Mesenchymal Stem Cells Relevance in Multicellular Bioengineered 3D In Vitro Tumor Models

In vitro 3D tumor microenvironment mimicking models are gathering momentum as alternatives to traditional 2D flat monolayer cultures due to their potential for recapitulating major cancer hallmarks. To fulfill such potential, it is crucial that 3D tumor testing platforms completely emulate in vitro the complex in vivo tumor niche and its cellular constituents. Mesenchymal stem cells (MSCs) are recognized to play a pivotal multi‐modulatory role in cancer, generating interest as biological targets and as key tumor suppressing, or tumor promoting effectors. This review discusses the biological influence of different types of MSCs in the tumor microenvironment and showcases recent studies that engineer 3D MSCs‐cancer cells co‐cultures as advanced in vitro therapy testing platforms. A special focus is given to MSCs‐cancer 3D co‐culture set‐up parameters, challenges, and future opportunities. Understanding cancer‐MSCs crosstalk and their underlying effects is envisioned to support the development of advanced 3D in vitro disease models for discovery of forefront cancer treatments.


Introduction
In vitro pre-clinical cancer models are mainly based on the use of 2D cancer cell monolayers and laboratory animal models. [1,2] Historically, both methodologies have been recommended by regulatory agencies to aid in the discovery and validation of anti-cancer therapies that are to be administered in human clinical trials. [1,3] However, while these platforms have contributed immensely to explore cancer development and biomarkers discovery, they are still inadequate approximations of human tumors. [4,5] Particularly, in 2D in vitro flat and monotypic cell culture models, cancer cells experience highly artificial environments often comprised of plastic or glass celladhesive surfaces, forcing cells to grow in an environment which lacks major extracellular matrix (ECM) and stromal cell components. [6] As a consequence, cultured cancer cells are exposed to an unnatural 2D spatial organization and non-physiological conditions, presenting a higher area of exchange with cell culture media in comparison to what naturally occurs in 3D tissues in vivo. [6] Such results in an abnormal morphology, loss of structural organization and cell polarization, with lower cell adhesion occurring, for example, due to loss of integrin-ECM interactions. [7] Therefore, cells phenotype and their response to different treatments is not representative of complex human tumors, thus creating a gap between in vitro/in vivo data correlation.
On the other hand, despite animal models constitute a more laborious and economically demanding alternative, they are more representative of the in vivo scenario than conventional 2D in vitro cell cultures. [8] A significant number of early stage preclinical in vivo drug screening studies are performed in small animal murine models. However, these often lack a correct representation of the tumor stroma and present expression variances in the structural homology of molecular targets, which can result in highly variable therapeutic responses. [8][9][10][11] Such variability is somewhat detrimental for the validation of candidate anti-cancer therapeutics and particularly limiting in the case of combinatorial cancer therapies high-throughput screening. Such treatment modality demands a higher level of reproducibility and predictability of safety/efficacy parameters as the cocktails of bioactive molecules could trigger a wide range of biological responses (e.g., additive toxicity, antagonism). The use of in vivo models in the context of combinatorial anti-cancer therapies entails significant ethical concerns due to the large number of test groups, not easily allowing high-throughput screening of different combinations. In addition, the correlation between in vivo animal studies and human clinical trials is very limited, thus evidencing the necessity of developing more realistic testing platforms that can provide robust preclinical data. [8,9,12] Hence, the development of in vivo-mimicking, and reproducible preclinical in vitro validation models to efficiently predict the biological performance of anti-cancer therapies may contribute to increase the translation speed of novel therapies from the bench-to-bedside. The bioengineering of such models must be based on similarities to human biology and disease specific features, so as to ensure higher predictability of preclinical research and exclude non-adequate anti-cancer therapeutic candidates prior to human clinical trials. [3,13] The demand for understanding cancer development and develop more advanced treatments has resulted on a growing number of studies exploring the potential of advanced threedimensional (3D) tumor models as more viable testing platforms. 3D multicellular tumor models are generally selfassembled compact cellular agglomerates that may be cultured in vitro during relatively short or prolonged periods of time using different techniques that include hanging drop, bioreactors, or organ-on-a-chip platforms. [14,15] Such platforms try to incorporate the spatial complexity, cellular heterogeneity, nutrient/pH characteristic of the in vivo tumor microenvironment. [14] In this context, various types of malignant cells were shown to possess gene expression patterns and phenotypes similar to those encountered in vivo when cultured in 3D. [16] Such evidence is an added-value to these models, since on one hand they provide the desired spatial distribution and on the other they ensure several biological functions found in human cancers. [17] This review summarizes the recent advances made in the field of 3D in vitro multicellular tumor models that more closely recapitulate the tumor microenvironment and its diverse components by including co-cultured cancer-stromal cells, including mesenchymal stem cells (MSCs). MSCs are now well-recognized to take multi-modal roles in oncological disease progression and metastasis, for which depending on MSCs tissue of origin and the type of tumor, they either play a beneficial or detrimental role. [18] This interesting duality is discussed considering the most recent literature reports and a critical perspective is given toward the development of MSCs-cancer 3D in vitro co-culture tumor models.

