Chromatin Reader Dido3 Regulates the Genetic Network of B Cell Differentiation

The development of hematopoietic lineages is based on a complex balance of transcription factors whose expression depends on the epigenetic signatures that characterize each differentiation step. The B cell lineage arises from hematopoietic stem cells through the stepwise silencing of stemness genes and balanced expression of mutually regulated transcription factors, as well as DNA rearrangement, in a complex process involving epigenetic remodeling. Here we report the impact on B cell differentiation of the lack of DIDO3, a reader of histone posttranslational modications, in the mouse hematopoietic compartment. We found reduced DNA accessibility in hematopoietic precursors, leading to a severe deciency specically in the generation of successive stages of B-cell differentiation. The expression of essential transcription factors and differentiation markers is impaired, as is the process of somatic recombination. DIDO3-decient cells show transcriptional alterations of a number of polycomb repressive complex 2, suggesting the involvement of DIDO3 in determining the specic activity of PRC2 in the B cell lineage, including VH-DJH rearrangement. CD11b-PeCy7, CD45.2-AlexaFluor700), and BD Pharmingen (CD11b-biotin, CD21-APC, CD23-PE, CD43-biotin, CD45.1-PE, IgG1-PE, IgG3-PE, Ly6G-PE, Nk1.1-APC). Biotinylated antibodies were visualized with streptavidin-eF450 (eBioscience), streptavidin-PE (Southern Biotech) or streptavidin-PE-CF594 (eBioscience). Cell death was monitored using annexin V-APC (Immunostep) and DAPI. In in vivo assays, the cell cycle was analyzed by measuring BrdU incorporation into cell DNA 18 h after intraperitoneal injection of 200 mg BrdU (Invitrogen). Cells were labeled following the protocol for the BrdU Staining Kit for ow cytometry (Invitrogen). Analytical cytometric data was acquired using a Gallios cytometer (Beckman Coulter).

subsequently controls B cell commitment at the transition to the pro-B cell stage, and is necessarily maintained to sustain B-lineage identity 6 .
In the CLP subpopulation, somatic rearrangements are first detected between the diversity (DH) and joining (JH) gene segments of the heavy chain (HC) locus 7 . CLP can mature into four different cell types: B lymphocytes, T lymphocytes, NK, and dendritic cells 8,9 . In mice, upregulation of B220 expression in a CLP subpopulation allows differentiation to prepro-B cells. Prepro-B cells determined for B cell differentiation are characterized by tyrosine kinase AA4.1 (CD93) expression 10 . Entry into the B lineage is marked by CD19 expression and the end of DH-JH rearrangement in the early pro-B state. IL7R signaling contributes to early pro-B cell survival and transient proliferation 11 and sequential reorganization of Ig genes 12 .
Heavy chain recombination culminates in the late pro-B stage with the productive rearrangement of VH heavy chain variable segments with DJH (VH-DJH). Productive rearrangement of Igh is essential for pre-BCR signaling, a crucial checkpoint for survival and further differentiation 13 .
Post-translational histone modifications constitute one of the epigenetic mechanisms that enables control of chromatin structure, DNA accessibility, and gene expression in a dynamic, cell type-specific manner. Proteins that specifically "read" these modifications must play important roles in development, differentiation, and tumorigenesis 14 .
Dido1 is a gene highly conserved throughout evolution; its orthologs have been identified in chordates and insects. The shortest coded protein isoform, DIDO1, includes nuclear location and nuclear export signals (NLS, NES), and a plant Zn-finger homeodomain (PHD); it interacts with chromatin mainly through di/trimethylated lysine of histone H3 (H3K4me2/3) 15,16 , and works as a reader of the chromatin state. DIDO1 nonetheless has yet to be assigned to known complexes. Both PHD and NLS, but not NES, are present in all three natural isoforms. DIDO2 is the least abundant and includes a transcription elongation factor S-II central domain (TFIIS2M), necessary for the nucleolytic activity of RNA polymerase II, and a Spen paralog and ortholog C-terminal (SPOC) domain of unknown function, but which is usually found in proteins with transcriptional co-repressor activity. DIDO3 is the main isoform and includes all domains found in DIDO2 as well as a coiled-coil (CC) region that probably has a tertiary or quaternary structural function. There are also proline-, alanine-, and arginine-rich low complexity regions in DIDO3.
All attempts to date to delete Dido1 have resulted in cell lethality. N-terminal truncation results in a shorter protein, expressed by an internal promoter, that is associated with aneuploidy, centrosome amplification, and DNA damage in the form of centromere-localized breaks 17 . Mice with the Dido1ΔNT mutation show hematological myeloid neoplasms, and alterations in DIDO1 are associated with the myelodysplastic syndrome in humans 18 .
Deletion of the Dido 3' terminal exon (E16) renders a truncated form of DIDO3. This deficiency leads to embryonic lethality; embryos die by gestation day 8.5 19 . Embryonic stem cells (ESC) derived from these DIDO3 C-terminal mutant embryos do not differentiate correctly, but retain their capacity for self-renewal.
Due to the embryonic lethality of Dido exon 16, we studied the effect of DIDO3 deficiency in the hematopoietic lineage using mice bearing a floxed exon 16 of Dido1 20 and a Cre recombinase gene under the control of the Vav1 promoter. In these mice, the B cell population is severely reduced by a primary defect in the pro-B cell stage progression. By analyzing the transcription levels of B cell-characteristic genes at distinct differentiation stages in these DIDO3-deficient mice, we detected a reduction in expression of B cell markers and of transcription factors specifically in prepro-B samples; there was no evident alteration in master regulators such as Spi1 (PU.1) and Ikzf1 (Ikaros). This finding suggests that DIDO3 is involved in the chromatin restructuring needed for the B cell differentiation pathway, and thus controls developmental cell plasticity and lineage fate decisions.

