The Role of Extracellular Matrix Stiffness in Megakaryocyte Development and Function

Abstract

The extracellular matrix (ECM) is a key acellular structure in constant remodeling to provide tissue cohesion and rigidity. Deregulation of the balance between matrix deposition, degradation and crosslinking results in fibrosis. Bone marrow fibrosis (BMF) is associated with several malignant and nonmalignant pathologies severely affecting blood cell production. BMF results from abnormal deposition of collagen fibers and enhanced lysyl oxidase-mediated ECM crosslinking within the marrow, thereby increasing marrow stiffness. Bone marrow stiffness has been recently recognized as an important regulator of blood cell development, notably by modifying the fate and differentiation process of hematopoietic or mesenchymal stem cells. This review surveys the different components of the ECM and their influence on stem cell development, with a focus on the impact of the ECM composition and stiffness on the megakaryocytic lineage in health and disease. Megakaryocyte maturation and the biogenesis of their progeny, the platelets, are thought to respond to environmental mechanical forces through a number of mechanosensors, including integrins and mechanosensitive ion channels, reviewed here.

Composition of the Bone Marrow Extracellular Matrix (ECM)

The bone marrow (BM) ECM is a non-cellular structure that provides physical support for tissue integrity, elasticity and hematopoesis^1–3. The ECM is vital in normal hematopoiesis and plays a role in pathological states. It is comprised of various matrix proteins, such as collagens, laminin, fibronectin, and fibrinogen, as well as various soluble proteins, such as cytokines, chemokines and secreted enzymes^1,4,5. Of the different ECM structural proteins, the main ones are collagens, which include fibrillary collagen (collagens I, II, III, V and XI) and non-fibrillary collagen, with the former providing stiffness to the ECM¹. Cellular adhesion to collagen has been found to promote bone marrow cells and megakaryocyte (MK) expansion^6–8. Similar to collagen, fibronectin is important for cellular interactions with the ECM, and plays a role in adhesion, migration, growth and differentiation^9–12. The ECM also contains proteoglycans with attached glycosaminoglycan (GAG) side chains. These GAGs bind to and sequester growth factors in the ECM and may act as a reservoir of these cell-impacting proteins^1,3. Heparan sulfate is a GAG involved in mobilization of HSCs¹³, and the GAG hyaluronic acid is important for normal hematopoiesis^14–16. Other ECM proteins, such as fibrillin-1, tenascin-C, agrin, thrombospondin and matrilin-4 are important for hematopoietic stem cell (HSC) proliferation and survival during homeostatic and stress conditions, as well as regulation of angiogenesis^17–21.

The ECM is in constant remodeling. Among ECM modifying proteins are matrix metalloproteinases (MMPs), tissue inhibitors of MMPs (TIMPs), plasmin, and lysyl oxidase (LOX). MMPs are a family of zinc-dependent endopeptidases that are the main enzymes responsible for ECM breakdown, while TIMPs counteract this breakdown by inhibiting MMPs²². Dysregulation of these counteracting proteins affects the ECM in some pathological states, such as multiple myeloma (MM) and primary myelofibrosis (PMF)^5,23. TIMP-3, not only inhibits MMPs in the BM, but it also plays a role in HSC expansion, differentiation and trafficking^24,25. Plasmin is a protein activated by urokinase plasminogen activator (uPA), which has been found to be expressed in mouse hematopoietic stem cells and progenitors and by human BM stromal cells^26–28. Plasmin degrades fibrin, fibronectin and laminin, while activating MMPs²⁷. Other components of hemostasis are also found in the ECM, such as plasminogen activator inhibitor 1 (PAI-1), tissue plasminogen activator (tPA), and fibrinogen. In the bone marrow, these factors are important for hematopoietic regeneration^29,30. PAI-1 inhibits the degradation of plasmin during hemostasis, and it has also been reported to be expressed in MKs³¹. Altered PAI-1 level is associated with hematological malignancies, including MM and myeloproliferative neoplasms (MPNs)^32,33. Fibrinogen is found in the vascular sinusoids in the BM³⁴. LOX is a copper-dependent secreted enzyme, the activity of which eventually leads to the crosslinking of collagen and elastin precursors, resulting in increased ECM stiffness³⁵. Hence, depending on the balance between matrix secretion and deposition, cross-linking and degradation, tissue stiffness may vary under physiological or pathophysiological conditions.

