**3. Blood Cancers**

Blood cancers, non-solid tumors or hematological malignancies are a collective term for neoplastic diseases of the hematopoietic and lymphoid tissues with a clinical presentation as leukemia, lymphoma or myeloma [25]. In humans, definitive adult hematopoiesis is established in the bone marrow (BM) and thymus [26]. Through this process, all blood cells arise from the hematopoietic stem cell (HSC). Precursor cells proceed through specific maturation steps before leaving the BM as mature circulating cells. Two major lineages exist: thrombocytes, erythrocytes, granulocytes, and dendritic cells are derived from the common myeloid progenitor (CMP) and make up the myeloid lineage, while Tand B-lymphocytes, plasma cells and natural killer cells arise from the common lymphoid progenitor (CLP) and compose the lymphoid lineage [26,27]. Lineage and maturation stages can be assessed by morphologic, immunophenotypic, genetic and cytochemical features. Hematological malignancies are categorized by the same methods according to cell origin, maturation or tumor characteristics [25]. The World Health Organization (WHO) has published a unified classification of neoplastic diseases of the hematopoietic and lymphoid tissues [28,29]. Two major classes of cell types are primarily affected: myeloid neoplasms or lymphoid neoplasms. And then, rare histiocytic, dendritic or mast cell neoplasms. Each of these major groups is categorized in di fferent subclasses, the most common of which are summarized in Figure 1.

**Figure 1.** Representation of BM hematopoiesis together with the simplified WHO classification of neoplastic diseases of the hematopoietic and lymphoid tissues [28,29]. Blue box contains lymphoid neoplasms, orange box myeloid neoplasms and green box rare histiocytic, dendritic or mast cell neoplasms. Abbreviations: del, deletion; HSC, hematopoietic stem cell; CMP, common myeloid progenitor; CLP, common lymphoid progenitor; MEP, megakaryocyte-erythrocyte progenitor; GMP, granulocyte-monocyte progenitor; TNKP, T-natural killer cell progenitor; BCP, B cell progenitor; NK, natural killer.

Although leukemias, lymphomas, and myelomas share some common features, they also have major di fferences, with similarities and di fferences in each disease group. Leukemia (a term derived from the Greek words "leukos" and "heima") refers to an excess of leukocytes in the body and originates in the BM. It can arise in either of two main groups -lymphocytes or myelocytes- and can be acute (a rapidly progressing form in which affected cells are very immature and unable to accomplish their function) or chronic (which progresses slowly from cells that are relatively differentiated but crudely functional). Lymphoma involves lymphocytes and is initiated in lymphoid tissues. Non-Hodgkin's lymphoma is the most prevalent form, with more indolent forms that progress slowly with well differentiated cells, and the more aggressive forms with less differentiated lymphocytes. Hodgkin's disease has different clinical features and is characterized by the presence of the distinctive Reed-Stenberg Cells. Myeloma are plasma cell disorders characterized by clonal proliferation of malignant cells, normally in the BM [30]. In leukemia, the cancerous cells are discovered circulating in the blood and BM, while in lymphoma cells tend to aggregate and form tumors in lymphatic tissues. Myeloma is mainly a tumor of the BM (Box 1). This review will focus on hematological malignancies with BM homing, how they are supported by BM niches, and explains how tissue engineers could take advantage of these interactions and architectures to recapitulate the malignant process in vitro.

## **Box 1.** Blood cancers.

"In leukemia, the cancerous cells are discovered circulating in the blood and BM, while in lymphoma cells tend to aggregate and form tumors in lymphatic tissues. Myeloma is mainly a tumor of the BM."

