Early studies investigating the properties of MDS-derived BM-MSCs were conducted in ex vivo expanded cells, as their in situ counterparts represent only a minor fraction of BM nucleated cells (16). Published data are at times conflicting and this might be due to the variability of experimental approaches and to patient heterogeneity (16). Within this context, cultured MDS-derived BM-MSCs have been reported to exhibit an irregular morphology ( Table 1 ) (17–22, 40), whereas other studies ( Table 1 ) found no morphological abnormalities in patient BM-MSCs as compared to their normal counterparts (23–30). Moreover, most studies ( Table 1 ) suggest that patient BM-MSCs do not differ in terms of immunophenotype from MSCs derived from healthy donors (25–30), although reduced expression of CD90, CD104 and CD105 has also been observed (17, 41).

A number of studies has demonstrated that MDS-derived BM-MSCs display defective growth potential (17–19, 21, 24, 30–32, 42, 43). This has been correlated with decreased expression of CD44 and CD49e (32) and at least in some cases it has been associated with increased cellular senescence (18, 19, 33) ( Table 1 ). The impaired proliferative capacity of MDS-derived BM-MSCs was corroborated by our study as well (20) and it was suggested that this could be attributed to the decreased expression of the canonical WNT signaling pathway and the concomitant up-regulation of the non-canonical pathway. More recently, Falconi et al., confirmed the down-regulation of the canonical WNT signaling pathway in patient-derived BM-MSCs (40).

As far as the differentiation potential of MDS-derived BM-MSCs is concerned, data are contradictory and these discrepancies might be explained by diversities in patient categories and methodologies across different studies. On one hand, there have been studies demonstrating that these cells do no differ in their capacity to differentiate towards osteoblasts (20, 27, 30, 31, 34), adipocytes (19, 20, 27, 30, 31) and chondrocytes (19, 27, 31) as compared with BM-MSCs derived from healthy donors ( Table 1 ). In contrast, reduced osteogenic (19, 21, 24), adipogenic (21) and chondrogenic differentiation (17) have also been reported ( Table 1 ). As regards the deregylated osteogenic capacity of BM-MSCs derived from MDS patients, Geyh et al. have demonstrated that it could be partly triggered by TGFβ1 (44). Furthermore, in line with the impaired in vitro osteogenic differentiation potential of patients-derived BM-MSCs, an early study on transiliac bone biopsies obtained from MDS patients demonstrated abnormalities in bone remodeling consisting of decreased number of osteoclasts and osteoblasts and decreased bone formation as evidenced by the diminished mineral apposition rate (45). Bone loss in individuals suffering from MDS was also shown in a more recent report (46), which provided evidence that osteoporosis was more prevalent in patients as compared with age-matched controls.

The question whether MDS-derived BM-MSCs can effectively support hematopoiesis has been addressed in vitro by co-culturing them with HSPCs. Results are inconsistent and this may due to patient heterogeneity and differences in the experimental protocols in between studies. Within this context, some authors have suggested that patient BM-MSCs are able to sustain the growth of both autologous/leukemic HSPCs (30, 34) and HSPCs derived from healthy individuals (27) ( Table 1 ). Muntion et al. (35), ( Table 1 ) have shown that exosomes may be involved in the crosstalk between patient-derived BM-MSCs and hematopoietic cells. More precisely ex-vivo expanded BM-MSCs from lower risk MDS patients have been demonstrated to secrete exosomes with a different microRNA cargo than their normal counterparts. Patient BM-MSC-derived exosomes are incorporated into CD34⁺ cells from healthy donors and alter their gene expression via microRNA transfer (such as miR-10a and miR-15a) and increase their clonogenic potential and viability. Whether BM-MSC derived exosomes similarly support clonal hematopoiesis in the MDS setting remains to be seen. On the other hand, many studies have demonstrated that MDS-derived BM-MSCs have an impaired potential to support normal HSPCs (18–21, 24, 29, 31, 44)( Table 1 ). This could be attributed to TGFβ1 (44) as well as to the defective expression of niche-derived molecules known to be involved in hematopoiesis, such as osteopontin, angiopoietin, jagged-1, kit ligand, hepatocyte growth factor (HGF), C-X-C motif chemokine ligand 12 (CXCL12), insulin like growth factor-1 (IGF1), insulin growth factor binding protein 2 (IGFBP2), thrombopoietin (TPO) (18, 19, 29). In addition, in CD271⁺ BM-MSCs from LR-MDS the up-regulation of inflammatory factors and inhibitors of hematopoiesis such as interleukin 6 (IL-6), IL-8 and C-C motif chemokine ligand (CCL3) ( 21, 36, 47), the transcription of which is increased secondary to activated NF-kB signaling (36), has been associated with attenuation of HSPC numbers and function ex vivo. Finally, CD73⁺ MDS-derived BM-MSCs were shown to have a negative impact on the clonogenic potential of autologous hematopoietic cells, as compared to the effect of CD73⁺ BM-MSCs derived from healthy donors on normal hematopoietic cells (48). This was associated with the increased expression of focal adhesion kinase, a protein involved in various cellular processes including survival, proliferation, differentiation and adhesion (49), in patient-derived BM-MSCs (48).

