Bone marrow (BM) niches are specialized microenvironments within bones where supportive cells forming the cellular niche allow the maintenance and differentiation of haematopoietic and mesenchymal stem cells (HSCs and MSCs, respectively) (1, 2). The cellular characterization of these complex microenvironments has been achieved mainly by means of elaborated genetic approaches targeting selected candidate factors, despite limitations of specificity inherent to this strategy, resulting in controversial conclusions. More recently, the combined application of state-of-the-art technologies including high resolution imaging, single-cell RNA sequencing (scRNAseq) and spatially resolved transcriptomics has led to an unprecedented insight into the molecular, cellular, and spatial organization of BM niches, and hierarchical structures and differentiation trajectories therein (3–6). On this basis, sinusoidal, arteriolar and endosteal niches appear to be distinguished not only by their specific location, but also by their unique cellular composition and molecular requirements. Moreover, increasing evidence demonstrate that alterations at different levels in the niche composition are associated with malignancies including myelodysplastic syndromes (MDS) (7) and myeloproliferative neoplasms (MPN) (8), with an additional effect on the osteo-haematopoietic niche deriving from medical treatments (9, 10). Remarkable efforts have been devoted also to translating this basic knowledge, mainly derived from murine models, into the development of relevant in vitro platforms to study the human counterpart and to test drugs in a patient-specific setting (11, 12). As a perspective, these tools may also serve to test the inherent therapeutic potential of the various niche components.

This expanding field has been covered by several comprehensive reviews. Here we will give an overview of the interplay between BM niches and tumour cells (TCs), focus preferentially on very last papers, and highlight therapeutic implications and perspectives of this mutual relationship, for the benefit of a wide audience ranging from neophytes to experts.

MSC and HSC lineage cells are responsible for the high dynamism of the bone tissue, and shape the BM niches, spatially defined microenvironments classified as endosteal and subendosteal (based on the distance from the inner bone surface, particularly at the metaphyseal spongiosa), arteriolar (close to the main vessels carrying blood into the BM), sinusoidal (next to vessels carrying blood out of the BM and forming a wide network within the BMME) and non-vascular (6), based on their location. Overall, niche cell composition is varied, with a distinct array of components in a specific spatial location. Osteoblasts (OBs), chondrocytes and endosteal fibroblasts are present only in the endosteal niche; arteriolar endothelial cells (ECs), smooth muscle cells, arteriolar fibroblasts localize to the arteriolar niche; a newly identified subset of CXCL12-abundant reticular (CAR) cells called Osteo-CARs, displaying a high expression of CXCL12 and osteolineage genes such as osterix (SP7) and lower leptin receptor (LEPR), localize to arteriolar and non-vascular niches; sinusoidal ECs are found in the sinusoidal and in the (sub-)endosteal niches, in line with sinusoids’ elongation through the entire BM cavity. In the proximity of sinusoids, another CAR subpopulation called Adipo-CAR, characterized by high expression of adipogenic lineage genes, resides, too. Interestingly, this subtle dissection of the BM cellular niche has highlighted the heterogeneity of the CAR cell population, specialized MSCs essential for the HSC maintenance and control at many developmental stages, recently isolated also from human adult BM (13). On the other hand, consensus on the panel of protein markers unique for murine and human MSCs is still lacking (see section A Role for MSCs in MPN) and establishing the relationship between MSC subsets described by various recent works is not straightforward (14).

At steady state, HSCs are mostly quiescent and located in perisinusoidal niches, while periarteriolar niches are important mainly for lymphopoiesis (15–17). Committed progenitors likely also exploit dedicated sinusoidal niches, as revealed by elegant approaches of in situ mapping, delineating an atlas of spatially and functionally distinct niches (6). For what pertains to endosteal niches, and the contribution of fully differentiated skeletal cells to HSC maintenance, opposed pieces of evidence are present in the literature with respect to osteoclasts (OCs) (18, 19); OBs have been the first cell population reported to support HSCs (20, 21), but later studies have clearly demonstrated that the major role is played by CAR cells, not by OBs (22–24), so OB contribution has yet to be clarified. Finally, osteocytes might influence specifically myelopoiesis by means of secreted molecules (25).