#### **4. Bone Marrow Microenvironment, Home of Hematological Malignancies**

#### *4.1. Healthy BM Niche: An Intricate and Precisely Organized Network Sustaining HSCs Homeostasis*

To ensure effective hematopoiesis throughout an individual's lifespan, HSCs are tightly regulated in their complex BM niche (BMN), which keeps a balance between maintaining the HSC population and producing mature immune and blood cells. Stromal cells, ECM and biochemical gradients orchestrate the regulation between quiescence and self-renewal that contribute to maintaining the HSCs vs. their activation, lineage commitment and terminal differentiation. The hematopoietic system has developed an adaptive response capacity in order to maintain homeostatic replacement of blood and immune cells (steady-state hematopoiesis), and also to rapidly increase differentiated cell production in a context of acute blood loss, infection and metabolic or toxic stress (emergency hematopoiesis). Aberrant prioritization of differentiation over HSC self-renewal and quiescence leads to the exhaustion of the HSC compartment, while inhibition of differentiation involves ineffective blood production. In both cases, the hematopoietic system becomes exhausted, leading ultimately to BM failure and hematological malignancies. Under physiological conditions, functional HSC heterogeneity appears to be controlled by spatially-different niches. HSCs with diverse differentiation biases (myeloid vs. lymphoid) or different fates (quiescent vs. activated) have been found to lodge in different anatomical locations under the control of different stimuli [31]. Despite the controversies, it is generally accepted that there are two main HSC niches in the BM: the endosteal BMN (EBMN) in the endosteum, with low vascularity (arteries and arterioles) and the central BMN (CBMN), with higher vasculature, arterioles and sinusoids, and enriched in HSCs. In both niches HSCs reside mainly in perivascular areas where endothelial cells, different stromal cells of mesenchymal origin, neurons or Schwann cells and mature blood or immune cells critically regulate their location, number and fate (Box 2). They are also governed by other factors such as the abundance of oxygen tension, shear flow or reactive oxygen species (ROS) (Figure 2).

**Figure 2.** Scheme of bone vasculature structure and healthy BMN. (**<sup>a</sup>**,**b**) BMN location and vasculature architecture. Arterial vessels that penetrate cortical bone merge and form the central artery. Arterioles branch from the central artery toward cortical bone and anastomose with the sinusoid, which then connects with the central vein in bone surface proximities. The vascular structure of the medullary canal in bones contributes to spatial delimitation of the BM HSC sub-niches. (**c**) Representation of cell populations (bold text) and their locations in BMN; main cell-cell interactions by direct contact and soluble factors and their role in HSC homeostasis (green) are detailed. Abbreviations: EBMN, endosteal bone marrow niche; CBMN central bone marrow niche; ROS, reactive oxygen species; HSC, hematopoietic stem cell; CMP, common myeloid progenitor; CLP, common lymphoid progenitor; aEC, arteriolar endothelial cell; sEC, sinusoidal endothelial cell; LepR+ CAR MSCs, leptin receptor expressing cells and abundant reticular mesenchymal stem cells; Nest<sup>+</sup> NG2+ MSCs, nestin and neural glial antigen expressing mesenchymal stem cells; SNS, sympathetic nerve fiber; nmSCs, non-myelinating Schwann cells; OBs, osteoblasts; OCs, osteoclasts; CXCL-12, CXC motif chemokine ligand 12; SCF, stem cell factor; ANG-1, angiopoietin 1; TIE-2, angiopoietin receptor; IL-18, interleukin 18; FGF-2, fibroblast growth factor 2; TGF-β, transforming growth factor beta; TPO, thrombopoietin; PF-4, platelet factor 4; JAG-1, Jagged-1; G-CSF, granulocyte-colony stimulating factor.

The central niche comprises > 90% of the BM volume, and shelters 85% of the HSCs. It contains most sinusoids and arterioles. Specific and di fferentiated functions are linked to arteriolar (aBMN) or sinusoidal (sBMN) BM niches. Sinusoidal areas are tightly related with myelopoiesis and contain CMPs and steady-state hematopoiesis. In sBMN mesenchymal stem cell (MSCs) subpopulations such as leptin receptor expressing cells and abundant reticular cells are mainly overlapping (LepR+ CAR MSCs) and produce crucial growth factors for HSC maintenance as CXC motif chemokine ligand 12 (CXCL-12) and stem cell factor (SCF). In arteriolar perivascular areas of CBMN and EBMN, nestin and neural glial antigen expressing cells (Nest+ NG2+ MSCs) are the most abundant MSCs subpopulation and also produce soluble factors involved in HSC maintenance, such as CXCL-12 and SCF. Sinusoidal (sEC) and arteriolar (aEC) endothelial cells have also been reported to provide soluble factors related to HSC maintenance such as CXCL-12, SCF, angiopoietin (ANG-1) or fibroblast growth factor 2 (FGF-2). However, they have phenotypical di fferences that determine di fferent functions; aECs form less permeable blood vessels. High permeability of adjacent vessels increases ROS intracellular levels in HSCs, increasing their migration capacity while compromising their self-renewal, hampering quiescence and accelerating di fferentiation. Sinusoids therefore promote HSC activation and lodge immature and mature leukocyte tra fficking to and from the BM, while more quiescent HSCs are found adjacent to arterioles. The ROSlow HSCs found in sinusoidal areas reside near megakaryocytes, which in the BM niche promote HSC quiescence through secretion of transforming growth factor beta (TGF-β), thrombopoietin (TPO) or platelet factor 4 (PF-4). sECs also express higher levels of E-selectin (adhesion molecule that supports HSC activation) than aECs [32,33]. Sympathetic nerve fibers innervate BM perivascular areas and contribute to their functional di fferentiation. Non myelinating Schwann cells (nmSCs) in sBMN regulate migration of HSCs by direct contact [31], while Nest<sup>+</sup> nmSCs maintain hibernating HSCs in aBMN [34]. nmSCs associated with sympathetic nerve fibers can promote HSC quiescence by activating latent TFG-β in EBMN and CMBN [33]. Sympathetic nerves release soluble mediators such as catecholamines from terminals. Noradrenaline reduces CXCL-12 production by di fferent niche-forming cells [35], inducing HSC activation (Figure 2c).

