The bone marrow stroma in hematological neoplasms—a guilty bystander

Introduction

The interaction between hematopoietic stem cells (HSCs) and bone marrow (BM) mesenchymal cells occurs within functional niches and is responsible for the maintenance of BM homeostasis. Two distinct niches with specialized microenvironments can be identified in the BM that are able to support localization, survival, and self-renewal of stem cells—the osteoblastic (or endosteal) niche and the vascular niche.¹ The osteoblastic niche, in close proximity to bone trabeculae and mainly composed of endosteal fibrocytes and osteoblasts, is associated with HSCs showing a maturation gradient towards the intertrabecular areas of the BM parenchyma.² The vascular niche displays a broad distribution along the BM parenchyma and is composed of endothelial cells and other vascular components, such as smooth muscle cells and adventitial reticular cells.³ The microenvironment of each niche is interconnected, forming a complex network, but the two niches differentially influence HSCs. The osteoblastic niche favors HSC homing during development through chemotactic signals that originate from the interaction between CXCL-12 (SDF-1) and its receptor CXCR-4 on HSCs. This process is recapitulated by transplanted HSCs and several neoplastic populations that display BM tropism.⁴ The osteoblastic niche also provides a nest for HSCs, in which signals from endosteal fibroblasts and osteoblasts maintain their quiescence while enabling self-renewal. These signals include cell–cell and cell–matrix adhesive interactions mediated by β1 integrins (α4, α7, and α9) and their ligands VCAM-1, type I collagen, tenascin-C, fibronectin, laminin, and osteopontin.⁵ In addition to cell adhesion pathways, the Notch pathway is involved in both BM homeostasis and in the pathogenesis of lymphoid and myeloid malignancies.⁶ In particular, engagement of Notch on HCSs prevents HSC differentiation, and contributes to their quiescence in the osteoblastic niche.^5,7 In addition to cellular components, soluble factors such as growth factors, cytokines, and matrix components also contribute to the composition of the BM osteoblastic and vascular niches to differently regulate HSC proliferation, mobilization, and differentiation dynamics (Table 1).

The vascular niche delivers maturation and activation signals to differentiating cells and controls the intravasation and extravasation of BM and foreign elements.² Blood vessels are evenly distributed throughout the BM parenchyma and considerable vascularity is also observed in the BM osteoblastic niche (Figure 1), where the unique composition of the extracellular matrix (ECM) and soluble mediator milieu renders the osteoblastic niche functionally different from that of the vascular niche. Cells that are characteristic of the vascular niche include endothelial cells, BM mesenchymal stem cells (BM-MSCs) that express the markers Nestin and CD146, and stromal cells expressing CD10, which all display different activation and adhesion phenotypes that influence their preference for selective hematopoietic elements.^8,9 Lineage fate can be determined by stromal cues. Multipotent, transformed CD34⁺ human cord blood cells transplanted into immunodeficient mice either lacking β2-microglobulin or expressing myeloid growth factors resulted in leukemias of lymphoid or myeloid lineages, respectively.¹⁰

The adipocyte component of the BM cannot be easily assigned to either of the two specialized niches. BM adipocyte content shows an inverse correlation with the hematopoietic cellularity in physiological (age-related changes) and pathological (BM failure) conditions. Adipocytes have an active role in regulating hematopoietic homeostasis through a direct suppressive influence on HSCs by reducing G-CSF and GM-CSF, and the release of adiponectin and tumor necrosis factor (TNF).¹¹

The BM stroma has a critical role in the maintenance, renewal, and differentiation of HSCs. In lymphoid malignancies, the crosstalk between the BM stroma and neoplastic cells eventually contributes to the establishment of a permissive microenvironment for clone survival and progression. In myeloid malignancies, the stromal involvement is frequently characterized by profound changes, including angiogenesis, fibrosis, and/or new bone formation (osteosclerosis), which may retain clinical and prognostic significance as predictors of BM failure.^12–15 We highlight the role of the BM stroma in the development and course of lymphoid, plasma cell, and myeloid malignancies. The direct interactions of neoplastic elements with the BM stroma, its clone-related modifications, and its emerging immunomodulatory role eventually favoring the generation and progression of clones are discussed.

