Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal stem cell disorders characterized by an extremely variable clinical course ranging from indolent disease with only minimal impact on survival of elderly patients to a very aggressive course with quick progression to acute leukemia [1]. On average, 30% of MDS patients develop overt acute myeloid leukemia (AML) during the course of the disease. The World Health Organization (WHO) classification identifies several MDS subtypes with a variable risk of AML transformation [1–3] (Table 1). There is a continuous search for the best model of individual risk-assessment for MDS patients. Several risk-related factors have been described, including clinical features, laboratory values, cytogenetics, bone marrow-related features, and recently established molecular markers [2, 4–8].

Table 1 Risk of AML transformation in myelodysplastic syndromes classified according to the WHO 2008 classification [2, 3]
Full size table

Mechanisms of disease progression and transformation from a chronic MDS phase to a more aggressive AML phase are still poorly understood. A mechanism based on an accumulation of cytogenetic and molecular aberrations during the course of disease, following a multi-step model of leukemogenesis, has been proposed (Fig. 1) [8]. However, most recently detected molecular changes occur only in a fraction of MDS patients and no universal molecular mechanism has as yet been found. The balance between apoptosis and proliferation within hematopoiesis, host-response related features, and stroma defects seem also to play a role.

Fig. 1
figure 1

Two-hit model of transformation from MDS to AML. Class I targeted genes are involved in signal transduction, while mutations in Class II genes affect transcription factors. Both Class I and Class II mutations can also lead to de novo AML [8]

Full size image

In the following review, we summarize current findings from clinical, laboratory, morphological, phenotypical, immunopathological, cytogenetic, and molecular research as related to the transformation from MDS to AML.

Prognostic scoring systems and assessment of risk for MDS–AML transformation

The International Prognostic Scoring System (IPSS), published in 1997, was based on the percentage of blasts in the bone marrow (BM) smear, number of cytopenias (defined as hemoglobin (Hb) levels less than 10 g/l, an absolute neutrophil count (ANC) of less than 1.5 × 109/L, and a platelet count of less than 100 × 109/L), and cytogenetic findings [5]. The IPSS could be applied to predict risk of AML transformation; 19% of low-risk patients, 30–33% of intermediate-risk, and 45% of high-risk patients died of AML in the primary cohort of Greenberg et al. [5] (Table 2). The clinical significance of IPSS has been verified in several studies [9–11] that also found IPSS to be of high predictive value for AML transformation [12]. Germing et al. [2] reported in 2005 that the addition of the lactate dehydrogenase (LDH) level to IPSS improved its prognostic value. MDS patients with elevated LDH had significantly higher risk of AML transformation. Malcovati et al. [6] proposed a prognostic scoring system based on the WHO2001 classification of MDS, cytogenetic findings, and transfusion requirements (WPSS). The WPPS defined five risk categories where very low-risk patients had only a 3% cumulative probability of AML transformation within 2 and 5 years, and for very high-risk patients, the respective probability was 80% and 84% (Table 2). The most recent MD Anderson prognostic scoring system is also based on the percentages of BM blasts and cytogenetics, but it does not take the WHO categories into consideration [7]. This system introduces other clinical data such as age, performance status, and transfusion requirements and applies a more detailed assessment of peripheral blood values, i.e., Hb, white blood cell count, and platelet levels. The MD Anderson system identifies low-risk patients in the IPSS high-risk group and patients with poorer survival within the IPPS low-risk group. However, no data concerning the assessment of the risk of transformation has been presented. Age ≥65 is considered as an adverse prognostic factor by the MD Anderson system. However, Nösslinger et al. [13] reported that within the IPSS high-risk category, women <65 years of age had the worst prognosis.

Table 2 Prognostic scoring systems and risk of AML transformation in MDS
Full size table

Morphological and immunophenotypic features related to disease progression in MDS

All above-mentioned prognostic scores include the number of BM blast evaluated in BM smears as a significant prognostic factor for AML transformation risk (Table 2). An arbitrary limit of 5% blasts has been used to define patients with refractory anemia with excess of blasts (RAEB) [1, 14]. A recent Chinese study indicated that refractory cytopenia with multilineage dysplasia (RCMD) patients with >3.5% blasts may have worse outcome [15]. Knipp et al. [16] reported the prognostic significance of increased peripheral blasts in patients with <5% BM blasts. Patients (16.2%) with at least 1% peripheral blasts suffered AML transformation by comparison to 6.8% of patients with less than 1% blasts in blood. Therefore, patients with ≥1% peripheral blasts have been included in the RAEB category in the 2008 WHO classification [1]. Several studies indicated that patients with multilineage dysplasia had a higher risk of AML transformation compared to those with unilineage dysplasia, which in part may be related to the higher incidence of multilineage dysplasia in patients with chromosomal changes [3, 17–19].

