The transcription factor DDIT3 is a potential driver of dyserythropoiesis in myelodysplastic syndromes

Abstract

Myelodysplastic syndromes (MDS) are hematopoietic stem cell (HSC) malignancies characterized by ineffective hematopoiesis, with increased incidence in older individuals. Here we analyze the transcriptome of human HSCs purified from young and older healthy adults, as well as MDS patients, identifying transcriptional alterations following different patterns of expression. While aging-associated lesions seem to predispose HSCs to myeloid transformation, disease-specific alterations may trigger MDS development. Among MDS-specific lesions, we detect the upregulation of the transcription factor DNA Damage Inducible Transcript 3 (DDIT3). Overexpression of DDIT3 in human healthy HSCs induces an MDS-like transcriptional state, and dyserythropoiesis, an effect associated with a failure in the activation of transcriptional programs required for normal erythroid differentiation. Moreover, DDIT3 knockdown in CD34⁺ cells from MDS patients with anemia is able to restore erythropoiesis. These results identify DDIT3 as a driver of dyserythropoiesis, and a potential therapeutic target to restore the inefficient erythroid differentiation characterizing MDS patients.

Fig. 1. Altered transcriptional profiles of HSCs in aging and MDS.

a Schematic representation of lesions taking place in aging and in MDS. Aging is characterized by transcriptional lesions of HSCs and can also present mutations in specific genes. MDS HSCs usually present genetic lesions but the transcriptional profile of these cells remains unexplored. Part of this figure was made in ©BioRender - biorender.com. b Schematic representation of the experimental approach: bone marrow specimens from young and older healthy adults as well as from MDS patients were obtained, and HSCs (CD34⁺ CD38^- CD90⁺ CD45RA^-) were isolated using FACS. The transcriptome of these cells was characterized using MARS-seq. Part of this figure was made in ©BioRender - biorender.com. c Principal component analysis (PCA) of the transcriptome of cells isolated in B. Healthy young (blue), older (yellow) adults, and MDS cells (red) are plotted. The percentage of variance explained by PC1 and PC2 principal component is indicated in each axis. d Venn diagram representing a partial overlap of genes deregulated in aging (young vs older healthy adults), in the transition to the disease (healthy older adults vs MDS) or between the more distant states (healthy young vs MDS). The number of common or exclusive DEGs is indicated in each area. e Bar-plot showing enriched biological processes determined by GO analysis of genes deregulated in aging or between HSCs from healthy older adults and MDS patients. The -log(p-value) for several statistically significant processes is depicted. Analysis was performed using DavidOntology. Modified Fisher exact test was used to calculate p-values. f Transcriptional dynamisms identified in HSCs in the aging–MDS axis (C1–C8). Left: dot plot depicting the median expression of genes of each cluster in each sample. Each color represents the different states: healthy young adults: blue, healthy older adults: yellow, and MDS: red. N = 37 biologically independent samples (n = 17 young adults, 8 older adults, and 12 MDS patients). The trend of each cluster is indicated with a line linking the median of each group. Center: heatmap showing z-scores for the expression profile of each cluster of genes in healthy young and older adults and in MDS samples. Right: Bar-plot indicating the number of genes per cluster. g Heatmap showing the statistical significance (log₁₀p value) of enrichment of the genes of each cluster in different biological processes, as determined by GO analysis. The different processes have been manually grouped into more general biological functions (right). Modified Fisher exact test was used to calculate p values.

Fig. 2. Different transcriptional dynamisms of HSCs in the aging-MDS axis associate with specific biological functions.

a GSEA analysis of genes from C1, C2, C5 and C6. The normalized enrichment score (NES) for several cancer-related signatures is depicted. Only signatures in which age-matched healthy controls were used as normal counterparts of tumor cells were considered. b Heatmap of z-scores of genes with known roles in the development of myeloid malignancies. The cluster to which they correspond, and the gene names are indicated on the left and right side of the heatmap, respectively. c Normalized expression of several genes from b in healthy young adults (blue), healthy older adults (orange) and MDS (red) samples. N = 37 biologically independent samples (n = 17 young adults, 8 older adults, and 12 MDS patients). Each point represents an individual and the mean +/− standard deviation (SD) is depicted. d Bubble plot representing statistically significant biological processes and pathways enriched in genes being specifically altered in MDS (C3: red dots; C4: blue dots). Bubble size depicts −log10(p value) and x axis represents GeneRatio. Analysis was performed using DavidOntology. Modified Fisher exact test was used to calculate p values. e GSEA analysis of genes from C3 and C4. The NES for several enriched signatures is shown. f Normalized expression of several genes from C3 involved in transcriptional regulation in healthy young adults (blue), healthy older adults (orange) and MDS (red) samples. N = 37 biologically independent samples (n = 17 young adults, 8 older adults, and 12 MDS patients). Each point represents a donor or patient and the mean +/− standard deviation (SD) is shown.