The EBMN is a much smaller niche (less than 10% of total BM volume) with 15% of the total HSC population. Several EBMN signals are related to the promotion of HSC quiescence, essential to preserve hematopoiesis normalcy through lifespan. It is related to lymphopoiesis and harbors CLPs [31]. Regulation of hematopoiesis by osteoblasts (OB) may depend on their di fferentiation state [36]. The contribution of the osteoblastic niche to HSC maintenance remains controversial due to OB population heterogeneity and varying degree of maturation. OBs produce cytokines and growth factors that promote HSC self-renewal and quiescence. For example, ANG-1 interacts with ANG-1 receptor TIE-2 on HSCs to promote quiescence and adhesion. Increased expression of Jagged-1 (JAG-1) in OBs simultaneously increases their number and HSC self-renewal through Notch-1 signaling activation in HSCs [36]. OBs also appear to regulate HSC homing and engraftment in the EBMN after HSC transplantation. Osteocytes, mature bone cells entrapped in the calcified bone matrix, control HSCs through the secretion of the granulocyte-colony stimulating factor (G-CSF). Osteoclasts (OCs), bone degrading cells of monocytic origin, also a ffect hematopoiesis [36].

Non-cellular elements such as oxygen level also a ffect HSCs in di fferent niches. The hematopoietic compartment is relatively hypoxic, which contributes to HSC pluripotency by mechanisms such as reducing intracellular ROS. The oxygen level close to arterioles is higher than in perisinusoidal regions [32,35]. The ECM produced by niche cells provides structural integrity and has a regulatory effect on niche-forming cells and HSCs [18]. The most abundant proteins are FN, collagens (COL) from I to XI, tenascin, osteopontin, thrombospondin or elastin. Also important are proteoglycans, which can present glycosaminoglycan side-chains such as HA, chondroitin sulfate, heparan sulphate or heparin. ECM interactions with cells are mediated by integrins or selectins and membrane-bound immunoglobulins such as the intercellular adhesion molecule 1 (ICAM-1) or vascular cell adhesion molecule 1 (VCAM-1) [32,37]. Proteoglycans and collagens have been shown to be essential for HSC maintenance; for example HA is required for in vitro hematopoiesis. As in cell-cell interaction, cell-ECM interactions are di fferentially regulated during steady-state or emergency hematopoiesis [37]. The mechanical properties of BM determined by ECM composition have also been shown to a ffect HSC fate. HSC cultivation in di fferent substrates of varying sti ffness and topographies have shown that matrix density and biophysical properties, e.g., specific presentation of adhesion ligands, contribute to HSC niche modulation, either directly or indirectly. ECM density and components also modulate direct cell-cell communications by conditioning secreted factors [37].

#### **Box 2.** Bone marrow microenvironment.

Healthy BMN in a nutshell. "HSCs maintenance vs. di fferentiation is tightly regulated in their complex BM niche ( ... ). HSCs reside mainly in perivascular areas, where ECs, di fferent stromal cells of mesenchymal origin, neurons or Schwann cells and mature blood or immune cells and factors as oxygen tension, shear flow or ROS regulate their location and fate ( ... ). The central BMN is related with myelopoiesis and steady-state hematopoiesis; sinusoidal areas promote HSCs activation and lodge immature and mature leukocyte tra fficking while more quiescent HSCs are found adjacent to arterioles. In bone proximities, the endosteal BMN is related with promotion of HSCs quiescence".

#### *4.2. BM Niche in Hematological Malignancies: When the Regulatory Machinery Becomes the Perfect Tumor Partner*

Multiple intricate interactions maintain hematopoiesis physiology and also the pathophysiology of hematological malignancies, which alter the BMN and its normal interactions and so contribute to tumor progression. Although BMN was suggested to influence tumor initiation twenty years ago its role is still not clear. However, there is significant evidence of BMN and malignant cell crosstalk being involved in tumor progression and resistance to therapies [31]. Transformed cells compete with HSCs to occupy the BMN during the progression of the disease, which disrupts physiological interactions, reduces the HSC population and disturbs normal hematopoiesis. For example, malignant cells have a dependence on canonical HSC niche pathways, such as CXCL12- C-X-C chemokine receptor type 4 (CXCR4). However, as the disease progresses transformed cells progressively become independent of BMN control.

