**Box 5.** Compartmentalization.

"Either by direct inclusion in the model by means of different design parameters or by indirect cell-mediated remodeling, compartmentalization seems necessary for functional di fferentiation of BMN".

#### *6.4. Stemness Maintenance Vs. Di*ff*erentiation Balance*

Another distinctive aspect of BMN is the balance between di fferentiation and maintenance of pluripotency, which is the key regulator of healthy and malignant BM niches (Section 5). Tumor cells

interact with many cell types in their environment, especially in hematological malignancies, and in a trustworthy model these cell types have to coexist, as in in vivo conditions. How can pluripotent or multipotent cells be maintained at the same time as their di fferentiated counterparts and the production of trophic and regulatory factors? (Box 6). Ex vivo expansion of HSCs has been widely studied, mainly for its implications in hematopoietic cell transplantation (HCT) and generating red blood cells ex vivo. Uncommitted primitive cells are needed to e fficiently repopulate the BM in HCT, which is still challenging since HSCs outside their niche tend to di fferentiate or become senescent [123]. The in vivo regulation of HSC self-renewal and di fferentiation is the reference pursued. Biomimetic co-cultures provide biological cues that liquid cultures, in the absence of stromal cells, barely replicate. Some studies maintain that the main factors that support HSC expansion rely on direct cell-cell interactions with di fferent BMN-forming cells, while for others direct contact is not required, so that soluble factors play a key role in HSC fate. However, in agreemen<sup>t</sup> with [123], we believe that the interplay between soluble factors and direct ECM and BMN-forming cell interactions allows niche cues to e fficiently regulate HSC fate. The choice of culture medium is crucial to determine HSC fate in vitro. Most approaches use supraphysiological concentrations of di fferent HSC-supportive cytokines such as SCF, TPO, ANG-1 or IL-6 directly added to the medium and/or secreted by the BMN-forming cells in the model. ECM components, such as FN, COL, laminin or proteoglycans help to regulate HSC fate by binding the growth factors produced by BMN-forming cells, favoring cell co-localization and biological cues. Exploiting ECM's ability to retain bioactive factors, Mahadik et al. bound SCF to a gelatin-based hydrogel to increase its bioactivity [124]. Di ffusion control of nutrients, oxygen and cytokines by the ECM can also lead to gradients that could provide regulatory cues to HSCs [72]. Nevertheless, the presence of stromal cells seems to be key to promoting HSC self-renewal in vitro. Gottschling et al. showed that the presence of MSC was enough to ensure self-renewal, while the activation of β1-integrins by FN was not [125]. Di fferent types of bioreactors have been used for HSC expansion ex vivo, including stirred tank suspension, perfusion chambers, fixed beds, airlift or hollow fiber reactors [123]. However, most of these rely on suspension cultures unable to maximize cell-cell and cell-ECM contact. The design of bioreactors coupled with biomaterial approaches has overcome this issue, such as that proposed in Sieber et al., whose 3-D co-culture model based on a hydroxyapatite-coated zirconium oxide sca ffold with human BM MSC inserted in a microfluidic device was able to support long-term HSCs [74].

The issue of the di fferentiation vs. stemness balance is of interest not only for HSCs, blood and immune cells, but also for MSCs and osteolineage cells of varying maturity. Several approaches incorporate MSCs and pre-di fferentiated OBs in the model to cover this heterogeneity, although a dynamic balance between MSC expansion and OB di fferentiation would be of grea<sup>t</sup> interest. Maintaining MSC stemness ex vivo is also still a challenge, some microcarrier bioreactors have been used as large-scale production systems [126], although TERM approaches do not ye<sup>t</sup> seem to have incorporated the dynamic balance between MSCs and OBs di fferentiation in BMN modeling. The initial steps have been taken, as there are many osteogenic sca ffolds that e fficiently generate well-di fferentiated OBs from initially seeded MSCs. However, the long-term co-existence of OBs, osteocytes and primitive MSCs seems to be di fficult to regulate, as mature osteolineage cells seem to promote osteogenic MSC di fferentiation and lead to progenitor exhaustion [127]. Osteogenic media are commonly used in in vitro approaches, however, when considering co-cultures with HSCs or even with blood cancer cells such as MM cells, the use of osteogenic media should be rethought, since some of the soluble factors necessary for osteogenic di fferentiation of MSC could alter other cells included in the model. For example, dexamethasone is a commonly used anti-MM treatment (although the concentrations used in osteogenic media are 10 times lower [128]). The main problem concerning the use of specific conditioned media to promote specific phenotypic cell commitment in an ideal BM model is how to control the di fferent e ffects of this media in di fferent zones, thus allowing the coexistence of undi fferentiated and mature cells. As this localized activation of cell di fferentiation does not seem to be possible by means of inductive media, we believe that TERM approaches have

