Myeloproliferative disorders and their effects on bone homeostasis: the role of megakaryocytes

Myeloproliferative neoplasms (MPNs) are a heterogeneous group of rare and chronic hematological diseases that arise from the clonal expansion of abnormal hematopoietic stem cells. According to the World Health Organization classification of 2016, the Philadelphia-negative, or classic MPNs consist of polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF).¹ PMF can be described as either prefibrotic (pre PMF) or overt PMF, and myelofibrosis (MF) can be either primary (ie, diagnosed without any preceding MPN) or secondary (developed after a diagnosis of PV or ET [post-MPN MF]).¹ ET is the most common classic MPN, followed by PV and PMF with an annual incidence of 1.03, 0.84, and 0.47 per 100 000, respectively.² The median survival is ∼18 years for ET, 15 years for PV, and only 4.4 years for PMF.^3,4

Driver mutations have been identified in the Janus kinase 2 (JAK2) gene and in the myeloproliferative leukemia virus oncogene (MPL) and calreticulin (CALR) gene that cause dysregulation of JAK-STAT signaling. The single-point mutation JAK2^V617F in exon 14 is present in ∼96%, ∼55%, and ∼65% of patients with PV, ET, and PMF, respectively,^4-6 and mutations in exon 12 of JAK2 have been found in ∼3% of PV cases.^4-7 Approximately 3% of ET cases and ∼10% of cases of PMF have mutations in the MPL gene,^4,6,8,9 and the presence of CALR gene mutations has been reported in ∼20% of ET and ∼25% of PMF cases.^4,6,10 Yet, further genomic analysis revealed that >50% of patients with MPN have additional genetic aberrations.¹¹ Various genes can be affected, such as SH2B3, DNMT3A, TET2, IDH1, IDH2, ASXL1, EZH2, U2AF1, SF3B1, SRSF2, ZRSR2, and TP53.^4,6,12

The genomic classification of patients has provided the opportunity for generation of promising prognostic models that could enhance treatment management.¹³ That MPN driver mutations can be detected very early in life before diagnosis, with JAK2^V617F and DNMT3A mutations being present even in utero, could provide early intervention and a better disease outcome.^14,15 Although the different forms of MPN have been extensively studied and reviewed in the context of clonal expansion, fibrosis, and other phenotypes, the influence of MPN subtypes on bone homeostasis has been debated. This review summarizes published knowledge on the effects of MPN on bone health and the possible mechanisms leading to these influences.

A brief survey of major regulators of bone development is given as the basis for probing the mechanisms of the effects of MPN on bone health.

Skeletal bone consists of cortical bone, which is the hard outer layer, and trabecular bone, which is found inside the tissue and has a large surface exposed to the bone marrow (BM) and blood flow. Bone homeostasis is based on a balance between formation by osteoblasts (OBs) and resorption by osteoclasts (OCs). Bone can be formed by either intramembranous or endochondral osteogenesis. In the former process, mesenchymal stem cells (MSCs) differentiate directly into OBs, generating the flat bones, whereas in the latter process, MSCs differentiate into chondrocytes that form a cartilage intermediate that must undergo ossification by OBs, thus giving rise to the long bones.¹⁶ In endochondral osteogenesis, the formed chondrocytes switch from a proliferating to a nonproliferating state, and they become hypertrophic and then apoptotic, enabling OBs to start the ossification process. In both intramembranous and endochondral osteogenesis, osteoblastic bone formation is the same.

The first step in skeletogenesis requires the MSCs to form condensations. This process is controlled by factors such as transforming growth factor-β (TGF-β) and members of the HOX family.¹⁷ In addition, interactions of cells with the extracellular matrix are involved in the initiation of MSC differentiation.¹⁷ As they start to differentiate, MSCs give rise to osteochondroprogenitors, which are multipotent cells capable of generating either chondrocytes or OBs. The decision of lineage fate depends on the expression of runt-related transcription factor 2 (RUNX2) and SRY-Box9 (SOX9) by the osteochondroprogenitors. The SOX9 transcription factor is dominant over RUNX2 and triggers the activation of chondrocyte-specific genes,¹⁸ whereas RUNX2 determines OB vs chondrocyte differentiation, and its expression is modified by regulators that can cause its stabilization and upregulation, surpassing SOX9 dominance.¹⁹ Nevertheless, RUNX2 requires osterix (OSX or SP7), another key osteogenic transcription factor, to activate OB-specific genes and ensure full differentiation to OBs.²⁰ 

