Modeling tumor cell adaptations to hypoxia in multicellular tumor spheroids

While the term hypoxia is used to describe a wide variety of oxygen concentrations [2, 7], it most often refers to the point at which oxygen concentrations have decreased beyond the threshold required for normal cell function. The majority of solid tumors will develop hypoxic regions due to a combination of rapid oxygen depletion, insufficient vascularization, and suboptimal tumor blood flow [2, 7]. For example, the consumption of oxygen by rapidly proliferating perivascular tumor cells can deplete the limited supply of available oxygen and prevent sufficient oxygenation of subsequent cell layers [8, 24‐26]. While intracellular oxygen is utilized in a variety of reactions, the majority of oxygen consumption is devoted to ATP production through glucose metabolism [26, 27] where oxygen serves as a terminal electron receptor during oxidative phosphorylation. In addition to consumption through intracellular processes, the physical distance between tumor cells and blood vessels also influences the development of hypoxia. Oxygen diffusion through tissue is limited to approximately 200 μm based on evidence from experimental and mathematical models [3, 28]. Hypoxia can be further exacerbated by the destruction of angiogenic vessels following cytotoxic or anti-angiogenic therapy [8, 29‐31]. Accumulating evidence now suggests that antiangiogenic therapy induces tumor hypoxia which provides a selective pressure for tumors to acquire a more aggressive phenotype leading to therapeutic resistance and tumor progression [29‐31]. Whether developed as a result of rapid tumor growth or in response to therapeutics, hypoxia is ultimately the result of an imbalance between oxygen availability, consumption, and the physical boundaries to oxygen diffusion inherent to a 3D tissue mass.

The effect of hypoxia on cells has traditionally been studied in monolayer culture. 2D (monolayer) hypoxia experiments are most typically performed by placing tumor cells in a gas-controlled chamber [2]. While experimentally straightforward, this method is unable to recreate clinically relevant aspects of tumor biology that can impact on tumor cell behavior and therapeutic response [13, 32]. For example, monolayer cells experience polarized cell adhesion and two dimensional contact with neighboring cells which results in abnormal cell spreading, alterations in the distribution of cell surface receptors, and selection for specific sub-populations of cells best adapted to in vitro growth [32]. It is also well established that the genomic profiles and therapeutic responses of tumor cells grown in 2D differ from those seen in solid tumors [9‐13, 33]. Studying hypoxia in vivo is challenging due to the high degree of variation in oxygen tensions within and amongst tumors, and limited ability to definitively identify regions of chronic versus acute hypoxia [2, 34]. For these reasons, there can be a disconnect between in vitro studies and the complex 3D environment of a tumor. MCTS may contribute to bridging this gap.

Several methods are used for generating spherical 3D cultures and are reviewed extensively elsewhere [8, 32, 35‐37]. Common techniques include the liquid overlay technique in which 3D culture is attained by incubating cell suspensions in plates coated with an inert substrate, and the “hanging drop” method wherein cell suspensions are cultured in suspended droplets such that gravity prevents cell attachment and favors cell-cell adhesion [8, 36, 38]. Several factors, including the size of the spheroids, can influence the behavior of MCTS [39]. Spheroid size and media composition dictate the viability and growth kinetics of 3D cultures due to the development of gradients in oxygen and metabolites (discussed below); therefore 3D culture requires careful design of growth conditions and analytical endpoints [39]. Excellent discussions on the practical considerations needed when designing and interpreting data with 3D culture models have been published elsewhere [12, 39].

There are four types of 3D spherical culture models that differ in some important ways (recently reviewed in [36]): MCTS, Tumor Spheres (TS), Tumor Derived Tumor Spheres (TDTS), and Organotypic Multicellular Spheres (OMS). MCTS are formed from established tumor cell lines and grown in standard culture media. MCTS can be composed of mono- or heterotypic cell populations, the latter being the co-culture of tumor cells with other cell types such as macrophages, endothelial cells, and fibroblasts (discussed later in this review). A variation on MCTS culture is the TS model which has been used as a method for growing tumor stem cells in 3D. These spheres are formed from clonal expansion of single cells suspended in non-adherent conditions supplemented with a specific compliment of growth factors in the culture media [36]. Unlike MCTS derived from established cell lines, TS culture represents a selective population of cells known to be aggressive and likely to contribute to tumor regrowth [36, 40]. Accordingly, TS may differ in drug response and growth kinetics relative to MCTS owing to the enrichment of the cancer stem cell population [40]. TDTS are similar to TS however they are formed from partially dissociated tumor tissues [36, 40, 41]. In the TDTS model, primary tumor cells are separated from non-tumor cell types and grown in non-adherent conditions. Relative to MCTS and TS, TDTS have been shown to more accurately recreate tumor growth and gene expression profiles. For example, in breast and colon cancer, TDTS mimic differentiation properties and growth kinetics of the parent tumors more accurately than MCTS from the same tumor background [36, 42]. TDTS therefore provide a suitable model for studying properties of individual tumors. The OMS model utilizes primary, non-dissociated, tumor tissue comprised of all cell types residing within the tissue at the time of excision which provides additional complexity. The presence of stromal cells can adversely affect tumor therapeutic response; therefor the OMS culture method provides a suitable model in which an individual tumor’s therapeutic response can be studied to predict an in vivo response [36, 43, 44].