Niche-driven transformations are mutations or functional alterations in BMN-forming cells that predispose to myeloid malignant tumors. The first indication was provided by two studies reporting that genetic ablation of the tumor suppressor gene retinoblastoma (Rb) [38] or retinoic acid receptor γ (Rarγ) [39] in mice induced myeloproliferative neoplasms (MPN), even though their development required inactivation of either of these genes in hematopoietic and BMN-forming cells. Development of MPN from Rarγ mutations needs increased levels of the tumor necrosis factor (TNF) [39], depicting a pro-inflammatory environment as a crucial triggering factor, together with BMN alteration. As inflammation is a hallmark of aging, and myeloid malignancies are more prevalent among the elderly, inflamed BMN could be used as a tumor-initiating factor. As BMN aging promotes myeloid biases at the expense of lymphoid di fferentiation, some authors sugges<sup>t</sup> that it facilitates pre-malignant clone growth by overstimulating myeloid cell expansion, which can lead to myeloid malignancies. This means that myeloid malignancies may develop over a period of years as a continuous process involving simultaneous mutation of pre-malignant cells caused by certain BMN alterations [40–42]. There is further evidence that BMN is a predisposing factor from the fact that many recipients contract leukemia after human clinical halogenic HSC transplantations as a result of the transformation of healthy HSCs into malignant clones [43].

BMN remodeling by malignant cells contributes to disease progression. Malignant cells alter the transcriptome, proteome and function of BMN-forming cells by means or secreted factors, exosomes or direct cell-cell contact and promote BMN changes towards angiogenesis and inflammation [44–46]. MSCs, adipocytes, OBs, ECs and sympathetic neurons or Schwann cells are also a ffected. The reprogramming of BMN-forming cells has been described in myeloid and lymphoid malignancies. In the case of myeloid malignancies, chronic myeloid leukemia (CML) cells activate MSCs through soluble factors like CC motif ligand 3 (CCL3) or TPO, and by direct cell-cell contact causing overproduction of functionally altered OBs that do not support normal HSC maintenance [47]. In lymphoid malignancies, T cell acute lymphoblastic leukemia (T-ALL) cells impair normal hematopoiesis with a dramatic loss of OBs [48]. In CLL, exosomes from neoplastic clones reduce the production of soluble HSC supporting factors, and CLL cell-derived vesicles a ffect immune cell function as natural killer cells and accelerate the transition of stromal cells towards cancer supportive phenotypes, up-regulating production of angiogenic factors by ECs and MSCs [49]. Sympathetic neurons, which innervate BMN, also become altered by malignant cells in mouse-model MPN. Transformed HSCs produce IL-β1, which damages sympathetic neurons and kill Schwann cells, leading to reduced CXCL-12 production and promoting mutant HSC proliferation [50]. Hypoxia and angiogenesis are two important traits of hematological malignancies, neovascularization induced by malignant cells provides increased nutrients and oxygen to supply the higher demand, but it also encourages the arrival of soluble factors that promote the survival, proliferation and chemoresistance of malignant cells. For example, vascular endothelial growth factor A (VEGF-A) secretion by transformed HSCs increases angiogenesis, but also stimulates its proliferation [51]. Under physiological conditions HSCs preferentially use glycolysis to avoid excessive ROS production and maintain quiescence. Malignant AML cells rely on di fferent regulatory mechanisms to gain metabolic plasticity. Unlike their normal counterparts, malignant cells tend to have a high glucose uptake and glycolytic rate in the presence of oxygen, known as the Warburg e ffect (aerobic glycolysis) [52]. Although this results in low ATP yields, the Warburg e ffect is an essential anabolic mechanism that allows cancer cells to manage cell growth and division [53]. However, the field of tumor metabolism is extremely heterogeneous and specific fractions of malignant cells exhibit increased reliance on oxidative phosphorylation [54]. Some studies report that malignant cells uptake functional mitochondria from niche-forming cells using endocytic pathways to satisfy the greater demand for energy, which provides a survival advantage as it promotes ROS detoxification, resistance to chemotherapy and the aggressiveness of the disease [53].