During their differentiation, OBs deposit bone matrix (osteoid) mainly composed by collagen type 1, and when the appropriate thickness is reached, some OBs get trapped and differentiate into osteocytes.²¹ Mineralization of the bone matrix is initiated at the stage of the late mature OBs and continues through the osteocyte level.²¹ Noncollagenous proteins produced by these cells regulate the mineralization process, which includes rapid mineral accumulation, carbonate substitution, collagen condensation, and water reduction.²¹ The formation of new bone, if not associated with normal mineralization, could lead to increased fragility.

Several signaling pathways are involved in bone development. In the canonical Wnt signaling pathway, translocation of β-catenin into the nucleus results in the transcription of specific genes that drive the osteoblastic precursor cells to become more mature OBs.²² Bone morphogenetic protein (BMP) signaling has various roles in the development and regulation of bone formation.²³ Several subtypes of BMP (BMP-2, -6. -7, and -9) have been demonstrated to induce differentiation of MSCs into OBs, and BMP2 has been shown to upregulate the expression of RUNX2 level. BMPs normalize cartilage development as well; therefore, this signaling pathway is crucial for both intramembranous and endochondral osteogenesis.²³ Similarly, fibroblast growth factors (FGFs), such as FGF-2, -3, -4, -9, and -18 are significant in chondrocyte and OB differentiation, as demonstrated by loss- or gain-of-function approaches.²⁴ 

Bone homeostasis also highly depends on the resorption processes, for which OCs are key. The receptor activator of the NF-κΒ ligand (RANKL) secreted by stromal cells or OBs binds to its receptor RANK on OC precursors, leading to OC differentiation. In contrast, osteoprotegerin (OPG), secreted by OBs, acts as a soluble decoy receptor that binds to RANKL and blocks the differentiation of OCs. Several hormones and cytokines regulate this process and are responsible for the final bone phenotype (reviewed in Zupan et al²⁵).

Considering potential differences in the bone phenotypes observed in male and female patients with MPN and in mouse models, attention should be given to the effect of estrogen or/and calcium homeostasis. Estrogen regulates bone metabolism on multiple levels by inhibiting OB apoptosis, OC formation, and activation of bone remodeling, as well as by increasing OB lifespan and OC apoptosis.²⁶ A decrease in estrogen level leads to augmented expression of macrophage colony-stimulating factor (M-CSF) and RANKL in OBs, and the production of interleukin-1 (IL-1), IL-6, and tumor necrosis factor-α (TNF-a) by T cells that enhance OC formation and function with a subsequent bone loss.²⁶ This finding is noteworthy, because women may have osteopenia at the age when they are affected by MPN.

Calcium is important for healthy bone, and its homeostasis is regulated by parathyroid hormone (PTH) and 1,25-dihydroxy-vitamin D (1,25(OH)2D), which in turn are regulated by serum calcium levels.²⁷ Calcium depletion is more common in menopausal women, whereas a decrease in serum calcium leads to increased PTH secretion by the parathyroid glands, causing increased calcium reabsorption in the kidneys, and augmented bone resorption for calcium release.²⁷ In parallel, PTH triggers the secretion of 1,25(OH)2D from the kidneys, causing increased calcium absorption in the gut, increased bone resorption, and a subsequent decrease in PTH secretion.²⁷ Calcium metabolism is interesting in the context of MPNs, because some patients have CALR mutations.¹⁰ Calcium is important for MK function,²⁸ and normal CALR is involved in its homeostasis. Normal CALR regulates store-operated calcium entry²⁸ by forming a complex with ER protein 57 and the endoplasmic reticulum calcium sensor stromal interaction molecule 1; however, a defective interaction of CALR mutants with these proteins cause the constitutive activation of store-operated calcium entry and an abnormal regulation of calcium.²⁹ These alterations may affect calcium availability and, subsequently, bone homeostasis.