Bone Marrow ECM Stiffness: Measurement and Natural Contributors

Several ECM components in the BM contribute to the stiffness of the tissue. This property is typically assessed by methods such as rheometry³⁶ and atomic force microscopy³⁷. Rheometry involves measuring the viscoelastic properties of a substance by looking at the relationship between deformations and stresses in order to calculate Young’s modulus, a measurement for stiffness³⁸. The limitation of rheometry is that it looks at the BM as a whole at the macroscopic level and cannot detect regional or cellular heterogeneity of stiffness³⁸. Other studies have used a micropipette aspiration method to measure stiffness by placing a glass capillary micropipette next to the tissue and applying vacuum pressure to measure the aspirated length, which is then used to calculate soft tissue stiffness³⁹. This method may be used in BM to find regional differences in stiffness; however, it faces the same limitations of rheometry in that it does not measure stiffness at the cell level. One way to measure ECM stiffness in BM at the cellular level involves atomic force microscopy (AFM) as it can provide a spatial resolution of almost 1 nm and can be used in solution and in living cells⁴⁰. AFM can also be used to measure stiffness of collagens and other ECM components⁴¹. Another method for measuring ECM stiffness at the cellular level is optical tweezers active microrheology⁴².

The stiffness of the ECM is heterogeneous throughout the BM³⁸, likely due to heterogeneous distribution of ECM components. Collagens are important contributors to ECM stiffness; Types I, II and III are the most abundant collagens in the ECM⁴³. For example, type IV and type I collagen are found mainly near the endosteal surface in the BM along with fibronectin, and fibronectin is also found in MKs and in central marrow^44,45. Type IV collagen is detected in the vasculature of the BM, while type I collagen is not normally found in those areas⁴⁵. ECM stiffness increases in a non-linear fashion as the concentration of type I collagen increases, while increasing concentrations of type III collagen mixed with type I collagen decreases stiffness^41,46. Fibronectin, a high molecular weight glycoprotein, appears to be an important regulator of ECM stiffness via contribution to the organization of collagen fibrils. In vitro experiments showed that fibronectin and integrins α₁₁β₁ and α₂β₁ are necessary to form a collagen matrix⁴⁷. This study showed that a collagen matrix did not form in a fibronectin-deficient mouse fibroblast cell line cultured in the absence of exogenous fibronectin, while the addition of fibronectin was sufficient for the collagen matrix to form. When this fibronectin-deficient cell line was transfected to express, either, α₁₁ or α₂ integrins, a collagen matrix formed even in the absence of fibronectin, however, this matrix was not as well organized. This suggested that both fibronectin and integrins α₁₁β₁ and α₂β₁ are necessary for forming a well-developed collagen matrix⁴⁷. Mice engineered to lack liver fibronectin developed extensive liver fibrosis following chronic liver injury, compared to matching controls. These null mice had more extensive and disorganized collagen deposition compared to mice with fibronectin⁴⁸.

LOX is a copper-dependent enzyme that causes oxidative deamination of lysine and hydroxylysine residues on collagen, forming aldehydes that spontaneously cross link, thereby increasing ECM stiffness^35,49. LOX has been implicated in the progression of several malignancies, including renal cell carcinoma (RCC), colorectal and breast cancers^50–52. Collagen-coated gels were found to be stiffer when treated with conditioned media of cultured primary RCC cells as measured by AFM. This stiffness was reduced when β-aminopropionitrile (BAPN), a LOX inhibitor⁵⁰. Augmented stiffness was also seen in vivo when colorectal cancer (CRC) cell lines were stably transfected with either vector, LOX or catalytically inactive LOX and implanted in nude mice. The tumors from LOX-transfected cells were stiffer than from cells with catalytically inactive LOX⁵¹. LOX is also upregulated in hematopoietic malignancies, including myeloproliferative neoplasms^4,53,54.