Aberrant phenotypes in BM blast population such as expression of lymphoid-associated markers CD7 (Fig. 2), CD4, and CD56 have been more often seen in MDS in transformation [20]. The aberrant expression of CD7 [20], overexpression of CD34 and CD36 [21], expression of Tdt [22], and the accumulation of phenotypic aberrancies, so called high flow score [23], have also been associated with the progression of disease in MDS.

Fig. 2
figure 2

Example of an emerging pathological blast population demonstrated by flow cytometry in a bone marrow from MDS patient. CD34+ cells are gated on CD34/Side scatter plot (upper right). Lower left plot shows that the population of CD34+ blasts partly displays an aberrant expression of CD7 (red dots). These cells can be found in the blast region (CD45 dim, left upper plot) and are only weakly CD33 positive (right lower plot). CD34+ CD7+ cells are not positive for CD56, but there is an increased CD56 expression in a subpopulation of granulopoietic precursors (right lower plot)

Full size image

The extent of blast infiltration has also been studied in BM biopsies by morphology and by CD34 immunohistochemistry (Fig. 3) that helps to evaluate topographic distribution of blasts [24–26]. Presence of abnormally localized immature precursors (ALIPs) and clusters of CD34 positive cells significantly increased risk of AML transformation, independent of IPSS. ALIPs and CD34 clusters were more often seen in MDS patients with RAEB. Causes of deaths in MDS patients who displayed CD34+ clusters were mostly related to leukemic transformation. The increased expression of CD34 in megakaryocytes was reported as a negative prognostic factor [27]. Clusters of CD117+ cells have also been described as ALIPs, but their significance in relation to AML transformation has not yet been determined [28].

Fig. 3
figure 3

Increased numbers and pathological distribution of CD34 cells in a bone marrow from MDS patient as shown by immunohistochemistry. Original magnification ×20

Full size image

Patients with high BM cellularity have a higher risk of transformation when compared to those with hypoplastic MDS [29]. Significant fibrosis detected by reticulin staining of BM biopsies from MDS patients also indicates an increased risk of AML transformation, independent of other variables [26, 30]. Literature data concerning a possible role of BM stroma-related factors in MDS–AML transformation is scarce. A variety of stromal defects may be present in various MDS subgroups, and some of these may facilitate transformation to AML [31]. Recent studies have shown that stromal factors can increase the susceptibility of stem cells to apoptosis in some MDS patients [32]. Fibroblasts derived from some MDS marrows, produced significantly higher levels of interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α). Macrophages from these MDS patients produced significantly higher levels of TNF-α than their normal counterparts [33]. Other MDS patients may have normal BM stroma. Since it has been demonstrated that following successful transplantation, fibroblasts (stroma cells) remain of host origin [34], MDS patients with normal BM stroma may have better chance to respond well to stem cell transplantation [35]. In MDS patients with del(5q), bone marrow stroma shows a decreased capacity to support hematopoiesis, which could be reversed by lenalidomide treatment [36].

Also, aberrant expression of vascular endothelial growth factor (VEGF) has been demonstrated in ALIPs of BM biopsies taken from MDS patients. It has been suggested that VEGF may provide signals reinforcing leukemia cell survival and this way contribute to transformation [37].

Cytogenetics

There is a consensus in the literature that complex karyotypes characterize a group of MDS patients with poor prognosis and a high-risk of transformation to AML [reviewed in [38]]. Most prognostic scoring systems consider MDS patients with three or more cytogenetic abnormalities within the same clone as having a complex karyotype while patients with five or more abnormalities are considered as a poor prognosis category in Medical Research Council AML trials. Within the group of MDS patients with complex karyotype, monosomal karyotype defined as the presence of two or more distinct autosomal monosomies or a single monosomy associated with a structural abnormality was related to poorer prognosis and shorter leukemia-free survival [39].