The BM stroma in lymphoid neoplasms

BM infiltration is a common feature of lymphoproliferative disorders, characterized by specific patterns depending on the disease entity. The relationship between BM infiltration and other histological parameters (including tissue architecture, stromal organization, and fibrosis) possibly reflects specific interactions with microenvironmental factors of the BM.^16–18 In most low-grade lymphomas, BM infiltration is associated with blood vessel proliferation and marked differences in the distribution of lymphoma and stromal cells.¹⁹ Aggregates in follicular lymphoma, lymphoplasmacytic lymphoma, and mantle-cell lymphoma seem to maintain contact, at least initially, with the BM stromal architecture, whereas aggregates in chronic lymphocytic leukemia (CLL) and marginal zone lymphomas predominantly displace the pre-existing stroma.¹⁹ Several factors, including adhesion molecules, contribute to the variation in BM-infiltration dynamics among lymphoid neoplasms. Indeed, leukemias and lymphomas that express BM-homing markers, such as CD44 (hyaluronan receptor) and/or CD22, but lack adhesive ligands required for the stable adhesion to the sinusoidal endothelium (E-selectin and P-selectin ligands), display a specific intrasinusoidal infiltration (Figure 1c).²⁰

Lymphoma cells that spread to the BM interstitium or form nodules are associated with the expression of a broader spectrum of adhesion markers, including P-selectin glycoprotein ligand-1 (CD162), and the integrins α_Lβ₂ and α₄β₁ (ligands of VCAM-1 and ICAM-1, respectively).²⁰ B-cell and T-cell leukemias and lymphomas characterized by an intrasinusoidal infiltration pattern in the BM, such as splenic marginal zone lymphomas, hairy-cell leukemia, and hepatosplenic T-cell lymphoma, frequently involve the spleen and the liver, where they display a prominent intravascular distribution.²¹ Such characteristic behavior might imply a neoplastic cell preference for the BM vascular niche, which shares several features with the spleen and liver sinusoidal compartment, characterized by fenestrated sinusoids abundantly exposing hyaluronan and prominent subendothelial reticular cells expressing CXCL12. The tight interaction with the vascular niche through the CXCL12/CXCR4 axis may represent a drug resistance mechanism, and anti-CXCR4 therapeutic strategies have been proposed.²²

Other lymphomas that infiltrate the BM, display a specific paratrabecular pattern in the osteoblastic niche (Figure 1d).^16,17 Non-Hodgkin lymphomas derived from germinal centers are often associated with large infiltrates, BM fibrosis, and megakaryocytic hyperplasia.¹⁷ The prototypic example is follicular lymphoma, commonly presenting with paratrabecular aggregates containing small T lymphocytes, probably recruited through CCL17/CCL22, and possibly sustaining the neoplastic clone.²³ A similar BM histopathology also extends to other B-cell malignancies (for example, T-cell/histiocyte-rich diffuse large B-cell lymphoma) and T-cell lymphomas of follicular derivation (angioimmunoblastic T-cell lymphoma).^24–26 The extracellular meshwork of lymphoid follicles is associated with the stromal follicular dendritic cell and microvascular network, and shares several components with that of the BM osteoblastic niche, including integrin-β₁, type I collagen, and matricellular protein expression.^27,28 These elements might provide a common adhesive background to neoplastic cells by favoring their location in the BM osteoblastic compartment, and signaling through the Notch stromal ligands Delta-like-1 and Jagged-1.²⁹

Thus, the BM stroma may condition the degree and pattern of lymphoid infiltration and contribute to neoplastic cell survival. Notably, besides selectively homing to specific favorable microenvironments, neoplastic lymphocytes can actively promote BM stroma modifications to produce an ad hoc neoplastic niche.³⁰ These modifications mainly include blood vessel formation, and changes in the adhesion profile of the stromal meshwork (Figure 2).