Once transformed, BMN favors malignancy though di fferent mechanisms. In some lymphoid malignancies such as CLL, B cell acute lymphoid leukemia (B-ALL) or mantle-cell lymphoma activation of survival and pro-inflammatory pathways, like the nuclear factor κB (NF- κB) pathway by altered BMN cells is necessary for malignant cell survival. This crosstalk is dependent on cell-cell contact and is also mediated by IL-1 α and IL-15. Previously malignant cells modify BMN cells such as MSCs, inducing protein kinase C (PKCβ) expression [55]. Protection from excessive ROS is a di fferent mechanism by which altered BMN protects malignant cells and promotes survival and chemoresistance. Altered BMN cells provide CLL or B-ALL cells with cysteine, used by malignant cells for production of glutathione for ROS detoxification, a key ability for survival, as chemotherapy e ffectiveness relies on ROS-induced DNA damage [56]. Metabolic reprogramming of malignant cells by altered BMN cells has also been reported. Altered BMN cells such as MSCs also promote immunosuppression and hamper the activity of e ffector lymphocytes by TGF-β, IL-10, prostaglandin E2 (PGE2) or arginase 1 or 2, and help to avoid attacks by malignant cells [57]. All these mechanisms combine to develop therapeutic resistance by malignant cells. The simple grafting of malignant cells onto specific sites could inhibit chemotherapy. CD44 has played a particularly important role in malignant cell interaction with the BMN in myeloid malignancies such as multiple myeloma, CML or AML [58–60]. CD44 binding with its E-selectin EC receptor mediates homing and grafting of malignant cells in CML [59] and enhances drug resistance in multiple myeloma [60]. Likewise, β1 integrins mediate adhesion to VCAM-1 or ICAM-1 molecules on stromal cells and induce CML cells adhesion to BMN. In AML, interaction between α4β1 integrin and VCAM-1 mediates chemoresistance towards activation of the NF-κB pathway in stromal cells [61]. Several studies report how the blockage of these interactions may help to sensitize malignant cells to conventional chemotherapy. In fact, the latest approaches to clinical treatment of hematological malignancies are combined therapies that not only attack malignant cells but also the altered BMN, or more specifically the crosstalk between supporting and malignant cells [62]. The importance of the environment in blood cancer pathogenesis is therefore undeniable (Figure 3). E ffective in vitro

modeling of hematological malignancies with BM homing, mainly myelomas and leukemias, goes through the inclusion into the model of different components from the native BMN.

**Figure 3.** Diagram of malignant cells and BM niche interplay guiding tumor initiation, progression and acquisition of resistance to therapies. Abbreviations: BM, bone marrow; BMN, bone marrow niche; Rb, Retinoblastoma; Rarγ, Retinoic acid receptor γ; HSC, hematopoietic stem cell; ROS, reactive oxygen species; CAM-DR, cell adhesion-mediated drug resistance.

#### **5. Advances in BM Models**

Engineering a BM analog would be important both for basic BMN research and for therapeutic strategies [63]. However, BM structural complexity, its spatially variable anatomy and the intricacies and dynamic character of its cellular interactions and regulation under physiological and especially under pathological conditions make BM and hematological malignancies particularly complicated to tackle for tissue and disease modeling. Despite the complications, TERM has developed tridimensional platforms attempting to overcome the limitations of conventional models, such as animal models or in vitro 2D culture. These platforms have been based on different approaches to provide structural support for HSCs and BMN-forming populations. BM is a complex machine with many different pieces (cellular components, ECM, soluble signals, oxygen level, tissue stiffness, etc.) working in an orchestrated manner to carry out its physiological roles. Different authors have chosen different approaches, in which some pieces conduct the niche's in vitro mimicry that governs the spatial and temporal regulatory signals. Approaches include naïve cultures of hematopoietic progenitors in association with a biomimetic scaffolding material or in a co-culture with specific niche-forming cells through the more sophisticated BM-on-a-chip devices. Nonetheless, the development of a functionally effective model capable of showing BMN diversity and dynamism with the translational potential for disease modeling is still challenging. No single approach has been adopted as the standard in the field [63–65], however, different models have succeeded in reproducing some particular and restricted aspects of specific BM contexts. For example, the first attempts at reproducing cell-cell interactions started with the co-culture of HSCs and different BMN-forming cells, such as ECs, OBs or MSCs [66–68]. Cell-ECM interactions have been extensively incorporated into the equation by the introduction of natural tridimensional substrates or by polymeric synthetic substrates with different biofunctionalizations, all of them commonly presented in the form of scaffolds or hydrogels [69–73] (Table 2). Some of these approaches aim to improve the platforms for in vitro expansion of HSCs and ex vivo platforms for blood production [71,73]. Others aim to reproduce certain complex aspects of native BMN, such as compartmentalization and differential regulation of HSC fate [69,70,72]. All of them represent simplified bio-inspired set-ups that have undoubtedly contributed to increasing our basic knowledge of the hematopoietic compartment. However, their complexity is still far from assimilating BM complexity. Micro-scale systems or microfluidic perfusion chips have been used to mimic BMN complexity. Usually termed as BM-on-a-chip platforms, they include combinatorial approaches of fluid flow, biomimetic sca ffolds and fine control of biochemical parameters (Table 3). These results are a step forward, as they succeed in mimicking functional compartmentalization, in providing knowledge that can be applied clinically or even reproduce the particular native behaviors of some blood cancers [74–78]. Nevertheless, what these di fferent strategies do have in common is that they show that progress in the field relies on interdisciplinary approaches (Box 3). Although micro-scale models o ffer greater control, their design flexibility and significant result assessment are complicated. For the present authors, the reviewed models that most closely resemble BMN complexity rely on genetically modified humanized animal models [79], or the series of studies based on, bioengineered and humanized, mouse ectopically implanted microenvironments [78,80–89]. This raises the issue of whether animals should be included in the biological complexity, or if less complete models derived exclusively from human factors should only be used.

