Flipping the switch in the stem cell niche

In this issue of Blood, Feng et al¹ describe a novel mouse model by which deletion of a single copy of the 9p21 tumor suppressor locus induces an aberrant bone marrow (BM) microenvironment, leading to loss of support for hematopoiesis, increased marrow fibrosis, and development of a fatal myelodysplastic/myeloproliferative (MDS/MPN) disorder.

The BM microenvironment, or niche, is essential for support and regulation of normal hematopoiesis and, likewise, is involved in sustaining and promoting leukemia.² Composed of diverse cell types with distinct functions, the BM microenvironment engages in intimate bidirectional cross talk with hematopoietic stem cells (HSCs) and their progeny. The array of cell types composing the niche include multipotent mesenchymal stem cells (MSCs) capable of trilineage differentiation into chondrocytes, osteoblasts, and adipocytes, as well as vascular endothelial cells and immune cells.

In myeloid neoplasms, cells from the malignant hematopoietic clone can remodel the microenvironment to form a self-reinforcing leukemic niche that maintains and promotes expansion of the malignant clone at the expense of normal hematopoiesis.^3,4 This is accomplished in part by secretion of inflammatory mediators that instruct MSCs in turn to produce factors promoting leukemic growth. Intriguingly, evidence that an aberrant BM microenvironment itself can initiate myeloid disease comes from studies employing genetic mouse models.² Activation of β-catenin specifically in osteoblasts⁵ and deletion of Dicer1 only in osteolineage cells⁶ both cause myeloid disease with transformation to acute myeloid leukemia (AML), highlighting the importance of osteoblasts as regulators of hematopoiesis.

Feng et al now add compelling new data on the role of the BM microenvironment. The p21 locus on chromosome 9 encompasses several tumor suppressor genes, among them MTAP, CDKN2A, and CDKN2B, which have been implicated in numerous solid cancers. Among hematopoietic neoplasms, monoallelic loss of 9p21 is found in pediatric acute lymphoblastic leukemia and is linked to reduced survival. To examine the effects of 9p21 loss on hematopoiesis in vivo, the authors generated a mouse strain with conditional deletion of the 9p21-syntenic locus. Perhaps surprisingly, 9p21^+/− haplodeficient mice did not develop acute leukemia but rather extensive marrow fibrosis resulting in a hypoplastic BM and extramedullary hematopoiesis in the spleen and liver. Hematopoiesis was profoundly disturbed with severe anemia, thrombocytopenia, and changes characteristic of myelodysplasia in megakaryocyte-erythroid as well as granulopoietic lineages. Haplodeficient mice also showed abnormal cortical bone formation, pointing to an altered BM microenvironment.

Because the mice developed extensive fibrosis and a hypoplastic BM, serial transplantation of BM cells proved difficult. The authors instead used mononuclear cells from the spleen, the site of extensive extramedullary hematopoiesis, to show that the MDS/MPN disease was fully transplantable. Most strikingly, reciprocal transplantation of donor wild-type (WT) hematopoietic cells into lethally irradiated haplodeficient 9p21^+/− recipients again resulted in the same MDS/MPN disorder with 100% mortality of the recipients, but transplantation of BM cells from young 9p21^+/− mice before disease onset into irradiated WT recipients showed no perturbation of hematopoiesis. To further pinpoint the cell of origin, the authors extended their study to specifically induce deletion of 9p21 in osteoblasts, which again was able to fully recapitulate the MDS/MPN phenotype.

Using single-cell RNA sequencing to dissect alterations of the stromal cell compartment, the authors show that deletion of the 9p21 locus leads to expansion of chondrocyte and osteogenic precursors as well as cells with a fibroblast phenotype, with a concomitant reduction in adipogenesis, thus shifting the balance of HSC supportive cells in the niche. Mechanistically, the cytokine milieu of the BM was altered, with increased secretion of CXCL13 and osteopontin and severely reduced levels of CXCL12. Although CXCL12 is a well-recognized cytokine necessary for support of normal hematopoiesis, CXCL13 has been linked to fibrosis in other organs but not previously described as a major player in hematopoiesis. Osteopontin, produced by osteoblasts, is involved in bone remodeling, inflammation, and fibrosis. Osteopontin is also a negative regulator of HSCs. Thus, loss of 9p21 induces reprogramming of MSCs to fibrosis-driving osteoprogenitors and a subsequent switch in the cytokine balance of the BM niche (see figure). This points to osteoblasts as culprits in myeloid disease initiation.

So, do these results translate to the human setting? To answer this question, the authors examined bone samples from patients with MDS. They found elevated levels of CXCL13 in some MDS samples and a correlation between decreased CDKNB2 gene expression and CXCL12 levels, suggesting there may indeed be a role for loss of tumor suppressors within the 9p21 locus in the MDS niche, but confirmation in a larger patient cohort is needed. As MSCs from patients with MDS are not clonally mutated,⁷ indirect mechanisms such as epigenetic changes may play a more important role in shaping an altered niche in human disease. Along this line, decades-old work identified CDKNB2 as a frequently epigenetically silenced gene in MDS.⁸ Treatment of MSCs from patients with MDS with hypomethylating agents restores normal methylation⁹ and reverts MDS MSCs to their normal function^9,10 but has not been linked to CDKNB2 expression.

Further questions remain. How do these data relate to driver mutations in HSCs, which are considered initiating events in clonal stem cell disorders such as MDS and AML? Crossing 9p21^+/− mice with mice carrying a heterozygous knockin allele of FLT3-ITD did not result in full-blown AML. FLT3-ITD is a late event in leukemogenesis and in itself only causes MPN in murine models. It would be interesting to see how HSCs harboring founding mutations such as DNMT3A, SF3B1, or others contribute to or alter the disease phenotype in this model.

In the clinic, BM fibrosis is notoriously difficult to treat. Although hypomethylating agents show clinical efficacy in MDS and MDS/MPN disorders, they do not alleviate fibrosis. Looking toward the future, more specific therapies directly targeting the disrupted stem cell niche and imbalances in its milieu may emerge as the BM landscape in myeloid disease becomes dissectible at a single cell level. The model from Feng et al offers the possibility of testing novel niche-based therapies to target fibrosis in vivo.

Conflict-of-interest disclosure: The author declares no competing financial interests.