Reports on bone phenotypes in MPNs vary, depending on the type of MPN, sex, and method and system of detection. Accordingly, such reports are surveyed here based on studies using experimental models and human data pertinent to PMF vs PV or ET.

Several investigations point to increased bone volume in PMF. In mouse models, expression of murine JAK2^V617F under the H2Kb promoter led to the generation of transgenic mice that recapitulate PV, ET, or PMF. The mice with a PMF-like phenotype showed a slight increase in bone cortical thickness, with formation of new bony trabeculae, which was more pronounced in older mice.³⁰ Similarly, other MK-driven MF mouse models assessed skeletal changes. GATA-1 and NF-E2 are transcription factors necessary for normal MK development, and mice deficient in those factors have an increased number of abnormal, immature MKs, with a striking increase in bone volume.^31,32 Although, MF almost always precedes osteosclerosis in patients, the increased bone volume occurs much earlier than MF in these mice. Moreover, the increased TGF-β1 expression in the MKs of Gata^low-mutated mice may be of great interest, because its inhibition restores MK maturation and reduces osteogenesis in the mice.³³ Based on these findings, it was suggested that the combination of JAK2 and TGF-β1 inhibitors could reduce disease manifestations. In these studies,^30-33 sex-matched mice were used for analyses, but possible differences between sexes were not probed. Female mice with experimentally induced TPO overexpression display myeloproliferative syndrome accompanied by MF and osteosclerosis.³⁴ Mice deficient in G6b-B, a receptor expressed on the surface of platelets and MKs, develop macrothrombocytopenia, MF, and osteosclerosis,³⁵ with females displaying a stronger bone phenotype than males. MK/platelet-specific G6b-B-knockout animals demonstrated bone phenotypes similar to the global G6b-B–knockout mice.³⁶ The role of the G6b-B gene in PMF was also supported by a publication, reporting that G6b-B loss-of-function mutations were found in a group of patients with PMF.³⁷ Together, these studies suggest that the bone phenotype is attributable to the development of PMF, regardless of the gene inducing the pathology.

Similar to conclusions derived from mouse studies, several human studies showed that patients with PMF have increased bone volume, but comparison according to sex was not fully performed.^38-45 In these later studies, bone volume was measured by various methods, as outlined in Table 1. In contrast, Farmer et al reported no significant differences in bone geometry, volumetric bone mineral density, and microstructure between patients with MF and controls, when using dual-energy X-ray absorptiometry (DXA) and high-resolution peripheral quantitative computed tomography.⁴⁶ That BM specimens from these patients showing osteosclerosis led the researchers to assume that peripheral scanning measurements may underrate the actual new bone formation and the biomechanical properties.⁴⁶ This result suggests that modes of bone assessment in PMF are indeed important, warranting further consideration.

The effect of PV and ET on bone is more obscure, and the literature shows that bone volume can be decreased, increased, or unchanged compared with controls, depending on the system examined. Mice receiving transplants of murine hematopoietic cells transduced with JAK2^V617F displayed a human-PV phenotype that progressed to secondary MF with time.⁴⁷ Histological examination of female mice in the early stages of the disease showed no difference in bone, compared with the controls, but increased bone cortical thickness and formation of newly bony trabeculae were present when the mice developed fibrosis secondary to PV, suggesting that development of MF is essential for a bone phenotype.⁴⁷ On the other hand, a significant loss of trabecular bone was observed in a JAK2^V617F-knockin mouse model that recapitulates PV.⁴⁸ 

Moreover, no definitive conclusion has been made in human studies. Similar to mice, patients with PV or ET either had attenuated cancellous bone³⁸ or showed no significant changes in the total area of trabecular bone.⁴⁰ One report stated that all categories of MPN, except newly diagnosed chronic myeloid leukemia (CML), had significantly higher trabecular volume than did the controls.⁴⁵ A meta-analysis concluded that the effect of PV, ET, and CML on bone density is negative, but is statistically significant only in CML, whereas MF (primary or secondary) is characterized by an increase in bone density.⁴⁹ Yet, there is an augmented risk of osteoporotic fractures in patients with ET, PV, or CML compared with the general population, but the mechanism remains to be elucidated.⁵⁰ 