It has been widely acknowledged that BM-MSCs possess broad immunoregulatory properties, involving cells associated with both innate and adaptive immunity (reviewed in (50)). A large body of evidence supports the role of immune abnormalities in MDS pathogenesis (reviewed in (51)) and accumulating data suggests that BM-MSCs may contribute herein ( Table 1 ). More precisely, BM-MSCs from a specific MDS subtype (refractory anemia) deficiently inhibit in vitro T cell activation and proliferation (25). Another study has demonstrated that BM-MSCs derived from lower risk MDS patients differ in terms of immunoregulatory properties as compared to those from higher MDS patients (28). More specifically, the former were associated with a lower T cell apoptosis, a less potent inhibitory effect and a lower T-regulatory cell inducible rate. In addition, BM-MSCs from lower risk MDS patients exhibit decreased capacity to inhibit dendritic cell maturation and proliferation as compared to those derived from higher MDS patients (37). Moreover, MDS-derived BM-MSCs, but not those derived from healthy donors, have been reported to induce naive normal monocytes to acquire the properties of myeloid derived suppressor cells and eventually down-regulate NK and T cell function (38). Taken together these findings suggest that MDS-derived BM MSCs have impaired immuneregulatory functions. The fact that in our study (31) BM-MSCs from MDS patients effectively inhibited T cell proliferation might in part be attributed to differences in patient distribution within the MDS subtypes, along with dissimilar experimental protocols as compared to other studies.

Whereas clonal cytogenetic abnormalities in hematopoietic cells are detected in approximately 40-70% of patients with de novo MDS, data regarding the presence of genetic aberrations in MDS-derived BM MSCs have been contradictory. While several studies (17, 27, 52), including ours (53) ( Table 1 ) have shown that patient ex-vivo expanded BM-MSCs harbor chromosomal aberrations, others report that these cells are normal in terms of cytogenetic analysis (25, 29, 34) ( Table 1 ). Interestingly, clonal chromosomal abnormalities detected in MDS-derived BM-MSCs consistently differ from those in hematopoietic cells from the same individual (53), thereby suggesting that patient-derived BM-MSCs and hematopoietic cells do not derive from the same clone. In support of this notion, Fabiani et al. (54) reported that mutations of epigenetic and spliceosomal genes in the BM-mononuclear cells from MDS patients were not present in the mesenchymal compartment.

As some chromosomal aberrations in MDS-derived BM-MSCs were not detected in earlier passages, but only in later passages and as cytogenetic abnormalities were also found in cultured BM-MSCs derived from healthy donors (53), the possibility that these alterations occurred due to ex-vivo expansion and/or culture conditions could not be excluded. This issue was clarified in a recent study (33) ( Table 1 ) investigating the occurrence of clonal mutations in patient and normal ex vivo expanded BM-MSCs by exome sequencing. MDS-derived BM-MSCs were shown to harbor increased mutational burden and distinct mutational signatures as compared to healthy BM-MSCs. However, highly recurrent mutations identified during culture could not be backtracked in primary -non expanded- stroma cells from the same patients (33). These findings suggest that there is no evidence for clonal mutations in the stroma compartment of MDS patients and that the mutations detected during ex vivo expansion of MDS-derived BM-MSCs are related to in vitro culture per se.

