Mimicking tumor hypoxia and tumor-immune interactions employing three-dimensional in vitro models

The tumor microenvironment (TME) encompasses a complex cluster of cells that are programmed to fuel initiation, progression, and metastasis of cancer [1]. The TME is composed of tumor cells, various secreted factors and extracellular matrix (ECM), which provides structural and biochemical support to the surrounding cells [2]. The cells that constitute TME include fibroblasts, mesenchymal stem cells, macrophages, lymphocytes, endothelial cells, epithelial cells, and pericytes [3]. Collectively, these cells exert their influence on tumor progression and ECM remodeling [3]. TME secreted factors, including cytokines, integrins, proteases, and microRNAs function as signals between cells, or tools in ECM remodeling [4]. Each of these components within the tumor has unique roles and collectively hinder immune recognition and response, promote malignancy, and diminish cancer cytotoxicity within the TME [5].

Hypoxia within the TME is a significant tumor feature that can influence tumor-immune interactions. Under physiological conditions where oxygen is abundantly available, the conserved proline residues of hypoxia-inducible factor 1-alpha (HIF-1α) subunits are ubiquitously degraded through hydroxylation of its alpha subunits by the oxygen-dependent enzyme HIF-prolyl-4-hydroxylases (PHDs) [6]. Following this, von Hippel–Lindau protein (pVHL), an E3 ubiquitin ligase, is bound to the hydroxylated HIF-1α hence catalyzing the proteasomal degradation of HIF-1α. During conditions of oxygen deprivation, as typically seen in hypoxia within the TME, HIF prolyl-hydroxylases are inhibited, thus perpetuating the presence of functional HIF-1α (Fig. 1) [6]. The undegraded HIF-1α is now translocated to the hypoxia-responsive elements (HREs) of its corresponding target genes, eventually upregulating their transcription [7]. As the mechanism to degrade HIF-1α is turned off, cells adapt their metabolism to function through the HIF-1α signaling biochemical cascade [8]. A direct consequence of this modulated metabolism functioning is overexpression of lactate dehydrogenases, pyruvate dehydrogenase kinases and other enzymes that regulate glycolysis. Increased expression of these enzymes leads to accelerated glucose uptake and glycolysis [9]. The effects of increased HIF-1α functioning also contribute to increased production and recruitment of pro-tumor entities such as myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and tumor-associated macrophages (TAMs) which modulate the surrounding tumor niche and compete with T cells and NK cells for glycolysis [10, 11]. Studies have conclusively shown that by reversing these metabolic changes, both T-cell glycolysis and the production of IFN-γ can be significantly restored [11]. A sustained presence of undegraded HIF-1α in the hypoxic tumor core can also correlate to differentiation checkpoints for Treg or Th17 cells, and an upregulation of PD-1 ligand expression on cancer cells. Reports have shown that Th17 cells particularly favor these increased glycolytic processes associated with undegraded HIF-1α under hypoxic conditions [12]. It has also been shown that by therapeutically blocking glycolysis during the differentiation stage of Th17 cells, Treg formation can be significantly promoted [13]. The collective influence of these changes correlates to several immune evasive consequences that stimulate the TME to either resist infiltration of cytotoxic T cells and NK cells [14, 15] or to inactivate the already infiltrated immune cells in the immunosuppressive microenvironment [16]. Several studies have highlighted that these evasive immune outcomes in the TME could potentially result in increased resistance to chemo and immune therapies [17, 18]. A recently published report has shown that there is a significant correlation between weakened immune infiltration and TME hypoxic scores [14]. This study identified a link between multiple gene signatures within the TME that are specific to tumor hypoxia and lymphocyte infiltration. The results from this study indicate that reduced CD8⁺ cell cytolytic activity driven by low oxygen could be of clinical significance as a prognostic factor for various tumor types, where hypoxia-driven CD8⁺ cell cytolytic activity could be associated with worse outcomes. A significant process driving hypoxic TME to influence immunosuppression is the enhancement of anaerobic glycolysis [15]. During hypoxia-driven anaerobic glycolysis in TME, tumor-infiltrating CD8+ T cells have to compete with host tumor cells for glucose, thus ultimately inhibiting their effective response [16].