The Dual Role of Mesenchymal Stem Cells in the Heterogeneous 3D Tumor Microenvironment
Mesenchymal stem cells (MSC) otherwise also known as multipotent mesenchymal stromal cells, were first described as fibroblast-like cells, about 50 years ago. [19] They are a subset of multipotent precursor stromal cells, with fibroblast-like morphology. MSCs represent a cell sub-population that can be found residing in the mesenchyme of a wide variety of tissues, including: i) umbilical cord Wharton jelly; ii) placenta; iii) peripheral and fetal blood; iv) adipose tissues (generally defined as adipose-derived stem cells (AT-MSCs)); v) skeletal muscle; vi) heart; vii) liver; viii) lung tissues; and iv) bone marrow. [20] Constituting a heterogeneous cellular population, MSCs have diverse morphologies, and are commonly identified through specific cell surface markers. Under the guidelines issued by the International Society of Cellular Therapy, MSCs in in vitro cultures must be able to adhere to plastic-treated surfaces in standard tissue culture conditions; with more than 95% of the population expressing CD105, CD90, CD73, and with less than 2% presenting positive CD45, CD14, CD34 or CD11b, CD79a or CD19, and major histocompatibility complex II (MHC-II) markers. Besides this, in vitro cultured MSCs must also show multilineage differentiation capacity. MSCs traditionally differentiate into chondrocytes, adipocytes, and osteoblastic lineages under controlled in vitro differentiating conditions. [21] A small sub-population of bone-marrow derived MSCs may still be negative for CD44, CD45, MHC I, MHC II, and c-kit, displaying the capacity to differentiate as well into nerve, pancreas and lung cells under certain culture conditions. [22] MSCs are reported to exhibit immunosuppressive and immunomodulatory properties, [23] as well as being able to migrate and induce modifications in damaged or inflamed tissues. [24] These abilities are normally explored under the context of cell-based therapies for tissue repair and regeneration, either via direct cell-cell contact or by means of paracrine signaling. [24] MSCs are also described to have high-tumor tropism, a process that is generally mediated by chemokines and growth factors. [25] Upon migrating into the tumor microenvironment (TME) and establishing contact with this complex, MSCs can be hijacked by cancer cells, either through cell-cell or paracrine interactions, leading to either beneficial or detrimental roles in the evolution of cancer. [26] For instance, MSCs recruited to the TME of breast cancers have shown to possess increased secretion of neovascularization and epithelial-mesenchymal transition (EMT) promoting factors, such as Notch1 and TGF-b1, thus promoting tumor growth and metastasis. [27][28][29] Positive interactions between cancer cells and MSCs are not restricted to a specific phenotype of MSCs. For example, umbilical cord derived MSCs have shown to greatly increase cholangiocarcinoma proliferation, migration and resistance ( Figure 1). [30] MSCs can also indirectly aid the evolution of cancer by regulating the activity of other TME cells. [25] For example, breast cancer-associated bone-marrow MSCs (BM-MSCs) are João F. Mano is a full professor at the Chemistry Department of University of Aveiro, and member of CICECO -Aveiro Institute of Materials. His current research interests include the use of biomaterials and cells toward the progress of transdisciplinary concepts to be employed in regenerative and personalized medicine. Prof. Mano has been coordinating or involved in many national and European research projects, including an Advanced Grant from the European Research Council.
known to increase the frequency of regulatory T cells, and decreasing the lysis activity of natural killer cells and cytotoxic T lymphocytes by the secretion of TGF-b1, resulting in a poorer prognosis due to unchecked tumor progression. [30,31] Contrariwise, MSCs have also been reported to exert antitumor effects, [32] with several studies having demonstrated the role of MSCs in increasing the growth and metastasis of tumors. [18,33] In an elegant a study performed by Quiao and coworkers, 2008, MSCs derived from fetal dermal tissue were shown to inhibit the proliferation of two human hepatocarcinoma cell lines (H7402, HepG2), when co-cultured. [34] Such resulted in an increased apoptosis for H7402 cells via the down regulation of the Wnt/b-catenin pathway as proposed by these researchers. [34] Interestingly, it is important to emphasize that this effect is closely correlated with MSCs tissue of origin ( Figure 2). As reported by Attar-Schneider et al., 2015, BM-MSCs can reduce both cellular proliferation, viability and migration of non-small lung cancer cells as a result of the down regulation of translation initiation factors and mitogen-activated protein kinases (MAPK) signaling. [35] However, the same is not true for Wharton jelly-derived MSCs which show antagonistic effects to those portrayed by BM-MSCs in lung cancer stem cells (L-CSCs), acting in a pro-tumorigenic manner. [32] Conversely, conditioned media derived from human fetal MSCs showed high levels of insulin growth factor binding proteins (IGFBP), which when used in hepatocellular carcinoma cell culture, ends up sequestering free IGF, inhibiting cancer cell proliferation. [36] Due to this anti-proliferative potential, MSCs have been proposed to be used as Trojan-horse-like vehicles for anti-cancer therapy by taking advantage of their tropism toward cancer cells. [37] By exploiting this feature, TREAT-ME1, the first worldwide clinical trial exploring the use of Herpes simplex virus genetically engineered MSCs for treatment of gastrointestinal tumors via RANTES/CCL5-ganciclovir prodrug therapy combination. [25] However, the lack of knowledge on how MSCs behave within the TME means that patients undergoing these treatments must be carefully monitored. Thus, understanding how MSCs from different tissue origins affect, and are affected, by tumors is of paramount importance for future therapeutic applications ( Figure 2).  www.advancedsciencenews.com www.biotechnology-journal.com