Conclusive evidence for the key role of the mesenchymal components of the BM microenvironment in the emergence of MDS-like disease in mice was firstly reported by Raaijmakers et al. (55). In this pivotal study, the authors demonstrated that targeted deletion of Dicer1 -an RNAse III endonuclease involved in miRNA biogenesis- from murine osterix-expressing osteoprogenitors, but not terminally differentiated osteoblasts, resulted in defective osteoblast differentiation and in the initiation and propagation of a form of myelodysplasia associated with leukopenia and lymphocytopenia. Osteolineage cells from mutant mice expressed significantly lower levels of the Schwachman-Bodian-Diamond syndrome (Sbds) gene (55). Mutations of this gene are found in Schwachman-Diamond syndrome (SDS) which is characterized by exocrine pancreatic dysfunction, cytopenias- especially neutropenia and bone abnormalities (56). Of note, patients with SDS are at risk for developing MDS and AML (56). Sbds deletion in murine osteoprogenitors resulted in leukopenia, lymphopenia, and myelodysplasia, thus recapitulating the phenotype of mice harboring Dicer1 deletion within the same cells (55). Consistent with these findings, reduced expression of Dicer1 mRNA and protein levels as well down-regulated Sbds gene expression were also reported in BM-MSCs from MDS patients (57). On the other hand, a recent study showed that none out of 121 individuals with germline pathogenic Dicer1 variants developed MDS or leukemia (58). Thus, the clinical significance of Dicer1 in the MDS setting remains elusive. In addition Sbds deletion in osterix+ mesenchymal stem/progenitor cells induces genotoxic stress in HSPCs via the inflammatory p53-S100A8/9-TLR signaling (59). Activation of this axis seems to have clinical relevance as it was also observed in a subset of lower risk MDS patients who were characterized by a significantly shorter progression free survival and leukemic evolution (59).

The NUP98 gene encodes a protein that is involved in RNA and protein transport across the nuclear membrane (60). NUP98 fusion genes have been identified in various hematologic malignancies, including MDS. The NUP98-HOXD13 fusion gene has been detected in patients with MDS (60) and mice expressing the hematopoietic compartment specific vav-driven NUP98-HOXD13 fusion transgene eventually develop MDS-like features (46, 60). Interestingly, these mice exhibit also a disrupted microenvironment, which has been reported to contribute significantly in disease progression (46, 61). Alterations in the BM microenvironment include increased number of osteoblasts, reduced numbers of osteoclasts and increased amount of non-mineralized bone (46). Notably, neutralization of the increased serum levels of fibroblast growth factor 23 (FGF23), a regulator of phosphate homeostasis and bone mineralization and an inhibitor of erythropoiesis, restored bone microarchitecture and improved osteoid mineralization and anemia (61). Elevated FGF23 levels and increased amount of non mineralized bone were also observed in samples from MDS patients, thereby corroborating the findings in NHD13 mice (61)

Impaired bone metabolism was also reported recently in the Abcg2 MDS/AML mouse model (62). In this model, expression of a mutant enhancer of zeste homolog 2 (EZH2-the catalytic subunit of polycomp repressive complex 2, one of the 10 most frequently mutated genes in MDS (63)) lacking the catalytic SET domain (EZH2-dSET) induces Abcg2 (a drug efflux transporter) up-regulation. Mice transplanted with syngeneic bone marrow cells retrovirally tranduced with EZH2-dSET developed an MDS-like disease (63). These mice were also shown to have reduced bone volume, due to decreased bone formation. Similar findings were observed in NHD13 mice and in MDS patients as well. Furthermore, the authors demonstrated that murine MDS/AML cells inhibited the commitment of BM-MSCs towards osteoprogenitors, thereby deregulating the capacity of MSCs to support normal hematopoiesis. This suppression was mediated via extracellular vesicles derived from MDS/AML cells (62). Collectively, this data add to the established notion that malignant hematopoietic cells are able to modify MSCs, so that the latter promote disease propagation and evolution.

The Apc^del/+ MDS mouse model is a conditional knock-out mouse based on the Mx1-Cre system the most commonly used system to delete the gene of interest in experimental hematology (64–66). Apc is a crucial negative regulator of the canonical b-catenin (Ctnnb1)/WNT-pathway. In the Apc^del/+ MDS mouse model the deletion of 1 Apc allele occurs not only in hematopoietic cells but also in BM stromal cells (65, 66). Apc haploinsufficiency in the mouse BM niche induces the development of myelodysplasia, which is characterized by severe macrocytic anemia (65). Interestingly, MDS is mediated via aberrant WNT signaling in the BM microenvironment (66). In support of this notion drug inhibition of the WNT/Ctnnb1 pathway by the anthelminthic agent pyrvinium delays and/or inhibits MDS development in Apc ^del/+ mice (66).