Effect of ECM Components and Stiffness on Bone Marrow Cell Adhesion and Development

Effect on megakaryocyte development and platelet formation

The effect of type I collagen on (megakaryocyte) MK development and function has been extensively studied. Type I collagen affects MK development by stimulating hematopoietic stem cells (HSCs) to differentiate through the megakaryocytic lineage^6,8. Type I collagen, however, inhibits proplatelet formation through activation of integrin α₂β₁ and downstream Rho/ROCK axis⁵⁵. Interestingly, MKs on N-acetylated type I collagen, which decreases mechanical tension, produce more proplatelet compared to type I collagen without this modification⁵⁵. In contrast to type I collagen, type III and IV collagens stimulate proplatelet production via the PI3K/Akt signaling pathway^44,56,57. The stiffness of type I collagen is greater than type IV collagen, and MKs plated on type I collagen remain spread after 16 hours of incubation, while MKs plated on type IV collagen extend proplatelets⁵⁷. MKs cultured on type IV collagen also show increased integrin β₁ activation and internalization at 3 and 8 hours of adhesion compared to type I collagen.

Fibronectin is another ECM component that is important to MK function and development⁵⁸. MKs adhere to fibronectin and proplatelet formation is increased by activation of its receptors, α₅β₁ (VLA-5) and α₄β₁ (VLA-4), on MKs^59,60. Fibronectin also modulates MK adhesion to type I collagen⁶¹. HSCs cultured on fibronectin produced more CFU-MKs as the matrix became stiffer, and blocking the fibronectin receptor integrin α₅β₁ abolished this increase⁶².

Other BM ECM components, such as fibrinogen, plasmin, PAI-1 and vitronectin, and thrombospondin-2 (TSP-2) have also been reported to affect MK function and proplatelet production^34,63,64. MKs adhere to fibrinogen via α_IIbβ₃₆₀. Fibrinogen is mainly found in the vascular sinusoids in the BM, and upon binding to it, MKs have increased proplatelet formation³⁴. PAI-1 inhibited MK adhesion to vitronectin, but its effect was reverse in the presence of tPA or uPA⁶⁵. Thrombospondin-2 (TSP-2) is a matricellular protein that is expressed by MKs but not platelets and has been found to reduce proplatelet formation and MK differentiation in vitro⁶⁶.

Glycosaminoglycans (GAGs), other important components of the ECM, play an important role in MK development⁶⁷. The GAG hyaluronic acid appears to be an important inhibitor of platelet production as depolymerazation of hyaluran is necessary for thrombopoesis; MKs with hyaluronidase-2 deficiency are less mature and make fewer proplatelets than hyaluronidase-2 containing MKs⁶⁸. However, in normal MKs, high molecular weight hyaluronic acid does not effect MKs and, therefore, may make a good ex vivo 3D scaffold⁶⁹. This is in contrast to the GAG dermatan sulfate, which increases proplatelet production in thrombopoetin (TPO)-stimulated MKs, compared to MKs cultured without this GAG⁷⁰.

LOX is a non-structural component of the ECM that affects MK proliferation, adhesion and function. LOX is expressed in immature normal MKs and downregulated as MKs mature⁵³. As mentioned above, the stiffness of the ECM is increased by LOX via cross-linking of collagen fibers³⁵, and LOX expression is upregulated in MKs of mouse or human primary myelofibrosis. LOX can also potentiate platelet derived growth factor (PDGF)-mediated MK proliferation by oxidizing and activating the PDGF receptor^53,71. Another role of LOX is the activation of the collagen receptor α₂β₁ on platelets, which leads to increased adhesion to collagen⁷². Reducing the stiffness of the ECM by inhibiting LOX crosslinking of collagen in mouse bone marrow, increased platelet level, supporting the notion that a less stiff matrix favors platelet biogenesis⁵⁷. LOX may affect MK development by mechanisms other than catalytic oxidation of receptors or the ECM. LOX is secreted as a 50-kDa pro-enzyme that is cleaved by BMP-1 (also expressed by MKs) to release the mature LOX enzyme and the 18-kDa propeptide (LOX-PP) that can enter cells^73,74. LOX-PP decreases MK polyploidy, possibly by decreasing the expression of cell cycle regulators⁷⁵.