Table 1 summarizes the bone phenotypes in human studies, including modes of measurement in each case. Indeed, different technologies used to assess bone have their limitations, and this could contribute to the above-described discrepancies in the literature concerning the effects of MPN on bone volume. Histomorphometry, usually applied to histological sections of transiliac bone biopsy specimens, measures structural and remodeling bone parameters, providing information on bone mass, structure, demineralized bone, extent of resorption cavities, and bone formation rate. Nevertheless, it is a complex and invasive technique with the possibility of large measurement errors related to observer variations.⁵¹ DXA is a noninvasive method, often performed on the lower spine and hips, that measures areal bone mineral density from which the risk of fracture can be estimated. However, it does not analyze cortical and trabecular bone separately, and it can under- or overestimate real bone density.⁵² Both histomorphometry and DXA are 2-dimensional representations of a 3-dimensional structure. Quantitative computed tomography (QCT) provides a tridimensional study of the bone and measures the true volumetric bone mineral density.⁵² Moreover, QCT assesses the structure and geometry of the bone and analyzes the cortical and trabecular areas separately.⁵² High-resolution QCT can also quantify the trabecular and cortical architecture.⁵² Importantly, bone changes may not be the same at different anatomical sites, and the fact that the methods used for bone assessment do not always examine the same sample site could also be responsible for variation in study results.

Several cytokines and growth factors have been consistently identified as upregulated in patients with MPN.⁵³ There is considerable evidence supporting a strong correlation between MF and increased TGF-β levels in patients with all 3 classic MPN forms.^54,55 The role of TNF-α in MPNs is also well established,^56,57 and it is associated with inferior overall survival in PV and PMF.⁵⁸ In addition, increased platelet-derived growth factor (PDGF) levels have been reported in urine of patients with PMF, ET, or PV,⁵⁹ and in circulating platelets and serum of patients with MPN and MF.⁶⁰ Interestingly, PDGF is higher in ET than PV, and it is also higher in patients with ET who are positive for the JAK2^V617F gene mutation than in those who are negative.⁶¹ Elevated levels of vascular endothelial growth factor (VEGF), IL-1β, IL-6, IL-8, and oncostatin M (OSM) were found in different forms of MPN.^56-58,61-64 Upregulation of lysyl oxidase (LOX), a regulator of the extracellular matrix, was also observed in MKs of patients with MPN, especially in PMF, compared with controls.^65,66 Finally, a very recent study highlighted the role of increased secretion of CXCL4/platelet factor 4 by MPN cells as a booster of fibrosis and a facilitator of upregulated inflammatory cytokines, such as TGF-β and TNF-α.⁶⁷ This study also showed that hematopoietic stem cells carrying MPN-associated mutations alter the gene expression profile in mesenchymal stromal cells.⁶⁷ The impact of this reprograming on OB differentiation is yet to be explored.

Of the several MPN-associated factors, some have been reported to lead to bone loss. IL-1β is a strong stimulator of bone resorption that exerts its function by increasing the expression of RANKL, through stimulation of OC development and activity, in stromal cells,^68,69 and/or by having a direct effect on OCs.⁷⁰ IL-1β also works in cooperation with TNF toward this effect.⁶⁹ Furthermore, migration of OBs is inhibited in the presence of IL-1β.⁷¹ IL-8 participates in the altered MK growth in MF,⁶² while having an osteoclastogenic function as demonstrated either directly on the number of OCs or indirectly through upregulation of RANKL.^72,73