As far as human MDS is concerned, in vivo evidence for the facilitating role of BM-MSCs was provided by a study using a xenograft model (67). More specifically, co-transplantation of CD34⁺ cells derived from lower risk MDS patients along with ex vivo expanded BM-MSCs from the same patients in NOD/LtSzcid-IL2rg^-/- (NSG) mice significantly increased engraftment of hematopoietic cells as compared to transplantation in the absence of BM-MSCs or with BM-MSCs derived from healthy donors. Furthermore, ex vivo expanded MDS-derived BM-MSCs were shown to differ from their normal counterparts regarding expression of genes related with osteogenesis, adipogenesis, fibrosis, inflammation, cell adhesion, extracellular matrix remodeling and cytokine signaling. Of note, normal MSCs acquired similar properties upon co-culture with patient-derived bone marrow cells, substantiating the issue of hematopoietic MDS cells inducing niche alterations, favoring malignant cell expansion and disease evolution (67). However, in contrast to the aforementioned xenograft model, femoral co-transplantation of BM mononuclear cells derived from MDS patients in combination with autologous or allogeneic patient-derived BM-MSCs or BM-MSCs derived from healthy individuals in NSG mice or in NSG mice with knock-in for human stem cell factor (SCF), granulocyte-macrophage colony-stimulating factor (GM-CSF), and IL-3, did not enhance MDS engraftment (68). The discrepancy between the two studies is not clear. Yet, the fact that in the latter study injected MSCs did not exhibit long-term engraftment suggests that human microenvironment was only transiently established, which might explain -at least in part- the lack of influence of MSCs on MDS engraftment

The contribution of human MDS-derived BM-MSCs for the engraftment and maintenance of MDS HSCs in mice was also corroborated in a recent study by Mian et al. (69). The authors injected patient derived HSCs into gelatin-based scaffolds that were previously seeded with autologous or allogeneic MDS-derived BM-MSCs and the construct was subsequently transplanted in mice. Engraftment of MDS HSCs was achieved in 94% of cases whereas persistent long-term engraftment within scaffolds was observed in 0.2-86%. Interestingly, patient-derived HSCs were shown to move out of the scaffold and home to an adjacent scaffold previously seeded with human BM-MSCs but not with murine BM- MSCs (69), thereby highlighting the critical interplay between bone marrow microenvironment and malignant cells in the MDS setting.

The aforementioned preclinical studies support the critical contribution of the BM-MSC compartment in the initiation and/or propagation of MDS and provide the rationale for its therapeutic targeting as a strategy to delay and/or halt disease evolution. In line with this notion, the beneficial effects of some of the therapeutic modalities currently applied in MDS may be exerted via amelioration of the BM-MSCs’ impairment, as will be outlined below ( Table 2 ).

Iron overload (IO) is a common finding in MDS patients that occurs as a result of ineffective hematopoiesis as well as red blood cell transfusions (76, 77). Aside from its deleterious effect on hematopoiesis (76, 77) IO has also been demonstrated to impair BM-MSCs. More precisely, Huang et al. (78) have demonstrated that BM-MSCs from iron overloaded higher-risk MDS patients exhibit decreased quantity, defective proliferation capacity and reduced osteogenic differentiation potential. Furthermore, they express lower levels of hematopoiesis-associated genes such as VEGFA, CXCL12, and TGFβ1 ( 78). In addition IO induces apoptosis in MDS-derived BM-MSCs via increased level of reactive oxygen species (ROS) and the ROS-associated Wnt/b-catenin pathway (78). IO has also been reported to induce mitochondrial fragmentation via ROS and the activation of the AMPK/MFF/Drp1 pathway in MDS-derived BM-MSCs. Of note, all these effects of IO in patient BM-MSCs are reversed, at least in part, by antioxidants and iron chelation. Moreover, in an IO mouse model, antioxidant and iron chelation have been shown to partially restore the defective BM-MSCs’ hematopoietic support. Iron chelation is commonly administered in MDS patients, especially in those with lower-risk MDS and its treatment is associated with improvement of cytopenias in a significant proportion of patients (70). The aforementioned data imply that in the MDS setting the beneficial role of iron chelation in hematopoiesis may also result from the restoration, to a certain extent, of the MSCs deficits and further support the application of this treatment in myelodysplasia.

On the other hand Hu et al. (71) have reported that IO upregulates the expression of IL-6, IL-8, TGFβ and VEGF in MDS-derived BM-MSCs through ROS upregulation and subsequent HIF-1a overexpression (71). Antioxidants and iron chelation down-regulated the levels of the aforementioned cytokines. These data provide the theoretical background for probing more deeply into the role of HIF-1a in the BM microenvironment in MDS and provide further justification for the use of HIF-1 inhibitors in this disorder, an issue that is currently under investigation (79). Of note, the discrepancies in terms of the effect of IO in the expression of VEGF and TGFβ between the study Huang et al. (56) and that of Hu et al. (71) may reflect differences in patient cohorts regarding disease categorization and IPSS risk classification.