MSCs Tropism Toward the Tumor Microenvironment
Mesenchymal stromal cells interactions with cancer cells and the tumor microenvironment result in part from the similarities existent between the TME and those of a regenerating wound. [18] Generally, the TME continuously produces and releases various cytokines and other signaling mediators, that act as chemoattractants for several immune cells such as monocytes, T lymphocytes, and mast cells. This establishes the grounds for a perpetual state of inflammation in the surrounding tissues. [38,39] Such signals derived from the tumor microenvironment, include: VEGF, TGF-b1, EGF, HGF, bFGF, PDGF, IL-8, neurotrophin-3, IL-1b, TNF-a, monocyte chemoattractant protein-1 (MCP-1 or CCL2), and C-X-C motif chemokine ligand 12 (CXCL12) also known as stromal cell-derived factor 1 (SDF-1) which act as homing beacons on various cells including MSCs which exhibit the capacity to accurately migrate via chemotaxis. [33] This tumor tropism toward primary and metastatic tumor sites by MSCs has been extensively observed both in vivo and in vitro in numerous cancers, such as those of breast, brain, colon, pancreas, skin, ovarian, and lung, and is characterized by MSCs infiltration. [26,33] The extent of MSCs migration appears to be directly influenced by the extent of chemoattractant signals released by the TME. Recent reports emphasize that factors such as increased inflammation, hypoxia, and cell dead (related or not to tumor treatment), increase MSCs migration to tumor tissues. [30] Such event is, for example, observed in radiotherapy treated breast cancer, which results in an increased release of release of TGF-b1 and platelet-derived growth factor BB (PDGF-BB) by tumor cells. [40] The underlying mechanism responsible for the tumor-directed migratory capacity of MSCs remains, however, to be fully uncovered. This is a crucial piece of information that must be explored in the future, before MSCsbased treatments can be translated into clinical practice. [26]

MSCs Conversion to Tumor-Associated MSCs À Role in Progression and Metastasis
Upon infiltrating on the TME, MSCs are converted into tumorassociated mesenchymal stem cells (TA-MSCs) ( Figure 3).
Depending on their original tissue, and on the type of cancer, TA-MSCs are known to acquire gene-expression profiles that exhibit increased secretion of tumor promoting factors. [41] TA-MSCs populations have been reported to increase the population of cancer stem cells (CSCs), and induce malignant cells to enter EMT, thereby promoting enhanced motility, invasiveness, and survival. [41] In turn, this can induce a refractory profile in the tumor, a factor that can lead to the establishment of metastasis. [42] TA-MSCs have been found to engage in multiple interactions with cancer cells [41] ; besides paracrine signaling, communication has also been achieved through the exchange of exosome cell-derived vesicles, [43] mitochondria, [44] and also cell membrane components. [45] Interestingly, these interactions have been reported to occur in a bidirectional mode. It is important to emphasize that the effects of the exchanged molecules vary with the type of cancer and MSCs population. [25] As previously stated, EMT requires the formation of a reactive stroma, capable of secreting specific EMT-inducing factors and rearranging the surrounding tumor-specific ECM. Various factors that promote the appearance of this reactive stroma such as fibroblast growth factor (FGF), hepatocyte growth factor (HGF), epidermal growth factor (EGF) are produced by MSCs upon impregnation into the tumor niche. [33,46] Some studies point out that this conversion from regular MSCs to TA-MSCs can also be performed in vitro through 2D co-culture of TA-MSCs with MSCs, indicating that naive MSC conversion can be augmented once in vicinity of the TME. [41] Besides directly increasing cancer cells EMT, MSCs also increase vasculogenesis or increase the secretion of growth factors, creating a suitable environment to support disease progression. [33,47] Moreover, as aforementioned, MSCs are capable of modulating the metabolism of cancer cells through the excretion of exosomes generally loaded with proteins, DNA, and non-coding RNA (ncRNA). [43] This effect can be particularly seen in the interactions that occur when cancer cells metastasize into the bone-marrow, a process in which MSCs secrete chemoattractant factors recognized by cancer cells. [41] In this fatal alliance between metastatic cancer cells and MSCs, factors such as MCP-1 (CCL2) and SDF-1 (CXCL12) appear to play a key role. [26,33] Once metastasis are established in the marrow, MSCs are hijacked into helping in the creation of a niche that protects the tumor against chemotherapeutic agents. [18,25,48] In summary, several factors affect MSCs-cancer cells interactions and MSCs TME interactions, including: i) the type and source of MSCs including their different cytokines expression; ii) the interactions with other elements of the TME; and iii) the in vitro and in vivo conditions in which MSCs are cultured. [49,50] The nature of these interactions and how they take place, either by direct or indirect interference with cancer Figure 3. MSC migrate toward the tumor, with different mechanism playing major roles in diverse cancers (glioblastoma -CCL2 and CCL25) (hepatocarcinoma -CCL15 and CCL20). Migrated MSCs suffer a conversion process, releasing a myriad of factors that aid the tumor by rearranging the surrounding ECM, promoting tumor angiogenesis, proliferation, immune suppression, and ultimately EMT and metastasis. [18,25,26,41,120] www.advancedsciencenews.com www.biotechnology-journal.com cells and surrounding stroma, requires further exploration in the future. With such knowledge holding great promise in the context of new therapies discovery, both at the development level by facilitating the discovery of key therapeutic targets, and at the drug-screening stage by allowing the development of more realistic in vitro 3D tumors.