It has been shown that complex karyotypes arise by stepwise accumulation of chromosomal changes and that chromosomal instability of CD34+ population precedes transformation to AML [40]. In the study of Bernasconi et al. [41], 77% of sequentially tested MDS patients who show cytogenetic evolution progress to AML in comparison to 30% of those patients without cytogenetic evolution. In a recent large study of patients with 5q- abnormality, there was a clear correlation between karyotype complexity and the risk of transformation. Patients with isolated 5q- had a 21.1% cumulative probability of AML evolution after 5 years, and patients with 5q- + 1 abnormality had 57.6% probability. Patients with 5q- and two or more aberrancies had a 100% risk of transformation [42]. Also, it has been shown that patients with del(5q) MDS who failed to achieve sustained cytogenetic remission after treatment with lenalidomide have an increased risk of clonal evolution and AML transformation [43]. Most patients who developed AML acquired complex karyotypes.

There was no similar correlation for patients with abnormalities of chromosome 7, which at least in adults carry a high risk of AML transformation, independent of the presence of other abnormalities [38]. A high risk of AML development in these patients may be related to up-regulation of genes which have been implicated in leukemic transformation, such as HOX9A, BRCA2, PRAME, BMI-1, and PLAB. Other up-regulated genes included genes promoting cell proliferation, such as cell cycle regulator SPHAR, the DNA replication check-point gene Rad17, and signal transduction gene TPO [44].

Molecular mechanisms of transformation

Two classes of molecular changes play a role in the development of MDS and transformation to AML (Fig. 1). Class I targeted genes are mostly involved in signal transduction and class II molecular changes affect transcription factors (Table 3, reviewed in [8]). Class II mutations [such as mutations of Tet2, RUNX1, or RSP14 haploinsufficiency in del(5q)] affect cellular differentiation but are probably insufficient to induce AML transformation but play a role in MDS initiation [45, 46]. Class I mutations (such as Flt3, RAS, KIT, IDH1, and NPM1) affect cell proliferation and survival and have been mostly found in MDS patients at progression to AML [8]. When several mutations were investigated in a cohort of MDS patients and AML patients without a complex karyotype or balanced translocation, mutations in the ASXL1 and CBL genes were frequent in RAEB MDS patients. Mutations in the TET2 gene were found both in MDS and in AML and could be associated with either ASXL1 or NPM1 mutations but not with a RUNX1 mutation. The latter could be combined with mutations in the ASXL1 but not the NPM1 gene. Mutations in FLT3, IDH1, IDH2, NPM1, and WT1 were found mainly in AML patients [47]. However, mutations of each of the above-mentioned genes occur only in a relatively small fraction of MDS patients, which points out the heterogeneity of the disease.

Table 3 Some molecular changes implicated in increased risk of AML transformation in MDS patients
Full size table

Overexpression of the Wilms Tumor gene (WT1) has been reported in MDS [48]. Longitudinal studies have shown that an increase in WT1 mRNA levels in blood was a strong predictor of a short period of time to AML transformation, independent of IPSS [49]. Interestingly, patients with high levels of anti-WT1 antibody had a significantly longer survival.

TP53 mutations have been mainly reported in high-risk and therapy-related MDS and more often in patients with complex karyotypes [50, 51]. An increased expression of p53 protein as detected by the immunohistochemistry in myeloid cells in the BM of MDS patients has been shown to precede AML transformation [52] and to correlate with TP53 mutations [53]. In a recent study using sensitive deep-sequencing technology, small TP53 mutated populations (median clone size 11%) could be demonstrated already at the time of diagnosis in 18% of patients with del(5q) MDS category [54]. These mutations were present years before disease progression and were associated with an increased risk of transformation to AML with 5-year cumulative incidence of leukemic evolution: 77% and 24% in mutated and un-mutated groups of patients, respectively. In patients with TP53 mutations, small numbers of p53 strongly positive cells in the BM could be demonstrated at diagnosis and increased at the time of progression (Fig. 4), indicating the expansion of TP53 mutated population [54]. These results indicate that patients with early stages of MDS can harbor sub-clones that can be resistant to therapy and give rise to AML. The TP53 mutated patients did not achieve complete cytogenetic response to lenalidomide, which indicates therapy resistance. However, AML transformation occurred also in patients without TP53 mutation, suggesting other mechanisms of progression.