RESULTS
To characterize the Dido1 expression pattern in different human and mouse tissues, cell lines, and cancer types, we retrieved publicly available data from web bio-sources. Publicly accessible expression data using the Genevisible tool (https://genevisible.com/search). The which showed >99% reduction in Dido1E16 DNA content (Suppl. Fig. 2A). Hematologic analysis showed a significant decrease in cellular components when Dido1∆E16 mice were compared with heterozygous littermate controls. White blood cell numbers dropped to 33% of wt (1.71 ± 1.32 vs. 5.11 ± 1.71 x 10 3 cells/ml, n = 8, p<0.001) and red cell numbers to 92% of wt values (9.65 ± 0.36 vs. 10.54 ± 0.51 x 10 6 cells/ml, n = 8, p<0.01). Within the leukocyte fraction, the lymphocyte subpopulation was responsible for the defect observed, with a reduction in cell numbers to 27% of wt (1.02 ± 1.20 vs. 3.84 ± 1.61 x 10 3 cells/ml, n = 8, p<0.0001)(Suppl. Fig. 2B). Peripheral blood samples from wt and Dido1∆E16 mice were analyzed by flow cytometry for EYFP reporter expression; >97% of circulating cells were EYFP-positive in both lymphoid and myeloid gates, as determined by forward scatter (FS) vs. side scatter (SS) gating (Suppl. Fig. 2B).
We analyzed the distribution of the distinct B cell precursor subpopulations, namely prepro-B, pro-B, pre-B, and immature B cells ( Fig. 2A). The number of B220 + cells was substantially lower in Dido1∆E16 mice than in the controls (Fig. 2B). The populations most affected were pre-B and immature B cells, although the number of prepro-B and pro-B cells was similar in mice of both genotypes (Fig. 2C).

