SF3B1 deficiency impairs human erythropoiesis via activation of p53 pathway: implications for understanding of ineffective erythropoiesis in MDS

The antibodies used for flow cytometry were as follows: mouse monoclonal antibody against human band 3 generated in our laboratory and labeled with FITC or APC as described previously [29]. Commercial antibodies used for flow cytometry were as follows: PE-CD235a (GPA), PE-CD34, FITC-CD36, and 7AAD (BD Pharmingen, USA); APC-α4 integrin (Miltenyi Biotec, USA); and PE-Cyanine7-IL-3R (CD123) and PE-Cyanine7-Annexin V (eBioscience, USA). The antibodies used for western blotting were as follows: rabbit anti-SF3B1 was from Abcam (USA); rabbit anti-MKRN1 antibody was from BETHYL (USA); rabbit anti-p53, rabbit anti-p21, rabbit anti-BBC3, and rabbit anti-BAX were from Cell Signaling (USA); monoclonal anti-actin antibody was purchased from Sigma (USA); HRP-conjugated goat anti-rabbit IgG was from Thermofisher (USA); and HRP-conjugated mouse anti-goat IgG was from Invitrogen (USA). SYTO-16 green fluorescent was from Invitrogen (USA).

CD34⁺ cell culture, fluorescence-activated cell sorting of erythroblasts, flow cytometry analysis, colony forming assay, cytospin preparation, western blotting analysis, and statistical analysis were performed as previously described [29, 30].

Full-length MKRN1 was cloned into the modified pRRLSIN.cPPT.PGK-IRES-GFP plasmid (abbreviated as pGFP in this study) [17]. Human full-length MKRN1 was amplified by PCR using pcDNA3-Flag-MKRN1 (Addgene, catalog number: 78751) as the template. The primer sequences for the amplification are in Additional file 1: Table S1. The resultant plasmid is abbreviated as pGFP-MKRN1.

To prepare the lentivirus for shRNA knockdown, 293T cells were co-transfected with packaging plasmid pCMV8.9, envelope plasmid pucMDG and pLKO1 vector which expresses shRNA against luciferase or targeted gene. All of the plasmids were purchased from Sigma-Aldrich. For MRKN1 overexpression, instead of pLKO1 vector, the pGFP-MKRN1 vector expressing full-length human MKRN1 was used. Supernatant containing viral particles was collected, the viral titers were measured, and shRNA-mediated knockdown was performed as previously described [17]. The SF3B1 knockdown shRNA sequences are listed in Additional file 1: Table S1.

The primers for SF3B1, MKRN1, TP53, CDKN1A, BAX, and BBC3 were obtained from Harvard primer bank, and the primers for the top differentially spliced transcripts were designed using Primer-BLAST. Eurofins MWG Operon LLC synthesized the primers. The sequences of all the primers are listed in Additional file 1: Table S1. The primer specificity was validated by documenting a single peak in the melt curve and further confirmed with agarose gel electrophoresis of amplified products. The real-time PCR was performed as previously described [17].

EdU kit was used for cell cycle measurement according to the manufacturer’s protocol. In brief, 1 × 10⁶ cells were incubated with 10 μM EdU for 2 h at 37 °C. After incubation, the cells were harvested and washed with 3 ml of 1% BSA in PBS. Cells were then fixed, permeabilized, and stained with EdU detection cocktail as well as 7AAD. The staining of EdU and 7AAD was analyzed by flow cytometry. Data were collected and analyzed using FlowJ, and the data are expressed as EdU fluorescence intensity versus 7AAD.

Lentivirus particles of MKRN1 expression construct were packaged as described above. SF3B1 shRNA lentivirus particles and pGFP or pGFP-MKRN1 overexpression lentivirus particles were co-transfected into CD34⁺ cells on day 2 of culture. Puromycin was added for selection of SF3B1 shRNA transduced cells. GFP⁺ cells (co-transduced with SF3B1-shRNA2/pGFP or SF3B1-shRNA2/pGFP-MKRN1) were sorted on day 7 by FACS. The sorted cells were cultured under erythroid differentiation condition.

CD34⁺ cells were cultured as previously described. On day 12, cells were treated with DMSO or PLK1 inhibitor (Selleckchem, USA; catalog no. S7720) dissolved in DMSO at the final concentration of 100 μM. Cytospins were prepared as described above.