been shown to be efficient enough to induce differentiation in specific areas of the model and promote pluripotency in others. Anderson et al. reviewed specific aspects of scaffolding material design to rationally target MSC cell fate [129]. Mechanotransduction is the process by which MSCs turn an adherent stimulus into a cellular response able to determine cell fate. For cells to adhere to a synthetic surface, the material has to replicate an ECM motif or absorb ECM proteins, so that by controlling the ability of a material to allow cell adhesion, the subsequent activation of different signaling pathways that control stem cell fate can also be defined [129]. There are important design parameters for this purpose: material chemistry, stiffness and nanotopography are known as the material/surface interface "triangle", as their variations control the scaffold's interactions with MSCs by conditioning the formation of focal adhesions. Binding the cell through focal adhesions to adhesive ligands on the materials creates tension and activates signaling that controls cell behavior and stem cell fate. Chemical functionality can be used to produce high or low adhesive areas for the cells to respond to stiffness, which affects their ability to create tension through focal adhesions. Topography can present the adhesion ligands to the cells in either a favorable or unfavorable way, again affecting adhesion and subsequent tension and signaling [129]. The chemistry of the surface and its conditioning of matrix protein deposition, which in turn regulates the presentation of cell adhesion motifs to cells, can be finely tuned with techniques such as dip pen nanolithography (DPN) to apply a surface chemistry to a precise substrate on the nanometer scale. For example, Curran et al. set out to optimize an arrangemen<sup>t</sup> of dots of "chemistry" to manipulate MSC behavior by creating specific patterns of -CH3, -NH2, -CO, and -CO2. They found that the functionalized -CH3 surface maintained stem cell markers while -NH2 dots increased adhesion and osteogenesis [130]. Precise control of chemical functionality in 3D is still challenging, however ongoing progress in nanofabrication techniques promises to make significant contributions in the near future [129]. ECM topography in vivo presents a native composition that provides the cells with behavioral cues. in vitro, the topographical cues influence on stem cell behavior has also been proved and explored by many scientists as a tool to manipulate MSC fate. Dalby et al. developed MSC growth substrates with random or highly ordered patterns and observed that osteogenesis increased in the disordered patterns as efficiently as in inductive media [131]. However, these studies imply a 2D character and a "static" nature, in the sense that they include a configuration of the material oriented to promoting a single effect on the cell fate (self-renewal or differentiation). While the stem cell niche and the BMN are dynamic microenvironments in which the balance between stemness and differentiation is regulated by the demand. Next-generation materials able to support self-renewal and differentiation with spatiotemporal control have attracted significant interest in recent years. For example, stimuli-respondent materials in which a cytocompatible stimuli such as light triggers material changes leading to alterations in cellular behavior, together with advances and tridimensional implementation of microfabrication techniques, will address the future need for niche-mimicking materials [129].

#### **Box 6.** Stemness maintenance vs. differentiation balance.

"In hematological malignancies, tumor cells interact with many cell types ( ... ). How to maintain in culture pluripotent cells at the same time as their differentiated counterparts? TERM approaches have shown to be efficient enough to, in a localized and differential manner, induce differentiation in specific areas of the model and promote pluripotency in others".

#### *6.5. Cell Culture Media Renewal and Composition*

Although culture media renewal is usually considered as routine in in vitro cultures, apart from its role as a supplier of nutrients and externally selected regulatory factors, a common problem in both static and bioreactor cultures is that this renewal may remove secreted factors essential for the relationship of the different cells in the culture system. We hypothesize that the effect of reducing active component concentration by renewing media can interfere with physiological behaviour. The dynamics and time lapses required for these processes have not ye<sup>t</sup> been clearly established. However, the effect of soluble factor dilution has been reported in particular applications; for example, spontaneous in vitro HSC differentiation was avoided by diluting secreted differentiation signals via proportional volume to cell number ratio [132]. Some studies on the use of conditioned media for several applications and culture duration for conditioned media generation from different stem cells found that culture conditions could vary from 16 h to 5 d [133]. The timelines of some mammalian cellular processes could help in understanding the timelines in which cells grow and communicate. For example in a human HeLa cell, diffusion of a protein across a cell will take 10 s in a 10 μm cell, while transcription of a 10 kbp gene takes approximately 10 min, and 1 min to translate a 300 aa protein [134]. This means that a cells' ability to reconstitute the removed soluble factors from the culture media is probably within these orders of magnitude. Some cellular responses to different stimuli may also be in the same order and are thus probably altered to some extent by medium renewal, e.g., the on-switch of apoptosis in HeLa cells, which has been reported to take between 9 and 29 min [135].