#### **Box 3.** BM in vitro existing models.

"Nowadays, the development of a functionally effective model with translational potential is still challenging ( ... ). Different authors have chosen different approaches, and several models succeeded in recapitulating some behaviors of BM specific contexts ( ... ) what these different strategies do have in common is that they show that progress in the field relies on interdisciplinary approaches ( ... )".

Narrowing the context down to hematological malignancies, due to the intricacy of the role of BMN in tumor initiation and progression some authors use healthy BMN models with deliberately injected malignant cells to study pathological conditions, while others attempt to mimic pathological niches by including altered ECM or soluble tumorigenic environment [90,91]. For example, major efforts have been made to optimize complex BMN models for growing leukemia cells [92]. Highly porous sca ffolds made from di fferent biodegradable and non-biodegradable polymeric materials, such as poly (l-lactic-co-glycolic acid), polyurethane (PU), poly (methyl- methacrylate), poly (<sup>d</sup>, l-lactide), poly (caprolactone), and polystyrene coated with COL I or FN have been tested as models to study AML biology and treatment [93]. Inclusion of cell-cell interactions has also been considered: co-culture of BMN-forming cells with AML cells in a decellularized Wharton's Jelly matrix (DWJM) showed higher resistance to chemotherapy than conventional suspension cultures and best resembled in vivo drug-resistance [94]. Di fferent ECM components of ECM are known to play a role in inducing drug-resistance in di fferent tumors: COL gel cultures induce drug resistance in di fferent tumors; lymphoma or myeloma cells adhering to FN acquired resistance to mitoxantrone or dexamethasone, respectively, and HA is associated with drug resistance in leukemia [95,96]. Di fferent authors have selected these biomolecules as sca ffolding for in vitro reproduction of drug resistance [92]. For example, COL, FN and HA are present in DWJM sca ffolding in the above-mentioned study on leukemic cells, generating a simplified but rationally designed platform for the study of ECM-induced drug-resistance [92,94,97]. Multiple myeloma (MM) is a hematological neoplasia in which the BMN's role in disease progression via elevated proliferation, migration and CAM-DR has been extensively reported [98]. Simplified TERM approaches contributed to this knowledge: co-cultivation of MM cells with BM-MSCs revealed that these cells co-modulated their phenotype and that BM-MSC secretomes and microvesicles (MVs) participate in this crosstalk [99–101]. MM cell lines cultured on decellularized ECM from normal donors (ND) or MM patients' BM-MSC recently showed that MM-MSCs' ECM promotes MM cells' MAPKs/translation initiation-dependent proliferation and migration, while normal donors' ND-MSCs' ECM has the opposite e ffect [91], an interesting result that supports the use of transformed in vitro conditions for modeling hematological malignancies. The more sophisticated platforms developed as healthy BMN have been used to study the behavior of malignant MM cell lines and e ffectively succeeded in reflecting resistance to conventional drugs, the main disadvantage in current MM clinical managemen<sup>t</sup> [76,102].



HSPCs, hematopoietic stem and progenitor cells; OB, osteoblast; PEGDA, poly (ethylene glycol) diacrylate; PU, polyurethane; HSC, hematopoietic stem cell; FN, fibronectin; TPO, thrombopoietin; SCF, stem cell factor; PVA, polyvinyl alcohol.