Functional analysis of bone marrow precursors
Signaling from the pre-BCR complex promotes the survival and proliferation of cells that have successfully rearranged the Igh locus. This is a first checkpoint in which those rearrangements that give rise to a non-functional heavy chain lead to death by apoptosis. In a second checkpoint, immature B cells whose BCR is activated by an autoantigen in the bone marrow can undergo receptor editing to modify its light chain, or they can be deleted by apoptosis. To detect cell death in distinct precursor populations, we used flow cytometry to analyze bone marrow samples from Dido1∆E16 and wt mice, labeled with appropriate antibodies and with annexin V and DAPI. Apoptosis levels were slightly higher, although nonsignificant, in Dido1∆E16 mice. In the immature B cell population (EYFP + B220 + CD19 + IgM low ), both the number of apoptotic cells and the annexin V mean fluorescence intensity (MFI) in bone marrow of Dido1∆E16 mice showed a significant increase relative to controls (Fig. 3A, B). This result suggests that DIDO3 affects the size of the immature B cell population in Dido1∆E16 mice by increasing the number of cells that undergo apoptotic processes at this stage. Our data support the idea that DIDO3 is necessary to overcome the BCR signaling-dependent checkpoint.
We used BrdU (5-bromo-2'-deoxyuridine) incorporation in DNA in vivo to identify and examine proliferating cells, and DAPI to determine DNA content and cell cycle stages. The percentage of cells that incorporated BrdU after 18 h post-intraperitoneal inoculation was quantified for pro-B/pre-B and immature/mature cells. Cells from Dido1∆E16 mice in pro-B/pre-B states showed BrdU incorporation similar to that of control mice (Fig. 3C). In the immature cell population, there was a non-significant increase in the percentage and the MFI of cells that incorporated BrdU (Fig. 3D)  In vitro differentiation assays were carried out to determine whether the Dido1∆E16 mouse B cell precursors showed alterations in their capacity to differentiate towards the B lineage. We purified cells lacking lineage-specific markers (lin -) by negative selection and plated them in a semi-solid matrix and medium optimized for B cell development, supplemented with IL7. In cultures from wild type mice, between 25 and 35 colonies were counted in plates seeded with 2 x 10 6 lincells/ml; between 5 and 10 smaller cell groups/plate were also registered. No colonies were found in plates seeded with linprecursors from Dido1∆E16 mice, although we detected some clusters consisting of a small number of disaggregated and undifferentiated cells (Suppl. Fig 3). B lineage cells (B220 + ) from the conditional mutant mice thus showed differentiation defects when stimulated in vitro, with blocking of proliferation and maturation towards B precursors in these conditions.

Hematopoietic precursors from Dido1∆E16 mice failed to repopulate the bone marrow of irradiated recipient mice
The small numbers of early hematopoietic precursors in Dido1∆E16 mice and their failure in in vitro differentiation assays led us to test whether their function was affected in vivo in the absence of DIDO3. As transplant recipients, we used nine lethally irradiated B6-CD45.1 mice (10 Gy); at day 1 post-irradiation, the mice received a retro-orbital injection of 150 µL of a mixture of total bone marrow cells (3.75 x 10 6 ) from CD45.2 + EYFP + Dido1∆E16 and CD45.2 + EYFP -Dido1 + donor mice, at a 1:1 ratio (Suppl. Fig. 4).
At day 21 post-inoculation, ~25% of circulating leukocytes were B cells, most of which were  4C). These data indicate that, relative to LSK-wt cells, the LSK-Dido1∆E16 cells and the remainder of the B lineage progenitor cells had a cell-autonomous disadvantage both in nesting and in repopulating the bone marrow of irradiated mice.

Although the proportion of LSK cells within the linpopulation is similar in wt and
Dido1∆E16 mice, adoptive transfer experiments showed a lower repopulation capacity, especially considering that this undifferentiated population was overrepresented in the mutant fraction of the inoculum. As DIDO1 was demonstrated to be a chromatin interactor that recognizes accessible regions of chromatin with H3K4me3 histone marks, we analyzed this population by Omni-ATAC-seq 22 . This technique detects differences in chromatin accessibility, as assessed by sensitivity to DNA digestion by transposase Tn5 in a process termed "tagmentation". The tagmented fragments were subjected to massive sequencing, and two biological replicas with accessible regions in the range of 10 4 were used for differential analysis between wt and Dido1∆E16 LSK cells.
Accessible regions were annotated with the ChIPseeker program 23 by associating the position of the accessible regions in each replica with the structure of the genes (promoter, 5'-UTR, exon, intron, 3'-UTR, or intergenic) in the mouse reference genome assembly GRCm38/UCSC mm10. After filtering low quality reads, an average of 52 million high quality reads were obtained, and 96.1% clean reads were mapped to the GRCm38/UCSC mm10 assembly using Bowtie2 (bowtie-bio.sourceforge.net/bowtie2/) and in parallel BWA-MEM 0.7.15 (http://bio-bwa.sourceforge.net). The largest percentage of accessible regions was concentrated in gene promoters and introns and in intergenic regions, in both LSK-wt and LSK-Dido1∆E16 cells (Fig. 5A). In the Dido1∆E16 samples, however, there was a relative reduction in the number of introns and intergenic regions relative to promoters, which was related to the preferential association of DIDO1 to enhancer sequences over promoters 24 .
A Gene Set Enrichment Analysis (GSEA, http://www.gsea-msigdb.org) of the ATAC-seq data was performed. GSEA is a computerized method that determines whether gene sets defined a priori show concordant and significant differences between two biological states 25 .
In seeking overlaps with genesets in the C2 collection ("curated genesets") of MSigDB, we found that the promoters with the least accessibility in LSK-Dido1∆E16 coincided with (1) genes with mixed epigenetic marks (H3K4me2 and H3K27me3) in mouse embryonic  Table 1). In accordance with these data, our previous microarray data 26 analyzed here by GSEA, indicate that genes silenced in DIDO3-deficient ESC correlate significantly with those silenced in mouse Suz12-deficient ESC; the genes overexpressed in these cells also coincided with those overexpressed in the Suz12 -/cells. There thus appears to be a relationship between epigenetic regulation of gene expression by PRC2 and the presence of DIDO3. An online tool (http://bioinformatics.psb.ugent.be/webtools/Venn/) was used to construct a Venn diagram that represents differentially-expressed genes (DEG) associated with distinct genesets by GSEA (Fig. 5B). There was also correlation with (5)  . The analysis showed that expression of genes related to differentiation towards the B lineage was non-significantly lower in LSK-Dido1∆E16 than in LSK-wt cells. Other genes such as Gata3 or Cdk6 were expressed more significantly in LSK-Dido1∆E16 cells, whereas Nanog levels were 16 times higher than in LSK-wt cells (Fig. 5D). Dido1∆E16 CLP samples showed reduced Ebf1 expression, whereas Nanog levels were normal.