## *6.6. Towards Personalized Medicine*

Research has shifted from gold standard treatments for a given cancer to finding solutions for patient-specific cancer subtypes or, in other words, personalized medicine [136]. Drug effectiveness differs greatly between individuals, a cure for one patient can be ineffective or harmful to another. When modeling hematological malignancies, the challenge lies in reproducing cell-cell and cell-ECM interactions in a 3D environment to regulate the signaling pathways leading to drug-resistance. However, inter-patient heterogeneity has driven the need for personalized medicine as individual tumors have different gene expression profiles, tumor microenvironments and behavior even within the same cancer subtype [136]. The challenge of recreating niche interactions leading to drug resistance has advanced to reproducing inter-patient differences in the model (Box 7). Personalized cancer therapy has historically focused on profiling tumor DNA, RNA, or protein as molecular biomarkers to predict patient response. However, these methods have not been able to predict therapeutic response [136], while functional assays based on integrating tumor cells into chemosensitivity and resistance assays have become a complementary method [137]. In 2017, Snijder et al. evaluated the effect of ex vivo drug sensitivity screening on the treatment of patients with refractory hematological malignancies using the Pharmacoscopy automated immunofluorescence microscopy-based platform on 48 patients, 17 of whom received treatment guided by this approach. Comparison of the benefit of Pharmacoscopy-guided treatment with the effect of previous treatments in the same patients showed a marked improvement in progression-free survival with the former, providing evidence of the promise of drug-response profiling in haemato-oncology [138]. Similar studies have served as the proof-of-concept of how phenotypic screening approaches to different blood cancers such as AML [139] or MM [140,141] could improve the selection of the right drug for the right patient at the right time. Although these approaches are still simplistic in terms of BMN mimicry, some of them include co-culture with BM stromal cells and ECM components like COL [141]. However, the most important factor is that they have been the pioneers in introducing multidisciplinary 3D biomimetic models that reproduce tissue architecture to revolutionize the clinical managemen<sup>t</sup> of cancer patients. Certain types of blood cancer such as MM would greatly benefit from these advances, as it is a treatable but incurable malignancy in which all the patients eventually relapse and the choice of their treatment now relies on clinical acumen instead of empirical personalized data [141].

Personalized drug resistance assays have thus shown promise, although they still remain extremely simplistic compared with BMN harboring cancer in vivo [136,142]. TERM engineered biomimetic models are rapidly progressing. Microscale models can provide unique functionality and controllability (e.g., enhanced spatial and temporal controls) and are emerging as practical tools to investigate tumor-stroma interactions [143]. Improvement of 3D substrate production techniques and their use for stem cell fate regulation [129,144], even on patient-derived ECM as scaffolding materials [14] and advancement in microscopy, flow cytometry and different evaluation techniques led the way to integrating patient samples into these in vitro models to assess therapeutic response

in biomimetic devices and overcome the limitations of immortalized cell lines. The latter are highly selected populations that do not reflect the heterogeneous tumor genetic and functional variability. Patient-derived primary cells would overcome this disadvantage, although their use is hindered by issues of patient sample acquisition, variability and the di fficulties associated with their culture. Integrating patient cells requires that part of the sample, which would ordinarily go to pathology, which is not always possible, is specially challenging in the case of hematological cancers in which extra mL of BM aspirate should be extracted. Secondly, it is di fficult to transfer the sample from the operating theatre to the in vitro platform while maintaining cellular viability. The most questionable aspect of this process is that the sample usually needs processing, which introduces additional variations in tissue architecture, microenvironment and cell selection. In addition, cells often undergo changes when removed from their natural environment (primary cells easily become senescent, and have a limited availability and lifespan). These are well known unsolved problems that hamper the development, optimization and validation of new assay platforms [136]. We also consider that other issues not widely included in the literature produce uncertainty and complicate the incorporation of primary cells into hematological malignancy models, as there are questions regarding the timing of cell expansion before seeding into the models, types of cells and number or proportions of cells. For example, how long does it take for a tissue sample from a patient to have the required number of cells of all the relevant types in a culture to be able to carry out drug resistance tests and ge<sup>t</sup> statistically significant results? Does a BMN aspirate from a patient e fficiently represent the proportions of key niche cells such as MSCs or HSCs?

Despite the progress made, it is undeniable that choosing the right cancer treatment is di fficult because of limited tools, money and time [136]. Important steps have been taken, as some authors have convincingly resembled acquired drug resistance in vitro with cell lines [76,102], but some issues need to be addressed until optimized models allow routine patient-specific drug testing studies. Personalization means that the model must consider inter-patient di fferences in drug response. In the authors' opinion interactions between tumor cells and the cellular and non-cellular components of the biological niche play a key role in determining these di fferences. This means that personalization is a key step towards obtaining the perfect blood cancer model. They also believe that the cells themselves, acting as the last-resort architects of the particular extracellular environment provided by the 3D biomimetic model, improve the inter-individual specificity of the biological niche and this cannot be artificially reproduced.

#### **Box 7.** Towards personalized medicine.

"The challenge lies in reproducing in a 3D environment the cell-cell and cell-ECM interactions which regulate the signaling pathways leading to drug-resistance ( ... ). Personalization means that the model must consider inter-patient di fferences in drug response, thus the challenge lies in reproducing di fferences from patient to patient ( ... ), in integrating patient samples in models to assess therapeutic response in biomimetic devices, and overcome the limitations of immortalized cell lines".
