To the Editor

Myelodysplastic syndrome (MDS) with myelofibrosis (MDS-MF) accounts for up to 50 % of cases of MDS [1], and differs from MDS without MF in terms of clinical performance, treatment tolerance, and survival [2]. MDS with severe MF (MF grade = 2–3) is considered as an independent risk factor for de novo MDS [3]; however, MDS-MF is not independently listed as a subtype of MDS according to the 2016 World Health Organization classification, indicating the need for further investigation.

We retrospectively enrolled 53 patients with MDS-MF (44 MF grade 1/MF1; 9 MF2 − 3) and 31 patients with de novo MDS without MF (MDS). There were no significant differences among the MDS, MDS-MF1, and MDS-MF2 − 3 groups in terms of age, sex, MDS subtypes, IPSS risk levels, and treatment strategies, except for a higher rate of poor karyotypes in the MDS-MF2 − 3 compared with the MDS and MDS-MF1 groups (with no difference between the MDS and MDS-MF1 groups) (Supplementary Table 1). Not only MDS-MF2 − 3, but also MDS-MF1 at IPSS low/int-1 risk, had a shorter leukemic transformation time compared to the MDS group (Fig. 1), suggesting that patients with low/int-1 MDS with even mild MF required chemotherapy to avoid disease progression. Detailed patients’ clinical data are shown in Supplementary file 1.

Fig. 1
figure 1

Leukemic-transformation time (LTT) curves. A LTT for all patients. Median LTT: 71.5 for MDS; 31.9 months for MDS-MF1; 5.6 months for MDS-MF2 − 3; MDS-MF2 − 3 vs. MDS p < 0.001; MDS-MF1 vs. MDS p = 0.045; MDS-MF2 − 3 vs. MDS-MF1p = 0.001. B LTT according to International Prognostic Scoring System (IPSS) risk categories. In the IPSS low/int-1 risk groups, median LTT: 71.5 months for MDS; 40.4 months for MDS-MF1; 2.8 months for MDS-MF2 − 3; MDS vs. MDS-MF1p = 0.029; MDS vs. MDS-MF2 − 3p < 0.001; MDS-MF2 − 3 vs. MDS-MF1p = 0.001. In the IPSS int-2/high risk groups, median LTT: 27.2 months for MDS; 17.8 months for MDS-MF1; 5.6 months for MDS-MF2 − 3; MDS vs. MDS-MF1p = 0.323; MDS vs. MDS-MF2 − 3p = 0.024; MDS-MF2 − 3 vs. MDS-MF1p = 0.131. Five Patients with IPSS int-2/high risk who refused chemotherapy; seven patients who received allogeneic-hematopoietic stem cell transplantation (4 MDS; 3 MDS-MF1) were excluded from LTT analysis. Two-tailed p-value < 0.05 by log-rank test was considered statistically significant. ^Another five patients who did not have cytogenetic analysis results (1 MDS; 3 MDS-MF1; 1 MDS-MF2 − 3) were excluded from IPSS-based LTT analysis

Full size image

Cytogenetic abnormalities or genomic mutations are related to the leukemic transformation of MDS-MF but cannot account for the role of MF in this process. The leukemia clonal revolution is influenced by many factors in the bone marrow (BM) microenvironment, including inflammation and abnormal immunity. Chronic inflammation is believed to promote malignant hematopoiesis in myeloproliferative neoplasms through pro-inflammatory /fibrogenic /angiogenic cytokines [4,5,6]. Reactive oxygen species also play a major role in tumor progression of myeloproliferative neoplasms [7]. However, the processes responsible for the clonal myeloproliferation of MDS-MF remain unclear. It is therefore necessary to determine the variations in gene expression levels responsible for the leukemic transformation of MDS-MF. We conducted single-cell sequencing of BM mononuclear cells (BMMCs) from a patient with MDS-MF2 − 3 (CN) in the MDS phase (CN1) and leukemic phase (CN2). A healthy donor (NC) and a patient with de novo acute myeloid leukemia (AML) with the same FAB subtype (M2) were used as controls. Patients’ clinical information and detailed methods are provided in Supplementary File 1. Cell clusters were identified based on 13,280 healthy cells and BMMCs from 40 patients with newly diagnosed AML in our previous study [8] and were checked with the Human Cell Landscape (http://bis.zju.edu.cn/HCL/index.html) established by our institute [9]. Marker genes for the cell clusters are listed in Supplementary Table 2.