TLDA analysis of gene expression in B cell precursors.
To analyze expression of several genes related to B cell development in the different precursor subpopulations, we used customized TaqMan Low Density Arrays, tested against preamplified samples of cDNA synthesized from total RNA from sorted subpopulations. The largest number of differences was detected in the prepro-B subpopulation (Fig. 6A), which suggests that a smaller number of cells is able to follow the B cell pathway at this lessdifferentiated stage. mRNA levels of transcription factors such as Tcf3, Foxo1, and Ebf1 As in almost every developmental process, the three "E-proteins" play a fundamental role in hematopoiesis. These transcription factors are characterized by a basic helix-loop-helix (bHLH) domain with capacity to homo-and hetero-dimerize with themselves and other HLH proteins, and to bind to DNA through E-boxes (CACGTG sequences). Tcf3 (which encodes E2A isoforms E12 and E47) and Tcf12 (which encodes HEB isoforms HEBcan and HEBalt) have been implicated in B cell development. To test the expression of the distinct isoforms, we designed specific primers for conventional RT-qPCR assays. As already detected in the TLDA assays, Tcf3 expression was lower in Dido1∆E16 than in wt prepro-B cells; this reduction was due equally to E12 and E47 isoforms (Fig. 6B). However, while the expression of the HEBcan isoform of Tcf12 (assessed by exon 1/2 levels) was not affected by DIDO3 absence, HEBalt isoform expression (exon X9 levels) dropped to 10% of the wt in Dido1∆E16 prepro-B cells ( Fig. 6C).
Tcf3 levels remained low in Dido1∆E16 pro-B cells (Fig. 6D), and we also found low Bcl2 and high levels of the cell cycle regulator Cdkn2a (p16, cyclin-dependent kinase inhibitor 2A).
E2A is necessary to maintain Ebf1 and Pax5 expression as well as the B cell genetic program, whereas BCL2 is an antiapoptotic factor, and CDKN2A inhibits the kinase-dependent kinase cyclin 4 (CDK4) and therefore blocks the cell cycle in G1.
In the pre-B Dido1∆E16 population, we again detected a pro-apoptotic environment, with low Bcl2 and high levels of cyclin D2 (Ccnd2), which codes for the regulatory subunit of CDK4 and CDK6, necessary for cell cycle G1/S transition. Our data also showed no silencing of genes that encode pre-BCR components such as Igll1, CD79a, and CD79b, or genes that inhibit pro-B cell differentiation such as Erg (Fig. 6E).
In immature B-Dido1∆E16 cells, we detected an increase in the mRNA levels of stemnessrelated transcription factors (Pou5f1 (Oct-4), Nanog, Kit, and Ly6a (Sca1)), in addition to high levels of pre-BCR surrogate light chain components such as Vpreb2 and Vpreb1, and of cell cycle regulatory genes such as cyclin-dependent kinase inhibitor Cdkn2b (p15INK4b). We also observed overexpression of Id3, which forms heterodimers with HLH proteins that lead to inhibition of DNA binding and thus, regulation of gene transcription (Fig. 6F). All these results suggested that Dido1∆E16 affects differentiation processes both in mouse hematopoietic precursors and in embryonic stem cells.