The continually evolving TME implements several biological events that influence tumor-immune interactions [19]. Physiologically, immune cells like CD8⁺ lymphocytes possess the capability to control cancer growth by exerting cytotoxic action over the cancerous tumor cells in a sustained manner [20]. However, the oxygen-deprived TME also enhances the secretion of immunosuppressive cytokines and chemokines that create a hostile environment for immune cell cytolytic activity [21, 22]. The collective influence of these components manipulates TME pathology and its role in chemoresistance [23]. Cancer cells within the tumor, along with other tumor-associated cells, control their surrounding hypoxic niche by enduring a dysfunctional metabolism leading to altered cell transdifferentiation, i.e., enhanced cancer-associated myofibroblast production [24], incorrect cytokine expressions [25], and altered ECM composition [26]. The utilization of three-dimensional (3D) in vitro models for deciphering the complex tumor immune communications yields dependable outcomes that help overcome therapeutic shortcomings [27]. In this review, we systematically review some of the current state-of-art 3D models that address the synergy between tumor hypoxia and tumor-immune interactions.

The complex nature of cancer models and their ability to recreate the TME varies widely, ranging from two-dimensional (2D) models to animal-based models. Although animal models are viewed as critical platforms to test anti-tumor drugs, mouse models are costly, relatively time-consuming, and the studies performed using mice are often not repeatable in human trials [28]. Also, to decrease animal distress and mitigate ethical concerns, scientists are tasked to find alternative methods that replace, reduce, and refine the use of animals [29]. To address these issues, 3D cell culture models are a reliable alternative, providing experimentally accessible human models to study the biological processes of cancer. Several 3D culture platforms such as spheroids, organoids, hydrogels, 3D scaffolds, 3D bio-printing, and microfluidics have attempted the recreation of certain aspects of tumor microenvironments present in tissues including the brain [30‐36], breast [37‐43], ovarian [44‐51], bone [52‐58], liver [59‐65], lung [66‐72], colon [73‐79] and thymus [80‐83], as illustrated in Fig. 2. Most of these studies have employed the stromal component of the tissue’s matrix to serve as the base platform. Various approaches taken to mimic several features of the TME in 3D models include the development of tumor spheroids [84], the design of scaffold-based TME models [85], patient-derived xenograft systems where cells isolated from patients are incorporated into rodents for investigative purposes [86], organoids that can be self-organized into desirable tissue phenotypes and mimic the functionality of an organ while expressing one or more cell types [87], highly maneuverable microfluidic systems [88], and more recently 3D bio-printing platforms that can recreate the TME under highly controllable parameters yielding tailored 3D tissue architecture [89]. Although each 3D model has its pros and cons, studying disease pathology in glandular or stratified tissues has become extremely convenient using these systems [37]. For example, organoids, with their capability to spontaneously support differentiation and self-organization of their host cells [90], have efficiently yielded outstanding research breakthroughs through stunning microscopy images, revealing critical molecular and cellular mechanisms that affect disease progression [91]. Organoids have earned the title of ‘Method of the Year 2017’ [92], and since then, their applications in therapeutic testing have seen a tremendous increase. However, most of the currently available 3D models have encountered a high attrition rate in terms of clinical translatability. This is due to their inability to generate patient-specific phenotypes, thus failing to accurately predict the degree of therapy response foreseen in specific patients or generate a tangible model to clinic therapeutic efficiency amongst different patients [93]. The most significant causes of these hurdles include the inability to entirely re-engineer a heterogeneous TME [94] and the lack of an approach to mimic the native streamlined modulation caused by spatial heterogeneity [95] within 3D models. Another important reason in vitro 3D models have failed to show clinical efficacy is due to the inter-patient variability [96]. Personalized approaches that can reinstate patient-specific cues could overcome these hurdles. The complex nature of the tumor-associated stromal components, including their organization, differences in oxygen content relative to tumor volume and disease location, and the immune system composition in each patient, collectively create many challenges for scientists and clinicians.

Approaches to recreate the TME in 3D models have evolved considerably over time. From early prototype models to the current state-of-the-art systems, 3D model designs have advanced immensely with our increased understanding of TME physiology. With time, the models have increased in complexity and served to recreate even more in vivo like parameters. As illustrated in Fig. 3a, although the idea of utilizing the third dimension in cell culture research has been well-practiced since the early 1900s, interest has spiked significantly over the last few years [97], which is reflected by the increase in the number of publications employing 3D models (Fig. 3b). The hanging drop and watch glass techniques are some of the first 3D design models that provided an efficient way to use non-adhesive cell culture technologies on pre-coated plates. The hanging drop method is especially helpful in promoting the spheroid formation of cells without attaching to the plastic surface of the plate [98]. Several other prospective 3D platform designs have surfaced after the successful isolation of collagen from rat tail. Now, it is possible to fabricate designs mimicking macromolecular networks in which various types of cells, such as mesenchymal cells can be incorporated. With the capacity to study cell-extracellular matrix interactions, the diagnosis and treatment of various pathophysiologies such as fibrosis and cancer have greatly improved. 3D models prepared using high-quality purified collagen type I, provide researchers valuable in vivo like suggestions leading to ideas that were never considered before. The introduction of organoids and Matrigel® matrix also significantly helped evolve the 3D cell culture model blueprints. The steep increase in the number of publications that used these two platforms of study reveals the interest generated by these systems during the early days of 3D cell culture. The compact nature of innovative microfluidic platforms could be thought of as the organoid of this decade. Exceptional microfluidic device designs that challenge unanswered biochemical questions look promising to serve as prototypes for a new era of 3D in vitro models.