Seventeen cell clusters and eight cell types were identified (Fig. 2A,B). We used “blast-like” cells for the differentially expressed genes (DEGs) and gene enrichment analyses to avoid the influences of other cell types. The top 20 DEGs and their reported functions are listed in Supplementary Table 3. The top 20 increased genes during leukemic transformation included some AML-related genes, such as CD52, SRGN, BEX1, BASP1, SPINK2, NEAT1, and CEACAM6, and some proinflammatory mediators, such as S100 family genes, RNASE3, and CYBB. In contrast, expression levels of many ribosomal protein genes were decreased (Fig. 2C). The S100 family comprises proinflammatory mediators associated with acute and chronic inflammation and neoplasm metastasis [10]. CYBB can produce superoxide, and trigger mitochondria transfer to stimulate BM stromal cells to form AML blast cells [11]. RNASE3 participates in nucleolysis, cell binding, lipid instability, cytotoxicity, and antibacterial activity [12]. However, none of these genes have previously been linked to leukemic transformation of MDS-MF. Gene enrichment analysis revealed that the upregulated pathways during leukemic transformation mainly contributed to inflammation /oxidation /energy metabolism-related signaling and tumor-related pathways (Fig. 2D). We also explored the DEGs between CN2 and M2 samples to detect the differences between secondary AML and de novo AML (Fig. 2E). Genes with higher expression in CN2 were significantly enriched in leukocyte transendothelial migration and the Rap1 signaling pathway (Fig. 2F). Leukocyte transendothelial migration is an inflammation pathway, while Rap1 acts as a molecular switch involved in many biological processes. S100A12, RNASE3, and CYBB were among the genes with higher expression levels in both CN2 compared with CN1, and in CN2 compared with M2 (Fig. 2G).

Fig. 2
figure 2

Distributions of cell clusters and single-cell gene expression patterns of blast-like clusters between CN1 and CN2, CN2 and M2. A Overall t-distributed stochastic neighbor-embedding analysis of bone marrow mononuclear cells from the four samples (NC, CN1, CN2, M2). Clusters indicated by different colors and numbers; cell types indicated by different colors of loops. B Overall t-distributed stochastic neighbor-embedding analysis of bone marrow mononuclear cells from the four samples. Samples were indicated by different colors and numbers. C Volcano plot of differentially expressed genes (DEGs) in blast-like cells between CN1 and CN2. The dots on the left represent higher expressed genes in CN2 and those on right represent lower in CN2. D Metascape Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of enriched terms of DEGs in blast-like cell clusters between CN1 and CN2. Color shows p value. E Volcano plot of DEGs in blast-like cells between CN2 and M2. The dots on the left represent higher expressed genes in CN2 and those on right represent lower in CN2. F Metascape KEGG pathway analysis of enriched terms of DEGs in blast-like cell clusters between CN2 and M2. Color shows p value. G Heat map of DEGs among CN1, CN2, and M2 in blast-like cells. Yellow: higher expression level; red: lower expression level. Samples labeled in different colors. CN: Initial of the MDS-MF2 − 3 patient for single-cell sequencing; CN1: The patient CN at her MDS phase; CN2: The patient CN at her leukemic phase; NC: Normal control; M2: The patient with de novo AML-M2

Full size image

In conclusion, this study revealed that MDS-MF with even mild MF had a higher risk of leukemic transformation than MDS without MF, suggesting that MDS-MF should have a different risk classification algorithm and may need special treatment. Inflammatory and oxidation activation may be essential processes, while S100 family genes, RNASE3, and CYBB might be key genes involved in the leukemic transformation of MDS-MF.