RNA-seq analysis of gene expression in pre-B cells
As the first checkpoint in B cell development is dependent on pre-BCR activation and takes place at the pre-B stage 29 , where we found the greatest effect of the absence of DIDO3 (seen as lack of progenitor cell progression), we sequenced the transcriptome (RNA-Seq) of cells in this phase.
Using pooled samples (2-3 mice/pool) to obtain sufficient starting material, we sorted pre-B cells by flow cytometry and purified RNA. Two biological replicates were used to construct libraries for massive sequencing. After filtering low quality reads, an average of 24 million clean reads were obtained and 94.91% clean reads were mapped to the mouse genome. A total of 7846 genes with more than three fragments per kilobase of transcript per million mapped reads (FPKM) was used for subsequent analyses. DEG were defined based on the absolute value of their fold change |FC| ≥1.5 (log2FC >0.6|) associated to a false discovery rate (FDR) ≤0.05. We found 120 genes with higher expression in wt pre-B cells, and 346 genes For a second analysis we applied the GSEA tool, using a ranked list of genes with base mean >1, ordered by FC. By analysis against the group of genesets defined as targets of specific transcription factors, we identified the genesets of targets for PRC2 components JARID2 and SUZ12 (p<0.001) (Fig. 7). Aebp2, a substoichometric PRC2 subunit that stimulates the K9 and K27 methylation of histone H3 30

DISCUSSION
Although DIDO3, which is expressed ubiquitously in the organism, presumably has a more general function in determining chromatin structure, its profound influence on the B cell population led us to focus our study on this cell lineage. Speculations on the molecular mechanism involved have been treated elsewhere 31,32,20,33 . Here we demonstrate that DIDO3 plays a role in HSC homing and differentiation, as well as in the later B cell differentiation process.
In lymphopoiesis B, Ebf1 is the key transcription factor for B cell specification and All the canonical E proteins are important for lymphoid development 37 . HEB (Tcf12) and E2 2 (Tcf4) are thought to contribute to the total E protein dose needed for B cell development, since the absence of either factor led to a decrease in pro-B cell numbers, but no obvious developmental block 38 . Of the E proteins, E2A is necessary to maintain Ebf1, Foxo1, and Pax5 expression, and hence the B cell genetic program. B cell development in Tcf3 +/-;Tcf12 -/mice is completely blocked at the CLP stage 39 . It is not known, however, whether this arrest is due to the lack of HEBcan, HEBalt, or both isoforms. In our case, neither is completely absent, but whereas HEBcan levels were normal, HEBalt levels dropped in Dido1∆E16 prepro-B cells to 10% of that in wt, which suggests an undescribed function for this isoform. Lack of E2A function results in an inability to rearrange Igh gene segments, which can be rescued with a single dose of E2A 40 . Although it remains to be determined whether this effect is dosedependent, it could explain our results with Dido1∆E16 pre-B cells, in which a substantial number of RNA reads correspond to the segment between D and J fragments, which was deleted efficiently in wt samples.
We also detected abnormal Igg2b expression in Dido1∆E16 pre-B cells. CpG DNA stimulates TLR9, with intense expression of activation-induced cytidine deaminase (AICDA) and production of Igg2b germ line transcripts in B cell precursors 41 , but we did not find Aicda or the TLR9-pathway components overexpression. We thus support the possibility of relatively large numbers of isotype-switched memory B cells 42 in Dido1∆E16 bone marrow, as were copurified in our pre-B cell sorting strategy.
In Dido1∆E16 pro-B cells, we found high levels of Cdkn2a, which acts as an inhibitor of the kinase-dependent kinase cyclin 4 (CDK4), and low levels of Bcl2, involved in blocking cell death (anti-apoptotic). Cdkn2a is a known target of EZH2 in several cell types 43 . Cdkn2a codes for p19ARF, which stabilizes p53, as well as p16INK4a, a CDK4 inhibitor 44 , both of which could affect the survival and expansion of adaptive lymphoid cells. CDK4 regulates cell progress in G1 and would be inhibited by excess CDKN2A.
EZH2, a core PRC2component, is needed to silence Cdkn2a in developing B lymphocytes prior to the preBCR checkpoint 45 . The Cdkn2a/p53 pathway must be repressed until preBCR selection, but also needs to be accessible to protect developing B cells from oncogenic transformation. Indeed, many early B lineage acute leukemias harbor inactivating mutations at the Cdkn2a locus. In contrast, germinal center B cell lymphomas have activating Ezh2 mutations; in these cells, EZH2 might contribute to tumorigenesis, in part through Cdkn2a repression. EZH2-dependent repression of Cdkn2a is therefore a critical control point for preBCR selection that could be manipulated by lymphoid cells to promote transformation 46 .
We found significant association of the transcriptional alterations in Dido1∆E16 cells with a number of PRC2 targets and genes bearing H3K27me3 epigenetic marks, which suggests that DIDO3 is involved in determining specific PRC2 activity in the B cell lineage. Moreover, loss of EZH2 is known to cause impaired distal VH-DJH rearrangement and Igh locus contraction 47 . There is thus a preference to use proximal V fragments as in our model, which reinforces the view that DIDO3 acts in coordination with PRC2.
We showed differential expression of the DIDO isoforms in MPD 18 , and others reported that BCR-ABL1 translocation (Philadelphia chromosome; Ph) in myelocytic leukemia (CML) might be involved in modulating DIDO1 expression in a JAK2V617F-independent manner 48 .
Disorders in multiple signaling pathways and genetic abnormalities combined with the Ph are essential for the evolution of different types of leukemia; however, why MPD evolve specifically into CML, acute myeloid leukemia, acute leukemia leukocytosis, or mixedphenotype acute leukemia is currently unclear. BCR-ABL1 kinase hyperactivity leads to the activation of signaling pathways and deregulation of cellular processes 49,50 , and hyperactivation of most of these pathways have been confirmed in CML and ALL mouse models. BCR-ABL1 activity channels mainly through JAK/STAT and PI3K/AKT/mTOR pathways, and through CCAAT/enhancer-binding proteins (C/EBP, which regulate normal myelopoiesis as well as myeloid disorders) 51 . Our Gene Ontology analysis of RNAseq data showed that, in cells lacking DIDO3, the expression of PI3K/mTOR and JAK/STAT pathway components, as well as Cebpb, are affected at the transcriptional level. Statistical analysis of DNA methylation showed that target genes of NANOG 52 , Polycomb-group target genes, and/or genes with an AACTTT promoter motif can have a methylation profile that seems to indicate predisposition to development of cancer, such as breast cancer 53