A plethora of supportive factors are provided in the niches ( Table 1 ). Recent evidence showed that the key niche factors SCF and CXCL12 can be modulated by the Caspase3/NLRP3 signaling, which extends understanding of regulatory mechanisms influencing haematopoiesis (40). Of note, the same molecular cue presented by diverse cells in BM niches may serve different functions in each compartment. For example, Himburg and colleagues recently demonstrated that the heparin-binding growth factor pleiotrophin, already known to promote HSC expansion in vitro and HSC regeneration in vivo, must be provided by LepR⁺ BM stromal cells for HSC maintenance during steady-state haematopoiesis, and by ECs for HSC regeneration after injury (irradiation) (32). Similar restrictions in growth factor provision apply to committed haematopoietic progenitors: for example, the maintenance of the pool of c-Kit⁺ haematopoietic progenitors requires (among other factors) SCF supply from LepR⁺ BM stromal cells and not from ECs (41). On the other hand, the same factor may elicit different effects on haematopoietic stem and progenitor cells (HSPCs) versus more downstream committed progenitors. For example, the secreted RNase angiogenin, expressed by BM osteolineage cells, has been recently demonstrated to restrict proliferation of primitive HSPCs, on one hand, and stimulate proliferation of myeloid-restricted progenitors, on the other, owing to a differential effect on RNA processing in the two subsets (42). Overall, this further demonstrates the cellular specialization within BM niches.

At the end of the 19th century, Stephen Paget postulated the “seed and soil” hypothesis that stated that TC (seeds) need a propitious medium (soil) to establish metastases. This concept can be applied also to malignant cells giving rise to haematological diseases. Most types of blood cancer, including acute and chronic leukaemia, myeloproliferative disorders and MDS, are primarily driven by accumulation of mutations in HSCs or in their progenitors. Growth and survival of the malignant clone is favoured by age-related or inflammation-driven changes in the BM microenvironment (BMME) (43, 44). On the other hand, BMME is affected by signals coming from mutated HSCs, in a bidirectional crosstalk (45).

Myeloproliferative neoplasms (MPN) and MDS are two paradigmatic and opposite examples of diseases caused by mutated HSCs that in turn can alter the surrounding microenvironment, although detailed mechanisms have not been completely defined. MPN include essential thrombocythemia, polycythaemia vera, and primary myelofibrosis, characterized by excess of platelets, erythrocytosis or myelofibrosis, respectively, with increased risk of thrombotic events and of leukemic transformation (46). They are due to mutations occurring at the HSC level in JAK2, MPL or CALR genes, all resulting in unregulated activation of the JAK/STAT pathway, although involvement of other genes (47) or microenvironmental factors contribute to disease initiation or progression.

MDS are a heterogeneous group of acquired clonal disorders of HSCs, characterized by ineffective haematopoiesis, peripheral cytopenia, genetic instability, and high risk of progression to acute myeloid leukaemia (AML) (48). MDS- and MPN-HSPCs display cell-intrinsic dysregulation of innate immune and inflammatory pathways, which in turn have an impact on the surrounding BMME. For example, MDS-HSPCs have aberrantly high expression of TLRs, which activate the adaptive immune system contributing to maintain an inflammatory environment that is detrimental for HSC function, characterized by increased local and systemic levels of IL-6, IL-1β, or type 1 IFN (49). The concomitant presence of chronic inflammation and of increased levels of anti-inflammatory proteins like TGF-β and TNF-α, contribute also to expand BM myeloid-derived suppressor cells (MDSCs), known to dampen T and natural killer (NK) cell anti-tumour activity (50). Other constituents of the BMME that are dysfunctional in MDS or MPN include ECs, within the vascular niche, and Schwann cells. MDS patients display increased BM microvascular density, likely due to the reported secretion of angiogenic growth factors from MDS cells. ECs from MDS patient manifest genetic, transcriptional and epigenetic modifications, along with secretion of supportive myeloid growth factors, further favouring the growth of the malignant clone (51). Regarding Schwann cells, their number is decreased in MPN compared to healthy donors (HDs), while it is markedly higher in MDS patients with severe fibrosis. Despite the relevance for the pathophysiology has not been demonstrated, one hypothesis is that Schwann cells enhance TGF-β activation, which contributes to the suppression of normal haematopoiesis as well as the promotion of BM fibrosis (52).