Fig. 3. DDIT3 overexpression promotes and MDS-like transcriptional state.

a Schematic representation of the experimental procedure: primary HSCs/CD34⁺ cells from healthy young donors were FACS-sorted and infected with a lentiviral plasmid harboring DDIT3, or a control plasmid. Two days after infection, transduced cells were sorted and their transcriptome was analyzed using MARS-seq. After 4 days of infection, transduced cells were also isolated and used to perform colony and liquid culture differentiation assays. The later were evaluated by flow cytometry and MARS-seq at different time points and by scRNA-seq at day 14 of differentiation. Part of this figure was made in ©BioRender - biorender.com. b Volcano plot of statistical significance (−log₁₀ (p value)) against fold-change (log₂ Fold-change) of gene expression between cells overexpressing DDIT3 and control HSCs. Green points represent genes with |FC | > 2, red points depict genes with |FC | > 2 and FDR < 0.05 and grey points indicate genes with no relevant changes in expression. Several genes with significant up- and down-regulation are indicated. Statistical values of differential expression were calculated with DEseq2, and plotted using EnhancedVolcano library. c, d GSEA of cells overexpressing DDIT3 and control HSCs. The NES for several signatures related to general biological processes (c) and cancer-related signatures (d) are depicted. e, f GSEA plots depicting the enrichment upon DDIT3 overexpression in gene signatures representing genes up- and down-regulated in HSCs from MDS patients when compared to those from young healthy adults. GSEA was performed with the GseaPreranked tool. The NES and adjusted p values are indicated for each signature.

Fig. 4. DDIT3 overexpression promotes a defect in erythroid differentiation.

a Bar and scatter plot indicating the number of colonies (BFU-E and GEMM, M, GM) obtained for control cells or cells overexpressing DDIT3 in three independent biological replicates. Data are presented as mean values +/− SD. p values were calculated with multiple t test corrected for multiple comparisons using the Holm–Sidak method. b Flow cytometry charts representing advanced erythroid differentiation (CD71 and CD235a markers; stages I–IV) for control cells and cell overexpressing DDIT3, at the indicated time points. c Box-plots (center line, median; box limits, 25th and 75th percentiles; whiskers, min to max) representing the percentage of cells observed in c in stages I–IV at 10 and 14 days of differentiation. Three independent biological replicates are depicted. Statistically significant differences are indicated, and were calculated with multiple t test corrected for multiple comparisons using the Holm–Sidak method. d UMAP plot of the transcriptome of cells subjected to ex vivo differentiation for 14 days. Cells have been clustered in different groups representing erythroid differentiation: CD34⁺ cells, burst forming unit/primitive erythroid progenitor cells (BFU), colony formation unit/later-stage erythroid progenitor cells (CFU), proerythroblast, early-basophilic erythroblast, late-basophilic erythroblast, polychromatic erythroblast, orthochromatic erythroblast, and reticulocyte. e Bar-plot representing the proportions of cells in each of the clusters described in f for control cells or cells overexpressing DDIT3 upon ex vivo differentiation for 14 days. f RNA velocity plotted in UMAP space for control or DDIT3-overexpressing cells in clusters defined in e. Streamlines and arrows indicate the location of the estimated future cell state. g UMAPs representing the length of velocity (from low-purple to high-yellow velocity length) for control and DDIT3-overexpressing cells. h Box plots representing the length of velocity for each erythroid differentiation state for control and DDIT3-ovexpressing cells. Box boundaries are the 25th and 75th quartiles, and the line within the box is the 50th quartile (median). Whiskers limits correspond to the largest and smallest values within 1.5 times the interquartile range (IQR). Dots represent values >1.5 times the IQR. p values were calculated using the Wilcoxon test. Biologically independent replicates (cells) for each differentiation state were: BFU: n = 6 (Ctrl), n = 4 (DDIT3); CFU: n = 62 (Ctrl), n = 67 (DDIT3); proerythroblast: n = 407 (Ctrl), n = 172 (DDIT3); early_basophilic: n = 176 (Ctrl), n = 76 (DDIT3); late_basophilic: n = 853 (Ctrl), n = 279 (DDIT3); polychromatic: n = 1,649 (Ctrl), n = 371 (DDIT3); orthochromatic: n = 2,918 (Ctrl), n = 413 (DDIT3); reticulocytes: n = 791 (Ctrl), n = 92 (DDIT3).