RNA was extracted from stage-matched luciferase control and SF3B1 knock down erythroid cells. cDNA Libraries were prepared using the standard illumina TruSeq kit and sequenced at Beijing Genomics Institute (BGI, China) using the Illumina HiSeq 4000 platform. Sequencing effort produced ~ 130 million paired-end 100 bp reads per sample. Quality control was performed on the sequenced reads, and low-quality reads were removed. The reads were then aligned with Tophat2 [31] short read aligner following the protocol detailed in Trapnell et al. [32]. Splice aware alignment was performed using cufflinks to assemble transcripts with the human hg19 genome as reference. Gene expression analysis was performed pair-wise between the luciferase and SF3B1 shRNA knockdown at each development stage using cuffdiff. Gene set enrichment analysis was performed using the pre-ranked list analysis against the curated gene sets of canonical pathways [33, 34]. Differentially expressed genes were extracted using the cummerbund R package. Differentially spliced genes where annotated and extracted using spliceR [35].

To explore the role of SF3B1 in human erythropoiesis, we first examined the expression of SF3B1 in highly purified erythroid cells at each distinct developmental stage derived from cord blood CD34⁺ cell cultures. SF3B1 is abundantly expressed at all stages of erythroid differentiation as assessed by RNA-seq analysis of SF3B1 transcript expression (Additional file 2: Figure S1). This pattern of expression of SF3B1 was further confirmed by real-time PCR (Additional file 2: Figure S1). Western blot analysis and analysis of proteomic data of terminally differentiated erythroblasts [36] showed that SF3B1 protein levels decreased in late-stage erythroblasts (Additional file 2: Figure S1).

Erythropoiesis is a complex process that can be divided into early-stage erythropoiesis and terminal erythroid differentiation. During early-stage erythropoiesis, HSC cells differentiate to generate the erythroid progenitors: BFU-E (burst-forming-unit-erythroid) cells that further differentiate to generate CFU-E (colony-forming-unit-erythroid) cells. We examined the effects of SF3B1 knockdown on early-stage erythropoiesis that occurs mainly from day 4 to day 7 of culture under our erythroid differentiation condition. On day 6 of culture, > 50% knockdown efficiency was achieved in erythroid cells by two independent shRNAs at the mRNA level (Fig. 1a) and at protein level (Fig. 1b). Importantly, SF3B1 knockdown severely inhibited erythroid cell growth (Fig. 1c). To examine the cause for this impaired cell growth, we measured the extent of cell apoptosis. The representative profiles of dual staining with 7AAD and Annexin V of cells at day 6 of culture are shown in Fig. 1d. Quantitative analysis reveals that the percentage of Annexin V⁺ cells was increased from ~ 5% for luciferase shRNA-transduced cells to ~ 20% for SF3B1 shRNA-transduced cells (Fig. 1e). The impaired growth of SF3B1 shRNA-transduced erythroid progenitors was further validated by their failure to give rise to BFU-E and CFU-E colonies compared to luciferase shRNA-transduced cells (Fig. 1f). To further confirm these findings, we performed colony-forming assays using cells sorted based on cell surface markers that define BFU-E and CFU-E stages. As shown in Fig. 1g, the sorted BFU-E and CFU-E cells also failed to give rise to erythroid colonies as a consequence of increased apoptosis. Taken together, these findings imply that SF3B1 knockdown impairs proliferation of erythroid progenitors.

We then examined the effects of SF3B1 knockdown on terminal erythroid differentiation, the process during which morphologically recognizable proerythroblasts derived from CFU-E cells undergo 4 to 5 mitosis to generate enucleated reticulocytes. The differentiation of CFU-E cells to proerythroblasts is characterized by the surface expression of the erythroid specific marker glycophorin A (GPA). As shown in Fig. 2a, while more than 40% of luciferase shRNA-transduced cells were GPA positive on day 7, less than 20% of SF3B1 shRNA-transduced cells were GPA positive, implying SF3B1 knockdown impairs differentiation of CFU-E to proerythroblasts. Differentiation of proerythroblasts to late-stage erythroblasts was monitored by flow cytometry using α4 integrin and band 3 as surface markers. SF3B1 knockdown delayed terminal erythroid differentiation that occurs from day 9 to day 13 in the culture system as reflected by the impaired downregulation of α4 integrin and delayed upregulation of band 3 (Fig. 2b). However, the expression patterns of α4 integrin and band 3 was similar between luciferase-shRNA-transduced and SF3B1-shRNA-transduced cells on days 15 and 17 of culture implying the eventual effective completion of terminal erythroid differentiation of surviving cells (Fig. 2b).