All the cell populations mentioned above cooperate in creating a unique milieu that favours TC immune evasion and promotes disease progression. Among all the different BMME components, here we will focus on MSCs, which can influence the malignant clone directly, through an altered HSC-supportive capacity, or by exerting immunosuppressive functions on innate and adaptive immunity cells, thus indirectly affecting malignant HSCs by favouring evasion from immunosurveillance.

In line with the “seed and soil” concept for metastases establishment, MSCs are an important “soil” component since they can enhance the metastatic ability of TC by strengthening their motility and invasiveness. Moreover, they create a metastatic niche at secondary tumour sites (81, 82). MSCs have been reported to promote gastric (83), lung (84) and breast cancer (BCa) (85) growth and metastasis via stimulation of epithelial to mesenchymal transition (EMT). For instance, OPN release by TC was found to induce MSC production of the chemokine CCL5, which in turn promoted CCR5-mediated BCa cell motility, invasiveness, and metastasis. CCL5 was also reported to be secreted in vitro by human BM MSCs in response to osteosarcoma (86) and BCa cells. MSCs release also factors such as TGF-β, IL-10, NO, PGE2, and IDO, implicated in immunomodulation and thus relevant in creating a TC favourable environment (87).

BCa cells entry into the BM may be facilitated by MSCs through Tac-1 regulation of SDF-1α and CXCR4 (88). The subsequent establishment of TC within the BM results in a pathological cellular crosstalk disrupting bone homeostasis which is mainly controlled by the RANKL/RANK axis (89–91). Whether and how OCs contribute to the pre-metastatic niche and TC bone tropism is largely unknown. RSPO2 and RANKL, secreted by BCa cells as recruiting factors for OC precursors, have been demonstrated to bind their receptor LGR4 through an autocrine/paracrine loop and stimulate the production of DKK1, which acts on OC precursors to promote OC differentiation and pre-metastatic niche formation (92). RANKL signalling is harnessed also in other contexts. For example, in Multiple Myeloma (MM), a plasma cell malignancy developing in the BM, TCs have increased RANKL and decreased OPG expression, resulting in enhanced OC bone resorption and the development of lytic bone lesions (93). Moreover, CXCL12-expressing fibroblasts have been associated with a cancer-promoting phenotype in BCa and aggressive solid tumours (94), pointing to a role in bone metastases (95).

Not only solid, but also haematological cancers remodel the bone microenvironment and generate bone metastases. For example, patients with Adult T-cell leukaemia/lymphoma (ATL) may have widespread osteolytic lesions and hypercalcemia, and a novel ATL mouse model has been recently generated to dissect disease mechanisms and heterogeneity (96). Osteolytic and/or osteosclerotic lesions are present in Hodgkin’s Lymphoma patients, too (97). Moreover, scRNAseq analysis of the BM stroma in a murine model of AML has demonstrated that TCs impair mesenchymal osteogenic differentiation and deregulate the expression of CXCL12 and KITL (3). As another example, MM hijacks the BM niche through direct cell-cell interaction and MM-derived EV-mediated signalling; in this respect, the oncogenic NOTCH receptors have been recently identified as part of the MM-EV cargo with pro-tumorigenic effect (98).

Intercellular communication, even on the long range, has been demonstrated to occur also through diverse types of EVs including exosomes (99). Exosome-mediated PKM2 transfer from prostate cancer (PCa) cells into BMSCs has been shown to promote premetastatic niche formation. Specifically, exosome-derived PKM2 increased CXCL12 production by BMSC in a HIF-1α-dependent fashion, which in turn enhanced PCa seeding and growth in the BM. Accordingly, targeting this axis diminished exosome-mediated bone metastasis (100). Using a bone metastatic model of enzalutamide-resistant PCa, Henrich et al. demonstrated that BM myeloid cells in vitro and in vivo did uptake the EVs released by PCa, leading to activation of NF-κB signalling, enhancing OC differentiation, and decreasing myeloid TSP-1 expression. Reducing BM myeloid cell cholesterol, through systemic administration of nanoparticle mimic of native HDL, prevented the uptake of PCa EVs and, consequently, reduced metastatic burden by 77% (101).