3. In Vitro 3D Tumor Models to Test Candidate Anti-Cancer Therapies 3D cell cultures of cancer cells take aim to recapitulate in vitro the complexity of human cancer allowing for better preclinical analysis of new pharmaceuticals, with the goals of permitting the evaluation of possible side effects, and the collection of meaningful compound behavior data before proceeding to human clinical trials. [51] When successfully engineered, 3D in vitro tumor models provide a more robust correlation with the in vivo biological performance of candidate compounds. [52] In order to fully mimic human tumors, the ideal in vitro 3D cancer culture model should reproduce in vitro all the interactions and selective pressures that occur in the human body. Unfortunately, the complexity of this task is tremendous and as such, a balance must be found between the model's capacity to faithfully recreate the TME, while assuring ease of analysis and lowering production costs to efficiently allow highthroughput screening methodologies. 3D models production methodologies extend over a gamut of techniques which can be grouped into: i) scaffold-free; ii) scaffold-based; and iii) combinatorial methodologies. Regardless of the production methodology, all these platforms have as a unifying element: the establishment of 3D multicellular structures, or microtissues, comprised of one or more cell types, either derived from immortalized cancer cells cultures or from patient's primary cells. [4,[52][53][54] 3D cell culture methods, range from monotypic cell line cellular spheroids (single cell type), for instance cultured in non-adherent substrates or through hanging-drop techniques, [55] up to complex co-culture systems comprising heterotypic cell lines (e.g., tumor stroma cells including fibroblasts) in a complex tumor ECM-like supporting matrix, or even associated with microfluidic systems for dynamic nutrient perfusion. [56] 3D multicellular tumor spheroids (3D MCTS) are seen as gold standard bioengineered microtissues for drug screening. [57] 3D MCTS are generally comprised by an actively proliferating outer cell layer and a necrotic core. [58,59] Moreover, upon aggregation into spherical structures MCTS acquire nutrient, pH, and oxygen gradients ranging from spheroids periphery to their core. [59] These characteristics are in accordance with those of solid tumors with more than 400-600 mm. [60,61] In these models, the lack of vasculature, similar to that of avascular solid tumors, and the establishment of a compact cell aggregate, hinders the mass transport of nutrients, metabolites and oxygen. [59] Additionally, their 3D nature confers them the ability to portray cell-cell biochemical interactions, tumor gene expression patterns, and even growth kinetics similar to those observed in vivo. [59] It has also been reported that 3D MCTS models promote cancer cells genetic/epigenetic modifications toward a more aggressive, tumorigenic, and multi-drug resistant phenotype. [62,63] Apart from these parameters, heterotypic 3D MCTS (resulting from co-culture of cancer-stromal cells), are able to mimic the deposition of ECM-components similar to those found in solid tumors. [57,59,64] An important parameter, since the ECM-based barrier that hinders O 2 and nutrients mass transfer also restricts the penetration of therapeutic agents in the tumor. The existence of ECM creates a suitable in vitro environment for testing and studying new pharmacological therapies and mimics in vivo conditions. [65] This correlation, resulting from the incorporation of diverse cell lines, is also observed in scaffoldbased models that provide an ECM-mimicking environment without requiring previous cell mediated ECM deposition. [64] The advantages and disadvantages of various 3D models production methodologies have been extensively discussed in several reviews. [51,54,56,59,[66][67][68][69][70][71] These different formulation technologies are addressed herein mainly in the context of heterotypic cancer-MSCs 3D in vitro models establishment.

Cancer-MSCs Co-Culture Relevance in 3D In Vitro Tumor Models
Irrespectively of 3D disease models production methodology, heterotypic cell co-cultures comprising cancer-stromal cells provide a better reproduction of the complex tumor microenvironment when compared to monotypic 3D models (comprised by cancer cells alone). [72] Different bi-or tri-co-culture systems combining cancer cells and stromal cells have thus far been employed for the study of cancer-TME specific interactions and drug screenings. [51,54,59] Recent findings regarding 3D co-culture of cancer-adipose cells, cancer-immune cells, cancer-fibroblasts, cancer-endothelial cells, and their physiologically relevant interactions have been reviewed in detail. [54] Stemming from this dual and multiple-cell culture concept, MSCs addition in in vitro 3D cancer disease models is still far from being fully explored, particularly when considering their important role in both cancer proliferation and metastasis. [73] In fact, having been found to play such an important role in cancer progression and therapy resistance, MSCs must be considered as key elements in the tumor milieu. [41] As such, their inclusion in in vitro 3D tumor models is a necessary step to achieve a more predictive model that closely mimics native in vivo conditions. So far, few articles have been published concerning MSCs interactions with cancer cells in in vitro 3D co-culture tumor models (Table 1), and their main findings will be addressed in the following sections.