We also monitored cell growth in our erythroid culture system from day 7 to 14. While cells transduced with luciferase-shRNA proliferated exponentially from day 7 to day 14 with cell numbers increasing from 1 million to ~ 90 million (Fig. 2c), the rate of growth was significantly impaired for cells transduced with SF3B1-shRNAs with cell numbers increasing from 1 million to only ~ 15 million. To define the mechanistic basis for the impaired cell growth, we measured the extent of cell apoptosis. The representative profiles of dual staining with 7AAD and Annexin V of cells at day 9 of culture are shown in Fig. 2d. The Annexin V⁺ population increased from ~ 3% for luciferase-shRNA-transduced cells to ~ 15% for SF3B1-shRNA-transduced cells. Quantitative analysis of apoptosis data from three independent experiments is shown in Fig. 2e. We also monitored changes in cell cycle by EdU incorporation assay, and the representative profiles of the cell cycle analysis are shown in Fig. 2f. Quantitative analysis of data from three independent experiments showed an increased G1 phase in conjunction with decreased S phase following SF3B1 knockdown (Fig. 2g). These findings demonstrate that the severely impaired growth of erythroblasts following knockdown of SF3B1 is a consequence of increased apoptosis and cell cycle arrest.

When examining the effects of SF3B1 knockdown on morphology of erythroid cells, we surprisingly found that many terminally differentiated erythroblasts exhibited abnormal nuclei following SF3B1 knockdown (Fig. 3a). Quantitative analysis of data derived from three independent experiments showed that while < 10% of control late-stage erythroblasts had abnormal nuclear morphology, knockdown of SF3B1 increased the percentage of cells with abnormal nucleus to ~ 30% (Fig. 3b). To determine the developmental stage at which this abnormal nuclear morphology occurs, we sorted erythroblasts at distinct developmental stages. As shown in Fig. 3c, the generation of abnormal nucleus occurred only at polychromatic and orthochromatic erythroblast stages but not at earlier stages of development. The percentage of cells with abnormal nucleus was ~ 30% for polychromatic erythroblasts and ~ 40% for orthochromatic erythroblasts (Fig. 3d).

To delineate the underlying molecular mechanisms for the observed changes in apoptosis and cell cycle following SF3B1 knockdown, we performed RNA-seq analysis on purified luciferase-knockdown and SF3B1-knockdown CFU-E cells. Bioinformatics analysis of the RNA-seq data revealed that 255 genes were differentially expressed, of which 167 genes were downregulated and 88 genes were upregulated. List of the differentially expressed genes is shown in Additional file 3: Table S2. To define the pathways affected by SF3B1 knockdown, we performed gene set enrichment analysis. Figure 4a shows the top five upregulated and the top five downregulated pathways. Notably, consistent with changes in apoptosis and cell cycle arrest, the highest upregulated pathway is the p53 pathway. As the p53-induced cell cycle arrest and apoptosis are mediated by cell cycle inhibitor p21 [37, 38] and by transcriptional activation of Bcl-2 family pro-apoptotic genes such as Bax and BBC3 (PUMA) respectively [39‐41], we compared the levels of expression of p53 and its downstream targets between control and SF3B1 knockdown cells. The RNA-seq data revealed that while there is no difference in p53 mRNA levels, the mRNA levels of Bax, p21, and BBC3 are increased in SF3B1-knockdown CFU-E cells (Fig. 4b). These mRNA expression levels were further confirmed by real-time PCR (Fig. 4c). Importantly, western blot analysis documented the increased protein levels of downstream targets of p53, p21, BAX, and BBC3, but surprisingly, p53 protein level was also increased in spite of lack of increase in its mRNA levels (Fig. 4d).