Mice
In brief, a mouse strain bearing an exon 16 flanked by loxP sites was generated and interbred with Cre VAV1 -producing mice heterozygous for exon 16 Supplementary Table 4).
Genomic annotations: The data were obtained, processed, and annotated using tools from R (www.r-project.org/) and Bioconductor (bioconductor.org). To determine the equivalent accessible regions in the replicas, the Intersect tool and the files produced by MACS2 were used. Genes, promoters, UTRs, introns, and exons were allocated to the accessible regions with the ChIPseeker program (github.com/GuangchuangYu/ChIPseeker).
Differential accessibility analysis of chromatin: The sequences (readings) in the accessible and equivalent regions of all replicates were counted and normalized with the StringTie 1.3.3 method (ccb.jhu.edu/software/stringtie). To calculate the differential accessibility of chromatin, the accessible and equivalent regions were analyzed with at least one reading in all replicas, with the DESeq2 (bioconductor.org/packages/DESeq2) and edgeR methods (bioconductor.org/packages/edgeR). Accessible regions with a fold change value >2 and a false discovery rate (FDR) <0.05 were considered statistically significant. We also used the HOMER tool in the differential accessibility analysis of chromatin. Gene Set Enrichment Analysis (GSEA, www.gsea-msigdb.org/) was used for the ATAC-seq data.

RNA-Seq
Pre-B cells extracted from the bone marrow of wt and Dido1∆E16 mice were isolated by sorting. RNA was purified using the RNeasy mini plus kit (Qiagen) and massive sequencing was performed at BGI Genomics (Hong Kong) using the HiSeq 2000 platform.
RNA-seq data analysis. Sequenced reads were processed in parallel using the Galaxy RNAseq pipeline (usegalaxy.org) and a combined approach with BWA-MEM 0.7.15, StringTie

Competing interests
The authors declare no competing interests.

Data availability
ATAC-seq and RNA-seq data are deposited in the Gene Expression Omnibus under accession code GSE157228. Private access token: gpunwiqctludnib

Supplementary figure 5. STRING analysis and clustering of differentially expressed genes between
Dido1∆16 and wt LSK cells.