Fig. 5. DDIT3 upregulation leads to a failure in the activation of erythroid differentiation programs.

a Bar-plot depicting GSEA analysis upon DDIT3 overexpression at day 14 of cells overexpressing DDIT3 and control HSCs at day 14 of erythroid differentiation. The NES for several signatures related to erythropoiesis and stem and progenitor profiles are shown. b Normalized expression of erythroid differentiation (top) and stem cell genes (bottom) of CD34⁺ cells at different time points of erythroid differentiation. N = 2 biologically independent samples. c Plot representing GSEA after differential expression analysis between cells overexpressing DDIT3 and control cells in scRNA-seq, done for the different stages of erythroid differentiation detected. We used the average logFC to rank all expressed genes. We tested for enrichment of signatures related to erythropoiesis and stem and progenitor profiles from MSigDB. GSEA was performed with the fgsea R package. The size of the dots represents NES absolute value, the color indicates the group in which processes are enriched (blue in control cells, red in DDIT3-overexpressing cells), and the intensity of the color depicts the p-value obtained for each geneset. d Gene expression trends of early hematopoietic progenitor genes (top) and erythroid differentiation factors (bottom) calculated by pseudotime are represented as a smooth fit (prediction of the linear model) with the SD of the fit shown in a lighter shade for control (blue) and DDIT3-overexpressing cells (red). p values showing the statistical differences between both trends of expression are indicated, and were calculated using the Wilcoxon test. e Heatmap showing the regulatory dissimilarity score between control and DDIT3-overexpressing cells at day 14 of ex vivo liquid culture differentiation in different stages of erythroid differentiation defined by scRNA-seq. f Ridge plot showing AUC scores for regulons of KLF1, ARID4B and SOX6 in control (blue) and DDIT3-overexpressing cells (pink) at different stages of erythroid differentiation. g Normalized expression of TFs guiding regulons showing decrease activity in DDIT3-overexpressing cells, at different day 14 of ex vivo erythroid differentiation. N = 2 biologically independent samples. h Analysis of putative transcription factor site enrichment in TFs guiding regulons with decreased activity in DDIT3-overexpressing cells using CiiiDER. Color and size of circles reflect p value of enrichment. Over-represented transcription factors of potential interest are depicted. i Immunoblot for CEBPB (different isoforms indicated) and DDIT3 after the immunoprecipitation with an anti-CEBPB antibody or an IgG control in cells transduced with a control or a DDIT3-overexpressing plasmid. These experiments were repeated three independent times with similar results.

Fig. 6. DDIT3 knockdown in MDS patients with anemia restores erythroid differentiation.

a Schematic representation of DDIT3 knockdown experiments in CD34⁺ cells from patients with MDS showing anemia. Cells were transduced with a control or a DDIT3-targeting shRNA, and after 2 days of infection, cells were subjected to ex vivo liquid culture differentiation over OP9 stromal cells. The differentiation state was evaluated by flow cytometry and MARS-seq analyses. Part of this figure was made in ©BioRender - biorender.com. b Flow cytometry charts representing advanced erythroid differentiation (CD71 and CD235a markers; stages I–IV) for cells from patient #13 harboring a control shRNA (shCtrl) or shRNAs targeting DDIT3, at the indicated time points. c Bar-plots representing the percentage of cells observed in b in stages I–IV at 7 and 13 days of differentiation. d Normalized expression of hemoglobin (top) and stem cell (bottom) genes of CD34⁺ cells from patient MDS13 transduced with a shRNA control or an shRNA targeting DDIT3 and subjected to 7 days of ex vivo differentiation. e Heatmap of z-scores of genes characteristic of proerythroblasts, early and late basophilic erythroblasts (left), and of genes expressed in poly- and orthochromatic stages (right), for cells from patient MDS13 transduced with a shRNA control, or an shRNA targeting DDIT3, and subjected to 7 of ex vivo differentiation.