Each model has distinct advantages and disadvantages (discussed elsewhere [12, 15, 32, 36]) but for maximal control and reproducibility of cell behavior, the MCTS model provides the best coupling of speed with which spheres can be generated, the ability to eliminate influence from non-tumor cell types, and the heterogeneous phenotype of tumor cells incorporated into the sphere. The properties of 3D culture discussed in this review however are applicable to most 3D systems because critical physical properties, such as the development of hypoxia, within the 3D environment are consistent amongst models [12, 32, 39, 45].

Spheroids are a useful in vitro model of avascular tumor spaces. Spheroids exceeding 400 μm in diameter develop a hypoxic core and activate known survival signaling pathways to maintain cell viability. Tumor cells grown as MCTS, TS, TDTS, and OMS display regional heterogeneity in tumor cell proliferation [44, 46, 47], metabolic activity [9‐11, 32, 44, 48, 49], and DNA damage repair (DDR) signaling [9, 50‐53]. In addition to the influence of the 3D microenvironment, the genetic background of the tumor cells and stromal cell populations also contribute to the specifics of spheroid regionalization [39, 54‐56].

Hypoxia in the 3D environment of tumors and MCTS induces the stable expression of hypoxia inducible factors (HIF). In vivo, HIF proteins transcriptionally promote angiogenesis through VEGF-A, glycolysis, and pH control through CA-IX [57]. Expression of HIF target genes occurs in tumor regions distal to blood vessels and is commonly used to identify hypoxic conditions [2]. Similar patterns have been observed in MCTS wherein HIF target genes (CA-IX, Glut1, VEGF-A) are expressed in the inner, hypoxic cell layers [8, 58‐61]. Functionally HIF expression in both MCTS and in tumors has been linked to cell survival through the repression of pro-apoptotic signaling, repression of proliferation, and the regulation of metabolic reprograming [58, 62]. In this way, HIF signaling contributes to regionalization of MCTS cell layers and the formation of microenvironments as a function of cell distance from the MCTS surface: limited diffusion of oxygen into MCTS establishes gradients in HIF-α stabilization and subsequently tumor cell behavior. Similarly, rapid proliferation by well oxygenated cell layers and the formation of physical barriers, through tight cell-cell contacts and ECM deposition, generates gradients in glucose, catabolites, and therapeutics [32, 63]. The following sections will discuss the ways in which gradient development within the MCTS model reflects tumor cell adaptations in the avascular tumor space in vivo. These adaptations are schematized in Fig. 1.

The MCTS discussed thus far were composed exclusively of tumor cells. However, tumors are diverse in terms of their micro-environmental composition as well as their cellular composition. The adaptation of tumor cells to a hypoxic environment is influenced by the activity of stromal cells including endothelial cells, fibroblasts, adipocytes, macrophages, monocytes, and other cell types [115]. As in the case of tumor cells, the hypoxic environment can affect stromal cell behavior [116]. Endothelial cells are known to increase proliferation in response to a hypoxic environment and to activate a DDR signaling cascade similar to that described for tumor cells [117‐119]. Accumulating evidence suggests that immune cell infiltration into the hypoxic tumor microenvironment increases angiogenesis and tumor metastasis [6, 120]. Both adipocytes and fibroblasts contribute significantly to tumor progression through the production of pro-proliferative cytokines and through production of extracellular matrix (ECM) proteins which confer chemo-resistance [6, 16, 56, 121‐124].

Towards accurately recreating the tumor microenvironment in vitro, various co-culture models have been developed wherein MCTS are composed of both tumor and stromal cells. The OMS model is perhaps the most representative form of co-culture system due to the inclusion of all resident stromal cell types within the excised tissue [36, 125]. OMS have been used as a model for multiple tumor types and accurately reflect tumor growth kinetics [36, 43, 44]. In glioblastoma for example, the OMS model was found to be superior to MCTS in its ability to recreate the immuno-histochemical profile of in vivo tissues including the expression of many proteins previously correlated to hypoxia-induced aggression (CD31, CD133, P-glycoprotein, and TIMP-1) [44]. However, the inclusion of multiple cell types in the OMS model makes it difficult to determine the influence of a specific stromal cell population on the adaptive tumor cell response [36, 125].

Heterologous spheroid culture in which tumor cells are combined with one or more stromal cell types has provided unique insights into stromal and tumor cell responses to hypoxia [15, 36, 126]. The most commonly used methods of heterologous culture can be broadly described as: 1) Spheroid confrontation culture: individual spheroids are formed from stromal cells and separately from tumor cells followed by joint culture in suspension or embedded within an ECM, 2) Spheroid-monolayer culture: pre-formed MCTS are placed atop confluent monolayers, 3) Heterologous spheroid co-culture: tumor and stromal cell suspensions are combined during the sphere forming process, 4) Spheroid co-cultures in vivo: pre-formed MCTS containing stromal cells are implanted in vivo [125].