Fig. 4
figure 4

Single cells with p53 overexpression are demonstrated by immunohistochemistry in a bone marrow of a patient with del(5q) MDS at diagnosis (upper panel). Virtually all bone marrow cells are strongly p53 positive at the time of AML transformation (lower panel)

Full size image

Global gene expression studies combined with detailed annotated pathway analyses and gene oncology analyses have identified multiple deregulated pathways in CD34+ hematopoietic stem cells in MDS patients [55]. In general, the most down-regulated pathway in MDS was the Wnt canonical pathway, which may lead to defective self-renewal of hematopoietic stem cells. The most up-regulated pathways were the “interferon signaling” pathway, which may be responsible for cytopenia and the “thrombopoietin signaling” pathway, which may be one of the causes of megakaryocyte abnormalities and thrombocytopenia. Major differences were found between low-risk MDS cases where most deregulated pathways were related to immune response and apoptosis and high-risk MDS where the most deregulated pathways were related to cell cycle check and DNA repair [55]. The results of this study support a model for MDS in which immune deregulation and apoptosis dominates in early MDS leading to ineffective hamatopoiesis while disruption of DNA damage check-points and increased genomic instability characterize advanced MDS leading to AML transformation (Fig. 5).

Fig. 5
figure 5

Multiple mechanisms involved in transformation from MDS to AML

Full size image

Another sign of genetic instability in MDS is reported shortened telomere length, especially in patients with complex cytogenetic abnormalities and advanced disease [56, 57]. If telomeres reach a critically short length, they may form dicentric chromosomes and undergo a break-fusion-bridge cycle that may lead to further genomic changes. A recent study applying the novel technique of combining chromosome banding and T/C-FISH showed that in MDS patients, telomere lengths were shorter in both normal and aberrant metaphases [58]. These results suggest that telomere shortening may be characteristic of all hematopoietic stem cells in MDS or that telomere shortening may be a predisposing factor for development of MDS. In MDS patients who accumulate further molecular changes providing a proliferative advantage, clonal cells may stabilize the aberrations by up-regulation of telomerase activity or other telomere-elongating mechanisms [58].

Epigenetic mechanisms in MDS progression

Epigenetic changes have been recognized in the past decade as major drivers prompting malignant phenotypes [reviewed in [59]]. There are three general molecular mechanisms carrying epigenetic information: DNA methylation, histone modifications, and RNA interference. Aberrant DNA methylation is catalyzed by DNA methyltransferases and occurs within “CpG islands” found in the promoter regions of >50% of human genes. “CpG islands” are CpG dinucleotide-rich regions with lengths of 4 kb or more. Hypermethylation of CpG islands of tumor suppressor genes is probably a progressive process that can confer a selective advantage for the survival of the transformed cell [60]. In MDS, aberrant methylation was seen in every studied BM sample, and the number of methylated CpG sites was significantly greater in high-risk MDS and MDS/AML in comparison to low-risk MDS [61]. Five genes were hypermethylated in more that 70% of patients (ALOX12, GSTM1, HIC1, FZD9, and HS3ST2). Of those, the FZD9 gene methylation was an independent predictor of decreased survival. Clinical outcome was poorest in patients with chromosome 7 deletion and aberrant methylation of the remaining allele [61]. Another study reported that methylation of the CTNNA1 gene promoter was found in 31% patients with AML with del(5q) but not in low-risk del(5q) MDS[62]. The CTNNA1 gene has been suggested as one of the candidate 5q tumor suppressor genes. Thus, in cases of chromosomal deletion, the aberrant DNA methylation of the remaining chromosome may silence the remaining allele of a recessive tumor suppressor gene and increase the risk of transformation.

Another target for methylation analysis in MDS is the inhibitor of DNA binding/inhibitor of differentiation gene (ID). The ID proteins form hetero-dimers with transcription factors and act as the dominant negative inhibitors of gene transcription [63]. Methylation of ID4, which is a putative tumor suppressor gene, has been found in 35.1% of MDS patients and patients with methylated ID4 progressed to AML more rapidly that those without methylation[64].