In xenograft models of B-cell and T-cell acute lymphoblastic leukemia (ALL), leukemic cells affect the BM vascular niche by promoting angiogenesis through the release of proangiogenic molecules, including VEGF, bFGF, angiopoietins, metalloproteinases, interleukin (IL)-6 and IL-8, and eventually establishing a niche that promotes ALL survival via IL-7 and IL-3.³⁰ The leukemic niche is characterized by low HSC support capability owing to down-regulated CXCL12 stromal expression, yet it outcompetes native stromal niches for HSC engraftment owing to production of stem-cell factor by neoplastic cells.³¹ In addition to ALL, increased BM angiogenesis is associated with unfavorable prognosis in several other lymphoid malignancies, including CLL.^32,33 In specific settings, such as in angioimmunoblastic T-cell lymphoma, overexpression of VEGF may not only sustain the increase in microvessel density, but also promote tumor cell proliferation and survival by autocrine and paracrine loops.³⁴ Expression of CD38 and CD49d adhesion molecules by the neoplastic clone in CLL has been associated with a more-aggressive disease course.³⁵ CD38 engagement by CD31 favors the synthesis and release of CCL3 and CCL4 chemokines by neoplastic cells, which induce the recruitment of CD68⁺ macrophages at sites of lymphoid infiltration.³⁵ Recruited macrophages influence the activation of BM stromal cells, inducing expression of VCAM-1 and release of proinflammatory cytokines such as TNF (Figure 2).³⁵ A strong prosurvival feedback signal is delivered to CLL cells by the concomitant CD38/CD31 and CD49d/VCAM-1 pathway. Similarly, other lymphoid neoplasms may benefit from the recruitment of cellular partners for the induction of stromal modifications at sites of infiltration. This process may occur in lymphoplasmacytic lymphoma and in angioimmunoblastic T-cell lymphoma, in which lymphomatous infiltrates may be enriched by mast cells attracted by neoplastic cell-derived CCL5 and CXCL13.³⁶ Lymphoma-infiltrating mast cells may foster neoplastic growth and survival through the CD40L/CD40 or OX40L/OX40 axes, and also contribute to the orchestration and maintenance of a supportive stromal microenvironment through the synthesis of mediators, including TNF, IL-6, matrix metalloproteinase-9, and interferon-γ.³⁶

Nodular BM infiltrates from mature B-cell lymphoproliferative disorders may foster a meshwork of mesenchymal cells that express markers of follicular dendritic cells, such as CD21, conferring to neoplastic aggregates the appearance of ‘true follicles’ (Figure 1e). Although the occurrence of a follicular organization in BM lymphoid infiltrates has never been related with specific biological features of the neoplastic clone, interplay between such follicular dendritic cells and neoplastic cells can be predicted in terms of cell adhesion (for example, mediated by ICAM-1 expression) and antigenic stimulation (Figure 1f). Interestingly, signals stemming from surface antigen receptors and signals derived from the actin network may support activation and survival of lymphoid cancer cells through a common pathway.^22,37

BM stromal changes associated with lymphoid neoplasms might also comprise reticulin fibrosis and fibroblastic proliferation. The former is often slight to moderate and associated with specific lymphoproliferative neoplasms, such as hairy-cell leukemia or peripheral T-cell lymphomas, whereas the latter is mainly observed in BM infiltrates from classic Hodgkin lymphoma.³⁸ Importantly, the degree of fibrosis may be helpful in differentiating non-Hodgkin lymphoma and benign lymphoid hyperplasia characterized by central localization and absence of considerable amounts of reticulin.³⁹

Available evidence indicates that neoplastic lymphocytes can shape the neighboring BM microenvironment by directly affecting stromal composition or co-opting other immune effectors that are active in tissue remodeling (Figure 2). Notably, such changes in the BM microenvironment eventually favor neoplastic cell survival and influence the clinical course of the disease.

The BM stroma in multiple myeloma

Although the BM is the primary site of multiple myeloma, it is not the site of origin. Emerging data indicate that the malignant plasma cell arises from a postgerminal center B cell that has undergone somatic hypermutation, lacks intraclonal variation, and expresses CD19 and CD27, consistent with a central memory B-cell phenotype.⁴⁰ Interestingly, several groups have identified critical cell populations for multiple myeloma BM engraftment, ranging from CD38^high/CD19⁺/CD138⁻ to CD19⁺/CD27⁺/CD138^{−.41,42,43,44} This latter population has self-renewal potential, which supports the hypothesis of a memory B cell as the cell responsible for sustaining the malignant phenotype in plasma cell dyscrasias. The presence of this population in the peripheral circulation raises the question of what mechanisms regulate trafficking and homing to the BM in plasma cell dyscrasias and how neoplastic-cell survival is maintained within the BM microenvironment.