Stiffness also plays a role in MK function and development. MKs cultured in 3D media that mimic BM ECM stiffness had higher ploidy levels than MKs cultured in liquid media⁷⁶. MKs cultured inside a 3D methylcellulose (MC) hydrogel of medium rigidity (30–60 Pa) had higher ploidy, higher demarcation membrane development and more proplatelet formation than MKs cultured in 2D on top of the gel or in liquid culture⁷⁶or in a 3D stiffer rigidity (300–600 Pa)⁷⁶. A separate study looked at MKs cultured on collagen-coated soft gels (300 Pa) or stiff gels (34 kPa) and found that MKs had higher ploidy on softer gels compared to stiffer gels, independent of the collagen concentration⁷⁷. However, this effect was abolished when the MKs were treated with a non-muscle myosin inhibitor⁷⁷. Ex vivo studies show that MKs forms more proplatelets on low to medium stiffness silk films than on high stiffness silk films, regardless if the silk films were coated with type I or type IV collagen^57,78. Increased stiffness also led to increased β1 integrin activation and internalization in MKs⁵⁷. Although some stiffness is needed for MK maturation, it appears that a stiff ECM is detrimental to MK development and proplatelet formation. ECM stiffness appears to control proplatelet formation via PI3K/Akt signaling pathways with decreased stiffness causing increased Akt phosphorylation while MKs had decreased proplatelet formation when treated with Akt inhibitor⁵⁷. Figure 1 includes an illustrative summary of the main ECM components that have the potential to affect stiffness.

Figure 1. Schematic illustration of megakaryocyte adhesion and fragmentation into platelets in the context of normal ECM stiffness.

As depicted here, type I collagen is normally found near the periosteum, and other collagen types, such as collagen IV, Glycosaminoglycans (GAGS) and fibronectin are spread throughout the niche. Metalloproteinases (MMPs) and tissue inhibitors of MMPs (TIMP) are present to remodel and turnover the ECM. Collagen I is less stiff than collagen IV and the rigidity of both is impacted by the level of a cross-linking enzyme, lysyl oxidase (LOX) secreted from low ploidy megakaryocytes and osteoblasts. Other components of hemostasis found in the ECM (and not illustrated here), such as plasminogen activator inhibitor 1 (PAI-1), tissue plasminogen activator (tPA) and fibrinogen were too found to impact megakaryocyte development through mechanisms outlined in the review.

Effect on blood stem cells and other bone marrow cells

The stiffness of the ECM is also important for HSC fate and differentiation^62,79–81. HSCs cultured on stiffer gels coated with fibronectin (mimicking the endosteal region of the BM) promoted the maintenance of myeloid progenitors, while laminin-coated gels promoted erythroid differentiation⁶². Laminin also increased the number of mature red blood cell (RBC) progenitors derived from human HSCs⁸². Further, culturing murine HSCs on three dimensional collagen matrices of different stiffness revealed that HSCs on stiffer matrices were more quiescent⁸¹. HSCs cultured on collagen gels showed increased viability when cultured on less stiff gels (44 Pa) than on stiffer gels (3.48 kPa)⁸³. Similarly, on collagen-coated stiff PA gels (136 kPa), HSCs showed decreased viability as the collagen concentration coating the gels was increased⁸³. Culturing cord blood derived-CD34+ HSCs on collagen gels led to decreased expansion but increased myeloid differentiation, compared to HSCs cultured in suspension⁸⁴. In a gene-expression assay, several growth factors, chemokines and cytokines were upregulated in HSCs cultured on collagen gels compared to suspension culture, including interleukins 8, 6 and 1β, TNFα, CXCL2, and CCL2 which are involved in HSC proliferation and mobilization⁸⁴. Similar to HSCs, human mesenchymal stem cells (MSCs) cultured on a type I collagen-coated gel matrix show different phenotypes depending on the rigidity of the matrix. MSCs cultured on a softer matrix appear more neurogenic and express more neurogenic genes, and those cultured on more rigid surfaces exhibit an osteogenic phenotype⁷⁹.⁸⁰

Fibronectin is another ECM component that is important for hematopoiesis. Fibronectin was found to induce apoptosis and decrease proliferation of human HSC lines via the fibronectin receptor, VLA-5, however it promotes erythropoiesis via VLA-4^85–87. Erythrocyte progenitors lose their attachment to fibronectin as they mature through down-regulation of VLA-5, which may allow them to detach from the ECM to be released into circulation⁸⁸.