MCTS in co-culture with endothelial cells mimic several processes during hypoxia-induced angiogenesis: oxygen deprivation within the MCTS core stimulates production and accumulation of vascular endothelial growth factors (VEGF) [61, 127] which provides a stimulus for endothelial cell recruitment. In vivo, endothelial cell invasion into tumors facilitates metastasis and release of tumor cells from the growth inhibitory effects of a hypoxic environment [6]. Other MCTS co-culture models recreate a similar process and have been used to improve our understanding of the relationship between endothelial cells and tumor cells [125, 128]. For example, placement of tumor spheroids atop confluent endothelial monolayers can result in infiltration by endothelial cells expressing tumor complimentary adhesion molecules or can result in tumor-mediated destruction of endothelial vessels, reminiscent of cell death observed in vivo following vessel cooption [125, 128]. Use of confrontation or heterologous co-culture models can successfully produce vascular networks within spheroids with an increase in tumor cell viability and drug resistance [22, 129]. The invasive potential of tumor cells following vascularization is also replicated through the use of heterologous sphere culture systems [56, 122, 130]. A recent study highlighted this by showing the formation of MCTS with both endothelial and tumor cells could lead to the formation of luminal vessels in which migrating tumor cells can be seen [130], mimicking the process of tumor intravasation.

Fibroblasts are also a critical stromal cell population that contribute to tumor progression through the release of proliferative and pro-metastatic growth factors and foster the creation of a drug resistant environment by depositing ECM which hinders drug diffusion [122, 125, 126]. Each of these characteristics are replicated in heterologous sphere culture [16, 56, 122, 124‐126]. Within a pro-angiogenic environment, fibroblasts take on a mural cell phenotype and augment angiogenesis through the release of VEGF, matrix metalloproteinases (MMP), and other growth factors [16, 56, 122, 124, 126]. In vitro this process can be studied through the “mini-tumor model” in which endothelial cells, fibroblasts, and tumor cells are co-cultured in 3D [122]. This tri-culture model replicated pathological formation of luminal vascular structures bordered by fibroblasts and tumor cells which are both dependent on angiogenic growth factors and responsive to anti-angiogenic therapeutics [122]. This mini-tumor model has significant potential for studying the refractory response of tumor and stromal cells following anti-angiogenic therapy and the specific mechanisms used to circumvent these treatments.

Immune cells are a major regulatory factor during tumor progression [5]. Within the hypoxic tumor environment, there are indications that immune cells facilitate tumor aggression through increased angiogenesis and immunosuppression [115, 131]. In vivo immune cells aggregate within hypoxic, peri-necrotic regions [131]. These observations have been mirrored in MCTS co-cultures and indicate potential utility for MCTS as an investigational tool to replicate in vivo conditions [56, 132]. In support of this, immune cell influence over angiogenesis has been described by a series of studies through the inclusion of monocytes in co-culture with tumor spheroids to show immune-dependent increases in production of VEGF, ECM degrading proteins, and in vivo angiogenesis through MCTS co-culture in skinfold window chambers [56, 131‐133]. A particular advantage to MCTS co-culture in modelling the hypoxic microenvironment is the accumulation of immunomodulatory metabolites and cytokines.

Levels of lactic acid are commonly high in hypoxic tumor regions due to increased glycolytic activity and HIF mediated production of lactic acid transporters [134]. In vitro studies show that hypoxia-induced increases in lactic acid and VEGF can decrease dendritic cell differentiation and maturation while increasing monocyte conversion to M2 macrophages, ultimately resulting in immunosuppression [56, 131, 135‐137]. The MCTS co-culture models are well suited to model a hypoxic microenvironment with high metabolite and growth factor levels. Hypoxic spheroids show higher levels of lactic acid which correlates with decreased immune cell invasion, reduced production of colony stimulating factor [131, 136] and decreased dendritic cell maturation in hypoxic conditions [131, 138].

Aside from instigating an aggressive phenotype, tumor hypoxia presents several challenges to therapeutic intervention [3, 139]. Hypoxia has been appreciated for its ability to reduce radiation effectiveness for many decades [24, 84, 140]. Since this phenomenon was described, a significant amount of evidence has amassed demonstrating the ways a hypoxic environment renders tumor cells privileged, or resistant, to multiple therapeutic modalities [3, 13, 22, 23, 44, 63, 88, 139, 141, 142]. Several studies have detailed the contribution of stromal cells to chemo-resistance in the MCTS model [14, 16, 122, 124] and technological advances are allowing more complex co-culture systems to be developed [15, 45]. The use of MCTS for screening of therapeutics has been extensively reviewed elsewhere [39].