RUNX2 and its regulated genes have been shown to facilitate the acquisition of osteomimetic features and enhance the bone metastatic potential of BCa cells, when overexpressed. Different EV proteins were identified mediating the specific recognition of tumour-derived EVs by OBs (CDH11) and the induction of the osteogenic premetastatic niche (ITGA5). These new markers were demonstrated to be responsible for the formation of a premetastatic niche, revealing a potential EV-based premetastatic niche blockage strategy (102).

Myelodisplastic and myeloproliferative disorders are characterized by vivid interactions with the osteo-haematopoietic niche. Thus, treatment strategies targeting not only malignant cells, but also the signalling pathways connecting both sides need to be developed to provide a more effective approach ( Figure 1 ). The importance of the BMME is highlighted by the high rates of graft failure and relapses as well as the prolonged time to stabilize the engraftment in MDS patients after allogeneic HSPC transplantation; in fact, the prerequisite of a success stands within the appropriate milieu provided by the BMME and the crosstalk that must be re-built with the haematopoietic cells. In this perspective, for example, the hypomethylating agents Azacytidine and Decitabine are effective not only on the leukemic clone through Wnt signalling inhibition, but also on bone cells improving bone metabolism and favouring bone formation (103, 104). Additionally, Azacitidine in combination with Magrolimab (anti-CD47 antibody) and APR-246 (Eprenetapopt) are exploited in high-risk MDS patients including those with TP53 mutations, which have a complete remission rate lower than 20% with the standard-of-care Azacitidine therapy and poor prognosis. Mechanistically, APR-246 covalently binds to mutant p53 leading to its thermodynamic stabilization, thus shifting the equilibrium toward a functional conformation restoring its activity (105). In a phase Ib/II study, combination treatment with Eprenetapopt and Azacitidine is well-tolerated yielding high rates of clinical response and molecular remissions in patients with TP53-mutant MDS and oligoblastic AML (106).

Another example of treatment affecting both compartments is ACE-11 (Sotatercept), an activin receptor IIA (ActRIIA) ligand trap, which binds activin and other ligand of the TGF-β family thus interfering with the SMAD pathway (107). Sotatercept influences differentiation of erythroid progenitors or precursors probably by modulating factors of the BM niche (108). Indeed, stromal cells showed alterations in the expression of various important genes and cytokines in response to the drug (109). For example, several secreted proteins with relevance for the regulation of erythropoiesis were upregulated (e.g., IGFBP2, angiotensin II, BMP6, TSP1) or suppressed (e.g., VEGFA, OSM, BMP2) in response to ACE-011 treatment. The analogous ACE-536 (Luspatercept) targets preferentially GDF-8 and GDF-11, corrects the anaemia associated with ineffective erythropoiesis in the NUP98-HOXD13 murine model (110) and has been recently shown to reinstate SDF-1-mediated haematopoietic support by MSCs, thus restoring ineffective haematopoiesis (10). Importantly, both drugs promote maturation of late-stage Epo-independent erythroid precursors and co-treatment with Epo-induced synergistic responses suggesting their use for the treatment of MDS-related anaemia (108), as well as for concomitant alterations of the osteo-haematopoietic niche. In fact, Luspatercept was recently approved for the treatment of β-thalassemia (111) and for MDS low-risk patients with ring sideroblasts who have failed or are ineligible for erythropoiesis-stimulating agents (112).

An alternative strategy to interfere with the niche in MDS could be to counteract the iron overload by means of exogenous hepcidin, transferrin, hepcidin analogues and signalling agonists, since evidence in patients suggested that iron chelation could improve erythropoiesis (113).

The MSC population could be also influenced by Ruxolitinib, a drug used to treat MPN. Ruxolitinib is a JAK1/2 inhibitor that reduces JAK-STAT signalling, which is altered in MPN cells due to the presence of somatic mutations in JAK2 (JAK2^V617F), CALR or MPL. The drug acts also on MSCs, by altering the expression level of fibrosis- and HSC maintenance-associated genes (such as LOXL2, SPARC and ADAMTS4, on one hand; and CDH2, CXCL12 and ANGPT1, on the other) and by modifying the cytokine profile, reducing MCP-1 and IL-6 secretion (114).