Hypermethylation of the CDNK2B (p15 INKAB) gene found at diagnosis in patients with low-risk MDS has also been strongly associated with AML transformation [65]. In a recent study, a prognostic significance of a combination of methylation profiles of ten selected genes was evaluated and the high methylation score was shown to be an independent predictor of shorter progression-free survival [66].

Histones are small proteins that form a core around which DNA is wrapped, forming nucleosomes. The best understood histone modifications are acetylation and methylation of specific residues. Histone modifications form a code that integrates gene activation/inactivation/silencing signals, so that transcriptional activity of a given promoter can be predicted by looking at the specific histone modifications [59]. Specific studies of histone modifications in MDS are rare, but mutations in the EZH2 gene that codes for histone methyltrasferase at chr.7q36.1 have been described [67]. There are a variety of histone deacetylase (HDAC) inhibitors (HDIs) in clinical trials as well as some drugs that can inhibit the activity of histone methyltransferases [68, 69].

Apoptosis and proliferation

Several studies have shown increased rates of programmed cell death (PCD, apoptosis) in the BM of MDS patients [reviewed in [70]]. Increased apoptosis, which exceeds proliferation rate, is mainly considered as a characteristic of the early-stage disease, while disease progression is associated with a reduction in apoptosis [71]. Several cytokines have been shown to be overexpressed in MDS, including TNF-α, transforming growth factor (TGF)-β, interferon (IFN)-γ, IL-6, and IL-1β [70]. TNF-α may be produced by macrophages [72] and the levels of TNF-α in BM plasma have been shown to correlate with pro-apoptotic Fas expression and apoptosis [73, 74]. TNF-α levels were lower in MDS patients with RAEB than in RA/RA-RS, which suggests a decreased role of TNF-α at progression. Other overexpressed molecules acting as negative regulators of hematopoiesis in MDS include FAS-ligand and TRAIL with their respective agonistic receptors [70]. At progression, there is a shift of balance in the signals and antiapoptotic/pro-proliferative signals prevail, including increased expression of Bcl-2 (Fig. 5) [71, 75].

Moreover, CD34+ BM cells from high-risk MDS patients or at progression showed higher levels of NF-κB compared to the early-stage of the disease. High NF-κB activity could lead to up-regulation of FLIP (FLICE-inhibitory protein) and other NF-κB dependent antiapoptotic regulators (such as Bcl-xL, Bcl-2, XIAP) followed by increased resistance of CD34+ cells to apoptosis in these patients [76].

Autoimmunity and host response

Several observations suggest that immune dysregulation plays a role in the pathogenesis of low-risk MDS and contributes to ineffective hematopoiesis in these patients. A proposed model of immune pathogenesis implied that CD8+ cells are stimulated by unique or overexpressed antigens, which leads to T cell receptor repertoire contraction through expansion of memory cells and repression of hematopoiesis trough cross-reactive antigens expressed on normal BM progenitors [77]. Both CD4 and CD8 subsets seem to be dysregulated in MDS. In younger MDS patients, reduced levels of naïve CD4+ T cells were associated with response to immunosuppressive therapy [78]. It has been suggested that loss of CD4+ cells in MDS may affect mainly regulatory T cells (T-regs), which are important for peripheral tolerance and prevention of an autoimmune process, while the numbers of CD3+ CD4+ IL-17 producing T cells were increased [79]. The levels of T regs in blood of high-risk MDS patients were higher than in the low-risk group and the levels further increased in patients who progressed to leukemia [80]. Also T-regs in patients with advanced MDS retained their functional capacity in contrast to T-regs from low-risk MDS patients that had a reduced ability to suppress immune responses. Moreover, patients with high-risk MDS have reduced natural killer (NK) cell function and reduced expression of activating NK receptors [81]. Thus, in high-risk MDS patients, impaired immune surveillance may contribute to progression and leukemic evolution.

Conclusions

Although great progress have been made in understanding molecular pathogenesis of MDS, the exact defects that make some patients quickly transform to aggressive and usually therapy-resistant AML phase are often not clear and are probably complex. Continuous research will bring more insight in the pathogenesis of various MDS categories, making individual risk assessment and tailored therapy for each patient possible.