As SF3B1 is the core component of the splicesome and is involved in the 3′ splice site recognition [42, 43], we analyzed the effects of SF3B1 knockdown on RNA splicing. We found that 288 transcripts were differentially spliced following SF3B1 knockdown in CFU-E cells. The list of differentially spliced transcripts is shown in Additional file 4: Table S3. Of the 288 differentially expressed transcripts, 198 could be categorized into different types of splicing events. The schematic presentation of types of altered splicing events is shown in Additional file 2: Figure S2. As shown in Fig. 5a, the most frequent splicing events noted were exon skipping/inclusion (ESI), alternative transcription start site (ATSS), and alternative transcription termination site (ATTS). We confirmed the expression of these differentially spliced transcripts by real-time RT-PCR (Fig. 5b). Amongst the differentially spliced genes, we focused our attention on Makorin Ring Finger Protein 1 (MKRN1) since it has been previously reported that the MKRN1 E3 ligase controls cell cycle arrest and apoptosis by regulating p53 and p21 [44]. As shown in Fig. 5c, there are two MKRN1 transcripts, which we termed large and small isoforms. While only the large isoform is predominantly expressed in control cells, SF3B1 knockdown resulted in decreased expression of the large isoform accompanied by increased expression of the smaller isoform (Fig. 5d).

The finding that the increased p53 protein level is not accompanied by increased mRNA level suggested that the increased levels of p53 protein could be due to impaired protein degradation. It has been well documented that p53 degradation is mediated by MDM2, the first identified E3 ubiquitin ligase for p53 [45, 46]. More recently, MKRN1 has also been identified as an E3 ligase that can target p53 for degradation [44]. Our finding that SF3B1 knockdown leads to isoform switch of MKRN1 strongly suggests a potential link between SF3B1 knockdown, MKRN1 splicing change, and p53 activation. To test this thesis, we first examined the protein levels of MKRN1. As shown in Fig. 6a, the large isoform of MKRN1 is significantly decreased in cells following SF3B1 knockdown. To provide further evidence that SF3B1 knockdown-induced changes are due to decreased expression of large isoform of MKRN, we ectopically expressed the large isoform of MKRN1in SF3B1-knockdown cells. As shown in Fig. 6b, transfection with GFP-MKRN1 but not GFP alone rescued the cell growth. Figure 6c shows that the apoptotic phenotype of SF3B1-knockdown cells were also rescued following ectopic expression of the large isoform of MKRN1. Moreover, as shown in Fig. 6d, ectopic expression of GFP-MKRN1 large isoform also rescued the protein levels of p53, p21, Bax, and BBC3.

In addition to apoptosis and cell cycle arrest, another phenotypic change noted during terminal erythroid differentiation is the generation of erythroblasts with abnormal nuclei at the polychromatic and orthochromatic erythroblast stages. To explore the molecular basis for these changes, we performed RNA-seq analysis on luciferase-knockdown and SF3B1-knockdown polychromatic and orthochromatic erythroblasts. Bioinformatics analysis revealed that 602 genes were differentially expressed between luciferase-knockdown and SF3B1-knockdown polychromatic erythroblasts and 461 genes are differentially expressed between luciferase-knockdown and SF3B1-knockdown orthochromatic erythroblasts. The list of these differentially expressed genes is shown in Additional file 3: Table S2. Gene set enrichment analysis revealed that the top five downregulated pathways are associated with mitosis/cytokinesis pathway and DNA replication pathway (Additional file 5: Table S4). As abnormal nucleus is most likely to be associated with defects in mitosis/cytokinesis, we focused our attention on this group of genes and found that several mitosis/cytokinesis genes such as the PRC1, PLK1, and RACGAP1 are indeed the top downregulated genes in both polychromatic and orthochromatic erythroblasts. Decreased mRNA levels of PRC1, PLK1, and RACGAP1 following SF3B1 knockdown were noted by RNA-seq analysis (Fig. 7a) and validated by real-time PCR (Fig. 7b). Importantly, treatment of late-stage erythroblasts by PLK1 inhibitor also resulted in the generation of abnormally nucleated erythroblasts (Fig. 7c). Quantitative analysis reveals that while < 10% of polychromatic and orthochromatic erythroblasts have abnormal nucleus in control group, PLK1 treatment increased the percentage of polychromatic/orthochromatic erythroblasts with abnormal nucleus to ~ 30% (Fig. 7d).