Many factors are involved in the crosstalk between multiple myeloma cells and the BM stroma, including soluble factors, such as chemokines, cytokines, growth factors, cell-surface receptors, ECM proteins, and cellular constituents of the BM stromal niches. BM stromal cells contribute to IL-6 synthesis, which in turn stimulates multiple myeloma-cell proliferation and confers chemoresistance through activation of pathways, such as PI3K/Akt, MAPK, and Janus-kinase/STAT3, and synergizes with adhesive β₁-integrin signals to increase survival of multiple myeloma cells.^45,46 Interestingly, IL-6 production by BM stromal cells requires multiple myeloma-cell binding and subsequent activation through nuclear factor kappa B signaling.⁴⁷ IGF-1 can overcome the effects of IL-6-blockade in multiple myeloma through binding to its receptor IGF-1R.⁴⁸ IGF-1R expression has been associated with more-aggressive disease and is found to a greater extent in the poor prognosis t(4;14) multiple myeloma.⁴⁸ High circulating levels of IGF-1 are associated with a dismal prognosis.⁴⁹ Through binding of IGFBP-3, IGF1 is targeted to multiple myeloma cells that express high levels of the cell surface adhesion molecule syndecan-1 (CD138). CD138 mediates IGF-1 uncoupling from IGFBP-3 and enables IGF-1 interaction with its receptor on multiple myeloma cells, in turn promoting cell survival and proliferation.⁴⁸ Furthermore, soluble CD138 also binds IGF-1, leading to further increased local delivery of IGF-1 to myeloma cells, representing a link between adhesion and soluble mediator-induced cell survival.⁴⁸

The BM stroma in myeloid neoplasms

Myeloid neoplasms are a heterogeneous group of clonal HSC disorders. Although alterations in the BM stroma in these neoplasms are considered secondary to cytokine effects, the BM stroma is not just an ‘innocent bystander’ but has an active role in the pathogenesis and disease progression, affecting both clinical course and response to therapy. The major morphologically recognizable BM stromal alterations include fibrosis (Box 1), angiogenesis, and osteosclerosis. Such stromal changes have been described in acute myeloid leukemia (AML), myelodysplastic syndromes, and myeloproliferative neoplasms.

Immunomodulation by BM stromal cells

BM stromal cells have a distinct immunological role from that of stromal cells of secondary lymphoid organs, which organize the memory response, or initiate the adaptive immune response, respectively.⁹⁸ Specialized BM stromal cells provide a survival niche for plasma cells and helper T lymphocytes. Similar mechanisms can operate to support survival of lymphoid malignant clones, whereas opposing mechanisms could enable hematological malignancies to escape from the control of the immune system.

Immune escape could occur in malignancies whose immunogenicity is confirmed by the identification of specific tumor-associated antigens. Chromosomal translocation can generate new antigens potentially perceived as non-self by the immune system. An example of this is the BCR-ABL fusion protein in CML, which is recognized by T lymphocytes.^99,100 Also, the proteins elastase and proteinase 3 induce cytotoxic T-cell responses against CML.¹⁰¹ In both AML and CML, the Wilms tumor gene product, WT1, is a target of cytotoxic T cells and a potential candidate for immunotherapy.¹⁰² These examples support the existence of immune suppression, which enables escape from immune control.

In the BM environment, BM-MSCs are capable of immune regulation in several settings, including transplantation, tolerance and, indeed, tumor evasion.¹⁰³ BM-MSCs exert their immunosuppressive functions directly on B cells, T cells, and dendritic cells via the release of specific soluble mediators, and indirectly through regulatory cells such as T regulatory (T_REG) cells and, probably, myeloid-derived suppressor cells (Figure 4).

BM-MSCs can inhibit the proliferation of T cells in response to alloantigens, mitogens, CD3 and CD28 as well as the proliferation and immunoglobulin secretion of reactive B cells to CD40 or IL-4.^104–106 The most relevant mediators of such effects include TGF-β, hepatocyte growth factor, IL-10, IL-6 and prostaglandin E2.¹⁰⁶ BM-MSCs also secrete galectin-1 and semaphorin-3A, which inhibit T-cell proliferation via neuropilin-1.^107,108 IL-6 derived from BM-MSCs, in particular, seems to inhibit the differentiation of CD34⁺ and monocyte precursors into mature dendritic cells, maintaining them in an immature, tolerogenic state.¹⁰⁹ Dendritic-cell differentiation from HSCs requires the activation of the Notch pathway and Wnt, whose ligands are produced by nearby stromal cells, a regulatory model that further implicates BM-MSCs in the immune response.¹¹⁰