Previous studies have demonstrated that ex vivo expanded MDS-derived BM-MSCs have significantly different DNA methylation patterns from their normal counterparts (19, 21)and are characterized by aberrant hypermethylation (19, 39). These findings set the stage for the investigation of the potential effect on patient MSCs of the hypomethylating agent azacitidine (AZA), which is widely used in higher risk MDS and may also be effective in lower risk MDS refractory to first line treatments (80, 81). AZA treatment has thus been shown to restore the aberrant hypermethylation pattern of MDS-derived BM-MSCs (39) to increase their proliferation potential and osteogenic capacity, and improve their ability to support HSPCs for in vivo engraftment (21). Another study (72) has reported that in patient BM-MSCs AZA regulates various genes involved in the support of hematopoiesis, essentially genes associated with IFN-γ and extracellular matrix receptor interaction pathways. This transcriptional modulation could account for the improved capacity of AZA-treated MDS-derived BM-MSCs to preferentially support healthy-derived HSPCs, over patient-derived HSPCs (72). AZA has also been shown to upregulate the expression of the serine protease inhibitor kunitz-type2 (SPINT2/HAI-2), an inhibitor of hepatocyte growth factor (HGF) activation, in patient-derived BM-MSCs (73). Based on data derived from the HS-5 stromal line, it has been hypothesized that AZA may down-regulate HGF and CXL12 production by MDS-MSCs and consequently increase cell-adhesion, proliferation and survival of MDS-derived HSPCs (73). The aforementioned findings provide evidence for the potential of AZA in targeting the defective BM-MSCs, thereby extending the effects of this therapeutic modality in the MDS-setting.

Decitabine is another hypomethylating agent which has been approved by the Food and Drug Administration (FDA) for the treatment of patients with MDS (82). Incubation of culture expanded BM-MSCs from MDS patients in the presence of Decitabine resulted in a significant decrease in the proportion of cells in the G0/G1 phases as compared to MDS-derived BM-MSCs incubated in the absence of the drug (control group). This effect was associated with a reduced gene expression of cyclin dependent kinase inhibitor 1A (CDKN1A) ( 74). Moreover the ability of BM-MSCs from MDS patients incubated in the presence of decitabine to induce the differentiation of T cells into Tregs was significantly reduced compared with control BM-MSCs and this was linked to decreased gene expression of programmed death-ligand 1 (PDL1) (74). Taken together these findings suggest that decitabine may improve, at least to a certain extent, the impaired properties of MDS-derived BM-MSCs

Lenalidomide is an immunomodulatory agent showing efficacy in lower risk MDS (80, 81). Of note, apart from acting directly on MDS cells, lenalidomide has been demonstrated to target the BM microenvironment as well (83), including BM-MSCs (18). More precisely lenalidomide reduces CXCL12 secretion by MDS-derived BM-MSCs and this has been hypothesized to induce the egress of dormant MDS cells from their niches and render them more sensitive to the drug. In addition, treatment of ex vivo expanded lower risk MDS-derived BM-MSCs with lenalidomide has been reported to improve their capacity to support normal clonogenic HSPCs (18).

The downstream key mediators of TGF-β superfamily signaling, SMAD2/3 are constitutively activated in MDS CD34⁺ cells and this has been associated with impaired late-stage erythroid maturation and subsequent anemia (84, 85). Luspatercept is a therapeutic agent that sequesters TGF-β superfamily ligands and consequently decreases SMAD2/3 activation thereby restoring erythroid maturation and improving anemia. It has been approved for the treatment of selected patient with lower risk MDS (84, 85). Interestingly, a recent study has provided evidence that luspatercept affects also the BM-MSCs (75). More precisely, it decreases SMAD2/3 activation in ex-vivo expanded MDS-derived BM-MSCs and increases CXCL12 secretion. Furthermore, pre-treatment of MDS-derived BM-MSCs with luspatercept improved the clonogenic potential of co-cultured HSPCs, their adherence and expression of CXCR4, the CXCL12 receptor, as well as their homing in zebrafish embryos. Finally, BM-MSCs derived from patients treated with luspatercept exhibited a greater capacity to sustain the clonogenic potential of normal, but not MDS-derived HSPCs (75).

As previously mentioned, MDS-derived hematopoietic cells have been demonstrated to impair BM-MSCs (62, 67). Therefore, it is tempting to speculate that the therapeutic effect of allogeneic HSPC transplantation, which remains the only curative option in MDS (80), may also be exerted via favorable modulations of the functions and properties of the mesenchymal compartment. This issue requires further investigation by comparative analyses on the characteristics of MDS-derived BM-MSCs prior and post allogeneic BM transplantation. To the best of our knowledge, such studies have not been reported thus far.