LOX is elevated in both human and mouse JAK2^V617F mutated MKs, the inhibition of which reduced MK adhesion to collagen⁶⁶ and the fibrotic phenotype in mouse models of PMF.⁷⁴ In contrast, LOX was shown to be necessary for normal OB proliferation and/or differentiation in studies using osteoblastic cell lines, primary mouse calvaria cells, and LOX-deficient mice.^75-78 Excess TNF-α, also an MPN-associated factor, downregulates LOX expression and affects OB fate.^76,79 Murine models treated with a LOX-inhibitor presented with altered collagen cross-linking patterns, reduced trabecular bone, and decreased cortical bone fracture toughness and strength.^80,81

Several of the released factors associated with MPNs exert a dual effect on bone, depending on concentration and timing, as well as possible effects on both OBs and OCs (Figure 1). For example, at early stages of MSC differentiation, TGF-β cooperates with BMP signaling to promote the osteoblastic lineage.⁸² Furthermore, it increases OB proliferation.⁸³ On the other hand, TGF-β promotes OC differentiation at a low dose but inhibits the process at a higher dose, by modifying the RANKL/OPG ratio.⁸⁴ PDGF (5 isoforms: PDGF-AA, PDGF-BB, PDGF-CC, PDGF-DD, PDGF-AB) has been considered to be a potential therapy in bone formation and healing of fractures.⁸⁵ It can promote OB proliferation, acting as a mitogen for BM stromal cells, with a chemotactic effect on OB-like cells.^86,87 Past studies concluded that PDGF is not a factor for OB differentiation and mineralization, even though it can promote their proliferation,^88,89 whereas a recent study demonstrated that continuous exposure to PDGF-BB increases osteogenic differentiation from BM stromal cells .⁹⁰ On the other hand, OC formation, osteoclastic activity, and OC precursor chemotaxis are enhanced by the presence of PDGF.⁹¹ Another factor, TNF-α, is known as a potent stimulator of bone resorption that stimulates osteoclastogenesis in a RANKL-dependent^68,92 or -independent manner.⁹³ In addition, TNF-α can both inhibit and promote osteoblastogenesis, and this contradictory effect can be explained based on the differentiation stage of the treated cells.⁹⁴ VEGF has a role in both intramembranous⁹⁵ and endochondral bone formation⁹⁶ by coupling angiogenesis with osteogenesis, and it stimulates OB differentiation in an autocrine or paracrine manner.⁹⁷ Moreover, it can control differentiation of MSCs by favoring osteoblastogenesis vs adipogenesis through an intracrine mechanism.⁹⁸ On the other hand, VEGF can also increase bone resorption by acting on OCs and enhancing their recruitment, differentiation, and survival.⁹⁹ Finally, both IL-6 and OSM have been shown to stimulate the function of OCs indirectly, by increasing the expression of RANKL and OPG, a protein with multiple effects, including on osteoporosis.¹⁰⁰ In contrast, other publications have described IL-6 as a suppressor of OC differentiation, and OSM can promote bone formation by inhibiting sclerostin.^101,102

The BM consists of at least 2 major parts: the vascular niche that contains most sinusoids and arterioles and the endosteal niche that includes OBs, OCs, and a specific osteoblastic subpopulation.¹⁰³ MKs are mainly located in the vascular niche; however, they have been associated with the endosteum, as well.^103,104 Thus, MK localization in the BM allows for these cells to potentially affect bone homeostasis through direct interaction with MSCs and/or through released factors (Figure 1). In addition, the atypical localization of MKs at the paratrabecular area and endosteal border seen in myelodysplastic and myeloproliferative bone marrow (Figure 2) could indicate that this altered distribution influences bone in these pathologies.^105-107 Accordingly, coculturing wild-type (WT) OBs with MKs derived from mice deficient in GATA-1 or NFE-2 transcription factors significantly enhanced OB proliferation, in accordance with a dramatic increase in trabecular bone volume observed in these mice.³¹ Interestingly, the use of conditioned medium from cultures of these gene-deleted MKs, not only failed to enhance OB proliferation, but caused a dose-dependent inhibition of this process, indicating that direct contact is necessary for the enhancement.³¹ A different study explored the effect of mechanical stresses on these processes, finding that the osteoblastic differentiation of MC3T3-E1 cells was decreased in the presence of coculture with the megakaryocytic cell line Meg-01 when exposed to fluid shear, as compared with cultures in a static condition.¹⁰⁸ 

