforce erythroid trajectory commitment, and cultured with 4-OHT in methylcellulose medium for 10 days. After 10 days in culture, colony formation was evaluated. (H) Bar plot of mRNA levels Mof and Hbb-bs in Mof-iKO and Mof-iKO plus exogenous Gata1 (indicated by “+”) (number of animals, n = 3). Relative expression was calculated as above. For the total number of colonies, see fig. S9E. (I) Colony quantification is represented by the stacked bar plot. Error bars represent means ± SEM from n = 4 independent biological replicates. Experimental P values were determined by one-way ANOVA, with statistical significance set as P < 0.05. (J) Graphical scheme showing the experimental design of HDACi rescue experiments. Mof+/+, Mof+/−, Mof-iKO, and control HSCs were sorted, treated with 1 mM Ex-527 treatment and cultured with (Mof-iKO and control) or without (Mof+/+ and Mof+/−) 4-OHT in methylcellulose medium for 10 days. After 10 days in culture, colony formation was evaluated. Mice genotypes are represented by the circles control (black), Mof-iKO (red), wild-type (gray), or Mof+/− (orange). Ex-527 treatment is indicated by the “+”. (K) Stacked bar plot representing the colony profile after Ex-527 treatment in primary culture and (L), upon serial plating. Error bars represent means ± SEM from n = 3 independent biological replicates (for total number of colonies, see fig. S9G). (M) As in (K) but in Mof+/+ and Mof+/− HSCs. Cells were treated with Ex-527 either immediately after FACS sorting (day 0) or after 5 days in culture (day 5). Error bars represent means ± SEM from n = 3 independent biological replicates. Enrichment was calculated by χ2 test (P < 0.05). A total number of colonies are shown in fig. S9H. Related to fig. S9.

Supplementary Materials

  • Supplementary Materials

    Temporal expression of MOF acetyltransferase primes transcription factor networks for erythroid fate

    Cecilia Pessoa Rodrigues, Josip Stefan Herman, Benjamin Herquel, Claudia Isabelle Keller Valsecchi, Thomas Stehle, Dominic Grün, Asifa Akhtar

    Download Supplement

    The PDF file includes:

    • Notes S1 and S2
    • Supplementary Data description
    • Legend for data file S1
    • Figs. S1 to S9
    • References

    Other Supplementary Material for this manuscript includes the following:

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article