**Figure 4.** TERM models of healthy BMN. (**a**) Co-culture of human HSPCs with MSCs in 3D COL hydrogel dissects two sub-populations of HSPCs. (i) Scheme of the 2-compartment system. (ii) HSC morphology by SEM at day 14 (adapted by kind permission of [69]). (**b**) Co-culture of HSPCs with mesenchymal stromal cells in PEGDA hydrogel coated with adhesive peptides mimicking trabecular bone. (i) Scheme of the approach. (ii) Cross-sections of hydrogels, SEM images (top panel) and fluorescence micrographs (lower panel) with Alexa Fluor 647-labeled BSA in red as tracer molecule to reveal interconnectivity of the pores. Scale bars, 20 μm (SEM) or 100 μm (FL). (iii) Pseudo colored SEM micrograph of a hydrogel seeded with primary MSC-BM and HSPCs. Purple: adherent MSC-BM, 28 μm in size; red: HSPCs, 12 μm in size. Scale bar, 20 μm (adapted by kind permission of [70]). (**c**) In vitro 3DTEBM used for culture of BM aspirates from MM patients. (i) Scheme of the approach; cultures were developed through cross-linking of fibrinogen with calcium; numerous cellular components, including MM cells, MM-derived stromal cells and endothelial cells were incorporated. (ii) Illustration of the BMN and the 3DTEBM with oxygen and drug concentration gradients (adapted by kind permission of [76]). (**d**) Cryogel-based COL coated carboxymethylcellulose scaffold seeded with murine BMN-forming cell line OP9 to generate a living, injectable stroma supportive for hematopoiesis, and with murine HPSCs, seeded scaffolds used as microcarriers for in vivo culture in mice (i).

(ii) Fluorescence image of CCM seeded sca ffold, blue: Hoechst (sca ffold); green: GFP (OP9 stromal cells); red: HSPCs (adapted by kind permission of [65]). Abbreviations: SEM, scanning electron microscopy; HSPC, hematopoietic stem and progenitor cell; PEGDA, Poly (ethylene glycol) diacrylate; HSC, hematopoietic stem cell; MSC, mesenchymal stem cell; BM, bone marrow; 3DTEBM, 3D tissue engineered bone marrow; MM, multiple myeloma; COL, collagen.

#### **6. Unresolved Questions in Modeling BM and Non-Solid Tumors**

As the review delves into the advances in modeling healthy or malignant BM, some basic questions regarding the main design principles of TERM approaches should be considered. The physiological BMN and the transformations leading to or sustaining blood tumors are extremely complex and it will be a major challenge for TERM to faithfully reproduce these artificially in a BM model. In our view, cells must raise the artificial BMN from a provided substrate, which should be as biomimetic as possible, and from the presence of the appropriate environmental conditions (oxygen, fluid flow, physico-chemical stimuli ... ) (Box 4). The promising approach proposed by Torisawa et al., consisting of an artificial substrate engineered in vivo in a biomimetic BM model for in vitro HSC culture, strengthens the idea that the cells themselves are the last-resort architects of model complexity. However, this model is implanted in mice and does not closely mimic the human situation. For more humanized models, TERM must propose alternative methods of cell-guided niche formation. Some of the critical issues of these approaches to modeling blood cancers are examined in this section (Figure 5).

**Figure 5.** Scheme of critical issues that should be addressed by TERM blood cancer model approaches.

**Box 4.** Unresolved questions in modeling BM and non-solid tumors.

"The physiological BMN and the transformations leading to or sustaining blood tumors are extremely complex and it will be a major challenge for TERM to faithfully reproduce these artificially in a BM model. In our view, cells must raise the artificial BMN from a provided substrate, which should be as biomimetic as possible, and from the presence of the appropriate environmental conditions (oxygen, fluid flow, physico-chemical stimuli ... )."

## *6.1. ECM Dynamics and Remodeling*

ECM is a highly dynamic structure continuously undergoing controlled remodeling [103]. In vivo, cells rebuild and remodel ECM constantly through synthesis, degradation, reassembly, cross-link or chemical modification of their di fferent components [104] and resultant biophysical, mechanical and

chemical ECM characteristics influence tissue homoeostasis [105]. ECM remodeling is therefore an important mechanism helping to regulate cell di fferentiation or the establishment and maintenance of stem cell niches, angiogenesis, bone remodeling or wound repair [104]. Deregulated ECM remodeling is associated with di fferent pathological conditions: abnormal ECM deposition and increased sti ffness have been broadly linked with solid tumors [106,107]. In tissue engineering approaches, when adherent cells are seeded on 3D substrates (or on 2D surfaces) preexisting adhesion ligands from substrate functionalizations or serum protein deposition are essential for initial cell attachment, after which the cells remodel and secrete their own ECM. This inherent ability of ECM remodeling to control cell behavior has also been reported and exploited in in vitro 3D models [108,109]. In these approaches, both sti ffness and degradability of the polymeric substrates are critical design variables, as metalloproteinases or other enzymes are commonly used by cells for ECM degradation, and natural or engineered proteolytically degradable materials have been extensively used in biomimetic approaches. Other groups have also proposed material-driven ECM remodeling. Salmerón-Sanchez et al. found that simple FN adsorption onto poly(ethyl acrylate) surfaces triggered FN organization in a fibrillar network similar to cell-assembled matrices [110]. These studies provide evidence of the cells' ability to direct ECM remodeling in vitro and to enhance this process towards biomimetic ECM configurations. However, it has also been reported that cells embedded in artificial matrices require externally-imposed parameters, including matrix sti ffness, which a ffect cell-mediated ECM remodeling by altering ECM regulatory genes [111]. This means that the initial matrix provided by the model will always guide cell behavior and influence cell-mediated ECM remodeling, perhaps reshaping the native output of this process. Therefore, the key question in hematological malignancies modeling is whether patient-derived malignant cells are able to can carry out this process or induce other cells in the model to do so.