Scaffold
Polystyrene particles added to collagen I to form a Gel None [27] BM-MSC Human breast cancer À MDA-MB-231 À MCF7 Scaffold Porous scaffold composed nanocrystalline hydroxyapatite (nHA) and Chitosan None [78] WJ-MSC Human hepatic cancer À HCCLM3 Contained spheroids Alginate beads Cisplatin [118] BM-MSC Human colorectal cancer À HCT8 À DLD1 À Colo320DM À HT29 À HCT116 À T84 Spheroids Forced floating 96-well plates coated with 0.7% SeaKem 1 GTG agarose None [87] WJ-MSC Human lung cancer À AC-LCSC-229 and 223 À SCC-LCSC-136 and 36 Spheroids Forced floating unspecified non-treated flasks to reduce cell adherence None [32] BM-MSC Human lung cancer À A549 Spheroids Hanging drop 25 000 cells per well in a tri culture method (ratios of 5:3:2 À A549 À endothelial lung cells À mesenchymal stem cells) Paclitaxel and gemcitabine [89] MM marrowderived MSCs Human multiple myeloma (MM) aspirates Scaffold Hydrogel composed by 0.5% PuraMatrix IMiDs, bortezomib, carfilzomib, doxorubicin, dexamethasone, melphalan [82] BM-MSC Human glioma À U87 À U373 Spheroids embedded into a scaffold Spheroids embedded in collagen I, laminin, or matrigel None [92] www.advancedsciencenews.com www.biotechnology-journal.com increasing EMT and ERK1/2 activity. These researchers used a lower ratio of MSCs to cancer cells (1:500 and 1:1000) when compared to other studies performed by Mcandrews and coworkers (MSCs:Cancer cells, 1:1 ratio [27] ), or Zhu and co-workers (MSCs:Cancer cells, 1:2 and 1:5 ratio [78] ), in which different ratios were employed to establish the co-culture of MSCs and breast cancer cells in hydrogel-based scaffolds. The data collected by Dittmer, suggests that low numbers of MSCs are capable of increasing breast cancer cell migration. [74] On another report, Klopp and co-workers, 2010, obtained a more organized and spherical 3D model upon inclusion of BM-MSCs when using the same cell lines as models. [79] The authors attributed these differences to the incorporation of MSCs prior to the formation of the aggregates and to the lack of serum-containing media. In a different approach, Dittmer also evaluated if the presence of MSCs in MCF-7 and MDA-MB-231 spheroids affected their susceptibility to 4 tyrosine kinase inhibitors. The interesting results that were obtained indicate that in both spheroids (MCF-7:MSCs and MDA-MB-231:MSCs), the effect of the administered therapeutic anticancer drugs was augmented, pinpointing the decreased resistance to alterations in ERK1/2 phosphorylation and PKCa. [75] Several subsequent studies, [27,78,80,81] using scaffold-based technologies for 3D models assembly provide evidence that BM-MSCs exert a positive effect in cancer cell survival and migratory capacity. In fact, cells co-cultured in tissue mimetic scaffolds including those based on: i) collagen hydrogels [27] ; ii) silk-fibroin scaffolds [81] ; or iii) porous chitosan scaffolds containing hydroxyapatite organotypic as bone mimetics [78] laden with undifferentiated or differentiated MSCs, evidence their capability to up-regulate the expression of EMT-related genes in cancer cells, to induce higher migratory capacity (Figure 4), and to enhance proliferation through different signaling factors. For example, through the release of cyclic AMP, [78] and TGF-b phosphorylation. [27] Moreover, MSCs are also able to increase cancer stem cell populations [80] and cancer cells resistance to therapeutics. In a study by Jakubikova and co-workers, the interactions between BM-MSCs and multiple myeloma (MM) were analyzed using a co-culture model derived from cellular populations extracted from patients in different stages of MM. [82] The results obtained by the authors demonstrated that in comparison to MM cells alone, the co-cultured cells exhibited resistance to an extensive panel of pharmacological agents, both novel and conventional, replicating clinical observations ( Figure 5).