BM-MSCs induce a T_REG cell fate from CD4 or CD8 T cells upon their interaction.¹¹¹ Once T_REG cells are activated, they suppress nearby lymphocyte and natural killer cells and as such can also offer escape from immune control to tumors. Tumor progression is associated with cell compartment, partially the expansion of the T_REG explaining the tumor-associated immune suppression; whether T_REG cells have a role in the initial phase of transformation and immune surveillance is less clear.¹¹² Regardless of their thymic origin or peripheral conversion, T_REG cells are endowed with plasticity and sensing different microenvironments can deflect toward the prevalent cytokine milieu including the inflammatory Th17.¹¹³ Even in the context of lymphoid neoplasms, such as angioimmunoblastic T-cell lymphoma, T_REG cells can be skewed towards Th17 cells, which in turn maintain the proinflammatory conditions and promote remodeling of the infiltrated tissue.³⁶ In hematological malignancies, increased numbers of T_REG cells have been reported in patients with MGUS, multiple myeloma, and AML.^114,115 AML cells can induce T-cell tolerance by directly converting CD4⁺CD25⁻ T cells into T_REG cells through indoleamine 2,3-dioxygenase,¹¹⁶ a mechanism that, originally described immunosuppressive in placenta, might characterize BM-MSCs.^117,118

The other immunosuppressive population potentially induced by BM-MSCs in the BM environment is that of myeloid-derived suppressor cells. In the presence of solid tumors, the BM responds and produces abundant immature myeloid cells unable to complete their differentiation program and collectively named myeloid-derived suppressor cells, which provide potent immune suppression.¹¹⁹ Factors inducing the expansion of myeloid-derived suppressor cells are prostaglandins, IL-6, GM-CSF, and indoledamine 2,3-dioxygenase produced by BM-MSCs, which indicates a potential complementary mechanism through which BM-MSCs might induce a tolerogenic BM microenvironment.¹¹⁹ Myeloid-derived suppressor cells can inhibit immune cell responses through different mechanisms: T-cell activation is suppressed by the production of arginase and reactive oxygen species, the nitration of the T-cell receptor, cysteine deprivation, and the induction of T_REG cells. Innate immunity is impaired by the down-regulation of IL-12 from macrophages, the increase in IL-10 from myeloid-derived suppressor cells, and the suppression of natural killer cell cytotoxicity. Antigen presentation is limited by the expansion of myeloid-derived suppressor cells at the expense of dendritic cells.¹²⁰

The outcome of the immunosuppressive functions of BM-MSCs in vivo has been investigated in animal models and human pathological settings. In rats and mice, administration of BM-MSCs ameliorated autoimmune myasthenia gravis or enabled successful immune escape of cancer cells in an allogenic background, respectively.^121,122 Moreover, BM-MSCs have been used in patients with severe graft-versus-host disease following solid organ and HSC transplantation, where they proved to be safe and showed substantial clinical efficacy in association with conventional immunosuppressive therapy.¹²³ The immunomodulatory properties of BM-MSCs discussed above may also impact on the establishment and progression of hematologic neoplasms. T_REG cells, which might be induced directly by BM-MSCs from effector cells, or recruited in the BM via the CXCL12/CXCR4 axis, may directly control the proliferation of the malignant clone and help maintain an homeostatic condition in the BM. They might also impair the establishment of an effective antitumoral immune response, thus favoring uncontrolled growth of the clones.^115,124 In myeloid malignancies, the detrimental contribution of BM-MSC-induced and/or BM-MSC-recruited T_REG cells might be enriched by the direct participation of T_REG cells to myeloid cancer-related stromal changes via the release of TGF-β, which directly contributes to remodeling of the BM stroma.

Conclusions

In hematological malignancies, the BM stroma is not an innocent bystander, but rather an accomplice of tumor development by favoring neoplastic cell growth and survival in the BM microenvironment, co-evolving alongside tumor progression, and forming part of the tumor microenvironment at extramedullary sites. Thus, treatment strategies that interfere with the major axes involved in crosstalk between neoplastic cells and BM stromal cells, such as monoclonal antibodies that target cell-adhesion or cytokine pathways, may prove effective in lymphoid malignancies, when combined with therapies that target the neoplastic clones. Efforts should also focus on limiting the stromal alterations that eventually determine BM failure in several types of myeloid neoplasms. The adoption of drugs that control the proliferation of BM stromal and vascular cells, such as tyrosine kinase inhibitors, proteasome inhibitors, and immunomodulatory agents could represent a promising perspective.