GAGs appear to play an important role in hematopoiesis. In mice treated with 5-fluorouracil (5-FU), injection of hyaluronic acid (HA) increased BM recovery as well as augmented IL-1 and IL-6 production by HSC, which support hematopoiesis¹⁵. Heparan sulfate is another GAG that is important for hematopoiesis⁸⁹. HSCs express Mac-1 (CD11b/CD18) and CD45 which mediate their adhesion to stromal heparan sulfate⁹⁰.

The remodeling of the ECM and the balance between MMPs and TIMPs is important for maintaining hematopoiesis. TIMP-3 is an endogenous inhibitor of MMPs and important regulator of HSC proliferation and trafficking, as well as bone turnover. Up-regulation of TIMP-3 in mice via BM transplantation of retroviral transduced HSCs increased BM and blood myeloid cell counts,, while it decreased lymphocyte levels²⁴. There was also increased HSC trafficking to the blood and spleen and late onset of fatal osteosclerosis in mice with up-regulated TIMP-3, compared to matching controls²⁴. The role of TIMP-3 in stimulating quiescent HSCs to proliferate appears to be independent of its MMP-inhibiting activity and may be due to direct inhibition of angiopoietin-1 signaling, which mediates HSC quiescence²⁵.

ECM Signaling to Mechanosensors Expressed in Megakaryocytes

Mechanical forces have been shown to be important in MK function and development through a role of shear stress, calcium and mechanosensitive ion channels. For example, shear stress in BM capillaries and sinusoids is important for proplatelet maturation and platelet release⁹¹. MKs express several integrins that sense the ECM around them and transduce signals to the cytoskeleton^92,93. Among these integrins, α₂β₁ binds to type I collagen and decreases the production of proplatelets via the Rho/ROCK pathway⁹⁴. Integrin α₂β₁ also regulates the assembly of actin stress fibers in MKs, which mediate cell migration⁹⁴. The Rho/ROCK pathway regulates thrombopoesis by phosphorylating myosin light chain (MLC), which enhances actin-dependent myosin motor activity⁹⁵. This regulation of thrombopoesis by the Rho/ROCK pathway may be important in controlling the timing of platelet formation by MKs and preventing premature platelet formation while the MKs are still in the osteoblastic niche⁹⁵. Indeed, patients with MYH9-related diseases (caused by mutations of the heavy chain of the non-muscle myosin) present with macrothrombocytopenia that is thought to be due to premature proplatelet formation in the osteoblastic niche and impaired migration^95–97. Given that type I collagen is found primarily in the osteoblastic niche and it is a significant contributor to ECM stiffness, it is logical that areas with large amounts of type I collagen have an unfavorable environment for platelet production^8,45,46. Therefore, the sensing of ECM stiffness is important for proper MK function.

Calcium plays an important role in MK function^98,99. Increases in cytosolic calcium are due to calcium release from intracellular stores in the endoplasmic reticulum (ER) followed by extracellular calcium influx via the Store-Operated Calcium Entry (SOCE) mechanism^100,101. Adenosine diphosphate (ADP) induces an increase in cystolic calcium in MKs, which activates downstream signaling important in MK adhesion an proplatelet formation (FAK, Src, ERK and Akt pathways)⁹⁸. Intracellular calcium store release into the cytosol appears to be sufficient to sustain proplatelet formation, however extracellular calcium influx is important in the regulation of the interaction between MKs and the ECM⁹⁸. Extracellular calcium is important in the phosphorylation of MLC, which has been shown to reduce proplatelet formation in MKs^55,98. Interestingly, adhesion of MKs to type I collagen increased cytosolic calcium, and inhibition of influx of extracellular calcium by blocking SOCE reduced MK adhesion to type I collagen^98,102. Calcium dysregulation has been implicated in MK pathologies, most notably the MPNs essential thrombocythemia (ET) and PMF where mutations in Calreticulin (CALR) have been found in 25% of cases without JAK2 or MPL mutations¹⁰³.¹⁰⁴

Extracellular calcium appears to be important in sensing the ECM in MKs, and MKs express mechano-sensitive ion channels that may play a role in platelet formation. One such mechano-sensitive ion channel is the transient receptor potential cation channel subfamily V member 4 (TRPV4)⁵⁷. This channel modulated the PI3K/Akt pathway in endothelial cells when physical stimuli were applied to their cell membrane¹⁰⁵. In MKs, TRPV4 senses a soft ECM and increased intracellular calcium influx, leading to increased PI3K/Akt pathway activation and proplatelet formation⁵⁷. Inhibition of TRPV4 in MKs adhered to type IV collagen decreased proplatelet formation⁵⁷. Mice treated with the LOX inhibitor BAPN had reduced BM ECM stiffness, higher platelet counts and increased TRPV4 activation compared to control, which highlights the importance of this ion channel in sensing soft ECM and increasing platelet formation in vivo⁵⁷.