Moreover, based on the reported hyperactive NF-kB signalling in MPN-MSCs, this pathway could serve as a target in a combined therapeutic approach against haematological malignancies, using for example a NF-kB inhibitor such as Bortezomib or Carfilzomib (66).

For what pertains to bone metastases, treatment is aimed at preventing disease progression and alleviating symptoms and may vary depending on the disease. The classical bone-targeting agents bisphosphonates and Denosumab (anti-RANKL antibody) have been shown to decrease the incidence of skeletal-related events in patients with MM and in those with bone metastases (regardless of the tumour type), but they do not replace the missing bone and, therefore, patients remain at risk of developing fractures, while the use of bone anabolic agents is not yet approved for routine clinical practice (115).

A wide range of agents studied in the last decade block OC bone resorption (116), including Everolimus (mTOR inhibitor), cathepsin K (a protease that degrades collagen during bone resorption) inhibitors (117), SRC tyrosine kinase inhibitors (118) and Cabozantinib (an inhibitor of receptor tyrosine kinases including VEGFR2 and MET) (119).

Emerging targets are the RSPO2/RANKL-LGR4 axis for inhibiting BCa bone metastasis (92), as well as metabolic factors like cholesterol homeostasis (101), that plays critical gate-keeping roles in regulating pro-metastatic signals by target cells at distant sites; both would motivate strategic diagnostic and therapeutic interventions aimed at preventing metastasis. Moreover, the exosome-induced CXCL12 axis could be another actionable pathway, based on the promising results in diminishing exosome-mediated bone metastasis (100). Notably, preventing CXCL12-CXCR4 interaction with the CXCR4 inhibitor Plerixafor (AMD3100) disrupted MM cell contacts with the BMME, thus leading to MM cell mobilization into the circulation (120).

In conclusion, up-to-date extremely powerful technologies have been increasingly unveiling the kaleidoscopic nature of BM niches and their changes in pathologic conditions. In this microenvironment, MSCs are a key component and attract specific interest for therapeutic purposes, even though thus far successful applications in the clinic are limited, while strategies to exploit their plasticity in situ could be explored to bring results closer to expectations. In the tumour setting, future research should better dissect the mechanisms underlying altered MSC function. A pro- and anti-tumorigenic function has been demonstrated for MSCs with respect to MPN cells. To the best of our knowledge, no evidence of a similar mechanism in MDS cells is present in the literature, while it would be worth investigating, particularly for therapeutic purposes.

In the framework of bone metastases, whether TC infiltration elicits long lasting effects on the BM resident cells and whether the BMME remains dysfunctional even after depletion of TC from the metastatic site, are open questions. In a translational perspective, a better understanding of the impact of TC infiltration on the BM milieu could reveal better therapeutic targets. A mechanism of cell-cell communication raising much interest lastly is the crosstalk mediated by EVs. Tumour-derived EV are potent mediators of pre-metastatic niche formation due to their pro-malignant molecular cargo and their propensity to target specific cell types, thus engineering EVs as drug carriers for targeted therapy is an attractive option. Last, combination of the most effective therapies that address different mechanisms, depending on the disease, is likely to be superior to any single therapy.

VG wrote the Abstract and the “A role for MSCs in MPN” chapter. LC and FF wrote the “BM Niches as Tumour Cell Factory in Acquired Blood Disorders: A Focus on MPN and MDS" and “A role for MSCs in MDS” chapters. And also: All the authors contributed to the “Conclusions and Perspectives” chapter. CN wrote the “The Bone-BMME: a tumour cell soil” and “Therapeutic implications” chapters. CS wrote the “Introduction” and “Composition of BM niches” chapters. All the authors contributed to the “Discussion and perspectives” chapter. LC generated the figure. All the authors contributed to the “Conclusions and perspectives” chapter. All authors contributed to the article and approved the submitted version.

This work was partially supported by the Italian Ministry of Health (grant RF-2018-12367680 to CS).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.