While creating 3D experimental models to study cell-tissue interactions, the importance of using a prototype that aims to study TME hypoxia has always been a top priority. In order to fully understand the numerous consequences by which a low oxygen niche within the TME can affect tumor pathogenesis, several factors need to be concurrently embedded in a single 3D system. A hypoxic microenvironment could modify gene expression patterns in tumor cells [99], promote angiogenesis [100], increase chemoresistance [101], and support metastasis [102]. Altered cell metabolism under hypoxia, through stabilized HIF-1α, decreases cell proliferation [103]. As cell proliferation is nourished by an oxygen-enriched environment, a hypoxic TME results in a problematic fraction of cells that can survive the hypoxic stress and start exhibiting invasive characteristics [104]. While most of the cancer drugs are designed to target rapidly proliferating cells, tumor cells in a hypoxic niche evade the therapeutic effect of these drugs by demonstrating low proliferation [105]. Low oxygen can result in an acidic environment, relative to the pH heterogeneity inside the TME [106], which inhibits drug uptake rates in cells due to the impediment of molecule diffusion as a result of unnecessary cell membrane charging [107]. Most cancer drugs are weakly basic drugs that become ionized in the acidic hypoxia-driven TME and hence fail to exert their therapeutic effects [108]. Moreover, an altered ECM produces an additional hurdle for cancer drugs and cytotoxic immune cell infiltration into the hypoxic TME [26].

The effect of TME hypoxia on tumor cells has been studied extensively in 2D monolayer cultures [109]. These 2D hypoxia experiments are mostly performed by incubating cells in incubators or gas-chambers, where the amount of oxygen can be controlled as needed. Although simple, these models fail to achieve all of the experimentally relevant aspects associated with in vivo conditions. For example, a 2D model fails to recreate TME oxygen gradients and creates a polarized two-dimensional system of either high or low oxygen [110]. On the other hand, studying hypoxia in vivo has been difficult due to the high degree of oxygen tension variations among tissues [111] and the limited ability to identify chronic versus acute hypoxic factors within the TME [112]. As a result of these failures, researchers have sought multiple tools for designing 3D platforms to study tumor hypoxia. Some of the advances made in the hypoxic 3D cell culture model realm are outlined below.

The need to develop dynamic TME models that allow researchers to study the extent of immune cell infiltration into the tumor corresponds to the demand to improve preclinical screening of immunotherapies. As the latest advancements in immunotherapy research are focused on the use of immune checkpoint inhibitors, these models would be essential to monitor the mechanisms by which TME influences immune evasion. Lately, the 2D models used to study tumor-immune interactions are being upgraded to 3D models as they help to reveal new immunotherapeutic perspectives [125]. Classic 2D models that have been used to study tumor-immune biology mostly employed permeabilized supports or Boyden chambers to investigate chemotaxis of immune cell migration and invasion [126]. Although these models have established interplay between numerous tumor-immune interactive aspects, they do not conclusively mimic the events that may challenge immune cell infiltrations into the tumor. Hence, there is a need to develop in vivo-like 3D multicellular models that seamlessly reiterate multiple tumor-immune aspects. The basis for an efficient 3D model that achieves this would depend on five critical hallmarks, as illustrated in Fig. 4: the choice of the biomaterial to create the 3D platform, the interplay of the various cellular heterogeneities, the addition of biochemical cues that would help maintain the homogeneous niche. These biophysical cues help retain the in vivo characteristics and the choice of immune cell types being selected for infiltration testing. The ideal 3D platform would need to be prepared by effectively balancing each of these hallmarks. Despite researchers modulating one or more of these five parameters: biomaterials: [26], cellular factors, biochemical factors, physical cues: [89] and immune system components: [127], very few have incorporated all five hallmarks into a single model.