Another family of mechano-sensitive ion channels that may play a role in MK and platelet function is the Piezo family. Piezo1 and Piezo2 are mechanically activated ion channels that induce nonselective cationic currents in response to mechanical forces^106–108. The Piezo proteins are activated by tension on the lipid membrane itself^107,108. Piezo2 appears to be important in mechanosensation and in light touch and pain perception, and is expressed in a subset of somatosensory neurons and Merkel cells^106,109. Loss of Piezo2 causes a loss in vibration detection, joint proprioception and touch discrimination, as well as congenital hip dysplasia in humans, while gain of function mutations in Piezo2 are associated with distal arthrogryposis type 5, which presents with congenital contractures, ophthalmopleiga and restrictive lung disease^109–111. In chondrocytes, both Piezo1 and TRPV4 are important for ECM mechanosensing, however, Piezo1 is responsible for calcium influx in response to cell membrane stretching¹¹². Piezo1 is expressed in human erythrocytes and platelets, and was found to play a role in vascular development^113–118. Gain-of-function mutations in Piezo1 are associated with hereditary xerocytosis, an autosomal dominant hemolytic anemia due to erythrocyte dehydration and an imbalance of intracellular cation concentrations^115,116. Loss of function of Piezo1 caused congenital lymphatic dysplasia in human patients, while loss of Piezo1 caused improper vessel formation in mice^117,119. Piezo1 is involved in shear-induced calcium influx that mediate ATP release in red blood cells, while ATP induces nitric oxide (NO) production and relaxation of endothelial cells¹¹⁴. There may be a role for Piezo1 in cellular proliferation. In one study, epithelial cells had increased proliferation in response to mechanical stretching via activation of Piezo1 and subsequent calcium influx and ERK1/2 phosphorylation¹²⁰. A recent study found that Piezo1 is expressed in human platelets and the Meg-01 MK cell line¹¹³. Extracellular calcium influx was induced in Meg-01 cells and platelets when exposed to fluid shear stress and when Meg-01 cells had their cell membranes mechanically deformed with a glass pipette. This influx was largely abolished by GsMTx-4, a Piezo1 inhibitor¹¹³. Interestingly, the addition of GsMTx-4 to whole blood for 30 seconds resulted in a reduction of thrombus formation under arterial flow on a collagen surface, though no difference was seen in platelet aggregation under low shear stress conditions¹¹³.¹¹²

The role of Piezo proteins in MK and proplatelet formation has not been studied yet, though it seems to be a promising mechanosensors for ECM stiffness. Piezo1 and other mechanosensitive ion channels may play a role in MK function by affecting ion currents¹²¹. One such channel is a depolarization-gated potassium-selective (Kv) channel. MKs express the α subunit (a pore forming subunit) of Kv, Kv1.3, which is encoded by the gene KCNA3¹²². Kv1.3 is important in early calcium influx in platelets, and Kv1.3 deficient mice had significantly higher platelet counts compared to WT, but no effects on megakaryopoeisis were seen¹²². Platelets and MKs also express calcium-activated potassium channels (KCa) which may help determine membrane potential and calcium influx^121,123. One such channel, KCa3.1, is important in stromal cell derived factor 1 (SDF-1) mediated platelet migration¹²⁴.

Bone Marrow Pathology Associated with a Fibrotic and Stiffer ECM

Given the importance of the ECM and cellular signaling evoked by it during maintenance of tissue homeostasis, it is not surprising that dysregulation of ECM components leads to tissue fibrosis that disrupts organ function and is an important cause of mortality and morbidity. Fibrosis associated with excessive ECM deposition in the BM (BMF) leads to impairment in blood cell production and sometimes extramedullary hematopoiesis (EMH). These fibers are essentially reticulin fibers composed of type III collagen that can be accompanied by type I collagen fibers. Semi-quantitative assessment of the severity of BMF is through a grading system on a scale of 0–4 for the Bauermeister system or 0–3 for the revised European Consensus system^125,126. This evaluation is based on the amount of reticulin fibers, their reticulation and the presence of collagen bundles. While no collagen fibers have been observed in healthy subjects, reticulin fibers may be present at a low grade^127,128. A strong increase in the quantity and reticulation of reticulin fibers beyond the normal range, associated or not with collagen bundles, is characteristic of fibrotic marrow disorders.