Whether MKs exert any effect on OCs was also examined.¹⁰⁹ To this end, OCs from spleen or BM of WT mice were cocultured with MKs that were generated from fetal liver of WT or GATA-1–deficient mice. Both types of MKs blocked OC development, and MK-conditioned medium completely inhibited OC biogenesis, suggesting a role for a yet unidentified soluble factor. In line with the mouse data, human MKs derived from normal donors were tested for an effect on the synthesis of type 1 collagen, OPG, and RANKL in human OBs, finding a decreased expression of RANKL and an increased expression of OPG and type 1 collagen, which is indicative of enhanced OB differentiation.¹¹⁰ Human MKs also inhibit the development and the activity of OCs, and this inhibition is stronger when the MKs act on OC precursors, rather than on more mature OCs.¹¹¹ It has been suggested that both cell-to-cell contact and MK-soluble factors are responsible for this result. No studies are available on the effect of human MPN MKs, or mouse JAK2^V617F, or other MPN-mutated MKs on OB or OC differentiation.

As to mechanisms by which MKs affect OBs, several pathways were explored. Because gap junction intercellular communication (GJIC) is necessary for cell proliferation and differentiation, its role in promoting MK-OB interaction was tested and confirmed by using two gap junction uncouplers.¹¹² However, an unexpected observation was that OB proliferation was enhanced after inhibition of GJIC in the presence of MKs. OB differentiation was decreased when MKs were cocultured with OBs for extended periods, and blocking GJIC did not alter this MK-mediated reduction. Subsequent studies pointed to integrins α3β1 and α5β1 and glycoprotein IIb as essential for MK-OB interaction.¹¹³ Signal transduction pathways that are implicated in MK-induced OB proliferation were identified as Pyk2 signaling cascades and the p38/MAPKAPK2/p90RSK kinase cascade.^114,115 Interestingly, the capacity of MKs to enhance OB proliferation weakens as the MKs age.¹¹⁶ In contrast, the number of MKs, per se, affected the phenotype observed.¹¹⁷ An increasing number of MKs in coculture experiments further increased OB proliferation and decreased OC formation, regardless of whether WT or mutant MKs were used. However, different maturation stages of MKs did not alter the aforementioned effects on OBs and OCs.¹¹⁷ The sex of the mice used in the cited publications was not specified.

There could be several reasons for the variability in reporting the effects of MPNs on patients’ bone health, with a major reason being the evolving methods of bone assessment. In addition, continuous updates in the classification system may result in misclassification of patients, topped by difficulty in recruiting the most appropriate controls. Usually, bone examination was conducted once, after diagnosis, whereas serial evaluations of the patient would be more informative for a correlation between disease progression and effect on bone. In some cases, patients may have taken medication that affects the bone, and in most studies, there is no comparison between men and women that could explain possible sex hormone effects. Larger studies that compare men with women and use individuals as their own controls via progressive bone density measurements are needed to better understand the bone phenotype in human MPNs. Future investigations could also include probing genes related to MF development for their effect on bone, such as NBEAL2, the mutations of which are present in patients with gray platelet syndrome who present with MF. Additional mouse studies could shed light on mechanisms by which MKs in different forms of MPN affect bone homeostasis directly or via secreted factors. Finally studying these processes in the context of bone regeneration after fracture, together could lead to developing means to ameliorate this side effect of the disease.

The authors thank Andrew Piasecki for help in editing the manuscript and Shinobu Matsuura for insights on some aspects of the review.

K.R. is an established investigator with the American Heart Association and is funded by National Institutes of Health, National Heart, Lung, and Blood Institute grant R01HL136363.

Contribution: A.K. searched the literature and wrote the first draft of the review, and K.R. added to the search and the content and finalized the review article, together with A.K.

Conflict of interest: The authors declare no competing financial interests.

Correspondence: Katya Ravid, Boston University School of Medicine, 700 Albany St, W-601, Boston, MA 02118; e-mail: kravid@bu.edu.