More recently, the development of an interesting MSCscancer cells scaffold-based 3D in vitro model for drug screening was also described by Han and co-workers. In this model, MSCs and human non-small cell lung carcinoma cells A549 cells were co-cultured in wells coated with a multi-layer film of chitosan and hyaluronan ( Figure 6). [83] In this layer-by-layer structured breast cancer cells evidenced a higher migratory capacity, with MDA-MB-231 evidencing the highest migration distance when co-cultured. This enhanced migratory capacity correlated with a higher expression of MTDH gene, which accompanied almost linearly increased co-culture ratios of MSCs and breast cancer cells. Adapted with permission. [78] Copyright 2015, Elsevier 1 . www.advancedsciencenews.com www.biotechnology-journal.com biomaterial, cancer cells and MSCs efficiently self-organized into spheroids, presenting a core-shell structure wherein A549 cells exhibited enriched twofold upregulation of EMT-related genes, in comparison with 2D cultures. This remarkable enrichment was also observed in conditioned medium experiments with 2D cultures exhibiting lower EMT and tumorigenicity associated factors than those of direct co-culture. [83] Interestingly, in vivo assays demonstrated that co-culture derived implants presented higher malignancy, therefore, highlighting the importance of MSCs cancer cell direct contact in tumor progression. On the other hand, the establishment of negative interactions between BM-MSCs and cancer cells has been reported in human colon cancer cell lines HT29 and HCT-116, which presented decreased proliferation when co-cultured in vitro in a threedimensional scaffold (of unmentioned composition) with BM-MSCs. [76] In this study, BM-MSCs were found to secrete PAI-1, a factor closely associated with aggressive colon cancer, [84] and showed to have a direct effect in cancer cell proliferation. In the case of HCT-116 cancer cells, BM-MSCs excreted PAI-1 was shown to yielded a negative influence in cellular proliferation, while with HT29 this interaction lead to a positive outcome, in which high concentrations of PAI-1 increased cell proliferation. [76] However, when taking into account all factors secreted by MSCs, these exerted a negative effect resulting in a decrease in cancer cell proliferation in the order of 50%. [84] This fact becomes extremely relevant when one considers that MSCs of different origins are known to express different levels of cytokines and signaling factors in the same context, [85,86] emphasizing once again the necessity of further studying the influence of MSCs from diverse sources. Still regarding human colon cancer Widder and co-workers, 2016, have demonstrated that BM-MSCs were able to increase HCT8 Ecadherin deficient cell line capacity to assemble into 3D spheroids, whilst increasing HCT8 cell proliferation. [87] This elegant study showcased a close relation between HCT8 and MSCs-specific secretoma, in comparison to that of fibroblasts (CCD18Co). Furthermore, differential MSCs and fibroblasts spatial distribution was observed amongst 3D spheroids, with external deposition in colorectal cancer cell line DLD1 spheroids and with homogeneous internal deposition in colorectal cancer cell line HCT8 spheroids. [87] The remarkable increase in spheroid adhesion provided by MSCs co-culture for HCT8 is proposed by the authors to be correlated with a b1-integrin dependent mechanism, and aided through ECM deposition specifically via the production of collagen I, which also highlights the importance of 3D in vitro disease models in the context of cancer cell-stromal cell interaction studies.

MSCs Role in Complex Heterotypic Co-Cultures and Organotypic Bone-Metastasis Models
Regarding the development of more complex heterotypic coculture models, the study by Beckermen and co-workers, established a tri-culture spheroid system comprised by MIA-PaCa-2 cells, primary fibroblasts, and HUVEC cells. [88] These spheroids were then immersed in a methylcellulose/collagen solution and transferred to a fibronectin-coated plate previously www.advancedsciencenews.com www.biotechnology-journal.com seeded with BM-MSCs in order to simulate and analyze MSCs migration toward tumor-associated blood vessels. The authors observed a positive effect on MSCs migration toward the tri-culture spheroids, in which VEGF appeared to be the prevalent factor with PDGF and EGF also displaying significant chemoattractant properties. [88] In vivo studies in mice demonstrated the capacity of ectopically injected BM-MSCs of migrating toward xenografts of human pancreatic cancer, in which BM-MSCs lead to an increase in the number of CD31þ early vascular structures, an event that directly correlates with the density of tumor vascularization, as was observed, with tumors with twice the vascularization of those in obtained in controls. [88] Another tri-culture was performed by Lamichane and co-workers, 2016, in which A549 cells were cocultured with human pulmonary microvascular endothelial cells and BM-MSCs. [89] The tracking of the different cellular populations was accomplished through transfection with lentiviral particles encoding red fluorescent (RFP) and green fluorescent (GFP) transgene reporters, and followed for 15 days. The results showed a decline in the endothelial population, and an accentuated increase both in MSCs, which primarily formed the core of the spheroid, and cancer cells. From the endothelial cells co-cultured in these spheroids only a small population remained viable, being hypothesized by the authors to have survived by close association with MSCs within the hypoxic regions of the necrotic core. The spheroids exhibited increased multidrug resistance markers expression (ABC-B1 mRNA and ROS). However, no increased drug resistance in the 3D models was obtained when compared to standard 2D co-cultures. In fact, only some concentrations showed a prevailing difference between models, with the 3D model presenting always an increased cellular viability in comparison to the 2D model. [89] The importance of the 3D organization in cancer cellmesenchymal communication was sophisticatedly addressed in a study performed by Pasquier and co-workers. [90] This report employed both ovarian and breast cancer cell lines and the authors demonstrated the existence of a ubiquitous mechanism of cytoplasmic material transference through the usage of tunneling nanotubules in spatially adjacent cells, using 3D spheroid tri-culture systems. [90] The results indicated that mitochondrial transference between cancer cells and stromal cells took place, preferably, between endothelial cells (EC) and cancer cells, also occurring in the same mode between explant tissue and co-cultured EC cells. This exchange endowed cancer cells with a proliferative advantage by rescuing them from erratic aerobic respiration dysfunctions and endowed them with a significant resistance to chemotherapeutic agents. [90] This report further demonstrates the necessity for analyzing cancer cells-MSCs interactions in a tumor stroma context. www.advancedsciencenews.com www.biotechnology-journal.com In addition to being used for the establishment of 3D drug testing platforms or in the study of the complex interactions with cancer MSCs have also been employed for establishment