A relationship between increased BM stromal fibers and disease has been most extensively studied in the myeloproliferative neoplasm, primary myelofibrosis (PMF)^129,130. PMF is a heterogeneous and clonal BCR-ABL1-negative hematopoietic stem cell malignancy characterized by BMF, ineffective hematopoiesis, MK proliferation and atypia, EMH, and splenomegaly. The last WHO diagnostic criteria from 2016 defined a pre-fibrotic PMF to distinguish it from other BCR-ABL-negative myeloproliferative neoplasms such as ET and polycythemia vera (PV) that may develop late secondary myelofibrosis^4,131,132. BMF is also observed in other hematological malignancies, such as acute myeloid leukemia (AML), acute lymphocytic leukemia (ALL), myelodysplastic syndrome (MDS) and chronic myeloid leukemia (CML).

Nonmalignant diseases associated with BMF can be of autoimmune origin, called primary autoimmune myelofibrosis (AIMF). In these cases, patients possess autoantibodies, but do not present a well-characterized autoimmune disorder. AIMF is marked by cytopenias without splenomegaly, and high grades of BMF with lymphocytic infiltration, but without associated osteosclerosis. AIMF can also be secondary to an established disorder such as systemic lupus erythematosus (SLE)¹³³. However, although isolated case reports are regularly presented, it is still an underdiagnosed disorder^134,135.

Infectious or inflammatory diseases may also be associated with BMF. This is the case for HIV or tuberculosis infection, visceral leishmaniasis, and pulmonary hypertension. In addition, disorders associated with vitamin D deficiency are also prone to development of BMF. Furthermore, exposure to toxins or radiations may lead to the development of BMF, as well as treatment with medications such as recombinant human IL-11 (rhIL-11)¹³⁶, or thrombopoietin receptor agonists^137–140.

Several studies have tried to correlate the amount and type of fibrosis with disease prognosis, but the conclusions vary. The presence of reticulin fibers alone, characteristic of mild fibrosis, does not seem to correlate with disease severity or co-morbidities. On the other hand, the presence of collagen fibers, representing higher grade of BMF, correlate with abnormal blood count and severity of the underlying disorders^126,130,141. Nevertheless, in both cases, provided that adequate treatment exists for the disease, reticulin fibrosis and even collagen fibrosis can decrease or resolve. This is the case for AIMF and SLE-associated myelofibrosis treated with corticoids¹⁴². Increased reticulin fibrosis due to CML is also reversible with Imatinib therapy, but not with interferon^143–145. In MPN patients, ruxolitonib or interferon treatment improves aspects of the pathology in some but not all patients^146,147. In cases where supportive care measures are ineffective, stem cell transplant eradicating the clone responsible for the neoplastic transformation has proven to be effective in resolving both collagen and reticulin fibrosis^148,149.

In the BM, the main cells responsible for the increase in ECM are fibroblasts, which are usually found in close association with collagen fibers^150,151. Rather than an increase in number, it is most probably their activation or differentiation into matrix-producing cells which is responsible for augmented ECM deposition^151,152. Fibroblasts respond to a number of fibrogenic factors present in the diseased marrow. Increases in cytokines such as interleukin (IL)-6, IL-12, IL-8, TNFα, IFNγ, as well as profibrogenic growth factors such as TGFβ, bFGF, VEGF and TGFβ1 have been implicated¹⁵³. MKs are primarily responsible for the secretion of many of these profibrotic, angiogenic and pro-inflammatory factors stored in their alpha granules^154–156. This has been well documented for PMF. In addition, non-myeloproliferative fibrosis has also been linked to MK hyperplasias as is the case for AIMF¹³⁵. Interestingly, the presence of MKs in the lungs was proposed to contribute to pulmonary fibrosis and pulmonary hypertension in systemic sclerosis¹⁵⁷. In this context, it is worth noting that pulmonary hypertension is diagnosed in some patients with PMF and other MPNs, and has been associated with pulmonary EMH and fibrosis¹⁵⁸. Yet, the degree of abundance and role of pulmonary MKs in health and disease, MPN included, require further study. In addition to MKs (and platelets), monocytes/macrophages could be alternative sources of fibrogenic growth factors. In PMF, monocytes are activated and overexpress IL-1 and TGFβ, suggesting that they also contribute to the fibrotic process. Table 1 summarizes the above-described physiological and pathological factors of BM stiffness.