## *6.2. Vascularization of the Model*

To answer the question of how far cells can reconstruct native BMN complexity from an artificial substrate we should consider not only cell-mediated ECM remodeling but also the cells' ability to create complex structures, such as vasculature, or reproduce the anatomical traits essential for tissue function, like compartmentalization. Vascularization of the model seems essential, as native BM blood vessels actively contribute to functional di fferentiation into subniches by their di fferent BMN-forming cell populations and generating di fferent biochemical gradients. In vitro vascularization is by itself a wide research field. Vascular networks have been generated in vitro by di fferent approaches to study angiogenesis, vasculogenesis and cancer metastasis [112]. Vasculogenesis is the process by which early capillary-like networks form in vivo during adulthood through recruitment of endothelial progenitor cells (EPCs) from BM [113]. Following the formation of a primary network, expansion and sprouting occur from existing vessels by the process termed as angiogenesis, which is key in solid tumor growth, and activated by hypoxic environments or shear flow [112]. This mechanical stimulus also modulates the process of arteriogenesis, a maturation step that contributes to arterio-venous di fferentiation [113]. Many in vitro models are now enabled by microfluidic-based techniques and flexible polymers such as PDMS that can produce well-defined micro-scale geometries. One of the important aspects dealt with in this review is that in vivo vascular organization depends on the surrounding microenvironment, as seen in vessels aligned parallel to muscle, radially in the retina and highly branched in the lungs [112]. The in vitro patterning of the substrate or the composition of ECM mimicry used to develop the micro vessels influence their organization [114], even systemic factors, such as media flow, may a ffect network morphology and function [115]. Two options arise as dominant strategies for vascularization of engineered tissues in vitro: sca ffold-based strategies using naturally derived and synthetically generated tube-like structures for guided vessel formation vs. the naturally formed cell-based strategies that rely on endothelial angiogenesis and vasculogenesis to form perfuseable networks [112]. ECs, EPCs, human dermal microvascular endothelial cells (HDMEC) or human umbilical vein endothelial cells (HUVECs), among others, have been used in in vitro angiogenesis and vasculogenesis models, always

with a supporting ECM component and proper soluble environment to enhance their attachment and function [112,113,116]. However, as MSCs and other stromal cells have been shown to interact with endothelial lineage cells during neovascularization, MSCs have been co-implanted with human ECs, improving vascular tissue formation [117,118]. The most successful models of healthy or malignant BM reviewed in this article in terms of vascularization are those in which complexity has been engineered by implantation in vivo in mice. Although the human vasculature structure and perivascular BMN can be resembled in humanized structures implanted in mice [80], these approaches are again limited by their uncertainty as to the specific role of mouse vasculature and the cytokines supplied by the mouse system to these humanized implanted structures. Focusing on non-solid tumors, di fferent authors have reported altered angiogenesis in association with di fferent hematological malignancies. Laroche et al. reported that the number of arterioles and capillaries increased in myeloma from its initial stages according to the gravity of the disease [119]. Increased BM angiogenesis and increased levels of associated factors, such as VEGF, have also been described in relation to AML, ALL, B-CLL or CML, among others, leading to the development of therapeutic applications based on disruption of this mechanism [120].

The remaining question concerning vascularization is thus whether there is an e ffective animal-independent cell-guided way to biomimetically vascularize human engineered models, or in its absence, whether it is preferable to sacrifice the complex vascular biomimetic networks and replace them with gradients and an unstructured or artificially conditioned cellular and soluble factors from vessels [68] in order to maintain fully human models. This is the case of in vitro models of hematological malignancies such as the MM model in De la Puente et al., which fully preserve humanization and recapitulate the polarized BM niche structure with the generation of biochemical gradients, showing a more hypoxic vs. a more normoxic niche with more endothelial cells (Figure 4c) [76,121]. The present authors believe that recent advances in the field of in vitro vascularization, particularly the intention to provide better support to organ-on-a-chip approaches, will sooner or later contribute to the compatibility of human engineered vessels with native-like characteristics.