of organotypic 3D in vitro bone metastasis models. Such approach takes advantage of the multilineage differentiation potential of BM-MSCs, namely their capacity to differentiate into osteoblasts. In this context, Jeon and co-workers, have developed a collagen I hydrogel laden with BM-MSCs differentiated into osteoblasts cocultured with vascular endothelial cells (HUVECs) and enclosed it in a microfluidic device (Figure 7). [91] The co-culture of osteodifferentiated MSCs and HUVECs promoted the formation of endothelium around the microfluidic channels and the deposition of bone-marrow specific biomolecules and matrix. This highly organotypic model was then used to evaluate the behavior of breast cancer cells that were injected in the channels of the hydrogelcontaining chip. The obtained results demonstrate that cancer cells in the tri-culture microfluidic model have a higher prometastatic potential, with higher rates of extravasation, migration and invasion. Such increased aggressiveness was evidenced by the breast cancer micro-metastatic pockets containing 4-60 cells that formed at 5 days of co-culture. [91] However, it is important to point out that due to the lack of MSCs population tracking during differentiation into osteoblasts it impossible to establish a direct connection between cells invasiveness and their interactions with possibly present non-osteogenic differentiated MSCs. This evidences the requirement for long-term cell tracking in these in vitro models, a parameter that will be further discussed.

Importance of MSCs Tissue of Origin and Cancer Heterogeneity
It is relevant to highlight that the origin of MSCs may also play an important role in cancer and consequently on the development of more advanced 3D in vitro disease models. In fact, a comparison between the effects of co-culturing Wharton jellyderived MSCs (WJ-MSCs) or BM-MSCs with lung cancer stem cells, either in 3D using multicellular spheroids or in 2D demonstrated contrasting effects between WJ-MSCs and BM-MSCs toward lung cancer stem cells (LCSC), with different subtypes of LCSC presenting different responses to WJ-MSC, especially when co-cultured in 3D. [32] Tumor heterogeneity may also play a role in cancer-MSCs interactions and hence influence the conclusions extrapolated from in vitro models. A recent study by Breznik and co-workers, 2017, demonstrated how phenotypically different cell lines used as models for glyoblastoma can lead to different MSC-cancer cells interactions and outcomes, with U87 cells showing decreased invasive potential, and U373 cells displaying the opposite, both in vivo and in vitro. [92] The researchers attributed this difference to alternatively favored metabolic pathways resulting from TGF-b release by MSCs. Thus underlining the importance of accounting for the heterogeneity and phenotypic variations present in actively proliferating tumors. [92] These results emphasize the necessity of considering the use of MSCs from several tissue origins, as well as different types of malignant cells to account for intrinsic in vivo tumor cellular heterogeneity.

Key Technical Parameters for Establishment of Cancer Cells-MSCs 3D In Vitro Co-Culture Models
The number of variables that can affect the interaction and coculture of such diverse cell populations makes achieving a valid co-culture model containing MSCs extremely difficult. In this context, it is important to emphasize that the relative lack of knowledge about the exact composition of both non-cellular and cellular components of the TME drastically reduces the ability to implement in vivo similar co-culture ratios. In addition, different cancer types are expected to possess diversified cell populations www.advancedsciencenews.com www.biotechnology-journal.com and thus, different ratios, such as those characteristic of brain and NSLC. [93,94] Moreover, variability may derive from different stages of the same cancer as verified in MM by Ref. [82] . The creation of in vivo mimicking in vitro 3D models for the study of cancer cells-MSC interactions have thus far relied in the usage of basic approaches. In a study by Zhu and co-workers, three diverse ratios (1:1, 1:2, and 1:5) were used for the co-culture of BM-MSC with breast cancer cells, the authors evaluated MTDH, an oncogene that promotes proliferation and enhances chemoresistance, [95] to be up-regulated in co-cultures containing higher ratios of MSC to cancer cells, when compared to lower MSC ratios. [78] In a study by Oerlecke and co-workers, 2013, as few as 1 MSC per 300 breast cancer cell was enough to obtain a significant increase in both Smad3 and CREB phosphorylation in BCC. [80] These findings showcase the necessity of correctly representing the diverse cellular populations of the TME in 3D laboratory models that aim to recapitulate the reality found in vivo, as different concentration of MSCs release dissimilar factors and cytokines that can lead to antagonistic responses. [25] In terms of cell culture specifications, MSCs are known to present diverse media requirements, as well as growth, proliferation, matrix deposition, and differentiation capabilities [22] when compared with other cellular components of the TME. With all these processes being affected by the previously referred factors, and by others such as the pre-existing scaffolds morphology and composition, which depending on the cell type might induce diversified phenotypes for example amongst cancer cells. [96,97] This complicates the establishment of a heterotypic co-culture model that suits all the necessities of the various cellular populations [98] ; especially when modified mediums for the coculture of cells are still rather underdeveloped. [99] Another problem that arises from these models is how co-culture populations can proliferate/change over extended periods of time, altering the initial experimental parameters as the trial progresses. [72] Such cell modifications can occur, for example, simply through the deposition of extracellular matrix, alteration/degradation of preexisting scaffolds for 3D culture, [54] or in a more complex facet, through the differentiation of the multipotent cells into diverse cellular populations otherwise not accounted for in the model. [100] When accounting for model alterations concerning the co-culture of MSCs and cancer cells, another variance that must be considered is how cancer cell cannibalism toward MSCs may affect the model progression. [101] This phenomenon, was recently addressed by using hanging drop-assembled 3D tumor spheroids which successfully replicated the relatively unstudied process of cellular cannibalism by MDA-MB-231 breast cancer cells (MDA-BCC) directed at MSCs surrounding populations in which cancer cells internalized and consequently degraded MSCs (Figure 8). Such resulted in the acquisition of unique molecular signature profiles by cannibal MDA-BCC cells, presenting an enriched molecular profile for pro-survival factors and tumor suppressor agents, as well as inflammatory mediators which marked these cells as detaining a senescence-associated secretory phenotype. [101] The authors were, thus, able to replicate in vitro a process of paramount importance in cancer cell metastasis and therapy survival, where MSCs once again are shown to have a vital role.