Table 1.

Physiological and pathological factors of BM fibrosis and stiffness

BM Components	Representative References
BONE MARROW MATRIX	1,4
Proteins: Fibrillar collagens I, II, III, V, XI ; Non fibrillar collagen IV ; Fibronectin ; Vitronectin ; Fibrinogen ; Laminin ; Fibrillin- A ; Tenascin-C ; Elastin ; Agrin ; Thrombospondin ; Matrillin-4
Proteoglycans: Heparan sulfate; Hyaluronan; Chondroitin sulfate; Dermatan sulfate; Keratin sulfate; Heparin
REMODELING FACTORS IMPACTING STIFNESS	23, 29–31
Degradation enzymes and inhibitors MMPs ; Plasmin ; tPA; TIMPs ; PAI-1	4, 53, 57, 74
Cross-linking enzyme: Lysyl oxidase (LOX)
FIBROGENIC SOLUBLE FACTORS	4, 154
Inflammatory cytokines: IL-1 -2, -6,-8,-12,- 13, -15; TNFa
Growth factors: TGFb; bFGF; VEGF; PDGF; BMP-2, -4, -5, -6	4, 73, 150, 152, 156
TREATMENT-INDUCING FIBROSIS
rhIL-11; TPO agonists	136, 138–140

Abbreviations are as in Figure 1.

Conclusions and Future Directions

Deregulation of ECM remodeling with excess matrix deposition, increased matrix cross-linking activity and defective matrix degradation, leading to increased stiffness, are associated with bone marrow pathologies. Important efforts will be needed to identify the mechanotransduction pathways used by MKs and other BM cells to sense such changes in the ECM. This is especially important considering that pathologies associated with BMF are often linked to platelet production defects. Not only platelet production but also platelet function may be affected by modifications of the mechanics of the vascular environment; as yet, unexplored area of investigation. The main function of platelets is to ensure hemostasis by forming a plug at the site of vascular injury that seals the breach and stops blood loss¹⁵⁹. Platelets are also involved in arterial thrombosis, which occurs in diseased vessels presenting evolved atherosclerotic plaques. Upon erosion and rupture of such a plaque, platelets accumulate and form a thrombus that can become occlusive, resulting in life-threatening ischemic pathologies such as stroke and myocardial infarction¹⁶⁰. The stiffness of the exposed sub-endothelium in healthy vessels far differs from that found in stiffer atherosclerotic vessels. An association between arterial stiffness and platelet activation has been reported¹⁶¹. This link was proposed to be indirect and appears to result from the reduced ability of atherosclerotic plaques to release normal levels of nitric oxide which maintain platelets in a resting state¹⁶². How stiff surfaces exposed after atherosclerotic plaque rupture modifies platelet function remains completely unknown and represents a very attractive research area which could provide some clues on why thrombus formation can become occlusive in diseased arteries as compared to healthy vessels. Recently, it was reported that platelets sense microenvironmental mechanical properties, including substrate stiffness, which results in biological signal responses¹⁶³. Evidence was provided that modifying the substrate stiffness of a fibrin surface increases platelet adhesion and spreading. Future studies are required to determine whether these observations are relevant to hemostasis and/or arterial thrombosis. They could help to identify novel mechanosensitive receptors which might represent interesting novel anti-thrombotic targets. Such a mechano-sensitive receptor, Piezo1, has recently been reported to be expressed by platelets. Piezo-1 was shown to promote Ca²⁺ entry and could participate in thrombus formation under arterial shear stress¹¹³. Future studies of other platelet mechanosensors could provide insights on how these cells sense both matrix stiffness and shear forces in hemostasis and arterial thrombosis.