A large number of tumor ECM components have been modified to construct the TME in 3D models, such as the use of biosynthetic collagen platforms [128] and fibrinogen based systems [119]. While these approaches have effortlessly recreated TME characteristics, scientists have also utilized many ECM-like materials as building block substitutes for creating 3D TME models. The choice of such an alternative material creates value in terms of availability and economic aspects. One such commonly used substance is silk fibroin protein procured from the cocoon of the silkworm Bombyx mori. Silk fibroin is being investigated extensively to produce an ideal TME building block for a 3D model [129]. The silk fibroin generates a non-immunogenic matrix after implantation, is highly maneuverable in terms of biodegradation, is stable in terms of robustness, and is abundantly available. Ultimately, the choice and combination of multiple parameters while constructing the platform, and the design blueprint, will determine the applicability and functionality of the 3D model. Some of the studies that address designing 3D approaches to recreate evasive tumor immune cell infiltration are listed as follows.

Despite the importance of connecting tumor hypoxia and reduced immune surveillance within a single model, very few studies have achieved unifying these two hallmarks. On this note, there is a significant study that highlights the advantage of 3D cell cultures in reproducing hypoxia-dependent changes observed in the TME. In particular, hypoxia was shown to negatively regulate MHC expression in a HIF-dependent manner as demonstrated by lower MHC expression in hypoxic 3D models but not 2D tumor cell cultures in vitro. Downregulation of MHC class I expression is a hallmark of cancer and leads to cancer escaping recognition and rejection by anti-tumor T cells [135]. Furthermore, in a prior study from our lab, we recently showed that 3D models engineering physio- and pathophysiological oxygen levels allow for a better understanding of the role of oxygen availability in tumor-immune interactions. In particular, an oxygen-deprived environment was shown to recapitulate known breast cancer hypoxia characteristics such as reduced cell proliferation, increased extracellular matrix protein expression, and immune evasion mechanisms. Additionally, CD8⁺ T cell infiltration was significantly impaired under pathophysiological oxygen levels, and the inhibition of HIF or PD-L1 was able to re-sensitize breast cancer cells to cytotoxic T cells [136]. One more study that must be highlighted builds on the collective understanding that hypoxia can aggravate immune evasion in the physiological TME [127]. This research employed highly efficient chimeric antigen receptor T (CAR-T) cells, coupled with a microfluidic platform that accurately generated gradients of oxygen to mimic TME hypoxia. Immune cell infiltration into the 3D model was observed to vary significantly based on the oxygen content, reflecting the already well-understood series of evasive immune events that occur as a result of low oxygen within the tumor. As recreating an oxygen profile in a 3D model is essential to determining the exact tumor-immune evasive mechanisms, the model reported in this study could accelerate hypoxia-specific immunotherapies encompassing specific biochemical mechanisms for each immune cell type.

The development of 3D ex vivo models is an area of rapid expansion, but the integration of immune cells in these cultures remains in its infancy. Further studies will warrant a better understanding of whether creating 3D in vitro models while mimicking tumor hypoxia and tumor-immune interactions would fare advantageous and therapeutically beneficial over classical in vivo animal models. Significant advances have been made in 3D tumor engineering toward development of the ideal TME model that may adequately accommodate multiple tumor properties. An ideal 3D TME model would function effortlessly by either being entirely inert from the patient immune variations or by implementing a personalized platform that is constructed using the patient’s cells. A 3D model that would bridge the gap between tumor hypoxia, immune evasion, and recreation of TME features as close as possible to in vivo conditions, must be suitable for all the different tumor cell types, without affecting inter-cell biology or the potentiation of immunogenic reactions. Although the advances have been noteworthy, all of the synthetic or biologically derived systems that have been fabricated with or without cancer cells have unfortunately hit a roadblock at some point in the developmental pipeline. The systems that have been developed in the presence of cellular growth are promising, but only from a proof of concept perspective. Also, the 3D models that have shown cellular growth have primarily been able to generate only specific tumor cell types: mutually exclusive populations of cancer, endothelial, stromal, or immune cells. Future studies that address the development of a system that can fuse all the different cell populations simultaneously are necessary. Results from several studies over the years have shed light on the understanding of solid tumor biology, thus forming the basis of TME modeling within a 3D in vitro niche. Despite extraordinary developments, most models recapitulate only some particular aspects of the TME; thus, further research and more holistic models are needed. As we advance in the field of 3D modeling of the TME, our methodologies and designs will hopefully become more superior. The synchronous blending of cellular growth, oxygen profile, immune cell interactions, and heterogeneity within the same 3D model would be a highly rewarding therapeutic advancement.