Apart from this, the various alterations that cellular populations ratios can suffer during culture must be carefully monitored, creating a necessity of tracking and analyzing each diversified cellular population in direct contact on 3D co-culture models. To this end, immunostaining techniques can be of use to a certain point, for example, when identifying non-stem cell populations or analyzing matrix deposition. [102] Unfortunately, for the differentiation of both cancer stem cells and MSC www.advancedsciencenews.com www.biotechnology-journal.com populations, difficulties might arise due to expression of overlapping receptors or other markers. [21,103,104] The use of cell tracking techniques, such as transfection with nonintegrative GFP transgenes via viral or non-viral vectors has been employed as an alternative. However, these methodologies are highly expensive and still present some technical constraints such as transient expression or influence in cells behavior. [73,88,105] Overall, the difficulty of tracking cellular populations hinders the ability to use more complex models for the analysis of specific cellular interactions. Such is the case with explant derived organoids, in which events such as the differentiation of MSCs lineages into osteo-differentiated MSCs must be prevented or dully tagged. [106,107] In summary, the establishment of a 3D MSC-cancer coculture testing platform encompasses several challenges, namely: i) different initial seeding ratios can lead to distinct results; ii) the spatiotemporal manner in which cells are introduced to the model can lead to diverse morphological organizations [79] ; with iii) diverse cells presenting diverse proliferative rates [108] ; iv) in vitro cellular division limits (Hayflick limit); v) and requirement of different types of medium; with all this diversity resulting in the need of vi) differentially tracking each cell population over time and space.

Conclusions
Upon analysis of different literature reports with 3D cultured MSCs, so far, they mainly focus on the use of co-cultured MSCstumor cells 3D spheroids. Only few articles take advantage of scaffold-based approaches and microfluidic chips, but mainly in the study of breast cancer metastasis into bone. [78,81,109,110] Developments in complex scaffold design for tissue engineering could be valuable in producing structures that could provide mechanical/biochemical cues and 3D spatial organization of MSCs and cancer cells. [111] Although the use of simple MSCscancer cells 3D co-cultures provides in vitro models compatible with high-throughput screening (one of the main requirements for developing models for pharmaceutics testing and discovery), several important TME associated factors are disregarded, including tumor-specific ECM, and other stromal cells-cancer cells bidirectional interactions, spatiotemporal mechanical, and biochemical cues. Clearly, this lack of ECM and stromal cells representation must be overcome if a representative model is to be achieved, possibly through the usage of complex functionalized scaffolds. [112] These scaffolds can be arranged in various forms in combination with microfluidic chips capable of providing a transport system in which mechanical cues obtained at different perfusion rates, combined with controllable biochemical signals, resembling those of an in vivo organism, allowing to increase the complexity of the designed 3D models. [98] It is also important to highlight that the widespread use of BM-MSCs populations in MSCs-cancer cell interactions can result in a bias toward the influence of this particular subset of cells. [25,85,86] In conclusion, further improvements to the currently available 3D in vitro tumor models will allow to develop valid platforms that permit the inclusion of MSCs in a way that more closely mimics in vivo conditions. While the analysis of individual interactions between these and other cells of the TME can provide a glimpse into the complex weave of communication and mutual regulation that takes place across the intricate setting present in malignant pathologies, ultimately the goal shall be to achieve a model that can be composed of both cancer cells, MSCs and other related stromal cells and immune system cells, in tissue specific ECM-like scaffolds. [113] This will require a combinatorial approach of current technologies, both in the fields of biomaterials and bioengineering, [99] in an effort to provide laboratory models that can be used for widespread screening of anti-cancer therapeutics in a number of different cancer types.